Intercellular Communication in the Nervous System [1 ed.] 0123750725, 9780123750723

Citation preview

INTERCELLULAR COMMUNICATION IN THE NERVOUS SYSTEM

This page intentionally left blank

INTERCELLULAR COMMUNICATION IN THE NERVOUS SYSTEM EDITOR-IN-CHIEF

ROBERT C. MALENKA Department of Psychiatry and Behavioral Sciences Stanford University School of Medicine Palo Alto, CA USA

Amsterdam • Boston • Heidelberg • London • New York • Oxford Paris • San Diego • San Francisco • Singapore • Sydney • Tokyo Academic Press is an imprint of Elsevier

Academic Press is an imprint of Elsevier 32 Jamestown Road, London NW1 7BY, UK 30 Corporate Drive, Suite 400, Burlington, MA 01803, USA 525 B Street, Suite 1900, San Diego, CA 92101-4495, USA Copyright ã 2009 Elsevier Inc. All rights reserved The following articles are US government works in the public domain and are not subject to copyright: AMPA Receptors: Molecular Biology and Pharmacology BDNF in Synaptic Plasticity and Memory Long-Term Depression (LTD): Endocannabinoids and Cerebellar LTD NMDA Receptors, Cell Biology and Trafficking No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher. Material in this work originally appeared in the Encyclopedia of Neuroscience, Ed. L.R. Squire, Elsevier Ltd, 2009. Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone (+44)(0) 1865 843830; fax (+44) (0) 1865 853333; email: [email protected]. Alternatively you can submit your request online by visiting the Elsevier web site at (http://elsevier.com/locate/permissions), and selecting Obtaining permissions to use Elsevier material Notice No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Catalog Number: 2009929652 ISBN: 978-0-12-375072-3 For information on all Elsevier publications visit our website at books.elsevier.com PRINTED AND BOUND IN SLOVENIA 09 10 11 12 13 10 9 8 7 6 5 4 3 2 1

CONTRIBUTORS

A Adamantidis

G Bernardi

Stanford University School of Medicine, Palo Alto, CA, USA

Universita` di Roma ‘Tor Vergata’, and IRCCS Fondazione Santa Lucia, Rome, Italy

G Aghajanian

T P Blackburn

Yale School of Medicine, New Haven, CT, USA

Helicon Therapeutics Inc., Farmingdale, NY, USA

S P H Alexander

N G Bowery

University of Nottingham Medical School, Nottingham, UK

GlaxoSmithKline, Verona, Italy

N J Allen

University of Rome ‘La Sapienza,’ Rome, Italy

Stanford University School of Medicine, Stanford, CA, USA

G Burnstock

R S Aronstam

University of Missouri – Rolla, Rolla, MO, USA G Aston-Jones

Medical University of South Carolina, Charleston, SC, USA D Atasoy

The University of Texas Southwestern Medical Center, Dallas, TX, USA

V Bruno

Royal Free and University College School of Medicine, London, UK D B Bylund

University of Nebraska Medical Center, Omaha, NE, USA L Cancedda

University of California at Berkeley, Berkeley, CA, USA G Casini

B A Barres

Universita` della Tuscia, Viterbo, Italy

Stanford University School of Medicine, Stanford, CA, USA

D Cervia

Universita` della Tuscia, Viterbo, Italy

G Battaglia

Istituto Neurologico Mediterraneo ‘Neuromed,’ Pozzilli, Italy

J-P Changeux

K L Behar

H Cline

Yale University School of Medicine, New Haven, CT, USA

Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA

M V L Bennett

G L Collingridge

Albert Einstein College of Medicine, New York, NY, USA

University of Bristol, Bristol, UK

F Bergquist

University of California at San Diego, La Jolla, CA, USA

University of Edinburgh, Edinburgh, UK

Institut Pasteur, Paris, France

J M Conner

v

vi

Contributors

A C Cuello

B A Grueter

McGill University, Montreal, QC, Canada

Vanderbilt University School of Medicine, Nashville, TN, USA

T C Cunnane

University of Oxford, Oxford, UK

E D Gundelfinger

M O Cunningham

Leibniz Institute for Neurobiology, Magdeburg, Germany

Newcastle University, Newcastle upon Tyne, UK G W Davis

University of California at San Francisco, San Francisco, CA, USA T M Dawson

The Johns Hopkins University School of Medicine, Baltimore, MD, USA V L Dawson

S Harris

Saint Louis University School of Medicine, St. Louis, MO, USA P G R Hastie

University of Bristol, Bristol, UK V Haucke

Freie Universita¨t Berlin, Berlin, Germany

The Johns Hopkins University School of Medicine, Baltimore, MD, USA

J M Henley

J S Dittman

A M Holohean

Weill Cornell Medical College, New York, NY, USA

University of Miami School of Medicine, Miami, FL, USA

P D Dodson

Geffen School of Medicine, Los Angles, CA, USA A J Doherty

University of Bristol, Bristol, UK S M Dravid

Emory University, Atlanta, GA, USA

University of Bristol, Bristol, UK

M O Huising

The Salk Institute for Biological Studies, La Jolla, CA, USA Department of Animal Physiology, Radboud University Nijmegen, Nijmegen, The Netherlands K A Jacobson

T M Egan

National Institutes of Health, Bethesda, MD, USA

Saint Louis University School of Medicine, St. Louis, MO, USA

P S Kaeser

B A Eipper

University of Connecticut, Farmington CT, USA

University of Texas Southwestern Medical Center, Dallas, TX, USA E T Kavalali

J D Elsworth

Yale University School of Medicine, New Haven, CT, USA S M Fitzjohn

University of Bristol, Bristol, UK M Frerking

Oregon Health and Science University, Beaverton, OR, USA Z-G Gao

National Institutes of Health, Bethesda, MD, USA

The University of Texas Southwestern Medical Center, Dallas, TX, USA B L Kieffer

IGBMC,CNRS/INSERM/ULP, Illkirch, France E Kim

Korea Advanced Institute of Science and Technology, Daejeon, South Korea H-C Kornau

Center for Molecular Neurobiology (ZMNH), University of Hamburg, Hamburg, Germany

J Garthwaite

University College London, London, UK D R Gehlert

Eli Lilly and Company, Indianapolis, IN, USA C Giaume

Colle`ge de France, Paris, France T Goetz

University of Aberdeen, Aberdeen, UK

A C Kreitzer

University of California at San Francisco, San Francisco, CA, USA F E N LeBeau

Newcastle University, Newcastle upon Tyne, UK L de Lecea

Stanford University School of Medicine, Palo Alto, CA, USA

Contributors vii S-H Lee

C A Meijas-Aponte

University of California at San Francisco, San Francisco, CA, USA

National Institutes of Health, Baltimore, MD, USA

Y-I Lee

State University of New York at Stony Brook, Stony Brook, NY, USA

The Johns Hopkins University School of Medicine, Baltimore, MD, USA J Lerma

Instituto de Neurociencias de Alicante Consejo Superior de Investigaciones Cientı´ficas– Universidad Miguel Hernandez, San Juan de Alicante, Spain J Lerma

Instituto de Neurociencias de Alicante Consejo Superior de Investigaciones Cientı´ficas-Universidad Miguel Hernandez, San Juan de Alicante, Spain Z Li

L M Mendell

N B Mercuri

Universita` di Roma ‘Tor Vergata’, and IRCCS Fondazione Santa Lucia, Rome, Italy A Merighi

University of Turin, Turin, Italy M P Meyer

Kings College London, London, UK A C Michael

University of Pittsburgh, Pittsburgh, PA, USA

Saint Louis University School of Medicine, St. Louis, MO, USA

W C Mobley

J Lisman

I Mody

Brandeis University, Waltham, MA, USA

Neuroscience Institute, Stanford, CA, USA

R-J Liu

David Geffen School of Medicine at UCLA, Los Angeles, CA, USA

Yale School of Medicine, New Haven, CT, USA

J-P Mothet

B J Lopresti

University of Pittsburgh, Pittsburgh, PA, USA

Institut National de la sante´ et de la Recherche Me´dicale U862, Bordeaux, France

D M Lovinger

R Narendran

National Institutes of Health, Rockville, MD, USA

University of Pittsburgh, Pittsburgh, PA, USA

B Lu

C C Naus

National Institutes of Health, Bethesda, MD, USA

University of British Columbia, Vancouver, BC, Canada

M Ludwig

University of Edinburgh, Edinburgh, UK B Lutz

Johannes Gutenberg University, Mainz, Germany

F Nicoletti

University of Rome ‘La Sapienza,’ Rome, Italy S H R Oliet

C Lu¨scher

Institut National de la sante´ et de la Recherche Me´dicale U862, Bordeaux, France

University of Geneva, Geneva, Switzerland

T S Otis

P J Magistretti

Geffen School of Medicine, Los Angles, CA, USA

Ecole Polytechnique Fe´de´rale de Lausanne (EPFL) and Centre Hospitalier Universitaire Vaudois, Lausanne, Switzerland

Y Paas

K L Magleby

University of Missouri – Rolla, Rolla, MO, USA

University of Miami School of Medicine, Miami, FL, USA

L Pellerin

R E Mains

O Peters

University of Connecticut, Farmington CT, USA

Charite´ University Medicine, Berlin, Germany

G Marsicano

R S Petralia

Johannes Gutenberg University, Mainz, Germany and U862 Centre de Recherche INSERM Franc¸ois Magendie Universite´ Bordeaux 2, Bordeaux, France

National Institute on Deafness and Other Communication Disorders, National Institutes of Health, Bethesda, MD, USA

Bar-Ilan University, Ramat-Gan, Israel P Patil

Universite´ de Lausanne, Lausanne, Switzerland

viii

Contributors

J C Petruska

A D Smith

State University of New York at Stony Brook, Stony Brook, NY, USA

University of Pittsburgh, Pittsburgh, PA, USA

M-M Poo

Stanford University, Stanford, CA, USA

University of California at Berkeley, Berkeley, CA, USA

J A Sobota

S Raghavachari

Duke University Medical Center, Durham, NC, USA A G Ramage

University College London, London, UK B R Ransom

University of Washington School of Medicine, Seattle, WA, USA L F Reichardt

University of California at San Francisco, San Francisco, CA, USA R J Reimer

Stanford University School of Medicine, Stanford, CA, USA B M Reuss

University of Go¨ttingen, Go¨ttingen, Germany R R Ribchester

University of Edinburgh, Edinburgh, UK J Rizo

University of Texas Southwestern Medical Center, Dallas, TX, USA

S J Smith

University of Connecticut, Farmington CT, USA S Takamori

Tokyo Medical and Dental University, Tokyo, Japan H Teng

Washington University School of Medicine, St. Louis, MO, USA N Teramoto

Kyushu University, Fukuoka, Japan S F Traynelis

Emory University, Atlanta, GA, USA A Triller

INSERM U789, Ecole Normale Supe´rieure, Paris, France R W Tsien

Stanford University Medical Center, Stanford, CA, USA M H Tuszynski

University of California at San Diego, La Jolla, CA, USA W W Vale

E Robles

The Salk Institute for Biological Studies, La Jolla, CA, USA

Stanford University, Stanford, CA, USA

C Vannier

R H Roth

Yale University School of Medicine, New Haven, CT, USA

INSERM U789, Ecole Normale Supe´rieure, Paris, France A Volterra

V C Russo

University of Lausanne, Lausanne, Switzerland

Royal Children’s Hospital and University of Melbourne, Parkville, VIC, Australia

R D Wassall

M Saarma

B Waterhouse

University of Helsinki, Helsinki, Finland

Drexel University College of Medicine, Philadelphia, PA, USA

V Scheuss

Max-Planck-Institute for Neurobiology, Martinsried, Germany C G Schipke

Charite´ University Medicine, Berlin, Germany

University of Oxford, Oxford, UK

R J Wenthold

National Institute on Deafness and Other Communication Disorders, National Institutes of Health, Bethesda, MD, USA M J Werle

University of Toronto,Toronto, ON, Canada

University of Kansas Medical Center, Kansas City, KS, USA

C R Slater

G A Werther

University of Newcastle upon Tyne, Newcastle upon Tyne, UK

Royal Children’s Hospital and University of Melbourne, Parkville, VIC, Australia

P Seeman

Contributors ix T C Westfall

Neuroscience Institute, Stanford, CA, USA

St. Louis University School of Medicine, St. Louis, MO, USA

P Wulff

University of Aberdeen, Aberdeen, UK

R S Wilkinson

Washington University School of Medicine, St. Louis, MO, USA G Wilson

University College London, London, UK D G Winder

Vanderbilt University School of Medicine, Nashville, TN, USA

Z Ye

University of Washington School of Medicine, Seattle, WA, USA H Yuan

Emory University, Atlanta, GA, USA W Zieglga¨nsberger

W Wisden

Max Planck Institute of Psychiatry, Munich, Germany

University of Aberdeen, Aberdeen, UK

M J Zigmond

N H Woo

University of Pittsburgh, Pittsburgh, PA, USA

National Institutes of Health, Bethesda, MD, USA C Wu

K Zito

University of California at Davis, Davis, CA, USA

This page intentionally left blank

CONTENTS

Contributors Contents Preface

v–ix xi–xv xvii–xviii

SECTION I: BASIC MECHANISMS OF SYNAPTIC TRANSMISSION (SYNAPTIC STRUCTURE AND ORGANIZATION) Synaptic Precursors: Filopodia E Robles, S J Smith, and M P Meyer

3

Presynaptic Development: Functional and Morphological Organization D Atasoy and E T Kavalali

11

Postsynaptic Development: Neuronal Molecular Scaffolds E Kim

19

Dendrite Development, Synapse Formation and Elimination H Cline

27

Synapse Formation: Competition and the Role of Activity L Cancedda and M-M Poo

31

Cell Adhesion Molecules at Synapses L F Reichardt and S-H Lee

38

Glia and Synapse Formation: An Overview N J Allen and B A Barres

46

Active Zone P S Kaeser

52

Calcium Channel Subtypes Involved in Neurotransmitter Release R W Tsien

59

SNAREs J Rizo

67

Synaptic Vesicles S Takamori

76

xi

xii Contents

Endocytosis and Presynaptic Scaffolds V Haucke and E D Gundelfinger

84

Postsynaptic Density/Architecture at Excitatory Synapses H-C Kornau

96

Synaptic Transmission: Models S Raghavachari and J Lisman

103

Glial Influence on Synaptic Transmission C G Schipke and O Peters

112

Neurotransmitter Release from Astrocytes A Volterra

120

Retrograde Transsynaptic Influences G W Davis

126

Synaptic Plasticity: Short-Term Mechanisms J S Dittman and A C Kreitzer

132

SECTION II: NEUROMUSCULAR AND GAP JUNCTIONS Neuromuscular Connections: Vertebrate Patterns of C R Slater

141

Neuromuscular Junction: Synapse Elimination R R Ribchester

150

Presynaptic Events in Neuromuscular Transmission H Teng and R S Wilkinson

158

Neuromuscular Junction (NMJ): Presynaptic Short-Term Plasticity of Neuromuscular Transmission A M Holohean and K L Magleby

168

Neuromuscular Junction (NMJ): Postsynaptic Basal Lamina M J Werle

174

Neuromuscular Junction (NMJ): Postsynaptic Events in Neuromuscular Transmission C R Slater

180

Gap Junction Communication C Giaume and C C Naus

189

Gap Junctions and Electrical Synapses M V L Bennett

193

Gap Junctions and Hemichannels in Glia Z Ye and B R Ransom

213

Gap Junctions and Neuronal Oscillations M O Cunningham and F E N LeBeau

219

SECTION III: AMINO ACID TRANSMITTERS AND RECEPTORS Glutamate S P H Alexander

229

Glial Energy Metabolism: Overview L Pellerin and P J Magistretti

239

Contents xiii

Transporter Proteins in Neurons and Glia T S Otis and P D Dodson

245

Vesicular Neurotransmitter Transporters R J Reimer

253

AMPA Receptors: Molecular Biology and Pharmacology S M Dravid, H Yuan, and S F Traynelis

260

AMPA Receptor Cell Biology/Trafficking P G R Hastie and J M Henley

268

NMDA Receptor Function and Physiological Modulation K Zito and V Scheuss

276

NMDA Receptors, Cell Biology and Trafficking R J Wenthold and R S Petralia

284

Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology F Nicoletti, V Bruno, and G Battaglia

292

Metabotropic Glutamate Receptors (mGluRs): Functions B A Grueter and D G Winder

302

Kainate Receptors: Molecular and Cell Biology J Lerma

308

Kainate Receptor Functions J Lerma

313

Long-Term Potentiation (LTP): NMDA Receptor Role A J Doherty, S M Fitzjohn, and G L Collingridge

321

Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDAR-Dependent Forms C Lu¨scher and M Frerking

327

D-Serine:

From Its Synthesis in Glial Cell to Its Action on Synaptic Transmission and Plasticity S H R Oliet and J-P Mothet

333

GABA Synthesis and Metabolism K L Behar

340

GABAA Receptors: Molecular Biology, Cell Biology, and Pharmacology T Goetz, P Wulff, and W Wisden

347

GABAA Receptor Synaptic Functions I Mody

355

GABAB Receptors: Molecular Biology and Pharmacology N G Bowery

360

GABAB Receptor Function A J Doherty, G L Collingridge, and S M Fitzjohn

368

Glycine Receptors: Molecular and Cell Biology C Vannier and A Triller

373

SECTION IV: AMINES AND ACETYLCHOLINE Dopamine J D Elsworth and R H Roth

383

xiv

Contents

Dopamine Receptors and Antipsychotic Drugs in Health and Disease P Seeman

392

Dopamine: Cellular Actions G Bernardi and N B Mercuri

410

Noradrenaline R D Wassall, N Teramoto, and T C Cunnane

414

Norepinephrine: Adrenergic Receptors D B Bylund

424

Norepinephrine: CNS Pathways and Neurophysiology G Aston-Jones, C A Meijas-Aponte, and B Waterhouse

430

Monoamines: Release Studies A D Smith, A C Michael, B J Lopresti, R Narendran, and M J Zigmond

442

Serotonin (5-Hydroxytryptamine; 5-HT): Neurotransmission and Neuromodulation A G Ramage

450

Serotonin (5-Hydroxytryptamine; 5-HT): Receptors T P Blackburn

456

Serotonin (5-Hydroxytryptamine; 5-HT): CNS Pathways and Neurophysiology G Aghajanian and R-J Liu

470

Cholinergic Neurotransmission in the Autonomic and Somatic Motor Nervous System T C Westfall

478

Cholinergic Pathways in CNS A C Cuello

486

Muscarinic Receptors: Autonomic Neurons R S Aronstam and P Patil

494

Nicotinic Acetylcholine Receptors J-P Changeux and Y Paas

503

SECTION V: NEUROPEPTIDES AND NEUROTROPHIC FACTORS Neuropeptide Synthesis and Storage J A Sobota, B A Eipper, and R E Mains

511

Neuropeptide Release F Bergquist and M Ludwig

519

Neuropeptides and Coexistence A Merighi

525

Opioid Peptides and Receptors B L Kieffer

532

Neuropeptide Y (NPY) and its Receptors D R Gehlert

538

Corticotropin-Releasing Hormone and Urocortins: Binding Proteins and Receptors M O Huising and W W Vale

544

Hypocretin/Orexin and MCH and Receptors A Adamantidis and L de Lecea

551

Peptidergic Receptors G Casini and D Cervia

557

Contents xv

Neuropeptides: Electrophysiology W Zieglga¨nsberger

564

Neurotrophins: Physiology and Pharmacology J M Conner and M H Tuszynski

570

Nerve Growth Factor J C Petruska and L M Mendell

576

Retrograde Neurotrophic Signaling C Wu and W C Mobley

584

BDNF in Synaptic Plasticity and Memory N H Woo and B Lu

590

GFL Neurotrophic Factors: Physiology and Pharmacology M Saarma

599

Insulin-Like Growth Factor Signaling and Actions in Brain V C Russo and G A Werther

609

Glial Growth Factors B M Reuss

617

SECTION VI: ATYPICAL NEUROTRANSMITTERS Adenosine K A Jacobson and Z-G Gao

627

Adenosine Triphosphate (ATP) G Burnstock

639

Purines and Purinoceptors: Molecular Biology Overview G Burnstock

648

P2X Receptors Z Li, S Harris, and T M Egan

658

Endocannabinoid Role in Synaptic Plasticity and Learning B Lutz and G Marsicano

664

Long-Term Depression (LTD): Endocannabinoids and Cerebellar LTD D M Lovinger

677

Nitric Oxide G Wilson and J Garthwaite

684

Role of NO in Neurodegeneration Y-I Lee, T M Dawson, and V L Dawson

690

Index

697

This page intentionally left blank

PREFACE

The brain is often considered an information storage and processing machine. It receives information from the external and internal worlds and processes this information to generate specific thoughts, feelings and behaviors, which are often based on the information stored from past experiences. The performance of these tasks depends entirely on the bewildering complexity of intercellular communication in the nervous system; the ability of millions of individual nerve cells to communicate with one another and process information via countless numbers (billions or trillions) of different types of synapses. Intercellular communication is not random but for any given task that the brain performs occurs in specific neural circuits composed of complex patterns of synaptically interconnected cells. In addition, while the vast majority of intercellular communication in the brain occurs via synapses, we now know that glia play a critical role in shaping intercellular communication and that neural circuit activity can also be influenced by additional factors that are not released at specific synaptic connections in a point to point fashion. This book presents a series of articles that attempts to summarize the key features of the major types of intercellular communication that occur in the mammalian brain. It is composed of articles obtained from the Encyclopedia of Neuroscience, a comprehensive and, at times, overwhelming compendium of 1465 chapters that summarize all aspects of modern neuroscience, from the molecular biology and crystal structures of ion channels to the neural circuit basis of higher cognitive functions such as attention. Our current understanding of the pathophysiology of a wide variety of disease states ranging from Alzheimer’s Diseases to schizophrenia is also covered in the Encyclopedia. Clearly, this is a valuable entity that provides a unique reference library but it is difficult to wade through if the reader wants an update on current understanding of a broad, but relatively defined topic such as intercellular

communication. This book fills this need in what I hope is a logical and comprehensive manner. As the synapse remains the primary point of information processing and communication between nerve cells, the book begins with a summary of current understanding of synapse formation and basic synapse structure and function. Several chapters in this first section address the important topic of the role of glia, which are now known to play important roles in synaptogeneis and synaptic transmission. The second section of the book summarizes current knowledge of the most classic of synapses, the neuromuscular junction as well as gap junctions, that play important roles in controlling neural circuit activity in key brain regions. We then move on, in the third section, to the major types of excitatory and inhibitory synapses, those using the neurotransmitters glutamate and GABA (as well as glycine). The subsection on glutamate is by far the longest in the book because in mammalian brains, >75% of all synapses are excitatory and use glutamate as their neurotransmitter. Glutamatergic synapses are the workhorses of the brain, carrying out virtually all of its critical tasks. They also are highly plastic so that the circuits in which they are embedded can be modified by experience and store these experiences as memories. Inhibitory, GABA-using synapses, on the other hand help shape circuit properties in important ways and thus are also an important topic. In Seciton IV, we move on to forms of intercellular communication that involve what are often called neuromodulators. These include the major amines, dopamine, norepinephrine and serotonin as well as acetycholine. Individual chapters review the molecular biology and pharmacology of the receptors for these neuromodulators, their anatomical distribution and connectivity, as well as their varied cellular physiological effects and nervous system functions. The large topic of neuropeptides and neurotrophic factors is reviewed

xvii

xviii

Preface

in Section V. This includes chapters on neuropeptide processing and release as well as on the receptors and actions of a large number of individual neuropeptides and neurotrophic factors. Current knowledge of the functions of neuropeptides and neurotrophic factors is advancing rapidly so it should not be surprising if the latest advances on certain restricted topics are not fully covered. Nonetheless, for the non-expert, this section will provide a comprehensive overview of this broad research area. The book concludes with so-called Atypical Neurotransmitters including the purines, ATP and adenosine, endocannabinoids and nitric oxide. Several of these are not released in the same manner as the classical neurotransmitters and exert their actions in a larger volume of brain tissue. Clearly, several topics of great relevance to Intercellular Communication in the Nervous System, such as Neuroendocrinology and Neuroimmunology, are

not covered in this book. It was felt that broad and important topics such as these deserve a more comprehensive treatment than could be accomplished herein. Nonetheless, we hope that the reader who reads this text cover to cover will leave with a thorough and current understanding of the detailed mechanisms by which intercellular communication in the brain occurs and how it functions to mediate the amazing tasks the brain accomplishes each and every day. We also hope that it serves as a valuable reference for readers to look up details on the many different highly specific topics that are covered by the 87 chapters. Both the publishers and I (the editor) think this is a compendium that will prove valuable to a broad array of neuroscientists and biologists. After delving into it, we hope you agree. Robert C. Malenka Editor-in-Chief

BASIC MECHANISMS OF SYNAPTIC TRANSMISSION (SYNAPTIC STRUCTURE AND ORGANIZATION)

A. Synapse Formation B. Synapse Structure and Function

3 52

This page intentionally left blank

Synaptic Precursors: Filopodia E Robles and S J Smith, Stanford University, Stanford, CA, USA M P Meyer, Kings College London, London, UK ã 2009 Published by Elsevier Ltd.

Filopodia as Synaptic Precursors The vertebrate central nervous system (CNS) contains billions of neurons connected by an even greater number of synapses. Despite this complexity, the CNS develops reliably and precisely. The goal of developmental neurobiologists is to understand the mechanisms that control the accurate formation of synaptic connections in the developing brain. This article will review evidence that axonal and dendritic filopodia are fundamental players in the assembly of functional neural circuits. As will become apparent during the course of this article, the highly motile nature of filopodia and the relationship of filopodia to synaptogenesis are central to their role in generating a precisely ordered nervous system. This article is structured, therefore, around the participation of filopodia in synapse formation, the regulation of filopodial motility, and the way such regulation is fundamental to the genesis of appropriate and ordered synaptic connections. In part one we concentrate on a phenomenological discussion of the role of filopodia in synaptogenesis and growth and branching of axonal and dendritic arbors. This is followed by a discussion of intrinsic and extrinsic factors that may regulate the formation of synaptic contacts by altering filopodial motility in developing neurons. Dendritic Filopodia as Spine Precursors

Dendritic spines were discovered in 1888 by Ramon y Cajal, who noticed that Purkinje cells’ dendrites were decorated with small thorns (‘espinas’). He later proposed that these small, submicrometer protrusions from the parent dendrite were the site of axodendritic contact. Cajal’s hypothesis was proved correct 60 years later, when the electron microscope was employed to show that spines were indeed sites of synaptic contact between axons and dendrites. In fact, more recent estimates suggest that more than 90% of excitatory axodendritic synapses in the mature CNS occur on dendritic spines. Spines must therefore be key elements in neuronal circuitry, and their structure suggests that one fundamental function of spines may be to bridge physical gaps between densely packed dendrites and axons. Hence, the mechanisms governing spine morphogenesis likely play a fundamental role in the selection

of synaptic partners and establishment of functional connectivity. A clue to the developmental origins of spines comes from the observations that prior to the overt presence of spines, many elongated filopodia extend from dendritic shafts. These filopodia are long (2–20 mm), thin (100 mM), both kinds of channels must be blocked in order to strongly inhibit neurosecretion. Circles, no channel inhibition; triangles, N-type channels blocked; upside down triangles, P/Q-type channels blocked; diamonds, both N- and P/Q-type channels blocked. Reproduced from Turner TJ, Adams ME, and Dunlap K (1993) Multiple Ca2þ channel types coexist to regulate synaptosomal neurotransmitter release. Proceedings of the National Academy of Sciences of the United States of America 90(20): 9518–9522, with permission.

64 Calcium Channel Subtypes Involved in Neurotransmitter Release

through either type of channel is sufficient to drive secretion, and the other type can be regarded as ‘spare channels.’ The synergistic or redundant actions of multiple channel types can be observed in the same experiment (Figure 4). The switch between cooperative and redundant behavior can be accomplished simply by prolongation of presynaptic action potentials with the Kþ channel blocker 4-aminopyridine. Prolongation of the action potential causes more activation of each type of Ca2þ channel, favoring its selfsufficiency.

The detailed composition of Ca2þ channel subtypes that mediate transmission can vary widely among individual nerve terminals.

neuromuscular junction and squid giant synapse, in which P- or P/Q-type channels are overwhelmingly prevalent. At some hippocampal inhibitory synapses, GABA release is mediated entirely by P/Q-type or by N-type Ca2þ channels. Neurotransmission at sympathetic neuron–effector junctions is governed almost completely by N-type channels. Differences in the reliance on various CaV2 family members take on additional significance when synaptic circuits are the target of pharmacological and possibly therapeutic intervention. For example, in the dorsal horn of the spinal cord, the main component in the circuitry of pain transmission/modulation, inhibitory synaptic transmission is dominated by P/Q-type Ca2þ channels; in contrast, N-type channels are more closely associated with glutamate release from the peptidergic primary sensory neurons that convey nociceptive information to the dorsal horn.

Not always Multiple Channel Types

Not always CaV2 Channels

Although neurosecretion is supported by more than one type of channel at most CNS synapses, at some synapses transmission may be heavily dominated by a single channel type. Examples include some preparations historically important for discovery of basic principles of neurotransmission such as mammalian

It is generally believed that Ca2þ channels of the CaV2 class are specialized to support transmitter release. This is true, but it is worth pointing out that in some cases, CaV1 and even CaV3 channels can also participate. For example, ribbon synapses in the retina and cochlea rely on CaV1.3 channels to trigger transmitter

Exceptions to Generalizations about Multiple Ca2+ Channel Types

w -CTx-GVIA (1 µM) w -Aga-IVA (1 µM) 120 4-AP 100 80 60 4-AP 40 Control 20 Control 0 −5 0 5 10 15 20 −5 0 5 10 15 20 25 30 a b

EPSP slope (V/s) c

PQ

Synaptic vesicle

PQ N

PQ N

R

d 4-AP (100 µM)

4-AP (100 µM)

w -Aga-IVA (1 µM)

w -CTx-GVIA (1 µM)

−1.6

PQ PQ

−1.2 −0.8 −0.4 0

0

60

120

180 Time (min)

240

300

360

Figure 4 Concerted or redundant actions of P/Q- and N-type channels, depending on strength of depolarizing stimulus. (a) Dependence of synaptic transmission on N-type Ca2þ channels changes markedly upon prolongation of the presynaptic action potential with 4-AP. Control graph (open circles) shows approximately 50% inhibition of basal transmission by o-CTx-GVIA. In 4-AP (solid circles), the effect of N-type channel blockade is barely evident. (b) 4-AP also reduces the degree of inhibition produced by P/Q-type channel inhibition with o-Aga-IVA. (c) Example of an experiment in which simultaneous blockade of N-type and P/Q-type Ca2þ channels with o-CTx-GVIA and o-Aga-IVA abolished transmission completely, even in the presence of 4-AP. (d) Schematic representation of a release site surrounded by multiple types of Ca2þ channels. (a–c) Reproduced from Wheeler DB, Randall A, and Tsien RW (1996) Changes in action potential duration alter reliance of excitatory synaptic transmission on multiple types of Ca2þ channels in rat hippocampus. Journal of Neuroscience 16: 2226–2237, with permission. (d) Reproduced from Cao Y, Piedras-Renterı´a ES, Smith GB, Chen G, Harata NC, and Tsien RW (2004) Presynaptic Ca2þ channels compete for channel type-preferring slots in altered neurotransmission arising from Ca2þ channelopathy. Neuron 43: 387–400.

Calcium Channel Subtypes Involved in Neurotransmitter Release

release. Furthermore, CaV2 channels can perform important neuronal functions distinct from excitation– secretion coupling, such as excitation–transcription coupling. Not always the Same Type of Ca2+ Channel throughout Development

At some synapses, N-type channels develop early but are gradually supplanted by late-blooming P/Q-type channels. Mammalian development might be echoing a phylogenetic progression seen at the neuromuscular junction, where lizards and frogs use L- and N-type channels, respectively, but mammals deploy P/Q-type channels.

What Are Multiple Ca2+ Channels Good for? Why are there different kinds of presynaptic Ca channels? Almost certainly it is not because of evolutionary pressure for different biophysical properties. Indeed, basic features such as extremely high Ca2þ selectivity, high open channel flux rate, and steeply voltage-dependent channel gating are very similar for P/Q-, N-, and R-type currents. In all cases, selectivity and permeation are conferred by a pore that is continually occupied by at least one Ca2þ ion, which prevents more abundant Naþ ions from rushing through. Despite overall similarities, there are substantial differences in channel regulation, and this is likely to be the main reason for channel diversity (Table 1). For example, P/Q- and N-type channels appear to differ in their susceptibility to modulation by G-proteins. N-type channels show a greater degree of modulation. On the other hand, P/Q-type channels are more susceptible to progressive increases with

Table 1 Differences between P/Q- and N-type channels with regard to their properties and contributions to synaptic transmission P/Q-type channels

N-type channels

Specific blocker

o-AgaIVA

Genetic diseases

Migraine, ataxia, epilepsy Late blooming Similar Weaker

o-CTx-GVIA, o-CTx-MVIIA ¼ Prialt No channelopathies found Early prominence Similar Stronger

Stronger

Weaker if at all

Development Voltage gating Modulation by G-protein-coupled receptors Facilitation by repeated stimuli

65

use, a phenomenon known as Ca2þ channel facilitation. Because of the steep power-law relationship between Ca2þ entry and secretion, even small increases in P/Q Ca2þ entry can have a profound effect on the strength of transmission. The presynaptic Ca2þ channels constitute a key convergence point for regulation of neurosecretion, either by neurohumoral modulation or by activity-dependent regulation. Much is known about the structural motifs in a1 subunits that support various forms of regulation, for example, by G-proteins and protein kinases.

Further Reading Bollmann JH and Sakmann B (2005) Control of synaptic strength and timing by the release-site Ca2þ signal. Nature Neuroscience 8: 426–434. Cao Y, Piedras-Renterı´a ES, Smith GB, Chen G, Harata NC, and Tsien RW (2004) Presynaptic Ca2þ channels compete for channel type-preferring slots in altered neurotransmission arising from Ca2þ channelopathy. Neuron 43: 387–400. Catterall WA (1996) Molecular properties of sodium and calcium channels. Journal of Bioenergetics and Biomembranes 28: 219–230. Dolphin AC (2006) A short history of voltage-gated calcium channels. British Journal of Pharmacology 147(supplement 1): S56–S62. Dunlap K, Luebke JI, and Turner TJ (1995) Exocytotic Ca2þ channels in mammalian central neurons. Trends in Neurosciences 18: 89–98. Ertel EA, Campbell KP, Harpold MM, et al. (2000) Nomenclature of voltage-gated calcium channels. Neuron 25(3): 533–535. Lisman JE, Raghavachari S, and Tsien RW (2007) The sequence of events that underlie quantal transmission at central glutamatergic synapses. Nature Reviews Neuroscience 8: 597–609. Llina´s R, Sugimori M, Hillman DE, and Cherksey B (1992) Distribution and functional significance of the P-type, voltage-dependent Ca2þ channels in the mammalian central nervous system. Trends in Neurosciences 15: 351–355. Miljanich GP (2004) Ziconotide: Neuronal calcium channel blocker for treating severe chronic pain. Current Medicinal Chemistry 11: 3029–3040. Nowycky MC, Fox AP, and Tsien RW (1985) Three types of neuronal calcium channel with different calcium agonist sensitivity. Nature 316: 440–443. Olivera BM, Miljanich GP, Ramachandran J, and Adams ME (1994) Calcium channel diversity and neurotransmitter release: The o-conotoxins and o-agatoxins. Annual Review of Biochemistry 63: 823–867. Piedras-Renterı´a ES, Barrett CF, Cao Y-Q, and Tsien RW (2007) Voltage-gated calcium channels, calcium signaling and channelopathies. In: Krebs J and Michalak M (eds.) Calcium: A Matter of Life or Death, pp. 127–166. Elsevier: New York. Schneggenburger R and Neher E (2005) Presynaptic calcium and control of vesicle fusion. Current Opinion in Neurobiology 15: 266–274. Takahashi T and Momiyama A (1993) Different types of calcium channels mediate central synaptic transmission. Nature 366: 156–158. Tsien RW and Barrett CF (2004) A brief history of calcium channel discovery. In: Zamponi G (ed.) Voltage-Gated Calcium Channels. New York: Kluwer/Plenum.

66 Calcium Channel Subtypes Involved in Neurotransmitter Release Turner TJ, Adams ME, and Dunlap K (1993) Multiple Ca2þ channel types coexist to regulate synaptosomal neurotransmitter release. Proceedings of the National Academy of Sciences of the United States of America 90(20): 9518–9522.

Wheeler DB, Randall A, and Tsien RW (1996) Changes in action potential duration alter reliance of excitatory synaptic transmission on multiple types of Ca2þ channels in rat hippocampus. Journal of Neuroscience 16: 2226–2237.

SNAREs J Rizo, University of Texas Southwestern Medical Center, Dallas, TX, USA ã 2009 Elsevier Ltd. All rights reserved.

Introduction Chemical synaptic transmission is mediated by neurotransmitters that are released by Ca2þ-triggered synaptic vesicle exocytosis. The exquisite spatial and temporal regulation of this process depends on a highly complex protein machinery that is formed in part by components with homologs in most types of intracellular membrane traffic, including N-ethylmaleimide sensitive factor (NSF), soluble NSF attachment proteins (SNAPs), SNAP receptors (SNAREs), Sec1/Munc18 (SM) proteins, and Rab guanosine triphosphatases (GTPases). Among these components, the SNAREs are believed to be at the heart of a conserved mechanism of membrane fusion by virtue of their ability to form tight SNARE complexes that bring two opposing membranes together. The neuronal SNAREs involved in neurotransmitter release are the synaptic vesicle protein synaptobrevin 2 (also known as vesicle associated protein 2 (VAMP2)) and the plasma membrane proteins syntaxin 1 (which in mammals includes two closely related isoforms, syntaxin 1A and 1B) and SNAP-25 (for synaptosomal associated protein of 25 kDa; no relation to SNAPs). Neurotransmitter release depends also on proteins such as synaptotagmin 1 and complexins that play specialized roles in the tight regulation of this process and are not generally involved in other types of intracellular membrane traffic. Several of these proteins bind to the neuronal SNAREs and probably function in the regulation of SNARE complex assembly and/or in conjunction with the SNARE complex. Hence, synaptobrevin 2, syntaxin 1, and SNAP-25 have unique properties that have been adapted to the regulatory requirements of synaptic exocytosis, in addition to conserved features that are shared with SNAREs from other membrane compartments. This article focuses on the structural, biochemical, and functional properties of the neuronal SNAREs, but it also describes key findings on other SNAREs that have helped us to understand which of these properties are general and which are specialized.

SNARE Structure SNAREs constitute a family of proteins characterized by sequences called SNARE motifs that comprise  60–70 residues and have a high propensity to

form coiled coils. Most SNAREs contain only one SNARE motif that is adjacent to a single C-terminal transmembrane (TM) region (e.g., synaptobrevin 2 and syntaxin 1). Some SNAREs contain two SNARE motifs connected by a long linker and do not have a TM sequence (e.g., SNAP-25) (Figure 1), but are attached to membranes through a posttranslational modification such as palmitoylation. Circular dichroism (CD) and nuclear magnetic resonance (NMR) studies showed that isolated SNARE motifs are generally unstructured and adopt a-helical conformations when they bind to other SNARE motifs to form SNARE complexes. A SNARE complex contains four SNARE motifs that assemble into a parallel four-helix bundle, as first demonstrated for the neuronal SNAREs by electron paramagnetic resonance (EPR) spectroscopy and X-ray crystallography (Figure 2(a)). The neuronal SNARE complex is very stable (it is resistant to sodium dodecyl sulfate (SDS) and requires high temperatures for denaturation). This stability is conferred in large part by contacts between many hydrophobic residues that are arranged in layers, with each SNARE motif contributing one hydrophobic side chain to each layer. However, a polar, or zero layer, in the middle of the bundle is formed by one arginine (from synaptobrevin 2) and three glutamine side chains (one from syntaxin 1 and two from SNAP-25) (Figure 2(b)). These features are highly conserved in four different subfamilies of SNAREs and have led to their classification into Qa, Qb, Qc, and R SNAREs, depending on their homology with the SNARE motifs of syntaxin 1 (Qa), SNAP-25 (Qb and Qc for its N- and C-terminal SNARE motifs, respectively), or synaptobrevin (R). This classification is less ambiguous for SNAREs involved in homotypic membrane fusion than the original classification into v- and t-SNAREs (for vesicular and target membrane SNAREs, respectively). The assembly of the SNARE four-helix bundle is most likely at the center of the mechanism of membrane fusion. The functional importance of the polar layer is currently unclear, but it has been suggested to play a role in aiding SNARE complex disassembly or in dictating the proper register for SNARE complex formation. In addition to a SNARE motif and a TM sequence, many SNAREs have an N-terminal region that spans more than half of its sequence. NMR studies have shown that the N-terminal regions of all SNAREs from the syntaxin subfamily, including syntaxin 1, contain an autonomously folded domain that adopts an antiparallel three-helix bundle structure, known as the Habc domain (Figure 2(a)). This domain is connected to the SNARE motif through a linker region and is preceded by a short N-terminal

67

68 SNAREs

sequence (NTS) (Figure 1). The N-terminal regions of SNAREs from other subfamilies may also contain Habc domains or autonomously folded domains with completely unrelated structures, such as longin domains or PX domains. In the assembled SNARE complex, the Habc domain of syntaxin 1 is flexibly linked to the

Synaptobrevin 2 TM

SNARE motif Syntaxin 1 NTS

SNARE motif

Habc domain

TM

SNAP-25 SNARE motif

SNARE motif

Figure 1 Domain structures of the neuronal SNAREs. The SNARE motifs, TM regions, and the helices of the Habc domain are represented by cylinders; other sequences are represented by black lines. NTS, N-terminal sequence; SNAP-25, synaptosomal associated protein of 25 kDa; SNARE, soluble N-ethylmaleimidesensitive factor attached protein receptor; TM, transmembrane.

four-helix bundle formed by the SNARE motifs. However, the Habc domain binds intramolecularly to the SNARE motif in isolated syntaxin 1, forming a closed conformation that is incompatible with the SNARE complex. These findings suggested that the Habc domain regulates SNARE complex formation and that syntaxin 1 must undergo a large conformational transition from a closed to an open conformation during exocytosis (Figure 3). However, only a few SNAREs from the syntaxin subfamily adopt a closed conformation, which thus appears to represent a specialization arising from unique regulatory requirements of neurotransmitter release and a few other membrane traffic processes. Conversely, some SNAREs outside the syntaxin family adopt closed conformations involving different types of autonomously folded N-terminal domains, indicating that intramolecular regulation of SNARE complex assembly is not restricted to syntaxins. The closed conformation of syntaxin 1 binds tightly to the neuronal SM protein Munc18–1 and its structure at atomic resolution (Figure 2(c)) was revealed by X-ray analysis of a syntaxin 1–Munc18–1 complex. This mode of interaction is not generally conserved

N c

a

V223 L50 V53 Q226

I171

Q53

R56

I230

Q174 L60

c L57

I178

b

N c

Figure 2 Three-dimensional structures of the neuronal SNAREs: (a) illustration of the three-dimensional structures of the syntaxin 1 Habc domain determined by NMR spectroscopy (Fernandez et al.) (left) and the neuronal SNARE complex formed by the SNARE motifs of synaptobrevin 2, syntaxin 1, and SNAP-25 determined by X-ray crystallography (Sutton et al.) (right); (b) stick models of selected layers forming the neuronal SNARE complex; (c) illustration of the structure of the closed conformation of syntaxin 1 bound to Munc18–1 determined by X-ray crystallography. In (a), the dashed curved represents the flexible linker between the Habc domain and SNARE motif of syntaxin 1. The color coding is the same as in Figure 1. In (b). the stick models include the polar layer formed by R56 of synaptobrevin 2, Q226 of syntaxin 1, and Q53 and Q174 of SNAP-25, as well as the preceding and ensuing layers, which are formed exclusively by hydrophobic residues. The backbone atoms are in the same color code as in Figure 1. Side chain atoms are in cyan for carbon, red for oxygen, and blue for nitrogen. In (c), the Habc domain of syntaxin 1, its SNARE motif, and the linker region connecting them are in orange, yellow, and black, respectively. Munc18–1 is in violet (Misura et al.). C, C-terminus; N, N-terminus; NMR, nuclear magnetic resonance; SNAP-25, synaptosomal associated protein of 25 kDa; SNARE, soluble N-ethylmaleimide-sensitive factor attached protein receptor.

SNAREs 69 Plasma membrane

N

Plasma membrane

N

Synaptic vesicle Figure 3 Diagram illustrating that syntaxin 1 must undergo a large conformational change during synaptic vesicle exocytosis. In the SNARE complex (right), the Habc domain of syntaxin 1 is flexibly linked to the four-helix bundle formed by the SNARE motifs of syntaxin 1, SNAP-25, and synaptobrevin 2. In isolated syntaxin 1 (left), the Habc domain binds intramolecularly to part of the SNARE motif, forming a closed conformation that is incompatible with SNARE complex formation. The color coding is the same as in Figure 1. N, N-terminus; SNAP-25, synaptosomal associated protein of 25 kDa; SNARE, soluble N-ethylmaleimide-sensitive factor attached protein receptor.

in other syntaxins and SM proteins, but many SM protein/syntaxin interactions depend on the syntaxin N-terminal region, particularly the NTS. Hence, it is likely that the general function of the syntaxin N-terminal region is related to coupling with SM protein function.

Evolution of Ideas on SNARE Function SNARE proteins have been investigated by a wide variety of techniques, providing a vivid example of the power of interdisciplinary research and a fascinating history with twists and turns that may continue for years to come. The first SNAREs were found in genetic screens for membrane traffic defects in yeast and in molecular cloning analyses of components from synaptic vesicle and synaptosomal membranes that identified syntaxin 1, synaptobrevin 2, and SNAP-25. The first evidence of the functional importance of these proteins for synaptic exocytosis was provided by the discovery in the early 1990s that they are the targets of clostridial neurotoxins, agents that potently inhibit neurotransmitter release. Around the same time, sequence analyses revealed a homology between these neuronal proteins and some of the proteins identified in the studies of membrane traffic defects in yeast. Moreover, NSF/Sec18p and SNAPs/Sec17p were identified as crucial factors for vesicle-mediated transport in yeast and mammalian cells. Homologies were also observed between yeast and neuronal Rab proteins, as well as between SM proteins involved in traffic at different membrane compartments. All these findings led to the now widely held notion that a conserved machinery mediates most types of intracellular membrane traffic. Crucial discoveries in the understanding of SNARE function were also the observations in 1993 that synaptobrevin 2, syntaxin 1, and SNAP-25 form a tight 7S complex (the SNARE complex) that binds to SNAPs and NSF to form a larger, 20S complex, and that adenosine triphosphate (ATP) hydrolysis by NSF

leads to SNARE complex disassembly. These findings, and the belief that NSF was directly involved in membrane fusion, led to the proposal that synaptobrevin 2, syntaxin 1, and SNAP-25 act as receptors for the membrane fusion apparatus and hence to their designation as SNAREs (for SNAP receptors). These results led also to a model of intracellular membrane traffic whereby binding of SNAREs in transport vesicles (v-SNAREs) to SNAREs on target membranes (t-SNAREs) mediates vesicle docking and target specificity (known as the SNARE hypothesis). In this model, the SNAREs were envisaged as binding in an antiparallel fashion and the disassembly of the SNARE complex by NSF/SNAPs was proposed to initiate fusion. The finding that clostridial neurotoxins inhibit release by cleaving SNAREs from the synaptic vesicle and plasma membrane was consistent with the notion that SNAREs are required on opposing membranes for fusion, which was further supported by subsequent studies of yeast vacuolar fusion. However, it was noted that the SNAREs could not play a role in vesicle docking because docking is not affected on SNARE cleavage by the neurotoxins or by genetic ablation of syntaxin 1 in Drosophila (note that very recent data have suggested that syntaxin 1 may actually function in docking; these contradictory conclusions that still need to be resolved probably arise because redundant docking mechanisms may exist and/or because of differences in the methodology used to prepare samples for electron microscopy (EM)). The genetic studies in Drosophila also revealed that syntaxin 1 is critical for release and, together with the neurotoxin data, led to the proposal that SNAREs function downstream of docking. Moreover, studies of exocytosis in PC12 cells and of homotypic vacuolar fusion in yeast showed in 1996 that NSF functions at a prefusion step rather than fusion itself, and the SNARE motifs of synaptobrevin 2 and syntaxin 1 were shown in 1997 to interact in an antiparallel fashion by EM and fluorescence resonance energy transfer (FRET). Because the SNARE motifs are

70 SNAREs

SNAREs and Membrane Fusion The minimal model of SNARE-mediated membrane fusion was based on reconstitution experiments that demonstrated lipid mixing between proteoliposome populations containing synaptobrevin 2 or syntaxin 1/SNAP-25. This model has been accepted by many researchers, but has also been strongly challenged by others for diverse reasons. On one hand, there were

a

b

c Ca2+

– –+ +

– + – +

+ + + +

+

adjacent to the TM regions of synaptobrevin 2 and syntaxin 1, which are anchored on the synaptic vesicle and plasma membrane, respectively, this key discovery led to the proposal that the energy of formation of the SNARE complex could be used to bring the membranes together and initiate membrane fusion (known as the zippering model; Figure 4(a)). In this model, disassembly of the SNARE complex by NSF/SNAPs after membrane fusion recycles the SNAREs for another round of fusion. The zippering model is attractive because of its similarity to the mechanism of viral fusion and has gained wide acceptance. A similar model that in addition postulates that the SNAREs constitute a minimal membrane-fusion machinery was proposed based on reconstitution experiments, but this minimal model has been strongly challenged by subsequent reconstitutions. Further debate about the central role of the SNAREs in membrane fusion predicted by both models emerged from analyses of mice lacking synaptobrevin 2 or SNAP-25. The synaptobrevin knockout revealed a strong impairment of evoked release and a less marked decrease in spontaneous or hypertonic sucrose-induced release; evoked release was also impaired strongly in SNAP-25-knockout mice, but spontaneous release was increased. Although these results reinforced the notion that the SNAREs function downstream of vesicle docking, the persistence of release in these mice contrasts with the total abrogation of any form of release in Munc18–1-knockout mice. It seems likely that the release remaining in the absence of synaptobrevin 2 and SNAP-25 may arise from functional redundancy with other isoforms of these proteins, but this hypothesis remains to be validated experimentally. Despite these ongoing debates, there is currently a general consensus that SNARE function is related to their ability to bring two membranes together and that synaptobrevin 2, syntaxin 1, and SNAP-25 play a critical role in neurotransmitter release. In addition to resolving these debates, recent and ongoing research on SNAREs is focusing on investigating the mechanism of SNARE complex assembly and studying how the roles of other key proteins are coupled to SNARE function.

+ + + +

d Figure 4 Models of SNARE function in membrane fusion: (a) original zippering model of SNARE-mediated membrane fusion based on EM data, showing that the SNARE motifs of syntaxin 1 and synaptobrevin 2 bind in a parallel fashion; (b) model showing how isolated assembling SNARE complexes might diffuse to the center of the space between the membranes, which requires much less energy than inducing membrane fusion; (c) same model as (b) with the diffusion prevented because the assembling SNARE complexes are bound to a bulky protein (in violet); (d) model illustrating the notion that complexin (in pink) may favor full or almost full SNARE complex formation and at the same time inhibit fusion to yield a metastable state. In (a), the model assumes that the SNARE motifs and TM regions of both proteins form continuous helices and that membrane fusion is forced as these continuous helices zipper from the N- to the C-terminus. In (b) and (c), the model assumes that the SNARE motifs and TM regions of synaptobrevin 2 and syntaxin 1 are connected by short flexible linkers. In (d), the metastable state (left) is predicted to be the substrate for synaptotagmin 1 to trigger neurotransmitter release by binding simultaneously to both membranes and to the SNARE complex, displacing complexin, on Ca2þ influx (right). Synaptotagmin 1 is in light blue, except for the C2B domain, which is in dark blue. The – and þ signs illustrate overall features of the surface electrostatic potential of the synaptotagmin 1 C2 domains that may prevent phospholipid binding in the absence of Ca2þ or activate phospholipid binding and membrane fusion upon Ca2þ influx. Positive charges at the C-terminus of the SNARE complex that may also help to bend the membranes to induce membrane fusion are not shown. The N-terminal region of syntaxin 1 is not shown in any of the models. EM, electron microscopy; SNARE, soluble N-ethylmaleimide-sensitive factor attached protein receptor; TM, transmembrane.

multiple technical problems with the initial reconstitution experiments and mixing of the proteoliposome contents without leakiness, a key feature expected for physiological membrane fusion, has never been

SNAREs 71

demonstrated; on the other hand, a fundamental problem from a biological perspective is that the minimal model does not account for the strict requirement of SM proteins for all types of SNARE-dependent membrane traffic, including the essential nature of Munc18–1 for neurotransmitter release. Moreover, from a biophysical point of view, the structural-energetic basis of the minimal model is unclear. A key aspect of the original zippering model was the prediction that the SNARE motifs and TM regions of synaptobrevin 2 and syntaxin 1 could form continuous a-helices, which would therefore exert mechanical force on both membranes to induce fusion (Figure 4(a)). However, the available structural data suggest that the SNARE motifs and TM regions are connected by short flexible linkers, which is expected to hamper transduction of the energy of SNARE complex formation into the energy required to bend the membranes and initiate fusion. Hence, as they assemble, the SNARE complexes could diffuse into the space between the membranes and hinder (rather than facilitate) fusion (Figure 4(b)). In addition, introduction of helix-breaking residues or the insertion of long flexible linkers in the sequences between the SNARE motifs and TM regions had little or only moderate effects on the lipid mixing observed in the initial reconstitution experiments. Because the long linkers should have strongly uncoupled SNARE complex formation from its action on the membranes, the observed lipid mixing probably arose in part from factors beyond SNARE complex formation (e.g., instability of the vesicles). Some of these factors became clear in subsequent reconstitution experiments. Thus, the proteoliposomes used in the initial reconstitutions may indeed have been unstable because they contained extremely high protein-to-lipid ratios (1:20 for synaptobrevin 2, which translates into a protein surface concentration comparable to a 12 mM concentration in bulk solution). Moreover, the proteoliposomes were prepared with a reconstitution method that relies on co-solubilization of proteins and lipids with detergent and yields heterogeneous vesicles (both in terms of size and protein density). Reconstitutions with this method using lower protein densities (e.g., 1:200) yielded less efficient lipid mixing, even though these densities are still comparable to a 1.2 mM protein concentration in bulk solution, and the observed lipid mixing could arise from populations of small vesicles with higher than average protein densities (note that the resulting high curvature destabilizes the vesicles because of membrane stress). Indeed, no significant lipid mixing was generally observed at these proteins densities when vesicles were prepared by a reconstitution method that involves the incorporation of detergent-solubilized SNAREs into preformed liposomes and yields more homogeneous proteoliposomes. Significantly, efficient

SNARE complex formation without lipid mixing was observed using proteoliposomes prepared by this method, as predicted from the structural arguments described in Figure 4(b). From the available reconstitution data, it seems clear that SNARE complex formation is not sufficient for membrane fusion, and lipid mixing can only be induced under conditions in which membranes are destabilized and a large number of SNARE complex force the two membranes to collapse. However, an important notion that emerged from some of the reconstitution experiments is that the TM regions of the SNAREs can destabilize the bilayer, which may contribute to the function of SNAREs in membrane fusion. Moreover, recent reconstitution data obtained with SNAREs, synaptotagmin, and complexin appear to reproduce events that may occur during neurotransmitter release. Hence, the reconstitution approach is providing insights into important features of the release machinery and, with the appropriate methodology, this approach promises to provide a critical tool to understand the mechanism of neurotransmitter release. Note also that, even though the minimal model now appears to be incorrect, the original idea arising from the zippering model that SNARE complex formation causes membrane fusion could still be valid without the requirement that the SNARE motifs and TM regions form continuous helices. Thus, a modified model predicts that the energy of SNARE complex assembly could be used to induce fusion if the SNAREs are kept apart by virtue of interactions with a bulky protein or proteins (Figure 4(c)), which would prevent the undesired diffusion of the SNARE four-helix bundle into the center of the intermembrane space (Figure 4(b)). This model remains to be tested but emphasizes the necessity of understanding the functions of additional factors that are crucial for release in order to fully understand the function of the SNAREs themselves.

SNARE Interactions with Other Conserved Components of the Fusion Machinery A wide variety of proteins have been shown to play a role in synaptic vesicle exocytosis. Elucidating how the functions of these proteins are coupled to SNARE function is thus critical to understanding the mechanism of release. A difficulty in this endeavor has arisen because of the stickiness of the SNAREs, which has led to the identification of many interactions that are probably irrelevant (for instance, more than 40 proteins have been described that bind to syntaxin 1). Nevertheless, the complementary information provided by genetic and electrophysiological experiments, together with structural studies, have helped to unravel

72 SNAREs

the nature of some of these interactions and to support their physiological significance. Among the interactions of the SNAREs with other conserved components of the intracellular membrane fusion machinery, those with SNAPs/NSF have a clear functional role, namely the disassembly of the SNARE complex after fusion to recycle the SNAREs for another round of fusion. Rab proteins, which have been implicated in vesicle docking or tethering, do not appear to interact directly with the SNAREs but have been shown to be functionally coupled to the SNAREs in diverse systems and may indirectly regulate the SNARE complex assembly. The most fascinating and at the same time enigmatic SNARE interactions with the conserved components of the fusion machinery are those with SM proteins. The strict requirement for SM proteins in all types of SNARE-dependent membrane fusion, their localization at sites of fusion, and their diverse interactions with SNAREs suggest that SM proteins may have a direct function in fusion. However, this function has remained elusive, in part because of the diversity of these interactions. Neuronal Munc18–1 was shown to bind tightly to the closed conformation of syntaxin-1, but this interaction hinders SNARE complex formation and hence does not explain the essential nature of Munc18–1 for neurotransmitter release. Conversely, the yeast plasma membrane SM protein Sec1p was found to bind to SNARE complexes instead of to the cognate syntaxin. Moreover, syntaxins from diverse intracellular compartments in yeast and mammals bind to SM proteins through their NTS, in an interaction that is compatible with SNARE complex assembly. Finally, the yeast vacuolar SM protein Vps33p appears to bind to SNARE complexes at least in part through the N-terminal PX domain of the SNAP-25 homolog Vam7p. This diversity of SM protein–SNARE interaction modes has probably emerged from distinct regulatory requirements of traffic in different membrane compartments, but it seems likely that a common binding mechanism between these two conserved protein families must exist. In this context, increasing evidence has suggested that all SM proteins bind to SNARE complexes, and a direct interaction between Munc18–1 and neuronal SNARE complexes was recently identified. Hence, Munc18–1 exhibits at least two different modes of interactions with the SNAREs: one with the syntaxin 1 closed conformation, which probably represents a specialization that evolved for the tight regulatory requirements of neurotransmitter release, and another with the SNARE complex, which may be conserved in all types of intracellular membrane traffic. The former interaction may play a role in stabilizing syntaxin 1 or

preventing this promiscuous protein from binding to other proteins. In addition, syntaxin 1 has been shown to form two different types of four-helix bundles with SNAP-25, containing two copies of syntaxin 1; these complexes represent kinetic traps that hinder SNARE complex formation, and binding of Munc18–1 to the syntaxin 1 closed conformation may prevent the formation of these traps. Because the closed conformation contains a four-helix bundle and Munc18–1 also binds to the SNARE complex, it has been proposed that Munc18–1 may provide a template for assembling the four-helix bundle formed by the SNARE motifs of syntaxin 1, SNAP-25, and synaptobrevin 2 in the same site. In this context, Munc18–1 could play the role of the bulky protein depicted in the model in Figure 4(c), but this model remains to be tested and the precise nature of the Munc18–1–SNARE complex interaction still needs to be defined. Moreover, alternative models can be envisaged for the active role of Munc18–1 in release. For instance, the SNARE motifs of synaptobrevin 2 and syntaxin can bind in an antiparallel fashion, which hinders fusion; Munc18–1 may ensure assembly of the SNARE complex in the proper, parallel orientation. The transition from the Munc18–1–syntaxin 1 complex to the Munc18–1–SNARE complex may be a central event in the priming reaction that makes docked synaptic vesicles readily releasable, but the mechanism of this transition is also unclear. Functional experiments in Caenorhabditis elegans have suggested that this transition may be mediated by Unc13/Munc13s and Unc10/Rab3 interacting molecules (RIMs), which are large proteins from presynaptic active zones with critical roles in release. Unc13/Munc13s were initially thought to bind to syntaxin 1, but recent evidence indicates that they do not form binary complexes with syntaxin 1. Hence, gaining insight into the biochemistry of these proteins, the interactions underlying the conformational transition of syntaxin 1, and the role of the Munc18–1–SNARE complex interaction will be crucial to the understanding of the mechanism of neurotransmitter release. It will also be critical to better understanding which aspects of the coupling mechanism between Munc18–1 and the neuronal SNAREs are conserved in other systems and may reflect the general function of SM proteins in membrane fusion, although it is becoming increasingly clear that this function is somehow coupled to SNARE complex assembly.

SNARE Interactions with Specific Components of the Release Machinery Interactions of the SNAREs with components of the exocytotic machinery that have specialized roles in

SNAREs 73

release also appear to be associated with the regulation of SNARE complex assembly, but may also attract factors that facilitate and/or inhibit membrane fusion to confer the exquisite Ca2þ sensitivity of neurotransmitter release (the fast, synchronous components of release is triggered in less than 100 ms after Ca2þ influx in some synapses). An example of the former category is the interaction of syntaxin 1 and SNAP-25 with tomosyn, a large protein with a SNARE motif homologous to that of synaptobrevin 2. Indeed, the crystal structure of tomosyn bound to syntaxin 1 and SNAP-25 revealed a four-helix bundle analogous to the SNARE complex but with synaptobrevin replaced by the tomosyn SNARE motif (Figure 5(a)). Therefore, formation of this complex is expected to inhibit release by preventing the assembly of functional SNARE complexes containing synaptobrevin 2. This prediction has been confirmed by genetic experiments in C. elegans, although tomosyn may play additional functions through the long sequences preceding the SNARE motif. These experiments revealed at the same time a functional interplay between tomosyn and Unc13/Munc13s that may control the balance between unproductive tomosyn–syntaxin 2–SNAP-25 complexes and productive SNARE complexes containing synaptobrevin 2. Diverse functional data have suggested that the SNAREs are directly or indirectly coupled to Ca2þ sensing during neurotransmitter release, including the

Figure 5 Structural basis for SNARE interactions: (a) with tomosyn; (b) with complexin. Crystal structures of the complex formed by the tomosyn, syntaxin 1, and SNAP-25 SNARE motifs (Pobbati et al.) (a) and the complexin–SNARE complex (Chen et al.) (b). Tomosyn is in salmon, complexin is in pink, and the remaining color coding is as in Figure 1. Note the similarity between the structures of the tomosyn–syntaxin 1–SNAP-25 complex (a) and the SNARE complex (Figure 2(a)), with the tomosyn and synaptobrevin 2 SNARE motifs occupying the same positions in the complexes. Note also that complexin forms an a-helix that binds to a groove between the syntaxin 1 and synaptobrevin 2 SNARE motifs in the SNARE complex (b). C, C-terminus; N, N-terminus; SNAP-25, synaptosomal associated protein of 25 kDa; SNARE, soluble N-ethylmaleimide-sensitive factor attached protein receptor.

alteration of the Ca2þ sensitivity of secretion in chromaffin cells caused by some point mutations in SNAP-25 and the finding that elevated Ca2þ can compensate the inhibition of release caused by botulinum neurotoxin A, which cleaves SNAP-25 close to its C-terminus. Whereas NMR studies showed that the SNARE complex does not contain specific Ca2þ-binding sites that could be directly involved in Ca2þ sensing during release, it has become increasingly clear that interactions of the SNAREs with synaptotagmin 1 and complexins are key for the Ca2þ-triggered step of neurotransmitter release. Synaptotagmin 1 is a synaptic vesicle protein that acts selectively as a Ca2þ sensor in synchronous release. This function depends on Ca2þ-dependent phospholipid binding to the two C2 domains that form most of the synaptotagmin 1 cytoplasmic region (the C2A and C2B domain), with the C2B domain playing a preponderant role. Many studies described interactions of the SNAREs with synaptotagmin 1, but it was unclear which of these interactions might be functionally relevant. Recent data may have yielded key insights into this issue. Thus, Ca2þ-bound synaptotagmin 1 was shown to bind simultaneously to the C-terminus of the SNARE complex and to phospholipids through the C2B domain, and the C2B domain was also found to interact simultaneously with two membranes on Ca2þ binding. Moreover, both the Ca2þ-bound C2B domain and the C-terminus of the SNARE complex are highly positively charged. These observations suggest that the SNARE complex and the C2B domain may cooperate in bringing the membranes together and in bending them to accelerate fusion (Figure 4(d), right). Although these ideas provide an attractive mechanism for coupling synaptotagmin 1 and SNARE function that explains the preponderant role of the C2B domain, their validity remains to be demonstrated and alternative models can be envisaged. For instance, it is also possible that the highly positive electrostatic potential generated by the SNARE C-terminus and synaptotagmin 1 may help to open the fusion pore if the synaptic vesicle and plasma membranes are already partially merged (forming a hemifusion state) before Ca2þ influx. Complexins are small soluble proteins that bind tightly to the neuronal SNARE complex in a Ca2þ independent manner. Knockout of the two major complexin isoforms in mice revealed a selective impairment of the Ca2þ-triggered step of synchronous release, but an excess of complexin can also inhibit release. X-ray crystallography and NMR spectroscopy showed that complexin binds in an antiparallel a-helical conformation to a groove between the SNARE motifs of synaptobrevin 2 and syntaxin 1 (Figure 5(b)), stabilizing the SNARE complex.

74 SNAREs

Because SNARE complexes are generally believed to assemble from the membrane-distal N-terminus and zippering toward the C-terminus is expected to be hindered by membrane repulsion, these findings led to a model whereby complexin helps to complete SNARE complex assembly, yielding a metastable state that is critical for the fast speed of synchronous release (Figure 4(d), left). Interestingly, recent reconstitution experiments showed that complexin inhibits SNARE-induced liposome fusion at a hemifusion state and that Ca2þ plus synaptotagmin 1 release the inhibition, resulting in a fast burst of fusion that cannot be observed with the SNAREs alone. In addition, simultaneous binding of the synaptotagmin 1 C2B domain to phospholpids and SNARE complexes was shown to displace a complexin fragment bound to these complexes. Altogether, these data suggest that complexin may indeed play an active role in release by promoting SNARE zippering toward the C-terminus and thus bringing the system closer to fusion (perhaps to hemifusion), but at the same time plays an inhibitory role to prevent full fusion before Ca2þ influx; such inhibition is then released by simultaneous binding of synaptotagmin 1 to Ca2þ, to both membranes, and to the SNARE complex to active release (Figure 4(d)). Clearly, there are alternative models to explain the available data, including the active function of complexin for release, but these data show that there is a fascinating interplay among the SNAREs, complexin, synaptotagmin 1, Ca2þ, and phospholipids that is probably crucial for the exquisite regulation of neurotransmitter release. The relation between Munc18–1, complexin, and synaptotagmin 1 binding to the SNARE complex remains to be studied.

SNARE Specificity The contribution of SNAREs to the specificity of membrane traffic has been another area of strong debate about SNARE function. This issue is important for interpreting the phenotypes of synaptobrevin 2 and SNAP-25 mice because, given the observation that neurotransmitter release can still be observed in these mice, an absolute specificity in SNARE pairing implies that the SNAREs are not essential for membrane fusion. Biochemical assays showed that some SNAREs have a tendency to bind to both cognate and noncognate SNAREs, and hence can be quite promiscuous. Conversely, reconstitution studies suggested a high specificity in SNARE-induced lipid mixing, but the noncognate SNARE pairings identified biochemically were not analyzed. A recent study showed that early endosome fusion is specifically mediated by a set of SNAREs in living cells, whereas the same set of SNAREs promiscuously induce liposome fusion

with SNAREs from the plasma membrane or late endosomes. These results can be interpreted in the framework of sequence analyses and the known structures of SNARE complexes. Because SNARE complex assembly involves protein–protein interactions, assembly must involve some degree of specificity. Moreover, the SNAREs are one of the most diverse among the protein families generally involved in intracellular membrane traffic (25 members in yeast and 36 in humans; only Rab proteins are more diverse). However, because coiled-coil interactions can be considerably promiscuous and the hydrophobic residues involved in SNARE complex formation are highly conserved in each subfamily, it is not surprising that some degree of nonspecificity exists in SNARE pairing. Hence, the picture that emerges is that SNAREs contribute to the specificity of traffic in distinct membrane compartments but that this specificity also involves other proteins. In particular, there is little doubt that an important contribution to specificity arises also from Rab protein interactions. Note also that the contribution of the neuronal SNAREs to specificity in neurotransmitter release most likely also involves interactions with factors specialized for release, such as synaptotagmin 1 and complexins.

Outlook Research on SNARE proteins has provided an emphatic example of the power of interdisciplinary approaches to studying protein function and of the difficulties of exactly pinpointing the specific roles of proteins that act as part of complex macromolecular assemblies. Hence, the divide-and-conquer approach has provided crucial clues to the understanding of the functions of the SNAREs and other components of the release machinery, but increasing evidence illustrates the complexity and cooperativity of this system. Clearly, a true understanding of the mechanism of release requires further studies in which more and more of these components are examined together in the presence of membranes. A critical foundation for these studies has been provided by the astonishing advances in this field made during the past 20 years. See also: Neurotransmitter Release from Astrocytes; Synaptic Vesicles.

Further Reading Brunger AT (2005) Structure and function of SNARE and SNARE-interacting proteins. Quarterly Review of Biophysics 38: 1–47.

SNAREs 75 Dulubova I, Sugita S, Hill S, et al. (1999) A conformational switch in syntaxin during exocytosis: Role of Munc18. EMBO Journal 18: 4372–4382. Hanson PI, Roth R, Morisaki H, Jahn R, and Heuser JE (1997) Structure and conformational changes in NSF and its membrane receptor complexes visualized by quick-freeze/deep-etch electron microscopy. Cell 90: 523–535. Jahn R and Scheller RH (2006) SNAREs – engines for membrane fusion. Nature Reviews Molecular and Cell Biology 7: 631–643. Lin RC and Scheller RH (1997) Structural organization of the synaptic exocytosis core complex. Neuron 19: 1087–1094. Link E, Edelmann L, Chou JH, et al. (1992) Tetanus toxin action: Inhibition of neurotransmitter release linked to synaptobrevin proteolysis. Biochemistry and Biophysics Research Communications 189: 1017–1023. Nichols BJ, Ungermann C, Pelham HR, Wickner WT, and Haas A (1997) Homotypic vacuolar fusion mediated by t- and v-SNAREs. Nature 387: 199–202. Poirier MA, Xiao W, Macosko JC, et al. (1998) The synaptic SNARE complex is a parallel four-stranded helical bundle. Nature Structural Biology 5: 765–769.

Rizo J, Chen X, and Arac D (2006) Unraveling the mechanisms of synaptotagmin and SNARE function in neurotransmitter release. Trends in Cell Biology 16: 339–350. Schiavo G, Benfenati F, Poulain B, et al. (1992) Tetanus and botulinum-B neurotoxins block neurotransmitter release by proteolytic cleavage of synaptobrevin. Nature 359: 832–835. Schoch S, Deak F, Konigstorfer A, et al. (2001) SNARE function analyzed in synaptobrevin/VAMP knockout mice. Science 294: 1117–1122. Schulze KL, Broadie K, Perin MS, and Bellen HJ (1995) Genetic and electrophysiological studies of Drosophila syntaxin-1A demonstrate its role in nonneuronal secretion and neurotransmission. Cell 80: 311–320. Sollner T, Bennett MK, Whiteheart SW, Scheller RH, and Rothman JE (1993) A protein assembly-disassembly pathway in vitro that may correspond to sequential steps of synaptic vesicle docking, activation, and fusion. Cell 75: 409–418. Sollner T, Whiteheart SW, Brunner M, et al. (1993) SNAP receptors implicated in vesicle targeting and fusion. Nature 362: 318–324. Sudhof TC (2004) The synaptic vesicle cycle. Annual Review of Neuroscience 27: 509–547.

Synaptic Vesicles S Takamori, Tokyo Medical and Dental University, Tokyo, Japan ã 2009 Elsevier Ltd. All rights reserved.

Introduction Communication between neurons or from neurons to their target tissues takes place at a specialized structure called the ‘synapse’ (Greek meaning ‘to clasp’). Synapses consist of two functionally and morphologically distinct components: the presynapse, from which neurotransmitter molecules are released, and the postsynapse, where specific receptors for the respective neurotransmitters are localized on the surface and the complex signaling cascades proceed. As such, one neuron can activate or inactivate connected neurons or target tissues, depending on the transmitter molecule they utilize for their signal transmission. Based on electrophysiological experiments pioneered by Katz and colleagues in the 1960s, it has been postulated that neurotransmitters are released from presynaptic terminals in discrete packets termed ‘quanta.’ Electron microscopy revealed that at presynaptic terminals, hundreds of small and round membranous vesicles – synaptic vesicles (SVs) – are clustered, which can be reasonably linked to the quanta (Figure 1(a)). Furthermore, biochemical experiments using isolated vesicles from mammalian brains have proven that SVs exhibit a variety of transport activities for major chemical transmitters. Based on these findings, it is generally believed that SVs store neurotransmitters and the fusion of SV membrane to the plasma membrane elicits the quantal release of neurotransmitters. Because of their important roles in basic neural functions, much effort has been focused on understanding the molecular machinery of SVs. This article focuses on general features of SVs in terms of morphology, biogenesis, and recycling and their overall molecular composition.

General Features of Synaptic Vesicles as an Organelle Clustered at the nerve endings, SVs are one of the most striking morphological hallmarks of presynaptic terminals in electron micrographs. They appear to be roughly homogeneous in size and shape, but it has been determined that some heterogeneity exists among them. Under certain experimental conditions, their shape appeared to be of two types – almost spherical and oval or flat. This difference may be an experimental artifact presumably from the fixation

76

process. Interestingly, this difference in shape may be associated with their transmitter content. It was found that most excitatory, asymmetric glutamatergic synapses contain spherical vesicles, and inhibitory, symmetric, GABAergic synapses contain the latter. Regardless of the functional significance, this observation has provided a possible morphological characteristic to distinguish glutamatergic SVs and GABAergic SVs. Due to their uniformity in size and abundance in the brain, it is feasible to isolate SVs with high purity and in large amounts – a prerequisite for biochemical and biophysical experiments (Figure 1(b)). Electron microscopic observations of isolated SVs from rat brains have revealed that the surface is uneven. They are decorated with one or more prominent globular structures and several spiky or amorphous substructures are observable under electron microscope. After proteolytic digestion these structures are removed and the rims of the lipid bilayer become clearly visible, demonstrating that the vesicles are coated with proteins on the surface (Figures 1(c) and 1(d)). As measured by electron microscopy, their size varies substantially: The diameter of the outer bilayer ranges from 35 to 50 nm, with an average peak at 42 nm. The total dry mass of an average SV was deduced from a combination of protein quantitation, lipid quantitation, and SV particle counting, resulting in approximately 30 attograms (ag) per vesicle, which consists of 17 ag of proteins and 12 ag of lipids. Assuming the thickness of a lipid bilayer is 4 nm, the average inner volume can be estimated as approximately 20
10–21 l, which provides enough space for approximately 1800 transmitter molecules at a concentration of 150 mM. Within these physical and molecular constraints, SVs are effectively equipped with a unique set of proteins and lipids necessary for executing the fundamental tasks in neurotransmitter release.

Biogenesis of Synaptic Vesicles Like other proteins of the secretory pathway, most SV proteins are synthesized at the endoplasmic reticulum and processed through the Golgi apparatus for maturation in the cell body of neurons (Figure 2(a)). Conceptually, subdomains which selectively collect the SV proteins bud off from the Golgi apparatus and then travel along the axon to the presynaptic terminals. However, since no vesicles are as small as mature SVs, and some larger nonuniform tubulovesicular structures are seen in the axons, SV constituents are thought to travel along the processes on heterogeneous membranes termed SV precursors.

Synaptic Vesicles 77

Figure 1 Morphology of synapse and synaptic vesicles. (a) A transmission electron micrograph of synapse. The presynaptic terminal (T) contains numerous synaptic vesicles (SVs) and mitochondria (Mit). A small portion of SVs is attached to electron dense structures at the presynaptic plasma membrane termed the active zone (AZ), where exocytosis of SVs takes place. Opposite the AZ, there are electron dense structures beneath the postsynaptic membrane termed postsynaptic density (PSD), where neurotransmitter receptors form signaling complexes. D, dendrite; SC, synaptic cleft. (b) An electron micrograph of synaptic vesicle fraction. SVs were purified from mouse whole brains, negatively stained, and imaged by transmission electron microscopy. Inset shows immunogold labeling of the vesicles with an antibody against an SV marker protein, synaptophysin. Approximately 95% of the membranous structures are labeled. (c, d) SVs before (c) and after (d) proteolysis imaged by cryo-electron microscopy. Scale bar ¼ 1 mm (a), 100 nm (b, inset), 20 nm (c). (a) Adapted from the George E. Palade EM Slide Collection at Yale University School of Medicine. (c, d) Reproduced from Takamori S, Holt M, Stenius K, et al. (2006) Molecular anatomy of a trafficking organelle. Cell 127: 831–846, with permission from Elsevier.

Once formed, the SV precursor needs a pathway to reach its destination. Microtubules are known to serve as the roads for transporting cargo. Motor proteins of the kinesin superfamily (KIFs) bind to both the microtubules and the cargo, and they control the directional transport of intracellular transporting cargo. Among the KIFs, KIF1A and KIF1Bb are known to participate in axonal transport of SV precursors. Gene disruption of either KIF1A or KIF1Bb results in reduction of SV density at the presynaptic terminals and therefore impaired neurotransmission. The KIF1A- and KIF1Bb-carrying SV precursors contain the major SV marker proteins, such as synaptophysin, synaptobrevin, synaptotagmin, and rab3A, but not presynaptic plasma membrane proteins, such as syntaxin 1 and SNAP-25. In addition, proteins that form the cytomatrix at the active zone, such as piccolo and bassoon, are transported to the

axon on distinct cargo, suggesting that segregation of the presynaptic proteins is initiated before they arrive at the terminals. The mechanism by which the SV constituents are selectively recruited into the SV precursor is unknown. It is possible that SV proteins have a specific signal sequence that is assembled and forms microdomains at the exit of the trans-Golgi network, but no such common signal sequence has been found. Furthermore, the molecular mechanism regarding how the constituents of the SV precursors can be recognized by KIF1A/1Bb proteins is poorly understood. When SV precursors arrive at the presynaptic terminal, they fuse with the plasma membrane where the main route for SV biogenesis is initiated. This route, the AP-2-dependent pathway, is mediated by several essential proteins for vesicle endocytosis, namely clathrin, dynamin, and AP-2. An alternative route by

78 Synaptic Vesicles

Golgi apparatus

SVprecursor KIF1A/KIF1Bb

Synaptic vesicle

ER Microtubules

Nucleus

Dendrite AZ precursor complex

Microtubules

Cell body (soma)

SV precursor complex Presynaptic terminal

Axon

a

1

NT

Early endosome

7⬘

7 H+

6 2

Docking 3

Exocytosis

Priming 4

5

Endocytosis 6⬘

Ca2+

b Figure 2 Biosynthesis and recycling of SVs. (a) Biosynthesis of SVs. SV proteins are synthesized in the cell body of neurons and the SV precursors bud off from the trans-Golgi network. The SV precursors are transported along the axon, guided by the microtubules and kinesin motor proteins (KIF1A and KIF1Bb). (b) SV cycle at presynaptic terminal. (1) SVs are filled with neurotransmitter. (2) The SVs are then transported to the active zone and docked to the presynaptic plasma membrane (docking) (3). The docked SVs become fusion competent by molecular events called priming (4). When an action potential arrives at the terminal, calcium influx through voltagedependent calcium channels triggers fusion of SV membrane with the plasma membrane (exocytosis), causing discharge of SV content (5). Exocytosed SVs are regenerated either by clathrin-independent fast endocytosis (6) or by clathrin-dependent slow endocytosis (60 ). The newly regenerated SVs are immediately refilled with neurotransmitters (7) or, in some case, they undergo fusion steps with early endosomes (70 ).

which SVs are generated from early endosomal membranes at the presynaptic terminal is dependent on AP-3. The latter pathway does not seem to account for the majority of SVs since there are no remarkable alterations in SV morphology and numbers in the

mocha mouse, which lacks functional AP-3. How, then, can SV proteins, but not plasma membrane residents, be recruited into newly generated SVs? Several amino acid sequence motifs or molecular determinants for each SV protein have been proposed to

Synaptic Vesicles 79

explain this mechanism. First, a dileucine motif located in the cytoplasmic tails of vesicular transporters (VMAT2, VAChT, and VGLUT1) is important for precise sorting and fast recruitment of those SV proteins. A VAChT mutant which lacks the dileucine motif was trapped at the plasma membrane of the cell body of the differentiated cholinergic SN56 cell lines, and the targeting of the mutant protein to neurites and varicosities (which resemble axon terminals) was dramatically reduced, indicating that the SV precursor cargo carrying VAChT is not directly transported through the axon but is delivered to the plasma membrane of the cell body before being transported along the axon. Second, the intravesicular N-glycosylation of synaptotagmin 1 is essential for its vesicular targeting. Interestingly, synaptotagmin 7, one of the synaptotagmin family members which resides preferentially on the plasma membrane, is targeted to vesicles when the intravesicular portion is replaced with a portion containing the N-glycosylation site of synaptotagmin 1. The mechanistic basis and involvement of other factors is not clear, but the interaction of glycoresidues of synaptotagmin 1 with other proteins at the cell surface may govern the targeting of synaptotagmin 1 to vesicles. It is likely that multiple factors, including protein– protein interactions and glycoresidue–protein interactions, modulate the sorting of SV proteins. The general principle of the segregation of SV proteins from plasma membrane proteins during SV biosynthesis has not been clarified.

The Synaptic Vesicle Cycle At the presynaptic terminals, SVs are not of ‘single use.’ They are regenerated at the terminals independent from protein synthesis in the cell body. The SV cycle can be outlined, with the uptake of neurotransmitter into SVs as a first step (Figure 2(b), step 1). Neurotransmitters in the central nervous system (CNS) are synthesized locally in the cytoplasm of the presynaptic terminals and are actively transported into SVs. Away from the plasma membrane, the majority of neurotransmitter-filled SVs are either diffusively floating in the cytoplasm or tethered with cytoskeleton components such as actin and spectrin (step 2). For exocytosis, SVs must come into physical contact with the plasma membrane (docking; step 3). SVs do not evenly dock to the whole area of the presynaptic plasma membrane but, rather, to a restricted area called the active zone. There, large cytomatrix proteins, such as bassoon, piccolo, and munc13, form huge protein complexes that appear as electron-dense structures on electron micrographs. Docked SVs are then transformed into fusion-competent SVs via a process called priming (step 4). As soon as an electrical stimulus

arrives at the terminal, voltage-dependent calcium channels at the active zone open, resulting in a rapid and local increase in Ca2þ concentration. Ca2þ ions trigger the fusion of the SV membrane with the plasma membrane in less than 100 ms. Since other exocytotic reactions (i.e., hormone secretion from endocrine cells) take much longer (seconds to minutes), there must be unique factors present only in neurons to perform such a rapid membrane fusion reaction. After exocytosis, SV components that are incorporated into the plasma membrane are retrieved to form a new SV by endocytosis. There are at least two kinetically distinct modes of endocytosis. The time constants of the fast and slow phase are approximately 1 and 10 s, respectively. Whereas the molecular machinery for the fast phase is not well understood, the slow phase is mediated by the formation of clathrin-coated pits. In both modes, GTP hydrolysis by the GTPase protein dynamin is indispensable for the fission of the invaginated membranes of newly formed SVs. Although various endocytosis-related proteins, such as AP-2, endophilin, amphiphysin, and synaptojanin, have been identified and implicated in controlling endocytosis, their precise roles are a matter of intense research. The reformed SVs then either recycle back and are refilled with neurotransmitters (step 1) or pass through the early endosomal intermediates before recycling back to step 1. An alternative pathway has been proposed that is similar to the exocytosis of secretory granules; that is, SVs do not fully collapse with the plasma membrane upon fusion but instead form a narrow and transient fusion pore which does not allow a complete discharge of neurotransmitter content. As soon as the pore closes, the half-empty SV can either be immediately engaged in another round of exocytosis or go back to step 1. The existence of such a rapid recycling mode in the CNS, termed the ‘kissand-run’ mechanism, is under debate. The SVs clustering at the presynaptic terminals can be divided into two functional pools. The first pool contains a small fraction of SVs (5–10% of the total SVs at the presynaptic terminals) that can be released rapidly by a brief high-frequency train of action potentials or by stimulation with hypertonic solution. This pool is thought to be release-ready and is therefore referred as to the readily releasable pool (RRP). The second pool, the reserve pool (RP), represents a vesicle fraction that does not immediate participate in exocytosis. Instead of participating in exocytosis, the RP vesicles replenish the RRP pool after the RRP vesicles undergo exocytosis. Both the amount of the RRP and the rate of replenishment of RRP with RP are critical parameters to determine the availability of vesicles for exocytosis, thereby affecting the

80 Synaptic Vesicles

characteristics of short- and long-term plasticity of a given neuron. Classically, the RRP was related to a fraction of morphologically docked vesicles at the plasma membrane and the RP was thought to be spatially distant from the plasma membrane. However, one study suggested that the RRP vesicles do not necessarily correlate with the morphologically docked vesicles but are randomly distributed in the SV cluster at the terminal, indicating that there are no correlations between the anatomical locations of SVs and their functional fusion competence. Mechanisms underlying how the mobility and the fate of an individual SV can be molecularly defined are uncertain.

Molecular Composition of Synaptic Vesicles With respect to SV functions as described previously, SVs should be equipped with two classes of essential protein components: transport proteins involved in neurotransmitter uptake into SVs and membrane trafficking proteins involved in the regulation of vesicle cycle and exo-endocytosis. A large body of work has focused on identifying protein components on SVs, leading to an almost complete identification of SV proteins. The stoichiometry of SV proteins may be flexible because many cytoplasmic proteins are temporarily attached to and detached from SVs depending on the status of a specific SV during recycling. On the other hand, a basic set of essential proteins must be present on all SVs for vesicle functions, and the numbers of individual proteins and the proportions within an SV might proteins within an SV to be maintained (Figure 3(a)). Furthermore, immunohistochemical studies on SV proteins have demonstrated that most of the major SV proteins consist of multiple isoforms whose expressions in the CNS are partially overlapping or mutually segregated, the combinations of which would create heterogeneity in protein composition in each SV. In contrast to the identification of SV proteins which became relatively handy with advancements in mass spectroscopy, understanding the mechanistic function(s) of each protein has been not so trivial, mainly because of technical limitations with regard to manipulating and measuring the intracellular events with high temporal and spatial resolution. The following sections introduce some of the essential and abundant SV proteins, which are stoichiometric components with at least one copy per SV. Proteins for Neurotransmitter Uptake

The accumulation of neurotransmitters in SVs is driven by a proton electrochemical gradient which

is generated by a vacuolar-type proton ATPase. The proton pump consists of at least 13 subunits and is the largest functional protein complex in the SV membrane. It contains two functional units – a larger peripheral protein complex (V1) which catalyzes ATP hydrolysis and an integral membrane protein complex (Vo) which builds up a ring structure in the membrane and mediates proton translocation. The molecular weight of the entire complex is approximately 800 kDa; therefore, a single complex accounts for approximately 10% of the total SV protein. An SV contains a single copy of the proton pump complex, which may be sufficient to energize neurotransmitter uptake into the SVs. With the aid of a proton electrochemical gradient, vesicular transporters specific for neurotransmittertype mediate neurotransmitter uptake into the vesicles. There are four distinct uptake systems for neurotransmitters in the CNS. Three isoforms of vesicular glutamate transporters (VGLUT1–3) transport glutamate into SVs. GABA and glycine share the same transporter, the vesicular inhibitory amino acid transporter (VIAAT; initially called vesicular GABA transporter (VGAT)). Two monoamine transporters (VMAT1 and VMAT2) transport all biogenic amines, with VMAT2 preferentially expressed in the brain. The uptake of acetylcholine is mediated by the vesicular acetylcholine transporter VAChT. All four transporter families belong to the solute carrier protein family, but there are no sequence homologies among the transporters for different neurotransmitter types. There are differences in energetics of the transport; some of them preferentially utilize the membrane potential (VGLUTs) and others use the pH gradient (VMATs and VAChT) as the main driving force. The expression of a particular transporter in the SV membrane is the ultimate determinant for the type of neurotransmitter which is released from a given neuron. Moreover, there are indications that the expression level of the transporter per SV modulates the amount of neurotransmitters accumulated in an SV, thereby regulating quantal size. In addition to the neurotransmitter transporters discussed previously, SVs contain a chloride channel to facilitate the acidification of SVs. The voltagedependent chloride channel ClC-3 was proposed to confer this activity on SVs. Although several channel activities, such as a cation selective channel, have been demonstrated by electrophysiological methods on reconstituted systems, the molecular identities of the activities have been elusive. The SV2 protein family and tetraspan vesicle membrane proteins (synaptophysin, synaptogyrin, and SCAMPs) have been identified as SV-specific proteins, and their predicted protein structures suggest their role as a transporter or a channel.

Synaptic Vesicles 81 16BAC-PAGE kDa

200 SV2 116 97

Synaptotagmin

SDS-PAGE

67

VGLUT

Synapsins

43

Synaptophysin

30 Rab proteins 20

Synaptobrevin

14 a

Synaptotagmin

Synaptobrevin

Synaptophysin V-ATPase

ClC-3

CSP

SNAP-29

SV2

VGLUT Munc-18 Rab

Trimeric GTPase b

Synapsins

Other transporter

Figure 3 Molecular composition of SV. (a) Protein composition of SVs visualized by 16-BAC/SDS two-dimensional gel electrophoresis. A total of 500 mg of SV proteins was applied. The spots containing the major SV proteins are circled. (b) Structural model of an average SV. Based on quantitative measurements, a three-dimensional model of an average SV was constructed. An average SV contains 70 copies of synaptobrevin, 30 copies of synaptophysin, 10 copies of neurotransmitter transporter (VGLUT), 8 copies of synapsin, 15 copies of synaptotagmin, 25 copies of Rab GTPase, and 1 or 2 copies of SV2, synaptogyrin, SCAMP, and V-ATPase. The numbers of phospholipids and cholesterol are estimated to be approximately 7000 and 6000, respectively.

82 Synaptic Vesicles

However, no clear transport activities associated with these proteins have been demonstrated. It should be noted that although not related to an intrinsic function of SV2s, they have been shown to function as a protein receptor for botulinum neurotoxin A. Proteins for Membrane Trafficking

Neural exocytosis is a tightly regulated process elicited by a rapid increase in Ca2þ concentration. Since Ca2þ ions induce neurotransmitter release within 100 ms, a series of complex enzymatic reactions could not be expected for exocytosis within the short time frame. Thus, SVs that undergo exocytosis not only should be docked but also primed for membrane fusion, waiting for the final stimulus for exocytosis. Two abundant integral membrane proteins on SVs, synaptobrevin (also referred as to VAMP2) and synaptotagmin 1, have been established as the most important players in Ca2þ-triggered exocytosis. Furthermore, SVs contain two abundant peripheral protein families, rabGTPases and synapsins, which are implicated in the modulation of vesicle pools. Synaptobrevin is a small integral membrane protein that contains a SNARE motif and is thus called a v-SNARE (vesicular SNARE, also referred to as R-SNARE because of a conserved arginine in the middle of the SNARE motif). Synaptobrevin forms a tight four-helix bundle, the SNARE complex, with two other SNARE proteins at the plasma membrane – syntaxin 1 and SNAP-25 (t-SNAREs named after target-SNAREs; also referred as to Q-SNARE, which has a conserved glutamine in the middle of the SNARE motif). The SNARE hypothesis regarding membrane fusion in general proposes that the assembly and tight complex formation of appropriate pairs of SNARE proteins pulls the two membranes close together so that the two membranes become competent for fusion. Analogously, SVs become ‘primed’ when synaptobrevin forms the SNARE complex with syntaxin 1 and SNAP-25. The importance of SNARE proteins in neuronal exocytosis is emphasized by the fact that the clostridial neurotoxins, a family of metalloproteases which specifically cleave the neuronal SNARE proteins, abolish Ca2þ-dependent neurotransmitter release while the docked vesicles remain unchanged by the toxin treatment. The formation of the SNARE complex is affected by other binding factors of the SNARE proteins. For instance, it has been proposed that Munc18 protein binds to syntaxin, keeping the syntaxin molecule in a close conformation and thereby preventing syntaxin from forming the SNARE complex. Biochemical experiments using the SNARE-reconstituted proteoliposomes suggest that the SNARE complex indeed mediates

membrane fusion. However, the liposome fusion with SNAREs alone is slow and spontaneously occurs without any regulation, indicating that other factors might play a role in synchronizing the fusion reaction. Synaptotagmin 1 is known as a vesicular Ca2þ sensor for exocytosis that accomplishes rapid membrane fusion in response to Ca2þ ions. It has an N-terminal highly glycosylated intravesicular domain and two cytoplasmic C2 domains that interact with Ca2þ ions. Ca2þ-triggered exocytosis in synaptotagmin 1-deficient neurons is severely impaired, indicating its role as the Ca2þ sensor for exocytosis. The affinity of C2 domains of synaptotagmin 1 for Ca2þ ions is very low, but it increases dramatically (up to 0.5–5.0 mM) when the C2 domains bind to phospholipids. Its apparent affinity to Ca2þ meets the requirement that it should have a Ca2þ concentration sufficient for triggering exocytosis in vivo (in the calyx of Held or in chromaffin cells). However, investigations of synaptotagmin 1-deficient neurons and chromaffin cells have suggested that synaptotagmin 1 is a Ca2þ sensor only for the fast, synchronous exocytosis and not for the slow, asynchronous component. It is not clear if other synaptotagmin isoforms (at least 15 members in mammals which differ in affinity to Ca2þ and subcellular localizations) function as a Ca2þ sensor for the latter. Rab proteins are known to be involved in a variety of intracellular membrane trafficking events. Among the Rabs, Rab3A, Rab5, and Rab11 are abundant isoforms on SVs. They associate with SVs in the GTP-bound forms, whereas they dissociate from SVs in the GDPbound state. The association and dissociation cycle of Rab3A occurs in parallel with SV exocytosis, although Rab proteins are not directly involved in exocytosis or membrane fusion. In Rab3A-deficient neurons, some forms of synaptic plasticity, such as long-term potentiation in hippocampal mossy fibers, are affected, indicating that Rab3A modulates the efficacy of neurotransmission, probably by changing the vesicle pools available for exocytosis. In addition to Rab proteins, the synapsin family represents abundant peripheral membrane proteins which may interact with SVs in an activity-dependent manner. The synapsin family consists of five isoforms (I/IIA, IIB, IIIA, and IIIB) and they form homo- or heterodimer on the surface of SVs. Originally, it was discovered as a neuron-specific abundant protein substrate for the cyclic AMP-dependent protein kinase (protein kinase A) at presynaptic terminals. Their function in synaptic transmission remains unclear. It has been proposed that synapsins function as an anchoring protein to cytoskeleton components to make a portion of vesicles immobile because they exhibit a binding property to actin, tubulin, and spectrin. Such a view is partially supported by the fact that although

Synaptic Vesicles 83

Ca2þ-triggered exocytosis is intact, short-term synaptic plasticity is impaired in the absence of synapsin 1 and 2. Therefore, synapsins might play a modulating role in controlling the vesicle pools at the terminals.

endocytosis has to be guaranteed by precise processes. More studies are needed to explore the structure– function relationships of each SV constituent to understand the function of this intriguing organelle.

Conclusion

See also: Active Zone; SNAREs; Vesicular Neurotransmitter Transporters.

Although significant progress has been made in understanding molecular mechanisms of neurotransmitter release, a mechanistic understanding of SVs is yet to be achieved. Until recently, even elementary quantitative information about SV components was lacking. Quantitative analyses of SV proteins and lipids have allowed, for the first time, the proposal of a structural model of an average SV shown in Figure 3(b). The model demonstrates that SVs are highly decorated with proteins. Numerous copies of the essential proteins for exocytosis, such as synaptobrevin and synaptotagmin, are present on a vesicle indicating that the surface densities of these proteins are not rate limiting for fusion. Furthermore, SVs contain large numbers of the neurotransmitter transporter, supporting a high-speed refilling of the vesicles upon strong repetitive stimulations. One exception among the essential proteins is the V-ATPase, which is estimated to have one or two copies per vesicle, further indicating that the retrieval of SV constituents by

Further Reading Becherer U and Rettig J (2006) Vesicle pools, docking, priming and release. Cell and Tissue Research 326: 393–407. Bonanomi D, Benfenati F, and Valtorta F (2006) Protein sorting in the synaptic vesicle life cycle. Progress in Neurobiology 80: 177–217. Fernandez-Chacon R and Sudhof TC (1999) Genetics of synaptic vesicle function: Toward the complete functional anatomy of an organelle. Annual Review of Physiology 61: 753–776. Jahn R (2004) Principles of exocytosis and membrane fusion. Annals of the New York Academy of Sciences 1014: 170–178. Montecucco C, Schiavo G, and Pantano S (2005) SNARE complexes and neuroexocytosis: How many, how close? Trends in Biochemical Sciences 30: 367–372. Rissoli SO and Betz WJ (2005) Synaptic vesicle pools. Nature Reviews Neuroscience 6: 57–69. Sudhof TC (2004) The synaptic vesicle cycle. Annual Review of Neuroscience 27: 509–547. Takamori S, Holt M, Stenius K, et al. (2006) Molecular anatomy of a trafficking organelle. Cell 127: 831–846.

Endocytosis and Presynaptic Scaffolds V Haucke, Freie Universita¨t Berlin, Berlin, Germany E D Gundelfinger, Leibniz Institute for Neurobiology, Magdeburg, Germany

Pathways of SV Cycling Clathrin-Mediated Endocytic Cycling of Presynaptic Vesicles

ã 2009 Elsevier Ltd. All rights reserved.

Introduction Neurons communicate with each other by the temporally and spatially controlled release of secretory molecules (the so-called neurotransmitters) via regulated exocytosis. Following diffusion across the synaptic cleft the released nonpeptide neurotransmitters and neuroactive peptides bind to and activate postsynaptic receptors, which then elicit a response within the postsynaptic cell. In the case of most fast-acting transmitters, such as glutamate, g-aminobutyric acid (GABA), or acetylcholine, signaling is elicited by ligand-gated ion channels, which are clustered at morphologically discernible zones specialized for chemical neurotransmission termed postsynaptic densities (PSDs). The PSD is opposed to a corresponding presynaptic element including the ‘active zone’ at which presynaptic neurotransmitter-bearing vesicles (SVs) are clustered. Presynaptic active zones are characterized by an electron-dense grid of scaffolding proteins interconnected with an actin-rich cytoskeleton, which among other functions helps to maintain a pool of vesicles docked at the presynaptic plasmalemma. To sustain chemical neurotransmission under conditions of high activity and to counter-balance net insertion of membrane by exocytic vesicle fusion, SVs undergo activity-driven cycles of calcium-triggered exocytosis and endocytosis within nerve terminals, commonly referred to as the SV cycle. Cycling of SVs must allow them to retain their specific biochemical identity, including the ability to store neurotransmitter by proton pump-driven neurotransmitter transporters, and to undergo further rounds of calcium-induced fusion with the presynaptic plasmalemma. The observed tight coupling between exocytic neurotransmitter release by vesicle fusion and compensatory endocytosis has resulted in a long and still unresolved debate regarding the precise molecular mechanisms involved in SV cycling. Here, we provide a brief summary of the pathways of SV cycling, the role of clathrin and its partner proteins in maintaining SV pools, and the temporal and spatial cues provided by scaffolding proteins and membrane lipids in maintaining presynaptic exocytic– endocytic membrane traffic.

84

Landmark studies in the early 1970s by Heuser and Reese showed that following fusion by complete collapse into the plasmalemma, SVs are retrieved by compensatory ‘clathrin-dependent endocytosis’ at specialized endocytic areas just outside the active zone. An overwhelming amount of genetic, morphological, biochemical, and physiological data suggests that clathrin-mediated endocytosis indeed constitutes an essential pathway of SV recycling, at least on the organismic level and over extended periods of time. Mutants in clathrin coat components, including the AP-2 complex or accessory and adaptor proteins such as dynamin, AP180, stoned B (the Drosophila ortholog of mammalian stonins 1 and 2), eps15, synaptojanin, endophilin, amphiphysin, or intersectin/DAP160, all display defects in neurotransmission owing, at least in part, to impaired SV endocytosis (Table 1). The most dramatic phenotypes have been observed following injection of dominant-negative domains or inhibitory peptides into the giant reticulospinal synapse of the lamprey. Following electrical stimulation, distinct endocytic intermediates and vacuolar structures accumulate within and around the active zones in conjunction with a partial or complete depletion of the recycling vesicle pool. Some of these intermediates resemble structures seen in neuromuscular junctions following intense stimulation or after genetic perturbation of protein function in temperature-sensitive alleles. The mechanistic details of this pathway will be discussed later. Kiss-and-Run Mode

Based on an apparent lack of correlation between the number of morphologically distinct stable endocytic intermediates and the synaptic endocytic activity, Ceccarelli and colleagues proposed an alternative model according to which SVs release their contents through the controlled opening of a narrow fusion pore, followed by rapid closure and refilling with neurotransmitter (Figure 1). This ‘kiss-and-run’ mode of regulated secretion has been convincingly demonstrated to occur by combined electrophysiological recordings and membrane capacitance measurements in neuroendocrine cells, which mostly secrete peptide hormones or biogenic amines from large secretory granules (SGs; also termed large dense core vesicles (LDCVs)). In the case of SG exocytosis, flickering of

Table 1 Endocytic proteins involved in presynaptic vesicle cycling and their interaction partners Domains and motifs

Interaction partner(s)

Proposed function(s)

Abp1

ADF homology domain SH3 domain

Linking actin cytoskeleton with endocytosis and CAZ

AP-2 (a, b2, m2, s2 subunits)

Trunk a-Appendage b2-Appendage b2-Hinge C-m2 N-BAR domain PWXXW, LLDLD motifs FXDXF, DXF motifs SH3 domain ANTH domain FXDXF, DXF motifs DLL motifs DnaJ domain DPF, WDW motifs WDW, DLL motifs HC-terminal domain LC GTPase domain GED domain PH domain N-BAR domain SH3 domain

F-actin PRDs of synaptojanin, dynamin, Piccolo, and synapsin1 PI(4,5)P2 FXDXF/DXW, WVXF motif proteins [DE]nX1–2FXX[FL]XXR motif proteins Clathrin terminal domain Yxx and basic motif membrane cargo; PI(4,5)P2 (Acidic) phospholipids; dimerization Clathrin terminal domain AP-2 via a-appendage PRDs of dynamin (& synaptojanin) PI(4,5)P2 AP-2 via a-appendage Clathrin Hsc70 AP-2 via a-appendage Clathrin PWXXW, LLDLD, DLL motif proteins HIP1, calmodulin, Hsc70 GDP/GTP Dynamin-GED PI(4,5)P2 (Acidic) phospholipids; dimerization PRD of dynamin and synaptojanin

Amphiphysins

AP180

Auxilin

Clathrin (HC, LCa/b) Dynamin 1

Endophilin

Plasmalemmal recruitment Coat assembly Cargo selection and coat assembly Scaffolding Membrane cargo selection Membrane curvature induction/sensing; scaffolding Coat assembly Vesicle fission Plasmalemmal recruitment Coat assembly Scaffolding Stimulation of ATPase Coated pit recruitment Assembly of scaffold Diverse Control of membrane fission Self-assembly; stimulation of GTPase activity Plasmalemmal recruitment Membrane curvature induction/sensing Coated pit maturation; fission and uncoating Continued

Endocytosis and Presynaptic Scaffolds 85

Endocytic protein

Endocytic protein

Domains and motifs

Interaction partner(s)

Proposed function(s)

Epsins

ENTH domain DPW motifs LLDLD motifs UIMs EH domains DXF motifs UIMs ANTH domain FXDXF, DXF motifs LLDLD motif EH domains SH3 domains Rho-GEF domain PTB DXF motifs PIP kinase domain WYSPL tail peptide WVXF motifs Stonin-homology domain m-Homology domain SAC1 domain 50 -Phosphatase domain FXDXF, WVXF motifs C2A C2B C2AB BAR SH3

PI(4,5)P2 AP-2 via a-appendage Clathrin terminal domain Ubiquitin (Ub) NPF motif proteins AP-2 via a-appendage Ubiquitin (Ub) PI(4,5)P2 AP-2 via a-appendage Clathrin terminal domain NPF motif proteins PRD domain proteins CDC42 PI(4,5)P2 AP-2 via a-appendage PI(4)P, ATP; Arf6-GTP FERM domain of talin AP-2 via a-appendage ? Synaptotagmin (1,2,9) 40 -phosphate-containing phosphoinositides PI(4,5)P2 AP-2 via a-appendage Ca2þ, acidic phospholipids PI(4,5)P2; AP-2 via subdomain B of m2 Stonin 2 via m-homology domain (Acidic) phospholipids; dimerization PRD of N-WASP and dynamin

Membrane curvature induction; coat assembly

Eps15, Eps15R

HIP1, HIP1R

Intersectin

Numb, Numb-like PIPK type Ig Stonin 2

Synaptojanin

Synaptotagmin

Syndapins

Scaffolding Ub-dependent cargo endoctyosis Endocytic protein network formation Coat assembly (edges of CCPs) Ub-dependent cargo endoctyosis Plasmalemmal recruitment Coat assembly Scaffolding; links actin with endocytosis Endocytic protein network formation CDC42-mediated actin polymerization Plasmalemmal recruitment; coat assembly Localized PI(4,5)P2 formation; regulation of PI(4,5)P2 synthesis at cell adhesion sites Coat assembly; coated pit recruitment ? AP-2-dependent recycling 40 -phosphoinositide phosphate hydrolysis PI(4,5)P2 hydrolysis, CCV uncoating coat assembly; coated pit recruitment Ca2þ-triggered membrane fusion membrane fusion; SV recycling SV recycling Membrane curvature sensing actin and dynamin-mediated fission

86 Endocytosis and Presynaptic Scaffolds

Table 1 Continued

Endocytosis and Presynaptic Scaffolds 87

Figure 1 Pathways of SV recycling. Schematic depiction of various proposed modes of synaptic vesicle (SV) recycling: a fast ‘kiss-and-run’ mechanism, where the vesicle connects only briefly to the plasma membrane without full collapse (‘kiss-and-run’) or a slow clathrin-mediated pathway, which either operates from large vacuolar infoldings (cisternae) or by direct recovery of vesicle membrane from plasmalemmal CCPs (‘clathrin-mediated endocytosis’). SVs may also arise from endosomes. Components of the cytomatrix assembled at the active zone (CAZ) together with actin function as molecular scaffolds in the spatial organization of the active zone. The SV cycle is paralleled by a cycle of phosphorylation and dephosphorylation of phosphoinositides, including PI(4,5)P2, that couple exocytosis and endocytosis.

a transient fusion pore precedes complete degranulation. However, in contrast to small clear SVs that undergo local recycling, SGs need to pass through the trans-Golgi network in order to allow refilling with secretory peptides generated from larger precursor proteins. The recent development of lipophilic fluorescent styryl dyes (FM dyes) that rapidly partition into membranes and exhibit a large increase in fluorescence within this hydrophobic environment and of pH-sensitive fluorescent proteins (so-called ‘pHluorins’; Figure 2) has provided the means to follow exocytic–endocytic cycling of SVs in realtime. FM1–43 dye-based single-vesicle tracking in dissociated hippocampal neurons in culture has revealed the existence of at least two types of release: small-amplitude events that show tightly clustered rate constants of dye release, and larger events with a more scattered distribution. The small-amplitude partial release events have been attributed to a pool of vesicles that undergoes cycling by rapid closure of a narrow, approximately 1 nm diameter, fusion pore. One would therefore have to assume that vesicles are targeted for partial release by specific factors that prevent the dilation and thus the complete opening of the

fusion pore. Whether vesicles undergoing transient opening and closure of the fusion pore remain docked (‘kiss-and-stay’) or undergo local cycling (as depicted in Figure 1) is under debate. The balance between partial kiss-and-run-type and full fusion events that may be followed by clathrin-dependent compensatory endocytosis can be shifted depending on the frequency of stimulation. While kiss-and-run exocytosis may prevail under conditions of low activity, high-frequency stimulation results in predominantly complete fusion events. Membrane capacitance measurements of giant terminals (e.g., goldfish retinal bipolar cells or the calyx of Held) have also provided evidence for two kinetically distinguishable cycling vesicle pools. However, all of these studies suffer from the lack of information on specific factors that target SVs for fast kiss-and-run exocytosis–endocytosis and that allow the application of genetic or biochemical tools to molecularly distinguish the proposed kiss-and-run mode from other pathways of SV endocytosis. Vacuolar Bulk Retrieval and Synaptic Endosomes

Extensive stimulation of the presynaptic neuron results in the massive insertion of SV membrane into

88 Endocytosis and Presynaptic Scaffolds

Figure 2 Real-time measurement of SV cycling using pHluorins. Schematic illustration (A) of how synapto-pHluorin can be used to probe SV cycling. Its fluorescence is quenched in the acidic vesicular lumen, but not when residing at the plasmalemma. SynaptopHluorin signals (B) during firing of action potentials. (a)–(c) synapto-pHluorin (spH) is recycled at boutons. (a) Time course of fluorescence intensity, averaged over 13 boutons expressing spH, following stimulation with a train of 600 action potentials at 20 Hz. The dark bar shows the duration of the stimulus. The decay of fluorescence was fit by a single exponential (solid line) with t ¼ 74 s. (b) Time course of average spH fluorescence in the same boutons as those used for (a) during alkalinization with NH4Cl (dark bar). (c) Time course of fluorescence intensity at the same boutons as in (a, b) during train of 600 action potentials followed by exposure to NH4Cl 30 s after the end of the electrical stimulus. NH4Cl-induced changes are completely reversible. (d) and (e) Endocytosis, not reacidification, is rate limiting during fluorescence decay. (d) Time course of the fluorescence intensity of spH-positive boutons (n ¼ 20) following electrical stimulation (dark bar). Exocytosis of spH causes a rapid increase in fluorescence, followed by a slow decay (solid line is a single exponential fit to the average fluorescence decay, t ¼ 68 s). (e) Time course of fluorescence intensity during brief exposures to acidic solution (hatched bars below trace) before and after electrical stimulation (dark bar). Exposure to acid during resting periods led to decreases in fluorescence (quenching), indicating the presence of a resistant surface pool of spH. The fluorescence after acid quenches was similar before and after electrical stimulation, indicating that most of the newly endocytosed vesicles were rapidly reacidified. Reprinted by permission from Macmillan Publishers Ltd: Nature Cell Biology (Sankaranarayanan S and Ryan TA (2000) Real time measurements of vSNARE recycling in CNS synapses. Nature Cell Biology 2: 197–204), copyright 2000.

Endocytosis and Presynaptic Scaffolds 89

the presynaptic plasmalemma. It therefore may not be surprising that at least in some experimental systems, such as the neuromuscular junction of frogs and snakes, parts of the presynaptic membrane can be internalized via large vacuolar structures and cisternae, in particular after chemical induction of neurotransmitter release by application of high concentrations of Kþ and calcium. Some of these vacuoles may still exhibit a narrow tubular connection with the plasma membrane and are sometimes seen to contain clathrin-coated buds at their cytoplasmic ends (Figure 1). Whether such cisternal invaginations are eventually consumed by clathrin- and/or dynamin-dependent processes remains unclear. Once having undergone fission from the plasmalemma, cisternae could undergo additional budding steps and thereby constitute a form of a specialized presynaptic endosome. In fact, early endosomal markers including the small GTPase Rab5 and the SNARE protein Vti1ab are present on SVs. Rab5 mutations in Drosophila interfere with efficient release during repetitive stimulation, suggesting that presynaptic endosomes could play an important functional role in maintaining SV pools.

Clathrin-Mediated SV Endocytosis Clathrin was first purified by Barbara Pearse more than 30 years ago, using coated vesicles isolated from pig brain. In fact, clathrin is most abundantly expressed in the central nervous system, where it is found to be particularly concentrated in presynaptic nerve terminals. The importance of clathrin for SV recycling is further underscored by the fact that clathrin-coated vesicles (CCVs) isolated from nerve terminals are highly enriched in SV proteins. Clathrin, the heterotetrameric adaptor complex (AP-2), and monomeric adaptors and accessory proteins (including epsin, eps15, AP180, HIP1/HIP1R, amphiphysin, endophilin, stonin 2, etc.) play an early role in coat formation. Recruitment of AP-2 to the plasma membrane is a cooperative and presumably highly regulated process involving interactions with phosphoinositides, membrane cargo, and a variety of AP-2a ear domain-binding partners. Many of these adaptor and accessory proteins also display higher expression levels in brain than in other tissues, perhaps owing to their increased half-lives. In addition, neurons contain endocytic protein isoforms, including splice variants of clathrin light chains and aA-adaptin, AP180, auxillin, intersectin, and dynamin 1. Much of what we know about the mechanism of CCV formation has been learned from nonneuronal systems or from structural studies on clathrin, adaptor, and accessory proteins or

domains thereof. In the following sections, we summarize these data and provide a tentative model for how clathrin, dynamin, and their binding partners could act at nerve terminals. Early Steps of Clathrin-Coated Pit Formation

CCVs are formed by the coordinated assembly of clathrin triskelia built from three tightly linked heavy and associated light chains onto the plasma membrane. The recruitment and polymerization of the outer clathrin layer is assisted by mono- and heterotetrameric adaptor proteins, which simultaneously bind to clathrin, to membrane lipids, and in many cases to transmembrane cargo proteins. In addition, there is a large reservoir of preassembled flat hexagonal clathrin lattices at the plasma membrane that, however, need to undergo a structural transition involving the formation of clathrin pentagons in order to accommodate a curved membrane bud. The most important clathrin adaptor is the heterotetrameric AP-2 complex comprising two large subunits (a and b2), a medium subunit (m2), and a small subunit (s2). The two large subunits together with s2 and the amino-terminal domain of m2 (N-m2) form the trunk or core domain of AP-2, and are joined by extended, flexible ‘hinges’ to the appendage or ear domains of a- and b2-adaptins. Since AP-2 associates with clathrin, a variety of accessory endocytic proteins, phosphatidylinositol 4,5-bisphosphate [PI(4,5)P2], and membrane cargo proteins, it has been postulated to serve as a main protein interaction hub during coated pit assembly. Many accessory proteins, such as epsins, AP180/CALM, and amphiphysin, also have an adaptor function by linking clathrin assembly to membrane bud formation. These mono- or dimeric adaptors possess a folded lipid-binding domain linked to a more flexible portion of the protein harboring short clathrin- and AP-2-binding motifs, which may aid stabilization of nascent clathrin-coated pits (CCPs) during the assembly process. During CCP assembly transmembrane cargo proteins are recognized by adaptor proteins, most notably the AP-2 complex, which bind to endocytic sorting motifs within their cytoplasmic tails. These motifs include tyrosine-based Yxx (where  is a bulky hydrophobic residue) and acidic cluster di-leucine motifs, which bind directly to distinct sites within the AP-2 core domain. Yxx motifs have been co-crystallized with the carboxyterminal portion of the AP-2 m-subunit (C-m2), to which they bind in an extended conformation. Cargo recognition by AP-2 requires the presence of PI(4,5) P2, which stabilizes the protein in an open conformation that enables cargo recognition by its m2-subunit. Consistent with this, clathrin/AP-2-coated pits were

90 Endocytosis and Presynaptic Scaffolds

shown to become stabilized in living cells upon encounter of cargo receptors, suggesting that the process of AP-2 recruitment and initiation of plasmalemmal CCPs is highly cooperative. Despite intense efforts, evidence regarding the presence of canonical Yxx- or di-leucine-type endocytosis signals within SV protein cytoplasmic domains remains scarce. However, C-m2 harbors a structurally unresolved binding site for basic internalization motifs found in a variety of multimeric membrane proteins, including the presynaptic vesicle protein synaptotagmin, the presumed calcium sensor in neuroexocytosis. Neuronal synaptotagmin isoforms also interact with the AP-2-binding m-homology domain containing adaptor protein stonin 2, which is capable of targeting synaptotagmin for clathrin/ AP-2-dependent internalization in neurons as well as in transfected fibroblasts. In addition to its interaction with the ear-domain of AP-2a, stonin 2 can bind to eps15 and intersectin, thereby linking synaptotagmin with other components of the clathrin-dependent endocytic machinery in neurons. Stonin 2 (and by analogy, stoned B in Drosophila) thus represents the first endocytic adaptor protein identified that is specifically dedicated to the endocytic internalization of a SV protein. Synaptotagmin may thus regulate both the exocytic and endocytic limbs of the SV cycle. In support of this hypothesis, it has been observed that genetic or chemical perturbation of synaptotagmin function by fluorophore-assisted light inactivation in mice, flies, or worms results in recycling defects and a partial depletion of SVs. Whether other SV proteins also interact with specific endocytic adaptor proteins or get co-sorted with synaptotagmin, that is, as part of a membrane microdomain, remains an open question. CCP Maturation and Vesicle Fission

After the clathrin lattice is formed, endophilin, epsin, eps15, amphiphysins, and other proteins are involved in membrane bending and clathrin rearrangements as coated pits progressively invaginate and mature. Partitioning of the amino-terminal amphipathic helix of the ENTH domain protein epsin and perhaps other components (i.e., the small GTPase Arf6) drives the acquisition of membrane curvature. Bin–amphiphysin–Rvs (BAR) domain proteins, such as amphiphysin and endophilin, may aid membrane bending, and function as curvature sensors that signal completion of the process. Through their SH3 domains, both amphiphysin and endophilin also interact with and recruit accessory enzymes, including the large GTPase dynamin and the phosphoinositide phosphatase synaptojanin, to the nascent vesicular bud. Dynamin is required for fission

of endocytic membrane vesicles by mechanochemically constricting (‘pinchase’) the vesicle neck. The eminent role of dynamin for SV recycling is best illustrated by the dramatic phenotype seen in shibirets mutants in Drosophila, which exhibit temperature-sensitive paralysis due to the accumulation of unbudded membrane infoldings and endocytic intermediates. Observation of CCPs dynamics using evanescent wave microscopy indicates that during fission, dynamin recruitment to coated pits is rapidly followed by recruitment of actin. Moreover, perturbation of actin disrupts the endocytic reaction with accumulation of coated pits with wide necks, suggesting a role of actin, actin-binding factors, and dynamin-interacting accessory proteins, such as Abp1 or syndapin, in promoting constriction of the neck and removal of endocytosed vesicles from the membrane. In lamprey, snake, and fly neuromuscular synapse the invagination of the membrane into pits, occurs at distinct ‘endocytic zones’ surrounding the active zones of exocytosis (termed the peri-active zone). FM1–43 photoconversion and serial section electron microscopy analysis revealed that labeled clathrin-coated endocytic vesicles were clustered significantly near active zones, consistent with local exocytic–endocytic recycling vesicle pools at this synapse. Together with the regulated turnover and synthesis of membrane phosphoinositides, actin and actin-binding proteins may thus provide spatial and temporal landmarks for SV endocytosis (see the next section).

Protein Scaffolds as Spatial Regulators of Vesicle Cycling Morphologically and functionally, the active zone can be divided into two parts: the core active zone, where regulated exocytosis (and kiss-and-run-type SV retrieval) takes place; and the peri-active zone, where clathrin-mediated endocytosis occurs (see the preceding section). At the ultrastructural level, the core active zone is characterized by the more or less regular array of electron-dense material, called the presynaptic grid, presynaptic particle web, or ‘cytomatrix assembled at the active zone’ (CAZ). During recent years, multiple molecular components – both CAZ-specific ones and those that are recruited through interaction with CAZ scaffolding proteins – have been identified and characterized. Functionally, the CAZ is thought to define the site of regulated neurotransmitter release by localizing presynaptic membrane proteins, including voltage-gated calcium channels and cell adhesion molecules, to organize steps of the SV cycle, including tethering and priming of SV, and to link the exocytic machinery

Endocytosis and Presynaptic Scaffolds 91

Figure 3 Molecular organization of the cytomatrix at the active zone (CAZ). The scheme depicts observed physical interactions between active zone-specific scaffolding proteins (black), associated proteins with putative structural functions (yellow), effector proteins (blue-green), actin cytoskeletal and associated elements (green), small modulatory molecules (gray), proteins involved in SV exocytosis (blue) and endocytosis (red), as well as presynaptic membrane proteins (pink). Some of the interactions for Piccolo and RIM were discovered in pancreatic b-cells and will have to be confirmed for the CAZ. For further details see Table 2. Note: the diagram neither reflects the relative sizes of the proteins nor their exact topographic localization within the presynaptic bouton. The arrow indicates that neurexins (a-forms) are involved in the localization of calcium channels.

with elements of the endocytic zone and with the surrounding actin cytoskeleton (Figure 3). In addition, CAZ elements turned out to be essential mediators of presynaptic plasticity. Molecular Organization of the CAZ

Relatively few CAZ-specific structural and effector proteins have been identified to date that are believed to constitute the scaffold of the CAZ and to mediate its structural and functional organization (black in Figure 3). These proteins belong to four different protein families: the Rab3-interacting molecules (RIMs), the mammalian Unc13 proteins (Munc13s), the two related giant CAZ scaffolding proteins Bassoon and Piccolo, and the ELKS/CAST proteins (Figure 3; Table 2). RIMs are multidomain proteins that were identified as effectors of Rab3, a small GTPase associated with SVs. In particular, the a-forms of RIM1 and RIM2 are important scaffolding molecules that interact with multiple other presynaptic proteins. These include isoforms of Munc13, a liaison that might be involved in

making SVs fusion-competent, as well as ELKS/CAST and Piccolo. Analysis of RIM1a-deficient mouse mutants revealed that this protein is involved in short- and long-term forms of presynaptic plasticity. For example, long-term potentiation at hippocampal mossy fiber terminals or parallel fiber synapses of the cerebellum involve protein kinase A-dependent phosphorylation of RIM1a and phosphorylationdependent binding of 14-3-3 proteins. Further interactions of aRIMs point to a central role of these proteins in active zone organization. They can bind presynaptic voltage-gated Ca2þ channels, either directly or via their binding to RIM-binding proteins, and the Ca2þ sensor synaptotagmin, which is involved in both exocytosis and endocytosis of SVs. The interaction with a-liprins, originally identified as a cytoplasmic adaptor for the receptor tyrosine phosphatase LAR, might serve the formation and maintenance of active zones, as suggested by work in invertebrates. Munc13 isoforms are involved in SV priming and the regulation of synaptic plasticity. They can link

Protein family Relevant members and synonymous names

Domains and motifs

Interacting protein

Proposed function(s)

RIMs – Rab3-interacting molecules

Zn finger

RIM1a; RIM2a,b,g; RIM3g; RIM4g; UNC10 (C. elegans)

Region between Zn finger and PDZ domain PDZ domain C2A

Rab3 Munc13-1, ubMunc13-2 14-3-3 cAMP-GEFII / Epac2 ELKS / CAST Piccolo N-type calcium channels RIM-BPs a-Liprin, N-type calcium channels synaptotagmin

Link to SV, tethering of SV, presynaptic plasticity SV priming Presynaptic plasticity Insulin secretion in pancreatic b-cells CAZ scaffolding CAZ scaffolding Channel anchoring and/or clustering CAZ scaffolding CAZ scaffolding, Anchoring and/or clustering of Ca channels CAZ scaffolding, organization of Ca sensing

Pro-rich motif btw. C2A & B C2B

RIM-binding proteins RIM-BP1 RIM-BP2

SH3 domains 2 and/or 3

RIMs N-type (Cav2.2) and L-type (Cav1.3) Ca channels

CAZ Scaffolding Channel anchoring and/or clustering

Piccolo–Bassoon family Bassoon and Piccolo

Zn fingers Between CC1 and CC2

PRA1 (prenylated Rab3 acceptor) CtBP/BARS Ribeye/CtBP2 ELKS / CAST Abp1 (actin-binding protein 1) GIT (ARF-GAP) Profilin cAMP-GEFII/Epac2 RIMs Piccolo L-type Ca channel a-RIMs

Link to SV? Membrane trafficking? Scaffolding of synaptic ribbons CAZ scaffolding Link to actin cytoskeleton, endocytosis Membrane trafficking Actin regulation Insulin secretion in pancreatic b-cells CAZ scaffolding, organization of SV cycle Homophilic interaction, scaffolding Channel anchoring in b-cells CAZ scaffolding,

Piccolo/Aczonin only

UNC13 proteins

CC3 N-terminal Q domain Between CC1 and CC2 Pro-rich region btw CC1 and 2 PDZ domain C2A C2A and C2B M13-1 and ubM13-2 N-term.

92 Endocytosis and Presynaptic Scaffolds

Table 2 Protein components of the cytomatrix assembled at the active zone

Munc13-1 bMunc13-2 ubMunc13-2 Munc13-3 Munc13-4 UNC13 (C.e.) Dunc13 (D.m.)

Conserved homology region (MUN)

C-terminus

ELKS/CAST proteins ELKS1A/ERC1A ELKS1B/ERC1B ELKS2/CAST/ERC2

CC regions

a-Type Liprins Liprin-a1–a4 SYD-2 (C.e.)

N-term CC region

C-terminus

C-terminal PDZ binding site

SV priming, control of SNARE complex formation

Bassoon, Piccolo a-Liprins RIMs Syntenin

CAZ scaffolding CAZ scaffolding CAZ scaffolding Active zone organization?

RIMs ELKS/CAST GIT (ARF-GEF)

CAZ scaffolding CAZ scaffolding Regulation of membrane trafficking and actin cytoskeleton Transport Regulation of membrane protein anchoring

Kif1A (kinesin motor) LAR receptor protein tyrosine Phosphatase b-Liprin CASK (MAGuK) MALS/Veli GRIP

Regulation of transsynaptic adhesion? Regulation of transsynaptic adhesion? Receptor transport and clustering/postsynaptic

Endocytosis and Presynaptic Scaffolds 93

SAM domains

Ca2þ-dependent plasticity Anchoring to actin/spectrin cytoskeleton Regulation of actin cytoskeleton ?

Calmodulin Spectrin b-spIIIS msec7-1 ARF-GEF DOC2a (double C2 domain protein) Syntaxin (in debate)

94 Endocytosis and Presynaptic Scaffolds

to the actin-spectrin cytoskeleton via binding to a b-spectrin isoform and to presynaptic membranetrafficking processes via the ARF guanine nucleotide exchange factor msec7-1. ELKS/CAST proteins display a very high content of coiled-coil structures and are considered as a major structural component of the CAZ. They can physically interact with RIMs as well as with Bassoon and Piccolo and thus might interconnect the major scaffolding proteins of the CAZ. Bruchpilot, an ELKS/ CAST-related protein in Drosophila, is responsible for the proper anchoring of particular specializations of the CAZ, so-called T-bars, to the active zone membrane. Via syntenin, ELKS/CAST proteins also connect to neurexins, which are specific cell adhesion molecules of the presynaptic membrane occurring as long a- and short b-forms involved in Ca2þ channel localization and linkage to postsynaptic neuroligins, respectively. Neurexins are additionally anchored to the presynaptic cytomatrix via a trimeric protein complex of CASK, Mint, and MALS/Veli, which in turn links to Ca2þ channels and a-liprins (Figure 3). Bassoon and Piccolo are considered as very large multidomain scaffolding molecules of the CAZ. Most of their interaction partners have still to be discovered. They both bind ELKS/CAST and the small prenylated Rab3 acceptor (PRA1), which potentially links into the SV cycle. In addition, Bassoon has been reported to interact with CtBP/BARS and its homolog RIBEYE, a specific component of synaptic ribbons in retinal photoreceptors and inner ear hair cells. The RIBEYE–Bassoon interaction was suggested to be essential for anchoring of ribbons to the presynaptic plasmalemma. The interaction of Bassoon (and potentially also Piccolo) with CtBP/BARS is of interest, as this protein has been implicated in vesicular fission from the trans-Golgi complex. By analogy, a similar function might be envisioned in the presynapse. Links of the CAZ to the Endocytic Machinery and the Actin-Based Cytoskeleton

While multiple interactions of the previously described CAZ proteins underscore the role of the presynaptic cytomatrix in organizing the apparatus for regulated exocytosis, a link to clathrin-mediated endocytosis is less clear. Specific interaction partners for Piccolo might be of particular interest in this context, as they can provide the basis for the physical linkage of exocytic and endocytic presynaptic processes. On the one hand, Piccolo can bind directly to Ca2þ channels and the guanine nucleotide exchange factor cAMP-GEFII/Epac2, as revealed from studies on pancreatic b-cells. Moreover, the C2A domain of Piccolo is discussed as a candidate low-affinity

Ca2þ sensor. These characteristics argue for a role in active zone organization and exocytosis. On the other hand, Piccolo can bind to the ARF-GTPase-activating protein GIT, which has been implicated in endocytic processes such as receptor internalization. The N-terminus of Piccolo specifically interacts with Abp1, an actin-binding factor directly regulating the GTPase dynamin. Indeed, the N-terminal Q-domain of Piccolo can interfere with endocytic processes in heterologous systems. Yet another link between Piccolo and the actin cytoskeleton is profilin, a small G-actin- and phosphoinositide-binding protein that is involved in local remodeling of the actin cytoskeleton. Thus, Piccolo might indeed be an important mediator between the neurotransmitter release apparatus and the neighboring endocytic machinery.

Regulation of the SV Cycle by Membrane Lipids Phosphoinositide Regulation of SV Cycling

Cycling of presynaptic vesicles requires the precise spatial and temporal regulation of protein–lipid interactions. Many synaptic proteins – including synaptotagmin1, calcium-dependent activator protein for secretion (CAPS), the Munc18-interacting proteins Mint-1 and Mint-2, voltage-gated P/Q-type calcium channels, and a variety of endocytic proteins such as AP-2, AP180, epsin, and dynamin (Table 1) – directly bind to and are regulated by [PI(4,5)P2]. SV cycling thus appears to be nested into a local cycle of phosphoinositide phosphorylation and hydrolysis. Accordingly, PI(4,5)P2 acts at multiple stages of the vesicle cycle. Knockout mice lacking PIP kinase type Ig, the major PI(4,5)P2-synthesizing enzyme at synapses, display defects in neurotransmitter release and endocytic recycling that cause synaptic depression. The activity of PIPK Ig is regulated by a variety of factors, including phosphatidic acid (PA), the small GTPases Arf6 and Rac1, and the actin cytoskeletonassociated adhesion protein talin. Association of PIPK Ig with these factors is dependent on their phosphorylation status, providing a means for the temporal and spatial regulation of phosphoinositide metabolism. PI(4,5)P2 is eventually consumed by synaptojanin-mediated dephosphorylation, resulting either in formation of PI(4)P and perhaps PI, cleavage via phospholipase C, or PI3 kinase-dependent synthesis of PI(3,4,5)P3. At least in neuroendocrine PC12 cells, PI(4,5)P2 appears to be concentrated within cholesterol-enriched microdomains near release sites, where it may aid vesicle docking and/or fusion. The recent observation that SV proteins remain clustered during their exocytic–endocytic journey, together

Endocytosis and Presynaptic Scaffolds 95

with the extremely high cholesterol content of SV membranes, suggests that cholesterol-enriched microdomains could serve to spatially organize the SV cycle, perhaps in part by locally concentrating PI(4,5)P2. At present we can only speculate about the exact mechanism of action of PI(4,5)P2 during vesicle fusion, but its role in clathrin-mediated synaptic vesicle endocytosis is much better understood. PI(4,5)P2 is an important factor in recruiting endocytic adaptor and accessory proteins to the membrane where these form a network of protein–protein interactions. The stability of this network critically depends on the PI(4,5)P2 content of the membrane, as suggested by the observation that CCVs accumulate in nerve terminals of synaptojanin knockout mice. These observations also indicate that PI(4,5)P2-hydrolysis may normally occur concomitantly with or directly after dynamin-mediated membrane fission. CCVs at presynaptic sites of synaptojanin knockout mice become trapped in a meshwork of actin filaments, consistent with the fact that PI(4,5)P2 regulates actin polymerization and drives the formation of actin comet tails that may help to propel endocytic vesicles away from the plasmalemma. Lipids and Membrane Deformation

SV fusion and the subsequent formation of endocytic clathrin-coated buds at the presynaptic peri-active zone involve radical geometric remodeling of the membrane in order to generate areas of different membrane curvature. While lipids with bulky polar headgroups and saturated or single fatty acid tails, including lysolipids and many glycolipids, promote positive curvature, lipids with compact headgroups and space-filling hydrophobic tails, such as PA and diacylglycerol (DAG), favor negative curvature. For example, lysophosphatidic acid (LPA) and PA, which are interconverted by LPA-acyl transferase and phospholipase A2 activities, respectively, favor opposite curvatures. Although differential distribution of distinct types of lipids between the two membrane leaflets may contribute to membrane deformation, it is generally assumed that membrane bending requires the action of proteins. During endocytic SV recycling, the forming bud must adopt a positive curvature at the bud center and negative curvatures at the edges. Endocytic proteins may act by one of several mechanisms to bend membranes.

As stated previously, epsin, a PI-(4,5)P2-binding clathrin accesssory protein, is able to partition into the cytoplasmic leaflet of the plasma membrane via PI (4,5)P2-induced formation of an extra a-helical segment that drives acquisition of positive curvature. Dimeric BAR domain-containing proteins, including amphiphysin or endophilin (Table 1), induce curved membranes by an additional amphipathic helix at their amino-terminal end, largely via electrostatic interactions of their concave surface with negatively charged membrane phospholipids. Endophilin has originally been proposed to be an LPA-acyl transferase, but this activity has recently been called into question. Highly curved membranes may become stabilized in addition by other scaffolding proteins, including clathrin itself, which forms a rigid basket around the emanating vesicular membrane bud. Moreover, the transmembrane domains of synaptic vesicle proteins could provide a barrier that prevents local areas of high curvature from lateral diffusion and thereby contribute to maintaining vesicle identity. Finally, protein–lipid interactions and the formation of microdomains might also underlie the choice between fast and slow modes of SV cycling that would require a tight control of fusion pore expansion and constriction. See also: Synaptic Vesicles.

Further Reading Dresbach T, Qualmann B, Kessels MM, et al. (2001) The presynaptic cytomatrix of brain synapses. Cellular and Molecular Life Sciences 58: 94–116. Galli T and Haucke V (2004) Cycling of synaptic vesicles: How far? How fast! Science Signal Transduction Knowledge Environment 264: re19. Gundelfinger ED, Kessels MM, and Qualmann B (2003) Temporal and spatial coordination of exocytosis and endocytosis. Nature Reviews Molecular Cell Biology 4: 127–139. Murthy VN and De Camilli P (2003) Cell biology of the presynaptic terminal. Annual Review of Neuroscience 26: 701–728. Royle SJ and Lagnado L (2003) Endocytosis at the synaptic terminal. Journal of Physiolology 553: 345–355. Shankaranarayanan S and Ryan TA (2000) Real time measurements of vSNARE recycling in CNS synapses. Nature Cell Biology 2: 197–204. Sudhof TC (2004) The synaptic vesicle cycle. Annual Review of Neuroscience 27: 509–547. Ziv NE and Garner CC (2004) Cellular and molecular mechanisms of presynaptic assembly. Nature Reviews Neuroscience 5: 385–399.

Postsynaptic Density/Architecture at Excitatory Synapses H-C Kornau, Center for Molecular Neurobiology (ZMNH), University of Hamburg, Hamburg, Germany ã 2009 Elsevier Ltd. All rights reserved.

Introduction Excitatory chemical synapses of the central nervous system are highly specialized structures which rapidly convey information from one neuron to the other by means of a soluble neurotransmitter, in most cases glutamate. The neurotransmitter is released from the presynaptic terminal upon action potential-evoked calcium influx, diffuses through the synaptic cleft, and opens ligand-gated ion channels on the surface of the postsynaptic cell, resulting in excitatory postsynaptic currents. Signal reception and processing is achieved in dendrites, which are often contacted by presynaptic terminals of thousands of neurons. Excitatory synapses are mostly located on spines, little mushroom-shaped structures protruding from the dendritic shaft. Dendritic spines form a compartment for processing of the individual synaptic input. In electron micrographs the presynaptic side is recognized by a high density of neurotransmitter-containing vesicles. On the postsynaptic side of the cleft a characteristic electron-dense thickening of the membrane, extending approximately 30 nm into the spine cytosol, is observed (Figure 1); this thickening, a structure that has attracted the interest of neurobiologists since the 1950s, is called the postsynaptic density (PSD). Given that synapses are crucial for information processing in the brain, the importance of the PSD, as the site where neurotransmitter receptors are connected to intracellular structures and where synaptic processing within each spine is initiated, is obvious. Molecular biology and biochemical approaches have identified many proteins of the PSD, including (1) adhesion molecules, receptors, and channels, (2) scaffolding, adaptor, and cytoskeletal proteins, and (3) signaling/regulatory proteins, including kinases, phosphatases, small GTPases, and similar molecules. Moreover, the studies revealed how these proteins physically interact with each other or with the postsynaptic membrane. Thus, we can draw preliminary models containing some of the major components of the PSD to get an idea of the way this macromolecular protein complex may function and how it is rearranged during development and in processes of synaptic plasticity.

central nervous system. PSDs are 25- to 50-nm-thick protein assemblies with a diameter of several hundred nanometers. High-resolution electron microscopy reveals a structured 5-nm-thick sheet, which might form the nucleus of the PSD, as well as filamentous structures which connect it to cytoskeletal elements within the cytoplasm of the dendritic spine and to organelles such as the smooth endoplasmic reticulum at the margin of the PSD. Due to their unique density and their resistance to nonionic detergents, PSDs can be purified as intact, disk-shaped structures by differential centrifugation protocols. These preparations serve as the protein sources for the molecular analysis of PSDs. Initially, individual PSD proteins were discovered by amino acid sequencing of protein bands of such PSD preparations after separation by gel electrophoresis. Alternatively, antibodies were raised against complex PSD preparations and used to screen cDNA expression libraries. Coimmunoprecipitates of known PSD components have been a source for finding new elements of this structure as well. Yeast two-hybrid screening, a genetic screen for interacting proteins, supplied a number of proteins physically binding to glutamate receptors or scaffolding proteins of the PSD. Recently, proteomics, which is generally regarded as a large-scale analysis of proteins in a given sample, involving efficient protein separation and highly sensitive protein identification techniques, has been applied to glutamate receptor complexes and to PSD preparations. The results suggest that the PSD contains several hundred different polypeptides. Mass spectrometry techniques have not only identified these proteins with great sensitivity, they have also allowed quantitative determinations. Thus, for some of the components, we know their stoichiometries. Using scanning transmission electron microscopy on a rat forebrain PSD fraction, the mass of an average PSD was measured and the number of copies of specific PSD proteins derived from their relative mass contributions was determined by quantitative gel electrophoresis (Table 1). Consistent with an independent approach using a green fluorescent protein-based calibration technique, these data suggest that a single PSD contains several hundred copies of each of its major polypeptides.

The Protein Composition of Postsynaptic Densities

How Postsynaptic Densities Are Studied

Glutamate Receptor Complexes

Electron microscopy initially identified PSDs as hallmarks of glutamatergic/Gray type I synapses in the

Glutamate receptors and associated proteins were among the first components identified to be part of

96

Postsynaptic Density/Architecture at Excitatory Synapses 97

Figure 1 The PSD, a structural element of excitatory synapses. The electron micrograph shows a representative excitatory synapse (Gray type I) at a synaptic spine (s) of an adult mouse hippocampus. Major synaptic structures are the presynaptic bouton (b) filled with synaptic vesicles (diameter
50 nm), the PSD (arrowheads), and the synaptic cleft in between. The electron micrograph was a kind gift of Dr. Michaela Schweizer, Central Service Facility of Morphology at the ZMNH, University of Hamburg, Hamburg, Germany.

Table 1 Dimensions of a PSD Mean diametera Depthb Mean massa Number of PSD-95 moleculesa


360 nm
25–50 nm
1.1 GDa
300

a

Data from Chen X, Vinade L, Leapman RD, et al. (2005) Mass of the postsynaptic density and enumeration of three key molecules. Proceedings of the National Academy of Sciences of the United States of America 102: 11551–11556. b Data from Valtschanoff JG and Weinberg RJ (2001) Laminar organization of the NMDA receptor complex within the postsynaptic density. Journal of Neuroscience 21: 1211–1217.

the PSD. Glutamate receptor channels mediate the principal postsynaptic functions – that is, changes in membrane potential upon binding of the excitatory neurotransmitter. The different types of ionotropic glutamate receptors are named for their specific

agonists. a-Amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptor channels are opened upon glutamate binding alone, whereas N-methylD-aspartate (NMDA) receptor activation requires simultaneous membrane depolarization (e.g., by activation of neighboring synapses). This depolarization relieves the extracellular magnesium block of the NMDA channel, allowing influx of sodium and, importantly, calcium through the channel pore. Calcium acts as a second messenger within the cell and is important for the induction of synaptic plasticity. A very abundant PSD protein involved in these processes and activated by calcium is calcium/ calmodulin-dependent protein kinase II (CaMKII). It is associated with the cytoplasmic region of the NMDA receptor, allowing it to sense the influx of calcium through this channel directly. Subsequent pathways involve a variety of structural rearrangements within the PSD and lead to a change in synaptic efficacy primarily expressed as an altered number of AMPA receptors in the postsynaptic membrane. Both native AMPA and NMDA receptor complexes are associated with a variety of scaffolding/adaptor proteins (e.g., postsynaptic density protein 95), which link the ionotropic glutamate receptors to other integral membrane proteins, to soluble signaling molecules, and to the spine cytoskeleton, and form the core lattice of the PSD. At the margin of the PSD, metabotropic G-protein-coupled receptors for glutamate (mGluRs) modulate the intraspinal calcium level. Proteomics has unraveled individual protein complexes for NMDA receptors, mGluRs, and AMPA receptors. The NMDA receptor complex is tightly linked to the core of the PSD by a series of adaptor proteins comprising postsynaptic density protein 95 family members (PSD-95; also called synapse-associated protein 90 (SAP90), PSD-93/Chapsyn-110, SAP102, and SAP97), guanylate kinase-associated proteins (GKAPs; also called SAP90/PSD-95-associated proteins, or SAPAPs), and Shanks (also called proline-rich synapse-associated proteins, or ProSAPs) (Figure 2). PSD-95 and its relatives encompass three PDZ domains, an SH3 domain, and a guanylate kinase domain, all of which function as protein–protein interaction domains. They belong to the family of membrane-associated guanylate kinases (MAGUKs), which function primarily at sites of cell-to-cell contact. Employing its PDZ domains, PSD-95 links the NMDA receptor to neuroligins, postsynaptic cell adhesion molecules crucial for synapse formation, and to several signaling molecules, among them neuronal nitric oxide synthase (nNOS) and synaptic GTPase-activating protein (synGAP), an abundant regulator of Ras and Rap GTPases. The GKAPs connect PSD-95 to the Shank

98 Postsynaptic Density/Architecture at Excitatory Synapses Synaptic cleft AMPAR

TARP

NMDAR

mGluR

Plasma membrane PSD-95 protein family

Homer

GKAP Shank sheet Spine cytoskeleton Figure 2 Simplified scheme of the molecular architecture of the PSD. The different types of glutamate receptors are linked to the Shank scaffold, a two-dimensional sheet, either by the PSD-95 family of proteins and by guanylate kinase-associated protein (GKAP) molecules, or by Homer dimers. The Shank scaffold connects the PSD with the spine actin cytoskeleton. The PSD-95 family proteins form a membrane-proximal layer of scaffolds, whereas the Shanks constitute a second, deeper layer. PDZ domains are depicted as blue circles. The C-terminal domains of the glutamate receptor subunits, the PSD-95 family members, as well as the Shanks, recruit numerous signaling molecules to the PSD. The PSD-95 family members connect the PSD with cell adhesion molecules governing the integrity with the presynaptic specialization. The Homer proteins link the metabotropic glutamate receptors (mGluRs) not only to the Shank scaffold, but also to the inositol 1,4,5-trisphosphate receptor within the smooth endoplasmic reticulum. For simplicity, all of these connections were omitted from the scheme. The figure does not reflect the real stoichiometry of the scaffolding proteins. AMPAR, a-amino-3-hydroxy-5methyl-4-isoxazole propionic acid receptor; TARP, transmembrane AMPA receptor regulating protein; NMDAR, N-methyl-D-aspartate receptor.

core of the PSD by binding to both the guanylate kinase domain of PSD-95, on one hand, and to the PDZ domain of a Shank molecule, on the other hand. Four closely related GKAP family members show distinct spatiotemporal expression patterns in the brain. Shanks (three related proteins) contain a number of protein interaction domains, including ankyrin repeats, an SH3 domain, a PDZ domain, a proline-rich region, and a sterile alpha allowing them to bind to a variety of proteins. Moreover, by self-association of their SAM domains they are able to form large sheets within the PSD. The resulting dense platform is associated with the actin cytoskeleton, the smooth endoplasmic reticulum, the different glutamate receptor complexes, and other channels and receptors. Shanks are also early components during synaptogenesis. The three classes of adaptor proteins play a crucial role within the PSD. They form an axis of scaffolds differing in their distance from the plasma membrane (Figure 2): PSD-95 is located approximately 12 nm cytoplasmic to the extracellular face of the postsynaptic membrane, like the C-termini of the NMDA receptor, whereas GKAP and Shank molecules are buried deeper in the cytoplasm (approximately 24–26 nm from the membrane). They recruit a number of signaling molecules to the PSD and connect the PSD with cytoskeletal elements. Proteomic analysis of the NMDA receptor complex has revealed a complexity of more than 100 proteins, including a variety of protein kinases and phosphatases, small G-proteins

and their regulators, and other signaling molecules, such as nNOS, phosphatidylinositol 3-kinase (PI3K), phospholipase C-g, Citron, and Arg3.1/Arc. In addition, several cell adhesion molecules and associated adaptors, including N-cadherins and b-catenin, which regulate dendritic spine morphogenesis, were found in the NMDA receptor complex. Finally, many cytoskeletal proteins are part of this complex. Type I metabotropic glutamate receptors are located at the margin of the PSD. These G-proteincoupled receptors lead to inositol 1,4,5-trisphosphate (IP3) production, which consequently activates the IP3 receptor and induces the release of calcium from intracellular stores. Key molecules of the mGluR protein complex are the Homer proteins, which are expressed from three genes. They can dimerize by a coiled-coil interaction and these dimers connect mGluRs to the Shank backbone of the PSD on the plasma membrane side, and IP3 receptors in the smooth endoplasmic reticulum to the Shank backbone on the cytoplasmic side, of the PSD. Thus, Homer proteins connect mGluRs and IP3 receptors to the PSD and link mGluRs and IP3 receptors, allowing efficient intracellular calcium release upon mGluR activation. In these assemblies the N-terminal EVH domains of the Homer dimers bind to a PPXXF motif in the receptors and to a proline-rich region in Shanks (Figure 2). (The EVH nomenclature derives from ‘enabled-VASP (vasodilator-stimulated phosphoprotein) homology.’) Interestingly, Homer 1a, an

Postsynaptic Density/Architecture at Excitatory Synapses 99

isoform lacking the coiled-coil domain, which is highly expressed as an immediate-early gene, can compete with the constitutive forms of Homer and thus with the link of mGluRs and IP3 receptors to the PSD. This may enable the cell to limit the release of intracellular calcium if required. The mGluR5 complex has been also subject to proteomic analysis. It provided a number of additional candidate proteins of mGluR complexes and confirmed the link to Homer, Shank, and IP3 receptors delineated herein. AMPA receptors mediate the majority of the fast excitatory neurotransmission in the brain. The efficacy of a synaptic connection is related to the number of AMPA receptors in the postsynaptic membrane. Thus, whereas NMDA receptors are mediating the induction of synaptic plasticity, AMPA receptors are crucial for its expression. The number of AMPA receptors in the postsynaptic membrane is indeed quite variable, ranging from zero to several hundred receptors per synapse, and can rapidly change upon usage of a synapse. AMPA receptors are rapidly cycling between plasma membrane and endosomal compartments. Thus, they need a unique set of associated proteins. The best documented interaction of AMPA receptors is their binding to the integral membrane protein stargazin. Stargazin was the first member of a family of four differentially expressed TARPs (for transmembrane AMPA receptor regulatory proteins), which function as auxiliary subunits of AMPA receptors. By two independent interaction sites, TARPs regulate both the trafficking and the gating of AMPA receptors. A C-terminal interaction of TARPs with PSD-95 connects AMPA receptors with the PSD and governs the synaptic localization of AMPA receptors (Figure 2). The rapid cycling of AMPA receptors requires the interaction with AP2 and clathrin. These and N-ethylmaleimide-sensitive fusion protein NSF, a multimeric ATPase playing an important role in membrane fusion, bind to specific motifs in the center of the short, intracellular C-termini of selected AMPA receptor subunits. Additional proteins interact with the very C-terminal amino acids of AMPA receptor subunits and affect the synaptic localization of AMPA receptors. These include an adaptor protein termed glutamate receptor-interacting protein (GRIP) and a protein originally identified as a protein interacting with C kinase, PICK1. Both of these proteins bind to the C-termini of AMPA subunits by PDZ domains. GRIPs (two related proteins) harbor seven PDZ domains and thereby can connect AMPA receptors to several cytoplasmic and membrane-standing proteins (e.g., the ephrin receptor). The PICK1 interaction with AMPA subunits is essential for the

AMPA receptor endocytosis involved in cerebellar long-term depression. Proteomic approaches to tackle the composition of AMPA receptor complexes identified a limited complexity as compared to NMDA receptor or mGluR complexes. However, AMPA receptor cycling between plasma membrane and endosomal locations may specifically require transient or weak interactions, which may be lost during purification of the receptor complex. Multiple interfaces between the individual glutamate receptor multiprotein complexes appear to exist: Homer (mGluR complex) can bind to Shanks in addition to mGluRs and IP3 receptors, and stargazin (AMPA receptor complex) binds to PSD-95 (NMDA receptor complex). Proteomic studies of PSD proteins have been performed on immunopurifications of glutamate receptors (brain protein complexes purified via peptides derived from the Cterminal sequence of the NMDA receptor subunit 2B, a target of PSD-95 family members) and on purifications of entire PSDs. The data from these different approaches have been compiled as our current view of the PSD composition. They show that in addition to the glutamate receptor complexes connected to the different scaffolding proteins we have to integrate a large number of additional integral membrane proteins, signaling/regulatory proteins, and cytoskeletal elements into our model of the PSD. Although the number of proteins identified in PSD fractions goes into the hundreds, only a few principal types of molecule classes can be covered here. Integral Membrane Proteins

Independent analyses of the PSD proteome revealed a variety of integral membrane proteins aside from glutamate receptors. These include not only other channels and receptors, but also cell adhesion molecules (e.g., neuroligins and N-cadherin) which directly link the PSD with the juxtaposed membrane specialization on the presynaptic side termed the active zone. Cell adhesion molecules govern several steps of synaptogenesis and may keep the active zone and the PSD in register during structural changes of the synapse. Regulatory Proteins

The importance of CaMKII as a very abundant PSD protein for the induction of synaptic plasticity has already been mentioned. However, many other types of protein kinases and phosphatases are found in the PSD as well. This is not surprising since phosphorylation and dephosphorylation events often mediate

100 Postsynaptic Density/Architecture at Excitatory Synapses

functional and structural modulation of central nervous system synapses. Small GTPases of the Ras superfamily, termed Ras and Rap, belong to the most important molecules involved in the integration of biochemical signals in the dendritic spine. Their activity determines the amount of AMPA receptors in the postsynaptic membrane. As GTPases they switch between an active, GTP-bound, and an inactive, GDP-bound, state. Activation requires a specific guanine nucleotide exchange factor (GEF); inactivation requires a GTPase-activating protein (GAP). PSD preparations contain several GEFs and GAPs for Ras and Rap, which themselves differ by the signals required for their activation. Small GTPases of the Rho family, including Rac and cdc42, directly influence the state of the spine cytoskeleton. For example, adhesion molecule-mediated recruitment of specific Rac GEFs activates Rac and leads to changes in actin polymerization. These GTPases have a strong impact on the morphology of dendritic spines and their regulators are found in PSD preparations. Cytoskeletal Proteins

The axis of scaffolds in the PSD (PSD-95 family members, GKAPs, and Shanks) is also a nucleus for interactions with the cytoskeleton in the spine. As already indicated in the section on the NMDA receptor complex, PSD preparations contain a variety of cytoskeletal proteins. They include actin, actinbinding proteins, tubulin, spectrin, cortactin, and many others. The proteomic approaches revealed several hundred different proteins in PSD preparations. However, it should be noted that this does not necessarily reflect the complexity at a single PSD. The protein composition of PSDs in different brain regions as well as in different synapses of a single neuron may considerably differ, and thereby may contribute to the large number of identified proteins.

PDZ Domains Like all macromolecular protein networks, the PSD is connected by modular protein–protein interactions. Analysis of the interacting protein sequences has revealed the presence of specific domains. A domain frequently found at sites of cell-to-cell contact is the PDZ domain (so named from PSD-95/discs-large/ zonula occludens-1), which is present in many different proteins in humans. PDZ domains bind, in most cases, to short peptide motifs at the C-termini of other proteins and are classified according to the

sequences they recognize. In addition, some PDZ domains can heterodimerize. The PDZ domain is also central in connecting different components of the PSD (Figure 2). It was first recognized in the prototype PSD scaffolding protein PSD-95 (hence, the name), which contains three N-terminal PDZ domains and uses two of them to bind to the C-termini of NMDA receptor subunits and the third PDZ and other domains to connect the receptor to a variety of PSD components. PDZ domains are essential for the function of many signaling and scaffolding proteins of the PSD. The scaffolding proteins often contain several PDZ as well as other interaction domains, each with a preferred binding partner, allowing the scaffold to coordinate large protein complexes.

Rearrangements of Postsynaptic Densities Developmental Assembly of Postsynaptic Densities

Most of the synapse formation in the rat central nervous system occurs during the first two postnatal weeks. It is initiated by an interaction between preand postsynaptic cell adhesion molecules within the synaptic cleft, including neurexins/neuroligins, cadherins, synCAMs, receptor tyrosine kinases, and others. Among the earliest proteins at synapses are PSD-95 family members and Shanks, suggesting that they play a major role in the assembly of the PSD. Their interactions with the other scaffolding and signaling molecules are thought to promote the formation of the PSD very rapidly. Simultaneously, NMDA receptors are recruited into the PSD, whereas the integration of AMPA receptors into the postsynaptic membrane requires presynaptic glutamate release. Early in development many synapses in the central nervous system lack AMPA receptors and are therefore called silent synapses. PSDs of postnatal day 2 rats contain already many of the proteins found in adult PSDs. However, during development, specific molecular changes are evident. These include a switch in the expression of NMDA receptor subunits from NR2B to NR2A and, simultaneously, of their interaction partners SAP102 to PSD-95, and an increase of AMPA receptors and CaMKII. Although during synaptogenesis the recruitment of key PSD proteins such as NMDA receptor subunits and Shanks appears to occur in a rather gradual manner, preformed transport packets containing either AMPA or NMDA receptors, or mobile clusters containing PSD-95, GKAP, and Shank, driven by actin transport, have been described. Thus, how the PSD is assembled during synapse formation requires further investigation.

Postsynaptic Density/Architecture at Excitatory Synapses 101 Activity-Dependent Changes at Postsynaptic Densities

The size, protein composition, and structure of the PSD are altered in response to developmental and environmental cues. One cellular model for learning and memory, long-term potentiation (LTP), involves a prolonged strengthening of a synapse in response to a high-frequency input. Calcium influx through NMDA receptors, which are open under these depolarizing conditions, allows activation of a number of signaling molecules of the PSD, including CaMKII, calcineurin, nNOS, and several regulators of small GTPases. The primary result is a change in the phosphorylation pattern of PSD proteins, among them the glutamate receptor subunits. Phosphorylation leads to an altered affinity of binding partners and concomitant stabilization or destabilization of the receptor molecules at the PSD. Activation of small G-proteins can trigger endo- or exocytosis of membrane proteins. As an example, additional AMPA receptor subunits are rapidly inserted into the plasma membrane during LTP, resulting in a direct increase of the postsynaptic response. In contrast, AMPA receptors are removed from the synaptic plasma membrane during long-term depression. Other small GTPases, upon activation, alter the polymerization state of actin in the dendritic spine. Thus, synaptic plasticity involves structural plasticity at the PSD. In addition to changes in the phosphorylation pattern, receptor localization, and cytoskeleton structure, local translation and targeted protein degradation contribute to the high capability of the PSD to rearrange. Activity-dependent translation of AMPA receptor subunits in dendrites may mediate synapsespecific modifications. The presence of mRNAs for other PSD components (e.g., Shanks) in dendrites suggests their local translation as well. Dendritic protein synthesis may contribute significantly to synaptic plasticity processes. Posttranslational modifications aside from phosphorylation affect the PSD structure. As an example, PSD-95 family proteins are differentially palmitoylated at two cysteine residues, with important consequences for their clustering ability. Changes in the PSD structure and composition can also result from altered patterns of transcription and splicing in the nucleus. Alternative splicing events affect the interactions of various PSD proteins by adding/removing sequences encoding specific interaction domains or motifs. For example, two forms of PSD-95, a palmitoylated one (see earlier) and one with an N-terminal L27 domain which enables additional interactions of PSD-95, result from alternative splicing. Rearrangements of the PSD may not only result from alterations in the amount or modification

of scaffolding proteins, but may also be influenced by their propensity to interact as a result of alterations in the ionic environment. Of interest in this context, it has been shown that the zinc concentration can, at least in vitro, directly influence the assembly of Shank fibers, a core element of the PSD.

The Physiological Role of Postsynaptic Densities What may be the function of this complex protein assembly? Based on its localization and content, the clustering, localization, and regulated trafficking of glutamate receptors, the compartmentalization of different glutamate receptor complexes, the spatial organization of signaling cascades, the dynamic regulation of the cytoskeleton, and the stabilization of the synaptic structure may be some of the roles the PSD plays. The ordered array of proteins into microdomains of the PSD governs the efficiency of the signal transduction pathways initiated by activation of the different kinds of glutamate receptors. Importantly, the PSD clusters the glutamate receptors at sites directly adjacent to the presynaptic release sites, which is crucial for fast chemical neurotransmission. Its connection to the receptors and cell adhesion molecules and to the cytoskeleton of the dendritic spine allows the PSD to coordinately respond to changes of presynaptic activity with alterations of the structure of the PSD and the morphology of the spine. Numerous links between the amount of PSD scaffolding proteins and the spine density and size as well as the glutamate receptor content of synapses have been experimentally established. An impaired architecture of the PSD would be expected to affect the balance of excitatory and inhibitory neurotransmission in the central nervous system and, therefore, to cause fatal diseases. However, relatively little is known about the connection between neurological disorders and the dysfunction of the PSD; maybe the high molecular redundancy of scaffolding proteins secures that central synapses will function even if one of the PSD-95 family member, GKAP, or Shank genes is mutated. Nevertheless, the Shank 3 gene is linked to a neurological disorder termed 22q13.3 deletion syndrome, characterized by mental retardation, delayed speech, and dysmorphic features. Mutations and chromosomal rearrangements in loci corresponding to PSD-95 and neuroligin genes are associated with autism. Mental retardation is linked to the genes for several additional PSD components and correlates with an abnormal morphology of dendritic spines and synapses. Therefore, studying the PSD may not only help to understand

102 Postsynaptic Density/Architecture at Excitatory Synapses

mechanisms of synaptic transmission, but may also provide insight into diseases of the central nervous system. See also: AMPA Receptor Cell Biology/Trafficking; LongTerm Potentiation (LTP): NMDA Receptor Role; Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology; NMDA Receptors, Cell Biology and Trafficking; Postsynaptic Development: Neuronal Molecular Scaffolds.

Further Reading Boeckers TM (2006) The postsynaptic density. Cell and Tissue Research 326: 409–422. Carlin RK, Grab DJ, Cohen RS, et al. (1980) Isolation and characterization of postsynaptic densities from various brain regions: Enrichment of different types of postsynaptic densities. Journal of Cell Biology 86: 831–845. Chen X, Vinade L, Leapman RD, et al. (2005) Mass of the postsynaptic density and enumeration of three key molecules. Proceedings of the National Academy of Sciences of the United States of America 102: 11551–11556. Cheng D, Hoogenraad CC, Rush J, et al. (2006) Relative and absolute quantification of postsynaptic density proteome isolated from rat forebrain and cerebellum. Molecular & Cellular Proteomics 5: 1158–1170. Collins MO, Husi H, Yu L, et al. (2006) Molecular characterization and comparison of the components and multiprotein complexes in the postsynaptic proteome. Journal of Neurochemistry 97 (supplement 1): 16–23. Funke L, Dakoji S, and Bredt DS (2005) Membrane-associated guanylate kinases regulate adhesion and plasticity at cell junctions. Annual Review of Biochemistry 74: 219–245. Gray EG (1959) Axo-somatic and axo-dendritic synapses of the cerebral cortex: An electron microscope study. Journal of Anatomy 93: 420–433. Gundelfinger ED, Boeckers TM, Baron MK, et al. (2006) A role for zinc in postsynaptic density assambly and plasticity? Trends in Biochemical Sciences 31: 366–373.

Harris KM, Jensen FE, and Tsao B (1992) Three-dimensional structure of dendritic spines and synapses in rat hippocampus (CA1) at postnatal day 15 and adult ages: Implications for the maturation of synaptic physiology and long-term potentiation. Journal of Neuroscience 12: 2685–2705. Husi H, Ward MA, Choudhary JS, et al. (2000) Proteomic analysis of NMDA receptor–adhesion protein signaling complexes. Nature Neuroscience 3: 661–669. Kennedy MB (2000) Signal-processing machines at the postsynaptic density. Science 290: 750–754. Kennedy MB, Beale HC, Carlisle HJ, et al. (2005) Integration of biochemical signalling in spines. Nature Reviews Neuroscience 6: 423–434. Kim E and Sheng M (2004) PDZ domain proteins of synapses. Nature Reviews Neuroscience 5: 771–781. Kornau HC, Seeburg PH, and Kennedy MB (1997) Interaction of ion channels and receptors with PDZ domain proteins. Current Opinion in Neurobiology 7: 368–373. Nicoll RA, Tomita S, and Bredt DS (2006) Auxiliary subunits assist AMPA-type glutamate receptors. Science 311: 1253–1256. Palay SL (1958) The morphology of synapses in the central nervous system. Experimental Cell Research 14: 275–293. Peng J, Kim MJ, Cheng D, et al. (2004) Semiquantitative proteomic analysis of rat forebrain postsynaptic density fractions by mass spectrometry. Journal of Biological Chemistry 279: 21003–21011. Scannevin RH and Huganir RL (2000) Postsynaptic organization and regulation of excitatory synapses. Nature Reviews Neuroscience 1: 133–141. Scheiffele P (2003) Cell–cell signaling during synapse formation in the CNS. Annual Review of Neuroscience 26: 485–508. Spacek J and Harris KM (1997) Three-dimensional organization of smooth endoplasmic reticulum in hippocampal CA1 dendrites and dendritic spines of the immature and mature rat. Journal of Neuroscience 17: 190–203. Sugiyama Y, Kawabata I, Sobue K, et al. (2005) Determination of absolute protein numbers in single synapses by a GFP-based calibration technique. Nature Methods 2: 677–684. Valtschanoff JG and Weinberg RJ (2001) Laminar organization of the NMDA receptor complex within the postsynaptic density. Journal of Neuroscience 21: 1211–1217. Ziff EB (1997) Enlightening the postsynaptic density. Neuron 19: 1163–1174.

Synaptic Transmission: Models S Raghavachari, Duke University Medical Center, Durham, NC, USA J Lisman, Brandeis University, Waltham, MA, USA ã 2009 Elsevier Ltd. All rights reserved.

Introduction Chemical signaling at a synapse occurs when a synaptic vesicle fuses with the presynaptic membrane in response to calcium influx through voltage-gated calcium channels. Vesicle fusion results in the formation of a fusion pore through which neurotransmitter packaged in the vesicle can escape into the synaptic cleft. The released neurotransmitter then diffuses through the synaptic cleft and binds to ligand-gated receptors on the postsynaptic membrane, which then lead to current flow across the postsynaptic membrane. To understand synaptic transmission, it is necessary to determine the spatiotemporal pattern of channel opening. However, physical limitations preclude direct observation of the dynamics of neurotransmitter release and receptor activation at single synapses. With the vast amount of biophysical and structural data that are now available, it is possible to make constrained models with sufficient biological realism that can be used to answer the following questions about the generation of synaptic responses: 1. How does the structure of the synapse shape the response? 2. How does the biophysics of receptor activation contribute to the response? 3. What is the influence of the kinetics of release of neurotransmitter and its reuptake or degradation? We first discuss modeling methods that can be used for the study of synapses and then discuss the application of these methods to the study of excitatory transmission at the frog neuromuscular junction and at central glutamatergic synapses.

Modeling Approaches The physical processes involved in chemical synaptic transmission are neurotransmitter diffusion and binding to receptors followed by conformational transitions of receptors, neurotransmitter binding to transporters for re-uptake (at central synapses), and neurotransmitter hydrolysis (in case of the neuromuscular junction).

The average time evolution of the neurotransmitter concentration in the cleft subject to these processes can be succinctly written as @Cðr; tÞ @Cðr; tÞ @Cðr; tÞ þ ¼ @t diffusion @t reaction @t @Cðr; tÞ þ ½1Š @t uptake

where C(r, t) is the concentration of the neurotransmitter at the spatial location r and time t. The diffusion of neurotransmitter is given by @Cðr; tÞ ¼ D2 Cðr; tÞ ½2Š @t diffusion

2

where r is the diffusion operator, and D is the diffusion constant (measured in mm2 s 1). These processes can be simulated on computers using two fundamentally different mathematical approaches. In the first, the molecular nature of the neurotransmitter and receptors is ignored. The neurotransmitter concentration and the probability of receptor activation are treated as continuous quantities that vary smoothly in space and time. The time evolution of these quantities is then modeled using analytic approximations or partial differential equations. A second approach respects the discrete, probabilistic nature of neurotransmitter diffusion and receptor interaction. The ligand and receptor molecules are represented individually and undergo random motions (corresponding to Brownian diffusion) and stochastic interactions and transitions (corresponding to the reactions). Continuum Methods

The main advantage of these continuum methods is their rapid implementation and the precise prediction of average quantities such as the time course of neurotransmitter clearance or receptor activation. Whereas early work using continuum models used simplified geometries, newer methods have allowed the incorporation of complex geometries, such as that of the neuromuscular junction. In continuum models, the equations for neurotransmitter diffusion, uptake, and binding are supplemented by equations for the concentrations of receptors in different states. This system of differential equations is subject to the appropriate initial conditions, specifying the concentration of the neurotransmitter and the receptor states at time t ¼ 0 at different spatial locations. The information on the synapse geometry is contained in the boundary conditions that govern the behavior of the neurotransmitter at reaction

103

104 Synaptic Transmission: Models

boundaries as well as the inert surfaces. The latter represent physical barriers that the diffusing neurotransmitter cannot cross. These are mathematically known as zero flux boundaries, where the spatial gradient of the concentration @C/@x ¼ 0. Uptake mechanisms or receptors located on the boundaries of the reaction space are instead modeled as absorbing boundaries where the concentration vanishes-that is, C(r, t) ¼ f(C)jsurface. In the absence of reaction and degradation/uptake terms in eqn [1], closed-form solutions of the diffusion equation are only possible for idealized geometries, which have a high degree of symmetry. Some example geometries include a cube, a cylinder, and a sphere. For example, a thin circular disk, with one face representing the presynaptic terminal and the other the postsynaptic membrane, is a good approximation of small central synapses. The solution of the diffusion equation in cylindrical coordinates is given in the form of Bessel functions, which can then, in principle, be used to study the dynamics of receptor opening. However, simple analytic models do not capture the complexity of the diffusion and reaction processes, and one must resort to numerical techniques for the solution of the full equation, including reaction terms and neurotransmitter uptake/degradation. A number of techniques have been developed for the solution of these equations under various approximations. Those relevant for synaptic transmission are discussed next. Finite difference methods Finite difference (FD) methods are a powerful set of techniques to solve nonlinear reaction-diffusion equations (e.g., eqn [1]). These methods represent the concentration C(x) by its values at discrete points in the reaction volume and approximate the derivatives (in space and time) as differences between adjacent points. This discretization results in a set of algebraic equations for the concentration of the diffusing neurotransmitter that can be solved using standard methods of linear algebra which have been implemented in software packages such as MATLAB. This approach is particularly useful for studying the synaptic current due to receptors with complex multistate kinetics as well as the effects of antagonists and modulators of receptors. Due to the long history of FD methods and their wide application, many optimization techniques are available for the simultaneous solution of the neurotransmitter diffusion, uptake, and receptor binding and even complex receptor kinetics. However, these methods are best suited for simplified geometries, but even here, the amount of discretization places severe restrictions on the accuracy of the method as well as its implementation on readily available personal computers.

Finite element methods Finite element (FE) methods are widely used in engineering applications to study how material structures behave under applied loads. These methods have been extended to a wide variety of partial-differential equations, including nonlinear reaction-diffusion equations (e.g., eqn [1]). In a strict mathematical sense, FE methods find a piecewise approximation to the solution of eqn [1]. This is done by subdividing the arbitrarily complex reaction space into small (finite) elements and decomposing the solution as a linear weighted sum of functions that are defined only over the finite elements. This decomposition then results in a nonlinear matrix equation that must be solved in order to determine the coefficients. In three-dimensional space, these elements are usually trapezoidal (other more complex shapes are possible). The accuracy of the method, as for the FD method, depends on the discretization (with finer subdivisions yielding better accuracy.) Similar to FD methods, the discretization can be adaptive, with high spatial resolution where the concentration of neurotransmitter varies rapidly. The diffusion of acetylcholine (ACh) in complex geometries can be implemented using publicly available FE packages to gain insight into the structurefunction relationship at neuromuscular junctions (NMJs). Similarly, the effect of vesicular release and acetylcholinesterase (AChE) distribution on miniature excitatory postsynaptic current (mEPSC) kinetics has also been studied using such models. The key advantage of the FE method is that the computational cost is proportional to the number of finite elements. Thus, a full endplate current (representing the release of approximately 200–300 quanta in a geometrically complex NMJ) can be simulated. However, the implementation of the reaction terms in FE models is nontrivial. The esterase and the receptors must be represented as spatially delimited sinks resulting in the boundary condition ^ðxÞD
Cðx; tÞ ¼ kCðx; tÞ n

½3Š

^(x) is the surface normal, and k is a constant where n that is related to the steady-state constant of ACh to the esterase or receptor. The kinetics of the ACh receptor are relatively simple (the only isomerization step involves the opening of the channel), which makes its implementation possible within the FE formalism. More complex reaction schemes (e.g., glutamate receptors) are much more difficult to represent using this method. This method has been used in a few cases to study the effect of differences in fold geometry on the kinetics of the mEPSC. Ultrastructural observations show that fast-twitch and slowtwitch muscles have different structures. Muscular

Synaptic Transmission: Models

dystrophy is also known to distort NMJ structure, leading to differences in mEPSC kinetics.

Continuum methods such as FD and FE outlined previously predict the average properties of chemical reactions. Such models are accurate (within the limits imposed by stability and accuracy considerations) as long as the number of reactions modeled is large. However, the chemical reactions that underlie neurotransmission at a single central synapse present a challenge to such models. Experiments show that the synaptic current at small central synapses is due to the opening of tens of receptors. This implies that each vesicle only contains a few thousand molecules of neurotransmitter, of which several tens participate in receptor activation. Finally, the amplitude of the individual synaptic response is often variable from trial to trial. These considerations imply that the stochastic or ‘noisy’ nature of chemical reactions must be accounted for in modeling synaptic transmission. The main source of this noise is the ubiquitous ‘thermal noise’ that characterizes the diffusive (Brownian) motion of molecules and conformational transitions of receptor proteins. The Monte Carlo method assumes that the physical processes in synaptic transmission are inherently probabilistic. The dynamics of the individual reactions are then followed by rolling a dice at each time step (hence the name Monte Carlo) to decide how much a given ligand molecule will move and whether a particular reaction will occur or not. Thus, variability is an inherent part of the simulation procedure. The average behavior of the system is obtained by averaging the results of the simulations across multiple trial runs. One strength of the Monte Carlo method is the high level of spatial and temporal detail that can be achieved. The heart of the Monte Carlo method is its approach to modeling the diffusion of neurotransmitter in complex reaction spaces. In general, a diffusing molecule at a point P at time t ¼ 0 has some thermal velocity and undergoes collisions with water molecules on the sub-picosecond timescale at room temperature. After many collisions (over a nanosecond timescale), the molecule may have moved to a new location in a random direction. The distribution of displacements from the original location can be calculated from the diffusion equation (eqn [1]). For diffusion in free space, the solution of this equation is given by Cðr;
tÞ ¼

Thus, the fraction of molecules in a spherical shell (of volume 4pr2dr) is

C0 ð4D
tÞ

2

3=2

er

=4D
t

½4Š

1

r2 =4D
t

ð4r2 drÞ ½5Š ð4D
tÞ which is the same as the probability of a radial displacement of a single molecule. Thus, diffusion of a single neurotransmitter molecule is simulated by choosing a random number with the previous probability distribution. A second random number specifies the direction of this displacement in three dimensions. When the diffusing molecule encounters an impenetrable barrier, it is reflected back into the reaction volume by an elastic collision. In this way, the zero flux boundary conditions specified previously are naturally handled. Note that collisions need not occur with nonreactive boundaries but also with reactive boundaries (representing a surface receptor). In this case, the reflection occurs with a probability that is related to the macroscopic on-rate constant of the ligand with its receptor. This probability is given by   1 p
t 1=2 ½6Š pb ¼ kon 2Nav Atile D fr ¼

Monte Carlo Approaches

105

3=2

e

where kon is the macroscopic on-rate constant for the ligand to bind to the receptor, Nav is Avogadro’s number, and Atile is the area of the surface element occupied by the receptor. Unimolecular reactions, such as conformational changes of receptors, also occur in a probabilistic manner, with the probability of any given transition at a time step given as " ! # X ki
t ½7Š pu ¼ 1 exp i

where ki is the macroscopic rate constants of the ith possible reaction for a receptor at any given state (e.g., a ligand can unbind from a receptor or the receptor might isomerize to a new state). The choice of the time step is dictated by the fact that the reaction probabilities must be 1 is the tortuosity of the medium, a measure of the hindrance (geometric or otherwise) experienced by the diffusing molecules. Measurements suggest that glutamate diffusion in the cleft is similar to its value in solution (decreased by approximately a factor of 2). In the hippocampus, electron microscopy studies show that most functional synapses seem to have glial processes at the perimeter of the axon-spine interface. At most CA1 synapses, the astrocytic processes only partially surround the synapse, allowing glutamate to freely escape into the extracellular space. The astrocytes make up nearly 10% of the total membrane in CA1 and are lined with glutamate transporters (proteins that move glutamate from the extracellular space into the glial cell) at a density of approximately 10 000 mm2. These transporters also bind glutamate with high affinity (10–20 mM) and serve to either buffer or remove synaptically released glutamate. From a modeling standpoint, it is not correct to incorporate the effect of transporters by slowing the diffusion of glutamate in the extracellular space. This is because the buffering action of the transporters means that glutamate actually binds to them, resulting in a loss of glutamate from the extracellular space. Instead, the effects of transporters should be included

by explicit modeling of the binding and unbinding of glutamate rather than a reduction of its diffusion constant. Kinetic Constraints

AMPA receptors act on a rapid timescale and display weak affinity to glutamate (the dissociation constant, KD  400 mM). In recent years, these channels have been cloned and expressed in heterologous systems in order to study their electrophysiological properties under controlled conditions, leading to a deep understanding of the structure and kinetic properties of AMPA channels. These channels are tetramers composed of several stoichiometric combinations of glutamate receptor subunits (GluR) 1–4. Thus, channels may be homomers composed of identical subunits (GluR14) or heteromers composed of distinct subunits in a strict stoichiometric manner (GluR12– GluR22, GluR22–GluR32, etc.). Depending on the subunit composition, AMPA channels may have distinct kinetic properties that shape channel kinetics and even permeability to different cations (GluR14 are calcium permeable). Each subunit has a binding domain which can bind glutamate independent of whether glutamate is bound to any of the other subunits (i.e., there appears to be no cooperativity in glutamate binding). Structural studies indicate that agonist binding causes a closure of the binding domain and initiates conformational changes that serve to activate the channel. It appears that at least two subunits need to have glutamate bound in order for the channel to conduct ions. Importantly, additional binding greatly speeds channel opening and enhances the single channel conductance (typical values of conductance are 5–8, 7–10, and 12–15 ps, depending on subunit occupancy). However, it is not entirely clear whether these conductance steps occur for channels natively expressed in synapses. Furthermore, it appears that glutamate binding and unbinding can occur from both closed and open states, but binding and unbinding from the open states is slow. A hallmark of glutamate channels is their ability to rapidly desensitize–that is, switch to a nonconducting conformation upon binding glutamate. The exact conformational changes that lead to desensitization are a subject of active research, but it is widely accepted that glutamate binding to the subunit can set in motion conformational changes leading to either activation or desensitization. Additionally, desensitization as well as recovery from desensitization show a concentration dependence that can be best explained by considering that each subunit is capable of desensitization upon binding glutamate (i.e., desensitization is not a cooperative phenomenon). These

Synaptic Transmission: Models

considerations lead to models of AMPA receptor gating that may typically have 40–50 distinct states. Once the state diagram for the receptors is chosen, the transition rates between these states can be determined by fitting the data on channel activation and desensitization produced by the controlled application of glutamate to patches excised from neurons. Many of these rate constants are fixed to be multiples of each other due to the assumption of subunit independence. Further reduction is possible due to the presence of loops in the state diagram (different paths through a loop should be equivalent according to the principle of microscopic reversibility). The model that best fits data from controlled glutamate application to AMPA channels from CA1 pyramidal cells suggests that at low concentrations, desensitization competes effectively with channel opening, whereas at higher concentrations the cooperative process of channel opening makes the opening rate much higher than the desensitization rate. Insights from Modeling

This model of receptor kinetics was used in a geometrically constrained model of a synapse to determine the influence of various parameters on synaptic transmission. Time course and amplitude of the mEPSC At large mushroom spines in CA1, the number of neurotransmitter molecules vastly exceeds the number of receptors, and neurotransmitter diffusion and clearance is rapid. Thus, the loss of neurotransmitter due to receptor binding can be largely neglected. Moreover, the uptake mechanisms at central synapses are located outside the cleft. With these simplifications, eqn [1] can be solved analytically for a cylindrical cleft where glutamate is released at the center of the top face of the disk with absorbing boundary conditions at the cleft edges. Modeling the escape of neurotransmitter from the vesicle through a narrow fusion pore leads to a brief ‘concentration spike’ during which a large amount of neurotransmitter is confined to a small volume of the cleft near the release site. This spike extends approximately 100 nm from the release site and is dissipated within 100 ms (i.e., the concentration declines by 1/e of its peak value). During this spike, the glutamate concentration builds up to nearly 3 mM near the fusion pore. The rapid diffusion of glutamate dilutes this spike. The efficacy of the spike in producing receptor occupancy is made clear by a calculation analogous to the neuromuscular junction. If we take the area of the spine to be the upper limit in the expression for tdiff (0.1 mm2 for larger spines), we obtain a diffusion time of approximately 10 ms, which is much shorter than

109

that in the NMJ. Assuming typical values of forward binding rate of 2  107 M s1, tbind is 80 ms. This implies that the rise time at central synapses would almost entirely be determined by the binding of glutamate to the receptors as well as the receptor kinetics. Solving the equations for receptor activation using the concentration of glutamate calculated previously results in a simulated mEPSC with an amplitude of approximately 10–15 pA and a 20–80% rise time of 200 kDa) when separated by apparent molecular mass in polyacrylamide gels. After heparitinase treatment the core protein is much smaller in size. HSPGs are found in all basal laminae and also associated with the cell surfaces. The three

Neuromuscular Junction (NMJ): Postsynaptic Basal Lamina 177

major HSPGs found in skeletal muscle extracellular matrix are perlecan, agrin, and collagen XVIII. Perlecan Perlecan has a widespread distribution. It is found in all basal laminae, and plays an important role in the storage and release of growth factors. There are three glycosaminoglycan (GAG) side chains extending form the N-terminal region. Collagen XVIII Collagen XVIII has been recently shown to be a HSPG and to be present in the basal lamina of skeletal muscle. Loss of collagen XVIII in Caenorhabditis elegans has been shown to disrupt neuromuscular junction formation. There are no reports showing that collagen XVIII is concentrated at the vertebrate neuromuscular junction, so it remains unclear whether it has a function there. Agrin Agrin is the most extensively studied HSPG at the neuromuscular junction, and it plays a pivotal role in the structure and function of the neuromuscular junction. Early experiments performed by UJ McMahan and colleagues were instrumental in revealing the fact that cell-signaling molecules were found in the extracellular matrix. When muscle fibers were damaged, in a way that left the extracellular matrix intact, the damaged portions of the muscle cells would be removed. In addition, the nerve terminal could be removed, leaving behind an empty basal lamina sheath. When the muscle cell regenerated by the proliferation and fusion of skeletal muscle satellite cells, the skeletal muscle would reform within the basal lamina sheath. The regenerated muscle cells would aggregate AChRs on their surfaces precisely at the spot that they contacted the previous synaptic site on the basal lamina. These experiments clearly showed that molecules stably bound to the synaptic basal lamina could direct the aggregation of AChRs on the surface of the muscle cell. Subsequent experiments revealed that the protein agrin could direct the aggregation of AChRs in culture. Mice that lack agrin do not develop neuromuscular junctions, and these mice die at birth. Agrin is a multifunctional component of the synaptic basal lamina. The C-terminal region of agrin has been most extensively studied, since this is the region that is sufficient to induce the aggregation of AChRs. The number of other molecules that interact with agrin at the neuromuscular junction is extensive. The N-terminal region of agrin has been shown to bind to laminin. Agrin will also bind to a-dystroglycan and thus helps to link the dystroglycan complex to the extracellular matrix. Agrin will also bind to integrins, and to neural cell adhesion molecules (N-CAMs) on the surface of the motor neuron and muscle cell.

In addition to these interactions, agrin will also bind to nidogen and collagen. Agrin clearly plays an important role in the structure and function of the neuromuscular junction, and this role is both structural (linking other proteins in the intracellular matrix together) and functional (acting as a cell-signaling molecule). AChE

The main form of acetylcholinesterase (AChE) at the neuromuscular junction and the synaptic basal lamina is the collagen-tailed form of AChE, ColQAChE. ColQ-AChE is bound to the heparan sulfate proteoglycan perlecan and will also interact with the muscle-specific tyrosine kinase receptor. Its main function is to remove ACh from the synaptic cleft. In fact, the ACh released from the nerve terminal must first run the gauntlet of AChE that is bound to the synaptic basal lamina, before it can bind to the AChR concentrated on the crests of the junctional folds.

Cell Surface and Membrane Receptors Dystroglycan Complex

The main components of the dystroglycan complex are a- and b-dystroglycan. These two forms are from a single gene, but have very different sizes and properties. The a-dystroglycan is 156 kDa in size; it is an extracellular protein that binds noncovalently to b-dystroglycan in the dystroglycan complex, on the one hand, and to laminin and/or agrin in the extracellular matrix, on the other. b-Dystroglycan is a 43 kDa glycoprotein and is an integral membrane protein. Numerous other proteins associate with the dystroglycans and together they form a large structure called the dystroglycan complex. The other members of the dystroglycan complex are the sarcoglycans (a-, b-, d-, and g-sarcoglycans), a-dystrobrevin, neuronal NO synthase (nNOS), and the syntrophins (a, b1, and b2). In addition, and of critical importance, the dystroglycan complex will bind to dystrophin or utrophin. On the intracellular side, the dystroglycan complex will bind to the proteins dystrophin or utrophin. In turn, dystrophin and utrophin will bind to the actin cytoskeleton. The protein utrophin is shorter than dystrophin, and utrophin is concentrated at the neuromuscular junction. The dystroglycan complex is found throughout the entire length of the skeletal muscle cell and is particularly concentrated at the neuromuscular junction and the myotendinous junction. The dystroglycan complex is greatly reduced in Duchenne muscular dystrophy patients, who suffer from muscle fiber breakdown and degeneration. Thus, the dystroglycan complex plays an important

178 Neuromuscular Junction (NMJ): Postsynaptic Basal Lamina

role in linking the intracellular actin cytoskeleton to the extracellular matrix. This link is undoubtedly important in the structural integrity of the muscle cell membrane and is probably critical to protect the membrane from tears that would result from the contractions of the skeletal muscle cell. At the neuromuscular junction the dystroglycan complex may also play an important role in cell-to-cell signaling. The binding of agrin to dystroglycan may be an important feature of the processes that regulate the induction and maintenance of the postsynaptic apparatus in muscle. The exact role of the agrin/dystroglycan complex interaction in postsynaptic signaling is still unknown. MuSK

The aggregation of AChRs induced by agrin depends absolutely on the presence of a muscle-specific tyrosine kinase (MuSK). MuSK is clearly a downstream mediator of agrin, and MuSK knockout mice have virtually the same phenotype as those lacking agrin. However, the direct binding of agrin to MuSK has not been demonstrated, and the exact choreography of the events that start with agrin and end with the activation of MuSK has not been determined. MuSK has a large extracellular domain and an intracellular kinase domain. More recently it has been found that MuSK is required for the anchoring of the ColQAChE complex to the synaptic site. It is also been shown that the Abl kinases (Abl1 and Abl2) are downstream of activated MuSK. Thus, MuSK plays a central role in the events leading to the proper formation of the neuromuscular junction. Integrins

Integrins are transmembrane protein complexes that form from heterodimers of a and b chains. Both chains are important for ligand binding. There are 18 a subunits and eight b subunits known, and currently 24 unique combinations of these have been identified. In addition, many of the subunits are alternatively spliced so that there are multiple variants of many subunits. Of particular interest is the a7 subunit, since it is found concentrated at the neuromuscular junction. The a7 subunit interacts with the b1 to form the a7b1 integrin. This integrin binds to laminin via the RGD (arginine-glycine-aspartate) domain on the b2 arm of the synaptic laminin. The a7 subunit is alternatively spliced, and there are at least six variants of the a7 integrin. While the binding of the a7b1 integrin to laminin is a structural link between the matrix and the cell, the integrins also participate in intracellular signaling. On the intracellular side the integrins play a key role in linking to the actin cytoskeleton. This linkage is

very similar to the molecular linkage found in a focal adhesion. Namely, the integrin binds to vinculin, a-actinin, and talin. All of these components are found aggregated at the postsynaptic site. In addition, integrins play an important role in regulating the activities of kinases. Of particular interest are the Src kinase family members Abl and Fyn, which have been shown to play a role in the assembly and maintenance of the postsynaptic apparatus. Thus, the integrins play an important structural and signaling role in the postsynaptic apparatus. Integrins also are likely to play a presynaptic role, since treatment of frog muscles with RGD-containing peptides will reduce the increase in miniature endplate potential release upon muscle stretch. In addition to the a7b1 integrins, the a3b1 integrins are found at the presynaptic active zones. The a3b1 integrins also interact with laminins in the synaptic basal lamina and can activate intracellular kinases. The precise role of the a3b1 integrin is still largely unknown, but it is clear, at least in frogs, that the binding of integrin to matrix also has an important presynaptic cell-signaling function. Cadherins

Cadherins are membrane-bound proteins that require calcium for their activity. They play an important role in the early development of the skeletal muscle cell. There are many cadherins, but the main ones produced in adult skeletal muscle are neural (N)-cadherin and muscle (M)-cadherin. Immunolabeling studies have revealed that both N-cadherin and M-cadherin are concentrated at the neuromuscular junction. The role of cadherins at the neuromuscular junction has been largely uninvestigated. N-Cadherin promotes neurite outgrowth in vitro and it is of interest that this effect is blocked by agrin. Both N-cadherin and M-cadherin play an important role in the formation of multinucleated muscle cells. They also may play a role at the synapse. The cadherins will form calcium-dependant bonds with cadherins on adjacent cells. Cadherins also interact on the intracellular surface with catenins. The a-, b-, and d-catenins have been detected at the neuromuscular junction. In other systems it is known that the a- and b-catenins form a link with the actin cytoskeleton. In addition, the liberation of b-catenin from cadherins can result in nucleocytoplasmic shuttling to alter gene transcription via the interaction of b-catenin with T cell factor (TCF). Similarly, d-catenin has also been shown to be present at the neuromuscular junction and interacts with the promoter region of rapsyn. Altogether, these results argue that the cadherin system functions at the neuromuscular junction to regulate cell structure and function.

Neuromuscular Junction (NMJ): Postsynaptic Basal Lamina 179

Proteases at the Neuromuscular Junction Matrix Metalloproteinases

The matrix metalloproteinases (MMPs) are a large family of enzymes whose main function is to degrade the extracellular matrix. There are over 25 metalloproteinases, each with preferred matrix substrates. Most MMPs are released from cells in a pro form and must be first activated by other proteases. While it is clear that the MMPs play a role in matrix development and remodeling, it is clear that they also have an influence on cell-to-cell signaling. For example, cleavage of the NC1 domain from type IV collagens is known to play a major role in the signaling of vascular development and remodeling. The strongest evidence for the involvement of MMPs at the neuromuscular junction is the fact that mice that lack MMP3 have increased junctional folds and have AChR receptors on their surface. This likely results from an accumulation of matrix molecules that would normally be removed by MMP3, particularly agrin. MMP2 and MMP9 have been shown to play an important role in the reinnervation of muscle following nerve damage. The fact that MMPs are present at the neuromuscular junction is not a surprise, since it is clear that the matrix must be remodeled to allow synaptic growth. The mechanisms that control MMP activation at the neuromuscular junction are still unknown. Tissue Inhibitors of Metalloproteinases

Balancing the activity of the MMPs are the tissue inhibitors of metalloproteinases (TIMPs). There are four TIMPs, and one of these, TIMP2, has been shown to be present at the neuromuscular junction. Mice that lack TIMP2 have altered neuromuscular junctions. TIMPs inhibit MMPs, but are also required for the activation of some MMPs. Thus, there is a delicate balance between the activity of MMPs and TIMPs. The control of matrix structure and function is likely to be central to the mechanisms that control synaptic structure and function.

Conclusion The synaptic basal lamina contains molecules that are common to all basal laminae throughout the body, and also a number of unique molecules. The synaptic basal lamina has both a structural role that links

the cytoskeletons of the synaptic components, and a cell-to-cell signaling role. The organization of the synaptic basal lamina is extremely precise, and the maintenance of this structure is undoubtedly important in the organization of the synapse. The synaptic basal lamina is a dynamic structure, and proteases are constantly sculpting this complex matrix. The control of matrix structure and function is likely to be central to the mechanisms that control synaptic structure and function. See also: Neuromuscular Connections: Vertebrate Patterns of; Neuromuscular Junction (NMJ): Postsynaptic Events in Neuromuscular Transmission.

Further Reading Barresi R and Campbell KP (2006) Dystroglycan: From biosynthesis to pathogenesis of human disease. Journal of Cell Science 119(Pt. 2): 199–207. Bixby JL, Baerwald-De la Torre K, Wang C, et al. (2002) A neuronal inhibitory domain in the N-terminal half of agrin. Journal of Neurobiology 50(2): 164–179. Burkin DJ and Kaufman SJ (1999) The alpha7beta1 integrin in muscle development and disease. Cell Tissue Research 296(1): 183–190. Fox MA and Umemori H (2006) Seeking long-term relationship: Axon and target communicate to organize synaptic differentiation. Journal of Neurochemistry 97(5): 1215–1231. Kummer TT, Misgeld T, and Sanes JR (2006) Assembly of the postsynaptic membrane at the neuromuscular junction: Paradigm lost. Current Opinion in Neurobiology 16(1): 74–82. Miner JH and Yurchenco PD (2004) Laminin functions in tissue morphogenesis. Annual Review in Cell and Developmental Biology 20: 255–284. Ortega N and Werb Z (2002) New functional roles for non-collagenous domains of basement membrane collagens. Journal of Cell Science 115(22): 4201–4214. Patton BL (2003) Basal lamina and the organization of neuromuscular synapses. Journal of Neurocytology 32(5–8): 883– 903. Rotundo RL, Rossi SG, Kimbell LM, et al. (2005) Targeting acetylcholinesterase to the neuromuscular synapse. Chemico– Biological Interactions 157–158: 15–21. Strochlic L, Cartaud A, and Cartaud J (2005) The synaptic musclespecific kinase (MuSK) complex: New partners, new functions. BioEssays 27(11): 1129–1135. Werle MJ and VanSaun M (2003) Activity dependent removal of agrin from synaptic basal lamina by matrix metalloproteinase 3. Journal of Neurocytology 32(5–8): 905–913. Yurchenco PD, Amenta PS, and Patton BL (2004) Basement membrane assembly, stability and activities observed through a developmental lens. Matrix Biology 22(7): 521–538.

Neuromuscular Junction (NMJ): Postsynaptic Events in Neuromuscular Transmission C R Slater, University of Newcastle upon Tyne, Newcastle upon Tyne, UK ã 2009 Elsevier Ltd. All rights reserved.

Essentials of Neuromuscular Transmission in Vertebrates The neuromuscular junction (NMJ) is the interface between the central nervous system and the muscles it controls. The contraction of muscle fibers is activated by depolarization of the muscle fiber membrane. When motor neuron firing is triggered, action potentials (APs) are initiated near the nerve cell body and then travel along the motor axon to the NMJ. There, the AP causes the release of neurotransmitter molecules that act locally to depolarize the postsynaptic muscle fiber. The ultimate effect of the transmitter is to make the muscle contract. In most vertebrate muscle fibers, an essential intermediate step in the process of neuromuscular transmission is the generation of an AP in the muscle fiber. This AP travels rapidly along the muscle fiber, activating contraction of the whole fiber at nearly the same time. This article describes how vertebrate muscles respond to transmitter released from the nerve and how the APs that activate muscle contraction are generated. In the invertebrates, most muscles do not generate APs and the details of muscle activation are therefore somewhat different. This is also true of a distinct class of electrically inexcitable muscle fibers in vertebrates, discussed briefly at the end of this article. Most vertebrate twitch muscle fibers are innervated at a single site, often roughly midway along the length of the fiber, by a single motor axon. Following a nerve AP, transmitter released from the nerve causes a local reduction in membrane potential, or depolarization, of the muscle fiber. This brief event (10 ms) is referred to as the endplate potential (EPP; Figure 1), since the NMJ is sometimes called the motor endplate. The EPP is a local event that has its maximum amplitude at the NMJ and declines to a low level within 1–2 mm along the fiber (see later). Since muscle fibers may be more than 10 cm long, some other event must account for the depolarization that activates contraction. This event is the AP (Figure 1(a)), which can propagate rapidly in an all-or-none fashion from the NMJ to the ends of the muscle fiber. APs are generally triggered when the membrane potential of the muscle fiber becomes less negative than about 60 to 55 mV. Thus, from a resting potential of 90 to 70 mV, the membrane must be

180

depolarized by at least 25 mV. In most vertebrates, the EPP is substantially larger than required to reach the AP threshold. This results in very reliable and complete activation of the contractile apparatus, leading to an ‘all-or-none’ twitch contraction, in each muscle fiber, by every motor nerve impulse. The postsynaptic events of neuromuscular transmission occur within the highly specialized region of the muscle fiber known as the postsynaptic apparatus. This region contains very high concentrations of the ion channels that mediate the response of the muscle fiber to the transmitter released from the nerve. In many vertebrates, there are also striking structural features of this region, including a highly folded surface membrane and an accumulation of specialized myonuclei with distinctive patterns of gene expression.

AChRs and the Generation of the Endplate Current The neurotransmitter acetylcholine (ACh) is released from the vertebrate motor axon terminal in multimolecular quanta. The ACh causes a depolarization by binding to, and thus opening, cation-selective ion channels in the muscle fiber membrane known as ACh receptors (AChRs; Figures 2(a) and 2(b)). When these channels open, the internal negativity of the muscle fiber drives the net entry of positive ions into the cell. It is this effect that gives rise to the depolarizing phase of the EPP. The AChRs present at the vertebrate NMJ are pentameric complexes of four types of subunits (Figure 2(a)). While the structures of these subunits are broadly similar there are also important differences and each subunit is encoded by a different gene. In the muscles of some lower vertebrates, and in immature or paralyzed mammalian muscles, each AChR molecule has two a-subunits and one each of b-, d-, and g-subunits. In many vertebrates, including mammals, the g-subunit is largely replaced by an e-subunit. Each subunit has a molecular mass of about 50 kDa, so the entire complex has a molecular mass of about 250 kDa. Every AChR molecule has two ACh binding sites, each of which includes part of one a-subunit. When no ACh is bound, the AChR channel is closed and no current flows through it. To open the channel, two ACh molecules must bind, one to each binding site. When this happens the channel opens rapidly (Figure 2 (b)). The open channel has a conductance of about 30–50 pS. In the ionic conditions found in an intact mammal, the entry of positive charge through this

Neuromuscular Junction (NMJ): Postsynaptic Events in Neuromuscular Transmission 181 15

Membrane potential (mV)

0 AP −15

−30 EPP

−45 −60

−75 a

3 ms EPC

mEPC 0

0

−100

−5 nA

nA −200

−10

2

mEPP

EPP −40

1 mV

−60

0 b

mV 0

50

100 ms

−80

0

10

20 ms

Figure 1 Electrical events associated with transmitter action at the vertebrate NMJ. (a) Intracellular recordings of endplate potentials (EPP) and action potentials (AP). When the EPP fails to reach the AP threshold (horizontal dashed line) no AP is generated. The smaller EPPs illustrated were recorded during partial block of AChRs by curare. The larger EPPs illustrated were recorded during block of NaV1 channels in the muscle fiber by a cone snail toxin. These EPPs would normally trigger an AP. (b) Miniature endplate currents (mEPC) result from the flow of current into the muscle fiber induced by a single quantum of ACh. The nerve impulse causes the release and action of many quanta, resulting in the much larger endplate current (EPC). The electrical properties of the muscle fiber convert these currents into equivalent changes in membrane potential, the miniature endplate potential (mEPP) and the larger endplate potential (EPP). Note that the potential changes are slower than the currents that give rise to them, a result of the membrane capacitance.

conductance represents an ‘inward’ current of about 2.5–4 pA (Figure 2(c)). After being opened, each channel closes spontaneously and abruptly, on average after about 1 ms (for AChRs containing an e-subunit). However, the duration of open times, even of the same channel on different occasions, varies considerably and randomly, with brief openings being the most common and long open times the least common. At the vertebrate NMJ, each ACh quantum consists of 5000–10 000 molecules. Each quantum causes the opening of 1000–2000 ACh-gated channels, resulting in a brief influx of a current of about 3–5 nA into the

muscle fiber. The current induced by a single quantum is known as a miniature endplate current (mEPC; Figure 1(b)). Although all the channels activated by a single quantum of ACh open almost simultaneously, the variability of the open times means that some close very quickly and others remain open for several milliseconds. As a result, the shape of the mEPC differs from that of the individual channels, having an abrupt rise to a maximum value and a slower exponential decay with a time constant of 1–1.5 ms (Figures 1(b) and 2(d)). In the case of AChRs containing a g-subunit (see earlier), the time constant is severalfold longer.

182 Neuromuscular Junction (NMJ): Postsynaptic Events in Neuromuscular Transmission b

a

e

d

a Out Membrane In

a

Open

Closed

b

Closed 2 pA

Open

20 ms

c 1 2 3 4 5 6

Open Closed

NMJ. The AChR channels opened by a single quantum do so within less than 100 ms. During this time ACh molecules, on average, can diffuse about 0.3 mm. Each quantum thus acts on about 0.25 mm2 of postsynaptic membrane. For most vertebrate NMJs, this is less than 0.1% of the membrane containing AChRs. As a result, the sites of action of individual quanta on the muscle fiber membrane are effectively separated from each other, allowing the effects of the individual quanta to sum to produce the much larger endplate current (EPC; Figure 1(b)). In different vertebrate species, the EPC may vary in amplitude from 200 to 2000 nA. Since the quanta released by a single nerve AP act almost simultaneously on the muscle, the time course of the EPC is usually very similar to that of the mEPCs.

From EPCs to EPPs The passive electrical properties of the muscle fiber determine how the current of the EPC is converted into the depolarization that is the EPP. A brief account of these properties is given in the following sections. Passive Cable Properties of Muscle Fibers

4 6 d 2

3

5

1 3 pA 2 ms

Figure 2 Acetylcholine receptors (AChRs) mediate permeability changes in the postsynaptic membrane. (a) The five subunits of the AChR associate to form a pore that spans the plasma membrane. (b) A model of the open and closed states of the AChR as viewed looking along the pore axis. Small relative movements of the subunits, induced by ACh binding, result in changes in the diameter of the ion pore that have a big impact on permeability. (c) Record of current flowing through a single AChR channel. Each event has an abrupt start and finish. Although the duration of the open states varies, the intensity of the current in different events is very similar. (d) Scheme of how single channel openings sum to give rise to a mEPC. In the upper part, each of six channels is shown to open simultaneously. Each channel stays open for a time that varies randomly about a mean value. Thus the six channels close at different times. In the lower part is shown a model mEPC consisting of the sum of the currents flowing through the six channels. It decays exponentially from the initial peak with a time constant equal to the duration of the mean open time of the individual channels. (b) Diagram provided by Nigel Unwin. (c) Reprinted by permission from Macmillan Publishers Ltd: [Nature] Mishina M, Takai T, Imoto K, et al. (1986) Molecular distinction between fetal and adult forms of muscle acetylcholine receptor. Nature 321: 406–411, copyright (1986).

At the NMJ of vertebrate twitch muscles, between 20 and 200 quanta are typically released by a single nerve AP. In general, the sites of release of these quanta are distributed randomly over the entire

The flow of current in muscle fibers is governed by the so-called cable properties of the cell. These derive from the relatively high electrical conductivity of the cytoplasm and the low conductivity (or high resistivity) of the plasma membrane. Current entering the cell at a point, such as the NMJ, may flow longitudinally in the cytoplasm or transversely through the membrane (Figure 3(a)). Only the latter pathway leads to a change in the membrane potential. The fraction of the current that takes each route depends on the relative resistance each pathway presents. In practice, the conductivity of the cytoplasm is roughly the same in most cells and the resistivity of the membrane is similar in many (but not all) skeletal muscle fibers. As a result, the dominant factor determining the distribution of current flow in different muscle fibers is their diameter. When the diameter is large, current can flow much more easily along the muscle fiber than across the membrane, and there is relatively little voltage change across the membrane at the site of current entry. In contrast, if the diameter is small, the resistance to longitudinal current flow is high and more current crosses the membrane close to the site of entry, giving rise to a larger local change in voltage. The peak amplitude of the potential change caused by a given intensity of current flow entering the cell at a single point is determined by the so-called input resistance of the muscle fiber. Typical vertebrate skeletal muscle fibers have ‘input resistances’ of

Neuromuscular Junction (NMJ): Postsynaptic Events in Neuromuscular Transmission 183 V

i

External medium

Cytoplasm

a

120 100 %Vo

80 60 40 20

Membrane potential (mV) c

−1.5

−1

−0.5 0 0.5 Distance (mm) Membrane potential (mV)

0 −2

b 30 15 0 −15 −30 −45 −60 −75

20 nA 3 ms

1

1.5

2

30 15 0 −15 −30 −45 −60 −75

d

Figure 3 The passive electrical properties of the muscle fiber shape the voltage change caused by current flowing across the membrane. (a) Diagram of a typical experiment. Two electrodes are placed within the muscle fiber, one to pass current into the cell (i ) and the other to measure the resulting change in transmembrane potential (V). Arrows show flow of current. The current across the membrane is greatest near the site of current injection and decreases with distance from that site. (b) Decay of transmembrane potential (Vm) with distance from the site of injection. The space constant (l) is the distance at which the potential has fallen to 36% (1/e) of its maximum (Vo). (c) Changes in membrane potential (upper part) in response to rectangular current pulses. Note the slower change in voltage due to the capacitance of the membrane. Horizontal dashed line shows AP threshold: depolarizations greater than this trigger an AP. (d) Potential changes in response to injection of current with a time course similar to that of the postsynaptic mEPC and EPC that occur during neuromuscular transmission. Note that the passive properties of the membrane convert these currents into potential changes very similar to mEPP and EPP. The action potential threshold is very similar to that observed with rectangular pulses. (c) Reproduced from Wood SJ and Slater CR (1995) Action potential generation in rat slow- and fast-twitch muscles. Journal of Physiology 486: 401–410, with permission from Blackwell Publishing. (d) Reproduced from Wood SJ and Slater CR (1997) The contribution of postsynaptic folds to the safety factor for neuromuscular transmission in rat fast- and slow-twitch muscles. Journal of Physiology 500: 165–176, with permission from Blackwell Publishing.

0.2–1 MO (see later). If the full amplitude of the EPC (e.g., 250 nA) were instantly converted into a depolarization, its value (for a resistance of 0.4 MO) would be 100 mV.

Spatial factors The decay of potential from a site of current injection (e.g., the NMJ) can usually be approximately described by a single exponential (Figure 3(b)). For many vertebrate muscle fibers, the

184 Neuromuscular Junction (NMJ): Postsynaptic Events in Neuromuscular Transmission

space constant of the decay – that is, the distance required for the potential to drop to 37% (1/e) of its value at the origin – is 0.5–2 mm. It is because of this that the ACh released from the nerve has little direct effect on the membrane potential more than a few millimeters away from the NMJ. Effects of membrane capacitance Like all cell membranes, the muscle fiber membrane has electrical capacitance as well as resistance. When positive charge enters the cell through the opened AChRs, it must bring about a reorientation of the charge stored in the molecular dipoles in the membrane before a potential difference develops across the membrane. This takes time. For an instantaneous change in current flow across the membrane of a mammalian muscle fiber, the voltage typically takes 3–5 ms to reach 84% of that, a measure of the so-called time constant of the muscle fiber (Figure 3(c)). For a flow of current which is brief relative to the membrane time constant, such as the mEPC, the capacitance distorts the time course of the resulting voltage change (Figure 3(d)). In addition, because the mEPC is short relative to the time constant of the membrane, the potential never reaches the maximum value that corresponds to the current at the peak of the mEPC. In practice, the distorting effect of the membrane capacitance means that the observed peak amplitude of the EPP is typically about 30–40 mV.

Initiation of the Muscle AP Most vertebrate skeletal muscle fibers, like neurons, have voltage-gated sodium channels in their membranes. The specific forms of these channels present in vertebrate skeletal muscle are of the class designated as NaV1 and are very similar in their structure and function to those in other excitable tissues. However, they differ in detail and are encoded by different genes which are normally expressed only in muscle cells. Each NaV1 channel contains a major poreforming a-subunit with a molecular mass of about 250 kDa and a smaller accessory b-subunit. Like most voltage-gated ion channels, NaV1 channels are largely closed at the normal resting membrane potential and open when the membrane potential becomes sufficiently less negative. In the case of NaV1 channels in mammalian muscle fibers, opening begins at a membrane potential of about 60 mV and is maximal at about 30 mV. In a vertebrate twitch muscle fiber, the action potential is normally initiated at a single point along its length, the NMJ. In order for this to happen, the charge entering at that point, and the depolarization it gives rise to, must open sufficient nearby NaV1

channels so that the inward current entering through them outweighs the flow of longitudinal current away from the NMJ and thus brings the surrounding membrane to threshold. In muscle fibers, this threshold is conventionally measured away from the NMJ using two closely spaced (1 mM)

[Na+]i

Astrocyte

Figure 3 Schematic diagram of gap junctions and hemichannels in astrocytes. Gap junctions are typically open because they are exposed to low intracellular [Ca2þ]; they mediate diffusion of molecules up to 1000 Da. Hemichannels (i.e., unopposed connexons) are gated closed under normal conditions by high extracellular [Ca2þ] (see text). When opened, hemichannels allow intracellular contents to equilibrate with extracellular fluid (e.g., glutamate leaves and Naþ enters).

216 Gap Junctions and Hemichannels in Glia Table 1 Gap junction coupling in mammalian glial cells Glial cell pair

Relative coupling strengtha

Connexin pairingsb

Astrocytes Astrocyte–oligodendrocyte Oligodendrocytes Schwann cells Microglia Mu¨ller cells Mu¨ller cell–astrocyte Ependymal cells

þþþþ þþ þþ þþ þ þþ þþ þþþ

Cx43–Cx43, CX43–Cx30, Cx30–Cx30, Cx30–Cx26, Cx26–Cx26 Cx43–Cx47, Cx30–Cx32, CX26–Cx32 Cx32–Cx32, Cx29–Cx29 Cx32–Cx32, Cx29–Cx29 Cx43–Cx43 Cx43–Cx43, Cx43–Cx45, Cx45–Cx45 Cx43–Cx43, Cx43–Cx45, Cx45–Cx45 Cx43–Cx43

a

Relative coupling strength is shown as a semiquantitative score between þ and þþþþ, with þ being the lowest strength of coupling. Connexin pairings are shown in order of decreasing frequency. For example, Cx43–Cx43 is the most frequent connexin pairing for gap junctions between astrocytes. b

such as glutamate application or depolarization can increase, while inflammatory cytokines decrease, astrocyte GJ coupling. The predominant connexin expressed by these cells is Cx43, but its expression varies by brain region. Not surprisingly, Cx43 is the most abundantly expressed connexin in the CNS. There are intriguing reports indicating that astrocytes form gap junctions with neurons. However, these junctions may be limited to immature cells and occur only during development. Mammalian astrocytes in situ typically show widespread dye coupling with each other. When the GJ-permeable dye Lucifer Yellow is injected into individual cortical astrocytes, as many as 100 adjacent cells are stained. Oligodendrocytes primarily form gap junctions with astrocytes. The coupling between oligodendrocytes and astrocytes is mainly mediated by GJs composed of Cx43 (astrocytes) and Cx47 (oligodendrocytes) (Table 1). Although functional coupling between oligodendrocytes is readily detected in vitro, actual GJs are only rarely seen using electron microscopy on freshly prepared whole brain. Based on the scarcity of morphological GJs, it has been suggested that communication between oligodendrocytes might be mediated by connections through the astrocytic syncytium. ‘Coupling strength’ measured electrically and by dye passage is much greater between astrocytes than between astrocytes and oligodendrocytes. In the peripheral nervous system, proliferating Schwann cells during development, or after peripheral nerve crush injury, primarily express Cx46 and are coupled to one another. When Schwann cells begin myelinating axons, they stop expressing Cx46 but increase expression of Cx32, which is confined to paranodal regions and Schmidt–Lanterman incisures. These so-called reflexic gap junctions connect the paranodal folds, perhaps facilitating ion and small-molecule movements to and from the tight, periaxonal space. The X-linked form of the hereditary neuropathy known as Charcot–Marie–Tooth disease is associated with mutations in the Cx32 gene.

Other specialized glial cells, such as ependymal cells, Mu¨ller cells of the retina, and Bergmann glia of the cerebellum, express connexins and GJs (Table 1). Activated, but not resting, microglia express Cx43 and show dye coupling in vitro. Interestingly, activated microglia can downregulate Cx43 expression and coupling in nearby astrocytes, probably via cytokine secretion, suggesting that astrocyte communication may be modulated by CNS inflammation mediated by microglia.

Functions of GJs and HCs The functions of connexins, and the channels that they form (both GJs and HCs), are still debated and undoubtedly vary by cell type. This uncertainty about function is especially true for GJs and HCs in glial cells. The following comments about function are directed mainly at astrocytes. In general, GJs may serve to coordinate the electrical and metabolic activities of cell populations, act to amplify the consequences of signal transduction, control intrinsic proliferative capacity, and help to orchestrate the complex events of embryonic morphogenesis. Roles of GJs in astrocytes include equilibrating intracellular ions among coupled cells (Figures 3 and 4(a)), regulating cell proliferation, participation in signal dispersion (Figure 4(b)), participation in Ca2þ wave propagation (Figure 4(c)), and involvement in brain injury. Because astrocytes are strongly coupled, they can experience GJ-mediated signal dispersion (or amplification). Consider a cell stimulated to produce an intracellular second-messenger molecule (e.g., cAMP, inositol 1,4,5-trisphosphate, Ca2þ) by binding a specific ligand (Figure 4(b)). If the cell is not connected to its neighbors by gap junctions it will act in isolation. If the cell is widely coupled to other cells, however, the diffusible junction-permeable second messenger has the potential of affecting many cells, effectively amplifying the initial signal. Astrocyte Ca2þ waves are propagated waves of elevated intracellular Ca2þ that may be elicited by

Gap Junctions and Hemichannels in Glia

217

Homeostasis

Mem. Pot. [Na+] [Ca2+ ] Amino acids Metabolites

Mem. Pot. [Na+] [Ca2+ ] Amino acids Metabolites

Mem. Pot. [Na+] [Ca2+ ] Amino acids Metabolites

a Stimulus

Response amplification

Receptor

cAMP IP3 Ca2+

cAMP IP3 Ca2+

Response

Response

cAMP IP3 Ca2+ Response

b Signaling interaction: astrocyte calcium waves

Glutamate receptors Glutamate . .. . . release .. .

PLC

..

PLC

.

..

ATP

. . .. . .. . . .

IP3 [Ca2+]i

.

. .

. . ... . . . .. . .. . . . .. . . . .. .. . . . . .. .... .. . .. . . .. . . Altered .. . . . . excitability . . . .. . . . . . .. . .

Stimulus (e.g., glutamate)

Astrocyte

Ca2+

Ca2+ stores

. ..

.

.

Neuron

c Figure 4 Illustration of possible functions of gap junctions and hemichannels. (a) In strongly coupled cell aggregates, the intracellular concentrations of ions and molecules less than 1000 Da in size (including amino acids, sugars, and second-messenger molecules) will tend to be similar due to free exchange across gap junctions. (b) Glial cells express many receptors for signaling molecules (e.g., neurotransmitters, cytokines). Coupling can amplify the physiological consequences of a chemical signal acting upon a single cell. Such stimuli often evoke production of second-messenger molecules, such as Ca2þ, inositol 1,4,5-trisphosphate (IP3), and cAMP, that can diffuse into adjacent cells via gap junctions; the coupled cells are then recruited to respond. (c) Gap junctions and hemichannels participate in astrocyte Ca2þ waves. Astrocyte Ca2þ waves are provoked in many ways, such as by the application of glutamate. Effective stimuli activate intracellular signaling pathways, such as those involving phospholipase C (PLC) and IP3, which cause Ca2þ release from Ca2þ stores. IP3 and Ca2þ diffuse to adjacent cells through gap junctions, creating a ‘Ca2þ wave’. Astrocytes also release ATP upon activation, probably through hemichannels. In turn, ATP activates purinergic receptors on nearby astrocytes, the predominant mode by which Ca2þ waves propagate. Increase in astrocyte [Ca2þ]i causes glutamate release, which modulates the excitability of neighboring neurons. Reproduced from Ransom BR and Ye ZC (2005) Gap junctions and hemichannels. In: Kettenmann H and Ransom BR (eds.) Neuroglia, 2nd edn., pp. 177–189. New York: Oxford University Press, with permission.

glutamate application, mechanical perturbation, or ischemia (Figure 4(c)). As the wave passes through astrocytes it can elicit glutamate release and alter the behavior of adjacent neurons; in essence, this event is a slow form of signaling (slow compared to action potentials and synaptic transmission in neurons). Ca2þ wave propagation is dependent on both gap junctions and extracellular signaling pathways,

including hemichannel-mediated ATP release; the relative contribution of these two mechanisms may vary in different brain regions. Gap junctions tend to restrict cell proliferation. During neoplastic transformation, gap junctions are usually lost and most malignant cells are not coupled. Most primary brain tumors derive from glial cells, and loss of GJs could plausibly contribute to

218 Gap Junctions and Hemichannels in Glia

CNS neoplasia. How GJs suppress mitotic activity is not clear. Genetic manipulation is one method of assessing the function of connexins. Animals null for Cx43 die at birth because of a cardiac defect. Conditional mutants have been created in which Cx43 is suppressed only in astrocytes. These animals live to adulthood and exhibit subtle neurological changes, including increased exploratory behavior, an anxiolytic-like state, and impaired motor capacities. In humans, mutations in Cx43 cause an autosomaldominant syndrome called oculodentodigital dysplasia (ODDD). At least half of these patients have neurological abnormalities, including spastic paraplegia and neurodegeneration (i.e., marked changes in cortex and white matter on magnetic resonance imaging), in addition to craniofacial and limb dysmorphisms. The exact pathophysiology of this condition remains uncertain and some mutations may cause a toxic gain of function or disturb the normal interactions known to occur between connexin and nonconnexin proteins. The roles of glial HCs are mainly a matter of speculation at this time. Possible roles include (1) release of signaling molecules such as ATP or glutamate, (2) modulation of membrane potential or ion gradients in a manner similar to conventional ion channels, and (3) mediation of cell injury under pathological conditions such as ischemia. HCs can open under pathological conditions, perhaps due to oxidative stress, and may release glutamate, which could contribute to excitotoxic injury.

Conclusions Connexins and the channels they form have been the focus of intense investigation for greater than a half century since their discovery. As more information about these fascinating channels has emerged, more questions have appeared. This is especially so in the brain, where these channels are widely expressed in glial cells. We lack conclusive insight about how they participate in brain function and pathological conditions. We are at a point, however, where critical hypotheses about their function can be formulated and tested using improved research techniques, including regional and cell-type-specific genetic manipulation of connexin expression. See also: Gap Junctions and Electrical Synapses; Gap Junctions and Neuronal Oscillations.

Further Reading Bennett MV, Barrio LC, Bargiello TA, et al. (1991) Gap junctions: New tools, new answers, new questions. Neuron 6: 305–320.

Bruzzone R, Hormuzdi SG, Barbe MT, et al. (2003) Pannexins, a family of gap junction proteins expressed in brain. Proceedings of the National Academy of Sciences of the United States of America 100: 13644–13649. Contreras JE, Sanchez HA, Eugenin EA, et al. (2002) Metabolic inhibition induces opening of unapposed connexin 43 gap junction hemichannels and reduces gap junctional communication in cortical astrocytes in culture. Proceedings of the National Academy of Sciences of the United States of America 99: 495–500. Cotrina ML, Lin JH, Alves-Rodrigues A, et al. (1998) Connexins regulate calcium signaling by controlling ATP release. Proceedings of the National Academy of Sciences of the United States of America 95: 15735–15740. Dermietzel R, Gao Y, Scemes E, et al. (2000) Connexin43 null mice reveal that astrocytes express multiple connexins. Brain Research – Brain Research Reviews 32: 45–56. Froes MM, Correia AH, Garcia-Abreu J, et al. (1999) Gapjunctional coupling between neurons and astrocytes in primary central nervous system cultures. Proceedings of the National Academy of Sciences of the United States of America 96: 7541–7546. Gomez-Hernandez JM, de Miguel M, Larrosa B, et al. (2003) Molecular basis of calcium regulation in connexin-32 hemichannels. Proceedings of the National Academy of Sciences of the United States of America 100: 16030–16035. Goodenough DA, Goliger JA, and Paul DL (1996) Connexins, connexons, and intercellular communication. Annual Review of Biochemistry 65: 475–502. Harris AL (2001) Emerging issues of connexin channels: Biophysics fills the gap. Quarterly Review of Biophysics 34: 325–472. Kettenmann H and Ransom BR (1988) Electrical coupling between astrocytes and between oligodendrocytes studied in mammalian cell cultures. Glia 1: 64–73. Levin M (2002) Isolation and community: A review of the role of gap-junctional communication in embryonic patterning. Journal of Membrane Biology 185: 177–192. Phelan P and Starich TA (2001) Innexins get into the gap. BioEssays 23: 388–396. Ransom BR and Ye ZC (2005) Gap junctions and hemichannels. In: Kettenmann H and Ransom BR (eds.) Neuroglia, 2nd edn., pp. 177–189. New York: Oxford University Press. Rouach N, Glowinski J, and Giaume C (2000) Activity-dependent neuronal control of gap-junctional communication in astrocytes. Journal of Cell Biology 149: 1513–1526. Saez JC, Berthoud VM, Branes MC, et al. (2003) Plasma membrane channels formed by connexins: Their regulation and functions. Physiological Reviews 83: 1359–1400. Saez JC, Contreras JE, Bukauskas FF, et al. (2003) Gap junction hemichannels in astrocytes of the CNS. Acta Physiologica Scandinavika 179: 9–22. Sontheimer H, Minturn JE, Ransom BR, et al. (1991) Cell coupling is restricted to subpopulations of astrocytes cultured from rat hippocampus and optic nerve. Annals of the New York Academy of Sciences 633: 592–596. Spray DC, Ye ZC, and Ransom BR (2006) Functional connexin hemichannels: A critical appraisal. Glia 54(7): 758–773. Unger VM, Kumar NM, Gilula NB, et al. (1999) Three-dimensional structure of a recombinant gap junction membrane channel. Science 283: 1176–1180. Ye ZC, Wyeth MS, Baltan-Tekkok S, et al. (2003) Functional hemichannels in astrocytes: A novel mechanism of glutamate release. Journal of Neuroscience 23: 3588–3596.

Gap Junctions and Neuronal Oscillations M O Cunningham and F E N LeBeau, Newcastle University, Newcastle upon Tyne, UK ã 2009 Elsevier Ltd. All rights reserved.

Introduction In addition to neuronal chemical neurotransmission, electrical transmission via gap junctions, which are protein channels that directly link the cytoplasm of adjacent cells, is now known to constitute an important type of cell-to-cell communication. Although initially thought to be important mainly in development, it is now clear that electrical signaling via gap junctions plays an important role in the adult central nervous system (CNS). Considerable advances have been made in understanding the contribution of electrical signaling, via gap junctions, to neuronal signaling, most notably the role of synchronizing oscillatory or rhythmic activity. Oscillatory neuronal activity underlies the production of the signals recorded with an electroencephalogram (EEG) and reflects the coordinated activity of large populations of neurons. An EEG of the brain reveals rhythmic oscillatory activity in distinct frequency bands that occur during different behavioral states, and can be altered in a number of neurological and psychological diseases. Work from many research groups over the past decade has highlighted the role of electrical signaling via gap junctions in the generation, and/or spread, of synchronous oscillatory network activity. Electrical coupling via gap junctions can occur in both neurons and glial cells throughout the CNS. However, in this article, we focus on electrical coupling in excitatory principal cells and inhibitory interneurons, mainly in the neocortex and hippocampus. We provide an overview of the role of electrical signaling via gap junctions in the generation of oscillatory network activity and discuss briefly the importance of gap junction function in neurological conditions such as epilepsy.

Historical Background The idea that neurons can form an interconnected network via direct cell-to-cell connections is not a novel concept. Proposed by Camillo Golgi in the late nineteenth century, the ‘reticular theory’ postulated that the nervous system could form a syncytium consisting of nerve fibers forming an intricate connected network and that the nerve impulse could propagate along such a diffuse network. However, at the same time, Ramo´n y Cajal’s ‘neuron doctrine’

hypothesized that the nervous system was composed of anatomically and functionally independent units (neurons) with chemical synapses connecting them. Today we understand that in the adult mammalian brain, each of these theories reflects an important aspect of the communication in neuronal networks.

Structure and Properties of Gap Junctions in the CNS Gap junctions form a continuous link between two separate neuronal structures (Figure 1) that allows the direct transmission of electrical signals between neurons. Neurons can be coupled together via gap junctions connecting soma-to-soma, dendrite-to-dendrite, or soma-to-dendrite. More recently, evidence has appeared of gap junctions linking neurons via axonto-axon connections. The major proteins that form gap junctions are termed connexins (Cxs). Different Cx subtypes are defined by their molecular mass (in kilodaltons). A number of different Cx subtypes have now been identified in the vertebrate CNS, including Cx26, Cx32, Cx36, Cx43, and Cx45, but of these only Cx36 forms have been shown to be specific for neurons. A functional channel connecting two different neurons is formed by the apposition of two connexons, one in the plasma membrane of each neuron. Each connexon contains six Cx proteins. A functional gap junction is thus composed of 12 Cx subunits. Diversity in the composition of connexons in a gap junction can arise, as hemichannels can either consist of the same Cx subtype (homomeric) or different Cx subtypes (heteromeric). Recently, a new family of gap junction proteins, the pannexins, has been identified in the CNS. Related to innexins, an invertebrate family of gap junction proteins, the so-called pannexins (Px) have been demonstrated to form functional gap junctions in recordings from paired Xenopus oocytes. Three Px subtypes have been described in the rat and human genome, with Px1 and Px2 expressed solely in the CNS. There are strong expression patterns for Px1 and Px2 in the cortex and hippocampus, but their functional role in network activity is unknown. Gap junctions are not simply inert pores between neurons, as they can be dynamically modulated in a number of ways. Evidence suggests that the number of gap junctions, the pattern of expression of Cx proteins, and conductance properties of the individual channels can all be modified by neurotransmitters, second messengers, calcium ion concentrations, intraand extracellular pH, and transmembrane voltage. This sort of modification occurs during development

219

220 Gap Junctions and Neuronal Oscillations Closed

Open

Connexon Connexin monomer

Plasma membranes

Intercellular space 2–4 nm space

Hydrophilic channel

Figure 1 Structure of a gap junction, showing a cluster of gap junctions (yellow) spanning two opposing plasma membranes (blue sheets). The intercellular space narrows at the point of the gap junction, bringing the two membranes into close contact. A gap junction consists of two hemichannels (one in each plasma membrane) that form a continuous channel between the two cells. Each hemichannel contains six connexin proteins, which, as a unit, are called a connexon. Insets show that a conformational change in the connexin proteins allows the gap junctions to open and close in response to modulators such as pH, calcium, and cAMP.

and in response to a variety of factors in the adult. Modulation allows for the complex, connexin-specific regulation of electrical signaling via gap junctions, although to date little is known about the mechanism.

Approaches to Studying Gap Junctions Evidence for electrical coupling mediated via gap junctions in the CNS comes from using a multidisciplinary approach. The CNS consists of excitatory principal cells and inhibitory interneurons, and there is evidence for electrical signaling in both neuronal populations. Electron microscopy and freeze-fracture studies have anatomically demonstrated gap junctions between both interneurons and pyramidal cells in the cortex. Early electrophysiological evidence for electrical signaling via gap junctions was obtained by recording from principal cells in the cortex and hippocampus. These studies also showed that certain dyes (e.g., Lucifer yellow) could pass from one principal neuron to another via gap junctions. In dye coupling experiments a specific dye is injected into one cell, and if that cell is coupled to other cells, dye transfer may be detected in the connected cells. However, dye coupling studies are complicated because different connexins may have different permeabilities to dyes, and dye transfer depends on the time for filling and the locations of the gap junctions. For this reason a lack

of dye coupling does not completely exclude the presence of functional gap junctions. Interneurons represent only 10–20% of the neuronal population and are thus difficult to record ‘blind’ with intracellular sharp electrode techniques. In order to surmount this issue experimentalists have utilized infrared differential interface contrast (IR-DIC) microscopy and whole-cell patch recording techniques. This has made it possible to identify interneurons visually from their basic morphology and thus conduct simultaneous whole-cell patch recordings from two or more interneurons. Using this technique electrical coupling between pairs of interneurons has been demonstrated. When current is injected into one cell to change the membrane potential, a potential change is also seen in the coupled cell (Figure 2). The ratio of the potential change in cell 1/cell 2 is termed the coupling coefficient or coupling ratio. As with the dye transfer studies, the absence of electrical coupling potentials cannot totally exclude the possibility that gap junctions may be present at sites too distant for current transfer to be detected with somatic recordings. Electrically coupled interneurons have now been shown to be widespread in many areas of the vertebrate CNS, including the hippocampus and neocortex. However, elucidating the functional significance of electrical signaling via gap junctions has been made difficult by the lack of specific agents to block gap

Gap Junctions and Neuronal Oscillations 221

Cell 2

Cell 1

2 mV

40 mV

Cell 2

2 mV

Cell 1 20 mV

Figure 2 Electrical coupling potentials, showing electrical coupling between fast-spiking interneurons. Paired whole-cell recordings of two fast-spiking interneurons from the hippocampal dentate gyrus region in the mouse are depicted. Traces show the voltage responses in cell 1 following depolarizing (upper) and hyperpolarizing (lower) current injections; voltage responses were also reflected in cell 2 but were smaller in amplitude.

junctions. Uncoupling agents that are currently used include heptanol, octanol, halothane, carbenoxolone, and 18a-glycyrretinic acid. However, these agents will block all gap junctions (including glial gap junctions), making it impossible to clearly identify the importance of any specific gap junctions. Recently, blockers for Cx36 have been described, including quinine and the more specific mefloquine. However, all of these compounds have been reported to have a range of nonspecific effects that could also affect neuronal activity. New strategies have been developed to study the role of electrical signaling via gap junctions using genetically engineered mice that lack specific connexin proteins (see later).

Evidence for Gap Junctions between Principal Neurons Dye coupling studies show that when one principal cell is injected with dye, often one or two other cells also fill with dye, suggesting direct coupling between the cells. Procedures that increase gap junction conductance (e.g., reduced extracellular calcium) tend to increase the number of dye-coupled neurons. Conversely, the degree of dye coupling is diminished when gap junction conductance is reduced by the application of gap junction blockers. In these and earlier studies, potentials resembling small action potentials (80 Hz) or very fast oscillations (VFOs). Gamma frequency activity can be recorded in many brain areas but is particularly prominent in the neocortex and hippocampus. It is associated with a number of functions such as sensory perception, attention, learning, and memory. Ultrafast oscillations in the hippocampus are also associated with memory functions, particularly the consolidation of declarative memories.

Role of Electrical Signaling via Gap Junctions in Generating Fast Network Oscillations A major methodological advance in the study of gamma frequency oscillations has been the ability to study this activity in vitro. Persistent gamma oscillations can be induced and sustained in brain slices of hippocampus and cortex by the bath application of drugs that can depolarize both pyramidal cells and interneurons such as kainate, muscarinic, or glutamate metabotropic receptor agonists. Gamma frequency activity cannot be generated in the absence of inhibitory neurotransmission, and several studies have identified specific interneuron classes, particularly fast-spiking basket cells, that fire at gamma frequency during the oscillation and so contribute to the generation of this activity. However, in addition to its role in GABAergic inhibition, electrical signaling via gap junctions also appears to be important for the generation of gamma frequency oscillations. Recent work suggests that electrical coupling between both interneurons and pyramidal cells is important for the generation of gamma frequency activity.

It had been known for some time that application of gap junction blockers (halothane, octanol, and carbenoxolone) can abolish gamma frequency oscillations, but these studies could not distinguish the contribution of electrical signaling via gap junctions in interneurons versus pyramidal cells. However, it is possible in vitro to generate a gamma frequency oscillation in which pyramidal cell activity has been blocked – the so-called interneuronal network gamma (ING). Computational modeling studies initially suggested that, along with chemical inhibitory transmission, electrical signaling via dendritic gap junctions between interneurons was important for synchronizing gamma frequency activity. Using the in vitro ING model it was found that blockade of gap junctions markedly reduced the synchrony of the oscillations, although some gamma frequency activity still remained. The contribution of electrical coupling between interneurons to fast network oscillations was tested further using Cx36 knockout mice that specifically lack functional electrical coupling between interneurons. Using the Cx36 mice, in which excitation is intact, it was found that the ability of hippocampal neuronal networks to generate gamma frequency activity was significantly reduced. Reductions in gamma frequency activity were seen both in vitro and in vivo. At a cellular level, recordings from principal cells and interneurons in vitro showed a disruption in the amplitude, rhythmicity, and coherence of inhibitory inputs. Under normal conditions, spikelets in interneurons increase the likelihood of full action potentials being generated in the coupled cells, thus enabling a network of interneurons to fire synchronously. These synchronous inhibitory outputs impinge on the postsynaptic principal cells and control the timing of the spike output, as the principal cells can only fire once the inhibition has decayed. The inability of interneurons in the Cx36 knockout mice to synchronize properly their firing via gap junctions leads to a reduction in the precision of their action potential timing and thus their rhythmic inhibitory output. Behavioral studies have shown that mice lacking Cx36 exhibit learning and memory impairments, an observation consistent with the idea that normal gamma frequency activity is important for these cognitive tasks. Importantly, the loss of Cx36 did not lead to the complete abolition of gamma frequency activity. A residual gamma oscillation was still present both in vivo and in vitro. In vitro it was shown that this remaining gamma frequency oscillation was blocked by the putative gap junction blocking agent, carbenoxolone. This suggested that some other form of

224 Gap Junctions and Neuronal Oscillations

gap junction-mediated signaling is also required for the generation of the gamma frequency activity, and this is now thought to arise from the pyramidal cell population. Electrical signaling via gap junctions between pyramidal cells was initially proposed to underlie the generation of certain types of ultrafast or ‘ripple’ activity. Using an in vitro brain slice model, highfrequency oscillatory activity, or ‘ripples,’ occurring at 80–150 Hz can be generated in the hippocampus in the absence of chemical synaptic transmission when using a nominally calcium-free bath medium. The lack of chemical synaptic transmission in these experiments supports the hypothesis that this activity is dependent on direct electrical connections between neurons. The ‘ripple’ activity is abolished in the presence of pharmacological blockers of gap junctions, such as halothane or carbenoxolone. At the level of individual neurons, intracellular recording from principal cells revealed that, concomitant with the extracellular field potential ‘ripple,’ the cells discharged either a full-action potential or a spikelet. The kinetic properties of these spikelets (as discussed earlier) were too fast to be due to dendrodendritic gap junctions, and so axo-axonal connections were proposed. In addition, in the Cx36 knockout mice, this ultrafast oscillatory activity was normal both in vivo and in vitro, suggesting that interneuronal electrical coupling did not underlie this activity. More recent work has now demonstrated that the ultrafast oscillations coexist with the gamma frequency activity. Ultrafast oscillations can be generated in the axon plexus of pyramidal cells in the hippocampus. It is currently thought that these highfrequency oscillations occur as a result of random ectopic action potentials (action potentials arising from regions other than the axon hillock) in the axon plexus. This random activity spreads throughout the axon plexus, via axo-axonic gap junctions, resulting in a barrage of fast excitatory synaptic potentials that can be recorded in the interneurons, and provides the necessary excitatory drive for the interneurons. The combination of chemical and electrical connections between the interneurons then allows a synchronized output back to the pyramidal cells at gamma frequency. Gap junctions are therefore required between the pyramidal cells to generate the drive to the interneurons, and the gap junctions between the interneurons are needed to temporally coordinate their activity.

Gap Junctions and Epilepsy The EEG is an important test in the diagnostic process of treating epilepsy. Traditionally, this approach

has been noninvasive, using small, metal electrodes attached to the scalp. This approach is satisfactory for the medical evaluation of the seizure activity. However, there are limitations to this approach. The advent of digital recordings and invasive human studies using depth electrode recordings or subdural electrode mats now allows the measurement of electrical signals with frequencies greater than those recordable with conventional EEG equipment. Using these techniques, ultrafast oscillatory activity has been observed in close proximity to seizure onset zones in both animal models and humans. Indeed, it has been suggested that this type of activity is a reliable electrophysiological marker of epileptogenicity. A common feature of focal epileptic seizure activity, in both humans and experimental models, is the presence of ultrafast oscillations. Thus fast ripples associated with interictal spikes currently provide information on localization and epileptogenicity of the tissue involved in their generation. Evidence from developmental studies further supports a role of gap junctions in epilepsy. Thus genetic linkage mapping demonstrates that Cx36 is a high-ranking positional and functional candidate gene for juvenile myoclonic epilepsy. A number of molecular biology studies have demonstrated that in neuronal tissue obtained from animal models and human cases of epilepsy there are obvious alterations in the abundance of connexins. Most of this work concerns the expression of connexins, which is limited to astrocytes. Several studies have examined protein and or mRNA expression, but the results from these studies yield conflicting results, with some reporting increases in connexin expression and others reporting decreases. Differences in a number of the experimental parameters applied could account for these discrepancies. It is obvious that more work is required to determine whether Cxs are modified in epileptic tissue, and if so in what directions. Moreover, if putative axo-axonic gap junctions are composed of pannexins, then attempts should be made to examine the expression and localization of such proteins in both human epileptic tissue and tissue from animal experimental models of epilepsy.

Conclusions Recent advances over the past decade have demonstrated that both chemical signaling and electrical signaling play a central role in cortical communication. It is generally widely accepted that chemical signaling allows for highly complex, specific, and modifiable neuronal communication. However, it is now apparent that electrical signaling mediated via gap junctions can also demonstrate a high degree of functional specificity. Thus, the combination of

Gap Junctions and Neuronal Oscillations 225

multiple coexistent synaptic and electrical networks serves to greatly enrich the processing capabilities on neuronal networks. See also: Gap Junctions and Electrical Synapses.

Further Reading Bennet MLV and Zukin RS (2004) Electrical coupling and neuronal synchronization in the mammalian brain. Neuron 41: 495–511. Buzsaki G and Draguhn A (2004) Neuronal oscillations in cortical networks. Science 304: 1926–1929.

Connors BW and Long MA (2004) Electrical synapses in the mammalian brain. Annual Review of Neuroscience 27: 393–418. Galarreta M and Hestrin S (2001) Electrical synapses between GABA-releasing interneurons. Nature Reviews Neuroscience 2: 425–433. Hestrin S and Galerreta M (2005) Electrical synapses define networks of neocortical GABAergic neurons. Trends in Neuroscience 28: 304–309. Schmitz D, Schuchmann S, Fisahn A, et al. (2001) Axo-axonal coupling: A novel mechanism for ultrafast neuronal communication. Neuron 31: 831–840. Whittington MA and Traub RD (2003) Interneuron diversity series: Inhibitory interneurons and network oscillations in vitro. Trends in Neuroscience 26: 676–682.

This page intentionally left blank

AMINO ACID TRANSMITTERS AND RECEPTORS

A. Glutamate

229

B. GABA

340

C. Glycine

373

This page intentionally left blank

Glutamate S P H Alexander, University of Nottingham Medical School, Nottingham, UK ã 2009 Elsevier Ltd. All rights reserved.

Introduction Glutamate is easily the most abundant transmitter in the nervous system and is reported to be the transmitter at 40% of all synapses in the brain. Indeed, it has been said that if a neuron fails to respond to glutamate, it is either not a neuron or it is dead. In itself, therefore, this abundance can lead to some issues. For example, a wide distribution of multiple pharmacological targets suggests that agents targeting the glutamate system will either lack target selectivity or impact on multiple bodily functions. However, it has become clear in recent years that the numerous receptors for glutamate and their localized expression have allowed the development of several potentially exciting therapeutic agents, although relatively few are sufficiently advanced as to reach a clinical setting. In common with many other neurotransmitters (g-aminobutyric acid (GABA), acetylcholine, 5-hydroxytryptamine, and adenosine 50 -triphosphate), the action of glutamate is mediated via both rapidly acting transmitter-gated channels (ionotropic receptors, Table 1) and more slowly acting seven-transmembrane (7TM), G-protein-coupled (metabotropic, Table 2) receptors.

Metabolism, Uptake, and Release Since glutamate is a component of every cell, there is no clear synthetic route that sets apart neurotransmitter glutamate. One intriguing aspect of glutamate metabolism, however, is glutamic acid decarboxylase activity. This enzyme performs an elegant switch in emphasis, by catalyzing conversion of the major excitatory neurotransmitter (glutamate) to the major inhibitory neurotransmitter (GABA). Glutamate Transport: Plasma Membrane and Vesicular Mechanisms

Due to the major roles of glutamate as an extracellular signaling molecule and as an intracellular building block for proteins combined with the toxicity associated with elevated extracellular levels, effective mechanisms exist for removal of glutamate from the extracellular fluid and also for concentrating the amino acid in synaptic vesicles. Thus, at least five cell-surface glutamate (and aspartate) transporters

have been defined as members of the solute carrier family 1 (SLC1), while three vesicular transporters (of the SLC17 family) have also been identified. The cell-surface transporters are multisubunit, homo- or heteropentameric structures in which each subunit appears to have 8TM spanning domains and at least one re-entrant loop that ‘dips’ into the plane of the membrane. One of the physiological roles for the avid glial glutamate transport system is to allow glial conversion of glutamate to glutamine, which is then exported via the extracellular space into neurons as a means of regenerating neurotransmitter glutamate. Determination of glutamate uptake in tissue slices or cultured cells is most often accomplished through the use of radiolabeled glutamate, as metabolism to other entities may confound direct assessment of glutamate accumulation. Detection of Glutamate Release

In order to define a role for glutamate as a mediator of a physiological or pathophysiological mechanism, it is fundamental to be able to measure changes in extracellular glutamate in nervous tissues. In vitro, this is commonly accomplished through the use of [3H]-D-aspartate as a surrogate marker, since the D-isomer is less likely to be metabolized. However, there are doubts as to whether [3H]-D-aspartate accumulates in synaptic vesicles and thus the assay may more accurately reflect cytoplasmic glutamate release. It is, therefore, preferable to assess glutamate release directly. One means is to label tissue with [3H]-L-glutamine, with a view to labeling preferentially the neuronal glutamate precursor pool. However, this approach has the drawback that a chromatographic separation of glutamate from glutamine and GABA is required, although it may be useful in some cases to identify simultaneously levels of both excitatory glutamate, as well as inhibitory GABA. A means of directly measuring glutamate release in vitro is through the use of high-performance liquid chromatography (HPLC) separation of extracellular fluid following amino acid derivatization with o-phthaldehyde. This has the potential advantage of allowing levels of glutamate and GABA (as well as other bioactive released amino acids such as glycine, aspartate, and taurine) to be assessed concurrently, although it has the disadvantage that a certain period of accumulation (usually 15–20 min) is required to obtain sufficient material to detect. Two ‘real-time’ in vitro assays for glutamate have been described: an enzyme-coupled spectrophotometric version making use of exogenous glutamate dehydrogenase activity and a glutamateselective electrode.

229

230 Glutamate Table 1 Ionotropic glutamate receptors: composition and pharmacology AMPA

Kainate

NMDA

Quisqualate GLUA1, GLUA2, GLUA3, GLUA4 Homo- or hetero-tetramers Naþ, Kþ, (Ca2þ) Polyamines, argiotoxin, Joro toxin

GLUK1, GLUK2, GLUK5, GLUK6, GLUK7 Homo- or hetero-tetramers Naþ, Kþ Polyamines

GLUN1, GLUN2A, GLUN2B, GLUN2C, GLUN2D, GLUN3A, GLUN3B Heterotetramers Naþ, Kþ, Ca2þ Dizolcipine, ketamine, phencyclidine, memantine, amantidine

Endogenous ligand(s)

L-Glutamate

L-Glutamate

L-Glutamate,

Selective agonists

AMPA, (S)-fluorowillardiine

ATPA, (S)-5-iodowillardiine, (2S,4R)-4-methylglutamate, LY339434, domoic acid UBP302

DL-(tetrazol-5-yl)

Other names Subtypes Composition Ions gated Channel blockers

Glutamate site L-Aspartate

Selective antagonist

NBQX, ATPO, LY293558, GYKI53655

glycine, homoquinolinic acid D-AP5, CGS19755

Glycine site Glycine, D-Serine (þ)HA966

5,7-DCKA, L689560, L701324

In vivo detection of extracellular glutamate has been conducted through the use of microdialy coupled to HPLC separation and o-phthaldehyde detection or alternatively through the use of a glutamate-sensitive electrode.

lining re-entrant segment (termed MD2 or p-loop) that inserts into the plane of the membrane from the cytoplasmic face (Figure 2) between TM1 and TM3. The cytoplasmic C-terminus is the site for interaction of a number of protein partners (vide infra).

Receptors and Signaling

Nomenclature and composition As ionotropic glutamate receptors are multi-subunit proteins with few selective pharmacological ligands, nomenclature of the individual subunits is based on physical (i.e., primary sequence) distinctions rather than pharmacological means. Of the eight subunits identified as NMDA receptor components, a minimal configuration in heterologous expression of the NMDA receptor appears to be two pairs of GLUN1 and GLUN2 subunits, although it appears more likely that, physiologically, NMDA receptors are mixtures of all three subunit groups (GLUN1, GLUN2, and GLUN3, encoded by the genes GRIN1, GRIN2A/2B/2C/2D, and GRIN3A/3B). The GLUN3 subunits appear to have a generally inhibitory influence on NMDA receptor function. Many of the genes encoding NMDA receptor subunits undergo alternative splicing, which can generate multiple isoforms with differing pharmacological profiles. AMPA receptors appear to be made up of any permutation of GLUA1, GLUA2, GLUA3, and GLUA4 subunits (derived from the genes GRIA1–4, respectively); although GLUA2 subunits appear to require co-expression with other AMPA receptor subunits for cell-surface localization. A feature of the AMPA receptor subunit GLUA2 is the ability to function as two versions inter-converted following RNA editing (resulting in a switch from a glutamine to arginine residue; a Q/R switch) from a channel highly permeable

Twenty-four separate gene products which make up the cell-surface receptors for glutamate have been identified and are divided into two major classes both structurally and functionally; rapidly responding ‘ionotropic’ transmitter-gated channels and slower responding ‘metabotropic’ 7TM receptors (Figure 1). The ionotropic receptors consist of three groups of receptors (yellow highlight in Figure 1), identified by the selective synthetic agonists a-amino-3-hydroxy-5methyl-4-isoxazole proprionic acid (AMPA), kainate, and N-methyl-D-aspartate (NMDA). Metabotropic glutamate (mGlu) receptors also consist of three groups as identified by similarities in primary sequence (blue highlight in Figure 1), known simply as group I, group II, and group III. Ionotropic Glutamate Receptors: AMPA, Kainate, and NMDA Glutamate Receptors

Rapid responses (30 mM), while activation of kainate receptors by AMPA is dependent on the constituent subunits, with the GLUK2 subunit appearing to confer AMPA sensitivity (Figure 3). The quinoxalinedione antagonists, such as NBQX (Figure 4) and CNQX, show some limited selectivity in distinguishing AMPA from kainate receptors (Table 1). Synthetic allosteric inhibitors of AMPA and kainate receptor function (GYKI5246 and NS3763, respectively) have also been described. Intriguingly, a number of natural toxins (e.g., from Joro spiders, Figure 4) have been described which function as channel blockers for AMPA (and, to a lesser extent, kainate) receptors. Presumably, these have evolved as a means of predators disabling insect prey, since glutamate is the neurotransmitter at the invertebrate neuromuscular junction, acting on postsynaptic AMPA-like receptors. In addition, aniracetam, cyclothiazide, and diazoxide are widely used to interfere with the channel desensitization characteristics of AMPA (and, to a lesser extent, kainate) receptors through an undefined allosteric interaction. Protein partners A key area of regulation of (particularly) AMPA and kainate receptors is desensitization and intracellular trafficking. An intracellular protein

O O

O

HO

O

O

HO

OH

HN

OH

NH2

O

N

AMPA

Kainate O

O H HO

NMDA

NH2

O

O

HO

OH O

OH NH

O

O

OH O

O

HO

NH

O

P

OH

N H

1S,3R-ACPD

HO O

O O HO

HO

OH

OH NH2 L-Glutamate

NH2

LSOP

NAAG

O

OH

O

NH2

L-Aspartate

Figure 3 Glutamate receptor agonists. The top row indicates the eponymous agonists at AMPA, kainate, and NMDA ionotropic glutamate receptors. The middle row indicates mGlu receptor agonists: the nonselective agonist 1S,3R-ACPD, the mGlu3-selective NAAG, and the mGlu4/7/8-selective L-serine-O-phosphate. The bottom row shows the endogenous excitatory amino acids, L-glutamate and L-aspartate.

234 Glutamate O HO

Cl

OH

P

O

OH

O

H N

O OH

NH2

NH O Cl DCKA

D-AP5

Ketamine OH

S NH2 O

O

O

H N

N H

UBP302 HO

O

O

N+ O O

O

HO

O H2N

O N

N O

JSTX3

O

NH2

S NH2 N

N

H O

LY341495

OH

NH

OH NH2

O HN

O

HN

NBQX H N

H N

N H

O

H N

O

Cl

OH

N

O

O

OH

O

P

OH

HO NH2 MTEP

MAP4

Figure 4 Glutamate receptor antagonists. The top row illustrates the three classes of NMDA receptor antagonist: antagonists competing for glutamate (D-AP5) and glycine (DCKA) binding sites, as well as an uncompetitive (open channel blocker, ketamine) antagonist. The middle row presents competitive AMPA (NBQX) and kainate (UBP302) receptor antagonists and a spider venom-derived AMPA channel blocker (JSTX3). The bottom line illustrates a nonselective mGlu receptor competitive antagonist (LY341495) and noncompetitive mGlu5 (MTEP) and competitive mGlu4/6 (MAP4) receptor antagonists.

called glutamate receptor-interacting protein (GRIP) appears to allow interaction between GLUA2 subunits and other proteins, including protein interacting with C kinase (PICK, which recruits protein kinase C) and kinesin (an intracellular molecular motor associated with the cytoskeleton). AMPA receptors also interact with Lyn, a protein kinase, which in turn signals via the extracellular signal-regulated kinase signaling pathway, leading to an increased expression of brainderived neurotrophic factor. This growth factor is associated with neurogenesis in the adult brain and its expression is reduced following exposure to stress. mGlu Receptors

mGlu receptors are members of the family C of 7TM receptors, which also includes GABAB and calciumsensing receptors. This family is made up of long polypeptides (1000 aa) and is characterized by the presence of a large extracellular N-terminus (600 aa, Figure 5). This portion of the receptor contains the agonist binding site and a cysteine-rich domain, which appears to stabilize the three-dimensional structure of the receptor. The 7TM domains bear little sequence homology with the classical biogenic amine receptors (members of the rhodopsin class,

family A), which make up the ligand-binding site in the plane of the plasma membrane. The N-terminus binding site is often described as a ‘Venus fly-trap’ arrangement in which the ligands bind within the hinge region (Figure 5). Conceptually, there is thus an obstacle to overcome as to how ligand binding at some distant extracellular point is able to influence G-protein activity across the hydrophobic plasma membrane. The means by which this appears to be mediated is through dimerization of the receptor, thereby allowing conformational changes in the extracellular region to be transduced through the membrane. A further consequence of the extended structure of the C family of 7TM receptors is the regulation of receptor activity through ligands binding at sites other than the cognate ligand-binding site. Thus, for many of the mGlu receptor subtypes, selective positive and/or negative allosteric modulators have been described which extends the potential for pharmacological/therapeutic exploitation considerably (vide infra). The cytoplasmic C-terminus is also enlarged in comparison with most members of the rhodopsin group of 7TM receptors and this is the site for interaction with a number of intracellular protein partners (vide infra).

Glutamate 235

GLU

GLU

Glutamate binding domains

Cysteinerich regions

Transmembrane domains

Protein:protein interaction

Figure 5 A dimeric mGlu receptor. Indicated are the extracellular N-terminus glutamate binding site, cysteine-rich conformation stabilizing regions, the transmembrane domains (TM1–7) and the intracellular C-terminus site for protein:protein interactions.

Eight subtypes of mGlu receptor have been identified. They are usually classified into three groups on the basis of similarities in primary sequence (Figure 1), signaling characteristics and pharmacology. Thus, group I mGlu receptors (mGlu1 and mGlu5) couple primarily to elevation of calcium, while group II (mGlu2 and mGlu3) and group III (mGlu4, mGlu6, mGlu7, and mGlu8) receptors evoke an inhibition of cAMP. Localization mGlu receptors are located throughout the brain and spinal cord, and are also found in the retina (mGlu6), pancreas (mGlu1, where they appear to subserve a role in the monitoring of nutritional status), and in taste buds (mGlu4, where they mediate sensation to the ‘umami’ taste). In neural tissues, there is evidence for differential neuronal and astrocytic locations of mGlu receptors. mGlu1 receptors appear to be exclusively expressed in neurons, while mGlu5 receptors are found on both astrocytes and neurons. Similarly, mGlu2 receptors are selectively neuronally expressed, while mGlu3 receptors have been identified on both neurons and astrocytes. Of the group III receptors, mGlu6 receptors appear to be exclusively retinal, while the remainder are primarily neuronally located. Pharmacology and Signaling

Although glutamate is an endogenous ligand at all of the mGlu receptors, the potency of glutamate varies widely. mGlu1, mGlu3, and mGlu5 receptors respond to submicromolar glutamate, while mGlu2, mGlu4, mGlu6, and mGlu8 receptors respond to glutamate at concentrations up to 10 mM. In contrast, mGlu7 receptors only respond to concentrations

of glutamate approaching the millimolar range, leading to the suggestion either that glutamate is not the principal endogenous ligand or that the mGlu7 receptor acts presynaptically as a glutamate autoreceptor, sensing synaptic release of glutamate and switching off further release as ‘toxic’ levels are reached. Metabotropic receptors may be activated selectively compared to the ionotropic glutamate receptors by 1S,3R-ACPD (Figure 3), while LY341495 is a useful antagonist with a similar selectivity (Figure 4). Selective agonists and antagonists exist for the three groups of, if not the individual, mGlu receptors (Table 2). Thus, quisqualate is a high potency group I agonist, although it also activates AMPA receptors potently, while LY393675 acts as a group I-selective antagonist. A useful group II agonist is LY354740, while EGLU antagonizes group II receptors, albeit with low affinity. L-AP4 is a nonselective group III agonist, while MAP4 blocks most of these receptors (Figure 4). Some members of the mGlu receptor family also respond to other endogenous ligands, raising the possibility of native activation in the absence of glutamate. Thus, N-acetylaspartylglutamate (also known as spaglumic acid) is a selective mGlu3 receptor agonist, while L-serine-O-phosphate activates group III mGlu receptors (Table 2 and Figure 3). A relatively unusual feature of the family C of 7TM receptors is the regulation of receptor activity through allosteric sites. Thus, both positive and negative allosteric modulators have been described for mGlu1, mGlu2, mGlu4, mGlu5, and mGlu7 receptors. This means (for example) that mGlu5 receptors can be activated by 2-chloro-5-hydroxyphenylglycine, a response which can be enhanced by 3,30 -difluorobenzaldazine,

236 Glutamate

which alone is ineffective. Activation of the mGlu5 receptor can be inhibited either competitively (3-hydroxy-6-methyl-N-(6-methylpyridin-2-yl)pyridine-2carboxamide) or noncompetitively (3-[(2-methyl-1,3thiazol-4-yl)ethynyl]pyridine, Table 2). The group I mGlu receptors, mGlu1 and mGlu5, couple primarily via activation of members of the Gq/11 family of G-proteins to phospholipase C-b and, hence, to elevation of intracellular calcium ion levels. The primary mode of signaling of the remaining mGlu receptors is via pertussis toxin-sensitive Gi/o family members. This results in inhibition of adenylyl cyclase activity and the subsequent reduction in cyclic AMP levels. Arguably of more importance in neuronal cells is the activation of potassium channels and the subsequent hyperpolarization, resulting (typically) in a reduction in transmitter release. The mGlu6 receptor, expressed in ON retinal bipolar cells, couples to activation of PDE6, a cyclic GMP-selective phosphodiesterase, thereby leading to the closure of cGMPactivated cation channels. Protein partners It is apparent that the expanded C-terminus of the mGlu receptors has a role in the interaction with intracellular proteins, beyond the archetypal G-proteins. In particular, the group I mGlu receptors are associated with an immediate early gene product, Homer, which is enriched at the postsynaptic density. Homer appears to play an anchoring role, bringing together elements of the signaling cascade, including mGlu1 and mGlu5 receptors, phospholipase C, cell-surface transient receptor potential canonical (TRPC) calcium channels, intracellular inositol 1,4,5trisphosphate receptors, and the phosphatidylinositol 3-kinase enhancer protein, PIKE. It has been suggested that the major function of Homer is to act as a ‘brake’ on group I mGlu receptor activity through the calcium mobilization pathway and to enhance coupling via PI 3-kinase. These protein partners may also be involved in bringing together mGlu and NMDA receptors, leading to a functional interaction, such that mGlu5 receptor activation enhances NMDA receptor function, while NMDA receptor activation reverses mGlu5 receptor desensitization. Physiological Functions

The primary physiological action of glutamate in the central nervous system (CNS) is as an excitatory transmitter, allowing neuronal pathways to function throughout the brain. At the same time, glutamate is thought to be the major transmitter underlying the acquisition of memory. The molecular correlates of memory acquisition appear to differ according to the

synapse and (possibly) the maturity of the animal. The phenomenon of long-term potentiation (LTP), which has been suggested to occur at every excitatory synapse in the CNS, is demonstrated as an increased efficacy of synaptic transmission following a period of high frequency stimulation. It is well described at synapses of the Schaffer collateral–commisural pathway in the hippocampus, in which high-frequency stimulation of the perforant path input from the entorhinal cortex causes a long-lasting enhancement of low-frequency activation of the CA1 pyramidal cell output pathway. Cerebellar long-term depression (LTD) is a model of synaptic memory in which the efficiency of synaptic transmission (activation of Purkinje neurons) is reduced following high-frequency stimulation (coincident activation of parallel and climbing fiber presynaptic pathways). One explanation of these changes in efficacy of synaptic transmission, which does not hold true at all synapses, is that the underlying mechanisms involve trafficking of AMPA receptors, such that LTP may be associated with an increase in postsynaptic AMPA receptor expression (possibly resulting from recruitment of cryptic receptors), while LTD may result from phosphorylation and internalization of AMPA receptors, reducing postsynaptic receptor numbers. This plasticity, therefore, is consistent with an increased and decreased efficiency of synaptic conductance, respectively. One of the most insightful means of determining physiological roles of particular proteins is though the use of gene disruption models, although a caveat with embryonic knockout models is that the possibility exists for compensatory mechanisms to lessen or alter the true role of the gene and protein in question. Table 2 identifies some of the findings from such gene disruption studies focusing on mGlu receptors. A common theme is an alteration in synaptic plasticity, observed with disruption of mGlu1, mGlu2, mGlu4, and mGlu5 receptors, further underlining the importance of neurotransmitter glutamate in memory acquisition.

Glutamate as a Peripheral Neurotransmitter There is good evidence that glutamate subserves an excitatory transmitter function in the gastrointestinal tract, with indications of glutamate release, ionotropic and mGlu receptors, and glutamate transporters in enteric ganglia. In addition, application of glutamate and its analogs has been shown to regulate gastrointestinal motility and solute secretion. Synapses in the enteric nervous system undergo

Glutamate 237

LTP-type plasticity, and it is thought that group I mGlu receptors mediate this phenomenon, at least in part.

Pathology Stroke, Ischemia, and Neurodegenerative Diseases

An overabundance of extracellular glutamate is associated with excitotoxicity of neurons, which appears to involve activation of NMDA receptors. The mechanism of excitotoxicity is not absolutely defined, but there is a crucial role of a prolonged elevation of intracellular calcium ions, leading to mitochondrial damage and apoptosis. The NMDA receptor antagonist dizocilpine reached an advanced preclinical level as a potential antistroke therapy, failing on the basis of damage to unstressed brain areas due to a narrow therapeutic window. This drug may have been a victim of its own success, in that the high affinity and low reversibility displayed as an open channel blocker allowed little differentiation between positive and harmful effects. The generation of a lower-affinity agent, memantine, which exhibits greater reversibility and an increased tolerability has allowed approval for its use in man. The current indication for memantine is for treatment of Alzheimer’s disease. The rationale for its use is that the prolonged presence of glutamate (albeit at lower concentrations) leads to protracted NMDA receptor activation and excitotoxicity of neurons and a cognitive deficit. By reducing NMDA receptor activation, it is hoped to reduce the excitotoxicity and reduce the cognitive deficit. The accumulation of plaques positive for b-amyloid in Alzheimer’s disease is associated with neurodegeneration and, since processing and secretion of g-secretase (an enzyme which cleaves the amyloid precursor protein) is enhanced by activation of group I mGlu receptors, these receptors have also become a focus for research in this area. Another neurodegenerative disorder in which glutamate overactivity has been implicated is Parkinson’s disease. The glutamatergic pathways involved may be either thalamostriatal or corticostriatial, and so it has been suggested that group II agonists or mGlu5 antagonists might be therapeutically useful by inhibiting glutamate release or action, respectively, at these synapses. Indeed, LY354740 has been observed to display efficacy in animal models of Parkinson’s, although the mGlu5 antagonist 2-methyl-6-(phenylthynyl) pyridine was ineffective. A further potential exploitation of glutamate signaling in combating the effects of neurodegenerative diseases, such as Alzheimer’s and Parkinson’s, is the use of ‘AMPAkines,’ agents which enhance AMPA

receptor function. As described above, a number of agents interfere with AMPA receptor function as allosteric modulators or by impeding desensitization. Aniracetam is an example of such an agent, which has been exploited as a nootropic substance, also known as ‘smart drugs.’ These agents have a positive profile in animal models and those in clinical and ‘street’ use are reported to enhance cognitive powers (concentration, memory, etc.); their use has yet to be approved in the treatment of Alzheimer’s, although piracetam is used to treat the involuntary twitches suffered by Parkinson’s patients. Epilepsy

In vitro and in vivo focal application of glutamate results in epileptiform activity. This action appears to be mediated by the ionotropic receptors, and postmortem examination of the brains of epileptics suggests a neurodegenerative profile consistent with overactivity of these receptors. One potential therapeutic avenue is the exploitation of mGlu4 and mGlu7 receptors, which have been suggested to be glutamate autoreceptors, acting presynaptically to reduce glutamate release at the synapse. As noted in Table 2, disruption of either of these receptors leads to changes in seizure susceptibility further highlighting the potential for exploitation of glutamate signaling in treating this disorder. Anxiety and Depression

In preclinical assessments for leads as anti-anxiety/ anti-depressant action, there are promising results for both mGlu2-selective agonists and mGlu5-selective antagonists. It is hypothesized that the mGlu2 agonists function by reducing frontal cortex transmitter release presynaptically, while the mGlu5 antagonists regulate postsynaptic glutamate excitability. Pain

Several observations suggest that mGlu receptors represent an excellent target for the treatment of pain. Members of all three groups of mGlu receptors are present at the primary afferent-spinal cord synapse or the subsequent ascending pathway. There is particular interest in group I mGlu receptors, partly because of evidence for diminished pain sensitivity in transgenic mice (Table 2), but also because of the efficacy of antagonists in multiple pain models. Schizophrenia

Schizophrenia is suggested to be the consequence of an imbalance of glutamatergic and dopaminergic signaling systems. Phencyclidine, which blocks

238 Glutamate

NMDA receptors (as well as dopamine uptake), induces schizophrenia-like symptoms in man. In animal models, the group II agonist LY354740 was observed to reduce some of the behavioral effects of PCP, with a distinct pattern to traditional neuroleptics. In addition, the observation that mGlu5-knockout mice have an altered behavioral trait observed in schizophrenics (Table 2) has led to heightened interest in the therapeutic exploitation of these receptors for this condition.

See also: AMPA Receptors: Molecular Biology and Pharmacology; Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDARDependent Forms; Long-Term Potentiation (LTP): NMDA Receptor Role; Metabotropic Glutamate Receptors (mGluRs): Functions; Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology.

Addiction

Further Reading

There is extensive interest in the use of group I antagonists and group II agonists as potential therapeutic agents in the treatment of drug addiction. In animal models, these agents reduce the symptoms of withdrawal from nicotine and morphine. In addition, there is evidence for sensitization of group II mGlu receptors in particular brain areas associated with drug dependence, following chronic exposure to morphine.

Alexander SPH, Mathie A, and Peters JA (2006) Guide to receptors and channels, 2nd edn. British Journal of Pharmacology 147 (supplement 3): S1–S183. Chen PE and Wyllie DJ (2006) Pharmacological insights obtained from structure–function studies of ionotropic glutamate receptors. British Journal of Pharmacology 147: 839–853. Dingledine R, Borges K, Bowie D, and Traynelis SF (1999) The glutamate receptor ion channels. Pharmacological Reviews 51: 7–61. Hinoi E, Takarada T, Tsuchihashi Y, and Yoneda Y (2005) Glutamate transporters as drug targets. Current Drug Targets CNS and Neurological Disorders 4: 211–220. Kew JNC and Kemp JA (2005) Ionotropic and metabotropic glutamate receptor structure and pharmacology. Psychopharmacology 179: 4–29. Lujan R, Shigemoto R, and Lopez-Bendito G (2005) Glutamate and GABA receptor signalling in the developing brain. Neuroscience 130: 567–580. Madden DR (2002) The structure and function of glutamate receptor ion channels. Nature Reviews Neuroscience 3: 91–101. Matute C, Domercq M, and Sanchez-Gomez MV (2006) Glutamate-mediated glial injury: Mechanisms and clinical importance. Glia 53: 212–224. Niswender CM, Jones CK, and Conn PJ (2005) New therapeutic frontiers for metabotropic glutamate receptors. Current Topics in Medicinal Chemistry 5: 847–857. Robbins TW and Murphy ER (2006) Behavioural pharmacology: 40þ years of progress, with a focus on glutamate receptors and cognition. Trends in Pharmacological Sciences 27: 141–148. Swanson CJ, Bures M, Johnson MP, et al. (2005) Metabotropic glutamate receptors as novel targets for anxiety and stress disorders. Nature Reviews Drug Discovery 4: 131–144. Watkins JC and Jane DE (2006) The glutamate story. British Journal of Pharmacology 147(supplement 1): S100–S108.

Cerebellar Ataxia

Animal models of disrupted mGlu1 receptors suggest many similarities with cerebellar ataxia, but of arguable greater significance is the observation of auto-antibodies against mGlu1 receptors in patients suffering from paraneoplastic cerebellar degeneration. These antibodies were observed to inhibit cerebellar function in vitro and in vivo arguing for a pivotal role for mGlu1 receptors in this disorder.

Concluding Remarks Although there has been an abundance of research investigating the physiological and pathophysiological roles of glutamate, as well as extensive medicinal chemistry programs to develop selective ligands, there is still an under-representation of therapeutic agents focusing on this pivotal transmitter.

Glial Energy Metabolism: Overview L Pellerin, Universite´ de Lausanne, Lausanne, Switzerland P J Magistretti, Ecole Polytechnique Fe´de´rale de Lausanne (EPFL) and Centre Hospitalier Universitaire Vaudois, Lausanne, Switzerland ã 2009 Elsevier Ltd. All rights reserved.

Introduction In the late nineteenth century, neuroanatomists described the morphological features of glial cells and their cytoarchitectural relationships with other elements of the central nervous system. Among glial cells, astrocytes are the most numerous and exhibit a number of striking characteristics. First, they occupy a strategic position, being usually located between blood vessels and neurons. Second, they possess specialized processes called end-feet, and these processes come into close contact with blood vessels, covering a large part of their surface. Third, they also ensheathe numerous synapses. Altogether, these characteristics suggested to those early neuroanatomists, and among them Camillo Golgi, that astrocytes could play a critical role in the allocation of metabolic substrates to neurons. This view was nicely formulated by Andriezen in 1893: The development of a felted sheath of neuroglia fibers in the ground-substance immediately surrounding the blood vessels of the Brain seems therefore . . . to allow the free passage of lymph and metabolic products which enter into the fluid and general metabolism of the nerve cells.

It is only recently, however, that such a theory experienced a revival through a series of experimental demonstrations that benefited from the advances made in the isolation of specific brain cell types and their use in vitro to study their metabolic characteristics. The current understanding of brain energy metabolism is still incomplete, but in recent years insights have been gained into a number of aspects related to glial energy metabolism.

Aerobic Glycolysis: A Versatile Metabolic Feature of Astrocytes to Cope with the Burden of Glutamate Recycling Several studies have demonstrated the large glycolytic capacity of astrocytes that is accompanied by a sizable production and release of lactate despite the sufficient amount of oxygen present and active oxidative phosphorylation. Not only is their basal glycolytic

rate elevated, but astrocytes also possess a large glycolytic reserve that can be mobilized under various conditions. This is the case under hypoxia or if oxidative phosphorylation is inhibited. But more important, this reserve can be used in specific physiological situations. Thus, it has been shown that enhanced sodium entry within an astrocyte, causing an elevation of intracellular sodium concentration, will lead to an increased glycolytic rate. Such a situation occurs notably when astrocytes are exposed to the excitatory neurotransmitter glutamate. Glutamate transport occurs in astrocytes via two glutamate transporters, glutamate aspartate transporter (GLAST) and glutamate transporter 1 (GLT1). Both depend on the Naþ gradient and carry glutamate, together with three Naþ ions, inside the cell. The increase in intracellular Naþ concentration that accompanies glutamate transport will activate Naþ/Kþ adenosine triphosphatase (ATPase), and as a consequence of adenosine triphosphate (ATP) consumption, an activation of glucose utilization will take place, together with enhanced lactate production (Figure 1). Such an activation of aerobic glycolysis in astrocytes appears to be facilitated by a few molecular characteristics. First, it has been shown that astrocytes have a large part of their pyruvate dehydrogenase enzyme present in a phosphorylated and thus inactive form. This characteristic may limit the entry of pyruvate into the tricarboxylic acid (TCA) cycle (Figure 2). Second, astrocytes appear to have very reduced levels of an essential component of the malate–aspartate shuttle called ARALAR 1. As a consequence, the transfer of reducing equivalents from the cytosol to the inner mitochondria is reduced, and cytosolic nicotinamide adenine dinucleotide (NADþ) regeneration is limited. To maintain a high glycolytic rate as observed in astrocytes, which requires a rapid regeneration of NADþ levels, the solution is to convert pyruvate to lactate via the lactate dehydrogenase enzyme. To facilitate this task, astrocytes prominently express the lactate dehydrogenase M isoform that is not inhibited by pyruvate and sustains high enzymatic rates. Moreover, astrocytes were found to express the monocarboxylate transporter (MCT) isoforms MCT1 and MCT4 that are particularly suited for lactate export from glycolytic tissues because of high Km values. An important question is whether such a metabolic response of astrocytes can be triggered by other neuroactive substances. The main inhibitory neurotransmitter, g-aminobutyric acid (GABA), is an interesting case since it is also transported by a Naþ-dependent mechanism in astrocytes and it is the most important neurotransmitter after glutamate. However, it was

239

240 Glial Energy Metabolism: Overview

Glutamatergic synapse

Astrocyte

Capillary

Lactate Glutamine

Synaptic vesicles

ADP ATP

Glucose

Glucose

Glutamate EAAT

G

H+ K+ Na+

Neuronal glutamate receptors

K+

Na/K ATPase

Glycolysis ATP ADP

Figure 1 Schematic representation of the mechanism for glutamate-induced glycolysis in astrocytes during physiological activation. EAAT, excitatory amino acid transporter. Modified from Pellerin L and Magistretti PJ (1994) Glutamate uptake into astrocytes stimulates aerobic glycolysis: A mechanism coupling neuronal activity to glucose utilization. Proceedings of the National Academy of Sciences of the United States of America 91: 1065–10629.

shown that GABA does not cause a similar enhancement of aerobic glycolysis in astrocytes. The explanation resides in a rapid reduction in GABA transport with increasing intracellular Naþ concentration before it reaches the threshold to activate glycolysis, a phenomenon not observed with glutamate transport. Several other neurotransmitters and neuromodulators have been tested, but all failed to activate glycolysis except noradrenaline. However, the mechanism by which noradrenaline stimulates glycolysis in astrocytes is different from that of glutamate but remains unclear. Another intriguing issue is whether the effect of glutamate is linked only to its uptake via glutamate transporters or whether activation of glutamate receptors (GluRs) can also participate in the metabolic effect. Indeed, astrocytes express both ionotropic and metabotropic glutamate receptors. Among ionotropic receptors, a-amino-3-hydroxy-5methyl-4-isoxazole propionic acid (AMPA)/kainate receptors are abundantly expressed, and all four subunit isoforms (GluR1–4) have been detected in astrocytes. Although activation of AMPA/kainate receptors with the specific agonist AMPA leads to a transient increase in intracellular Naþ concentration, this is not sufficient to cause an activation of aerobic glycolysis. In this case, rapid desensitization of the receptors prevents the Naþ influx from activating

Glycogen

Glucose

Glucose-6-phosphate Pentose phosphate pathway

Glycolysis

Pyruvate

Energy

TCA cycle

NADP NADPH

LDH

Lactate

CO2 H2O

Figure 2 Metabolic pathways of glucose. LDH, lactate dehydrogenase; NADP, oxidized nicotinamide adenine dinucleotide phosphate; NADPH, reduced nicotinamide adenine dinucleotide phosphate; TCA, tricarboxylic acid cycle.

metabolism. Kainate causes a more important intracellular Naþ change and a small but significant stimulation of aerobic glycolysis. This effect could be explained by the fact that kainate, in contrast to

Glial Energy Metabolism: Overview

AMPA, leads to only a partial desensitization of AMPA/kainate receptors. One particularity of AMPA/kainate receptors is that they contain allosteric sites allowing certain agents to modulate their degree of desensitization/deactivation. In recent years, a series of compounds have been developed on the basis of their ability to prevent desensitization/deactivation of AMPA/kainate receptors. It is quite interesting that they were shown to exhibit neuroprotective and cognitive enhancement effects in various models showing impaired cognitive functions. Although the main explanation proposed for these effects is related to their action on neuronal AMPA/kainate receptors, it has also been suggested that a potentiation of glutamate-induced aerobic glycolysis in astrocytes occurs and participates in the beneficial effects of these drugs. Thus, enhancing the glycolytic response of astrocytes might represent a valuable therapeutic target for the future. A new dimension of the role of enhanced aerobic glycolysis in astrocytes was revealed recently. It has been known for a while that astrocytes communicate with each other through Ca2þ waves that propagate in a regenerative manner along the astrocytic network. The specific role of these traveling Ca2þ waves is still not entirely clear. But in parallel with Ca2þ, Naþ waves have been described. These Naþ waves arise from glutamate released by one astrocyte on elevation of intracellular Ca2þ and taken up by its neighbors. As a consequence of this Naþ influx, a metabolic wave of enhanced aerobic glycolysis will ensue among the astrocytic network. It is purported that such a mechanism could be used to coordinate metabolism and provide adequate energy substrate supply to activated neuronal ensembles, irrespective of their neurotransmitter phenotypes. If confirmed in vivo, such a phenomenon would have important implications not only for an understanding of neuroenergetics but also for the analysis of images obtained with brain imaging techniques that rely on metabolic signals.

Glycogen: Helping to Face Increased Energy Demands Glycogen has been viewed traditionally as an energy reserve maintained to face periods of low energy supply. This notion has been inherited from studies in peripheral organs such as the liver and muscle. However, the amount of glycogen contained in the brain is relatively small: approximately 10 times less than in muscle and 100 times less than in liver. Moreover, under physiological conditions, energy supply to the brain is maintained within narrow limits such that the idea of maintaining an emergency reserve for pathological situations appears unlikely. Rather, a series of

241

observations argues in favor of a role for glycogen as an energy buffer rather than an energy reserve in the central nervous system. Glycogen turnover was shown to be enhanced in vivo under sustained stimulation. In contrast, glycogen was found to accumulate in the brain under anesthesia. These observations suggest that glycogen utilization is tightly linked to neuronal activity. It is interesting that at the cellular level, glycogen is found almost exclusively in astrocytes. Since it is purported that the vast majority of energy expenditures are the consequence of neuronal activity, it may seem paradoxical that the sole energy reserve is located in a nonneuronal cell type. However, several lines of evidence indicate that astrocytes respond to neuronal signals and mobilize their glycogen stores to release a metabolite for use by neurons as an energy substrate. It has been shown that the brain, and in particular astrocytes, do not possess glucose-6-phosphatase and thus cannot release glucose as a product of glycogenolysis. Rather, it has been demonstrated that lactate constitutes the main substrate formed from glycosyl residues arising from glycogenolysis and released by astrocytes. Among the various neuroactive substances shown to induce glycogenolysis are the neurotransmitters noradrenaline and serotonin, the neuropeptide vasoactive intestinal peptide (VIP), and the neuromodulators adenosine and ATP. It has also been shown that elevated Kþoccurring as a consequence of neuronal activity can efficiently cause glycogenolysis in astrocytes. In addition to glycogenolysis, several of these neuroactive substances promote a delayed but massive resynthesis of glycogen in astrocytes that takes place over several hours. Such resynthesis is critically dependent on the expression of a key protein called protein targeting to glycogen. This protein serves as a scaffold that binds not only glycogen but also all the enzymes involved in the regulation of glycogen synthesis and degradation. It is also massively induced in astrocytes following treatment with noradrenaline or VIP (Figure 3). Although its purpose is still a matter of debate, this mechanism might be essential to replenish glycogen stores and thus ensure adequate levels in order to face subsequent periods of high activity. In the optic nerve, glial glycogen has been shown to play a critical role in maintaining neuronal activity, in particular by providing lactate to meet the energetic demands of the axon.

The Pentose Phosphate Pathway: Energy Fluxes and Neuroprotection ATP is not the only form of metabolic energy, as reducing power is needed in addition to ATP to energize

242 Glial Energy Metabolism: Overview NA

ATP Long-term effects

Astrocyte Adenylate cyclase

S

Short-term effects

cAMP

40

PTG GS activity

Resynthesis of glycogen

0 0.0

0.5

1.0

1.5

2.0

Time (h)

Glycogenolysis Glycogen (nmol / mg–1 prot.)

PTG mRNA

Min

20

Nucleus DNA

500 400 300

H

200 100 0 0

2

4

6

8

24

48

Time (h)

Figure 3 Regulation of glycogen metabolism by noradrenaline. Similar regulatory mechanisms are activated by vasoactive intestinal peptide and by adenosine. ATP, adenosine triphosphate; cAMP, cyclic adenosine monophosphate; GS, glutamine synthase; mRNA, messenger RNA; NA, nicotinamide adenine; Prot, protein; PTG, protein targeting to glycogen.

several important cellular processes. This reducing power is provided by the reduced form of nicotinamide adenine dinucleotide phosphate (NADPH). The processing of glucose through the pentose phosphate pathway produces NADPH (Figure 2). A critical cellular process for which NADPH is needed is the scavenging of reactive oxygen species (ROS). The superoxide radical anion (O2 ) hydrogen peroxide (H2O2), and the hydroxy radical (HO.) are three ROS generated by the transfer of single electrons to molecular oxygen as by-products of several physiological cellular processes. A considerable contribution to the generation of ROS is the oxidative metabolism of glucose taking place in the mitochondrial electron transfer chain associated with oxidative phosphorylation see below). Other ROS-generating reactions include the activities of monoamine oxidase, tyrosine hydroxylase, nitric oxide synthase, and the eicosanoid-forming enzymes lipoxygenases and cyclooxygenases. Reactive oxygen species are highly damaging to cells because they can cause DNA disruption and mutations, as well as activation of enzymatic cascades, including proteases and lipases, that can eventually lead to cell death. The coordinated activity of NADPH and glutathione is essential in protecting cells against ROS-mediated damage, or oxidative stress. Scavenging of ROS is ensured by the sequential action of superoxide dismutase (SOD) and glutathione peroxidase. Thus, two superoxide anions are converted by SOD into H2O2, still an ROS. Glutathione peroxidase converts H2O2 into H2O and O2

at the expense of reduced glutathione, which is regenerated by glutathione reductase in the presence of NADPH. Scavenging of ROS provides another example of metabolic cooperation between neurons and astrocytes. Glutathione is a tripeptide (g-L-glutamylL-cysteinylglycine (GSH)) synthesized through the concerted action of two enzymes, g-GluCys synthase, which combines glutamate and cysteine to yield the dipeptide g-Glu Cys, and glutathione synthase, which adds a glycine to the dipeptide to yield GSH. The glutathione content and reducing potential are considerably higher in astrocytes than in neurons; this fact, combined with the much higher oxidative activity of neurons than of astrocytes, makes neurons more vulnerable to oxidative stress as well as highly dependent on astrocytes for protection. Indeed, a cooperativity between astrocytes and neurons appears to exist for glutathione metabolism; astrocytes release GSH, which is cleaved by the ectoenzyme g-glutamyltransferase, releasing CysGly. The dipeptide is transported into neurons (note that neurons cannot take up GSH), providing two precursors for GSH synthesis. Glutamate, the third precursor of GSH, is also provided by astrocytes to neurons in the form of glutamine, from which glutamate is produced through the action of glutaminase. This astrocyte–neuron metabolic cooperation aimed at controlling the damaging effect of ROS generated by cellular activities may be critical for neuroprotection. Indeed, neurons, being highly oxidative cells

Glial Energy Metabolism: Overview

dependent on astrocytes for scavenging ROS, are at high risk for neurodegeneration. Indeed, dysfunction of such neuroprotective mechanisms has been described in certain neurodegenerative diseases, such as a familial form of amyotrophic lateral sclerosis, due to an SOD mutation. Evidence for a decrease in GSH content in the substantia nigra has also been described in Parkinson’s disease.

The TCA Cycle, Glutamate Metabolism, and Anaplerosis The main metabolic pathway for ATP production is the TCA cycle. While this pathway is active in both neurons and astrocytes, neurons, being richer than astrocytes in mitochondria where the TCA cycle occurs, are overall more oxidative than astrocytes. In addition to its central role in ATP production through its coupling to oxidative phosphorylation, the TCA cycle contributes to another critical aspect of astrocyte–neuron metabolic cooperation, related to glutamate metabolism. Synaptically released glutamate is removed rapidly from the extracellular space by a transportermediated reuptake system that is particularly efficient in astrocytes. This mechanism contributes in a crucial manner to the fidelity of glutamate-mediated neurotransmission. Indeed, glutamate levels in the extracellular space are low ( ∆y

NT H+

−60 mV

nH+

MA

nH+ H+ H+

H+

∆pH = ∆y

2H+

2H+

∆pH < ∆y Glu–

GABA, gly

ACh

nH+

nH+

nH+

Substrate charge Protons per cycle

+1 2

+1 2

0 1−2

−1 1−2

Net charge per cycle

+1

+1

+1 − 2

+2 − 3

Figure 1 Stoichiometry and bioenergetic dependence of the vesicular neurotransmitter transporters. A vacuolar type Hþ-ATPase on synaptic vesicles generates a proton electrochemical gradient. The gradient consists of a pH gradient and a transmembrane potential of approximately 60 mV (inside positive). The different transporter families have different dependencies on the two components of the gradient – VMATs and VAChT are most dependent on the pH gradient, VGAT can be driven by either the pH or electrical gradient, and VGLUTs are most dependent on the electrical gradient. For monoamine and acetylcholine transporters, the stoichiometries have been determined, but for VGAT and VGLUTs the specifics of energetic coupling remain to be determined. Dc, electrical gradient; ADP, adenosine diphosphate; ACh, acetylcholine; ATP, adenosine triphosphate; GABA, g-aminobutyric acid; Glu, glutamate; gly, glycerin; MA, monoamine; NT, neurotrophin; Pi, inorganic phosphate; VAChT, vesicular acetylcholine transporter; VGAT, vesicular GABA transporter; VGLUT, vesicular glutamate GABA transporter; VMAT, vesicular monoamine transporter.

the sequestration of the toxin inside vesicles, away from its primary site of action in mitochondria. The cloned vesicular monoamine transporter VMAT1 is predicted to have 12 transmembrane domains and is structurally related to the bacterial drug-resistance transporters. In addition to mediating the uptake of MPPþ, VMAT1 also transports dopamine, norepihephrine, epinephrine, and serotonin. VMAT1 is expressed by cells of the adrenal medulla, by neurons in sympathetic ganglia, and by other nonneural cells that release monoamines. In contrast, neuronal populations in the nervous system express the closely related VMAT2. The substrate specificity for the two isoforms is similar, and the apparent affinities for the monoamines are in the low micromolar range for both transporters, though VMAT2 has a slightly higher apparent affinity for all monoamines. In addition, only VMAT2 appears able to transport histamine at physiological concentrations, consistent with its expression by histamine-releasing mast cells. The high affinity of the VMATs may reflect a need to keep cytosolic concentrations of the relatively toxic monamines low. Transport by the VMATs involves the exchange of two lumenal protons for one protonated molecule of transmitter carrying a net single positive charge and,

thus, the outward movement of one net positive charge. For a vesicular pH gradient of  1.5 pH units and membrane potential of  60 mV, this predicts the accumulation of transmitter inside vesicles at concentrations 104–105-fold greater than those in cytoplasm. This is consistent with the micromolar and near-molar catecholamine concentrations measured in the cytosol and vesicles, respectively. Because osmotic forces limit the free transmitter concentration to  150 mM, a fraction of the measured concentration, it has been suggested that the formation of macromolecular complexes consisting of a proteinaceous core, ATP, and the catecholamines reduces the effective osmolarity inside the vesicle. The broad specificity of these transporters may be of particular importance in epinephrine-releasing cells. In these cells, VMATS transport dopamine into secretory vesicles, where the transmitter is converted to norepinephrine by the action of the dopamine b-hydroxylase, an enzyme restricted to the vesicle lumen. Norepinephrine then exits the vesicle, most likely through the action of VMATs, into the cytoplasm where it is converted to epinephrine by the action of phenylethanol-amine-N-methyltransferase. Epinephrine is then transported back into vesicle,

256 Vesicular Neurotransmitter Transporters

again presumably by VMATs. Although these last two steps probably occur through heterologous exchange (e.g., lumenal norepinephrine is exchanged for cytoplasmic epinephrine with no net proton movement), epinephrine released by exocytosis passes through VMATs three times. Two well-characterized inhibitors act on the VMATs: reserpine and tetrabenazine. Both VMATs are irreversibly inhibited by reserpine, which initially showed clinical promise as an antihypertensive medication. The depressive effects of reserpine led to its more limited use but helped to formulate the original monoamine hypothesis of affective disorders. Reserpine appears to interact with the transporters near the site of substrate recognition. Tetrabenazine reversibly inhibits VMAT2, but is markedly less effective as an inhibitor of VMAT1. It has a limited clinical use in the treatment of some movement disorders. Amphetamines and 3,4-methylenedioxy-N-methylamphetamine (MDMA, or extasy) act as weak bases to reduce the vesicular DpH. Loss of this driving force for the vesicular transporters leads to the efflux of the lumenal contents into the cytoplasm and subsequent release through reversal of the plasma membrane transporters. The specificity of their effects (amphetamines lead to dopamine and, to a lesser extent, norepinephrine release, whereas MDMA leads to serotonin release) is thought to be due to selective uptake by the plasma membrane transporters. The association of polymorphisms in the VMAT1 gene with an increased risk for bipolar disorder also suggests a link between vesicular monoamine transport and molecular psychopathology. A major mechanism for the regulation of vesicular neurotransmitter transport appears to involve changes in protein trafficking. VMATs undergo phosphorylation by casein kinase, and this posttranslational modification influences their retrieval from maturing large dense-core vesicles (LDCVs). Because sorting to LDCVs versus synaptic vesicles determines the site and mode of transmitter release, the regulation of transporter trafficking has great potential to influence signaling. In particular, differential trafficking of VMAT2 in midbrain neurons can lead to the release of dopamine from exocytosis at their cell bodies and dendrites, if trafficked to LDCVs, or axon terminals, if trafficked to synaptic vesicles. Regulation of VMAT activity by the heteromeric G-protein Gao2 has also been demonstrated. It appears that intralumenal monoamines activate the G-protein and downregulate uptake. This may be a mechanism to assure consistency in quantal size and may also serve to limit the efflux of the neurotransmitter through reverse transport once a vesicle has been filled.

The Vesicular Acetylcholine Transporter Is Structurally Related to the Vesicular Monoamine Transporters The electric ray organ of Torpedo californica was an early source of synaptic vesicles. Subsequent to the determination that these vesicles are filled with acetylcholine, a specific activity for acetylcholine transport was characterized. The molecular identification of the transporter, however, required genetic studies in Caenorhabditis elegans. A screen for nematode mutants resistant to aldicarb, an inhibitor of acetylcholinesterase, identified unc17, a gene encoding a protein closely related to the VMATs. Based on the similarity to the VMATs, it was predicted that unc17 functions as an acetylcholine transporter and that the resistance phenotype results from the reduced release of acetylcholine. The binding of vesamicol, a known inhibitor of vesicular acetylcholine transport, was also found to be absent in membrane fractions from the unc17 animals. Subsequent studies on the mammalian ortholog demonstrated that the protein is indeed a vesicular acetylcholine transporter. In vertebrates, the vesicular acetylcholine transporter VAChT is expressed in cholinergic neurons in the central, peripheral, and autonomic nervous systems, including the basal forebrain neurons, lower motor neurons, and parasympathetic neurons. In an interesting genomic structure, the VAChT gene lies within the first intron of the gene encoding cholineacetyltransferase, the enzyme that catalyzes the synthesis of acetylcholine from choline and acetyl CoA. This structure, which suggests a spatial component to transcriptional regulation of factors conferring cholinergic characteristics to a neuron, is conserved from C. elegans to humans. Like VMATs, VAChT recognizes a cationic substrate and depends primarily on DpH. VAChT has been postulated to have a similar stoichiometry of one cytosolic AChþ exchanged for two luminal Hþ. In contrast to VMATs, which have substrate affinities in the low micromolar range, the mammalian VAChT exhibits an apparent affinity of approximately 1 mM. The higher affinity of VMATs may reflect a need to keep cytoplasmic levels of potentially toxic monoamine transmitters low. The measured quantal size at the neuromuscular junction ( 10 000 molecules) is approximately fivefold larger than that expected for the 150 mM vesicular acetylcholine concentration predicted by osmotic limits. This suggests that, as with monoamine storage, mechanisms exist to limit the osmotic effects of acetylcholine in synaptic vesicles.

Vesicular Neurotransmitter Transporters 257

The vesicular acetylcholine transport can be inhibited by vesamicol and several related compounds. Vesamicol competitively inhibits transport by binding to a cytoplasmic domain on VAChT with a Kd of approximately 5 nM. Vesamicol binding can be used to estimate transporter number, but neither vesamicol nor its analogs are currently used clinically.

Vesicular GABA and Glycine Transport Activities Are Mediated by a Single Protein The primary excitatory and inhibitory neurotransmitters in the mammalian brain are the amino acids glutamate and GABA. There are no sources of readily purified glutamatergic or GABAergic vesicles, as there are for monoaminergic and cholinergic vesicles (e.g., chromaffin granules and electric organ synaptic vesicles, respectively). Early biochemical studies defining vesicular transport systems for these neurotransmitters were, therefore, carried out on mixed synaptic vesicles purified from either the bovine or rodent brain. As for VMATs and VAChT, molecular characterization of the vesicular glutamate transporters (VGLUTs) and vesicular GABA transporter (VGAT) relied on additional information from molecular genetic studies. The molecular identity of the vesicular VGAT was determined from a screen for C. elegans genes involved in GABAergic synaptic transmission. One gene identified in the screen, unc47, encodes a multitransmembrane domain protein that is localized to synaptic vesicles in GABAergic neurons. The functional characterization of a vertebrate ortholog confirmed that the protein mediated vesicular GABA transport. As predicted from studies with purified synaptic vesicles, VGAT mediates the uptake of GABA with an apparent affinity in the low millimolar range and uptake can be driven by DpH or Dc. VGAT also recognizes glycine as a substrate, but with a lower affinity, and is therefore also referred to as for vesicular inhibitory amino acid transporter (VIAAT). VGAT shows no sequence similarity to VMATs or VAChT. Rather, it belongs to a large family of amino acid transporters. This family includes the proteins responsible for the transport activities biochemically defined as amino acid transport system N and system A. The latter is responsible for much of the active amino acid uptake by mammalian cells. Characterization of the function and cellular and subcellular localizations of these transporters suggests that the system N transporters and system A transporters mediate glial glutamine release and neuronal uptake, respectively, in the glutamine–glutamate cycle. This intercellular metabolic pathway is involved in the

recycling of synaptically released glutamate and, to a lesser extent, GABA. VGAT is expressed in GABAergic and glycinergic neurons as well as in the pancreas and has been reported to be expressed in the glial cells of the pineal gland. The targeted disruption of the VGAT gene in mice leads to a marked reduction in the synaptic release of GABA and glycine. The loss of the transporter leads to embryonic lethality, failures in gut withdrawal, and formation of a cleft palate. It is suggested that these defects are secondary to the loss of inhibitory neurotransmission and resulting paralysis. Vesicular GABA transport can be competitively inhibited by amino acids, including glycine and b-alanine. Transport can also be competitively inhibited by g-vinyl GABA, a derivative of GABA. g-Vinyl GABA has been used in the clinical treatment of epilepsy and is known to inhibit GABA transaminase, an enzyme that metabolizes GABA. The mode of action for g-vinyl GABA as an antiepileptic drug is thought to be through its effects on GABA transaminase, but the inhibition of vesicular GABA transport could have an effect by increasing the nonvesicular release of GABA.

Vesicular Glutamate Transporters Define Glutamatergic Neurons Glutamate is the principal excitatory neurotransmitter in the mammalian central nervous system. Three unique vesicular glutamate transporters (VGLUT1, -2, and -3) have been identified. VGLUT1 was initially isolated in a screen for mRNA transcripts upregulated in cultured neurons by subtoxic doses of the excitotoxin N-methyl-D-aspartate (NMDA). Structural similarity to a class of proteins characterized as inorganic phosphate transporters led to an initial impression that the protein was a Naþ-dependent phosphate transporter and to a temporary designation of the protein as brain-specific Naþ dependent inorganic phosphate transporter (BNPi). Although subsequent studies have demonstrated that the primary function of the protein is synaptic vesicle glutamate transport, some controversy remains regarding the role of the protein in phosphate transport. The expression of VGLUT1 is limited to a subset of glutamatergic neurons in the brain. The highly homologous protein VGLUT2, also initially characterized as a phosphate transporter and named differentiationassociated Naþ-dependent inorganic phosphate transporter (DNPi), has a nearly complementary pattern of expression with one of the two proteins present in all established glutamatergic neurons. In general, VGLUT1 predominates in the neocortex and cerebellar cortex, whereas VGLUT2 predominates in the

258 Vesicular Neurotransmitter Transporters

brain stem nuclei, thalamic nuclei, and cerebellar deep nuclei. The septal nuclei, nuclei of the diagonal band, and hypothalamus also express VGLUT2. Although all cortical layers express VGLUT1, layer IV of frontal and parietal cortex and layers IV and VI of temporal cortex also express VGLUT2. Conversely, VGLUT2 predominates in the thalamus, but certain thalamic nuclei such as the medial habenula express VGLUT1. In the hippocampus, dentate gyrus granule cells express only VGLUT1, whereas pyramidal neurons from CA1 through CA3 express VGLUT1 as well as lower levels of VGLUT2. In the amygdala, the medial and central nuclei express VGLUT2, and the lateral and basolateral nuclei express VGLUT1. The third vesicular glutamate transporter, VGLUT3, is expressed in neurons not classically considered glutamatergic. Immunohistochemical and in situ studies indicate VGLUT3 is expressed in GABAergic, serotonergic, dopaminergic, and cholinergic neurons as well as astrocytes. Recent findings suggest that VGLUT1 and VGLUT2 are expressed in astrocytes as well, but these studies are in isolated and cultured cells, not in intact tissue. Glutamate uptake by the VGLUTs depends primarily on Dc rather than DpH, with apparent affinities in the low millimolar range. Interestingly, the VGLUTs do not appear to recognize aspartate. This is consistent with studies of synaptic vesicles and suggests that aspartate is not readily accumulated in synaptic vesicles through an active transport system. Vesicular glutamate transport shows a biphasic dependence on chloride with an optimum at 2–10 mM, and recent studies with purified VGLUT1 suggest that glutamate transport requires Cl . Like VMAT transport, VGLUT transport also appears to be regulated by Gao2; specifically, Gao2 reduces the chloride dependence of transport. The role of chloride in vesicular glutamate transport is further complicated by the finding that the expression of VGLUT1 appears to increase the vesicular Cl conductance. Although the VGLUTs exhibit similar transport activities, they are expressed in cells with very different properties. For example, compared to VGLUT1containing neurons, VGLUT2 neurons, in general, have a lower firing rate and a higher probability of release. The three isoforms also appear to have different subcellular localizations that could influence glutamate-release characteristics. Most notably, VGLUT3 is expressed in vesicular structures within the dendrites of some neurons in the hippocampus and striatum, suggesting a role in retrograde signaling. Further, two polyproline domains, which are present in the C-terminal cytoplasmic tail of VGLUT1 (but not VGLUT2 or -3), mediate interactions with the endocytic protein endophilin. This interaction may

regulate synaptic vesicle recycling and the mode in which VGLUT1 is internalized. Recent studies indicate that VGLUT1 and VGLUT2 are present in multiple copies on synaptic vesicles. Correcting for the number of vesicles containing the transporters, it is estimated that there are  9 copies of VGLUT1 and  14 copies of VGLUT2 in the synaptic vesicles in which they are expressed. Although not directly demonstrated, a similar stoichiometry probably exists for the other vesicular neurotransmitter transporters. Several compounds that inhibit vesicular glutamate transport have been identified. These include the dyes Evans blue and rose Bengal. In addition, the stilbene derivative 4,40 -diisothiocyanatostilbene-2,20 -disulfonic acid (DIDS), a compound commonly used as an inhibitor of anion channels, inhibits vesicular glutamate transport. Most known inhibitors have a limited utility because they are membrane impermeant, with the exception of rose Bengal. No inhibitors unique for specific isoforms have been identified.

Neuromodulators Are Also Stored in Vesicles and Released through Exocytosis Synaptic vesicles mediate the release of neuropeptides and small molecules other than the classic neurotransmitters. Neuropeptides enter the lumen of the secretory pathway in the endoplasmic reticulum through cotranslational translocation and are sorted to the secretory vesicle pathway, where they undergo processing to form the biologically active species. After release, it is believed that neuropeptides are degraded and not repackaged. Of the synaptically released small-molecule neuromodulators, zinc and ATP are the best characterized. NMDA and GABA receptors contain binding sites for zinc, and zinc exerts a direct effect on excitatory and inhibitory neurotransmission. ATP activates both ionotropic and G-protein-coupled receptors. As with the classical neurotransmitters, the exocytotic release of these compounds requires transport into synaptic vesicles. The multitransmembrane domain protein ZnT3 has been implicated in zinc uptake by synaptic vesicles. ZnT3 belongs to a family of zinc transporters and localizes to synaptic vesicles. Mice deficient in ZnT3 show a loss of zinc staining from hippocampal neurons, and the expression of ZnT3 in PC12 cells increases vesicular zinc staining. Although ZnT3 transport has not been directly demonstrated, these findings strongly support a role for ZnT3 in synaptic vesicle zinc transport. The phenotype of ZnT3-deficient mice is mild, with the most striking abnormality being an increased susceptibility to seizures. Chromaffin granules, platelet dense-core vesicles, and synaptic vesicles contain concentrations of ATP

Vesicular Neurotransmitter Transporters 259

many fold higher than cytosolic concentrations, suggesting active vesicular uptake. ATP transport has been demonstrated in chromaffin granules and synaptic vesicles, and the process appears to depend on DmHþ. It has generally been assumed that ATP is costored only with monoamines and acetylcholine, as an anion to balance to cationic charge of those transmitters. However, the extent of ATP storage and release by different neuronal populations remains unknown, and the proteins responsible for ATP uptake by secretory vesicles have not been identified. One of the first synaptic vesicle membrane proteins to be identified was SV2. This protein is a multitransmembrane domain protein with limited structural similarity to the VMATs and VAChT. The protein was identified in synaptic vesicles from T. californica, and three isoforms (A, B, and C) have been identified in mammals. The expression patterns of the vertebrate proteins show partial overlap and together cover essentially all neurons. Targeted disruption of isoform A and both isoforms A and B demonstrates that these proteins are crucial for normal brain function in mice. Although a specific biochemical function was not been defined by these studies, it has been suggested that SV2 might be involved in regulating the size of the readily releaseable pool of vesicles or Ca2þ homeostasis at the nerve terminal. SV2 is thought to be the binding site for botulinum A toxin entry into neurons and to be a binding site for the antiseizure medication levetiracetam. It has also been suggested that the long sugar chains on SV2 serve as a smart-gel that regulates the release of neurotransmitters for the fused vesicle by limiting diffusion (Table 1).

Summary The vesicular uptake of neurotransmitters is a requirement for the quantal release of neurotransmitters. Earlier biochemical studies that identified four primary activities have been complimented by the molecular identification of unique proteins that mediate vesicular neurotransmitter transport activities. The cloned transporters can be divided into three structural families – the VMATs and VAChT; VGAT; and the VGLUTs. All the transporters are driven by the proton electrochemical gradient generated by the vacuolar Hþ-ATPase. Multiple copies of a given transporters are present on a single synaptic vesicle, suggesting that expression levels and regulated trafficking may play roles in modulating synaptic vesicle filling. A greater number of transporters in the vesicle increases the rate of filling and, if there is a significant leak, also leads to an

increase in the steady-state concentration of transmitter achieved. Although the potential for the modulation of the vesicular neurotransmitter transporter activities has been established, the extent to which neurotransmitter accumulation and synaptic transmission are regulated by vesicular transporter density, G-protein-mediated modulation, and the co-storage of other small molecules such as zinc and ATP remains to be determined. See also: Glutamate; Synaptic Vesicles.

Further Reading Bellocchio EE, Reimer RJ, Fremeau RT Jr., and Edwards RH (2000) Uptake of glutamate into synaptic vesicles by an inorganic phosphate transporter. Science 289(5481): 957–960. Brunk I, Holtje M, von Jagow B, et al. (2006) Regulation of vesicular monoamine and glutamate transporters by vesicleassociated trimeric G proteins: New jobs for long-known signal transduction molecules. Handbook of Experimental Pharmacology 175: 305–325. Carlson SS, Wagner JA, and Kelly RB (1978) Purification of synaptic vesicles from elasmobranch electric organ and the use of biophysical criteria to demonstrate purity. Biochemistry 17(7): 1188–1199. Fremeau RT Jr., Voglmaier S, Seal RP, et al. (2004) VGLUTs define subsets of excitatory neurons and suggest novel roles for glutamate. Trends in Neuroscience 27(2): 98–103. Johnson RG Jr. (1988) Accumulation of biological amines into chromaffin granules: A model for hormone and neurotransmitter transport. Physiological Reviews 68(1): 232–307. Lein ES, Hawrylycz MJ, Ao N, et al. (2007) Genome-wide atlas of gene expression in the adult mouse brain. Nature 445(7124): 168–176. Liu Y, Peter D, Roghani A, et al. (1992) A cDNA that suppresses MPPþ toxicity encodes a vesicular amine transporter. Cell 70(4): 539–551. McIntire SL, Reimer RJ, Schuske K, et al. (1997) Identification and characterization of the vesicular GABA transporter. Nature 389(6653): 870–876. Nishi T and Forgac M (2002) The vacuolar (Hþ)-ATPases – nature’s most versatile proton pumps. Nature Reviews Molecular Cell Biology 3(2): 94–103. Reimer RJ and Edwards RH (2004) Organic anion transport is the primary function of the SLC17/type I phosphate transporter family. Pflugers Archiv 447(5): 629–635. Stobrawa SM, Breiderhoff T, Takamori S, et al. (2001) Disruption of ClC-3, a chloride channel expressed on synaptic vesicles, leads to a loss of the hippocampus. Neuron 29(1): 185–196. Takamori S, Holt M, Stenius K, et al. (2006) Molecular anatomy of a trafficking organelle. Cell 127(4): 831–846. Voglmaier SM, Kam K, Yang H, et al. (2006) Distinct endocytic pathways control the rate and extent of synaptic vesicle protein recycling. Neuron 51(1): 71–84. Wojcik SM, Katsurabayashi S, Guillemin I, et al. (2006) A shared vesicular carrier allows synaptic corelease of GABA and glycine. Neuron 50(4): 575–587.

AMPA Receptors: Molecular Biology and Pharmacology S M Dravid, H Yuan, and S F Traynelis, Emory University, Atlanta, GA, USA Published by Elsevier Ltd.

Introduction Glutamate, one of the fundamental amino acid building blocks of proteins, is also a major excitatory neurotransmitter in the central nervous system (CNS). Neurons synthesize and package glutamate into presynaptic vesicles for release into the postsynaptic cleft. Synaptically released glutamate that diffuses across the 30 nm distance encounters a series of transmembrane postsynaptic proteins that comprise the glutamate receptor family. One class of glutamate receptors, metabotropic glutamate receptors, comprises transmembrane proteins that have extracellular clamshell-like domains that bind glutamate. When activated by glutamate binding, these G-protein-coupled receptors shift intracellular concentrations of signaling molecules to control a diverse set of cell properties. A second class of glutamate receptors, ionotropic glutamate receptors, comprises transmembrane proteins that contain an ion conduction path through the plasma membrane, as well as an array of clamshell-like extracellular ligand-binding domains, some of which bind to glutamate. Mammalian ionotropic glutamate receptors are ligand-gated ion channels encoded by 18 genes, and are subdivided into four major families on the basis of agonist pharmacology and sequence homology. These four receptor classes are known as amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA), kainate, N-methyl-D-aspartate (NMDA), and d receptors. This article focuses on the structure and function of AMPA receptors. The AMPA receptor family is composed of four genes encoding the GluR1–GluR4 subunits (sometimes called GluRA–GluRD). In humans the chromosomal location of GluR1, GluR2, GluR3, and GluR4 encoding genes is 5q33, 4q32–33, Xq25–26, and 11q22–23, respectively. AMPA receptors were the first class of glutamate receptor cloned by screening a rat brain cDNA library for expression of kainate-activated ion channels in Xenopus laevis oocytes. After initial identification of GluR1, GluR2–GluR4 were rapidly identified by homology screening. There is about 70% sequence homology among different AMPA receptor subunits. A great deal of information now exists about AMPA receptor structure and function, and it could be argued that more is known about the structure of AMPA receptors than any other class of glutamate receptor.

260

Expression of AMPA Receptors AMPA receptors are abundant and widely distributed in the central nervous system. Hippocampus, outer layer of cortex, basal ganglia, olfactory regions, lateral septum, and amygdala of the CNS are all enriched with GluR1, GluR2, and GluR3 subunits. In contrast, GluR4 expression is lower in many regions of the CNS except cerebellum, thalamus, and brain stem, where the expression is high. Immunoprecipitation studies have shown that the pyramidal cells of the hippocampus expressed AMPA receptors composed of GluR2 receptor in complex with either GluR1 or GluR3 subunits. GluR1 homomeric receptors, which have unique ion permeation properties, are thought to be expressed in select neuronal populations. The expression of AMPA receptors is developmentally regulated. The GluR2 subunit appears as early as embryonic day 16 in rats whereas other receptors are upregulated later during the development. The GluR2 subunit can also be selectively altered during synaptic plasticity as well as during CNS injury, such as global ischemia. These changes in receptor subunit composition are known to change functional receptor properties. AMPA receptors are present both postsynaptically and presynaptically. AMPA receptors are present on the synaptic membrane; however, 60–70% of AMPA receptors are present intracellularly. Glial cells also express AMPA receptors, which appear to be involved in glutamate-induced cell death. Activation of glial AMPA receptors also leads to release of ATP or nitric oxide.

Topology and Assembly of AMPA Receptor Subunits All of the ionotropic glutamate receptors share a common topology, which consists of an extracellular N-terminal domain, a ligand-binding domain, three transmembrane domains (M1, M3, and M4), a cytoplasm-facing reentrant membrane loop (M2), and an intracellular C-terminal domain (Figure 1(a)). AMPA receptors are composed of approximately 900 amino acids and have a molecular mass of 105 kDa. The location of the N-terminus and Cterminus was first deduced by use of specific antibodies. Because the N- and C-terminal regions were located on the opposite ends of the polypeptide chain, it was proposed that AMPA receptors had an odd number of membrane-spanning domains. Further studies delineated the membrane topology, and showed that the M2 segment is a reentrant loop.

AMPA Receptors: Molecular Biology and Pharmacology 261 Amino terminal domain

S2

N-terminal

S1

Domain 2

Domain 1

Domain 1

Glutamate

C

Flip/flop

J

D

K

R/G site I

Q/R site

Extracellular

F

1

3

4

2

Domain 2

Intracellular PKC/CaMKII site a

PKA site

C-terminal H

To TM123 Linker From TM123

COOH b

Figure 1 Structure of the AMPA receptor subunit. (a) Transmembrane topology of the AMPA receptor, indicating flip/flop site, Q/R site, R/G site, and phosphorylation sites (PKA, protein kinase A; PKC, protein kinase C; CaMKII, Ca2þ/calmodulin-dependent protein kinase II). (b) Crystal structure of the ligand-binding domain of the GluR2 subunit, with glutamate bound in the cleft formed by clamshells (Protein Data Bank code 1FTJ). The lime-colored structure is domain 1 and the violet-colored structure is domain 2 (TM, transmembrane domain).

This transmembrane topology shares a parallel organization to potassium channels, which are now considered a model for AMPA receptor membrane domain structure. The pore diameter of AMPA receptors is about 0.8 nm and, in contrast to potassium channels, permits the entry of Naþ and, for some subunit combinations, Ca2þ. The N-terminal domain in AMPA represents up to 45% of the mature polypeptide, but its function is poorly understood. Hypothesized functions of this domain include receptor assembly, allosteric modulation of the ion channel (similar to NMDA receptors), and binding of a second ligand. Contradictory to other subunits, the GluR4 subunit can form normally functioning homomeric channels even in the absence of the N-terminal domain. The semiautonomous ligand-binding domain has been studied in detail using X-ray crystallography. The GluR2 agonist-binding domain is composed of two discontinuous peptide segments of approximately 150 amino acid residues. The first segment (S1) is adjacent and N-terminal to the M1 domain, whereas the second segment (S2) is located between M3 and M4 domains (Figure 1(a)). The agonistbinding domains of the AMPA receptors share similarities in sequence and structural arrangement with the ligand-binding site of several bacterial periplasmic amino acid-binding proteins. Like potassium channels, AMPA receptors assemble as tetramers. Studies in recombinant and native receptors suggest that AMPA receptors assemble as

dimer-of-dimers in a two-step manner. First, the monomers interact through the N-terminal domain to form dimers. Next, the dimers combine via the membrane domains to form the tetramer.

AMPA Receptor Function A remarkable step forward in understanding the AMPA receptor structure and function occurred in 1995 when a water-soluble mini-receptor that included only the agonist-binding core was described. This was a fusion protein consisting of the S1 and S2 domains of GluR4 joined together via a short hydrophilic linker peptide that replaces the membranespanning regions. The engineered agonist-binding domain functionally reproduced the AMPA-binding properties of the GluR4 receptor. This concept paved the way for the subsequent production of soluble agonist-binding core and later generation of crystals for X-ray diffraction, which ultimately could be produced by careful refolding of the denatured ligand-binding core. The first crystal structure of the GluR2 ligandbinding domain complexed with kainate revealed a bilobed clamshell-like shape, with agonist bound deep in the cleft formed by the two lobes. Subsequent descriptions of crystal structures of the nonliganded form (apo state) as well as forms complexed to a variety of ligands were obtained. These studies revealed that the clamshell was geometrically opened widest in the apo state. In the glutamate-bound state, the clamshell was 21 more closed than in the apo

262 AMPA Receptors: Molecular Biology and Pharmacology

state (Figures 1 and 2). Competitive antagonists of the receptor stabilized the open-cleft state. Glutamate makes a number of hydrogen-bonded contacts within the binding pocket with both upper (D1) and lower (D2) domains. The upper domain of the clamshell-like structure is formed by segment S1 and the C-terminal portion of segment S2. The N-terminus of S2 forms the lower D2 domain. It has been proposed that when glutamate first encounters the ligand-binding pocket, it initially docks or interacts with the D1 domain, which then promotes the rotation of the D2 domain toward D1 and induces closure of the clamshell. The closed conformation is stabilized by glutamate, itself forming a cross-domain bridge, as well as a number of other hydrogen bonds that form between the domains during glutamate binding. Water appears to persist inside of the agonist binding pocket, particularly for some agonists that interact directly with it. This closure of the cleft within the isolated ligand-binding domain is considered to be analogous to domain closure in the full-length native receptors.

Domain closure has been hypothesized to pull the linker that connects the S1 and S2 domains. In native receptors, this pull or strain is considered to be the force responsible for rearrangement of the pore-forming membrane domains, leading to the opening of the transmembrane conduction path. Thus, each subunit can bind glutamate and undergo conformational changes that contribute to dilatation of the pore. In agreement with this view, single-channel analysis of intact receptor shows at least three conductance states that have been proposed to correspond to two, three, or four liganded subunits within a receptor complex. Pore conductance is therefore conceptually related to the fraction of subunit occupancy and activation within each receptor complex. This suggests that subunits can make incremental contributions to pore opening, gradually shifting the unitary conductance through the channel to higher levels as more subunits become activated. The observed concentration dependence of conductance levels supports this idea. This view of fourfold rotational symmetry is also supported by the similarities of the pore-forming

AMPA

AMPA

Domain 1

AMPA

Domain 2

Figure 2 Conformational changes induced by agonist binding to the AMPA receptor. Crystal structure of the ligand-binding domain of the GluR2 subunit shown in nonliganded (apo –state; Protein Data Bank code 1FTO) and liganded (AMPA bound state; Protein Data Bank code 1FTM) states. Agonist binding in the cleft leads to closure of the clamshell (top).

AMPA Receptors: Molecular Biology and Pharmacology 263

AMPA receptor domain to potassium channels. However, several other features of AMPA receptors point to a twofold rotational symmetry for the tetrameric receptor complex. First, the glutamate-binding domains crystallize as dimers. Second, the D1–D1 dimer interface plays a role in channel activation by forming a structural scaffold that leads to the movement of D2. However, excessive tension can trigger rearrangement of the D1–D1 interface, leading to AMPA receptor desensitization. Mutations that affect desensitization lie on the interface. Additionally the AMPA receptor desensitization inhibitor cyclothiazide (see below) also acts on the dimer interface. Third, the reactivity of cysteine-modifying reagents on cysteine residues inserted into the M3 domain by site-directed mutagenesis fits well with a twofold symmetry rather than a fourfold symmetry. However, additional structural information about the full AMPA receptor complex would be needed to ascertain receptor symmetry. Partial agonists are typically considered as ligands that induce a response at the maximally effective concentrations, which is lower than that of the endogenous ligand glutamate. The mechanism of action of partial agonists has been studied in AMPA receptors using a series of 5-substituted willardiines, which show lower efficacy compared to glutamate. The 5-substituted willardiines differ in only a single atom at the same position in the molecule. Structural and functional studies suggest that the degree of domain closure of the agonist binding cleft forms the basis for the partial agonist action at the AMPA receptor. Specifically, a combination of crystallographic and functional data show that the degree of domain closure of the clamshell is correlated with the efficacy of the agonist at an individual subunit such that agonists that induce less domain closure appear less effective in opening the channel. However, other conformational rearrangements of the agonist binding domain can also impact agonist efficacy.

Posttranscriptional Modification of AMPA Receptors Alternative RNA Splicing

All four AMPA subunits undergo alternative RNA splicing in the C-terminal half of the M3–M4 loop, leading to so called flip/flop splice variants. The locations of flip/flop splicing for GluR1, GluR2, GluR3, and GluR4 are 742–793, 736–787, 740–791, and 737–788, respectively. In rats expression of the flip variant predominates up to postnatal day 8, but in adults both forms are expressed to a similar extent in many regions. The flip splice variant endows receptors with a diminished form of desensitization and a

faster rate of recovery from desensitization as compared to the flop splice variant. GluR4-flop has the fastest desensitization ( 7.0 allowed the measurement of dynamic changes to the surface expression of proteins. Because intracellular compartments are more acidic than the extracellular environment, super eclipitic phluorin (SEP) fluorescence is suppressed until the protein is surface expressed. This technology was used to show that, in response to NMDAR activation, endocytosis of synaptic GluR2 is preceded by the removal of extrasynaptic receptors (Figure 2). An obvious reason for this is that the endocytosis machinery is located at extrasynaptic sites. Quantum dot-tracking experiments show that the proportion of mobile GluR2 increases following neuronal activity, representative of increased lateral diffusion from synaptic anchors. This suggests that, rather than the regulation of endocytosis itself, the limiting step in generating LTD through internalization is the release of AMPARs from the PSD. Protein Interacting with C Kinase 1 Releases AMPARs from the Synapse

Protein interacting with C kinase (PICK1) interacts with a number of neuronal receptors, channels, and transporters. A single PDZ domain enables an interaction with short AMPARs leading to LTD. This occurs because PICK1 competes for the GRIP/ABP binding site on GluR2, uncoupling the AMPARs

from the synaptic scaffold. This is controlled by protein kinase C (PKC) phosphorylation of GluR2 at Ser880, which prevents GRIP/ABP but not PICK1 binding. Phosphorylation is facilitated by PICK1 itself, which binds PKCa to target the kinase to AMPARs. PKCa binding removes an intramolecular interaction, exposing the BAR domain of PICK1. A second phosphorylation switch with the same function has recently been found at Tyr876, and this residue must be Src phosphorylated in response to drug treatment for internalization to occur. NSF hydrolysis of adenosine triphosphate (ATP) disrupts the PICK1–GluR2 interaction in complexes containing a-SNAP. This stabilizes the AMPARs at the surface. b-SNAP, however, prevents this dissociation and causes receptor internalization. Thus, in addition to actively inserting AMPARs at the synapse, NSF activity prevents their internalization. The presence of a BAR domain in PICK1 suggests this protein can recruit AMPARs to clathrin-coated vesicles (CCVs) in constitutive endocytosis in which AP2 does not seem to be involved. Ca2þ binding to PICK1 increases the affinity of the GluR2 interaction. This has obvious implications for LTD, in which Ca2þ concentration is elevated in spines. The increase in affinity may be another switch in the balance between constitutive cycling and longlasting endocytosis of AMPARs. PICK1 is not the only Ca2þ-sensing molecule implicated in the endocytosis of AMPARs during LTD. The neuronal calcium-sensor hippocalcin binds AP2 in a Ca2þdependent manner, and the Ca2þ-sensing region of hippocalcin is required for the generation of LTD. It is thought that Ca2þ binding, which exposes a myristyl tail in hippocalcin, recruits AP2 to the membrane, enabling AMPAR sorting into CCVs. Hippocalcin also complexes with transferrin receptors but in a Ca2þ-independent manner, suggesting the function of this protein is modified for a specific role in AMPAR endocytosis. Degradation of Synaptic Scaffolds Accompanies AMPAR Internalization

PSD-95 is ubiquitinated following synaptic NMDA activation. Truncated PSD-95 mutants lacking the ubiquitination motif prevent NMDA-induced GluR2 internalization. One function of ubiquitination is as a signal for proteasomal protein degradation, suggesting that the synaptic scaffold may have to be dismantled for AMPAR release. Although this has not yet been shown for the proteasome pathway, calpain cleavage of PSD-95, SAP-97, and GRIP1 has been shown. AMPARs are themselves cleaved at the C-terminus by calpain, allowing their release from the PSD.

Co-localization with EEA1 (%)

AMPA Receptor Cell Biology/Trafficking

60

GluR1 GluR2 GluR3

50 40 30

Resting

NMDA 7

10 0

5

10

15

20

25

GluR1

30

1

Co-localization with Syntaxin 13 (%)

60 40

2

30

GluR3

20

40 35 30 25 20 15 10 5 0

NEEP21 Calpain/ proteasome

6

10 0

Co-localization with Lamp1 (%)

GluR2

50

0

a

After NMDA

b

20

0

273

5

10

15

20

25

3

30

E L

R 4

0

10

20 30 40 Time (min)

50

PSD-95

5

All conditions NMDA AMPA

60

c

Figure 2 AMPAR internalization and recycling: (a) experiments using recombinant HA-tagged AMPAR subunits; (b) virally expressed pHluorin-GluR2 labeling resting, with NMDA, and after NMDA; (c) main steps in AMPAR recycling. In (a), experiments using recombinant HA-tagged AMPAR subunits demonstrate that they are differentially sorted on internalization. All HA-tagged AMPAR subunits are internalized following the application of 50 mM NMDA for 8 min. All subunits colocalize with the early endosome marker EEA1 after 10 min, but only GluR1 continues to accumulate in this compartment up to 30 min following NMDA treatment. Both GluR2 and GluR3 pass through the syntaxin 13-positive recycling endosome and accumulate in the Lamp1-positive late endosome prior to degradation, whereas GluR1 remains in the recycling endosome. In (b) under resting conditions virally expressed pHluorin-GluR2 labels both punctate spines (red, purple) and diffuse shaft regions (blue). Treatment with 50 mM NMDA causes an immediate loss of the diffuse staining, representative of GluR2 endocytosis. This is followed by the loss of punctate staining after approximately 10 min. (c) Shows a schematic of the main events in AMPAR recycling: (1) AMPARs are released from the PSD by switching binding partners accompanied by degradation of the PSD; (2) receptors exit the spine by lateral diffusion to sites of clathrin endocytic machinery; (3) clathrin-mediated exocytosis transports AMPARs to the early endosome; (4) under all conditions, GluR3 is sent to the late endosome prior to degradation, and GluR2 enters this pathway after NMDA treatment but is sent to the recycling endosome following AMPA treatment; (5) GluR1 constitutively enters the recycling endosome, but its rate of exit from this compartment is increased by NMDA; (6) GluR2 exit from the recycling endosome is promoted by an interaction with NEEP21; (7) GluR2 is rapidly recycled to the synapse following AMPA treatment or in the absence of activity. AMPA, a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid; E, early endosome; EEA1, early endosome marker; GluR, glutamate receptor; HA, peptide derived from human influenza haemagglutinin; L, late endosome; Lamp1, lysosome associated membrane protein 1; NEEP21, neuron-enriched endosomal protein of 21 kDa; NMDA, N-methyl-D-aspartate; PSD, postsynaptic density protein; R, recyclingendosome. (a) from Lee SH, Simonetta A, and Sheng M (2004) Subunit rules governing the sorting of internalized AMPA receptors in hippocampal neurons. Neuron 43: 221–236. (b) Adapted from Ashby MC, De La Rue SA, Ralph GS, Uney J, Collingridge GL, and Henley JM (2004) Removal of AMPA receptors (AMPARs) from synapses is preceded by transient endocytosis of extrasynaptic AMPARs. Journal of Neuroscience 24: 5172–5176.

Small GTPase Signaling Triggers AMPAR Endocytosis in LTD

NMDAR activation of the Rap1–p38MAPK pathway via the NR2B subunit causes internalization of GluR2/3 heteromers. Activation of the Rap2–c-Jun N-terminal kinase (JNK) pathway causes the internalization of GluR1/2 heteromers in depotentiation, which is the resetting of synapses that have undergone LTP back to the baseline. Rab5 overexpression specifically depresses AMPAR-mediated excitatory postsynaptic currents (EPSCs) through the removal of GluR1–3 from dendritic spine surfaces. A dominant negative mutant of Rab5 prevents LTD induction.

Rab5 activation is likely to occur downstream of p38MAPK activation because this regulates its interaction with Rab-GDI. As would be expected from the localization of Rab5 to endosomes, these effects of overexpression occur downstream of Ser880 phosphorylation and release from the PSD. Fate of Internalized AMPARs Is Subunit Specific

GluR1 internalizes rapidly and independently of activity, supporting the LTP model in which GluR1containing subunits are inserted and then replaced by GluR2/3 heteromers. Overexpressed GluR2 and GluR3 are internalized on agonist treatment, leading

274 AMPA Receptor Cell Biology/Trafficking

to GluR2 being recycled, but NMDAR activation targets GluR2 to lysosomes. GluR1 is recycled and GluR3 is directed to endosomes independent of drug treatment (Figure 2). These subunit rules appear to be determined by the NSF binding site, with GluR2 being dominant over GluR1. NSF may promote the recycling of the receptors, and this further supports a model in which NSF activity or binding to GluR2 is reduced during LTD, allowing targeting to lysosomes. Overall, these data suggest that NMDAR activation during LTD leads to the degradation of GluR2-containing receptors, consistent with an activity-dependent depletion of AMPARs at the synapse. However, exposure to the sodium-channel blocker tetrodotoxin (TTX), which prevents action potentials and therefore evoked presynaptic glutamate release, reverses the effects of NMDA and AMPA. The physiological relevance of this is not clear, but it does highlight the importance of the prior experience of the synapse in determining the effect of plasticity protocols. It has been suggested that this might represent a form of homeostatic plasticity known as synaptic scaling, in which neuronal sensitivity is altered in line with the overall sensitivity of the network. TTX-treated neuronal cultures have been used to show the role of proteins other than NSF involved in directing AMPARs through the endocytic pathway. Suppression of neuron-enriched endosomal protein of 21 kDa (NEEP21) reduces the rate of GluR1 recycling in response to NMDA. GluR2 internalizes to a NEEP21 positive compartment following NMDA treatment. The presence of Rab4 and syntaxin 13 in this compartment suggests that NEEP21 promotes trafficking to the recycling endosome. In neurons with suppressed NEEP21, AMPAR-mediated EPSCs are decreased, but this effect can be overridden with dynamin mutants to block endocytosis. Because LTP is blocked in the absence of NEEP21, this protein may also have a role in trafficking receptors from intracellular holding pools. The forward trafficking of GluR2 from the early endosome to the recycling endosome is mediated by the formation of a GluR2– GRIP1–NEEP21 complex. Although it is assumed that an as yet unidentified GluR1–NEEP21 complex interacting protein mediates the recycling of GluR1, the lack or low availability of such a protein may explain the retarded recycling of GluR1 to the cell surface.

Summary The mechanisms of AMPAR trafficking, which in turn influence synaptic efficacy, are complex. However, overall patterns of AMPAR forward traffic,

recruitment, and cycling have been established. The insertion of AMPARs at newly formed synapses and at synapses that have received a signal to undergo LTP depends on GluR1. Insertion of these GluR1containing AMPARs is preceded and/or accompanied by an upregulation of slot proteins at the synapse that hold the AMPARs at the PSD until such time as they are replaced by constitutively cycling GluR2/3containing AMPARs. During LTD, AMPARs are released from the synapse by alterations to their scaffold binding partners, which are then degraded. AMPARs are then endocytosed at sites separate from the PSD. The fate of the receptors endocytosed by LTD appears to differ. GluR2/3-containing receptors are degraded, whereas GluR1-containing receptors are slowly recycled to extrasynaptic sites. Although this model can explain many of the features of plasticity, clearly a great deal of work remains to be done to obtain a full mechanistic understanding of how AMPARs are regulated. Given the crucial role of these receptors in brain function and dysfunction, we believe that gaining this information is a goal of fundamental importance. See also: AMPA Receptors: Molecular Biology and Pharmacology; NMDA Receptor Function and Physiological Modulation; NMDA Receptors, Cell Biology and Trafficking; Transporter Proteins in Neurons and Glia.

Further Reading Ashby MC, De La Rue SA, Ralph GS, Uney J, Collingridge GL, and Henley JM (2004) Removal of AMPA receptors (AMPARs) from synapses is preceded by transient endocytosis of extrasynaptic AMPARs. Journal of Neuroscience 24: 5172–5176. Ashby MC, Ibaraki K, and Henley JM (2004) It’s green outside: Tracking cell surface proteins with pH-sensitive GFP. Trends in Neuroscience 27: 257–261. Bredt DS and Nicoll RA (2003) AMPA receptor trafficking at excitatory synapses. Neuron 40: 361–379. Choquet D and Triller A (2003) The role of receptor diffusion in the organization of the postsynaptic membrane. Nature Reviews Neuroscience 4: 251–265. Collingridge GL, Isaac JT, and Wang YT (2004) Receptor trafficking and synaptic plasticity. Nature Reviews Neuroscience 5: 952–962. Dingledine R, Borges K, Bowie D, and Traynelis SF (1999) The glutamate receptor ion channels. Pharmacological Reviews 51: 7–61. Gerges NZ, Backos DS, Rupasinghe CN, Spaller MR, and Esteban JA (2006) Dual role of the exocyst in AMPA receptor targeting and insertion into the postsynaptic membrane. EMBO Journal 25: 1623–1634. Horton AC and Ehlers MD (2003) Neuronal polarity and trafficking. Neuron 40: 277–295. Lee SH, Simonetta A, and Sheng M (2004) Subunit rules governing the sorting of internalized AMPA receptors in hippocampal neurons. Neuron 43: 221–236.

AMPA Receptor Cell Biology/Trafficking Malenka RC and Bear MF (2004) LTP and LTD: An embarrassment of riches. Neuron 44: 5–21. Palmer CL, Cotton L, and Henley JM (2005) The molecular pharmacology and cell biology of alpha-amino-3-hydroxy-5-methyl4-isoxazolepropionic acid receptors. Pharmacological Reviews 57: 253–277.

275

Perestenko PV and Henley JM (2003) Characterisation of the intracellular transport of GluR1 and GluR2 a-amino-3hydroxy-5-methyl-4-isoxazole propionic acid receptor subunits in hippocampal neurons. Journal of Biological Chemistry 278: 43525–43532.

NMDA Receptor Function and Physiological Modulation K Zito, University of California at Davis, Davis, CA, USA V Scheuss, Max-Planck-Institute for Neurobiology, Martinsried, Germany ã 2009 Elsevier Ltd. All rights reserved.

Introduction With key roles in essential brain functions ranging from the basics of excitatory neurotransmission to the complexities of learning and memory, the N-methyl-D-aspartate (NMDA) receptor can be considered one of the fundamental neurotransmitter receptors in the brain. Named for its most potent exogenous agonist, the NMDA receptor has been thoroughly characterized via electrophysiological and pharmacological techniques, as well as through molecular manipulations and transgenic knockout strategies. The NMDA receptor belongs to a family of ionotropic receptors for the excitatory amino acid glutamate and is characterized by high affinity for glutamate, a high unitary conductance, high calcium permeability, and a voltage-dependent block by magnesium ions. this article focuses on the basic biophysical properties and physiological functions of NMDA receptors and how these are modulated by various signaling molecules and biochemical cascades under physiological conditions.

Biophysical Properties of NMDA Receptors During excitatory neurotransmission, presynaptic release of glutamate activates glutamate receptors in the postsynaptic membrane, resulting in the generation of an excitatory postsynaptic potential (EPSP). Contributing to the EPSP are two classes of glutamate receptors, the non-NMDA receptors (a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA)/kainate receptors) and the NMDA receptors (Figure 1). The rise times for NMDA receptor currents (10–50 ms) are much slower than those of non-NMDA receptors (0.2–0.4 ms). NMDA receptors also deactivate with a slower time course (50–500 ms vs. 2 ms for non-NMDA glutamate receptors), much longer than the time course of glutamate in most synaptic clefts (1.2 ms). Therefore, during synaptic transmission, the non-NMDA glutamate receptors provide rapid depolarization in response to neurotransmitter release, and the NMDA receptor kinetics determine the duration of the synaptic current. Activation of NMDA receptors by the neurotransmitter glutamate requires glycine as an essential

276

coagonist. Glutamate and glycine molecules bind different subunits of the receptor; two of each are thought to be required for maximum activation of the receptor. Recent experiments have shown that D-serine can also bind at the glycine site and might represent another physiological coagonist. Compared with glutamate receptors of the non-NMDA type, which have low-affinity binding sites for glutamate (EC50 500 mM), NMDA receptors bind glutamate with high affinity (EC50 1 mmol l1). Despite this high affinity for glutamate, NMDA receptors are not saturated during synaptic transmission at synapses of cortical pyramidal neurons. NMDA receptor activation leads to opening of an ion channel that is selective for cations, resulting in the influx of Naþ and Ca2þ ions and efflux of Kþ ions. Although most glutamate receptors are cation selective, few are permeable to calcium ions. Its exceptional calcium permeability is the first of two key properties of the NMDA receptor that form the basis for its regulatory role in synaptic plasticity. Entry of calcium into the postsynapse via the NMDA receptor (Figure 2) permits coupling of electrical synaptic activity to biochemical signaling via activation of Ca2þ-dependent enzymes and downstream signaling pathways. In this way, calcium influx through the NMDA receptor can lead to long-term changes in synaptic strength and other cellular modifications, including alterations in synaptic structure or connectivity. The NMDA receptor has a high single channel conductance (30–50 pS) compared with that of other glutamate receptor types (4–15 pS). The open probability of agonist-bound receptors has been estimated to range between 0.04 and 0.3, and open times can vary from 0.1 to 8 ms. The molecular events that underlie the opening of NMDA receptors are predicted to include two independent agonist-binding steps preceding a single, concerted conformational change that results in channel opening. Channel closing is thought to be controlled both by the unbinding rate of glutamate and by receptor desensitization. In fact, NMDA receptors can enter long-lived desensitized conformations in which glutamate is bound but the channel remains closed. The second key biophysical property of the NMDA receptor, and the one that bestows its proposed role in learning and memory, is that it is blocked by magnesium ions in a voltage-dependent manner. At resting membrane potential, NMDA receptors are blocked by magnesium ions; however, if excitation by synaptic inputs causes sufficient depolarization of the neuron, the Mg2þ block is relieved and those NMDA receptors which have glutamate bound will open (Figure 3).

NMDA Receptor Function and Physiological Modulation 277

Figure 1 NMDA receptor component of excitatory postsynaptic potentials (EPSPs) and excitatory postsynaptic currents (EPSCs). (a) Unitary EPSP (‘EPSC’) recorded in a L2/3 cortical pyramidal neuron. The NMDA receptor-mediated component (subtraction) was obtained by subtracting the non-NMDA receptor-mediated component (a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPAR) component) when NMDA receptors were blocked with the NMDA receptor antagonist AP5. (b) Unitary EPSC (‘EPSC’) recorded in a L2/3 cortical pyramidal neuron. The NMDA receptor-mediated component (subtraction) and the non-NMDA receptor mediated component (AMPAR component) were determined as in (a). Reproduced from Blackwell Publishing: figure 6A and B in Feldmeyer D, Lu¨bke J, Silver RA, et al. (2002) Synaptic connections between layer 4 spiny neuron-layer 2/3 pyramidal cell pairs in juvenile rat barrel cortex: Physiology and anatomy of interlaminar signaling within a cortical column. Journal of Physiology 538(3): 803–822.

∆F/F

a

l

b

c

Figure 2 Synaptic activation of NMDA receptors causes calcium entry into spines. (a) Two-photon fluorescence image of a dendritic branch of a hippocampal CA1 pyramidal neuron loaded with the calcium indicator Oregon Green BAPTA. Scale bar ¼ 1 mm. (b) Sequences of line scans in which the vertical dimension corresponds to the line indicated in (a) and the horizontal dimension represents time. Below: Relative change in fluorescence (DF/F(t)) at the bottom spine (averaged over window indicated by bracket) during calcium entry evoked by synaptic stimulation (time indicated by arrowhead). The neuron was voltage clamped at positive potential to isolate NMDA receptor-mediated Ca2þ influx. Scale bars ¼ 250 ms, 50% DF/Fmax. (c) Average EPSC (bottom) and corresponding DF/F(t) (top) at the same spine as in (b). Scale bars ¼ 25% DF/Fmax, 100 pA, 25 ms. Reprinted by permission from MacMillan Publishers Ltd.: Nature (Mainen ZF, Malinow R, and Svoboda K (1999) Synaptic calcium transients in single spines indicate that NMDA receptors are not saturated. Nature 399: 151–155) copyright 1999.

Since opening of the NMDA receptor requires simultaneous activation by glutamate and depolarization to relieve the magnesium block, the NMDA receptor can act as a coincidence detector for pre- and postsynaptic activity. Glutamate and glycine affinities, calcium permeability, ion conductance, channel kinetics, and sensitivity to Mg2þ block of the NMDA receptor are all determined in part molecularly, by alternative splicing and subunit composition. Most of these biophysical properties of NMDA receptors can also be modulated by various ionic and molecular interactions and signaling pathways (described below).

NMDA Receptor Physiological Function The outstanding physiological function of NMDA receptors is the coupling of electrical to biochemical signaling in neurons by mediating calcium influx in response to synaptic activity. However, NMDA receptors also serve important functions in electrical neurotransmission alone. Despite the voltage-dependent magnesium block, NMDA receptors can contribute significantly to the amplitude of unitary evoked postsynaptic potentials. The slow kinetics of NMDA receptor-mediated excitatory postsynaptic currents (EPSCs) facilitates temporal summation and reduces

278 NMDA Receptor Function and Physiological Modulation Control Mg-free

Mg-free

10 µM GLU Mg 500 µM

30 pA +60 20 +40 10

−80

−60

−40 −20

+20

+40

+60

0

mV

10 pA 4s

10

−40

20 −60 a

b

Na+

Ca2+

Glu

Na+ Glycine

Ca2+ Glycine

Glu Mg2+

Extracellular Mg2+

Plasma membrane

Vm

Vm

Cytosolic

c

K+

K+

Figure 3 Magnesium block of NMDA receptors. (a) Voltage dependence of glutamate-induced currents in Mg2þ-free (circles) and Mg2þ-containing solution (500 mM; squares). (b) Voltage and Mg2þ-dependence of NMDA receptor current noise. Reprinted by permission from MacMillan Publishers Ltd: Nature (Nowak L, Bregestovski P, Ascher P, et al. (1984) Magnesium gates glutamate-activated channels in mouse central neurons. Nature 307: 462–465) copyright 1984. (c) Schematic of the mechanism of the magnesium block. At resting potential, the pore of the NMDA receptor channel is blocked by magnesium ions. Upon depolarization, the magnesium ions are removed from the pore, and the channel can pass current.

dendritic filtering of synaptic inputs. In addition, coactivation of multiple synapses can trigger NMDA receptor-dependent dendritic spikes by generating sufficient depolarization to overcome the magnesium block. Such dendritic spikes cause nonlinear synaptic integration, which enhances the computational power of neurons and may allow neurons to detect the synchrony of inputs (Figure 4). The NMDA receptor plays an essential role in brain plasticity, that is, the ability of the brain to change in response to external stimuli, such as during learning and memory. First investigated pharmacologically, chronic blockade of the NMDA receptor by

infusion of an antagonist into the ventricles of rats was shown to impair spatial learning. Since that time, similar behavioral experiments have been performed using knockout mice lacking NMDA receptor subunits from specific subregions of the hippocampus. These experiments have repeatedly provided evidence for the importance of the NMDA receptor in learning and memory processes. The role of the NMDA receptor in learning and memory is attributed to its ability to regulate excitatory synaptic transmission. The NMDA receptor has been shown to play an essential role in both the strengthening of synapses, through long-term

NMDA Receptor Function and Physiological Modulation 279 Paired-pulse stimulation Control APV Individual EPSP Arithmetic sum Combined EPSP

AB

5 mV 50 ms

Measured peak EPSP (mV)

a

Pre Post

200

150

100

50

Paired-pulse, APV

20

0 15

−200

−150

−100

−50

0

50

Pre/post spike interval (ms) 10 5

Control APV

0 0

b

Normalized EPSP slope (%)

250

5 10 15 20 Expected peak EPSP (mV)

Figure 4 NMDA receptors in nonlinear synaptic integration. (a) Somatic voltage responses of a layer 5 pyramidal neuron to stimuli from two electrodes (A and B) placed 30 mm apart on a single dendritic branch display nonlinear summation (left panel). With application of the NMDA receptor antagonist, APV, the summation is linear. EPSP, excitatory postsynaptic potential. (b) Expected vs. actual peak somatic responses plotted for different stimulus intensities before (black) and after (blue) application of APV. NMDA receptor blockade linearizes synaptic summation. Reprinted by permission from MacMillan Publishers Ltd: Nature Neuroscience (Polsky A, Mel BW, and Schiller J (1994) Computational subunits in thin dendrites of pyramidal cells. Nature Neuroscience 7: 621–627), copyright 1994.

potentiation (LTP), and the weakening of synapses, through long-term depression (LTD). LTP and LTD are proposed to be cellular mechanisms underlying learning and memory. Via its role as a coincidence detector, the NMDA receptor is capable of signaling coincident pre- and postsynaptic activity. Detailed studies of the relationship between the temporal correlation of pre- and postsynaptic activity and the resulting type of plasticity have led to the discovery of spike-timing dependent plasticity. If the presynaptic action potential (AP; and thus synaptic activity) repetitively precedes the generation of the postsynaptic AP by less than 50 ms, the synapse is potentiated; however, if the postsynaptic AP occurs within 50 ms before the presynaptic AP (and thus independently of presynaptic activity), the synapse is depressed (Figure 5). With respect to the more complex and irregular patterns of neuronal activity that are likely found in vivo, it has been suggested that both rate and timing determine the direction of neuronal plasticity.

Figure 5 Spike-timing-dependent synaptic plasticity. Change in excitatory postsynaptic potential (EPSP) slope induced by repetitively paired pre- and postsynaptic spikes in layer 2/3 of rat visual cortex. Insets depict the sequence of spiking. Pre before postsynaptic spiking leads to potentiation while post before presynaptic spiking leads to depression. Used from The American Physiological Society: figure 1 in Dan Y and Poo MM (2006) Spike timingdependent plasticity: From synapse to perception. Physiological Reviews 86: 1033–1048.

The NMDA receptor also acts as a coincidence detector during development, where it plays a critical role in the maturation of synapses and the activitydependent establishment of topographic maps in the brain. By acting to strengthen neighboring connections of similar activity patterns, the NMDA receptor enforces the principle that ‘cells that fire together, wire together.’ In addition, early in development, many excitatory synapses appear to contain only NMDA receptors, and therefore these synapses are thought to be at first silent (no EPSP in response to synaptic stimulation). It is hypothesized that only through coincident activity sufficiently robust to depolarize the postsynaptic cell and relieve the voltage-dependent Mg2þ block of the NMDA receptors does the insertion of non-NMDA receptors at these synapses occur. While presynaptic NMDA receptors have been detected by immuno-electron microscopy, their function remains far less established compared with their postsynaptic counterparts. As autoreceptors, presynaptic NMDA receptors have been shown to modulate synaptic transmission in the spinal cord, cerebellum, and entorhinal cortex and to play a role in cerebellar and cortical LTD. Calcium influx through the NMDA receptor is thought to be responsible for its roles in LTP and LTD and for both neuroprotective and neurotoxic effects. Although these various effects may at first seem contradictory, the explanation lies in the spatial

280 NMDA Receptor Function and Physiological Modulation

and temporal patterns of calcium influx. Calcium influx at synaptic sites is thought to activate signaling pathways that lead to LTP or LTD, depending on the temporal pattern and amplitude of the calcium transient. Synaptic calcium influx also leads to activation of cyclic adenosine monophosphate response element binding protein (CREB), a transcription factor, resulting in activation of gene expression. One of the genes upregulated in response to activated CREB is brainderived neurotrophic factor (BDNF), which acts to promote prosurvival programs. In contrast, calcium influx through extrasynaptic NMDA receptors is thought to lead to CREB deactivation and inhibition of BDNF expression, which can have deleterious effects on cell health. In fact, overstimulation of NMDA receptors results in neuronal excitotoxicity, which has been implicated in the loss of neurons associated with ischemic stroke; Alzheimer’s, Parkinson’s, and Huntington’s disease; and amyotrophic lateral sclerosis.

Desensitization

As the NMDA receptor is a ligand-gated ion channel, it can display a decrease in conductance despite the continuous presence of agonist, a process referred to as desensitization. Three different types of desensitization have been reported for NMDA receptors, a glycine-sensitive, a glycine-insensitive, and a calciumdependent type. The glycine-sensitive desensitization refers to the transition of the NMDA receptor into a glutamate-bound closed state, reflecting a negative allosteric interaction between the glutamate and the glycine binding sites, which results in a reduced glycine affinity in the presence of high glutamate concentrations. High glycine concentrations can overcome this type of desensitization. Glycineinsensitive desensitization has been observed only in dialyzed cells and excised membrane patches and thus does not appear to be physiologically relevant. The calcium-dependent desensitization (also referred to as calcium-dependent inactivation) provides a negative feedback loop by which calcium entering the cell via NMDA receptors in turn leads to the desensitization of the receptor (Figure 6), although calcium from other sources (voltage-gated calcium channels or release from intracellular stores activated by second messenger cascades) has the same effect. In a current model, calcium influx activates calmodulin to displace a-actinin2 from the NMDA receptor. The resulting dissociation of the NMDA receptor from the cytoskeleton is proposed to cause a conformational change in the receptor, which reduces its open probability.

Physiological Modulation of NMDA Receptors Although glycine and/or D-serine are essential coagonists of the NMDA receptor, and thus, strictly speaking, not modulators, their cytosolic levels nevertheless regulate NMDA receptor activation. One of two amino acids appears to predominate, dependent on the brain region. The cytosolic concentrations are controlled by release and uptake by both neurons and glia. Stimulated release of glycine and D-serine can be induced by depolarization or non-NMDA glutamate receptor activation. In the spinal cord, glycine spillover from inhibitory synapses has been shown to enhance NMDA receptor currents.

Endogenous Allosteric Modulators

NMDA receptor function is fine-tuned by various forms of allosteric modulation involving endogenous extracellular substances such as zinc ions, protons, polyamines,

Inactivation NMDA EPSCs, 2.7 mM [Ca]o

Control 10 µM NMDA 0.2 mM [Ca]o

Recovery 10 µM NMDA, 0.2 mM [Ca]o

7th stimulus 1 min 0.5 nA 1s

1st stimulus

1 nA

100 ms

0.5 nA 4 min

1s

Figure 6 Calcium-dependent desensitization of NMDA receptors. Whole cell currents of cultured rat hippocampal neurons (left and right panels) were evoked in low calcium (0.5 mM) solutions by application of NMDA. During a train of seven excitatory postsynaptic currents (EPSCs) evoked in 2.7 mM calcium, the NMDA receptor mediated EPSC was inactivated to 50% of control (middle). The whole cell current was inactivated to a similar degree and recovered within 4 min. Used from The American Physiological Society: figure 3 in Rosenmund C, Feltz A, and Westbrook GL (1995) Calcium-dependent inactivation of synaptic NMDA receptors in hippocampal neurons. Journal of Neurophysiology 73: 427–430.

NMDA Receptor Function and Physiological Modulation 281

and reducing and oxidizing agents (Figure 7). Zinc and polyamines mediate a voltage-dependent block of NMDA receptors, which is weaker but appears to involve the same intrapore residues as block by magnesium. Some of the voltage-independent effects of allosteric modulators are mutually interdependent, and structural evidence suggests convergence by overlapping binding sites or common downstream structural modifications. Zinc, protons, and oxidizing agents inhibit NMDA receptor function, while polyamines and reducing agents cause potentiation by modifying channel open frequency and time or agonist or modulator affinities. These modulations are considered to be physiologically relevant during synaptic transmission, when synaptic activity can cause shifts in pH, co-release of Zn2þ occurs at certain synaptic terminals (e.g., mossy fiber-CA3), and the amine histamine is released from modulatory afferents (e.g., those arising from the anterior hypothalamus). An important aspect of the negative regulation of NMDA receptor function by pH and oxidizing agents is that it can limit or delay cell damage under pathological conditions such as stroke and ischemia. Phosphorylation

NMDA receptors can be phosphorylated by the serine/ threonine kinases protein kinase C (PKC), protein kinase A (PKA) and calcium/calmodulin-dependent protein kinase II (CaMKII), as well as by the tyrosine kinases Src and Fyn. In general, phosphorylation enhances NMDA receptor function (Figure 7).

A significant percentage of NMDA receptor subunits in the brain are estimated to be phosphorylated by PKC or PKA at one or more sites (10–70%, depending on the region of the brain). Phosphorylation by PKC reduces the affinity for extracellular magnesium ions and increases the open probability. Calcium influx through the NMDA receptor itself can enhance the potentiation mediated by PKC. However, in some preparations, phosphorylation by PKC reduced NMDA receptor-mediated responses by preventing NMDA receptor subunit clustering. Less is known about the regulation of NMDA receptors by PKA and CaMKII. Active CaMKII directly associates with the NMDA receptor, and this association is required for some forms of activity-driven synaptic potentiation. PKA activation has been shown to increase the fractional Ca2þ influx through the NMDA receptor; however, some of the effects of PKA activation on NMDA function appear to be indirect. In contrast, fewer NMDA receptor subunits on neural membranes are phosphorylated on tyrosine residues (2–4%). Phosphorylation by Src enhances NMDA receptor function (Figure 8) by reducing the potency of Zn2þ block of recombinant NMDA receptors expressed heterologously; however, evidence is lacking for this mechanism in hippocampal or spinal cord neurons. An alternative view proposes that phosphorylation of NMDA receptor subunits by Src-family kinases could affect downstream signaling proteins, either by recruiting them to the NMDA receptor, by

Na+

Ca2+ Glycine

Glu Mg 2+ Modulations that decrease current

Zn 2+ po yam nes

Modulations that increase current

K+ Zn2+

H+ Polyamines

Cys reduction

Polyamines

Cys oxidation

Zn2+ Ser/Thr dephosphorylation (PP1, 2A, 2B/calcineurin)

Tyr dephosphorylation

P -Ser/Thr phosphorylation

(PKC, PKA, CaMKII)

Tyr- P phosphorylation (Src, Fyn)

Figure 7 Schematic overview of modulators of NMDA receptor functions. Activation of the NMDA receptor requires binding of glutamate (Glu) and glycine (or serine) together with membrane depolarization to release the magnesium block (center). Modulators causing depression of NMDA receptor currents (left) include Hþ, Zn2þ, polyamines, reducing agents, and the protein phosphatases PP1, PP2A, and PP2B/calcineurin. Polyamines and Zn2þ cause a voltage-dependent block similar to that caused by Mg2þ. Modulators causing potentiation of NMDA receptor currents (right) include external polyamines, oxidizing agents, and the protein kinases PKC, PKA, CaMKII, Src, and Fyn. For details see text.

282 NMDA Receptor Function and Physiological Modulation

protecting them against degradation, or through blocking the assembly of signaling complexes. Srcmediated phosphorylation appears to play a role in LTP induction and seems to be modulated by BDNF. Phosphorylation by Fyn may target the NMDA receptor to the plasma membrane by antagonizing its interaction with spectrin. In some cases, Src activity is required for PKCmediated activation of NMDA receptor currents, placing Src downstream of PKC on the same pathway. PKC stimulation leads to the activation of the nonreceptor protein tyrosine kinase, cell adhesion kinase-b, which activates Src by disrupting the intramolecular interactions that maintain Src in a low-activity state. Activation of Src then leads to the potentiation of NMDA receptor currents. The effects of Src on NMDA receptors can be modulated by H-Ras, a small guanosine triphosphate (GTP)-binding protein, which binds Src and inhibits its activity. Hippocampal pyramidal neurons from H-ras homozygous null mice displayed enhanced NMDA receptor synaptic responses. The activation of NMDA receptors by phosphorylation can be reversed by serine and threonine phosphatases 1, 2A, and 2B (calcineurin) and endogenous tyrosine phosphatases. Inhibition of endogenous protein tyrosine phosphatase activity led to potentiation of NMDA receptor currents, indicating that these phosphatases participate in determining the basal phosphorylation level of the NMDA receptor.

Peak currents (nA)

1.5

NMDA

1.0 Src

Control

0.5

0

Src 0.5 nA 2

4 6 8 10 12 Time (min)

0.25 s

Figure 8 The protein tyrosine kinase pp60c-src potentiates NMDA currents. Whole cell recordings were made from mouse cultured hippocampal neurons. Peak currents evoked by rapid application of NMDA (100 mM) are plotted in the graph on the left. The tyrosine kinase pp60c-src (30 U per milliliter) was actively perfused through the recording electrode. The period of the perfusion is indicated by the horizontal bar on the graph. On the right, representative currents recorded before (Control) and after (Src) intracellular perfusion with pp60c-src are shown. NMDA was applied as indicated by the trace above the currents. Reprinted by permission from MacMillan Publishers Ltd: Nature (Wang YT and Salter MW (1994) Regulation of NMDA receptors by tyrosine kinases and phosphatases. Nature 369: 233–235), copyright 1994.

Calcineurin can be activated by calcium influx through the NMDA receptor itself. Modulatory Signaling Receptors

Modulation of NMDA receptors by kinases and phosphatases occurs downstream of a number of membrane receptors, the largest group of which are the G-protein-coupled receptors (GPCRs). GPCRs are integral membrane proteins which transmit extracellular signals to the cytoplasm by coupling to heterotrimeric GTP-binding proteins. Activation of a subset of GPCRs, including metabotropic glutamate receptors (mGluRs), muscarinic acetylcholine receptors, m opioid receptors, lysophosphatidic acid (LPA) receptors, and the protease-activated receptor PAR1, enhances NMDA receptor currents. Those GPCRs that are coupled to Gq proteins, such as mGluR5, m1 muscarinic receptor, and the LPA receptor, are thought to act through a PKC-Src signaling pathway, and mGluR1 also activates Src, but through a PKCindependent pathway. A second set of GPCRs, those coupled through Gs, are thought to enhance activity of the NMDA receptor through activation of Src-family kinases via PKA and receptor for activated C kinase 1 (RACK1). One such GPCR, the pituitary adenylate cyclase activating polypeptide (PACAP) receptor, is activated by the neuropeptide PACAP(1–38), which has been shown to potentiate NMDA receptor-mediated excitatory postsynaptic field potentials in hippocampal slices. Finally, NMDA receptors have been shown to interact directly with G-protein-coupled dopamine receptors. Depending on the class of dopamine receptor, activation of dopamine receptors has been reported both to enhance and to inhibit NMDA receptor currents. NMDA receptor currents are also influenced by downstream signaling from receptor protein tyrosine kinases, including the EphB receptors, platelet-derived growth factor (PDGF) receptors, and insulin receptors. EphB receptors interact directly with NMDA receptors in cultured neurons, and their activation by ephrinB2 increases NMDA receptor-dependent calcium responses via the activation of Src. Insulin has been shown to enhance NMDA receptor activity in hippocampal slices via a mechanism that requires PKC and tyrosine kinase activity. In contrast, application of PDGF depresses NMDA receptor currents in hippocampal pyramidal neurons via a phospholipase C-IP3 receptor-cyclic AMP-PKA pathway that inhibits Src activity. Cytokine receptor activation has been shown to enhance NMDA receptor currents. One group of cytokines, leptins, enhances NMDA receptor-mediated EPSCs in hippocampal slices via activation of the leptin receptor Ob-Rb. The leptin-mediated potentiation was

NMDA Receptor Function and Physiological Modulation 283

reduced by inhibitors of PI3-kinase, mitogen-activated protein kinase, and Src-family kinases. Another cytokine, interleukin (IL)-1b, enhanced the rise in intracellular calcium following application of NMDA to cultured hippocampal neurons, a response mediated by the IL-1RI receptor. NMDA Receptor Complex

The ability of modulators to influence NMDA receptor currents is dependent on their abundance and proximity to the receptor. Proteomic characterization identified synaptic NMDA receptors as part of a remarkably large macromolecular signaling complex called the NMDA receptor complex. NMDA receptors were found linked to receptors, adhesion molecules, scaffolding proteins, signaling molecules, cytoskeletal proteins, and various novel proteins, in complexes lacking non-NMDA receptors. Among the signaling proteins were kinases, phosphatases, GTPases, and GTPase-activating proteins. Physical linkage of these receptors, signaling molecules, and the cytoskeleton to the NMDA receptor is an important way to facilitate rapid modulation of the receptor in response to synaptic activity. See also: D-Serine: From its Synthesis in Glial Cell to its

Action on Synaptic Transmission and Plasticity; Kainate Receptor Functions; Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDARDependent Forms; Long-Term Potentiation (LTP): NMDA Receptor Role; NMDA Receptors, Cell Biology and Trafficking.

Further Reading Cull-Candy SG and Leszkiewicz DN (2004) Role of distinct NMDA receptor subtypes at central synapses. Science’s STKE: Signal Transduction Knowledge Environment 255: re16.

Dan Y and Poo MM (2006) Spike timing-dependent plasticity: From synapse to perception. Physiological Reviews 86: 1033–1048. Dingledine R, Borges K, Bowie D, et al. (1999) The glutamate receptor ion channels. Pharmacological Reviews 51: 7–61. Edmonds B, Gibb AJ, and Colquhoun D (1995) Mechanisms of activation of glutamate receptors and the time course of excitatory synaptic currents. Annual Review of Physiology 57: 495–519. Feldmeyer D, Lu¨bke J, Silver RA, et al. (2002) Synaptic connections between layer 4 spiny neuron-layer 2/3 pyramidal cell pairs in juvenile rat barrel cortex: Physiology and anatomy of interlaminar signaling within a cortical column. Journal of Physiology 538(3): 803–822. Husi H, Ward MA, Choudhary JS, et al. (2000) Proteomic analysis of NMDA receptor-adhesion protein signaling complexes. Nature Neuroscience 3: 661–669. Kotecha SA and MacDonald JF (2003) Signaling molecules and receptor transduction cascades that regulate NMDA receptormediated synaptic transmission. International Review of Neurobiology 54: 51–106. Lester RA and Jahr CE (1992) NMDA channel behavior depends on agonist affinity. Journal of Neuroscience 12: 635–643. Mainen ZF, Malinow R, and Svoboda K (1999) Synaptic calcium transients in single spines indicate that NMDA receptors are not saturated. Nature 399: 151–155. Nowak L, Bregestovski P, Ascher P, et al. (1984) Magnesium gates glutamate-activated channels in mouse central neurons. Nature 307: 462–465. Polsky A, Mel BW, and Schiller J (2004) Computational subunits in thin dendrites of pyramidal cells. Nature Neuroscience 7: 621–627. Rosenmund C, Feltz A, and Westbrook GL (1995) Calciumdependent inactivation of synaptic NMDA receptors in hippocampal neurons. Journal of Neurophysiology 73: 427–430. Salter MW and Kalia LV (2004) Src kinases: A hub for NMDA receptor regulation. Nature Reviews Neuroscience 5: 317–328. Schneggenburger R, Zhou Z, Konnerth A, et al. (1993) Fractional contribution of calcium to the cation current through glutamate receptor channels. Neuron 11: 133–143. Vanhoutte P and Bading H (2003) Opposing roles of synaptic and extrasynaptic NMDA receptors in neuronal calcium signalling and BDNF gene regulation. Current Opinion in Neurobiology 13: 366–371. Wang YT and Salter MW Regulation of NMDA receptors by tyrosine kinases and phosphatases. Nature 369: 233–235.

NMDA Receptors, Cell Biology and Trafficking R J Wenthold and R S Petralia, National Institute on Deafness and Other Communication Disorders, National Institutes of Health, Bethesda, MD, USA Published by Elsevier Ltd.

Introduction The N-methyl-D-aspartate (NMDA) receptor is present at the postsynaptic membrane of nearly all glutamatergic synapses, where it is found with other glutamate receptors, particularly a-amino-3-hydroxy5-methyl-4-isoxazole propionic acid (AMPA) receptors. It is a key player in synaptic plasticity, and its malfunction has been implicated in a wide range of neurological and psychiatric disorders ranging from ischemia to schizophrenia. It is unique among neurotransmitter receptors since it requires activation by two agonists, glutamate and glycine, which interact with the NR2 and NR1 subunits, respectively. The ion channel of the NMDA receptor is normally blocked by magnesium, and the channel opens only after the neuron is depolarized by activation of AMPA receptors, thus allowing the NMDA receptor to serve as a coincidence detector. Finally, the ion channel of the NMDA receptor is permeable to calcium, a universal intracellular signaling molecule. These properties have made the NMDA receptor one of the most intensely studied molecules in the brain. Since neurons are highly compartmentalized, forming synaptic connections with thousands of other neurons, and the NMDA receptor has a restricted distribution within a neuron, trafficking of this receptor is central to its functional regulation. Here we define trafficking as the nonrandom movement of a protein within a neuron to, or from, a site where it is required for function. This process undoubtedly depends on interactions with other molecules, particularly other proteins, lipids, sugars, and small molecules.

Structure and Subunits Three subunits of the NMDA receptor complex have been identified. A single NR1 subunit has eight splice variants. There are four NR2 subunits (NR2A–D) and two NR3 subunits. NR2A and NR2B are the most abundant NR2 subunits, with NR2B being predominant early in development and NR2A appearing later, along with a general decrease in NR2B expression. All subunits have the same topology, with three transmembrane domains, a reentrant loop, an extracellular N-terminus, and a cytoplasmic C-terminus,

284

which is unusually long in all NR2 subunits. Supporting several previous functional and biochemical reports, the recent determination of the crystal structure of the extracellular domains of the NMDA receptor indicates a tetrameric protein made up of two NR1 and two NR2 subunits. In a tetramer, the NR2 subunits can be the same, such as two NR2B subunits, or different, such as one NR2B and one NR2A, in this complex. While NR3 can also assemble into complexes with NR1 and NR2, the function of NR3 remains unclear, and it has been reported that NR3 can assemble with NR1 alone to produce a functional glycine receptor. The NMDA receptor subunits interact with a large number of other proteins, which modify their function and influence their trafficking (Table 1). The subunit composition determines the functional properties of the NMDA receptor. This is dependent mostly on the NR2 subunits forming receptors, with kinetics ranging from the slowest (containing NR2D) to the fastest (NR2A). Thus, channels with faster kinetics allow less calcium to enter the neuron, and would probably be less effective in developing plasticity. When NR3 is assembled into a complex with NR1 and NR2, it decreases channel function. NR3A knockout mice are normal, but show enhanced NMDA responses early in development.

NMDA Receptors in the Endoplasmic Reticulum All NR2 subunits and some NR1 subunits, depending on the splice variant, are retained in the endoplasmic reticulum (ER) in their unassembled form and are released after assembly (Figure 1). This is a standard mechanism of quality control to assure that unassembled or misfolded proteins do not reach the cell surface. For the NR1 subunit, ER retention is controlled by an RXR motif (X indicates any nonacidic amino acid) in the C-terminal domain of the major splice variant, NR1-1. NR1 splice variants (NR1-2 and NR1-4) that lack this retention motif are not retained, and others (NR1-3) that contain this motif plus the C20 cassette are also not retained. Thus, only the NR1-1 splice variant is retained in the ER; this is the major splice variant, accounting for about 60% of the total NR1. The C20 cassette causes the subunit to terminate in the amino acid sequence – STVV, and this motif is required for overriding the RXR retention motif. This is a postsynaptic density/disc/zonula occludens-1 (PDZ) binding motif and also a site of interaction for components of the coat protein complex II (COPII),

NMDA Receptors, Cell Biology and Trafficking Table 1 Proteins that bind directly to NR1 and/or NR2 NR1a

NR2

PSD-95 (to NR1-3, NR1-4) SAP102 (to NR1-3, NR1-4) PSD-93 (to NR1-3, NR1-4) SAP97 (to NR1-3, NR1-4) a-Actinin Tubulin Spectrin Myosin regulatory light chain Dopamine D1 receptor CaMKII NADH dehydrogenase subunit 2 (ND2)b Calmodulin Neurofilament-L Yotiao EphB receptors Apolipoprotein E receptor 2 SALM1

PSD-95 SAP102 PSD-93 SAP97 a-Actinin Tubulin Spectrin Myosin regulatory light chain

Sec23/24 of COPII

Dopamine D1 receptor CaMKII NADH dehydrogenase subunit 2 (ND2)b S-SCAM (¼ MAGI-2) CIPP mLin-7 (¼ Veli ¼ Mals) Phospholipase C-g Rack1 m subunit of adaptor protein complexes (AP1–4) a1-Chimerin RasGRF-1 Cyclin-dependent kinase-5 (cdk5)b

a

NR1-3 and NR1-4 are splice variants. cdk5 may bind directly to NR2A, and ND2 appears to bind Src to NMDA receptors, but the exact nature of these interactions has not been determined. b

285

which is involved in ER-to-Golgi trafficking. Thus, the PDZ or COPII interaction, or both, may control the ER export of these NR1 splice variants and override the retention that is mediated by the RXR motif. It remains unclear why some NR1 splice variants can exit the ER unassembled, or if they actually do in neurons, since there is no evidence that NR1 homomeric complexes exist in neurons or serve any function. NR1 subunits lacking retention motifs may be more readily exported from the ER assembled with NR2. The mechanism by which NR2 subunits are retained in the ER is more complex than that of NR1 and not well characterized. There appears to be an ER retention signal(s) in the C-terminus although the specific site has not been identified. There is also an additional mechanism for ER retention since deletion of the C-terminus does not allow the remainder of the molecule to reach the cell surface unassembled, as is the case for NR1; however, this C-terminus-deleted construct can assemble with NR1 to produce functional receptors. Equally poorly understood is the mechanism by which the retention signals on NR1 and NR2 are negated to allow the assembled complex to exit the ER. Mutual masking of the retention motifs in the C-terminus, as has been demonstrated for some other proteins, does not

Large excess of NR1 NR1 NR2A NR2B PDZ protein

Rapid degradation (t1/2 = 1–2h) of unassembled NR1 subunits

NR1 splice variants NR1-2, NR1-3, and NR1-4 can exit the ER unassembled

Assembly

ER Exit. NR2A and NR2B require HLFY motif in C-terminus Early PDZ protein association

Figure 1 Assembly of the NMDA receptor in the ER. All NR2 subunits and the NR1-1 splice variant are retained in the endoplasmic reticulum (ER). Assembly negates retention and allows exit from the ER. NR1 subunits are synthesized in excess of NR2, and most are retained in the ER and rapidly degraded. In heterologous cells, some splice variants that lack the ER retention signal present in the C1 cassette, or contain the C20 cassette, can exit the ER unassembled, probably as homodimers. It is not known if this is also the case in neurons or if it is of functional significance. Assembly can lead to receptor complexes with two identical or two different NR2 subunits. Receptors can assemble and be functional in the ER based on MK801 binding, but export requires a four-amino-acid segment (HLFY in NR2A and NR2B) immediately following transmembrane domain 4.

286 NMDA Receptors, Cell Biology and Trafficking

appear to play a role in the NMDA receptor, since subunits with a deletion of either the NR1 or NR2 C-terminal tails alone can form functional receptors. An interesting finding that further complicates our understanding of the ER retention of NR2 is that the NR2 subunit does not need to be intact to form functional receptors. The NR2A subunit can be divided into two polypeptides in the extracellular loop between transmembrane domains 3 and 4 (TM3 and TM4). Co-expression of these two fragments with NR1 produces functional receptors on the cell surface. A single neuron can contain a variety of NMDA receptor complexes (such as NR1/NR2B, NR1/ NR2A, and NR1/NR2A/NR2B), and the question arises whether or not the formation of these complexes simply reflects the availability of subunits, or is regulated in some way. There is a large excess of NR1 subunits synthesized, which would ensure that most NR2 subunits would find an NR1 binding partner and form a functional receptor. This fits with the fact that the particular NR2 subunit controls the properties of the receptor.

Trafficking of NMDA Receptors in the Dendrite NMDA receptors are concentrated at postsynaptic sites on dendrites while synthesis occurs predominantly in the cell body, requiring a mechanism to transport receptors from the cell body to synapses. This could be done in two ways. Receptors could be delivered to the plasma membrane at the cell body, and the receptors could reach the dendrite by lateral diffusion or through some form of regulated transport while they are in the plasma membrane. Alternatively, receptors can be delivered to dendritic sites through association with intracellular transport vesicles. The latter is assumed to be the major mechanism of delivery, but lateral movement to the synapse is probably also necessary for movement into the synapse after delivery to an extrasynaptic location (Figure 2). NMDA receptors can associate early in the biosynthetic pathway with PDZ proteins, including members of the PSD-95 family of membrane-associated guanylate kinases (MAGUKs), which are transported in the dendrite as a complex with the NMDA receptors. The MAGUK allows the indirect association of the NMDA receptors with multiple other proteins that can influence the trafficking of the NMDA receptor. The exocyst, a complex of eight proteins, associates with the receptor through a PDZ interaction with the PSD-95-family MAGUK, SAP102 (synapse-associated protein), and perhaps with other PDZ proteins; disruption of this interaction will dramatically reduce the

Myosin complex Actin filament

Kinesin KIF1Ba Microtubule

Adaptor complex?

Adaptor complex?

mPins + Ga i Kinesin Exocyst complex

mLin complex Kinesin KIF17

Figure 2 Proposed complexes involved in the transport of NMDA receptors. Two kinesins that indirectly interact with NMDA receptors are KIF17 and KIF1Ba, and additional kinesins may also play a role. Delivery of receptors to spines is likely to involve myosins, which remain to be identified. Both Sec8, probably in a complex with other exocyst proteins, and mPins bind to SAP102 and other related PDZ proteins (MAGUKS) and may target the NMDA receptor/PDZ protein complexes to the plasma membrane. Also, G-protein signaling via the interaction of mPins and Gai may regulate the effect of mPins on NMDA receptor trafficking and postsynaptic spine structure.

surface expression of receptor. Another trafficking molecule, mPins/LGN (mammalian homolog of Drosophila partner of inscuteable) interacts with the receptor complex through the Src homology 3/guanylate kinase (SH3/GK) motifs of the MAGUK and disruption of this interaction also reduces NMDA receptor surface delivery. mPins/LGN interacts with a number of additional proteins, including Gai, suggesting a possible link between G-protein signaling and NMDA receptor trafficking. Transport vesicles carrying NMDA receptors appear to be largely distinct from those carrying AMPA receptors, suggesting that these two receptor populations are packaged into distinct populations of transport vesicles as they exit the trans-Golgi network (TGN). In young neurons transport in dendrites involves multiple rounds of endo- and exocytosis with the plasma membrane and association of the transport vesicles with SAP102. Two motor systems

NMDA Receptors, Cell Biology and Trafficking

have been identified in the transport of NMDA receptors and MAGUKs. The kinesin KIF17 associates indirectly with NMDA receptors through a complex with mLin-7, mLin-2, and mLin-10 and may mediate the transport of NMDA receptors along microtubules. MAGUKs can associate directly with the kinesin KIF1Ba, making it a candidate for transporting NMDA receptors associated with SAP102. Myosin motors, which can mediate transport of proteins along actin filaments, are also likely to play a role. NMDA receptors are regulated by myosin light chain kinase and show a direct interaction with myosin regulatory light chain. The NMDA receptor, therefore, is transported as part of a macromolecular complex. However, unlike the presynaptic active zone, the postsynaptic density, a structure with which the NMDA receptor is associated at synapses, is not transported as a completely preassembled complex. The NMDA receptor complex may remain intact as it is delivered to the synapse, but a more likely scenario is that the addition and removal of proteins are a normal parts of the trafficking process to change parameters as the local environment changes.

Trafficking of NMDA Receptors at the Synapse Synaptic Delivery and PDZ Proteins

The steps involved in delivering NMDA receptors present in transport vesicles in the dendrite to the synapse are largely unknown. This process may involve an intermediate stop in an endosomal compartment, where they may mix with recycling receptors, or direct delivery to the plasma membrane. The site of delivery is also unknown but is likely adjacent to the synapse rather than directly at the synapse. This will require movement within the plasma membrane, and these receptors may represent the extrasynaptic pool of NMDA receptors (discussed later). The number and composition of NMDA receptors at the synapse are relatively stable compared to those of AMPA receptors. However, the mechanism regulating synaptic NMDA receptors is not well understood. The number of receptors is not based on availability, since overexpression of NR2 subunits does not increase the synaptic number of receptors, but does increase the extrasynaptic number, indicating a local mechanism of control for synaptic receptors. The NR2/PSD-95 interaction was the first PDZ interaction identified (subsequent studies have shown that the other three members of the PSD-95 family of MAGUKs, SAP102, SAP97, and PSD-93, also can interact with the NR2 subunits), and it generally has been assumed that synaptic NMDA receptors are tethered to the PSD

287

through this PDZ interaction. A number of studies have supported this by showing that the PDZ binding domain is required for an NR2 subunit to be admitted to the synapse. However, recent studies have shown that the NMDA receptor/PSD-95 interaction is not straightforward. Knockout of PSD-95 or its overexpression does not change synaptic NMDA receptors, but overexpression does increase AMPA receptors. Studies have also shown that an NR2A subunit lacking its PDZ binding motif can enter the synapse, although NR2B subunits without the PDZ binding motif cannot. A variety of adhesion proteins are found associated with NMDA receptors at synapses and these proteins may help regulate NMDA receptor distribution and function. Postsynaptic EphB2 receptors are activated by their presynaptic ligand, ephrin-B, and bind directly to the extracellular domain of NR1, affecting NMDA receptor function directly and subsequent dendrite arborization indirectly. SALM1 can bind to the extracellular domain of NR1 and this and some other SALMs also can associate with NMDA receptors via a mutual PDZ-mediated binding to MAGUKs of the PSD-95 family. SALM1 can enhance surface expression of transfected NR2A. A number of other adhesion factors do not bind directly to NMDA receptors but instead form associations with them at the synapse and may affect NMDA receptor trafficking and activation and subsequent neuronal structure and function; examples are neuroligin, cadherin/catenin, and a number of members of the L1/NrCAM and NCAM (neural cell adhesion molecule) families of adhesion proteins. Neuroligin binds to neurexin in the presynaptic terminal and to PSD-95 and other NMDA receptor-associated MAGUKs in the postsynaptic density. Neuroligin localization at synapses may help determine if developing synapses become excitatory or inhibitory. In fact, a number of adhesion proteins such as neuroligin, cadherin, NCAM, and EphB receptors may mediate the earliest stages of synaptogenesis that are required for subsequent trafficking of MAGUKs and NMDA receptors to the nascent postsynaptic membrane. Clathrin-Mediated Internalization and Regulation of NMDA Receptors

The number of molecules of any protein on the cell surface can be regulated by internalization, and several studies have now shown that a number of factors can increase the internalization of NMDA receptors. These include synaptic activity, activation of type 1 metabotropic receptors, and exposure to the agonist glycine. In addition, run down of extrasynaptic receptors appears to depend on internalization.

288 NMDA Receptors, Cell Biology and Trafficking

YEKLSSIESDV AP2 PDZ NR1/NR2B NR2

PDZ protein

NR2 Y Fyn

Fyn

AP2 complex

P Y

Anchor

a

NR1/NR2B NR2A-containing receptors are enriched at the synapse

NR1/NR2A

NR2B-containing receptors are enriched at the extrasynaptic membrane

NR2B-containing receptors are recycled

b

NR2A-containing receptors are degraded

Figure 3 (a) Model for regulation of NR2B-containing receptors at the synapse. Internalization is mediated by the interaction of AP2 with the YEKL motif in the distal C-terminus of NR2B. At the synapse, Y1472 in this motif is phosphorylated by Fyn kinase and is unable to interact with AP2. Fyn is recruited by PSD-95, which associates with NR2B through its PDZ binding domain. Loss of phosphorylation of Y1472 facilitates the interaction with AP2 and leads to removal from the synapse and internalization. Retention of NMDA receptors at the synapse may involve additional anchors. (b) NR2A- and NR2B-containing receptors are regulated differently at the synapse. Several studies have suggested that NR2A-containing receptors are more abundant at the synapse while NR2B-containing receptors are more abundant at extrasynaptic sites. Several factors could contribute to this differential localization. Receptors can enter the synapse by lateral diffusion in the plasma membrane. A filter may selectively exclude NR2B-containing receptors from the synapse. Alternatively, NR2B-containing receptors may be more likely to be removed from the synapse through internalization (as described in panel (a)). Finally, internalized NR2A-containing receptors are more susceptible to degradation while NR2B-containing receptors are recycled. If they enter the extrasynaptic pool, the recycled receptors would contribute to the enrichment of NR2B-containing receptors in this pool.

NMDA receptors can be removed from the plasma membrane through a clathrin-dependent route involving the adaptor protein AP2 (Figure 3). Clathrindependent internalization requires the recognition of a particular motif in its cytoplasmic domain by the adaptor protein. The NR2B subunit has such a motif, YEKL, near its C-terminus, that has been shown to play a role in the surface stability of the receptor. Interestingly, a similar motif in the NR2A subunit is

not involved in the internalization of this subunit, but rather an upstream dileucine motif (also a substrate for the AP2 adaptor) was shown to play such a role. The differences in the trafficking of these two subunits are highlighted in the different pathways followed by the internalized receptors. While both initially enter early endosomes, NR2B is trafficked to recycling endosomes while NR2A is trafficked to late endosomes. Therefore, NR2B-containing receptors are more likely

NMDA Receptors, Cell Biology and Trafficking

to be recycled and NR2A-containing receptors are more likely to be degraded after internalization. Does internalization play a role in regulating the number of synaptic NMDA receptors? It is unlikely that internalization per se regulates the number of receptors, but the availability of endocytic motifs such as the YEKL motif on NR2B may. This motif contains a tyrosine, Y1472, which is a major substrate for Fyn kinase. In the striatum, synaptic membranes have a high level of phosphorylated Y1472 while intracellular membranes have a low level of phosphorylated Y1472. In cultures of cerebellar granule cells and hippocampal neurons, using transfection of NR2B subunits with mutations in this motif, it was shown that mutations that block the interaction of AP2 increase the number of synaptic receptors. Furthermore, while the PDZ binding domain of NR2B is required for synaptic localization, NR2B subunits that lack this domain and have mutations of the YEKL motif still localize to the synapse. One explanation is that the PDZ interaction may not be critical for anchoring the receptor complex to the postsynaptic protein complex, but rather plays a role in maintaining the phosphorylation of Y1472. PSD-95 has been shown to interact with Fyn kinase, and, at the synapse, PSD-95 may recruit Fyn kinase, thus maintaining the phosphorylation of Y1472 and preventing its interaction with AP2 and its subsequent internalization. The fact that the internalization of NR2A is not affected by mutations in a homologous motif, YKKM, indicates that different factors control the surface stability of NR2A-containing receptors. The YEKL and dileucine motifs near the distal end of the C-terminus of NR2B and NR2A, respectively, are not the only endocytic motifs found on the NMDA receptor. Another tyrosine-containing motif is present in the proximal part of the C-terminus, near the last transmembrane domain. This motif was functionally characterized in receptors with the NR2A subunit, but it is present in all NR2 subunits and the NR1 subunit. Since mutation of neither this motif nor the distal motif can block all internalization, there are likely other motifs present in the NR1 and NR2 C-termini that mediate clathrin-dependent internalization. Nonclathrin mechanisms may also exist. Extrasynaptic Receptors

NMDA receptors are found not only at the postsynaptic membrane but have also been reported on glia, presynaptic terminals, the perisynaptic membrane, and the extrasynaptic membrane. Most extensively studied are perisynaptic and extrasynaptic receptors, which may be functionally important and activated when glutamate that is released at the synapse is not

289

sufficiently sequestered and is able to diffuse outside the synapse. While most neurons contain both synaptic and extrasynaptic receptors, in a few specific cases such as the retinal ganglion cell, NMDA receptors are found only outside the synaptic cleft. In other neurons, NR2A-containing receptors are preferred at the synapse and NR2B-containing receptors are preferred at the extrasynaptic membrane, although the separation is not complete. This subunit preference may reflect functional differences between these two pools of receptors. This, however, is difficult to reconcile with studies that show diffusion of receptors from the extrasynaptic plasma membrane into the synapse. In one study, synaptic receptors were blocked with the use-dependent antagonist, MK801, and recovery of function was monitored. The recovery of function was surprisingly rapid, and controls ruled out sources other than the extrasynaptic membrane. The preferential localization of NR2A- and NR2B-containing receptors would be consistent with these findings if there was a filter that favors the entry of NR2A into the synapse or the removal of NR2B from the synapse. NMDA receptors do recycle, although not as fast as AMPA receptors, and the enrichment of NR2B in the extrasynaptic membrane could reflect this. As noted earlier, internalized NR2B receptors are more likely to be recycled and NR2A receptors are more likely to be degraded. If receptors are recycled to the extrasynaptic membrane, the difference in intracellular trafficking could favor NR2B at the extrasynaptic membrane. Such a mechanism would also assume that there is an exchange between the synaptic and extrasynaptic pools of receptors. The functional consequences resulting from activation of these receptors is also an area of intense study. Activation of extrasynaptic receptors promotes cell death. A series of controversial studies has indicated that NR2A-containing receptors mediate longterm potentiation (LTP) and that NR2B-containing receptors mediate long-term depression (LTD). This requires more work, and it will be interesting to determine if position of the receptor – for example, synaptic versus perisynaptic – plays a role. Developmental Changes in NMDA Receptors

During development, most neurons begin expressing predominantly NR2B, which then gradually decreases as NR2A expression increases. The functional implication of this switch is that NMDA receptors in young animals have slower kinetics (due to a preponderance of NR2B), and receptors in older animals are faster (due to a preponderance of NR2A). It has been proposed that this change, in which ‘adult’ NMDA receptors are faster and admit less calcium, may be related

290 NMDA Receptors, Cell Biology and Trafficking

to the decrease in synaptic plasticity in older animals. While most neurons continue to express NR2B in older animals, and therefore have a mixture of NR2A- and NR2B-containing receptors and perhaps those containing both subunits, cerebellar granule cells completely stop synthesizing NR2B. During development, NR2A and NR2C expression increases in these neurons. Trafficking plays a role in the preferential localization of subunits during development. A series of elegant studies has shown that, within hours, visual experience can cause synaptic NMDA receptors in the visual cortex to become more NR2A-like. Thus, either NR2A-containing receptors are preferentially added or NR2B-containing receptors are preferentially lost, or both. While this could depend somewhat on new protein synthesis, it is unlikely that such a change could be effected in the time period, and a more likely explanation is that NR2A-containing receptors are recruited from extrasynaptic or intracellular sites. Similar transitions from NR2B- to NR2A-containing receptors may be occurring at other synapses during development, or even at mature synapses. Thus, the mechanism underlying this switch is fundamentally important to NMDA receptor trafficking.

Phosphorylation and Trafficking of NMDA Receptors As discussed in the preceding section, Fyn kinase phosphorylation has a major impact on trafficking of the NR2B subunits. The NMDA receptor is a substrate for several other kinases, including other members of the Src family of protein tyrosine kinases, calcium/ calmodulin-dependent protein kinase II (CaMKII), cyclin-dependentkinase-5 (cdk5),cyclic AMP-dependent protein kinase A (PKA), protein kinase C (PKC), casein kinase II, and others. The NMDA receptor directly interacts with CaMKII at the postsynaptic density. This interaction is important for the translocation of CaMKII to the synapse. CaMKII has been shown to interact with all three of the common subunits, NR1, NR2A, and NR2B, through their C-termini, although the affinities of the interactions and the functional implications are very different. A recent study in which the interacting domains on NR2A and NR2B were switched showed that the CaMKII interaction with the NR2B subunit is required for LTP. Phosphorylation of NMDA receptor subunits affects the trafficking of the individual subunits. The subcellular distribution of the NR1 subunit expressed alone in heterologous cells can be regulated by PKC phosphorylation of the serine residues in the C1 domain. These amino acids are near the ER retention signal, and mutations that mimic phosphorylation greatly increased surface expression of NR1/tac chimeric constructs.

Phosphorylation by several other kinases, particularly casein kinase II and PKC, affects the trafficking of the assembled receptor. The NR2B subunit terminates in ESDV, which is a PDZ binding domain. Serine-1480, present in this domain, is a substrate for casein kinase II, and phosphorylation of Ser1480 blocks interaction with PDZ proteins. Therefore, this may be an important mechanism for regulating receptor/PDZ interactions. PKC activation increases the delivery of NMDA receptors to the cell surface.

Degradation of NMDA Receptors While trafficking can regulate local amounts and distributions of a protein effectively, altering either synthesis or degradation can effect global changes in proteins. Like other proteins, the NMDA receptor is continually degraded and, at steady state, replenished by synthesis of new receptors. In cultured cerebellar granule cells, the half-life of assembled NMDA receptor is about 1 day. Unassembled NR1 subunits, which are synthesized in excess compared to NR2, are retained in the ER and degraded rapidly, with a half-life of 1–2 h. As presented previously, there is a developmental change in the relative amounts of NR2A and NR2B, which could be achieved by altering synthesis or degradation. Since mRNA expression often matches protein levels, it suggests that synthesis is the main controlling factor. The degradation of NMDA receptors may not be entirely random, but may depend on the subunit composition and activity of the neuron; degradation, therefore, could change the subunit balance or the total number of receptors at a particular synapse. As noted above, receptors containing NR2A or NR2B follow different pathways after internalization. Recent studies have shown that ubiquitination plays a major role in the removal of postsynaptic proteins, including NMDA receptors. Covalent attachment of ubiquitin can cause internalization of membrane proteins and degradation in the proteasome. Several proteins of the PSD are ubiquinated, including PSD-95. NR1 can be ubiquitinated on its extracellular domain. In this case, activity-dependent NMDA receptor degradation may be mediated by retrotranslocation followed by ubiquitination, although the mechanism is not well understood. See also: Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDAR-Dependent Forms; Long-Term Potentiation (LTP): NMDA Receptor Role; NMDA Receptor Function and Physiological Modulation; Postsynaptic Density/Architecture at Excitatory Synapses.

NMDA Receptors, Cell Biology and Trafficking

Further Reading Ali DW and Salter MW (2001) NMDA receptor regulation by Src kinase signaling in excitatory synaptic transmission and plasticity. Current Opinions in Neurobiology 11: 336–342. Barria A and Malinow R (2002) Subunit-specific NMDA receptor trafficking to synapses. Neuron 35: 345–353. Barria A and Malinow R (2005) NMDA receptor subunit composition controls synaptic plasticity by regulating binding to CaMKII. Neuron 48: 289–301. Dunah AW and Standaert DG (2001) Dopamine D1 receptordependent trafficking of striatal NMDA glutamate receptors to the postsynaptic membrane. Journal of Neuroscience 21: 5546–5558. Ehlers MD, Tingley WG, and Huganir RL (1995) Regulated subcellular distribution of the NR1 subunit of the NMDA receptor. Science 269: 1734–1737. Kornau HC, Schenker LT, Kennedy MB, et al. (1995) Domain interaction between NMDA receptor subunits and the postsynaptic density protein PSD-95. Science 269: 1737–1740. Lavezzari G, McCallum J, Dewey CM, et al. (2004) Subunitspecific regulation of NMDA receptor endocytosis. Journal of Neuroscience 24: 6383–6391. Li B, Chen N, Luo T, et al. (2002) Differential regulation of synaptic and extra-synaptic NMDA receptors. Nature Neuroscience 5: 833–834. McIlhinney RAJ, LeBourdelles B, Molnar E, et al. (1998) Assembly intracellular targeting and cell surface expression of the human N-methyl-D-aspartate receptor subunits NR1a and NR2A in transfected cells. Neuropharmacology 37: 1355–1367. Mu Y, Otsuka T, Horton AC, et al. (2003) Activity-dependent mRNA splicing controls ER export and synaptic delivery of NMDA receptors. Neuron 40: 581–594. Nong Y, Huang YQ, Ju W, et al. (2003) Glycine binding primes NMDA receptor internalization. Nature 422: 302–307. Perez-Otano I and Ehlers MD (2005) Homeostatic plasticity and NMDA receptor trafficking. Trends in Neuroscience 28: 229–238. Prybylowski K, Chang K, Sans N, et al. (2005) The synaptic localization of NR2B-containing NMDA receptors is controlled by interactions with PDZ proteins and AP-2. Neuron 47: 845–857.

291

Quinlan EM, Philpot BD, Huganir RL, et al. (1999) Rapid, experience-dependent expression of synaptic NMDA receptors in visual cortex in vivo. Nature Neuroscience 2: 352–357. Rao A and Craig AM (1997) Activity regulates the synaptic localization of the NMDA receptor in hippocampal neurons. Neuron 19: 801–812. Roche KW, Standley S, McCallum J, et al. (2001) Molecular determinants of NMDA receptor internalization. Nature Neuroscience 4: 794–802. Rumbaugh G and Vicini S (1999) Distinct synaptic and extrasynaptic NMDA receptors in developing cerebellar granule neurons. Journal of Neuroscience 19: 10603–10610. Sans N, Prybylowski K, Petralia RS, et al. (2003) NMDA receptor trafficking through an interaction between PDZ proteins and the exocyst complex. Nature Cell Biology 5: 520–530. Sans N, Wang PY, Du Q, et al. (2005) mPins, the mammalian homologue of Drosophila Pins, modulates PSD-95 and SAP102 trafficking and influences NMDA receptor surface expression. Nature Cell Biology 7: 1179–1190. Setou M, Nakagawa T, Seog DH, et al. (2000) Kinesin superfamily motor protein KIF17 and mLin-10 in NMDA receptorcontaining vesicle transport. Science 288: 1796–1802. Sprengel R, Suchanek B, Amico C, et al. (1998) Importance of the intracellular domain of NR2 subunits for NMDA receptor function in vivo. Cell 92: 279–289. Standley S, Roche KW, McCallum J, et al. (2000) PDZ-domain suppression of an ER retention signal in NMDA receptor NR1 splice variants. Neuron 28: 887–898. Tovar KR and Westbrook GL (2002) Mobile NMDA receptors at hippocampal synapses. Neuron 34: 255–264. Vissel B, Krupp JJ, Heinemann SF, et al. (2001) A use-dependent tyrosine dephosphorylation of NMDA receptors is independent of ion flux. Nature Neuroscience 4: 587–596. Washbourne P, Bennett JE, and McAllister AK (2002) Rapid recruitment of NMDA receptor transport packets to nascent synapses. Nature Neuroscience 5: 751–759. Wenthold RJ, Prybylowski K, Standley S, et al. (2003) Trafficking of NMDA receptors. Annual Reviews of Pharmacology and Toxicology 43: 335–358.

Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology F Nicoletti and V Bruno, University of Rome ‘La Sapienza,’ Rome, Italy G Battaglia, Istituto Neurologico Mediterraneo ‘Neuromed,’ Pozzilli, Italy ã 2009 Elsevier Ltd. All rights reserved.

Background The first challenge to the general belief that all glutamate receptors were ligand-gated ion channels was the demonstration by Sladeczek, Pin, Recasens, Bockaert, and Weiss that glutamate stimulates polyphosphoinositide (PI) hydrolysis in cultured striatal neurons. Similar data were obtained in brain slices, where the PI response to glutamate is maximal in the early postnatal life and declines progressively during postnatal development. The term ‘metabotropic glutamate (mGlu) receptors’ was introduced in 1987 by H Sugiyama and colleagues, who examined responses to excitatory amino acids in Xenopus oocytes injected with rat brain mRNA. Cloning of the first mGlu receptor subtype by S Nakanishi and colleagues in 1991 was the milestone for any further development of the field.

Classification of mGlu Receptors mGlu receptors form a family of eight subtypes subdivided into three groups based on their sequence similarities and G-protein coupling.

Homologous desensitization of group I mGlu receptors involves phosphorylation mechanisms mediated by either PKC or G-protein-coupled receptor kinases (GRKs). Different GRK isoforms, including GRK2, GRK4, and GRK5, have been found to desensitize mGlu1 receptors in recombinant cells. GRK4 is also required for desensitization of native mGlu1 receptors expressed by cerebellar Purkinje cells. While desensitization by GRK4 is mediated by receptor phosphorylation, desensitization by GRK2 is independent of phosphorylation and may involve a direct interaction with activated Gaq. The mGlu5 receptor is selectively desensitized by members of the GRK2 family (GRK2 and GRK3) through a mechanism that involves phosphorylation of the Thr840 residue in the C-terminus domain of the receptor. The Regulator of G-protein signaling, RGS4, which acts as effector antagonist by increasing the GTPase activity of Gaq, inhibits mGlu1 and mGlu5 receptor signaling. Group II mGlu Receptors

Group II includes mGlu2 and mGlu3 receptors, which are coupled to Gi/Go proteins (Figure 1). Their activation inhibits cAMP formation, inhibits L- and N-type voltage-sensitive Ca2þ channels, activates Kþ channels, and stimulates the MAPK and PI3K pathways. The latter two effects are presumably mediated by the bg subunits of Gi. Activation of group II mGlu receptors can also amplify the stimulation of cAMP formation by b-adrenergic receptor agonists in cultured astrocytes, and the stimulation of PI hydrolysis by mGlu1/5 receptor agonists in brain slices.

Group I mGlu Receptors

Group III mGlu Receptors

Group I includes mGlu1 (variants: mGlu1a, -1b, -1c, -1d, and -1e) and mGlu5 (variants: mGlu5a and -5b) receptors, which are primarily coupled to Gq/G11 proteins (Figure 1). Secondary coupling involves Gs and the pertussis toxin-sensitive Go. Activation of both subtypes in recombinant cells stimulates PI hydrolysis with an ensuing formation of the two intracellular second messengers, inositol 1,4,5-trisphosphate (IP3) and diacylglycerol (DAG). IP3 releases Ca2þ from intracellular stores, whereas DAG activates protein kinase C (PKC). Activation of group I mGlu receptors can also stimulate cyclic adenosine monophosphate (cAMP) formation, arachidonic acid release, the mitogen-activated protein kinase (MAPK) pathway, and the phosphatidylinositol 3-kinase (PI3K) pathway. The mGlu1 receptor is negatively coupled to a variety of Kþ channels and to the C1 type of transient receptor potential (TRP) channels.

Group III includes mGlu4, mGlu6, mGlu7 (variants: mGlu7a and -7b), and mGlu8 (variants: mGlu8a, -8b, and -8c) receptors, which are all coupled to Gi/Go proteins (Figure 1). Activation of mGlu4, mGlu7, and mGlu8 receptors inhibits cAMP formation and N-type voltage-sensitive Ca2þ channels. In cells expressing mGlu4 receptors, GRK2 attenuates agonist-stimulated MAPK activation without affecting inhibition of adenylyl cyclase. Activation of mGlu6 receptors inhibits cGMP-dependent cation channels in ON bipolar cells of the retina (see later).

292

Structure and Mechanisms of Activation of mGlu Receptors mGlu receptors belong to class C G-protein-coupled receptors (GPCRs), which also includes g-aminobutyric acid B (GABAB) receptors, the Ca2þ-sensing receptor

Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology 293

Groups

Subtypes

Main transduction pathways Glutamate

I

mGlu1a mGlu1b mGlu1c mGlu1d mGlu1e

PtdIns-4,5-P2

PLC

Gq

DAG Ins-1,4,5-P3 PKC

Ca2+

Ca2+ Glutamate K+ channel

mGlu5a mGlu5b

II

mGlu2

-

Gq

Glutamate

mGlu3 Gi/o mGlu4

-

AC cAMP ATP

III

mGlu7a mGlu7b

Glutamate Ca2+ channel

mGlu8a mGlu8b mGlu8c

-

Gi

Glutamate

III

mGlu6

PLC

Go

PtdIns-4,5-P2

DAG Ins-1,4,5-P3 PKC

Ca2+

Ca2+ Figure 1 The mGlu receptor family. PtdIns-4,5-P2, phosphatidylinositol 4,5-bisphosphate; Ins-1,4,5-P3, inositol 1,4,5-trisphosphate; DAG, diacylglycerol; PKC, phosphokinase C; PLC, phospholipase C.

expressed parathyroid cells, the taste receptors that detect sweet and umami, and some pheromone receptors. mGlu receptors are formed by (1) a large N-terminus extracellular domain (about 600 amino acids) containing the glutamate binding site and a Cys-rich region, (2) a seven-transmembrane (7TM) domain, and (3) an intracellular C-terminus domain that varies in length depending on the receptor subtype/splice variant. mGlu1a, mGlu5a, and mGlu5b receptors have long C-terminus domains, which favor their interaction with adaptor and scaffolding proteins (see later). In the mGlu1b and -1d variants, the 313 C-terminus amino acids are substituted by 20

(mGlu1b) and 22/26 (mGlu1d human/rat) amino acids. The mGlu1e isoform corresponds to the soluble extracellular domain of the mGlu1 receptor. The mGlu5b variant differs from mGlu5a for the presence of 39 amino acids inserted 49 residues downstream of the seventh TM domain. The existence of truncated forms of mGlu3 and mGlu6 receptors, corresponding to the extracellular portion of the receptors, has been demonstrated in human tissue. In the mGlu7b variant, the last 16 amino acids of mGlu7a are substituted by 23 different residues. The elegant report by Kunishima and colleagues in 2000 has afforded insights into the structure

294 Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology

and mechanism of activation of mGlu receptors. Glutamate binds to a ‘venus fly trap’ (VFT) region of the N-terminus domain constituted by two globular lobes separated by a hinge. The lobes oscillate between a close/active conformation and an open/ inactive conformation. Agonist binding to the hinge facilitates the closure of the two lobes, whereas binding of competitive antagonists stabilizes the open conformation. Similarly to other class C GPCRs, mGlu receptors function as dimers. Dimerization involves a hydrophobic interaction between lobe I of the VFT in each monomer, and is stabilized by a disulfide bond. The use of receptor chimeras containing the C-terminus domains of type 1 and type 2 GABAB receptors (which form obligatory heterodimers) has allowed to establish that the mGlu receptor dimer can be activated by a single molecule of orthosteric agonist, but that binding of two molecules (one to each monomer) is required for full receptor activation. In contrast, it appears that only one 7TM domain is required for effective G-protein activation. The 7TM domain contains the binding sites for positive allosteric modulators (‘potentiators’ or ‘enhancers’), which amplify receptor activation only in the presence of an orthosteric agonist – that is, in an activity-dependent manner. One enhancer molecule per dimer is needed for full amplification of mGlu receptors. The 7TM domain also bears the binding sites for negative allosteric modulators, which inhibit mGlu receptors independently of the concentrations of ambient glutamate. mGlu receptors interact with a variety of adaptor and scaffolding proteins, which regulate receptor expression and coupling with other receptors or enzymes. Homer proteins, which contain both postsynaptic density/disc/zonula occludens (PDZ) and Ena/ vasodilator-stimulated phosphoprotein homology-1 (EVH-1) domains, bind to a proline-rich motif in the C-terminal region of mGlu1a and mGlu5 receptors. Constitutively expressed, long isoforms of Homer proteins (Homer 1b, -1c, -2, and -3) have a C-terminal coiled-coil domain that mediates self-multimerization and allows the interaction of mGlu1a/5 receptors with other Homer-interacting proteins, such as the IP3 receptor (Figure 2(a)). This interaction is disrupted by the short inducible isoform of Homer (Homer 1a), which lacks the coiled-coil domain. Interestingly, Homer 1a is encoded by an early inducible gene, which is expressed in response to synaptic activation. The interaction between Homer and the long isoform of the PI3K enhancer, PIKE, mediates the stimulation of PI3K by mGlu1 receptor activation. mGlu1 receptors functionally interact with ephrin-B2, a member of the ephrin/Eph receptor family of transmembrane proteins which mediate processes of cell-to-cell interaction

during development and in the adult life. This interaction might have interesting implications for processes of developmental plasticity. Other proteins interacting with group I mGlu receptors are the E3 ubiquitin ligase, Seven in absentia homolog 1a, protein phosphatase 1c, tubulin, and tamalin. Protein phosphatase 2C binds selectively to, and dephosphorylates, mGlu3 receptors. mGlu7 receptors interact with calmodulin, protein interacting with PKC 1 (PICK 1), glutamate receptor-interacting protein 1 (GRIP 1), syntenin, a-tubulin, filamin A, and the catalytic g subunits of protein phosphatase 1C. A simultaneous or exclusive binding to these proteins tightly regulates mGlu7 receptor signaling. mGlu8 receptors interact with sumo1 and other components of the sumoylation cascade, such as ube2a, Pias1, Piasg, and Piasxb.

Cell Biology, Pharmacology, and Implications for Human Pathology The three groups of mGlu receptor ligands (Figure 3) have potential clinical applications, as outlined in Table 1. mGlu1 Receptor

The mGlu1 receptor consists of 1194 amino acids in humans and 1199 amino acids in rats and mice; accession numbers are Q13255 (human), P23385 (rat), and P97772 (mouse). The gene is named GRM1 (human) and Grm1 (rat, mouse); chromosomal locations are 6q24 (human), 1p13 (rat), and 10 A2 (mouse). At a subcellular level, mGlu1 proteins are mostly found in postsynaptic elements at the periphery of the postsynaptic density (PSD). This contrasts with the distribution of N-methyl-D-aspartate (NMDA) and a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptors, which are localized in the central region of the PSD. mGlu1 receptors are expressed by Purkinje cells of the cerebellar cortex, where they participate in the induction of longterm depression (LTD) at the parallel fiber–Purkinje cell synapse. LTD is a particular form of synaptic plasticity that, in the cerebellum, underlies motor learning. Mice with genetic deletion of mGlu1 receptors are ataxic and show persistent multiple climbing fiber innervation of cerebellar Purkinje cells. Interestingly, anti-mGlu1 receptor antibodies are found in a subset of patients with Hodgkin’s lymphoma and paraneoplastic ataxia. The mGlu1 receptor is also expressed in the olfactory bulb, pars compacta of the substantia nigra, hippocampus, and thalamic nuclei. In the hippocampus, activation of mGlu1 receptors inhibits GABA release and mGlu1 receptor antagonists are protective against excitotoxic and hypoxic/ischemic neuronal death. Unexpectedly, that

Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology 295

mGlu1/5

Glut

NMDAR

mGlu1/5

TRPC PSD95 Dynamin3

GKAP

Shank Cortactin IP3R

PIKE L

Ho mer-long

ER

F-actin

Ho mer-short

a

Photoreceptor

mGlu4/7/8 mGlu2/3

Glut

Na+

Glut

cGMP-dependent channel

EAAT mGlu3

DAG

mGlu5

Calcineurin (PP2B)

+ +

TGFb, NGF

b

mGlu6 Go

Ins-1,4,5-P

ON-bipolar cell

Glia

TGFb, NGF

PLC

c

Figure 2 Synaptic localization of mGlu receptors. (a) mGlu1 and mGlu5 receptors are found at the periphery of postsynaptic elements. mGlu1a, mGlu5a, and mGlu5b receptors interact with scaffolding and adaptor proteins, including Homer. (b) mGlu2/3 and mGlu4/7/8 receptors are predominantly found in presynaptic terminals, where their activation inhibits neurotransmitter release; mGlu3 and mGlu5 receptors are also found in glial cells. (c) Activation of mGlu6 receptors by the glutamate released from photoreceptor cells inhibits the activity of ON bipolar cells. Glut, glutamate; IP3R, inositol 1,4,5-trisphosphate receptor; ER, endoplasmic reticulum; EAAT, excitatory amino acid transporter; PP2B, protein phosphatase 2B; TRPC, transient receptor potential-C. (a) Adapted from Fourgeaud L (2005) Addicted to Holmer? Journal of Neuroscience 25(42): 9555–9556.

ectopic expression of mGlu1 receptors in melanocytes promotes the development of multiple melanomas in mice. Because mGlu1 receptors are found in human melanomas but not in benign nevi, it is possible that receptor activation contributes to neoplastic transformation of melanocytes. There are no selective agonists of mGlu1 receptors. 3,5-dihydroxyphenylglycine (DHPG) activates

both mGlu1 and mGlu5 receptors, although it has no activity at other glutamate receptor subtypes. The following drugs are examples of competitive mGlu1 receptor antagonists (in the rank order of affinity): LY367385 ¼ 4C3HPG > 4CPG > AIDA. Compounds Ro01–6128 and Ro67–7476 are mGlu1 receptor enhancers with nanomolar potency, whereas BAY36–7620 and CPCCOEt are negative allosteric

296 Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology Orthosteric ligands of mGlu receptors OH

OH

OH

H HO2C

CO2H

Cl H2N

CO2H

H2N

DHPG L-Glutamate (Non-selective, low potency) (mGlu1/5 Ago)

CO2H

H H

2R,4R-APDC (mGlu2/3 Ago)

H2N

NH2

O HO2C H2N

H H2N

CO2H

O

PO(OH)2 H

Me HO2C

CO2H H2N

L-AP4 (mGlu4-8 Ago)

EGlu (mGlu2/3 Ant)

PO3H2

O

H

CO2H

H2N

LY379268 (mGlu2/3 Ago)

PO3H2

PO3H2

CO2H H

NH2

LY354740 (mGlu2/3 Ago)

CO2H

LY367385 (mGlu1 Ant)

CO2H

O

H

CO2H

CO2H

H H2N

AIDA (mGlu1 Ant)

H

CO2H

NH

CO2H

H2N

CO2H

CHPG (mGlu5 Ago)

H

HO2C

CH2

Me

H

NH2

H2N

CO2H

HO2C

L-SOP (mGlu4-8 Ago)

CO2H

H2N

CO2H

R,S-PPG (mGlu4-8 Ago)

LY341495 (pan mGlu Ant)

MSOP (mGlu4-8 Ant)

Positive allosteric modulators of mGlu receptors

N

O

H N

O

O

N

F

H N F

DFB (mGlu5)

Ro 67-4853 (mGlu1)

Ro 01-6128 (mGlu1)

O

N N

O

S

O

N

H2O

S

O

NH2

O

F F

OH

O

N

O O

N

O HN

F

3-MPPTS (mGlu2)

APPES (mGlu2)

(−)-PHCCC (mGlu4)

Negative allosteric modulators of mGlu receptors OH

T

N A

S

H N

N

N NH

O

O

O

N

O

O

Cl

T

O

CPCCOEt (mGlu1)

N

OEt

BAY36-7620 (mGlu1)

MPEP (mGlu5)

MTEP (mGlu5)

Fenobam (mGlu5)

Figure 3 Selected mGlu receptor ligands. Ago, agonist; Ant, competitive antagonist.

modulators of mGlu1 receptors. (Abbreviations: LY367385, (E)-2-methyl-6-stryrylpyridine(–)-2oxa-4-aminocyclo[3.1.0]exane-4,6-dicarboxylic acid; 4C3HPG, 4-carboxy-3-hydroxyphenylglycine;

4CPG, 4-carboxyphenylglycine; AIDA, (R,S)-1aminoindan-1,5-dicarboxylic acid; Ro 01–6128, ethyl diphenylacetylcarbamate; Ro 67–7476, (S)-2-(4fluorophenyl)-1-(toluene-4-sulfonyl)pyrrolidine;

Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology 297

Table 1 Potential clinical applications of mGlu receptor ligands Group

Application

Group I

Chronic pain (mGlu1/5, Ant/NAM) Anxiety disorders (mGlu5, Ant/NAM) Drug addiction (mGlu5, Ant/NAM) Fragile X syndrome (mGlu5, Ant/NAM) Parkinsonism (mGlu5, Ant/NAM) Schizophrenia (mGlu5, PAM) Brain ischemia (mGlu1, Ant/NAM) Generalyzed anxiety disorders and panic attack (mGlu2/3, Ago/PAM) Drug addiction (mGlu2/3, Ago/PAM) Schizophrenia (mGlu2/3, Ago/PAM) Chronic pain (mGlu2/3, Ago/PAM; mGlu2 inducers) Epilepsy (mGlu4/7/8, Ago/PAM) Parkinsonism (mGlu4, Ago/PAM) Anxiety (mGlu8, Ago/PAM)

Group II

Group III

Ago, agonist; Ant, competitive antagonist; PAM, positive allosteric modulator; NAM, negative allosteric modulator.

CPCCOEt, 7-(hydroxyimino)cyclopropa[b]chromen1a-carboxylate ethyl ester.) mGlu5 Receptor

The mGlu5 receptor consists of 1212 amino acids in humans and 1203 amino acids in rats; accession numbers are P41594 (human) and P31424 (rat). The gene is named GRM5 (human) and Grm5 (rat, mouse); chromosomal locations are 11q14.2 (human), 1q32 (rat), and 7 D3 (mouse). Activation of recombinantly expressed mGlu5 receptors induces oscillatory increases in intracellular Ca2þ release. This property, which is not shared by mGlu1 receptors, relies on the presence of a Ser residue in the C-terminus domain, which is phosphorylated by PKC, and places the mGlu5 receptor at the core of basic processes in cell biology. Similar to the mGlu1 receptor, the mGlu5 receptor is localized in postsynaptic elements, where it is physically linked to NMDA receptors via a chain of interacting proteins, which include PSD-95, Shank, guanylate kinaseassociated protein (GKAP), and Homer (Figure 2(a)). Activation of mGlu5 receptors facilitates the opening of NMDA-gated ion channels, and activation of NMDA receptors amplifies mGlu5 receptor function by inhibiting receptor desensitization. Because of their functional interaction with NMDA receptors, mGlu5 receptors are involved in the induction of long-term potentiation (LTP), a putative electrophysiological substrate of associative learning. mGlu5 knockout mice show a defective LTP in the hippocampus and an impaired spatial learning. Activation of mGlu5 receptors also mediates a nonNMDA-dependent form of LTD in the hippocampus. mGlu5a and mGlu5b variants show an opposite developmental pattern of expression in the central

nervous system (CNS). mGlu5a receptors are abundantly expressed early after birth, and mediate the robust PI response to excitatory amino acids in the first 2 weeks of postnatal life. Expression of mGlu5b receptors increases with age and is prominent in the adult hippocampus, striatum, and cerebral cortex. mGlu5 receptors are also found in peripheral cells, including thymocytes, hepatocytes, melanocytes, cells of the male germinal line, osteoblasts, and insulinoma cell lines, where their function begins to be explored. Remarkably, mGlu5 receptors are present in both embryonic and neural stem cells, and their activation supports proliferation and self-renewal of these cells. A dysfunction of mGlu5 receptors has been associated with the fragile X syndrome, the most frequent form of inherited mental retardation. This syndrome is caused by the absence of the fragile X mental retardation protein (FMRP), an RNA-binding protein which functions as a negative regulator of protein synthesis. LTD mediated by mGlu5 receptors is enhanced in mice lacking FMRP, and pharmacological blockade of mGlu5 receptors partially corrects neurological and behavioral abnormalities in these mice. mGlu5 receptors are selectively activated by 2-chlorohydroxyphenylglycine (CHPG), which, however, is a relatively weak agonist. No selective competitive antagonists are currently available. 2-Methyl-6-(phenylethynyl)pyridine (MPEP), 6-methyl-2-(phenylazo) pyridin-3-ol (SIB-1757), and (E)-2-methyl-6-stryrylpyridine (SIB-1893) are prototypes of a growing list of negative allosteric modulators of mGlu5 receptors, which includes 3-[(2-methyl-1,3-thiazol-4-yl)ethynyl] pyridine (MTEP) and the anxiolitic drug, fenobam. Compounds 3,3’-difluorobenzaldazine (DFB), S-(4-fluorophenyl)-{3-[3-(4-fluorophenyl)-[1,2,4] oxadiazol-5-yl]-piperidin-1-yl}methanone (AD X47273), 3-cyano-N-(1,3-diphenyl-1H-pyrazol-5-yl) benzamide (CDPPB), and N-{4-chloro-2-[(1,3dioxo-1,3-dihydro-2H-isoindol-2-yl) methyl]phenyl}2-hydroxybenzamide (CPPHA) are example of mGlu5 receptor enhancers which amplify receptor function in an activity-dependent manner. mGlu5 receptor ligands are of potential interests for the treatment of psychiatric and neurological disorders, such as anxiety, schizophrenia, parkinsonism, and neuropathic pain. Systemic administration of MPEP produces anxiolytic effects in a number of tests, including fear-potentiated startle, auditory and contextual fear conditioning, the elevated plus maze, conflict tests, and stress-induced hyperthermia. These effects are mediated by mGlu5 receptor blockade in the hippocampus and in the central nucleus of the amygdala, two regions that are actively involved in the acquisition and expression of fear conditioning. Pharmacological blockade or genetic

298 Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology

deletion of mGlu5 receptors disrupts prepulse inhibition of the acoustic startle reflex in rodents, thus mimicking the deficit in sensorimotor gating typical of schizophrenic patients. Administration of MPEP also enhances the cognitive deficit induced by the psychotomimetic drug, phencyclidine. Hence, mGlu5 receptor enhancers are currently developed as potential antipsychotic agents. mGlu5 receptors are highly expressed in all stations of the basal ganglia motor circuit, and regulate the ‘indirect pathway’ of motor control by opposing the action of dopamine released from nigrostriatal terminals. The activity of this pathway is negatively regulated by the dopamine released from nigrostriatal fibers, which acts on D2 inhibitory receptors located on medium-size spiny GABAergic neurons that project to the external portion of the globus pallidus (GPe). In striatal projection neurons, mGlu5 receptors physically interact with A2A adenosine receptors, and act synergistically with them in antagonizing the inhibitory control exerted by D2 receptors. In addition, activation of mGlu5 receptors inhibits GPe neurons by a mechanism of cross-desensitization with mGlu1 receptors, and stimulates neurons of the subthalamic nucleus. Pharmacological blockade or genetic deletion of mGlu5 receptors protects mice against nigrostriatal damage produced by the parkinsonian toxin, 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP), or the psychostimulant, methamphetamine. This raises the attractive possibility that mGlu5 receptor antagonists/negative modulators behave as antiparkinsonian agents by relieving motor symptoms and reducing the ongoing degeneration of nigrostriatal neurons at the same time. mGlu5 and mGlu1 receptors are implicated in the pathophysiology of chronic pain. Both receptor subtypes are localized on peripheral unmyelinated sensory afferents, and their activation increases the sensitivity to noxious heat, a phenomenon termed ‘thermal hyperalgesia.’ Activation of mGlu5 receptors in peripheral nociceptors stimulates PI hydrolysis and arachidonic acid release from DAG, leading to prostaglandin formation. Autocrine or paracrine stimulation of prostanoid receptors by prostaglandins in nociceptors results into an increased sensitivity of vanilloid TRPV1 receptors, thus inducing thermal hyperalgesia. Group I mGlu receptors also contribute to the regulation of nociceptive transmission and plasticity in the dorsal horns of the spinal cord, and perhaps in other stations of the pain neuraxis. Antagonists/negative modulators of mGlu5 and mGlu1 receptors produce analgesic effects in models of inflammatory and neuropathic pain. Finally, mGlu5 receptors have been implicated in the synaptic mechanisms underlying drug addiction. These receptors positively regulate brain reward

function and drive drug consumption. Blockade of mGlu5 receptors by MPEP elevates the intracranial self-stimulation threshold, decreases cocaine and nicotine self-administration, and attenuates naloxoneinduced somatic signs of morphine withdrawal. mGlu2 and mGlu3 Receptors

The mGlu2 receptor consists of 872 amino acids in humans, rats, and mice; accession numbers are Q14416 (human) and P31421 (rat). The gene isnamed GRM2 (human) and Grm2 (rat, mouse); chromosomal locations are 3q21.31 (human), 8q32 (rat), and 9 (mouse). The mGlu3 receptor consists of 877 amino acids in humans and 879 amino acids in rats and mice; accession numbers are Q14832 (human), P31422 (rat), and Q9QYS2 (mouse). The gene is named GRM3 (human) and Grm3 (rat, mouse); chromosomal locations are 7q21.1–q21.2 (human), 4q32 (rat), and 5A1-h (mouse). mGlu2 and mGlu3 are highly homologous and have a similar functional and pharmacological profile. Both receptors are predominantly localized in the preterminal region of the axon, and, if present in glutamatergic nerve terminals, can only be activated by amounts of glutamate sufficient to escape the clearance mechanisms around the synapses (Figure 2(b)). Activation of group II mGlu receptors attenuates neurotransmitter release as a result of a reduced cAMP formation and inhibition of voltage-sensitive Ca2þ channels. In brain slices, activation of mGlu2/3 receptors fails to stimulate PI hydrolysis per se, but amplifies the stimulation of PI hydrolysis mediated by group I mGlu receptors. The functional significance of this mechanism is unknown. mGlu3 receptors are also found in astrocytes, where their activation stimulates the MAPK and PI3K pathways and leads to the production of tropic factors (Figure 2(b)). Both mGlu2 and mGlu3 receptors are activated by 2R,4R-4-aminopyrrolidine-2,4-dicarboxylic acid (APDC; (a selective agonist with micromolar affinity), and by the cyclopropan derivatives, 2S,20 R,30 R)2-(20 ,30 -dicarboxycyclopropyl)glycine (DCG-IV) and (2S,10 S,20 S)-2-(carboxycyclopropyl)glycine (L-CCG-I). Both drugs activate mGlu2 and mGlu3 receptors in the mid-nanomolar range; however, DCG-IV is also an NMDA receptor agonist, whereas L-CCG-I can also activate group I mGlu receptors. (1S,2S,5R,6S)(þ)-2-Aminobicyclo[3.1.0]hexane-2,6-dicarboxylic acid (LY354740) and (–)-2-oxa-4-aminocyclo[3.1.0] hexane-4,6-dicarboxylic acid (LY379268) are conformationally constrained glutamate analogs in which the glutamate backbone is locked into a fully extended state by incorporation into a bicycle[3.1.0]

Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology 299

hexane ring system. Both compounds are potent and selective mGlu2/3 receptor agonists with affinity in the low nanomolar range, and are systemically active. LY354740 and its prodrugs are currently under clinical development for the treatment of anxiety and panic disorders (see later). (1R,2S,5S,6S)-2-Amino6-fluoro-4-oxobicyclo[3.1.0]hexane-2,6-dicarboxylic acid monohydrate (MGS0028) is another potent and selective agonist of mGlu2/3 receptors. Compounds (2S,10 S,20 S)-2-(9-xanthylmethyl)-2-(20 -carboxycyclopropyl)glycine (LY341495) and (1R,2R, 3R,5R,6R)-2-amino-3-(3,4-dichlorobenzyloxy)-6fluorobicyclo[3.1.0]hexane-2,6-dicarboxylic acid (MGS0039) are mGlu2/3 receptor antagonists with nanomolar affinity. However, LY341495 recruits other mGlu receptor subtypes at micromolar concentrations and is considered as a pan-mGlu receptor antagonist. There are only a few tools that can help discriminate mGlu2 from mGlu3 receptors. No subtype-selective mGlu2 receptor agonists are available. N-Acetyl-aspartyl-glutamate (NAAG) is an endogenous compound that activates mGlu3 receptors and has no activity at mGlu2 receptors. However, NAAG displays a low potency and is also active at NMDA receptors. A growing number of positive allosteric modulators, such as 2,2,2-trifluoro-N-[3-(2-methoxyphenoxy)phenyl]-N-(3-pyridinylmethyl)-ethanesul fonamide (3-MP PTS) and 2,2,2-trifluoro-N-[3(cyclopentyloxy)phenyl]-N-(3-pyridinylmethyl)ethanesulfonamide (cyPPTS), selectively amplify responses mediated by mGlu2 receptors. No mGlu3 receptor enhancers are currently available. The development of mGlu2/3 receptor agonists or enhancers is aimed at the treatment of a number of psychiatric disorders, such as anxiety disorders, schizophrenia, and drug addiction. The agonist LY354740 is effective in all animal models of anxiety, and is also active in humans in preventing CO2-induced anxiety and panic attacks and relieving symptoms of generalized anxiety disorder (GAD). Remarkably, pharmacological activation of mGlu2/3 receptors does not produce adverse effects, such as sedation, ataxia, and dependence, that are typical of benzodiazepines and other CNS depressants, although it may disrupt memory processing. mGlu2 and mGlu3 receptors are highly expressed in limbic regions, and may regulate transmission at the synapses between the basolateral nuclei and the central nucleus of the amygdala. A final effect of LY354740 appears to be an enhanced GABAergic activity within the central nucleus of the amygdala, as shown by an increased Fos protein expression in GABAergic neurons. It appears that both mGlu2 and mGlu3 receptors are necessary for the anxiolytic effect of LY354740, although selective mGlu2 enhancers can also relieve anxiety.

mGlu2/3 receptor agonists or mGlu2 receptor enhancers are active in models predictive of antipsychotic activity, including the disruption of working memory, motor activity, sensory–motor gating (only mGlu2 enhancers), and glutamate efflux induced by phencyclidine or amphetamine. In addition, there is a functional antagonism between mGlu2/3 receptors and serotonin 5-HT2A receptors, which mediate the psychotomimetic action of LSD and other hallucinogenic drugs. Thus, there is a common pathway for the action of mGlu2/3 receptor agonists and atypical antipsychotic drugs, such as clozapine or olanzapine, which act as 5-HT2A receptor antagonists. Although there is no association of polymorphism in the mGlu2 receptor gene with schizophrenia, an increased expression of mGlu2/3 receptors is found in the prefrontal cortex of schizophrenic patients. Thus, one expects that the brain of schizophrenic patients is highly responsive to drugs that activate mGlu2/3 receptors. mGlu3 receptors are present in astrocytes and microglia, where their activation stimulates the synthesis of neurotropic factors, such as transforming growth factor-b (TGF-b), nerve growth factor (NGF), and brain-derived neurotropic factor (BDNF). An increased secretion of TGF-b from astrocytes mediates the neuroprotective activity of group II mGlu receptor agonists in culture. mGlu3 receptors are also found in neural stem cells and human glioma cells, and their blockade inhibits cell proliferation by dampening the stimulation of the MAPK and the PI3K pathways. Group II mGlu receptors have been recently shown to play a role in the synaptic adaptations that occur during the development of drug dependence. Activation of mGlu2/3 receptors negatively regulates brain reward function, as agonists elevate the intracranial self-stimulation threshold and decrease relapse to drug-taking behavior during abstinence, thus regulating the negative affective state observed during withdrawal. Finally, mGlu2/3 receptors mediate analgesic effects via a variety of mechanisms, including a reduced sensitivity of TRPV1 channels in peripheral nociceptors, and a reduced neurotransmitter release at the synapses between primary afferent fibers and second-order neurons in the dorsal horn of the spinal cord. L-Acetylcarnitine, a drug currently used for the treatment of painful neuropathies, induces the expression of mGlu2 receptors in dorsal root ganglia neurons by amplifying the activity of transcription factors of the nuclear factor-kappa B (NF-kB) family. mGlu4, mGlu7, and mGlu8 Receptors

The mGlu4 receptor consists of 912 amino acids in humans, rats, and mice; accession numbers are Q14833 (human), P31423 (rat), and Q68EF4 (mouse). The gene is named GRM4 (human) and

300 Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology

Grm4 (rat, mouse); chromosomal locations are 6p21.3 (human), 20p12 (rat), and 17 A3.3 (mouse). The mGlu7 receptor consists of 915 amino acids in humans, rats, and mice; accession numbers are Q14831 (human), P35400 (rat), and Q68ED2 (mouse). The gene is named GRM7 (human) and Grm7 (rat, mouse); chromosomal locations are 3p26.1 (human), 4q42 (rat), and 6 E3 (mouse). The mGlu8 receptor consists of 908 amino acids in humans, rats, and mice; accession numbers are O00222 (human), P70579 (rat), and P47743 (mouse). The gene is named GRM8 (human) and Grm8 (rat, mouse); chromosomal locations are 7q31.3–q32 (human), 4q22 (rat), and 6 A3 (mouse). These three receptor subtypes are preferentially, although not exclusively, located in the terminal region of the axons and share the ability to inhibit neurotransmitter release at various synapses in the CNS. Their location in the vicinity of the active zones of neurotransmitter release makes these subtypes as putative glutamate autoreceptors (Figure 2(b)). L-2-Amino-4-phosphonobutanoate (L-AP4) and the endogenous compound L-serine-O-phosphate (L-SOP) are prototypical orthosteric agonists of all group III mGlu receptors. (þ)-4-Phosphonophenylglycine (4-PPG) activates receptors with the following rank order of potency: mGlu8 > mGlu4 >> mGlu7. (S)3,4-Dicarboxyphenylglycine (DCPG) is an agonist with greater affinity for mGlu8 than for mGlu4, whereas (1S,3R,4S)-1-aminocyclopentane-1,3,4-tricarboxylic acid (ACPT-I) activates mGlu4 and mGlu8 receptors with equal potency. Neither DCPG nor ACPT-I has any effect at mGlu7 receptors. (R,S)-aCyclopropyl-4-phosphonophenylglycine (CPPG) and (S)-a-methyl-2-amino-4-phosphonobutanoic acid (M AP-4) are competitive antagonists with higher affinity for mGlu8 receptors. CPPG has no apparent activity at mGlu7 receptors. (–)-N-Phenyl-7-(hydroxyimino) cyclopropa[b]chromen-1a-carboxamide (PHCCC) is an enhancer with high selectivity for mGlu4 receptors. At high concentrations, PHCCC may also recruit mGlu1 receptors because of its similarity with CPCCOEt (see earlier). A selective agonist of mGlu7 receptors, N,N0 -dibenzhydrylethane-1,2diamine dihydrochloride (AM N082), has recently been synthesized. This compound is atypical because it does not bind to the VFT (like all classical mGlu receptor agonists), but to a site located in the 7TM domain of mGlu7 receptors. Activation of mGlu4 receptors is neuroprotective and mGlu4 receptor agonists/enhancers have been proposed as potential therapeutic agents in Parkinson’s disease for their ability to reverse parkinsonian symptoms in animal models. This action appears to be mediated by the inhibition of GABA release in

the GPe, which lies along the ‘indirect pathway’ of the basal ganglia motor circuit. Activation of mGlu4 receptors is also protective against experimental parkinsonism induced by MPTP in mice. mGlu4 receptors are highly expressed by cerebellar granule cells, and are also found in medulloblastoma cells, which derive from granule cell neuroprogenitors. Interestingly, activation of mGlu4 receptors inhibits proliferation of medulloblastoma cells and limits the spontaneous growth of medulloblastomas in the cerebellum of tumor-prone mice. The function of mGlu7 and mGlu8 receptors is only partially known. Both receptors have been proposed as drug targets in treatment of anxiety and stress-related disorders, but there are some caveats. The mGlu7 receptor agonist, AMN082, relieves anxiety, but genetic deletion of mGlu7 receptors does the same. In contrast, mGlu8 knockout mice show an increased level of anxiety, similar to that found in animals subjected to stressful conditions. Finally, mGlu4, mGlu7, and mGlu8 receptors have been considered as potential targets for the experimental therapy of epilepsy. Drugs that activate mGlu4 or mGlu8 receptors reduce generalized seizures in epilepsy-prone mice and rats and chemically induced seizures. In addition, mice lacking mGlu7 or mGlu8 receptors show an increased susceptibility to chemically induced seizures. Expression of mGlu4 receptors is increased in the dentate gyrus of mice with a low susceptibility to seizures and of patients with temporal lobe epilepsy. In contrast, activation of mGlu4 receptors is required for the induction of absence seizures in mice. mGlu6 Receptor

The mGlu6 receptor consists of 877 amino acids in humans and 871 amino acids in rats and mice; accession numbers are O15303 (human), P35349 (rat), and Q5NCH9 (mouse). The gene is named GRM6 (human) and Grm6 (rat, mouse); chromosomal locations are 5q35 (human), 10 (rat), and 11 B1.3 (mouse). This is a peculiar receptor; it is exclusively localized in the dendrites of ON bipolar cells of the retina and responds to the glutamate released from rod and cone photoreceptor cells in the dark. Activation of mGlu6 receptors leads to hyperpolarization of ON bipolar cells by closing cGMP-dependent cation channels. This effect may involve a receptor coupling with Go and activation of the Ca2þ-dependent protein phosphatase, calcineurin (Figure 2(c)). Light reduces glutamate stimulation of mGlu6 receptors, leading to opening of cation channels and depolarization of ON bipolar cells. Mice lacking mGlu6 receptors show a loss of ON response but unchanged responses to light. Interestingly, mutations in the gene encoding

Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology 301

the mGlu6 receptor are associated with night blindness and abnormal electroretinogram ON responses in humans. The mGlu6 receptor shows the same pharmacological profile of other group III mGlu receptors, being activated by L-AP4, L-SOP, and 4PPG, and antagonized by CPPG, MSOP, and MAP-4. See also: Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDAR-Dependent Forms; Metabotropic Glutamate Receptors (mGluRs): Functions.

Further Reading Alexander GM and Godwin DW (2006) Metabotropic glutamate receptors as strategic targets for the treatment of epilepsy. Epilepsy Research 71: 1–22. Bear MF, Huber KM, and Warren ST (2004) The mGluR theory of fragile X mental retardation. Trends in Neurosciences 27: 370–377. Bruno V, Battaglia G, Copani A, et al. (2001) Metabotropic glutamate receptor subtypes as targets for neuroprotective drugs. Journal of Cerebral Blood Flow and Metabolism 21: 1013–1033. Conn PJ, Battaglia G, Marino MJ, et al. (2005) Metabotropic glutamate receptors in the basal ganglia motor circuit. Nature Review Neuroscience 6: 787–798. DeBlasi A, Conn PJ, Pin J, et al. (2001) Molecular determinants of metabotropic glutamate receptor signaling. Trends in Pharmacological Sciences 22: 114–120. Fourgeaud L (2005) Addicted to Holmer? Journal of Neuroscience 25(42): 9555–9556. Kenny PJ and Markou A (2004) The ups and downs of addiction: Role of metabotropic glutamate receptors. Trends in Pharmacological Sciences 25: 265–272.

Kunishima N, Shimada Y, Tsugji Y, et al. (2000) Structural basis of glutamate recognition by a dimeric metabotropic glutamate receptor. Nature 407: 971–977. Pollock PM, Cohen-Solal K, Sood R, et al. (2003) Melanoma mouse model implicates metabotropic glutamate signaling in melanocytic neoplasia. Nature Genetics 34: 108–112. Nakanishi S (1992) Molecular diversity of glutamate receptors and implication for brain function. Science 258: 597–603. Nicoletti F, Wroblewski J, Iadarola MJ, et al. (1986) Excitatory amino acid recognition sites coupled with inositol phospholipid metabolism: Developmental changes and interaction with alpha 1-adrenoceptors. Proceedings of the National Academy of Sciences of the United States of America 83: 1931–1935. Pin JP, Kniazeff J, Liu J, et al. (2004) Allosteric functioning of dimeric class C G-protein-coupled receptors. FEBS Journal 272: 2947–2955. Schoepp DD, Jane DE, and Monn JA (1999) Pharmacological agents acting at subtypes of metabotropic glutamate receptors. Neuropharmacology 38: 1431–1476. Sladeczek F, Pin JP, Recasens M, et al. (1985) Glutamate stimulates inositol phosphate formation in striatal neurones. Nature 317: 717–719. Sugiyama H, Ito I, and Hirono C (1987) A new type of glutamate receptor linked to inositol phospholipid metabolism. Nature 325: 531–533. Swanson CJ, Bures M, Johnson MP, et al. (2005) Metabotropic glutamate receptors as novel targets for anxiety and stress disorders. Nature Review Drug Discovery 4: 131–144. Varney MA and Gereau RW 4th (2002) Metabotropic glutamate receptor involvement in models of acute and persistent pain: Prospects for the development of novel analgesics. Current Drug Targets CNS Neurological Disorders 1: 283–296.

Relevant Website http://www.iuphar-db.org – IUPHAR Receptor Database.

Metabotropic Glutamate Receptors (mGluRs): Functions B A Grueter and D G Winder, Vanderbilt University School of Medicine, Nashville, TN, USA

G-protein-independent ways through as yet unclear mechanisms.

ã 2009 Elsevier Ltd. All rights reserved.

Structure of mGluRs

Metabotropic Glutamate Receptors In the mid-1980s, glutamate was shown to stimulate inositol 1,4,5-trisphosphate (IP3) production, indicating the potential existence of a ‘metabotropic,’ or non-ion-channel-forming, glutamate receptor. This led to the distinction between ionotropic and metabotropic glutamate receptors (mGluRs). There are eight subclasses of mGluRs that have been identified and cloned. mGluRs belong to family C of the G-protein-coupled receptor superfamily, which includes calcium-sensing and g-aminobutyric acid B (GABAB) receptors and act to modulate neuronal excitability in many brain regions. Eight separate gene products (mGluRs 1–8) comprise the mGluR family of receptors, several of which can exist as alternatively spliced variants. These subtypes are classified into three groups based on pharmacology, similarities in coupling mechanisms, and sequence homology (Table 1). mGluRs of the same group show about 70% sequence identity, whereas between groups this percentage decreases to about 45%. Group I receptors (mGlu1 and mGlu5) are linked to Gq, whereas, group II (mGlu2 and mGlu3) and group III (mGlu4, mGlu6, mGlu7, and mGlu8) receptors are coupled to Gi/o. Intracellularly, the large C-terminus of mGluRs can interact with a variety of signaling systems. G-Protein-Coupled Receptors

G-protein-coupled receptors (GPCRs) are seventransmembrane-domain (7TM) proteins that interact with G-proteins. Receptor activation leads to associated guanosine triphosphatase (GTPase) activity and ultimately to activation or inhibition of an effector protein. G-protein regulation of effectors, particularly in neurons, occurs through two broad mechanisms. One is direct G-protein binding to ion channels. More traditionally, G-proteins function to activate or inhibit second-messenger-forming enzymes, leading to direct second messenger binding to ion channels, second messenger activation of kinases or phosphatases that phosphorylate/dephosphorylate ion channels, or alterations in protein– protein interactions involving an ion channel. These mechanisms constitute the major routes for mGluR signaling, although it should be noted that these receptors have also been reported to signal through

302

mGluRs are composed of two main domains, a large extracellular domain (ECD), and a 7TM region, and share very little sequence homology with other GPCRs. Proposed models confirmed by X-ray crystallography suggest the unique ECD contains two globular lobes separated by a cleft which contains the orthosteric ligand-binding site. While glutamate is commonly thought to be the primary ligand at this site, it is important to note that many excitatory amino acids are present in neural tissue, many of which can also activate mGluRs. In addition, allosteric binding sites have also been identified in the 7TM region. The second intracellular loop of mGluRs infers selectivity for G-protein coupling while a highly conserved third loop plays a crucial role in G-protein activation. Interactions among the first loop, third loop, and C-terminal tail are thought to control efficacy of G-protein coupling. The intracellular C-terminal region of mGluRs plays a major role in regulation and trafficking. Regulation of mGluR Signaling

mGluR signaling is tightly regulated by both homologous and heterologous mechanisms. Both agonistinduced (homologous) and heterologous forms of regulation of mGluRs have been observed. Phosphorylation by several different kinases can alter mGluR signaling. For instance, protein kinase C (PKC)dependent phosphorylation of mGluRs, particularly group I mGluRs, is thought to serve as a major mechanism of desensitization. b-Arrestin-induced desensitization of mGluRs has also been described. The bg subunits of G-proteins are thought to target G-protein-coupled receptor kinases (GRKs) to mGluRs, resulting in b-arrestin-mediated sequestration and ultimately desensitization. Group I mGluR expression at the plasma membrane and the functional coupling of group I mGluRs to intracellular signaling molecules are dependent upon scaffolding proteins. The Homer family of proteins is a major example of this. There are three genes encoding Homer proteins. Homer 1a is an immediate-early gene product while Homer 1b/c, Homer 2, and Homer 3 are constitutively expressed. Homer proteins link group I mGluRs through the C-terminal tail to signaling molecules such as the IP3 receptor, and to scaffolding components of the postsynaptic density, such as Shank. It is suggested that Homer

Metabotropic Glutamate Receptors (mGluRs): Functions

303

Table 1 Classification of mGluRs Family receptor

Coupling

Group/subtype-selective pharmacological agentsa

Group I mGluR1

Gq-coupled

mGluR5

Gq-coupled

Agonists: DHPG, ACPD, quisqualate Antagonist: LY393675 Inverse agonist (allosteric antagonist): LY367385 Agonists: DHPG, ACPD, quisqualate, CHPG Inverse agonist (allosteric antagonist): MPEP

Group II mGluR2

Gi/o-coupled

mGluR3

Gi/o-coupled

Group III mGluR4

Gi/o-coupled

mGluR6

Gi/o-coupled

mGluR7

Gi/o-coupled

mGluR8

Gi/o-coupled

Agonists: DCG-IV, LY354730, ACPD Antagonist: LY341495 Agonists: DCG-IV, LY354730, ACPD Antagonist: LY341495 Agonists: L-AP4, L-SOP Antagonists: MSOP, MAP4 Agonists: L-AP4, L-SOP Antagonists: MSOP, MAP4 Agonists: L-AP4, L-SOP Antagonists: MSOP, MAP4 Agonists: L-AP4, L-SOP, DCPG Antagonists: MSOP, MAP4

a ACPD, (1S,3R)-1-amino-1,3-cyclopentanedicarboxylate; L-AP4, L-2-amino-4-phosphonobutyric acid; CHPG, (R,S)-2-chloro-5-hydroxyphenylglycine; DCG-IV, (2S,20 R,30 R)-2-(20 ,30 -dicarboxycyclopropyl)glycine; DCPG, (S)-3,4-dicarboxyphenylglycine; DHPG, 3,5-dihydroxyphenylglycine; MAP4, a-methyl-2-amino-4-phosphonobutyrate; MPEP, 2-methyl-6-(phenylethynyl)pyridine; MSOP, methylserine-O-phosphate; L-SOP, L-serine-O-phosphate.

proteins may regulate the expression and function of group I mGluRs at multiple levels, including targeting, surface expression, clustering, physical linkage to other synaptic and subsynaptic complexes, and modulation of constitutive activity. Interestingly, biochemical evidence suggests the C-terminus of Homer 2a interacts with the Rho family of small GTPases, in a GTP-dependent manner. Homer proteins have also been shown to form complexes with Shank proteins, which act as scaffolding proteins and link group I mGluRs with other proteins in the postsynaptic densities (PSDs), such as Dynamin2, an important molecule implicated in endocytosis, and a-amino-3hydroxy-5-methyl-4-isoxazole propionic acid receptor (AMPAR)–glutamate receptor-interacting protein (GRIP) complexes. Expression of mGluRs

mGluRs have unique but overlapping distributions throughout the central nervous system (CNS). In addition, mGluRs have been reported to function in the peripheral nervous system and elsewhere. For example, mGluRs have been shown to play an important role in nociceptive transmission, as will be described in a later section. Glutamate Sources from Which mGluRs Can Be Activated

mGluRs are often located in the areas adjacent to the synapse. Functionally, their location suggests that they

become activated in situations of repeated stimulation of afferents, which results in considerable glutamate accumulation in the synaptic cleft coupled to spillover to extrasynaptic sites. In addition to the frequency and duration of synaptic activity, the extracellular concentration of glutamate also depends on the efficacy of neuronal and glial uptake, which is dictated by the number of glutamate transporters (GluTs). GluTs, along with passive diffusion, are the primary means of glutamate removal from the extracellular space. The ability of mGluRs to participate in signaling is thought to be tightly regulated by GluT activity. Another potentially important contributor to extracellular glutamate levels is the cystine–glutamate exchanger. The nonsynaptic cystine–glutamate antiporter exchanges extracellular cystine for intracellular glutamate with a minimal contribution of synaptic glutamate release. It has been suggested that one form of cocaine-induced neuroadaptation may be due to a decrease in basal levels of extracellular glutamate in the N-acetylcysteine (NAc) after several weeks of withdrawal from cocaine, an effect attributed to a reduced function of the cystine–glutamate transporter.

mGluR Function Neuromodulation

Neuromodulation is a subtle influence on synaptic efficacy or neuronal excitability. Neuromodulation

304 Metabotropic Glutamate Receptors (mGluRs): Functions

can also involve longer-lasting changes such as alterations in gene expression. The ‘fine tuning’ of neuronal function by neuromodulation makes receptors and pathways attractive therapeutic targets for many diseases and disorders. mGluR activation serves as a major source by which glutamatergic transmission can modulate synaptic efficacy, neuronal excitability, and gene expression. Synaptic Plasticity

Persistent alterations in synaptic efficacy initiated by transient stimuli are referred to as forms of synaptic plasticity. Changes in synaptic efficacy can occur by presynaptic mechanisms such as altered neurotransmitter release as well postsynaptic mechanisms. Activation of mGluRs can play an important role in facilitation of multiple forms of synaptic plasticity. mGluR activation can induce long-term depression (LTD) at many glutamatergic synapses. Further activation, primarily of group I mGluRs at some synapses, can positively modulate N-methyl-D-aspartate receptor (NMDAR)-dependent long-term potentiation (LTP) induction.

Group I mGluRs Group I receptors (mGluR1 and mGluR5) are linked to activation of phospholipase C (PLC), leading to phosphoinositide hydrolysis, calcium release, and PKC activation. Activation of group I mGluRs can also indirectly regulate adenylyl cyclase, leading to increased cyclic adenosine monophosphate (cAMP) formation, as well as cyclic guanosine monophosphate (cGMP) accumulation, phospholipase A, and extracellular signal-regulated kinase (ERK) activation. Although highly homologous, studies suggest mGluR1 and mGluR5 receptors have distinct functions in regulating synaptic transmission. This implies differential regulation of these receptors and potential divergence in effector systems. Conversely, at some synapses only one type of group I mGluR is functionally important: for instance, only mGluR1 has been shown to function in the ventral tegmental area and cerebellum. Synaptic Locus of Group I mGluRs

Mechanisms underlying group I mGluR function vary across brain regions and synapses. Much information on the function of these receptors can be gained by determining the synaptic locus of action of these receptors. Immunohistochemical and electrophysiological results reveal that group I mGluRs localize and function presynaptically, postsynaptically, and on glia. However, group I mGluRs are predominantly

located at the periphery of the postsynaptic density (perisynaptically). Activation of group I mGluRs causes postsynaptic effects such as neuronal depolarization, excitation, and spike frequency adaptation in a number of brain regions. For instance, selective activation of group I mGluRs with the agonist 3,5-dihydroxyphenylglycine (DHPG) increases postsynaptic membrane excitability in hippocampus, cortex, striatum, amygdala, subthalamic nucleus, and hypothalamic nuclei. Increases in neuronal excitability via activation of group I mGluRs are thought to occur in large part through modulation of Kþ channels. Specific activation of group I mGluRs induces net inward currents by inhibiting Kþ channel conductance in neurons in the hippocampus and hypothalamus. For instance, Kþ channels are a major target of mGluRs, and many diverse types of Kþ channels are inhibited following activation of mGluRs. Activation of mGluRs has been shown to block the IAHP (IAHP is an outward, Ca2þ-activated, Kþ current with slowly activating and inactivating (2 s) kinetics) and to thereby reduce accommodation of spike firing in hippocampal CA1 pyramidal cells, CA3 pyramidal cells, the dentate gyrus, cultured cerebellar Purkinje cells, and basolateral amygdaloid nucleus. Activation of group I mGluRs has also been shown to modulate voltage-gated calcium channels (VGCCs). Alterations in VGCCs can occur either presynaptically, consequently changing vesicular release probability, or postsynaptically, leading to changes in intracellular calcium levels. Although mGluR-dependent modulation of channels has been shown to be a consequence of direct G-protein-linked action – for example, inhibition of Ca2þ channels – many other effects occur as a result of activation of intracellular messenger pathways. Group I, in contrast to group II, mGluRs can use several distinct signal transduction pathways to inhibit Ca2þ channels, both Ca2þ intracellular-independent and -dependent mechanisms. In addition to the postsynaptic modulation of synaptic transmission by alteration of neuronal excitability, group I mGluRs have been reported to function presynaptically. One example is in the hippocampus, at the level of the synapses between Schaeffer collaterals and CA1 pyramidal cells. Consistent with presynaptic function, an increase in paired pulse facilitation has been observed upon activation of mGluR5, suggesting a decrease in vesicular release probability. Retrograde Signaling

Cannabinoids Of great interest in the mGluR field was the finding that endocannabinoid synthesis can be

Metabotropic Glutamate Receptors (mGluRs): Functions

triggered by activation of group I mGluRs. Retrograde signaling through the endocannabinoid system helped explain biophysical evidence suggesting a presynaptic change in probability of neurotransmitter release, with immunochemical data suggesting that group I mGluRs are primarily localized to postsynaptic structures. Use of pharmacological tools as well as genetic manipulations has provided evidence for a role of arachidonic acid metabolites – more specifically, the 12-lipoxygenase metabolite of arachidonic acid, 12(S)hydroxyeicosa-5(Z),8(Z),10(E),14(Z)-tetraenoic acid (HETE) – in inducing group I mGluR LTD in the hippocampus. Direct application of this metabolite mimicked and occluded mGluR5-dependent LTD at these synapses; mGluR5-dependent LTD in the hippocampus is absent in mice lacking the ‘leukocyte type’ 12-lipoxygenase, and group I mGluR LTD-inducing stimuli promote 12-lipoxygenase activity. A suggested potential downstream target of 12-lipoxygenase is p38 kinase. p38 kinase, like ERK, is a member of the mitogen-activated protein kinase (MAPK) family and has been implicated in group I mGluR LTD. For instance, in the dentate gyrus, group I mGluR LTD is dependent on p38 kinase signaling and in the CA1 of the hippocampus activation of group I mGluRs results in AMPA receptor endocytosis (which will be discussed in greater detail in a subsequent section) and is dependent on p38 kinase signaling as well as protein synthesis. Kinases Involved in Group I mGluR Function

Group I mGluRs are linked to the activation of multiple kinases. Consistent with coupling of group I mGluRs with Gq and Ca2þ signaling, activation of these receptors can lead to activation of the Ca2þ/diacylglycerol (DAG)-dependent PKC family. mGluR-induced activation of PKC can lead to phosphorylation of a multitude of downstream targets, ultimately resulting in neuronal modifications such as LTP or LTD. One such target of mGluR-stimulated PKC activity is the MAPK/ERK pathway. The ERK pathway is a signaling cascade that plays a crucial role in a variety of cell regulatory events, including cell proliferation, differentiation, and survival, as well as an involvement in long-term synaptic changes and behavior. In neurons, ERK activation has been shown to be involved in processes associated with synaptic remodeling and long-term changes in synaptic efficacy. These processes include protein synthesis, changes in gene expression, dendritic spine stabilization, modulation of ion channels, and receptor insertion. The ERK cascade is activated by a variety of extracellular agents, including growth factors, hormones, and neurotransmitters, and serves as an important

305

point of convergence for the PKC and protein kinase A (PKA) pathways. PKC regulates ERK activity through an interaction with either Ras or Raf-1, leading to activation of mitogen-activated protein extracellular kinase (MEK) and consequently ERK. Additionally, it was discovered that DAG, another second messenger product of PLC activity, is capable of activating ERK independent of PKC activation. Consistent with a role for PKA and PKC in some forms of group I mGluR LTD, group I mGluR LTD in the hippocampus, bed nucleus of the stria terminalis (BNST), and cerebellum is dependent on ERK activation. In addition to these kinases, receptor tyrosine phosphatases have also been shown to play an integral role in the group I mGluR-induced removal of AMPARs from the synapse, resulting in LTD of excitatory synaptic transmission. Pharmacology of Group I mGluRs

The recent development of mGluR-selective ligands (e.g., phenylglycine derivatives acting either as agonists or as antagonists) has helped begin the elucidation of the functional roles of mGluRs in brain and behavior. The pharmacological profile of group I mGluRs has been determined in mammalian heterologous expression systems on cloned mGluRs. The rank order of potency of the most common agonists is quisqualate > DHPG ¼ glutamate > (1S,3R)-1amino-1,3-cyclopentanedicarboxylate (ACPD). The specific agonist for group I mGluRs used in the studies presented in the following sections is DHPG, which is devoid of activity at other mGluRs. An agonist for mGluR5 – (R,S)-2-chloro-5-hydroxyphenylglycine (CHPG) – has been reported, although it is not very potent. The first antagonist described for group I mGluRs was a-methyl-4-carboxyphenylglycine (MCPG). However, MCPG also antagonizes group II mGluRs. There are several drugs that selectively inhibit mGluR5; the most widely used is the noncompetitive antagonist 2-methyl-6-(phenylethynyl)pyridine (MPEP), with the caveat that at high concentrations it blocks NMDA receptors. Group I mGluRs and Plasticity

Group I mGluRs have been shown to potentiate either NMDAR- or AMPAR-mediated responses. However, a major function of group I mGluR activation is to produce a persistent weakening of glutamatergic transmission at synapses in the CNS. There appear to be at least two mechanisms through which this is accomplished. One involves the recruitment of endocannabinoid signaling and presynaptic alterations, and has been observed in the dorsal and ventral striatum.

306 Metabotropic Glutamate Receptors (mGluRs): Functions

A second major mechanism is through endocannabinoid-independent signaling mechanisms, and has been heavily studied in the hippocampus and cortical regions. Both presynaptic and postsynaptic mechanisms have been described for endocannabinoidindependent forms of group I mGluR LTD. Group I mGluRs Can Alter iGluR Function

Group I mGluRs have been shown to modulate iGluRs in several ways. Activation of mGluRs can lead to changes in neuronal excitability and therefore influence NMDAR activity. For instance, membrane depolarization via group I mGluR stimulation can enable enhanced NMDAR-mediated synaptic plasticity. Additionally, through activation of kinases/phosphatases, group I mGluRs can alter the signaling properties of iGluRs and can influence trafficking of these receptors into and out of the synapse. PKC phosphorylates NMDA and AMPA receptors. The activation of PKC by mGluRs might therefore modulate the phosphorylation state of these two ionotropic receptors, resulting in a potentiation of the response. Activation of group I mGluRs can lead to a removal of AMPA receptors from the membrane, resulting in a depression of synaptic transmission. More specifically, following activation of group I mGluRs there is a reduction in the density of postsynaptic AMPARs.

mGluR2 and mGluR3, with an EC50 of 1050 nM in the rat cortex, hippocampus, and striatum, and 10 or 30 nM in cells expressing recombinant mGluR2 or mGluR3, respectively. LY341495 ((2S)-2-amino-2-[(1S,2S)-2carboxycycloprop-1-yl]-3-(xanth-9-yl) propanoic acid) is an antagonist of mGluRs, with nanomolar potency against group II mGluRs and micromolar potency against group III mGluRs. LTD of excitatory transmission can be induced by activation of group II mGluRs in several brain regions. For example, group II mGluR LTD has been shown presynaptically in the basolateral amygdala (BLA), the nucleus accumbens, the BNST, and the striatum. In contrast, it was shown that stimulation of thalamic inputs to the lateral nucleus of the amygdala induces a group II mGluR-mediated postsynaptic LTD. Additionally, in the dentate gyrus and the medial PFC it was found that group II mGluR activation induced a postsynaptic LTD that was PKA and PKC dependent. Behavioral studies of mice lacking mGluR2 or mGluR3 support earlier findings of a role of these receptors in anxiety behaviors. Consistent with the role of these receptors in various animal models of anxiety/stress, group II mGluR agonists have been considered promising candidates as therapeutic agents for the treatment of anxiety as well as other brain disorders.

Group III mGluRs Group II mGluRs Group II mGluRs (mGluR2/3) are widely distributed throughout the CNS, where they are moderately to highly expressed in brain regions that are commonly associated with anxiety disorders, including the hippocampus, prefrontal cortex (PFC), and amygdala. mGluR2 is generally localized at the periphery of the presynaptic terminal, suggesting a role for these receptors at instances of spillover of glutamate beyond the synaptic cleft in response to repetitive stimulation. By contrast, mGluR3 is more diversely localized, including both pre- and postsynaptic localization on neurons, as well as relatively heavy localization on glial cells. Group II mGluRs most commonly function presynaptically to decrease the probability of synaptic vesicle release, but at multiple synapses within the CNS postsynaptic mechanisms have been reported. Voltagedependent Ca2þ channels are likely to be involved in the presynaptic inhibition mediated by mGluRs. Additionally, it has been suggested that Kþ channels mediate some of the effects induced by activation of group II or III mGluRs. LY354740 ((1S,2S,5R,6S)-(þ)-2-aminobicyclo[3.1.0] hexane-2,6-dicarboxylic acid) and several related compounds are potent and selective agonists at

Like group II mGluRs, group III mGluRs are coupled to inhibition of cAMP production, and modulation of ion channels. However, in contrast to the other groups, group III mGluRs are primarily located at presynaptic active zones at the axon terminal. As with the group II mGluRs, the group III mGluRs inhibit neurotransmitter release, resulting in a reduction in glutamatergic or GABAergic synaptic transmission. Unfortunately, there are few pharmacological tools with the desired receptor subtype selectivity ideal for testing the significance of the subtypes of group III mGlu receptors. L-2-Amino-4phosphonobutanoate (L-AP4), the most commonly used selective group III agonist, has a high affinity for mGluR4, mGluR6, and mGluR8 and a low affinity for mGluR7. One of the most promising drugs available for group III mGluRs is the positive allosteric modulator of mGluR4, (–)-N-phenyl-7(hydroxyimino)cyclopropa[b]chromen-1a-carboxamide (PHCCC). (S)-3,4-Dicarboxyphenylglycine (DCPG), a more recently described group III mGluR agonist, is a relatively specific mGluR8 agonist. While group III mGluR function has been reported in brain regions such as the BLA, central nucleus of the amygdala, BNST, and hippocampus using an agonist

Metabotropic Glutamate Receptors (mGluRs): Functions

pharmacology approach, behavioral analysis of specific receptor mutant mice has also proved useful. In combination with the limited pharmacology, mutants of mGluR4, mGluR7, and mGluR8 have all been shown to have behavioral phenotypes. For instance, consistent with a dysregulation of the hypothalamic– pituitary–adrenal (HPA) axis, mice with targeted deletion of mGluR7 show changes in animal behavior paradigms predictive of antidepressant and anxiolytic action. They also exhibit a deficit in fear responses and learning paradigms. Interestingly, administration of the mGluR4 allosteric potentiator, PHCCC, also has anxiolytic-like effects, implicating a role for this receptor in anxiety as well. mGluR4 knockout (KO) mice have altered glutamate and GABA release and, interestingly, mGluR4 activation protects against neurodegeneration. mGluR8 has been implicated in conditioned fear, and similar to the other group III mGluRs, mice lacking mGluR8 have an increased anxiety phenotype. Of the mGluRs, mGluR6 has the most restricted expression. These receptors are expressed postsynaptically at ON bipolar cells in both rod and cone systems and function as the main excitatory receptor responsible for relaying synaptic signals in these retinal cells. Consistent with their important functional role in visual processing, genetic targeting of mGluR6 impairs detection of visual contrasts.

Diseases Many psychiatric as well as neurological diseases have been linked to alterations in neuronal excitability via the glutamatergic system. As mGluRs function to regulate a large number of synapses, and are known to play a role in pathophysiological conditions, compounds that act on these receptors are prime targets for a number of therapeutic applications. For instance, drugs targeting mGluRs have been shown to prevent certain forms of pain and anxiety. There are positive potential therapeutic effects of mGluR compounds in treating Parkinson’s disease as well as the conditions of drug dependence and withdrawal. Also, activating or potentiating mGluRs may help in treatment of schizophrenia and Alzheimer’s disease. An interesting model suggests a specific role for group I mGluRs in fragile X mental retardation. Intense investigation reveals a role for mGluRs in nociception both centrally and peripherally. For instance, group I mGluRs have been implicated in the processes of central sensitization and persistent nociception, whereas activation of group II mGluRs is effective against neuropathic or inflammatory pain.

307

Pharmacology/Therapeutic Utility of mGluRs An exciting new direction in the pharmacological targeting of mGluRs has been the development of subtype-selective allosteric modulators. An advantage of allosteric modulators over traditional compounds is that allosteric modulators rely on activation by the endogenous transmitter released in an activity-dependent manner. mGluRs are prominent targets as both positive and/or negative allosteric modulators with unique pharmacological properties. Allosteric modulators, as compared to orthosteric ligands, appear to show a unique degree of selectivity. See also: Glutamate; Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDARDependent Forms; Long-Term Depression (LTD): Endocannabinoids and Cerebellar LTD; Long-Term Potentiation (LTP): NMDA Receptor Role.

Further Reading Anwyl R (1999) Metabotropic glutamate receptors: Electrophysiological properties and role in plasticity. Brain Research Reviews 29: 83–120. Bordi F and Ugolini A (1999) Group I metabotropic glutamate receptors: Implications for brain diseases. Progress in Neurobiology 59: 55–79. Conn PJ and Pin JP (1997) Pharmacology and functions of metabotropic glutamate receptors. Annual Review of Pharmacology and Toxicology 37: 205–237. DeBlasi A, Conn PJ, Pin J, et al. (2001) Molecular determinants of metabotropic glutamate receptor signaling. Trends in Pharmacological Sciences 22: 114–120. Pin JP and Duvoisin R (1995) The metabotropic glutamate receptors: Structure and functions. Neuropharmacology 34: 1–26. Rouse ST, Marino MJ, Bradley SR, et al. (2000) Distribution and roles of metabotropic glutamate receptors in the basal ganglia motor circuit: Implications for treatment of Parkinson’s disease and related disorders. Pharmacology & Therapeutics 88: 427–435. Schoepp DD (2001) Unveiling the functions of presynaptic metabotropic glutamate receptors in the central nervous system. Journal of Pharmacology and Experimental Therapeutics 299: 12–20. Sladeczek F, Pin JP, Recasens M, et al. (1985) Glutamate stimulates inositol phosphate formation in striatal neurones. Nature 317: 717–719. Sugiyama H, Ito I, and Hirono C (1987) A new type of glutamate receptor linked to inositol phospholipid metabolism. Nature 325: 531–533. Tanabe Y, Masu M, Ishii T, et al. (1992) A family of metabotropic glutamate receptors. Neuron 8: 169–179.

Kainate Receptors: Molecular and Cell Biology J Lerma, Instituto de Neurociencias de Alicante Consejo Superior de Investigaciones Cientı´ficas – Universidad Miguel Hernandez, San Juan de Alicante, Spain ã 2009 Elsevier Ltd. All rights reserved.

It is now well established that there are three types of ionotropic glutamate receptors (GluRs), whose nomenclature reflects their preferred ligands: AMPA (a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid) receptors, NMDA (N-methyl-D-aspartate) receptors, and kainate receptors (KARs). The cloning of many GluRs at the beginning of the 1990s and the discovery of their structural relationships confirmed the legitimacy of this pharmacological subdivision. Likewise, their cloning has enabled us to advance our understanding of the biophysical properties and the physiological role of each receptor subtype in the mammalian brain. Indeed, analyzing just a few of these GluR subunits emphasized that true KARs are ion channels with a stronger preference for this agonist rather than for AMPA and that they display rapidly desensitizing responses. Thus, defining the molecular biology of KAR subunits represents a real breakthrough in the study of these receptors, establishing the foundations for us to better understand their physiology.

The Diversity of Kainate Receptors It is now accepted that all ionotropic GluRs share a conserved transmembrane topology and stoichiometry. Like the AMPA receptors (AMPARs) and NMDA receptors (NMDARs), KARs are tetrameric combinations of the GluR5, GluR6, GluR7, KA1, and KA2 subunits. Of these, GluR5/6/7 can form functional homomeric or heteromeric receptors, whereas KA1 and KA2 participate in functional receptors only when accompanied by GluR5/6/7 subunits. In the ionotropic GluRs, each monomer carries its own ligand binding site and it contributes with a specific amino acid stretch to the channel lumen. In addition to this hydrophobic sequence, each subunit has three transmembrane segments (M1, M3, and M4) arranged in such a manner that the N-terminal domain of each protein lies extracellularly and the C-terminal region lies within the cell. As in the case of AMPARs, the glutamate binding site of KARs is formed by residues that are distributed throughout the distal N-terminal domain (called S1) and the loop between M3 and M4 (called S2). Indeed, this latter segment contains specific residues that confer

308

the receptor with high or low sensitivity to various agonists (AMPA and kainate), as well as to ions that participate in the channel gating. According to the rules governing subunit assembly and the different arrangements that give rise to functional receptors, a wide diversity of KARs may be expressed in the brain. Of the five KAR subunits that have been cloned so far, some combinations do not render functional channels and others are not specific to kainate. For example, the GluR5 subunit forms receptors that can be activated not only by glutamate but also by kainate, domoate, and AMPA. When activated by kainate these receptor channels desensitize relatively slowly, whereas when they bind glutamate their desensitization is faster and almost complete. In contrast, GluR6 subunits form homomeric channels with fast-desensitizing kinetics when activated by either agonist. Unlike the GluR5 receptors, GluR6 homomeric channels are not sensitive to AMPA or to its derivative ((RS)-2-amino-3(3-hydroxy-5-tert-butylisoxazol-4-yl) propanoic acid (ATPA) (Figure 1(b)). Functional homomeric receptors can also be formed by GluR7 subunits, these displaying a very low affinity for glutamate and complete insensitivity to AMPA and domoate. Each of these three subunits also form heteromeric assemblies with KA1 or KA2, and the resulting heteromeric receptors present slightly different biophysical and pharmacological channel properties. Moreover, GluR5, GluR6, and GluR7 may also combine with one another to form functional receptors. The sequence alignment reveals a homology of 75–80% between GluR5, GluR6, and GluR7, whereas KA1 and KA2 present 68% of similarity. Homology, however, drops to 45% when these two subfamilies of KARs are considered. The homology of KAR subunits with AMPARs is also approximately 40%, although they are not able to form heteromeric receptors.

Kainate Receptor Subunits Are Subject to Alternative Splicing In addition to the inherent variation in KARs, alternative splicing generates a number of different isoforms of KAR subunits (Figure 1(a)). Indeed, two different GluR5 variants have been found, one that contains an extra segment of 15 amino acids in the N-terminal region (GluR5-1) and another that lacks this insert (GluR5-2). Furthermore, the GluR5-2 subunit presents four different C-terminal domains. The C-terminus of the 2a variant is 49 amino acids shorter than the variant originally described (which was then renamed 2b) due to the introduction of a premature stop codon. Variant 2c contains an extra in-frame exon

Kainate Receptors: Molecular and Cell Biology 309 GluR6 N-term

M1 M2 M3 S1 I/V Y/C

M4

K823

ATPA 1 mM

54 aa

S2

Q/R

Glu 1 mM

C-term GluR6a

20 pA

GluR6b

15 aa

200 ms GluR5 N-term GluR5-2 GluR5-1

15 aa insert

Glu 1 mM

Q824 C-term M4 HY826

M1 M2 M3 Q/R Q

-2a

CLS....

..VA875

CLS....

..VA904

29 aas

-2b 20 pA

-2c

200 ms

GluR7 N-term

M1 M2 M3

Q825 M4

Glu 30 mM C-term 64 aa

Q 13 aa insert

42 aa

a

300 pA 50 ms

GluR7b

M4

AMPA 1 mM

C-term

Dom 100 µM 200 pA

2s

GluR6/KA2 ATPA M1 M2 M3

50 pA 200 ms

300 pA 50 ms

GluR7a

KA1 and KA2 N-term

ATPA 100 µM

ATPA 100 µM

1 µM

Q

10 µM b

10 pA 1s

100 pA

Figure 1 Kainate receptor (KAR) subunits: (a) structure of subunits; (b) electrophysiological recordings of responses to different agonists of the corresponding homomeric receptors recombinantly expressed in HEK293 cells. In (a), the KAR subunits are all made of a chain of approximately 900 amino acids (100 kDa) that spans the membrane three times (hydrophobic segments M1, M3, and M4). Thus, the N-terminal domain of the protein is extracellular and the C-terminal domain lies intracellularly. The M2 hydrophobic sequence does not cross the membrane, but it is thought to dip into the membrane forming a loop and forming the pore of the channel. Two isoforms of GluR6 subunits exist, -a and -b, differing in their C-terminal domain. The GluR5 can be found as two different variants: GluR5-1, which contains 15 extra amino acids in the N-terminal region, and GluR5-2, which lacks this insert. GluR5-2 presents three additional possibilities generated by a splice of the C-terminal domain. GluR5-2a is 49 amino acids shorter than the GluR5-2b because of the introduction of a premature stop codon. GluR5-2c contains an extra in-frame exon that makes it 29 amino acids longer than the 2b isoform. GluR7 can be present as two isoforms with different C-terminals due to a 13-amino-acid insertion (blue segment), which produces a new open reading frame and a completely different C-terminus (red segment). In contrast, no splice variants have been found for the KA1 and KA2 subunits. GluR5 and GluR6 subunits undergo mRNA editing at the M2 (Q/R site), generating a mixture of edited and nonedited subunits. GluR6 presents two additional editing sites in M1. In every case, a single receptor channel is believed to be composed of four subunits. In all KARs, the binding domain is formed by two peptide segments, S1 and S2, which are situated before the M1 (S1) and linking M3 and M4 (S2). In (b), responses of GluR6 are also shown when this subunit heteromerizes with KA2 subunits (bottom). GluR, glutamate receptor. ATPA, (RS)-2-amino-3(3-hydroxy-5-tert-butylisoxazol-4-yl) propanoic acid; Dom, domoic acid.

that makes it 29 amino acids longer than the 2b isoform. Finally, the last 15 amino acids of variant 2d (not shown), from Q824 onward, are completely unrelated to any other variant, and significantly this last isoform was isolated from human tissue. Two splice variants of the murine GluR6 have also been found and, again, these differ in their C-terminal domains. Furthermore, in the rats and humans two splice variants of the GluR7subunit have been discovered. In this case,

the insertion of 40 nucleotides out-of-frame in the C-terminal sequence of GluR7a leads to the production of a shorter protein, GluR7b, which bears no significant homology in this region to any of the known ionotropic GluRs. In contrast, the KA1 and KA2 subunits do not seem to undergo alternative splicing. Despite the existence of this repertoire of isoforms, the role of the different KAR splice variants (e.g., in receptor trafficking) is just starting to be determined.

310 Kainate Receptors: Molecular and Cell Biology

Some Kainate Receptor Subunits Undergo mRNA Editing The structural repertoire of KAR subtypes is even larger because the pre-mRNA of the KAR subunits GluR5 and GluR6 can undergo editing at the so-called Q/R site of the second membrane domain (M2). As in AMPARs, the Q to R substitution in homomeric KARs assembled either from GluR5 or GluR6 decreases the permeability of the receptor to calcium. Furthermore, this substitution transforms the rectification properties of these receptors from inwardly rectifying to linear or slightly outwardly rectifying, and it reduces the unitary conductance of the channels. As in AMPARs, inward rectification arises from the blockade of the ion channel by internal polyamines, such that the presence of an arginine (R) residue not only drastically reduces calcium permeability but also abolishes ion channel block by polyamines. In addition, GluR6 alone presents two further sites in the first transmembrane domain (M1) that are susceptible to editing, where I567 and Y571 could be changed to V and C, respectively. Although the role of these M1 editing sites remains unclear, editing within the M2 domain of GluR6 KARs has recently been shown to exert a significant effect on synaptic physiology.

Kainate Receptors Are Widely Distributed throughout the Brain Most of the studies defining the distribution of KARs in the brain come from in situ hybridization analysis of mRNA. KAR subunits are widely expressed throughout the nervous system, although there are a few known cell types that exclusively express a given subunit. For instance, GluR5 and KA2 are the only subunits that have been detected in the dorsal root ganglion (DRG) neurons. GluR5 transcripts are also abundant in the Purkinje cells of the cerebellum as well as in the subiculum, the piriform and cingulate cortices, and the hippocampal and cortical interneurons. On the other hand, GluR6 is the most abundant subunit in cerebellar granule cells, the striatum, the dentate gyrus, and the CA1 and CA3 regions of the hippocampus. The mRNA encoding GluR7 is expressed at low levels throughout the brain, but it is particularly prominent in the deep layers of the cerebral cortex, in the striatum, and in the inhibitory neurons of the molecular layer of the cerebellum. The KA1 subunit is present at low levels in the dentate gyrus, in the amygdala, and in the entorhinal cortex, being more abundant in the CA3 region. In contrast, KA2 mRNA can be found in virtually every part of the nervous system. Although the different KAR subunits are already present in the embryo, most of these expression

patterns emerge during the early postnatal period. Thus, the amount of GluR5 mRNA peaks between P0 and P5, and it then begins to decline toward the levels found in adults by P12. Similarly, the patterns of GluR6 and KA1 expression observed within the hippocampal formation begins to change at around P0 (GluR6) and P12 (KA1). We are still awaiting the development of specific antibodies against various KAR subunits. Although in situ hybridization is informative, it is not capable of revealing the subcellular distribution of a given subunit. Relatively specific antisera against the KAR subunits GluR6/7 and KA2 have been used to detect KARs, and to date, the most convincing studies have been carried out in the rat retina. In this tissue, the localization of these subunits at the synapse has been investigated by electron microscopy and the selective synaptic distribution of KARs in both plexiform layers of the retina has been observed. Whereas GluR6/7 could be found in horizontal cell processes postsynaptic to both rod spherules and cone pedicles, the KA2 subunit was found only in postsynaptic densities of the cone pedicles in the dendrites of off-bipolar cells. Anti-GluR6/7 and less specific anti-GluR5/6/7 antibodies have also been used to study other neurons in the brain. Although the limitation of these antibodies, particularly the latter, must be taken into account when interpreting the data obtained, they appeared to map KARs to fibers and synaptic terminals, as well as to dendrites and postsynaptic membranes. Hence, it seems that KARs may be found in both pre- and postsynaptic locations. In particular, GluR5/6 and KA2 have been found on parallel fibers and GluR5/6 immunoreactivity has been identified in a large population of terminals that form axospinal and axodendritic asymmetric synapses in the monkey striatum.

Regulation of the Surface Expression of Kainate Receptors Although more recent experiments have been directed toward understanding how KARs are targeted to specific membrane domains, the mechanisms controlling surface expression remain unclear. Retention in the endoplasmic reticulum (ER) plays a rate-limiting role in KAR trafficking. As in most GluRs, the residues important for ER retention or exit are proximal to Psd-95, Dlg, and ZO1 (PDZ) binding motifs present in the C-terminal domain. A number of proteins containing PDZ motifs have been shown to interact with KAR subunits, although most of these are promiscuous, also binding to AMPARs and, possibly,

Kainate Receptors: Molecular and Cell Biology 311

NMDARs. These include postsynaptic density protein (PSD-)95/ synapse-associated protein (SAP-)90, CASK, glutamate receptor interacting protein (GRIP), protein interacting with C kinase 1 (PICK1), and syntenin. However, PDZ proteins such as PICK1, GRIP, and PSD-95 do not appear to play a significant role in the exit of GluR5 or GluR6 from the ER but, rather, they have been implicated in the insertion and/or maintenance of AMPARs and KARs at the synapse. Indeed, it was recently shown that the number of KARs in the membrane surface can rapidly be altered through protein–protein interactions. These PDZ proteins appear to regulate kainate and AMPARs in a distinct manner. Disrupting the interaction with GRIP decreases synaptic responses through KARs while concomitantly increasing the AMPAR-mediated component. Similarly, interfering with the interaction of PICK1 with KARs and AMPARs depresses the KAR-mediated excitatory postsynaptic current (EPSC), leaving the current mediated by AMPARs unaltered. It is likely that the synaptic expression of KARs is tightly regulated through interactions with still unknown scaffolding proteins, as well as through proteins that influence the trafficking and targeting of KARs. For instance, the rules governing the polarized targeting of KARs to

GluR6

The Atomic Structure of Kainate Receptor Binding Domain Has Been Resolved One of the most exciting advances in the structure– function relationships of KARs has been the achievement of high-resolution crystal structures for the ligand-binding core of some subunits. This has provided a detailed understanding of agonist recognition not only for KARs but also for other ligand-gated ion channels. Indeed, the crystal structure of the GluR5

GluR5 B

NH2

axons and dendrites remain unknown. The study of mice that fail to express different subunits has not demonstrated a link between subunit composition and the compartment to which KARs are targeted. Nevertheless, a variety of trafficking determinants in kainate receptors may promote either membrane expression or intracellular sequestration. The identification of the proteins involved (chaperones) should shed light on the mechanisms that regulate KAR signaling and function. Indeed, a recent study has identified the co-atomer protein complex I (COPI) vesicle coat as a critical mechanism for the retention of KA2 subunits in the ER, such that assembly of KA2 subunits with other KAR subunits to form heteromeric receptors reduces their association with COPI.

B

NH2 J

J

A C

D

K

F

C

D

K

F

I

Extracellular

I COOH

COOH

G

G H

a

A

To TM1 TM2

H

Intracellular To TM1 TM2 b

Figure 2 Atomic structure of glutamate-bound GluR6 and GluR5 binding domains: (a) ribbon representation of GluR6 and GluR5 crystal structures at resolution 1.65 and 2.1 A˚, respectively; (b) suggested molecular architecture of glutamate ionotropic receptors. In (a), S1, S2, and loops 1 and 2 are colored blue, gold, and green, respectively. The linker shown in gray replaces the TM segments 1 and 2. The C-terminus of GluR5 appears partially disordered (discontinued line) in the crystal. The bound glutamate molecule is illustrated with ball and stick representations. The three ligand binding pocket side chains shown are conserved in all kainate receptors. Letters A through K name the different a-helical structures loops in the molecule. In (b), the architecture is composed of the crystal structures of the mGluR1 ligand-binding domain dimer (green), the GluR5 ligand-binding core dimer (S1, cyan; S2, orange), and two subunits from the KcsA potassium channel tetramer (gray). An intact glutamate receptor is believed to assemble as a dimer of dimers. GluR, glutamate receptor; mGluR, metabolic glutamate receptor. (a) Adapted from Mayer ML (2005) Crystal structures of the GluR5 and GluR6 ligand binding cores: Molecular mechanisms underlying kainate receptor selectivity. Neuron 45: 539–552. (b) Adapted from Mayer ML (2006) Glutamate receptors at atomic resolution. Nature 440: 456–462.

312 Kainate Receptors: Molecular and Cell Biology

and GluR6 KAR ligand binding cores complexed with glutamate and other agonists, such as kainate and quisqualate, have been shown at a resolution of 1.65 ˚ (GluR5) (Figure 2(a)). As in the (GluR6) and 2.1 A AMPARs and NMDARs, the structure of the KAR binding domain revealed a two-domain closed shell motif linked by b strands. Glutamate binds in a cavity that is isolated from the external solution. For example, agonist binding induces shell closure; this evokes a conformational rearrangement, leading to the opening of the channel. Interestingly, the cavity is smaller in GluR6 than in GluR5 but larger than that of the AMPAR subunit GluR2, due to the presence of unique side chains in KAR subunits. Nevertheless, the mechanism by which glutamate binds to GluR5, GluR6, and GluR2 is nearly identical. Like GluR2, the GluR6 domain associates as a dimer with many conserved interdimer contacts. Thus, an examination of the atomic structure explains why some residues are critical for GluR5-selective ligands (e.g., ATPA). Binding of these agonists to GluR6 is prevented by steric occlusion because the cavity in GluR6 is too small to accommodate the agonists that, like ATPA, bind to GluR5. In Figure 2(b), a recently suggested molecular structure for ionotropic glutamate receptors, including KARs, is presented. It seems that ionotropic GluRs are formed by the fusion of discrete segments, whose ancestors could be found as bacterial proteins. The structure has been built by taking the solved crystal structures of both the N-terminal domain of metabotropic GluRs and the agonist binding domain of KARs. So far, the structure of the GluR ion channel is unknown at the atomic resolution. However, based on the fact that the ion channel of ionotropic GluRs presents remarkable sequence homology with bacterial potassium channels, the crystal structures of a bacterial potassium channel, the KcsA potassium channel, has been incorporated as the channel domain. Although it is likely that a KAR presents a similar arrangement, much work remains to finally solve the structure of the GluR channel. It is expected that this rapidly developing area will soon provide the real structure at the atomic resolution. See also: AMPA Receptor Cell Biology/Trafficking; AMPA Receptors: Molecular Biology and Pharmacology; Kainate Receptor Functions; NMDA Receptors, Cell Biology and Trafficking.

Further Reading Bettler B, Boulter J, Hermans-Borgmeyer I, et al. (1990) Cloning of a novel glutamate receptor subunit, GluR5: Expression in the nervous system during development. Neuron 5: 583–595. Jaskolski F, Coussen F, and Mulle C (2005) Subcellular localization and trafficking of kainate receptors. Trends in Pharmacological Science 26: 20–26. Kohler M, Burnashev N, Sakmann B, and Seeburg PH (1993) Determinants of Ca2þ permeability in both TM1 and TM2 of high affinity kainate receptor channels: Diversity by RNA editing. Neuron 10: 491–500. Lerma J, Paternain AV, Naranjo JR, and Mellstro¨m B (1993) Functional kainate selective glutamate receptors in cultured hippocampal neurons. Proceedings of the National Academy of Sciences of the United States of America 90: 11688–11692. Lerma J, Paternain AV, Rodriguez-Moreno A, and Lopez-Garcia JC (2001) Molecular physiology of kainate receptors. Physiological Reviews 81: 971–998. Madden DR (2002) The structure and function of glutamate receptor ion channels. Nature Reviews Neuroscience 3: 91–101. Mayer ML (2005) Crystal structures of the GluR5 and GluR6 ligand binding cores: Molecular mechanisms underlying kainate receptor selectivity. Neuron 45: 539–552. Mayer ML (2006) Glutamate receptors at atomic resolution. Nature 440: 456–462. Paternain AV, Herrera MT, Nieto MA, and Lerma J (2000) GluR5 and GluR6 kainate receptor subunits coexist in hippocampal neurons and coassemble to form functional receptors. Journal of Neuroscience 20: 196–205. Paternain AV, Morales M, and Lerma J (1995) Selective antagonism of AMPA receptors unmasks kainate receptor-mediated responses in hippocampal neurons. Neuron 14: 185–189. Petralia RS, Wang YX, and Wenthold RJ (1994) Histological and ultrastructural localization of the kainate receptor subunits, KA2 and GluR6/7, in the rat nervous system using selective antipeptide antibodies. Journal of Comparative Neurology 349: 85–110. Schiffer HH, Swanson GT, and Heinemann SF (1997) Rat GluR7 and a carboxy terminal splice variant, GluR7b, are functional kainate receptor subunits with a low sensitivity to glutamate. Neuron 19: 1141–1146. Seeburg PH (1996) The role of RNA editing in controlling glutamate receptor channel properties. Journal of Neurochemistry 66: 1–5. Swanson GT, Feldmeyer D, Kaneda M, and Cull-Candy SG (1996) Effect of RNA editing and subunit co-assembly single-channel properties of recombinant kainate receptors. Journal of Physiology 492: 129–142. Wisden W and Seeburg PH (1993) A complex mosaic of highaffinity kainate receptors in rat brain. Journal of Neuroscience 13: 3582–3598.

Relevant Website http://www.ebi.ac.uk – European Bioinformatics Institute.

Kainate Receptor Functions J Lerma, Instituto de Neurociencias de Alicante Consejo Superior de Investigaciones Cientı´ficas-Universidad Miguel Hernandez, San Juan de Alicante, Spain ã 2009 Elsevier Ltd. All rights reserved.

After finding specific kainate receptors (KARs) in central neurons, physiologists suffered many problems trying to elucidate the role of these receptors in the nervous system, principally due to the lack of pharmacological tools to activate or antagonize KARs. Indeed, many of the pharmacological agonists and antagonists that activate KARs also interact with a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptors (AMPARs). This lack of pharmacological specificity has hindered our understanding of KAR function for several years. Thus, the discovery that 2,3-benzodiazepines, and particularly GYKI 53655 (LY300168 or the active isomer LY303070), antagonize AMPARs but not KARs represented a significant breakthrough in the field. More recently, other compounds that specifically act on KAR subunits have been developed and mice deficient for KAR subunits have been generated, paving the way for the study of the synaptic physiology of KARs. As a result, KARs have been seen to play a role in synaptic transmission, influencing both neuronal excitability and information transfer in the brain. An interesting characteristic of KARs that has emerged is that they use two forms of signaling: a canonical pathway that involves ion flow and another, more unexpected noncanonical signaling pathway that links KAR activation to G-proteins and second-messenger cascades.

Elementary Properties of Kainate Receptor Channels KARs are nonselective cation channels. Therefore, the single-channel conductance of some KAR assemblies has been determined. According to stationary noise analysis, the single-channel conductance of KARs lies in the picosiemen range with values of 2.9 and 5.4 for homomeric glutamate receptor (GluR)5(Q) or GluR6(Q), respectively. A more detailed study of channel openings revealed three subconductance levels that are approximately two or three times the size of the smallest current. Like AMPARs, the replacement of Q by R in the channel domain as the result of RNA editing reduces the single-channel conductance of KARs by more than one order of magnitude. Furthermore, the conductance is also altered

by heteromerization. For instance, the elementary conductance of GluR5(R) channels is less than 200 fS, but it increases to 950 fS on co-assembly with KA2. Similarly, GluR6(R) homomers have a single-channel conductance between 230 and 260 fS, whereas the value for GluR6(R)/KA2 heteromers lies between 570 and 700 fS. The effect of KA2 on the conductance of unedited homomeric receptors is more subtle, with increases to 4.5 and 7.1 pS, for GluR5 and GluR6, respectively. Interestingly, native receptors display similar values, a conductance of 2–4 pS having been measured in dorsal root ganglion (DRG) cells and three subconductance levels, similar to those found for GluR5(Q) or GluR5(Q)/KA2, have been identified. These findings are in accordance with the abundant expression of GluR5 in DRG neurons. In addition, the conductance of KARs expressed in immature proliferating cerebellar granule neurons has been determined to be 1 and 4 pS, the channels spending most of their open time in the 4 pS state.

Desensitization of Kainate Receptor Currents Rapid desensitization is one of the major characteristic features of KARs. The time course of current decay in the continued presence of an agonist follows a single or double exponential decay. The speed of desensitization is dependent on the receptor subtype and on the cell type being analyzed. GluR6 homomers and native hippocampal KARs have desensitization time constants of 11–13 ms, a value similar to that found for recombinant channels in excised patches or lifted cells. However, KARs recover slowly from the desensitized state, and this recovery is dependent on the nature of the agonist. Indeed, whereas the recovery from desensitization by a pulse of kainate takes over 1 min, the receptor requires only 15 s when glutamate is the agonist. The subunit composition also affects recovery from desensitization; thus, whereas GluR5 homomeric receptors recover in about 1 min, the time constant for the recovery of GluR5/KA2 heteromers is just 12 s. These dramatic differences in the time scale of desensitization and recovery imply that the equilibrium between the two states is strongly displaced toward the inactive state. In other words, the receptors spend most of their time desensitized. This contradicts the fact that low concentrations of kainate and glutamate have a striking effect on native KARs, as seen in neurons in brain slices. There is still no clear explanation for this behavior, which is not predicted from rapid

313

314 Kainate Receptor Functions

activation and inactivation kinetics. However, it is possible that unknown proteins interact with native KARs, altering their gating properties, as seen for other receptors and channels.

Pharmacological Aspects As already mentioned, a major difficulty in understanding KAR function has been the lack of a specific pharmacology for these receptors. For instance, kainate, albeit showing a clear preference for KARs, has a significant action on AMPARs at relatively low doses. Indeed, there is only a 5–30 times difference between the apparent affinity of these two receptors for kainate. There has also been some progress concerning selective KARs agonists. Some derivatives of willardiine have been developed as selective KAR agonists (Table 1). Among them, (S)-5-iodowillardiine shows a 130-fold selectivity for KARs (median effective concentration (EC50) ¼ 140 nM). SYM 2081, the (2S-4R) diastereomer of 4-methylglutamate, displays a potency three orders of magnitude higher for KARs than for AMPARs, both in binding and in functional assays, but the selectivity of this molecule for kainate over N-methyl-D-aspartate receptors (NMDARs) is significantly lower. Although its pharmacological profile is incomplete, SYM 2081 does not seem to show the subunit specificity observed for ATPA and for (S)-5-iodowillardiine; it elicits rapidly desensitizing currents on GluR5 and GluR6 homomeric channels. Due to this property, this compound has also been used as a functional antagonist of the KARs because it efficiently desensitizes KARs at low concentrations (Table 2). Similarly, the prototypic non-NMDAR antagonists, 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), 6,7dinitroquinoxaline-2,3-dione (DNQX) and 6-nitro-7sulphamoylbenzo[f]quinoxaline-2,3-dione (NBQX), just show poor selectivity between AMPARs and KARs, although it has been shown that low concentrations of NBQX do not have any action on KARs. However, the separation of KAR-mediated responses from the AMPAR-mediated currents is possible by using 2,3-benzodiazepine GYKI 53655 (Lilly’s code LY300168), which is selective for antagonizing AMPARs. Therefore, in the presence of GYKI53655, or its active isomer LY303070, CNQX could be used as a KAR antagonist (Table 2). Several other very useful compounds have been developed lately. In particular, LY382884 ((3S, 4aR, 6S, 8aR)-6-((4-carboxyphenyl) methyl-1,2,3,4,4a,5,6,7,8,8a-decahydro isoquinoline3-carboxylic acid), seems to antagonize KARs at concentrations that do not affect AMPARs or NMDARs. Actually, LY382884 is a selective antagonist at neuronal KARs containing the GluR5 subunit (Table 2).

Table 1 Kainate receptor agonists Compound

Selectivity

ATPA S-5-iodowillardine Kainate Domoate AMPA Me-Glutamate (SYM 2081)

GluR5-contaning; GluR6/KA2 GluR5-containing; GluR6/KA2; GluR7/KA2 All assemblies GluR5- and GluR6-containing. GluR5-containing; GluR6/KA2; GluR7/KA1 GluR5- and GluR6-containing

AMPA, a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid; GluR, glutamate receptor.

Table 2 Kainate receptor antagonists Compound

Selectivity

Mechanism

CNQX

All assemblies; more potent at AMPAR GluR5-containing GluR5-containing Homomeric GluR5 GluR5-containing GluR5-containing and homomeric GluR6 GluR5- and GluR6-containing

Competitive

LY382884 UBP296 NS3763 LY377770 Kynurenate SYM2081

Competitive Competitive Noncompetitive Competitive Competitive Desensitization

AMPAR, a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor; CNQX, 6-cyano-7-nitroquinoxaline-2,3-dione; GluR, glutamate receptor.

Kainate Receptors Have a Fundamental Role in Controlling Synaptic Neurotransmitter Release The discovery that GYKI 53655 (or LY303070, the active isomer) antagonizes AMPARs but not KARs paved the way for the study of the participation of KARs in synaptic responses. Nowadays, there is considerable evidence suggesting that KARs may be situated on both sides of the synapse. A considerable amount of effort has been devoted to determining the role of presynaptic KARs in controlling transmitter release, particularly in the hippocampus. In this structure, KARs are widely distributed in presynaptic buttons, and they can bidirectionally regulate the release of glutamate at the mossy fibers to CA3 synapses. The mild activation of KARs enhances the release of glutamate, whereas when they are more intensely activated glutamate release is inhibited. As such, a proportion of the high-frequency-dependent facilitation, a hallmark of the mossy fiber synapses, can be attributed to the activation of presynaptic KARs by synaptically released glutamate, as if low concentrations of exogenous kainate had been applied (Figure 1(b)). Therefore, the characteristic short-term plasticity of mossy

Kainate Receptor Functions 315

Dentate gyrus granule cells Mossy fiber

Mossy fibers

b

CA3 neuron

CA3 pyramidal neurons

–GYKI = EPSCAMPA EPSCAMPA +GYKI = EPSCKAR

+KAR antagonist

40 ms

a Without KAR antagonist c

d

40 ms EPSCAMPA

5 pA EPSCKAR

25 ms

e Figure 1 Proposed roles of kainate receptors (KARs) in synaptic transmission: (a) mossy fiber-CA3 synapses; (b) synaptic response of the mossy fibers to CA3 pyramidal cells; (c) presynaptic role; (d) postsynaptic current; (e) postsynaptic responses at the thalamo-cortical synapse. These can be isolated in the presence of APV (to block NMDA receptors) and GYKI53655 (to block AMPA receptors). In all cases, the KAR-mediated synaptic responses are much slower than those observed for AMPA receptors and usually present one-tenth their amplitude (in (d), both responses have been normalized). In (c), presynaptic KARs can enhance or inhibit glutamate release, depending of the degree of activation. The best-established example is the synapse made by mossy fibers on to CA3 pyramidal neurons, where KARs seem to be responsible of a part of the high-frequency facilitation, a characteristic of these synapses. KARs also mediate synaptic responses of Schaffer collateral to interneurons (not shown). In (e), EPSCs mediated exclusively by AMPA or KARs have been isolated by minimal stimulation. AMPA, a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid; EPSC, excitatory postsynaptic current; Glu, glutamate; NMDA, N-methyl-D-aspartate. (d) Adapted from Lerma J (2003) Roles and rules of kainate receptors in synaptic transmission. Nature Reviews Neuroscience 4: 481–495. (e) From Kidd FL and Isaac JT (1999) Developmental and activity-dependent regulation of kainate receptors at thalamocortical synapses. Nature 400: 569–573.

fiber–CA3 synapses is partly mediated by the longlasting activation of a kainate autoreceptor. This mechanism has been shown to impose associative properties on mossy fiber long-term potentiation (LTP) because the activity in neighboring mossy fiber synapses, or even associational/commissural synapses, influences the threshold for inducing LTP. However, higher concentrations of kainate, instead of facilitating mossy fiber synaptic transmission, depress it. Interestingly, this phenomenon is reproduced by synaptic activity, such that a brief conditioning of tetanus on associational-commissural fibers enhances the mossy fiber responses, whereas

prolonged tetanus depresses it. The exact mechanism involved in the presynaptic facilitation of glutamate release is unclear. One explanation for the facilitation of glutamate release is that presynaptic KARs depolarize synaptic terminals to a level at which further release is enhanced. Also, it has been claimed that a Ca2þ-induced Ca2þ release from the intracellular stores is required for this phenomenon to occur. Therefore, such a mechanism remains unclear. How the larger activation of KARs might lead to the inhibition of glutamate release in mossy fiber buttons also remains a matter of debate. It could be the case that the large depolarization of synaptic terminals inactivates

316 Kainate Receptor Functions

the generation of action potentials, leading to a failure in release. However, recent data show that inhibition (but not facilitation) of release by presynaptic KARs depends on a G-protein-dependent mechanism. This indicates that KARs act independent of their ion channel activity. Therefore, it is possible that the threshold for activating one or other KAR signaling pathway could determine the physiological response. It is now widely accepted that presynaptic KARs also control the release of the inhibitory neurotransmitter g-aminobutyric acid (GABA). Although KAR stimulation may enhance the release of GABA at

connections between the hippocampal interneurons, stimulation of these receptors inhibits GABA release at interneuron–pyramidal cell synapses. Thus, data indicate that the spillover of glutamate from adjacent terminals could activate KARs at presynaptic GABA terminals (Figure 2(b)). The specific role that the receptors controlling GABA release play in hippocampal synaptic networks remains unclear. However, the infusion of kainate in vivo reduces the efficiency of recurrent inhibition, producing a state of hyperexcitability that leads to the generation of recurrent epileptic spikes. Accordingly, KAR antagonists should be able to prevent and/or abolish epileptic

IPSCs Control CA1 pyramidal neuron Kainate 0.3 µM

GABAergic terminal

CA1 interneuron

c Stratum oriens single stimulus

Schaffer collaterals

IPSCs

a Glu

Schaffer collateral stimuli d EPSCs +GYKI

Schaffer collateral

IAHP

−GYKI

Control Kainate

b

e

f

40 mV

1s

Figure 2 Role of kainate receptors (KARs) in neurotransmitter synaptic release: (a) CA1 pyramidal neurons receive excitatory inputs from the Schaffer collateral and inhibitory actions from interneurons; (b) illustration of an excitatory and an inhibitory synapse on a CA1 pyramidal cell dendrite; (c) a low concentration of kainate selectively reduces the GABA release, as evidenced by a reduction in the evoked IPSC, under pharmacological isolation of GABAA-mediated responses; (d) glutamate could spill over to activate presynaptic KARs present in GABAergic terminals such that the IPSCs (induced by a single shocks to the stratum oriens) are reduced after a train of stimuli is applied on the Shaffer collateral pathway; (e–f) synaptically released glutamate could activate postsynaptic KARs. Synaptically released glutamate could also activate presynaptic (not shown) KARs. Although the activation of postsynaptic KARs does not induce any synaptic current ((e), þGYKI), the afterhyperpolarization current ((f), IAHP) is depressed. KARs have been shown to exert all these actions by activating G-proteins. EPSC, excitatory postsynaptic current; GABA, g-aminobutyric acid; Glu, glutamate; IPSC, inhibitory postsynaptic current. (d) Adapted from Min MY, Melyan Z, and Kullmann DM (1999) Synaptically released glutamate reduces gamma-aminobutyric acid (GABA)ergic inhibition in the hippocampus via kainate receptors, Proceedings of the National Academy of Sciences of the United States of America 96: 9932–9937. (f) From Lerma J (2006) Kainate receptor physiology. Current Opinion in Pharmacology 6: 89–97.

Kainate Receptor Functions 317

activity. This has indeed been demonstrated when epilepsy is induced in rats by a pilocarpine injection. In this model, KAR antagonists (e.g., LY377770, Table 2) not only abolish the established seizure activity but also prevent the development of seizures when they are applied beforehand. Therefore, these and other data indicate that the excitability of the network may be tightly controlled by the extracellular concentrations of glutamate through the transient and/or tonic activation of KARs. One important issue that has remained a matter of intense debate over the last few years relates to the mechanism by which KARs affect release at these inhibitory synapses. There is evidence indicating that GABA release is inhibited by a G-protein- and protein kinase C (PKC)-dependent mechanism (Figure 2(b)). It is possible that the large barrage of inhibitory postsynaptic current (IPSC) received by pyramidal cells on kainate perfusion may cause some short circuiting, but it is now generally accepted that this is not sufficient to account for such inhibition. Indeed, it should be taken into account that KARs may regulate CA1 GABAergic circuits through two distinct and opposing mechanisms. Whereas the activation of somatodendritic KARs may increase the activity of GABA interneurons, presynaptic KARs may diminish inhibition. These two effects are mediated by receptor populations that are functionally, physically, and pharmacologically distinct, one of which is located in the somatodenritic and/or axonal compartments and the other at the presynaptic terminals directly inhibiting GABA release. Likewise, there is considerable data that point to a prominent role of KARs in the regulation of neuronal excitability through the modulation of the GABAergic system in a variety of brain nuclei, such as the amygdala, the neocortex, the striatum, and the cerebellum.

Kainate Receptors Mediate Part of the Postsynaptic Response at Certain Synapses Synaptic release of glutamate can also activate postsynaptic KARs, as seen primarily in synapses established by mossy fibers on CA3 neurons. Subsequently, KARmediated synaptic responses were demonstrated in Schaffer collaterals–hippocampal CA1 interneurons, at the parallel fiber–Golgi cell synapses, in the basolateral amygdala, in Purkinje cells of the cerebellum, at thalamo-cortical connections on cortical interneurons (Figure 1(e)), in neocortical pyramidal neurons, and in the synapse established by cones on bipolar cells in the retina. Similarly, a synaptic component mediated by KARs has been found in the synapse between afferent sensory fibers and dorsal horn neurons in the spinal

cord. With the exception of excitatory postsynaptic currents (EPSCs) in the retina, the KAR-mediated EPSCs all displayed common characteristics such as a low amplitude (approximately an order of magnitude lower than the component mediated by AMPA receptors) and a longer-lasting time course than the EPSC mediated by the AMPA receptor (Figure 1(d)). What could be the functional significance of a small and long-lasting EPSC for neurons? Although it is not definitively determined, a slow small EPSC, such as that mediated by KARs, offers the possibility of integrating excitatory inputs over a wide time window. Thus, the relative contribution of KARs to synaptic drive could be significant. It has been calculated that KARs elicit sufficient charge transfer so as to exert a substantial impact on the activity of interneurons. Indeed, the kinetics of KAR-mediated EPSPs are sufficiently slow to allow substantial tonic depolarization even during modest presynaptic activity. Such tonic depolarization could account for the observation that single afferent inputs are very effective in driving interneuron spiking. Accordingly, KARs contribute to the ongoing ionotropic glutamatergic transmission in the hippocampus. Therefore, neurons may display unique integration properties in areas where KARs are expressed. Why native and recombinant KARs present different kinetic properties has not been convincingly explained. However, KARs take direct part of the glutamatergic synapse because these receptors are activated by quantal release of glutamate at both mossy fibers–CA3 synapses and excitatory synapses on CA1 interneurons. In addition, there is evidence that AMPARs and KARs can be segregated at different synapses. This seems to be true at least in CA1 interneurons and at thalamo-cortical synapses, although not in individual mossy fiber synapses. These studies allowed the kinetics of elementary KAR-mediated EPSC in CA1 interneurons to be determined, revealing that they are still slower than the pure AMPARmediated EPSC. Therefore, there should be factors such as the dissimilar composition of the synaptic receptors or accessory proteins that may explain the difference between the response kinetics of the native and recombinant receptors studied so far. The retina constitutes one of the systems in which KARs have been shown to display similar properties to those found in recombinant receptors, specifically at the synapse from cones to the off bipolar cells. In this system, the kinetics of the synaptic responses fits well with the activation and desensitization properties of KARs determined in recombinant systems. In these synapses, the slow kinetics of recovery from desensitization of KARs is exploited in processing the afferent signals rather than suffering the constrictions

318 Kainate Receptor Functions

of the faster time course of AMPARs. In the dark, a cone releases sufficient neurotransmitter to desensitize most postsynaptic KARs, leaving just a small persistent postsynaptic current. Because the postsynaptic KARs desensitize rapidly and recover from desensitization slowly, little recovery can occur during the brief hyperpolarizations that are produced in the cones by repetitive light stimuli. Therefore, the large increase in conductance that would saturate the membrane voltage if the rate of transmitter release by the cone was high is prevented. Actually, signaling between the cones and the three morphological types of off bipolar cells occurs through distinct postsynaptic receptors. Whereas two cell types use KARs with distinct properties, the other type uses AMPARs. Each receptor recovers from desensitization with a particular time course, and this property contributes to determining the temporal characteristics of synaptic transmission in the retina (Figure 3). Recent studies have demonstrated that KARs are also involved in long-term synaptic plasticity (LTP). Indeed, mossy fiber LTP may be prevented by the GluR5 antagonist LY382884. This antagonist,

K+ channels

however, is ineffective for preventing LTP induced by tetanic stimulation of associational/commissural fibers in the CA3 region or at CA1 synapses. Therefore, a role for KARs in the induction of LTP seems to be specific for the mossy fiber pathway, although pharmacological data are at odds with results obtained from knockout mice. The use of KAR subunit-deficient mice confirmed that KARs play a critical role in mossy fiber LTP. However, mossy fiber LTP was reduced, but not abolished, in GluR6knockout mice, whereas GluR5-deficient mice exhibited normal LTP at these synapses. Clearly, more investigation is needed to understand the molecular mechanisms by which KARs may play a permissive role at mossy fibers to induce a permanent change in release properties. Also interesting is the fact that at thalamo-cortical synapses in the neonatal barrel cortex, LTP expression is associated with a reduction in the KAR-mediated component of the synaptic response. Thus, as recently confirmed, KAR-mediated synaptic transmission can undergo a form of long-term depression (LTD) that is mechanistically distinct from the LTD seen for AMPARmediated transmission. Kainate receptor

Ca2+ channels

1 Kainate

Na+

N-type Ca2+ current DG Kainate Recovery Control 3

Ca2+

PKC IP3 2

200 pA 20 ms

Intracellular stores

Figure 3 Canonical and noncanonical signaling of kainate receptors (KARs). As ligand-gated ion channels, the activation of KARs evokes an inward current, through which they depolarize the membrane (1). However, they can also set in motion a noncanonical signaling pathway triggered by the activation of a pertussis toxin-sensitive G-protein (G). The signaling pathway involves phospholipase C (PLC), leading to the synthesis of diacylglycerol (DG) and inositol 1,4,5, triphosphate (IP3). Although IP3 could release Ca2þ from intracellular stores, as revealed by confocal Ca2þ imaging (2), DG activates protein kinase C (PKC), which in turn inhibits Ca2þ channels, as illustrated by the direct recording of Ca2þ currents (3). This pathway is also responsible for the inhibition of the Ca2þ-dependent Kþ current, accounting for the hyperpolarization of the membrane after repetitive firing (afterhyperpolarization). These effects of kainate do not seem to require ion channel receptor activity, providing evidence for the existence of a metabotropic KAR that inhibits Ca2þ and Kþ channels. Adapted from Rozas JL, Paternain AV, and Lerma J (2003) Noncanonical signaling by ionotropic kainate receptors. Neuron 39: 543–553.

Kainate Receptor Functions 319

Dual Signaling of Kainate Receptors It was recently demonstrated that KARs and other ionotropic glutamate receptors may transmit signals both through ion flux and by receptor coupling to G-proteins, which in turn may regulate voltagedependent Ca2þ channels (Figure 3). This was initially described in the hippocampus, and it was later confirmed in other tissues. Indeed, activation of KARs in cultured DRG cells, which almost exclusively express GluR5 and KA2 KAR subunits, induces the release of Ca2þ from intracellular stores in a G-protein-dependent manner (Figure 3). Interestingly, the concomitant activation of PKC inhibited voltagedependent N-type calcium channels. This noncanonical signaling is independent of ion channel activity and provides insights into the dual signaling of KARs because both pathways depend on a common ionotropic subunit, GluR5. But the way in which an ion channel becomes coupled to a G-protein is still unclear, although it seems likely that intermediate or linker proteins may assist in the process. Unfortunately, none of the proteins that interacts with KARs identified to date seem to carry out this role, and thus, these intermediate elements remain to be identified. In addition, this property does not seem to be restricted to a given KAR subunit. Postsynaptic KARs can also inhibit hyperpolarizing currents through similar signaling mechanisms (Figure 2(f)), and this seems to involve KA2 subunits rather than GluR6 or GluR5 subunits. There is compelling evidence indicating that synaptically released glutamate could activate postosynaptic KARs in CA1 and CA3 pyramidal neurons that do not involve the generation of inward currents. Rather, these receptors reduce the typical afterhyperpolarization current seen after cell firing (IAHP). Inhibition of after hyperpolarization has a great impact on the neuron excitability, such that the response to a given input is drastically altered. Afterhyperpolarization is induced by the activation of Ca2þ-dependent Kþ channels, curtailing repetitive firing. The reduction of such a current causes the neuron to generate large burst of spikes, largely increasing the neuronal output. Together with the inhibition of GABA release, this represents an example in which KARs may exert a striking effect on neuronal excitability by a noncanonical signaling. Therefore, temporal integration of neuronal inputs may be modulated by KARs not acting as ion channels, just by attenuating hyperpolarizing conductances such as IAHP and GABA receptor-mediated current. Although exogenous kainate could facilitate or inhibit transmitter release, depending on the concentration and type of synapse, cumulative data indicate that the rule is that the inhibitory activity of KARs is linked to this noncanonical signaling. However, the variety

of signals activated by these metabotropic KARs indicates a certain lack of predictability in the coupling of KARs to G-proteins. Hence, much work remains to be done to clarify when KARs may work through classical or noncanonical signaling pathways. Although not without their problems, the availability of specific drugs as well as of knockout mice for each of the KAR subunits allows us to clarify which subunit participates in specific KAR-mediated responses.

Involvement of Kainate Receptors in Neuropathology The implications of KARs for nervous system pathology is being underscored but not without uncertainties. For instance, a contribution of KARs to pain is supported by pharmacological data because the GluR5-selective antagonist LY293558 has demonstrated preclinical and clinical efficacy in models of pain. Indeed, GluR5 subunits are essential for KARmediated responses in DRG neurons and for presynaptic regulation of transmitter release in the spinal dorsal horn. However, nociceptive thresholds are largely unaffected in GluR5- and GluR6-deficient mice. Only one report found decreased response to capsaicin and formalin in GluR5-deficient mice. Such a profile raises the possibility that KAR antagonists could be effective for the treatment of certain forms of pain and acute migraine. Similarly, excitotoxicity is triggered by the enhanced activation of GluRs. Although kainate was demonstrated to be a neurotoxin long ago, there is no compelling evidence for a direct role of KARs in this process. The most sensitive area in the brain to kainate-induced excitotoxicity is the hippocampal CA3, a region rich in high-affinity KARs. However, in GluR6-knockout mice, kainateinduced CA3 toxicity is attenuated but not eliminated. Recently, KARs activation has been linked to the activation of c-jun N-terminal kinase 3 (JNK) and mixed lineage kinase 3 (MLK3) that seem to link GluR6-containing KARs to the cell death pathway activated after ischemic insults. The best-established brain disorder in which KARs are implicated is the excitatory imbalance linked to epilepsy. Certainly, kainate injections have provided an animal model for studying human temporal lobe epilepsy. The inhibition of GABA release leading to recurrent epileptic activity may account for the acute epileptogenic effect of kainate. Recently, it has been observed that the synaptic response mediated by KARs provides a substantial component of the excitatory transmission at functional aberrant synapses formed by sprouting of glutamatergic fibers, a hallmark of human temporal lobe epilepsy. Thus, these

320 Kainate Receptor Functions

results raise the possibility of designing antiepileptic therapies based on KAR antagonists. A number of studies have examined the link between the expression of KARs genes and/or singlenucleotide polymorphisms (SNPs) to disorders with a genetic background with disparate results. No strong associations have been found between the expression of genes coding for KAR subunits (GluR6 and GluR7 (GRIK2 and GRIK3 in human nomenclature)) and obsessive–compulsive disorder patients. Only GRIK2 SNP I867 was transmitted less readily than expected. Interestingly, this variant has also been associated with autism. An allele of the GluR6 gene is associated with the early age of onset of Huntington’s disease; GluR6-mediated excitotoxicity has been implicated in the pathogenesis of Huntington’s disease. Allelic variants of GRIK1 (GluR5), but not GRIK2 (GluR6), appear to contribute a major genetic determinant to the pathogenesis of juvenile absence epilepsy and related phenotypes. Interestingly, the GluR5 gene is located at chromosame 21q22.1, and physical mapping situates it near the familial amyotrophic lateral sclerosis APP and SOD1 regions, making it a possible candidate for familial amyotrophic lateral sclerosis and other diseases. Whether GluR5 influences the phenotypes associated with partial trisomy or monosomy of chromosome 21 remains to be determined. Other recent results have indicated that there is a disequilibrium of GRIK2 (GluR6) transmission in autism and schizophrenia. However, no linkage has been established with GluR5 (GRIK1) SNPs and their haplotypes in schizophrenia. Whether the described SNPs may impact the function and/or trafficking of KARs is yet to be determined. Therefore, a definitive role of KARs in these disorders is possible but not demonstrated. See also: AMPA Receptor Cell Biology/Trafficking; AMPA Receptors: Molecular Biology and Pharmacology; Kainate Receptors: Molecular and Cell Biology; NMDA Receptors, Cell Biology and Trafficking.

Further Reading Castillo PE, Malenka RC, and Nicoll RA (1997) Kainate receptors mediate a slow postsynaptic current in hippocampal CA3 neurons. Nature 388: 182–186. Christensen JK, Paternain AV, Selak S, Ahring PK, and Lerma J (2004) A mosaic of functional kainate receptors in hippocampal interneurons. Journal of Neuroscience 24: 8986–8993. Contractor A, Swanson G, and Heinemann SF (2001) Kainate receptors are involved in short- and long-term plasticity at mossy fiber synapses in the hippocampus. Neuron 29: 209–216. Cossart R, Epsztein J, Tyzio R, et al. (2002) Quantal release of glutamate generates pure kainate and mixed AMPA/kainate EPSCs in hippocampal neurons. Neuron 35: 147–159.

Coussen F, Perrais D, Jaskolski F, et al. (2005) Coassembly of two GluR6 kainate receptor splice variants within a functional protein complex. Neuron 47: 555–566. Epztein J, Represa A, Jorquera I, Ben-Ari Y, and Crepel V (2005) Recurrent mossy fibers establish aberrant kainate receptoroperated synapses on granule cells from epileptic rats. Journal of Neuroscience 25: 8229–8239. Frerking M, Malenka RC, and Nicoll RA (1998) Synaptic activation of kainate receptors on hippocampal interneurons. Nature Neuroscience 1: 479–486. Frerking M and Ohliger-Frerking P (2002) AMPA receptors and kainate receptors encode different features of afferent activity. Journal of Neuroscience 22: 7434–7443. Kidd FL and Isaac JT (1999) Developmental and activitydependent regulation of kainate receptors at thalamocortical synapses. Nature 400: 569–573. Lerma J (2003) Roles and rules of kainate receptors in synaptic transmission. Nature Reviews Neuroscience 4: 481–495. Lerma J (2006) Kainate receptor physiology. Current Opinion in Pharmacology 6: 89–97. Lerma J, Paternain AV, Rodriguez-Moreno A, and Lopez-Garcia JC (2001) Molecular physiology of kainate receptors. Physiological Reviews 81: 971–998. Melyan Z, Lancaster B, and Wheal HV (2004) Metabotropic regulation of intrinsic excitability by synaptic activation of kainate receptors. Journal of Neuroscience 24: 4530–4534. Min MY, Melyan Z, and Kullmann DM (1999) Synaptically released glutamate reduces gamma-aminobutyric acid (GABA) ergic inhibition in the hippocampus via kainate receptors. Proceedings of the National Academy of Sciences of the United States of America 96: 9932–9937. Mulle C, Sailer A, Perez-Otano I, et al. (1998) Altered synaptic physiology and reduced susceptibility to kainate-induced seizures in GluR6-deficient mice. Nature 392: 601–605. Paternain AV, Morales M, and Lerma J (1995) Selective antagonism of AMPA receptors unmasks kainate receptor-mediated responses in hippocampal neurons. Neuron 14: 185–189. Rodriguez-Moreno A, Herreras O, and Lerma J (1997) Kainate receptors presynaptically downregulate GABAergic inhibition in the rat hippocampus. Neuron 19: 893–901. Rodriguez-Moreno A and Lerma J (1998) Kainate receptor modulation of GABA release involves a metabotropic function. Neuron 20: 1211–1218. Rozas JL, Paternain AV, and Lerma J (2003) Noncanonical signaling by ionotropic kainate receptors. Neuron 39: 543–553. Rubinsztein DC, Leggo J, Chiano M, et al. (1997) Genotypes at the GluR6 kainate receptor locus are associated with variation in the age of onset of Huntington disease. Proceedings of the National Academy of Sciences of the United States of America 94: 3872–3876. Schmitz D, Mellor J, Breustedt J, and Nicoll RA (2003) Presynaptic kainate receptors impart an associative property to hippocampal mossy fiber long-term potentiation. Nature Neuroscience 6: 1058–1063. Smolders I, Bortolotto ZA, Clarke VR, et al. (2002) Antagonists of GLU(K5)-containing kainate receptors prevent pilocarpineinduced limbic seizures. Nature Neuroscience 5: 796–804.

Relevant Website http://www.ebi.ac.uk – European Bioinformatics Institute.

Long-Term Potentiation (LTP): NMDA Receptor Role A J Doherty, S M Fitzjohn, and G L Collingridge, University of Bristol, Bristol, UK ã 2009 Elsevier Ltd. All rights reserved.

Introduction The subtype of glutamate receptor known as the N-methyl-D-aspartate (NMDA) receptor is the trigger for both long-term potentiation (LTP) and long-term depression (LTD) at the majority of synapses in the central nervous system (CNS), where long-lasting synaptic plasticity has been observed. The role of NMDA receptors (NMDARs) in synaptic plasticity was first identified, and has been most extensively studied at, the synapses made between CA3 and CA1 pyramidal neurons in the hippocampus, the Schaffer collateral–commissural pathway. While most is known about LTP at these CA1 synapses, it seems that the principles discovered here apply elsewhere in the brain. The following discussions pertain to LTP at CA1 synapses.

Glutamate Receptors and Synaptic Transmission It is commonly considered that a-amino-3-hydroxy-5methyl-4-isoxazole propionic acid receptors (AMPARs) mediate synaptic transmission and NMDARs trigger alterations in synaptic efficiency. However, the reality is more complex. AMPARs do indeed mediate the majority of the response to low-frequency stimulation. However, NMDARs contribute significantly to high-frequency synaptic transmission. Given that in the conscious animal CA3 neurons often discharge in high-frequency bursts, then NMDARs are likely to play a major role in the transmission of synaptic information within the CA1 area. It is also sometimes mistakenly considered that only AMPARs mediate the response that is modified during LTP. While it is true that it is often the AMPAR-mediated response that is studied, it is equally true that the NMDARmediated component of synaptic transmission has the capacity to undergo LTP.

Second, this block is highly voltage dependent. Third, the NMDAR-mediated conductance has slow kinetics, compared with the AMPAR-mediated response. Fourth, the NMDAR not only passes monovalent cations but also has a significant permeability to Ca2þ. During low-frequency transmission there is rapid activation of AMPARs. However, while L-glutamate binds to NMDARs, they are largely blocked by Mg2þ, which is present in the synaptic cleft at around 1 mM (Figure 1). This prevents NMDARs from contributing appreciably to the synaptic response. During high-frequency transmission there is a sustained depolarization, due in part to the temporal summation of AMPARmediated excitatory postsynaptic potentials (EPSPs), and this alleviates the voltage-dependent Mg2þ block to enable NMDARs to contribute to the synaptic response. Since they have a slow decay phase, the NMDAR-mediated EPSPs summate very effectively.

Induction of LTP Induction by High-Frequency Stimulation

Conditions that enable the activation of NMDARs can result in the induction of LTP. The most commonly used induction stimulus is a continuous high-frequency train (tetanus), sometimes delivered more than once. This provides the depolarization to transiently relieve the Mg2þ block of NMDARs. The Ca2þ that then enters the neuron through the NMDARs provides the initial trigger for the plastic change. Although a tetanus typically comprises 100 stimuli or more (e.g., 100 Hz, 1 s), far fewer stimuli can induce LTP, provided that they are appropriately timed. One such protocol is the delivery of brief high-frequency bursts (e.g., four shocks at 100 Hz) with an interburst frequency of around 5 Hz (i.e., an interburst interval of 200 ms). This pattern of activity mimics the physiological activity that occurs during the theta rhythm, a frequency of activation observed in the electroencephalogram (EEG) when animals are exploring their environment. The minimal induction protocol is the delivery of a single priming stimulus followed approximately 200 ms later by a single burst (which could comprise as few as two stimuli, though typically four are used). Thus, under optimal conditions LTP can be induced by three to five stimuli.

NMDAR Properties Govern Their Role in Synaptic Transmission

Induction by Pairing

The NMDAR has unique properties that govern its role in synaptic processing. First, it is strongly antagonized by micromolar concentrations of Mg2þ.

LTP can also be induced by pairing. This is achieved by maintaining low-frequency stimulation but depolarizing the neuron artificially by injecting current through

321

322 Long-Term Potentiation (LTP): NMDA Receptor Role

No stimulation

Basal stimulation

High-frequency stimulation

AMPAR NMDAR Mg2+

Figure 1 N-Methyl-D-aspartate receptors (NMDARs) are activated by high-frequency stimulation. Under basal conditions, NMDAR activity is blocked due to voltage-dependent binding of Mg2þ ions. A single stimulus results in the release of L-glutamate and activation of a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptors (AMPARs) but does not provide sufficient depolarization to repel the Mg2þ ions. High-frequency stimulation results in a more sustained depolarization of the postsynaptic terminal, leading to the release of the Mg2þ block and an increase in cytosolic Ca2þ ions. The rise in Ca2þ concentration may be localized to the activated synapse, ensuring that only the activated synapse undergoes long-term potentiation.

the recording electrode. Typically, approximately 50 pairings of low-frequency stimulation and depolarization are employed. LTP can also be induced by the pairing of appropriately timed presynaptic and postsynaptic action potentials. The common feature of all of these induction protocols is that they enable sufficient depolarization to transiently relieve the Mg2þ block of NMDARs. Once this occurs, it enables Ca2þ to enter the synapse through the NMDARs in sufficient amounts to trigger the biochemical cascades that result in enhanced synaptic transmission. NMDAR Activation without LTP

However, activation of NMDARs does not necessarily induce LTP. Under conditions whereby the LTP process has been saturated (by prior periods of LTP induction), protocols that normally induce LTP result in no change in synaptic strength. In addition, certain patterns of activation can activate NMDARs but induce longterm depression rather than LTP. If a pathway has already undergone LTP, then these same patterns of activation may induce depotentiation (DP) of the potentiated response. Typically, NMDAR-dependent LTD and DP are induced by prolonged periods of low-frequency stimulation (1–3 Hz). Finally, an intermediate frequency (e.g., 5 Hz) will neither induce LTP nor LTD, while still activating NMDARs.

Features of NMDAR-Dependent LTP Input Specificity

NMDAR-dependent LTP has several hallmark features. Input specificity refers to the property that if two inputs are activated and converge on a population of neurons, or indeed a single neuron, but an induction protocol is delivered to just one of the inputs, then only that input will undergo LTP. The other input typically will not change its efficiency of transmission, or under certain circumstances may be depressed via a process known as heterosynaptic depression. Input specificity is an important property. It means that the unit of modification is the activated input, not the entire neuron. This implies that the unit of information storage can be as small as a single synapse. Input specificity is due to the requirement for L-glutamate to activate NMDARs and the fact that the Ca2þ that permeates NMDARs does not spread very far. Therefore, when a neuron is sufficiently depolarized, those synapses that are activated by synaptically released L-glutamate can undergo LTP. Nearby neighbors may also undergo LTP if either the synaptically released L-glutamate ‘spills over’ in sufficient amounts to activate enough NMDARs or if the Ca2þ spreads beyond the activated spine. In its purest form, LTP is restricted to the activated inputs.

Long-Term Potentiation (LTP): NMDA Receptor Role Cooperativity

Cooperativity refers to the need to stimulate multiple afferent fibers to induce LTP. Experimentally this is observed as the failure of a ‘weak tetanus’ to induce LTP, while a ‘strong tetanus’ that activates more fibers is able to induce LTP. Cooperativity can be explained by the need for multiple inputs to be activated around the same time to provide sufficient depolarization to remove enough of the Mg2þ block to enable the induction of LTP. This property provides a threshold of activity that has to be exceeded for synaptic plasticity to occur. CA3 neurons typically fire in brief, synchronous bursts and this is presumably to exceed the cooperativity threshold to enable LTP to occur at CA1 synapses. Associativity

Associativity is an extension of cooperativity, but whereby the ‘strong tetanus’ is delivered to an independent input. There is a critical timing window in which the strong input can be delivered after the weak input and still be effective. Associativity is explained by the weak input providing the necessary L-glutamate but insufficient depolarization, whereas the ‘strong input’ provides the (additional) depolarization to sufficiently alleviate the Mg2þ block. The associative depolarization could be provided by a second glutamatergic input or by a different neurotransmitter system that provides or enables depolarization at the critical time. Experimentally, the depolarization can be provided by the recording electrode (in a pairing experiment) and this can permit a single input onto a single neuron to undergo LTP. Thus the requirement for synchronous activity to induce LTP is to provide sufficient depolarization. In conclusion, the Mg2þ block and Ca2þ dependence of the NMDAR explain the properties of input specificity, cooperativity, and associativity.

Role of g-Aminobutyric Acid Inhibition in LTP Role of Postsynaptic g-Aminobutyric Acid Receptors

g-Aminobutyric acid (GABA) inhibition plays a pivotal role in LTP. When several afferents are activated at a low frequency the EPSP is mediated essentially exclusively by AMPARs. This is only because of synaptic inhibition, which is also activated via the excitation of feed-forward interneurons. If GABA receptors are blocked pharmacologically, then the AMPA-mediated EPSP is prolonged and NMDARs contribute much more substantially to the later part of a composite EPSP. Thus, it is the hyperpolarization by GABA to maintain, and indeed intensify, the Mg2þ block that

323

prevents NMDARs contributing to any appreciable extent during low-frequency transmission. Thus, blocking GABA inhibition can facilitate the induction of LTP. However, the inhibition of GABA is not just some pharmacological curiosity. During the more physiological patterns of induction (priming and theta) there is an activity-dependent depression of GABA inhibition that is required to enable sufficient activation of NMDARs. This activity-dependent depression is mediated largely by the activation of a presynaptic GABAB autoreceptor. Role of GABAB Autoreceptors

During low-frequency stimulation, GABA is released and evokes a biphasic inhibitory postsynaptic potential (IPSP), mediated by GABAA and GABAB receptors, to restrict the synaptic activation of NMDARs (Figure 2). Some of the GABA that is released activates a presynaptic GABAB autoreceptor, which inhibits subsequent GABA release. During low-frequency transmission this has no effect on GABA transmission, since the effect fully recovers within a second or so. But during high-frequency transmission the autoreceptor mechanism leads to a depression in GABA inhibition. The maximum effect occurs after around 200 ms, which explains the effectiveness of this interval for theta and priming-induced LTP. The GABAB autoreceptor mechanism is essential for the induction of LTP under these conditions, since antagonizing this receptor completely blocks priming-induced LTP. However, the depression of synaptic inhibition only occurs for a few stimuli as the synapse undergoes an activity-dependent facilitation, and, when sufficient stimuli are delivered, a depolarizing potential is generated. Therefore when more artificial induction protocols are used, such as tetanus, blocking GABAB autoreceptors has no effect on the induction of LTP. Given that in the exploring animal there are brief periods of synchronized high-frequency activity occurring at the theta frequency, it is likely that GABAB autoreceptors are critically involved in the induction of LTP in situ.

Signaling Mechanisms Involved in LTP The induction of LTP requires an elevation of postsynaptic Ca2þ. This Ca2þ enters the synapse via the activated NMDARs, where the Ca2þ signal may be greatly magnified by Ca2þ-induced Ca2þ release from intracellular stores. It then activates various kinases. Several protein kinases have been implicated in the induction of LTP. These include Ca2þ/calmodulindependent protein kinase II (CaMKII), protein kinases A and C (PKA, PKC), protein tyrosine kinase (PTK), mitogen-activated protein kinase (MAPK), and

324 Long-Term Potentiation (LTP): NMDA Receptor Role

Excitatory terminal (glutamatergic)

AMPAR NMDAR

Inhibitory interneuron (GABAergic)

GABAAR GABABR

EPSP

a

Excitatory terminal (glutamatergic)

AMPAR NMDAR Inhibitory interneuron (GABAergic)

GABAAR GABABR

EPSP

b Figure 2 Role of g-aminobutyric acid (GABA ergic) inhibition in the induction of long-term potentiation. (a) Activation of glutamatergic inputs onto CA1 hippocampal neurons is accompanied by the activation of inhibitory GABAergic interneurons. These in turn provide an inhibitory input onto the pyramidal cells. GABA release is delayed with respect to glutamate release, as there are two synapses for the signal to pass through, rather than one. This results in a rapid depolarization due to a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor (AMPAR) activation followed by a rapid inhibition due to GABAA receptor (GABAAR) activation, ensuring the maintenance of the Mg2þ block. Sustained inhibition is then maintained by activation of postsynaptic GABABRs. Simultaneously, activation of presynaptic GABAB autoreceptors results in a temporary inhibition of GABA release (EPSP, excitatory postsynaptic potential). (b) A burst of high-frequency stimulation while GABA release is reduced (200 ms after the priming stimulus) results in sustained postsynaptic depolarization, overcoming the weakened inhibitory inputs. This sustained depolarization results in the release of Mg2þ from the N-methyl-D-aspartate receptor (NMDAR) and the influx of Ca2þ ions, leading to the induction of long-term potentiation.

phosphatidylinositol 3-kinase (PI3K). Why the need for multiple kinases? It seems that different kinases may be important at different developmental stages and for different phases of LTP. A full understanding of how kinases lead to persistent alterations in synaptic strength during LTP is far from complete.

Expression Mechanisms Involved in LTP Phases of LTP

Numerous studies have attempted to determine how AMPAR-mediated synaptic transmission is enhanced in LTP, and multiple mechanisms have been identified.

Long-Term Potentiation (LTP): NMDA Receptor Role

It is clear that different mechanisms occur at different stages of the LTP process. For example, the earliest phase of LTP, often referred to as short-term potentiation (STP) and most prevalent when a high-frequency tetanus is used to induce the process, probably involves an increase in L-glutamate release. The next phase, commonly referred to as early-phase LTP (E-LTP), most probably involves an alteration in AMPAR number or properties. A similar alteration in AMPAR function may also account for the next, protein synthesisdependent phase of LTP, known as late-phase LTP (L-LTP). However, the expression mechanisms of LTP seem also to be developmentally regulated, with presynaptic mechanisms appearing to play a more important role in E-LTP within the first week of life. Mechanisms of LTP Expression

Two ways in which AMPARs function at synapses may be directly altered during LTP have been identified. There can be an increase in the number of receptors or an increase of the efficiency of the receptors (Figure 3). In the latter case, the mechanism involves an increase in the single-channel conductance of the receptors, so that they pass more current in unit time. The extent to which this results from a modification of receptors already inserted in the synapse – for example, by phosphorylation, or by the exchange of high- for low-conductance receptors – is not yet known. One idea is that AMPARs lacking the GluR2 subunit can be transiently inserted during LTP. This has the effect of both increasing singlechannel conductance and providing an activitydependent Ca2þ signal, by permeation through the now Ca2þ-permeable AMPARs. This Ca2þ signal

Baseline response

325

might then drive the insertion of additional Ca2þimpermeable AMPARs. A number of proteins that interact directly with AMPARs have been found to affect their targeting and stabilization at synapses, or their functional properties. These include two membrane-associated guanylate kinases (MAGUKs) – synapse-associated protein 97 (SAP97) and postsynaptic density-95 protein (PSD95), the latter of which binds to an accessory AMPAR protein known as a transmembrane AMPA receptor protein (TARP). Another protein that binds to AMPARs and may be involved in LTP is PICK1 (‘protein interacting with C kinase 1’). Elucidating how these and other interacting proteins, together with small guanosine triphosphatases (GTPases) and protein kinases, function to regulate AMPARs during LTP is the focus of much current interest.

Metaplasticity Metaplasticity is a term that is used to refer to the plasticity of synaptic plasticity. With respect to NMDAR-dependent LTP this can be manifest in several ways. For example, a stimulus protocol (e.g., continuous low-frequency stimulation) that has no effect on baseline transmission can cause LTD of synaptic transmission following the induction of LTP. When the depression is conditional on prior LTP it is referred to as depotentiation. Two forms of depotentiation have been identified; one is dependent on the activation of NMDARs and the other is dependent on activation of metabotropic glutamate receptors (mGluRs). Similarly, two forms of LTD of baseline transmission (sometimes referred to as de novo LTD)

Increase in receptor number

Increase in single channel conductance AMPAR NMDAR Mg2+ Phosphorylation

Figure 3 Expression of long-term potentiation. There are multiple mechanisms that result in the long-term increase in synaptic strength that come under the umbrella of long-term potentiation. The number of a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptors (AMPARs) present in the postsynaptic membrane may be increased by exocytosis from an intracellular pool, or the activity of those AMPARs already present may be modified by, for example, phosphorylation. Additional mechanisms include the increase in glutamate release (NMDAR, N-methyl-D-aspartate receptor).

326 Long-Term Potentiation (LTP): NMDA Receptor Role

have been identified that are dependent on activation of NMDARs or mGluRs. The induction of LTP, while enabling depotentiation, can also inhibit the induction of NMDAR-dependent de novo LTD. The induction of NMDAR-dependent LTP can be affected by the prior history of synaptic activity. For example, prior activity at NMDARs can result in inhibition of subsequent LTP, while prior activity of mGluRs can result in facilitation of LTP. Finally, it should be noted that the NMDARmediated component of synaptic transmission is itself plastic – both LTP and LTD have been demonstrated. See also: AMPA Receptors: Molecular Biology and Pharmacology; Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDARDependent Forms; Metabotropic Glutamate Receptors (mGluRs): Functions; Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology; NMDA Receptor Function and Physiological Modulation; NMDA Receptors, Cell Biology and Trafficking.

Further Reading Alford S, Frenguelli BG, Schofield JG, et al. (1993) Characterization of Ca2þ signals induced in hippocampal CA1 neurones by the synaptic activation of NMDA receptors. Journal of Physiology 469: 693. Ault B, Evans RH, Francis AA, et al. (1980) Selective depression of excitatory amino acid induced depolarizations by magnesium ions in isolated spinal cord preparations. Journal of Physiology 307: 413–428.

Bliss TV and Collingridge GL (1993) A synaptic model of memory: Long-term potentiation in the hippocampus. Nature 361: 31–39. Bliss TVP, Collingridge GL, and Morris RGM (2007) Synaptic plasticity in the hippocampus. In: Andersen P, Morris RGM, Amaral DG, et al. (eds.) The Hippocampus, pp. 343–474. Oxford, UK: Oxford University Press. Collingridge GL, Kehl SJ, and McLennan H (1983) Excitatory amino acids in synaptic transmission in the Schaffer collateralcommissural pathway of the rat hippocampus. Journal of Physiology 334: 33–46. Collingridge GL, Isaac JT, and Wang YT (2004) Receptor trafficking and synaptic plasticity. Nature Reviews Neuroscience 5: 952–962. Dale N and Roberts A (1985) Dual-component amino-acidmediated synaptic potentials: Excitatory drive for swimming in Xenopus embryos. Journal of Physiology 363: 35–59. Davies CH, Starkey SJ, Pozza MF, et al. (1991) GABA autoreceptors regulate the induction of LTP. Nature 349: 609. Derkach VA, Oh MC, Guire ES, et al. (2007) Regulatory mechanisms of AMPA receptors in synaptic plasticity. Nature Reviews Neuroscience 8(2): 101–113. Herron CE, Lester RA, Coan EJ, et al. (1986) Frequency-dependent involvement of NMDA receptors in the hippocampus: A novel synaptic mechanism. Nature 322: 265–268. MacDermott AB, Mayer ML, Westbrook GL, et al. (1986) NMDAreceptor activation increases cytoplasmic calcium concentration in cultured spinal cord neurones. Nature 321: 519–522. Nowak L, Bregestovski P, Ascher P, et al. (1984) Magnesium gates glutamate-activated channels in mouse central neurones. Nature 307: 462–465.

Relevant Website http://www.bris.ac.uk – University of Bristol, Medical Research Council Centre for Synaptic Plasticity.

Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDAR-Dependent Forms C Lu¨scher, University of Geneva, Geneva, Switzerland M Frerking, Oregon Health and Science University, Beaverton, OR, USA ã 2009 Elsevier Ltd. All rights reserved.

How Is LTD Induced? Although N-methyl-D-aspartate receptor (NMDAR-) and metabotropic glutamate receptor (mGluR-) longterm depression (LTD) can both be elicited in vivo, the mechanisms of LTD have been best studied in the acute brain-slice preparation, where it can be induced with a number of protocols. At most synapses, NMDAR-LTD is induced by a modest activation of NMDARs, whereas stronger activation leads to longterm potentiation (LTP). Exceptions to this rule may apply to the parallel fiber (PF)–Purkinje cell synapse in the cerebellum. Activity-Dependent Induction

For NMDAR-LTD, prolonged low-frequency afferent stimulation (e.g., 900 stimuli at 1–5 Hz) has proven to be very efficient. Brief low-frequency afferent stimulation coupled to weak postsynaptic depolarization to –40 mV (to promote a modest relief of the Mg block of the NMDAR) can also efficiently induce NMDARLTD. In addition, NMDAR-LTD can be induced by the correlated activation of the pre- and postsynaptic neuron in many systems in which the firing of the postsynaptic neurons precedes the presynaptic action potential (AP) that releases the glutamate. The proposed mechanism to explain this spike timing-dependent LTD is that the AP in the postsynaptic cell backpropagates to the synapse, relieving the Mg block of the NMDAR. If the back-propagating AP precedes the release of glutamate, a low level of NMDAR activation is generated that leads to LTD. To induce mGluR-LTD, a wide variety of stimulation protocols have been used. Typically mGluR-LTD requires a higher stimulation frequency than NMDAR-LTD (up to 300 Hz) or burst firing (e.g., several repetitions of five stimuli at 66 Hz), most likely due to the extrasynaptic location of mGluRs. In the hippocampus, a typical induction protocols consists of stimulation at 5 Hz for 3 min or the delivery of pairs of presynaptic APs in rapid succession. Chemical Induction

Because activity-dependent LTD is generally specific to the small fraction of synapses that receive that activity,

there has also been interest in chemical induction of LTD with receptor-selective agonists. This approach is particularly useful for studying the mechanisms underlying both NMDAR-LTD and mGluR-LTD because such treatment uniformly changes transmission in the majority of synapses. NMDAR-LTD can be chemically induced by the exposure of a slice with NMDA for a few minutes. This chemically induced NMDAR-LTD occludes with NMDAR-LTD that is induced by synaptic stimulation, suggesting that the two overlap substantially. Similarly, mGluR-LTD can be induced by bath application of the drug dihydroxyphenylglycol (DHPG), a selective agonist of group I mGluRs. As with NMDARLTD, mGluR-LTD induced chemically occludes with mGluR-LTD induced by synaptic activity.

What Receptors Are Involved in the Induction of LTD? An interesting idea, introduced about 6 years ago, is that the subunit composition of activated NMDARs determines whether LTD or LTP is elicited. Based on the use of subtype-selective NMDAR antagonists, it has been reported that the NR2A-containing NMDARs selectively elicit LTP but not LTD, whereas NR2B-containing receptors selectively elicit LTD but not LTP in both the hippocampus and cortex. However, results from genetic manipulations suggest that NR2B can contribute to LTP. Several studies have also failed to replicate the reported NMDAR-subtype dependence of LTP or LTD in the hippocampus. LTD is, similarly, not obviously subunit-dependent in the anterior cingulate cortex. mGluR-LTD requires the activation of mGluRs, but there are several mGluR subtypes (mGluR1–8) and the particular mGluR leading to LTD appears to depend on the preparation in which it is induced. DHPG, a selective agonist of group I mGluRs (mGluR1 and mGluR5) elicits mGluR-LTD, which occludes with synaptically induced plasticity in many different systems, suggesting that in most cases mGluR-LTD is mediated by mGluR1, mGluR5, or both. The selective mGluR1 antagonist CPCOOEt blocks synaptically induced mGluR-LTD in the ventral tegemental area (VTA), and mGluR1 is also implicated in LTD in the cerebellum and neostriatum. mGluR5, on the other hand, has been shown to be the principal receptor responsible for mGluR-LTD in the hippocampus, cortex, and nucleus accumbens, although recent data in knockout mice indicate that the two

327

328 Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDAR-Dependent Forms

receptors may work in synergy. The activation of group II mGluRs (mGluR2 and mGluR3) has also been implicated in the CA3 and dentate gyrus of the hippocampus, as well as some regions of the cortex and the amygdala.

Intracellular Events Triggered by the Induction of LTD The crucial event triggered by NMDAR activation during NMDAR-LTD is calcium influx (see Figure 1). The requirement for modest, but not strong, NMDAR activation suggests that a modest increase in calcium can lead to NMDAR-LTD, whereas strong increases give rise to NMDAR-LTP. This notion is supported by manipulations of intracellular calcium in which weak and prolonged calcium stimuli lead to LTD, whereas rapid and robust calcium stimuli lead to LTP. The exact determinants of the time course and spatial extent of the calcium signal still have to be worked out, and it remains unclear whether [Ca] levels alone are sufficient to determine whether LTP or LTD is induced. However, an attractive explanation for the bidirectional effects of calcium is that the calcium-dependent effectors that are required for NMDAR-LTD are very sensitive to calcium, whereas the effectors for NMDAR-LTP are less so.

mGluR1/5

NMDAR

GluR2

GluR1

Ca2+

Gq PICK1 PKCa +

Kinase

+

Calcineurin

ser-880 NSF

GRIP/ABP

−

PKA

+

ser-845

PICK1 Endocytosis

PSD-95 SAP-97

AKAP

Conductance recycling

Figure 1 Schematic of intracellular signaling pathways involved in postsynaptic NMDAR-LTD and mGluR-LTD. Both forms of LTD involve the phosphorylation of ser-880 on the C-terminal tail of GluR2, which eventually increases endocytosis of mobile AMPARs. Mobility is conferred by exchanging GRIP/ABP for PICK1. NMDAR-LTD is, in addition, associated with a dephosphorylation of ser-845 on the C-terminal tail of GluR1, leading to decreased conductance and recycling. AKAP regulates the balance between constitutive PKA phosphorylation and activitydriven dephosphorylation of ser-845. AKAP, A-kinase associating protein; GRIP/ABP, glutamate receptor interacting protein/AMPA receptor binding protein; mGluR, metabotropic glutamate receptor; NMDAR, N-methyl-D-aspartate; receptor; NSF, N-ethylmaleimide sensitive factor; PICK1, protein interacting with C kinase 1; PKA, protein kinase A; PKC, protein kinase C; PSD, postsynaptic density protein; SAP, synapse-associated protein; ser, serine.

Consistent with this line of reasoning, NMDARLTD requires the calcium/calmodulin-dependent phosphatase calcineurin, which has a very high affinity for calcium-bound calmodulin (100 pM). The protein phophatase 1 (PP1) is also required for at least some forms of NMDAR-LTD. NMDAR-LTD leads to a dephosphorylation of ser-845 of the a-amino-3hydroxy-5-methyl-4-isoxazole propionic acid receptor (AMPAR) subunit GluR1. Thus, NMDAR activation leads to the activation of PPs, which leads to AMPA receptor dephosphorylation. A caveat, however, is that there is much yet to be understood regarding how phosphatase activity is required for NMDAR-LTD; activity-dependent NMDAR-LTD requires both calcineurin and PP1, but chemically induced NMDARLTD does not appear to require PP1. Moreover, the phosphatase that actually directly targets GluR1 has not yet been identified. mGluR-LTD depends on the activation of G-proteins of the pertussis-toxin-insensitive Gq and Ga11 family (see Figure 1). Group I mGluRs couple to phosphoinositide signaling via Gq. Gq seems to predominate in mGluR-LTD at most synapses, including in area CA1 of the hippocampus and at cerebellar PF synapses, where LTD was strongly reduced in Gq / mice, whereas in Ga11/ mice the reduction was much smaller. The events that happen downstream of Gq activation, however, are less clear. Gq is coupled to phospholipase C (PLC), which triggers the synthesis of inositol triphosphate (IP3) and protein kinase C (PKC). This ultimately leads to release of calcium from intracellular stores. PKC is clearly involved in mGluR-LTD in the cerebellum and the VTA, but in the hippocampus mGluR-LTD seems not to require PKC or even PLC. The involvement of several other signaling pathways has been suggested, including p38 mitogen-activated protein (MAP) kinase, the extracellular signal-regulated kinase, tyrosine phosphatases, and phosphoinositide-3-kinase. Further study to define the intracellular cascades activated by mGluRs in the hippocampus will be of interest.

Expression Mechanisms Underlying NMDAR- and mGluR-Dependent LTD There is a general agreement that at least some forms of NMDAR-LTD and mGluR-LTD are expressed as a postsynaptic reduction in AMPAR function, and evidence suggests that both forms of LTD act primarily via the clathrin-mediated endocytosis of AMPARs. However, there are several differences between these two types of LTD, and a number of significant questions remain to be addressed.

Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDAR-Dependent Forms 329 AMPAR Trafficking

AMPAR Subunits and LTD

The first evidence that NMDAR-LTD was due to a withdrawal of AMPARs from the cell surface came from a series of studies demonstrating that NMDAR activation could lead to the internalization of AMPARs in cultured hippocampal neurons and that NMDAR-LTD in hippocampal slices was blocked by disrupting interactions between key components of the endocytic machinery (the proteins amphiphysin and dynamin). LTD is abolished by blocking the interactions between AMPAR subunits and the AP2 complex, which links to clathrin and is involved in the formation of endocytic-coated pits. In addition, LTD is impaired by viral transfections of the dominant-negative form of the small GTPase Rab5, which is thought to be involved in the trafficking of vesicles between the plasma membrane and endosomal compartments. Thus, NMDAR-LTD requires several components of endocytotic machinery at different stages of the endocytic pathway, and NMDAR activation leads to the withdrawal of AMPARs from the cell surface. Some evidence also implicates receptor endocytosis in mGluR-LTD, although the issue has received less direct experimental attention. In the cerebellum, the postsynaptic infusion of peptides that interfere with the amphphysin–dynamin interaction blocks mGluRdependent LTD. Activation of group I mGluRs has also been found to cause a reduction in AMPAR clustering at synapses in hippocampal cell cultures that is, again, blocked by interfering with the amphiphysin – dynamin interaction. Thus, mGluR-LTD appears to lead to loss of synaptic AMPARs via receptor internalization. However, not all data are readily compatible with this model. NMDAR-LTD leads to a reduction in postsynaptic responsiveness to exogenous agonists, as does mGluR-LTD in the cerebellum. Surprisingly, however, mGluR-LTD induced by DHPG application in the hippocampus does not. One possible explanation is that in the hippocampus, mGluR-LTD leads to a selective internalization or dispersion of synaptic AMPARs while leaving extrasynaptic AMPARs unaffected. In fact, with both types of LTD the time course and location of endo- and exocytosis remains to be fully identified. In hippocampal neuronal cultures, NMDAR activation induces a rapid lateral movement that precedes internalization at extrasynaptic sites. Several studies also implicate a presynaptic reduction of transmitter release probability in mGluRLTD, particularly in hippocampus. It remains unclear whether mGluR-LTD has both presynaptic and postsynaptic components that operate in tandem, or whether AMPAR loss preferentially affects synapses which have a high release probability.

Because AMPARs are heteromultimeric complexes that can be composed of any of four different subunits, GluR1–4, there has been considerable emphasis on the possible differential involvement of specific AMPAR subunits in the expression of LTD. GluR2 The extreme C-terminus of GluR2 and GluR3 contains a postsynaptic density protein (PSD-) 95, Dlg, and ZO1 (PDZ) ligand sequence that is recognized by several proteins with PDZ domains (the glutamate receptor interacting protein (GRIP), the AMPA receptor binding protein (ABP), and the protein interacting with C kinase 1 (PICK1)). The binding of GluR2 to GRIP/ABP appears to be important in stabilizing AMPARs at the synapse and may also be involved in retaining AMPARs in intracellular pools. Blocking the interaction between GluR2 and PICK1 blocks the removal of GluR2 from the plasma membrane and recent data suggest a model in which the binding of PICK1 allows the mobility of AMPARs. The GluR2/ PICK1 complex can be disassembled by ATPase activity of the N-ethylmaleimide sensitive factor (NSF), which binds to GluR2 (but not GluR3) at a site upstream of the PDZ ligand. In addition to binding with GluR2, PICK1 also binds to the a isoform of PKC and to GRIP/ABP. This complex of PICK1, PKC, and GRIP/ABP recruits PKCa to synapses. GluR2 has a PKC phosphorylation site (serine 880 (ser-880)) that is near the PDZ ligand sequence, and the phosphorylation of GluR2 at ser880 disrupts the interaction between GluR2 and GRIP/ABP but not the interaction between GluR2 and PICK1. Thus, a simple model is one in which LTD induction activates PKCa, leading to the PICK1-mediated targeting of PKCa to synapses. This would promote phophorylation of ser-880 on GluR2, causing GluR2 to release GRIP/ABP and bind to PICK1, promoting lateral diffusion and ultimately internalization. An elegant series of studies suggest that just such a mechanism accounts for the mGluR-dependent LTD in the cerebellum. Cerebellar LTD requires PKC, and in particular PKCa. LTD induction leads to the phosphorylation of ser-880, and genetic manipulations that disrupt the PKC consensus site on GluR2, as do disruptions of the PDZ ligand site on PKCa that mediates its interaction with PICK1. PDZ domain-mediated interactions among GluR2, PICK1, and PKCa also appear to be required for cerebellar LTD. Peptides that disrupt the PDZ interactions between GluR2 and PICK1 attenuate cerebellar LTD without affecting basal transmission. Cerebellar LTD is also abolished in knockout mice lacking either GluR2

330 Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDAR-Dependent Forms

or PICK1. More specific evidence of the necessity for GluR2–PICK1 interaction is provided by the observation that cerebellar LTD is also impaired by genetic manipulations that disrupt the PDZ ligand on GluR2 or the PDZ domain on PICK1. Finally, interactions between NSF and GluR2 appear to be essential for the synaptic incorporation of AMPARs that are competent to undergo cerebellar LTD. Transfection of GluR3 into Purkinje cells from GluR2 knockout mice does not rescue cerebellar LTD, nor is the transfected GluR3 incorporated into synapses. However, a mutated version of GluR3 that contains the NSF binding site of GluR2 is incorporated into Purkinje cell synapses from GluR2 knockout mice, and these synaptically incorporated AMPARs are competent to undergo cerebellar LTD. Do similar mechanisms account for mGluR-LTD observed at other synapses or for NMDAR-LTD? It is clear that at least some synapses express a form of mGluR-LTD that cannot occur via a loss of GluR2containing AMPARs – at excitatory synapses on to dopamine cells in the VTA, for example, mGluR-LTD causes a shift in the biophysical properties of the AMPAR excitatory postsynaptic current (EPSC) that suggests a selective loss of AMPARs that lack the GluR2 subunit. At first glance, NMDAR-LTD also seems unlikely to act via similar mechanisms because NMDAR-LTD does not require PKC activity. However, NMDARLTD can lead to PKC-independent phosphorylation of ser-880 through an as yet unidentified kinase, raising the possibility that a similar mechanism can be engaged by the same phosphorylation mediated by a different kinase. Moreover, peptides that interfere with the interaction between GluR2 and PDZ domain-containing proteins do impair LTD, although there is disagreement about whether GRIP/ABP or PICK1 is the essential mediator of these effects. Peptides that disrupt the interaction between GluR2 and NSF have also been observed to cause a rundown of synaptic transmission and a concomitant impairment of LTD. However, a few notes of caution are in order. The NSF binding site on GluR2 overlaps substantially with a binding site for the AP-2 adaptor complex that links to clathrin and is involved in the formation of endocytic-coated pits. This AP-2 binding site is also present in GluR1 and GluR3, and the peptides used in initial studies of the NSF–GluR2 interaction also disrupt the interaction between AMPARs and AP-2. Experiments with more selective peptides suggest that AP-2 interactions with AMPARs are required for NMDAR-LTD, whereas NSF–GluR2 interactions are required for the rundown of basal transmission. Mutated versions of GluR2 in which the interactions with GRIP, ABP, and PICK1 are abolished can still

undergo normal NMDA-induced internalization in hippocampal cultures, and a mutated version of GluR2 in which the phosphorylation of ser-880 is disrupted is still competent to undergo LTD in hippocampal slices, albeit less efficiently than wild-type GluR2. More decisively, NMDAR-LTD is unaffected in GluR2 knockout mice, and it is actually enhanced in GluR2–GluR3 double-knockout mice. This makes it difficult to argue that either GluR2 or GluR3 is a critical component of NMDAR-LTD. One possible explanation is that the peptides used to selectively interfere with GluR2 function are not as specific as previously thought. This possibility is supported by the difficulties in distinguishing between NSF and AP-2 interactions with GluR2 and also by the observation that the peptides that interfere with GluR2– PDZ domain interactions can also impair PDZ domain interactions with distantly related kainate receptor subunits. GluR1 An alternative subunit-selective mechanism for NMDAR-LTD is suggested by the observations that the ser-845 residue on GluR1 is dephosphorylated by exogenous NMDA in cell culture or by more conventional LTD induction in hippocampal slices. Ser-845 is constitutively phosphorylated by protein kinase A (PKA), and PKA activation has little effect on AMPAR function under basal conditions, presumably because of basal phosphorylation. However, the inhibition of PKA can depress AMPAR-mediated synaptic transmission and occlude LTD. PKA and calcineurin are colocalized by the A-kinase associating proteins (AKAPs), which are targeted to synapses by interactions with the scaffolding proteins PSD-95 and synapse-associated protein (SAP-)97. An accumulating body of evidence suggests indicates that AKAPs are dynamically regulated during LTD to shift the equilibrium between PKA and calcineurin localization at synapses. This determines the phosphorylation state of ser-845, and manipulations that interfere with PKA binding to AKAPs cause a depression of AMPAR-mediated transmission that occludes LTD. Further experiments to selectively examine the role of ser-845 in NMDAR-LTD will be of interest. How might ser-845 dephosphorylation be involved in LTD? Phosphorylation of ser-845 can enhance AMPAR function by increasing the peak open probability of the channel, so calcineurin-mediated dephosphorylation of GluR1 could directly contribute to LTD expression even in the absence of changes in receptor trafficking. However, dephosphorylation of ser-845 appears to be downstream of receptor internalization and prevents the recycling of AMPARs after they have been internalized. Thus, current evidence suggests that the dephosphorylation of GluR1 is a

Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDAR-Dependent Forms 331

fairly late event in the mechanisms underlying LTD that may have more to do with preventing AMPAR reinsertion than with the withdrawal of AMPARs from the synapse. Maintenance of LTD

The previous section focused on the mechanisms by which AMPAR function is depressed during LTD. We now turn our attention to the mechanisms that allow the depression to persist long after the induction event is over. The answers to this question are still largely unknown, although the available data are sufficient in some cases to draw some surprising conclusions that deserve mention. In the case of NMDAR-LTD, some evidence is consistent with the idea that LTD is maintained by the change in phosphorylation of the AMPAR itself. As discussed previously, dephosphorylation at ser-845 on GluR1 prevents internalized AMPARs from recycling back to the synapse, suggesting that the phosphorylation state of this site may determine whether LTD persists; consistent with this idea, PKA activation can lead to a reversal of LTD following the stable induction of LTD, as can synaptic activity. Synaptically induced de-depression can be reduced by the inhibition of PKA, providing further evidence for dephosphorylation of PKA targets as a mechanism that is responsible for the maintenance of NMDAR-LTD. However, it need not be the case that NMDARLTD is maintained by the duration of modifications to AMPARs. An alternative is that the number of AMPARs at the synapse is limited by proteins that tether the AMPAR at the postsynaptic density; a decrease in the amount of this tether would preclude AMPAR reinsertion at synapses following LTD induction even if the AMPARs were recycled to the cell surface because there would be no mechanism to hold them at the synapse. The protein PSD-95 has many properties that are consistent with such a tether. Manipulations that increase or decrease PSD-95 expression cause parallel changes in the synaptic AMPAR-mediated response. This may be relevant to the maintenance of NMDAR-LTD, because NMDA applied to hippocampal neurons in culture leads to the ubiquitination and subsequent proteosomal degradation of PSD-95. This loss of PSD-95 is dependent on calcineurin and blocked by PKA activation, similar to LTD; moreover, the inhibition of proteosome function prevents the NMDA-induced reduction in surface expression of GluR1 and GluR2. NMDA also actively recruits proteosomes to dendritic spines. Further characterization of this potential mechanism for LTD maintenance will be of interest. In the case of mGluR-LTD, it has long been known that the LTD induced by DHPG can be reversed

long after stabilization by the application of group I mGluR antagonists. This indicates, surprisingly, that the expression mechanisms underlying mGluR-LTD are maintained by constitutive activation of an mGluR following LTD induction; in the absence of mGluR activity, the depression is quickly reversed. An alternative model that has some experimental support is that mGluR-LTD might be maintained by the synthesis of new proteins that stabilize LTD expression. Both LTD and the internalization of AMPAR subunits are blocked by inhibitors of mRNA translation. This potential link between translation of new proteins and mGluR-LTD became a source of considerable excitement when it was found that mice lacking the fragile X mental retardation protein (FMRP) have enhanced mGluR-LTD in both the hippocampus and cerebellum. The loss of function of FMRP is responsible for the fragile X syndrome in humans, an X-linked form of mental retardation. FMRP is an RNA-binding protein, although it remains a topic of debate whether FMRP represses or stimulates translation. Regardless of the precise action of FMRP on translation, the observation that FMRP enhances mGluR-LTD seemed at first to clearly indicate a role for changes in protein synthesis as an essential part of mGluR-LTD. Surprisingly, however, subsequent analysis of FMRP knockout mice indicated that mGluR-LTD in these mice is not only enhanced but also independent of protein synthesis. Similarly, the inhibition of proteosome function has recently been found to block mGluR-LTD in wildtype mice but not in FMRP knockout mice. These results are difficult to reconcile with an integral role for protein synthesis or degradation in mGluR-LTD, and further study will be of interest. See also: Long-Term Potentiation (LTP): NMDA Receptor Role; Metabotropic Glutamate Receptors (mGluRs): Functions; Metabotropic Glutamate Receptors (mGluRs): Molecular Biology, Pharmacology and Cell Biology; NMDA Receptor Function and Physiological Modulation; NMDA Receptors, Cell Biology and Trafficking.

Further Reading Anwyl R (2006) Induction and expression mechanisms of postsynaptic NMDA receptor-independent homosynaptic long-term depression. Progress in Neurobiology 78: 17–37. Bear MF (2003) Bidirectional synaptic plasticity: From theory to reality. Philosophical Transactions of the Royal Society of London. B. Biological Sciences 358: 649–655. Bear MF, Huber KM, and Warren ST (2004) The mGluR theory of fragile X mental retardation. Trends in Neuroscience 27: 370–377. Bellone C and Lu¨scher C (2005) mGluRs induce a long-term depression in the ventral tegmental area that involves a switch of the subunit composition of AMPA receptors. European Journal of Neuroscience 21: 1280–1288.

332 Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDAR-Dependent Forms Carroll RC, Beattie EC, vonZastrow M, and Malenka RC (2001) Role of AMPA receptor endocytosis in synaptic plasticity. Nature Reviews Neuroscience 2: 315–324. Cull-Candy S, Kelly L, and Farrant M (2006) Regulation of Ca2þpermeable AMPA receptors: Synaptic plasticity and beyond. Current Opinion in Neurobiology 16: 288–297. Dan Y and Poo MM (2006) Spike timing-dependent plasticity: From synapse to perception. Physiological Reviews 86: 1033–1048. HanleyJG(2006)MolecularmechanismsforregulationofAMPARtraffickingbyPICK1.Biochemical Society Transactions 34:931–935. Jorntell H and Hansel C (2006) Synaptic memories upside down: Bidirectional plasticity at cerebellar parallel fiber-Purkinje cell synapses. Neuron 52(2): 227–238. Lu¨scher C, Nicoll RA, Malenka RC, and Muller D (2000) Synaptic plasticity and dynamic modulation of the postsynaptic membrane. Nature Neuroscience 3: 545–550.

Malenka RC and Bear MF (2004) LTP and LTD: An embarrassment of riches. Neuron 44: 5–21. Malinow R and Malenka RC (2002) AMPA receptor trafficking and synaptic plasticity. Annual Review of Neuroscience 25: 103–126. Meng Y, Zhang Y, and Jia Z (2003) Synaptic transmission and plasticity in the absence of AMPA glutamate receptor GluR2 and GluR3. Neuron 39(1): 163–176. Oliet SH, Malenka RC, and Nicoll RA (1997) Two distinct forms of long-term depression coexist in CA1 hippocampal pyramidal cells. Neuron 18: 969–982. Segal M (2005) Dendritic spines and long-term plasticity. Nature Reviews Neuroscience 6(4): 277–284.

D-Serine:

From Its Synthesis in Glial Cell to Its Action on Synaptic Transmission and Plasticity

S H R Oliet and J-P Mothet, Institut National de la sante´ et de la Recherche Me´dicale U862, Bordeaux, France ã 2009 Elsevier Ltd. All rights reserved.

Introduction When thinking about brain signaling, neurophysiologists traditionally refer to the transfer of information from one neuron to another. Accordingly, the chemical synapse is viewed as the structure in which such communication occurs between the pre- and postsynaptic elements and where neurotransmitters are released from the presynaptic bouton to activate specific receptors localized on the postsynaptic neuron. This model, however, needs to be revised in view of the findings describing the role of glial cells as the third element of the chemical synapse. At the anatomical level, astrocytic processes enwrap up to 60% of the synaptic volume. Despite such an intimate structural relationship, astrocytes were believed to ensure only a housekeeping function at synapses. There is now accumulating evidence indicating that glial cells, and particularly astrocytes, contribute actively to synapse development, synaptic transmission, and neuronal excitability. These results fuel the emerging concept of the tripartite synapse, which considers astrocytes as an integral part of central and peripheral synapses. According to this model, glial cells first sense synaptic activity through a broad variety of ion channels, transporters, and receptors expressed at their surface. Synaptic activation of glial cells then triggers intracellular second-messenger pathways, including Ca2þ increases that vary according to the synaptic inputs solicited and the glial receptors involved. In turn, activation of these second-messenger pathways induces the release of active substances termed gliotransmitters (by analogy to neurotransmitters). These gliotransmitters mediating astrocyte-to-neuron signaling include glutamate, taurine, and adenosine triphosphate (ATP), which can be hydrolyzed in adenosine. Another major gliotransmitter which could have a major role in brain signaling is D-serine. The discovery of this amino acid in the brain has forced us to reconsider the dogma that only L-isomers of amino acids occur in mammals. Organic molecules, and therefore biological molecules such as D-amino acids, are based on the chemistry of the carbon atom. Carbon atoms can have up to four bonded groups attached to them in three-dimensional space,

forming a tetrahedron. Because of this structure, carbon-containing molecules can have the same four constituents, yet can differ in their structure by the location of the four groups in space. This property of carbon-based molecules is known as chirality. As an example, our hands are mirror images of one another, but they cannot be superimposed on one another. Similarly, carbon atoms with four different groups attached occur in two forms, known as enantiomers, that are not superimposable mirror images of one another (see Figure 1). Although the chemical and physical properties of L-amino acids and D-amino acids are extremely similar, only L-amino acids seemed to have been selected from the origin of life on the primitive Earth. In this chemical evolutionary step, D-amino acids were eliminated, and it was thought that all living organisms were composed only of L-amino acids. This asymmetry in biology is assumed to be a feature of fundamental physics because it turns out that natural L-amino acids are more stable than their ‘unnatural’ mirror images, D-amino acids. Until the last 30 years, it was believed that D-amino acids were excluded from living systems except for D-amino acids in the cell walls of microorganisms. Now, biologists have discovered that nature can deal with at least two D-amino acids, D-serine and D-aspartic acid, in higher organisms. Of these two atypical amino acids, D-serine is potentially very important for the nervous system because it is likely be the major endogenous ligand for the strychnineinsensitive glycine-binding site of N-methyl-D-aspartate receptors (NMDARs), key receptors for excitatory transmission and cognitive functions. Since the late 1980s, it has been established that activation of NMDARs required glutamate and a co-agonist that was first identified as glycine. Yet significant levels of D-serine are present in rodent brain areas enriched in NMDARs, suggesting that D-serine might be an alternative to glycine for NMDAR activation. D-Serine and serine racemase (SR), an enzyme that synthesizes D-serine from L-serine, are mainly, if not exclusively, localized to astrocytes. That glial-derived D-serine is an endogenous ligand of NMDARs and was unambiguously demonstrated by the use of D-amino acid oxidase (DAAO), an enzyme that selectively degrades D-serine but not glycine. Under conditions in which D-serine levels are dramatically reduced with DAAO, NMDAR activity is markedly impaired in different brain regions. It is now obvious that glial cells, through the release of D-serine, contribute actively to NMDAR function in the mammalian brain.

333

334

D-Serine:

From Its Synthesis in Glial Cell to Its Action on Synaptic Transmission and Plasticity

COOH

H

C

R

NH2

COOH

R

C

H

NH2

Figure 1 The chirality in your hands. The term chiral (from the Greek word for hand) is used to describe an object which is not superimposable on its mirror image, for example, our two hands. This property is based on the fact that carbon atoms (C) bind four groups – COOH, NH2, H, or any other radical (R) – in a threedimensional space. Such a tetrahedral hybridized carbon with four different groups always forms a chiral center. Conversely, a substrate carbon atom is not chiral if two out of four of its groups are identical, as it is in the case of the amino acid glycine.

Localization of D-Serine D-Serine is present in significant amounts in the brain of rodents and humans, where its levels (500 mM) are up to one-third of the total free serine pool. Its distribution is heterogeneous with the highest concentrations in the telencephalon and the developing cerebellum. This pattern of distribution throughout the rat brain resembles that of NMDARs. In contrast, glycine immunoreactivity does not correspond to NMDAR distribution excepted for some regions where D-serine is also present. Detailed analysis of D-serine staining shows that it is mostly present in astrocytes that ensheath synapses. In the developing cerebellum, D-serine is localized to Bergmann glia, whereas in adults it declines to negligible levels. Recent investigations have found that microglia cells as well as Schwann cells contain significant amount of D-serine and SR, which are also found in Mu¨ller glial cells in the retina and in the supporting cell of the vestibular sensory epithelium. Whether D-serine and SR are also present in other types of glial cells, such as oligodendrocytes, pituicytes, tanycytes, or ependymal cells, remains unknown. Although at very low levels, neuronal immunoreactivity for D-serine has been still observed in the neurons of the cerebral cortex, the brain stem, and the olfactory bulb.

Synthesis and Degradation of D-Serine The idea that D-amino acids and particularly D-serine serves specific roles in the brain was strengthened by

the discovery of its metabolic pathway. High levels of D-serine in the mammalian brain are generated by the activity of a pyridoxal 50 -phosphate (vitamin B6)-dependent enzyme, SR. This enzyme not only converts L- into D-serine, but also converts D- into L-serine, with a lower affinity. The enzymatic-catalyzed racemization of L-serine proceeds by removal of a proton from the asymmetric C–H bond of the amino acid to form a carbanion intermediate. The trigonal carbon atom of the carbanion, having lost the original asymmetry, then recombines with a proton to regenerate as an inverted tetrahedral structure corresponding to D-serine. The distribution of SR closely resembles that for D-serine, with the highest expression in the hippocampus and corpus callosum and very low levels in the amygdala, subthalamic nuclei, and the thalamus. Regarding cellular distribution, detailed analysis indicates that SR, like D-serine, is mostly confined to astrocytes, although substantial expression of SR is present in some neurons. However, due to the numeric preponderance of astroglia cells over neurons, glial cells remain the principal source of D-serine in the brain. SR activity can be negatively regulated by a series of cellular compounds. Thus, glycine and a series of metabolites related to L-aspartic acid competitively inhibit the enzyme. Because glycine concentrations in astrocytes are approximately 3–6 mM, glycine would constitutively inhibit SR activity unless glycine and SR show different compartmentalizations within the astrocyte cytosol. The metabolic pathway for D-serine degradation remains more elusive. Mammalian D-serine can be metabolized by the peroxisomal flavoprotein DAAO, an enzyme highly present in astrocytes of the hindbrain and cerebellum. Adult DAAO-deficient mice display increased D-serine levels, especially in areas where it normally occurs at low levels. However, although DAAO protein is present in the forebrain, an area enriched in D-serine, levels of the D-amino acid appear relatively unchanged in this region in DAAOdeficient mice. This result suggests that other mechanisms are probably implicated in regulating D-serine concentrations in this brain area. SR and DAAO may not work in isolation because SR and DAAO activities are controlled in opposite ways by nitric oxide (NO). Although NO enhances SR activity, it decreases that of DAAO, thus downregulating the intracellular levels of D-serine. In turn, D-serine inhibits NO synthase in glial cells and stimulates the production of NO in neurons. Neuronal NO could thus represent an inhibitory feedback mechanism tightly regulating D-serine metabolism in astrocytes and thereby preventing its overproduction and excessive stimulation of NMDARs.

D-Serine:

From Its Synthesis in Glial Cell to Its Action on Synaptic Transmission and Plasticity 335

Release of D-Serine Glial cells detect changes in their environment through a broad variety of ion channels, transporters, and receptors expressed at their surface. Specific activation of metabotropic and non-NMDA glutamate receptors on astrocytes induces the efflux of D-serine from these cells in culture (Figure 2). Whether activating these glutamate receptors is the only stimulatory pathway for D-serine release remains to be determined. Similarly, the conditions during which the activation of glial glutamate receptors, and, consequently, D-serine release occur in response to afferent synaptic activity are unknown. Interestingly, release of glutamate from glial cells also occurs when metabotropic and non-NMDA glutamate receptors are stimulated. An important issue for D-serine release, and for gliotransmission in general, is the identification of the molecular pathway responsible for the efflux of active substances from cells that were long thought to be passive and/or unexcitable. Astrocytes, like all eukaryotic cells, use secretory lysosomes to transport new membrane and proteins to the plasma membrane during constitutive exocytosis. Most cell types also possess a pathway of regulated exocytosis in which secretory vesicles undergo

Astrocyte

Actio

n pot

entia

l

Glutamatergic terminal

AMPAR KainateR mGluR D-Serine

AMPAR NMDAR Ca2+

Figure 2 Induction of D-serine release at glutamatergic synapses. Synaptically released glutamate diffuses locally to neighboring glial processes and binds to AMPA (AMPAR), kainate (kainaiteR) and/or metabotropic (mGluR) receptors. This triggers the release of D-serine within the synaptic cleft, enabling NMDAR activation on the postsynaptic neuron. AMPA, a-amino-3-hydroxy5-methyl-4-isoxazole propionic acid; AMPAR, AMPA receptor; kainaiteR, kainaite receptor; mGluR, metabotropic glutamate receptor; NMDAR, N-methyl-D-aspartate receptor.

Ca2þ-dependent fusion with the plasma membrane. Until recently, Ca2þ-regulated exocytosis was considered as a hallmark of neurons in the nervous system. Both constitutive and regulated secretory pathways require specialized proteins to bring together the membranes of the vesicles with the plasma membrane. The soluble N-ethylmaleimide-sensitive factor (NSF) attachment protein receptors (SNAREs) are leading candidates for mediating membrane fusion and most of them are present in glial cells. In addition, glial cells contain also synaptic-like microvesicles (SLMV) and large dense-core vesicles (LDCV), as revealed by electron microscopy and confocal microscopy analyses. Results obtained over the last decade using electrophysiological, biochemical, and imaging techniques suggest that astrocytes use a Ca2þregulated SNARE-dependent exocytosis to release glutamate from SLMVs but also ATP and peptides from LDCVs. In cultured astrocytes, activation of metabotropic and non-NMDA receptors effectively triggers a Ca2þ- and SNARE-dependent release of D-serine. In this experimental model, modifying extracellular or intracellular Ca2þ concentrations or depleting thapsigargin-sensitive Ca2þ stores, considerably impairs D-serine release. These results reveal that activation of non-NMDARs is the major source of extracellular Ca2þ and that mobilization of Ca2þ from internal stores is necessary for D-serine release. This is consistent with the observation that both inositol triphosphate (IP3)- and caffeine/ryanodinesensitive Ca2þ stores control the release from glial cells of glutamate, another gliotransmitter. Furthermore, inhibiting exocytosis with clostridial toxins or preventing amino acid uptake into vesicles with concanamycin or bafilomycin, impairs the release of D-serine from astrocytes. These results are thus consistent with a vesicular storage and release of the D-amino acid. Interestingly, astrocytic glutamate also appears to be stored in, and released from, vesicles. If glutamate and D-serine were co-stored in the same vesicles, this would be the perfect gliotransmitter cocktail to activate NMDARs. This hypothesis is indeed supported by the observation that some D-serine immunoreactivity is found in vesicles expressing the vesicular transporter for glutamate. The final proof that D-serine and glutamate are co-stored and co-released, however, awaits further investigations. The existence of a vesicular pathway does not exclude the possibility that other routes for D-serine release coexist in glial cells. Such nonvesicular pathways could involve processes that have been linked already to gliotransmitter release such as hemichannels, P2X7 receptors, volume-sensitive channels, and reversal of transporters (Figure 3).

336

D-Serine:

From Its Synthesis in Glial Cell to Its Action on Synaptic Transmission and Plasticity

D

D

a

b

Figure 3 Concentration of D-serine in the synaptic cleft and the number of NMDARs available: (a) control, (b) with glial withdrawal in lactating rats. In (a), the concentration of D-serine is sufficient to make most NMDARs on magnocellular neurons available for activation by synaptically released glutamate. In (b), glial withdrawal causes a reduction of synaptic D-serine levels, thereby diminishing the number of NMDARs available. NMDAR, N-methyl-D-aspartate receptor.

Another important issue regarding D-serine release, and more generally gliotransmission, is to know whether specialized sites for its release exist in astrocytes. Regarding D-serine, it is not clear whether its release is restricted to the fine processes apposed to NMDARs or whether it can occur from any compartment of the glial cell. Interestingly, astroglial cells have functionally distinct compartments, known as microdomains, where localized high Ca2þ increases occur and which tightly enwrap synapses. Furthermore, electron microscopy analysis has revealed that glutamate-containing vesicles are found in glial processes that contact dentritic spines bearing NMDARs. It is thus tempting to speculate that D-serine release occurs from such localized glial microdomains in close vicinity to the synapses and to NMDARs. This hypothesis is supported by electrophysiological recordings obtained in the hippocampus and in the hypothalamus demonstrating that the level of occupancy of the NMDAR glycine site is higher at synaptic than at extrasynaptic receptors. Although this could simply reflect the proximity of synaptically released glutamate, which is the putative trigger for D-serine release, it could also indicate the existence of microdomains for gliotransmitter release localized at the synapses. D-Serine

its hydrolysis by specific ectonucleotidases located in the extracellular compartment, and/or through cellular uptake as occurring for glutamate. Injection of D-serine into the lateral ventricle results in a preferential accumulation of the amino acid in glial cells, as expected if glial uptake is the main process involved in the clearance of the D-amino acid. Several putative candidates for D-serine transport have been identified on the membrane of glial cells but also on neurons. Glial cells express a Naþ-dependent transporter with low affinity for D- and L-serine and with characteristics similar to those of the alanine–serine–cysteine transporter (ASCT) system, which carries D-serine in cultured astrocytes as well as in isolated retina. Another neutral amino acid transporter, the ASC-1, has also been identified in neurons, and its cellular localization on axon terminals and dendrites suggests that it could contribute to the synaptic clearance of þ  D-serine. Finally, a novel Na /Cl sensitive transporter has been described in rat brain synaptosomes which, in contrast to the ASCT system, has limited affinity for other neutral amino acids, including cysteine and alanine. It is thus conceivable that multiple transport systems, some of which may still be awaiting identification, contribute concomitantly to the regulation of D-serine concentrations in the extracellular space.

Clearance

Like for most neurotransmitters, the signaling action of D-serine should be terminated by its clearance from the extracellular space. This could occur through metabolic breakdown, as it is the case with ATP and

D-Serine Contribution to Synaptic Transmission and Plasticity

NMDARs are peculiar ionotropic receptors in the sense that they require not one but two different

D-Serine:

From Its Synthesis in Glial Cell to Its Action on Synaptic Transmission and Plasticity 337

agonists to be activated: glutamate and glycine. NMDARs are tetramers formed by the association of two NR1 and two NR2 subunits. On the NR1 subunits, there is a strychnine-insensitive binding site for glycine. This so-called glycine site can be activated not only by glycine but also by D-serine, with a higher potency according to the subunit composition of NMDARs. The discovery that D-serine occurs naturally in the mammal brain has thus completely changed our view regarding the identity of the endogenous ligand for NMDAR glycine-binding site in the brain. Although glycine was first identified as such, it is obvious today that in several brain areas the endogenous ligand of the NMDAR glycine site is D-serine and not glycine. And because D-serine is released from glial cells, these cells are likely to play an active, if not essential, role in regulating NMDARdependent processes, including synaptic transmission, synaptic plasticity, rhythmic activity, activation of second messenger pathways, gene expression, and pathophysiological phenomenon such as excitoxicity and neurodegeneration. The first indication that D-serine could play a role in NMDAR function came from the observation that the D-amino acid was present in astrocytes ensheathing neurons bearing NMDARs. In vitro studies then revealed that D-serine was released from astrocytes on activation of their glutamatergic receptors. These findings strongly suggested that, in some regions of the brain, glutamate released from nerve terminals triggered D-serine release from glial cells, which, in turn, could modulate NMDARs localized on adjacent neurons. The functional role of D-serine was first studied in the hippocampus, where high densities of D-serine and NMDARs occur in the subiculum as well as in the CA1 and CA3 regions. In hippocampal neurons co-cultured with atrocytes, specific enzymatic degradation of D-serine with exogenous DAAO considerably reduces agonist-evoked and spontaneous NMDAR-mediated currents. Because DAAO does not affect glycine levels, this result unambiguously demonstrates that endogenous D-serine is required for NMDAR activity in this brain area. Similar results were obtained in the retina, where Mu¨ller glial cells appear to control NMDAR-mediated neurotransmission through the release of D-serine, as revealed by the inhibitory effect of DAAO on the NMDAR-mediated responses evoked by an agonist or light in retinal ganglia cells. The ability of D-serine to control NMDARdependent synaptic transmission has been confirmed through the use of a naturally occurring mouse strain lacking DAAO activity. In DAAO-deficient mice, the levels of D-serine are very high in the brain stem and spinal cord. As expected, NMDAR-mediated excitatory postsynaptic currents recorded from dorsal horn

neurons in the spinal cord are significantly augmented in these animals. Knockout mice for the transporter ASC-1 provides another experimental model for studying the relevance of D-serine in glutamatergic neurotransmission; these mice display NMDAR-dependent hyperexcitability, presumably resulting from elevated extracellular D-serine associated with a deficient clearance of the amino acid. Because in many brain areas NMDARs are responsible for the induction of long-term potentiation (LTP) and long-term depression (LTD), the cellular substrates for learning and memory in the mammalian brain, it is then important to know whether D-serine, and thus astrocytes, contributes to such forms of synaptic plasticity in the brain. The contribution of D-serine in LTP was shown in hippocampal cell cultures and brain slices where reducing D-serine levels with DAAO dramatically compromised the ability of high-frequency stimulation to induce LTP in CA1 pyramidal cells. This result confirms that D-serine, rather than glycine, is the endogenous ligand of NMDARs in this brain region, not only during basal transmission but also when synaptic inputs are stimulated intensively. Interestingly, it is commonly believed that senescence is associated with impaired NMDAR-dependent synaptic plasticity and notably LTP. In agreement with this hypothesis, a deficient LTP is observed in the hippocampus of a senescenceaccelerated mouse strain. This senescent-related deficiency in synaptic plasticity appears to be due to an impaired D-serine metabolism, as indicated by the reduced production of the D-amino acid measured in the hippocampi of senescent rats. In agreement with this observation, the hippocampal LTP is completely rescued when the D-serine is supplied to the tissue. The key role of D-serine in governing NMDAR activity in the hippocampus is further supported by the demonstration that it is the major, if not the only, endogenous ligand involved in NMDAR-mediated neurotoxicity in this structure. Neuronal death is indeed prevented when slices are treated with serine deaminase, another enzyme that degrades D-serine, whereas supplying the slices with glycine oxidase (GO), an enzyme that degrades specifically glycine, does not affect NMDAR-dependent neurotoxicity. Taken together, these findings suggest that astrocytic D-serine modulates NMDAR-dependent neurotransmission and synaptic plasticity in different brain regions. Because D-serine is synthesized and released from astrocytes, its action in these different regions depends on the astrocytic environment of neurons. Indeed, it is now accepted that glial coverage of neurons is not static and that it undergoes profound reversible anatomical remodeling in different

338

D-Serine:

From Its Synthesis in Glial Cell to Its Action on Synaptic Transmission and Plasticity

areas under different physiological and/or pathological conditions. The hypothalamo-neurohypophysial system (HNS) constitutes certainly the most striking example of such anatomical remodeling that can be observed physiologically because it occurs during lactation, parturition, and chronic dehydration. This system is made of magnocellular neurons, localized in the hypothalamic supraoptic and paraventricular nuclei, whose axons project to the neurohypophysis where their hormonal content, namely oxytocin and vasopressin, can be directly released in the bloodstream. The morphological plasticity of the HNS is characterized by a pronounced reduction in astrocytic coverage of oxytocin-secreting magnocellular neurons, which is entirely reversible on the cessation of the stimulation. Both SR and D-serine are found at high levels in this brain structure, and their expression is strictly restricted to astrocytes. The HNS thus provides a remarkable model for studying the physiological impact of glial-derived D-serine in the context of synaptic transmission and synaptic plasticity. As reported for the hippocampus and the retina, treating hypothalamic slices with the highly specific enzyme DAAO dramatically reduced NMDARmediated synaptic responses recorded in magnocellular neurons, whereas degrading specifically glycine with GO has no effect, thereby indicating that D-serine is the endogenous co-agonist of NMDAR in the HNS. Accordingly, D-serine levels within the synaptic cleft, and thus the level of occupancy of the glycine site of synaptic NMDA receptors, is reduced under conditions in which the astrocytic coverage of neurons is diminished. This results in NMDAR-mediated synaptic responses of smaller amplitude in lactating than in control animals, a difference that disappears when saturating concentrations of D-serine are supplemented. Because the reduction of D-serine concentrations within the synaptic cleft caused by glial withdrawal lowers the number of synaptic NMDARs available for activation, the activity-dependence of phenomena such as LTP and LTD, whose induction depends on NMDAR activation, is modified. The neuron–glia remodeling causes a shift of the activity-dependence of long-term synaptic changes toward higher activity values, similar to what can be observed when NMDARs are partially blocked with pharmacological agents. Therefore, the glial environment of neurons, through its capacity to provide D-serine, has a profound impact not only on NMDAR-mediated synaptic transmission but also on the direction and magnitude of NMDAR-dependent long-term synaptic plasticity. Such astrocyte-mediated metaplasticity is likely to exist at all synapses where endogenous D-serine is involved in regulating NMDARs.

Conclusion It is now clear that D-serine fulfils all criteria that identify it as the leader of a new class of brain messengers, the D-amino acids. Glial-derived D-serine plays essential roles in the mammalian brain. By modulating NMDARs, this amino acid contributes actively to the transfer and storage of information. Because of this role in governing NMDAR function, glial cells may become prime targets in pathological events that cause neurons to degenerate. It is well documented that over- or downstimulation of NMDARs are implicated in a large number of acute and chronic degenerative disorders, including stroke, epilepsy, peripheral neuropathies, Parkinson’s disease, Alzheimer’s disease, Huntington’s disease, and in psychiatric disorders such as schizophrenia. Whereas previous therapeutic approaches centered directly on NMDARs have been associated with deleterious side effects, the discovery of D-serine as the predominant endogenous ligand of the NMDARs thus provides new strategies for the development of drugs that target NMDAR function indirectly through the modulation of D-serine metabolism, for example, by inhibiting SR or DAAO. See also: Glial Influence on Synaptic Transmission; Glutamate; Glycine Receptors: Molecular and Cell Biology; Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDAR-Dependent Forms; Long-Term Potentiation (LTP): NMDA Receptor Role; NMDA Receptors, Cell Biology and Trafficking.

Further Reading Araque A, Parpura V, Sanzgiri RP, and Haydon PG (1999) Tripartite synapses: Glia, the unacknowledged partner. Trends in Neuroscience 22(5): 208–215. Boehning D and Snyder SH (2003) Novel neural modulators. Annual Review of Neuroscience 26: 105–131. Fields RD and Stevens-Graham B (2002) New insights into neuronglia communication. Science 298(5593): 556–562. Fujii N (2002) D-Amino acids in living higher organisms. Origins of Life and Evolution of Biospheres 32(2): 103–127. Haydon PG and Carmignoto G (2006) Astrocyte control of synaptic transmission and neurovascular coupling. Physiological Review 86(3): 1009–1031. Hirrlinger J, Hulsmann S, and Kirchhoff F (2004) Astroglial processes show spontaneous motility at active synaptic terminals in situ. European Journal of Neuroscience 20(8): 2235–2239. Johnson JW and Ascher P (1987) Glycine potentiates the NMDA response in cultured mouse brain neurons. Nature 325(6104): 529–531. Laming PR, Kimelberg H, Robinson S, et al. (2000) Neuronal-glial interactions and behaviour. Neuroscience and Biobehavioral Reviews 24(3): 295–340. Mothet JP, Parent AT, Wolosker H, et al. (2000) D-Serine is an endogenous ligand for the glycine site of the N-methyl-D-

D-Serine:

From Its Synthesis in Glial Cell to Its Action on Synaptic Transmission and Plasticity 339

aspartate receptor. Proceedings of the National Academy of Sciences of the United States of America 97(9): 4926–4931. Mothet JP, Pollegioni L, Ouanounou G, Martineau M, Fossier P, and Baux G (2005) Glutamate receptor activation triggers a calcium-dependent and SNARE protein-dependent release of the gliotransmitter D-serine. Proceedings of the National Academy of Sciences of the United States of America 102(15): 5606–5611. Panatier A, Theodosis DT, Mothet JP, et al. (2006) Glia-derived D-serine controls NMDA receptor activity and synaptic memory. Cell 125(4): 775–784.

Shleper M, Kartvelishvily E, and Wolosker H (2005) D-Serine is the dominant endogenous coagonist for NMDA receptor neurotoxicity in organotypic hippocampal slices. Journal of Neuroscience 25(41): 9413–9417. Theodosis DT (2002) Oxytocin-secreting neurons: A physiological model of morphological neuronal and glial plasticity in the adult hypothalamus. Frontiers in Neuroendocrinology 23(1): 101–135. Volterra A and Meldolesi J (2005) Astrocytes, from brain glue to communication elements: The revolution continues. Nature Reviews Neuroscience 6(8): 626–640.

GABA Synthesis and Metabolism K L Behar, Yale University School of Medicine, New Haven, CT, USA ã 2009 Elsevier Ltd. All rights reserved.

Introduction g-Aminobutyric acid (GABA) is the major inhibitory neurotransmitter in mammalian brain and is found widely throughout the central and peripheral nervous systems. In the neocortex GABAergic neurons are plentiful, constituting 15–30% of all neurons. GABA serves both metabolic and trophic functions, in addition to its role as a neurotransmitter, influencing the migration of neurons and astroglia to their target locations in the cortex. During early brain development GABA elicits excitatory (depolarizing) rather than inhibitory (hyperpolarizing) postsynaptic responses. Later in development GABA influences the synaptic organization and fine-tuning of local circuits. Because of GABA’s many important roles – as neurotransmitter, neuromodulator, trophic factor, and cellular metabolite – GABA metabolism may impact many aspects of brain function. Brain tissue is highly heterogeneous, and compartmentation of the enzymes of the GABA shunt between neurons and astroglia plays an important role in the function of GABA as a neurotransmitter. The discovery of metabolic compartmentation in the early 1960s to the late 1970s laid the conceptual foundation for our current understanding of brain glutamate and GABA metabolism. Studies using 14C- and 15N-labeled substrates revealed that brain glutamate metabolism was not homogeneous and could be separated kinetically into distinct pools – a large pool labeled from [14C]glucose and a small, rapidly turning over pool labeled from [14C]acetate, [14C]butyrate, and [14C] bicarbonate. The subsequent discoveries that glutamine synthetase (GS; glutamate–ammonia ligase, EC 6.3.1.2) and pyruvate carboxylase (PC; EC 6.4.1.1) were astroglial enzymes provided key information that the large and small pools of glutamate corresponded to metabolism in neurons and astroglia, respectively. Glutamatergic and GABAergic neurons are included within the definition of the ‘large pool,’ although glutamate levels in the respective neurons differ greatly. Most of the 8–12 mmol g 1 of glutamate measured in brain tissue is present in glutamatergic neurons (80%), with GABA neurons (2–10%) and astroglia (10%) containing much less. Astroglia play a critical role in glutamatergic and GABAergic function by maintaining low synaptic and interstitial levels of glutamate and GABA through active uptake

340

and metabolism. Astroglia also provide glutamine, and potentially other precursors, necessary for the resynthesis of these neurotransmitters. The study of brain glutamate and GABA metabolism has been advanced significantly in recent years through the use of nuclear magnetic resonance (NMR) spectroscopy in conjunction with 13C- and 15N-labeled substrates. The use of isotopically labeled substrates in these studies allows metabolic pathways of GABA and glutamate synthesis and neuron/glial substrate trafficking to be examined noninvasively and in unprecedented detail, revealing new insights into these pathways and their functional interactions.

GABA Synthesis and Degradation in the GABA Shunt GABA synthesis and degradation occur at the level of the tricarboxylic acid (TCA) cycle. GABA, a product of glucose metabolism, is synthesized almost exclusively from glutamate, a product of the TCA cycle intermediate, a-ketoglutarate, in an a-decarboxylation catalyzed by the pyridoxal-50 -phosphatedependent enzyme, glutamate decarboxylase (GAD; EC 4.1.1.15). GABA is degraded to succinate by the concerted actions of two enzymes: pyridoxal-50 phosphate-dependent GABA transaminase (GABA-T; 4-aminobutyrate transaminase, EC 2.6.1.19) and nicotinamide adenine dinucleotide (NAD)-dependent succinate-semialdehyde dehydrogenase (SSADH; EC 1.2.1.24). In the first step of catabolism, GABA nitrogen is transferred to a-ketoglutarate by transamination, forming succinic-semialdehyde and glutamate. In the second step succinic-semialdehyde is oxidized to succinate with reduction of NADþ to NADH2; succinic-semialdehyde is a short-lived intermediate and is present at low cellular concentrations. The pathway of GABA metabolism from a-ketoglutarate to succinate forms a metabolic shunt, by passing two enzymes of the TCA cycle, a-ketoglutarate dehydrogenase and succinyl-CoA synthetase (Figure 1). Glucose oxidation through the GABA shunt is slightly (3%) less energy efficient than through the complete TCA cycle, resulting in one less molecule of guanosine triphosphate (GTP) for each molecule of a-ketoglutarate traversing the shunt. GABA synthesis occurs only in neurons, whereas GABA catabolism occurs both in neurons and glia. Thus, the resynthesis of glutamate from a-ketoglutarate during transamination with GABA does not occur totally in the same cells in which GABA was formed. Because GABAergic and other neurons are incapable of de novo synthesis of glutamate precursors, due to

GABA Synthesis and Metabolism 341 Glucose Lactate

Pyruvate Acetyl-CoA

Citrate

TCA cycle Succinate

NADH2 SSADH

a-Ketoglutarate GTP NADH2 GABA-T

Glutamate

Succinic semialdehyde

GABA

GAD

histidine) and g-hydroxybutyrate (GHB). Homocarnosine is synthesized from GABA and histidine by homocarnosine synthetase, and hydrolysis of this dipeptide by a serum carnosinase regenerates GABA and histidine. GHB is synthesized from succinic-semialdehyde by a specific NADPH-dependent succinic-semialdehyde reductase, and oxidation back to succinic-semialdehyde is catalyzed by GHB dehydrogenase. Tissue levels of GABA, homocarnosine, and GHB can be elevated substantially by pharmacologic inhibitors or gene deficiencies in GABA-T and SSADH.

CO2 NAD+

Figure 1 Metabolic pathway depicting the synthesis and catabolism of g-aminobutyric acid (GABA) in the GABA shunt and its relationship to glucose metabolism and the tricarboxylic acid (TCA) cycle. The GABA shunt bypasses two enzymatic steps of the TCA cycle, resulting in one less molecule of guanosine triphosphate (GTP) formed for each molecule of a-ketoglutarate metabolized to succinate through the shunt. GABA synthesis occurs in GABAergic neurons but GABA catabolism occurs in neurons and astroglia. Astroglial glutamate precursors (e.g., glutamine) are required to replenish GABA removed and metabolized by astroglia. GAD, glutamic acid decarboxylase; GABA-T, GABA transaminase; SSADH, succinic semialdehyde dehydrogenase; NAD, nicotinamide dinucleotide; CoA, coenzyme A.

the absence of the necessary anaplerotic enzymes, glutamate precursors must be supplied by astroglia to maintain TCA cycle intermediates (and GABA) at constant levels. The concentration of GABA in brain tissue ranges from 1 to 5 mmol g 1, with the variation in concentration reflecting the density of GABA neurons and their terminals. GABA levels are very low in the extracellular fluid and in cells which do not express GAD. The GABA concentration within GABAergic terminals may reach 50–100 mM, although much of this is contained in synaptic vesicles and is inaccessible to GABA-T. The concentration of GABA in the cytoplasm and the mitochondrial intermembrane space where GABA-T is located is not well known, but levels are likely to be high and saturating for GABA-T, which has a relatively low Km (1 mM). Thus, in neurons GABA levels are likely to be regulated by changes in synthesis and GAD activity rather than catabolism. In contrast to neurons, in astroglia where GABA levels are low and nonsaturating for the glial transaminase, GABA level is likely to regulate the rate of catabolism. The GABA shunt also generates precursors for synthesis of a limited number of other compounds, the most studied of which are the neuromodulator/ neurotransmitters, homocarnosine (g-aminobutyryl

GABA Transporters Are Expressed in GABAergic Neurons and Astroglia GABA is transported across the plasma membranes of neurons and astroglia by electrogenic GABA transporters (GATs), which facilitate the reversible symport of GABA with two Naþ ions and one Cl ion. These transporters, like the electrogenic glutamate transporters, require the energy of ATP to maintain the Naþ ion gradient that provides the driving force and direction of GABA transport. GATs operate at much slower rates than do glutamate transporters, so that efficient clearance of GABA from the synapse requires relatively high transporter densities. The rate of GABA transport is determined by the number of functional GATs residing on the cell surface which are recruited from the cytoplasm. The distribution of GATs between cell surface and cytoplasm is regulated by extracellular GABA levels, protein kinase C, and ligands of G-protein-coupled receptors.

Pathways of GABA Clearance Following Release: Neuronal Reuptake and Glial Uptake GABA clearance following release into the synaptic cleft occurs both by reuptake into presynaptic terminals and by transport into surrounding astroglia. Isotopic studies which follow the labeling of GABA from different precursors cannot distinguish between the different uptake pathways because GABA metabolism in neurons and glia leads to the same labeling patterns. Thus, the relative quantitative importance of these two pathways to total GABA clearance in vivo is unknown and this issue remains a vexing problem. Studies using cell cultures have shown that GABA transport capacity and uptake are greater in neurons than in glia, supporting a predominant role for reuptake in GABA clearance from the synapse. Although reuptake has been suggested to be more energetically efficient because new GABA

342 GABA Synthesis and Metabolism

synthesis is not required, the activity of GABAergic neurons and vesicular loading of GABA has been shown to involve new GABA synthesis, thus weakening a rationale for reuptake. GATs are reversible and can function in GABA release as well as uptake. GATs in interneurons operate closer to their equilibrium potentials than do glutamate transporters, providing more favorable conditions for calcium-independent, nonvesicular transport of GABA in either direction. Under depolarizing conditions GABA transport reversal can lead to loss of GABA from the cell, but it has not been shown whether the recovery of GABA levels during repolarization involves reuptake or new synthesis from glial precursors. GABAergic synapses are closely surrounded by glial end-processes possessing high densities of GATs, and astroglia possess all the enzymes needed to degrade GABA. Studies of isolated glial fractions, astrocyte cultures, and intracellular recordings of glial GABA transporter currents in tissue slices show that glia have a substantial capacity to transport GABA. In situ levels of extracellular GABA and the occupancy of postsynaptic GABA receptors are regulated by astroglial transporters. Inhibitors of astroglial transporters show greater antiepileptic efficacy than do neuronal transport inhibitors of GABA, suggesting that glial clearance is quantitatively significant. These observations are consistent with the high rate of GABA synthesis from glutamine (23% of total glutamate plus GABA cycling) measured in rat cortex in vivo using 13 C NMR. Whether glial GABA transport is more important in paracrine signaling or synaptic transmission is unclear at present, but such differences may be related to the functions of the two isoforms of GAD, as discussed in a later section.

Precursors of GABA Synthesis

glutamate and ammonia by phosphate-activated glutaminase (PAG; EC 3.5.1.2), a highly regulated enzyme, present on the outer surface of the inner mitochondrial membrane, which provides glutamate for GABA synthesis to recharge cytoplasmic and vesicular pools. Release of GABA from neurons followed by uptake and metabolism to succinate in the astroglia completes the GABA shunt while regenerating glutamate (by transamination of a-ketoglutarate with GABA). Thus, de novo synthesis of a-ketoglutarate by anaplerosis is not required to replenish the GABA catabolized in astroglia. However, GABA catabolism requires an equivalent flow of acetyl-CoA from glucose-derived pyruvate or other substrates (e.g., acetate) to maintain the availability of a-ketoglutarate for transamination with GABA. The glutamate generated by transamination (GABA-T) can be converted to glutamine by GS and is released for uptake and replenishment of glutamate and GABA in neurons. In contrast to GABA, which requires oxidative metabolism in the TCA cycle, much of the glutamate removed by glia (70–80%) following release from glutamatergic neurons is converted directly to glutamine. A schematic of GABA and glutamate neurotransmitter cycling between GABAergic and glutamatergic neurons and astroglia is shown in Figure 2. Although glutamine has been shown to be a major glial precursor of GABA in vivo and in vitro, studies of isolated nerve terminal preparations indicate that glial TCA cycle intermediates (e.g., a-ketoglutarate or malate) might also contribute to GABA synthesis. In addition, the role of phosphate-activated glutaminase in GABA synthesis has been challenged; immunohistochemical findings in situ have shown that many neocortical interneurons which co-express certain peptides do not stain positively for PAG, suggesting that a substrate other than glutamine (e.g., a-ketoglutarate) might be utilized for GABA synthesis in these neurons.

The Glutamate/GABA–Glutamine Cycle

GABA Synthesis from Extracellular Glutamate

Glutamate precursors for GABA synthesis can be produced directly from the metabolism of glucose in neurons. However, glucose metabolism cannot replace glutamate and GABA carbon lost from neurons during activity due to the absence of the necessary anaplerotic enzymes (e.g., PC), which are expressed only in the astroglia. GABAergic neurons import the needed precursors from astroglia mainly in the form of glutamine. The astroglia which surround GABAergic (and glutamatergic) neuronal processes express high densities of the N-type electroneutral glutamine transporter (SN1), while GABAergic neurons express both of the A-type electrogenic glutamine transporters (SA1 and SA2). Glutamine is hydrolyzed to

In addition to glial precursors, extracellular glutamate can enter some GABA neurons through the EAAC1 glutamate transporter, where it can be used in GABA synthesis. This pathway has been observed in the hippocampus and thalamus but not in the cortex. Regional differences in GABA precursors might be related to the apparent heterogeneity seen in PAG immunostaining for different populations of GABAergic neurons. For interneurons using extracellular glutamate in GABA synthesis, glial precursors must ultimately replenish this glutamate in an extended loop, whereby glutamine is redirected first through the glutamatergic neuron and released as glutamate before uptake and conversion to GABA.

GABA Synthesis and Metabolism 343

Pyr

Glc

PDH (Vpdh(a))

Veff

Ac Glna

(VAc)

Glnecf

Pyr/lac

Glc

AcCoA

PC (VPC)

PDH (Vpdh(Gab)) Cit

OAA

PAG

GS (Vgln)

TCA cycle(a) (VTCA(a))

Mal

a-KGa

Glua

Glun(ecf) (Vcyc(Gab/Gln))

Suc

Gabaecf

GABA-T + SSADH (Vcyc(Gab/Gln)) a

Glc

Glnn

Astroglia

Pyr

PDH (Vpdh(a))

ECF

(Vcyc(Gab/Gln))

Glun GAD (VGAD) GABA

a-KGn

Suc GABA-T+ SSADH (VShunt)

GABAergic neuron

Veff

Ac

Glnecf

Glna

(VAc)

Glnn

Pyr/lac

AcCoA

PC (VPC)

TCA cycle(n) (VTCA(Gab))

Glc

PDH (Vpdh(Glu)) Cit

OAA

TCA cycle(a) (VTCA(a))

Mal

PAG (Vcyc(Glu/Gln))

GS (Vgln) (Vcyc(Glu/Gln)) a-KGa

Glua

Glu(ecf)

Glun

a-KGn

TCA cycle(n) (VTCA(Glu))

Suc

b

Astroglia

ECF

Glutamatergic neuron

Figure 2 Depiction of GABA and glutamate cycling with glutamine between GABAergic and glutamatergic neurons with astroglia via extracellular fluid (ECF). (a) GABAergic neuron and astroglial trafficking in the glutamate/GABA–glutamine cycle. (b) Glutamatergic neuron and astroglial trafficking in the glutamate–glutamine cycle. Enzymes and their catalyzed fluxes: Ac, acetate; AcCoA, acetyl-coenzyme A; aKG, a-ketoglutarate; Cit, citrate; GAD, glutamic acid decarboxylase; Glc, glucose; Gln, glutamine; Glu, glutamic acid; GS, glutamine synthetase; lac, lactate; Mal, malate; OAA, oxaloacetic acid; PAG, phosphate-activated glutaminase; PC, pyruvate carboxylase; PDH, pyruvate dehydrogenase; Pyr, pyruvate; Suc, succinate; VGAD, rate of GABA synthesis; Vshunt, rate of GABA shunt in GABAergic neurons; Vcyc(Gab/Gln), rate of glutamate/GABA–glutamine cycling; Vgln, rate of glutamine synthesis; Veff, rate of glutamine efflux; Vpdh(Gab) and Vpdh(a), rate of acetyl-CoA formation from pyruvate in GABAergic neurons and astroglia, respectively; VAc, rate of acetyl-CoA formation from acetate.

GABA Synthesis from Polyamines

Alternative but quantitatively minor pathways of GABA synthesis from putrescine and other polyamines have been described, but these pathways account for only a small fraction (1%) of synthesis in the mature brain. Polyamines may play important but largely unknown roles in GABA metabolism in the developing brain and retina. Putrescine and other

polyamines increase in response to stress, but a potential link to GABA synthesis has not been reported.

Nitrogen Homeostasis in Brain Glutamate and GABA Metabolism The operation of the glutamate/GABA–glutamine cycle requires the efficient transfer of nitrogen as well as

344 GABA Synthesis and Metabolism

carbon between neurons and astroglia. Nitrogen fixed as glutamine and released from astroglia for the synthesis of glutamate and GABA in neurons must ultimately return to the glia to permit continuous operation of the neurotransmitter cycles. Ammonia is freely diffusible through cell membranes, and the ammonia generated by PAG in the neurons could provide, in principle, the amide-N needed for glutamine synthesis from glutamate, part of which is acquired through uptake of extracellular glutamate or is produced by transamination with GABA. However, astroglia also undergo significant de novo synthesis of glutamate (anaplerosis) to replace glutamate and GABA lost by oxidation and diffusion, which may approach 18–30% of total glutamine synthesis, thus requiring a source of amine-N. Glutamate dehydrogenase (GDH; EC 1.4.1.2) is highly expressed in astroglia but is not believed to be sufficiently active in the direction of glutamate formation under normal physiological conditions. Recent studies suggest instead that amine-N for de novo glutamate (and glutamine) synthesis from a-ketoglutarate occurs by transamination with alanine or a branched-chain amino acid (BCAA), such as leucine. Two different nitrogen carrier cycles have been proposed – one involving an exchange of alanine for glutamine in the glia with the product of transamination, pyruvate, returning to the neuron as lactate (alanine/lactate shuttle), and the other involving an exchange of a BCAA (leucine) for glutamine with the product of transamination, branched-chain keto acid (a-ketoisocaproic), returning to the neuron. Alanine and leucine are regenerated in the neurons by transamination with glutamate, the latter formed by GDH and the ammonia from the PAG reaction. Both neurons and astroglia express the necessary transaminases needed to support the operation of these cycles, alanine aminotransaminase (AAT; EC 2.6.1.2) and branched-chain-amino-acid transaminase (BCAT; EC 2.6.1.42). Distinct cytosolic (neuronal) and mitochondrial (astroglial) isoforms of BCAT have been found. Studies of the BCAAs and their transaminases, together with findings of significant 15N labeling of brain glutamate in rodents during intravenous infusion of [15N]leucine in vivo, indicate that as much as 25% of brain glutamate nitrogen may be derived from leucine. While studies of nitrogen shuttling between neurons and astroglia have focused mainly on glutamatergic neurons, leucine is likely to be a source of glutamate nitrogen in some GABAergic neurons. GABAergic Purkinje neurons in the cerebellum and basket neurons in the hippocampus express BCATc in their cell bodies. In Purkinje neurons, PAG expression is low, suggesting that GABA synthesis involves substrates other than glutamine. In these neurons and others potentially not expressing PAG,

leucine transamination of a-ketoglutarate derived from astroglia during GABA catabolism or de novo synthesis could supply glutamate needed for GABA synthesis. Whether the GABA synthetic pathways operating in the many different classes of interneurons (e.g., chandelier, double bouquet, Martinotti neurons) differ from one another is largely unknown, but such differences could play an important role in their function.

Relationship of Cytoplasmic and Vesicular GABA to Different Release Pathways Studies of cultured neocortical neurons enriched in the GABAergic phenotype demonstrate the presence of cytoplasmic and vesicular GABA pools; one is released by agonists of the glutamate N-methyl-D-aspartate (NMDA) receptor and is calcium independent (cytoplasmic GABA), and the other is released by 55 mM potassium and is calcium dependent (vesicular GABA). The vesicular and cytoplasmic GABA produced from glutamine can be distinguished metabolically when neurons are depolarized in a manner related to the pathway of GABA release. Vesicular GABA produced from glutamine involves metabolic processing in the TCA cycle and equilibration of the glutamate carbon skeleton with endogenous mitochondrial glutamate. In contrast, glutamate produced directly from glutamine appears to be a better precursor of GABA released through reversal of membrane transporters (i.e., cytoplasmic GABA). Neuronal firing patterns may also play an important role in the mode of GABA release and thus the functional expression of the GAD isoforms, as discussed in the following section.

Relationship of the GAD Isoforms to Cytoplasmic and Vesicular GABA Synthesis GAD is expressed as two major isoforms, of 65 and 67 kDa, which are the products of two different genes. The two isoforms of GAD are distributed differently within GABAergic neurons: GAD67 is present throughout the cytoplasm, with more in the cell bodies, whereas GAD65 is highly enriched in axon terminals and is associated with synaptic vesicles. GAD67 appears to play the major role in basal GABA synthesis, and knockout of the gene for this isoform results in a major loss of tissue GABA. GAD67 expression is regulated both by transcriptional and by posttranscriptional mechanisms, whereas GAD65 expression is regulated at the transcriptional level and by short-term (kinetic) mechanisms. GAD65 knockout mice display electrophysiological features consistent with reduced vesicular release of GABA. GAD65 binding to synaptic vesicles was shown recently to involve a multiprotein

GABA Synthesis and Metabolism 345 GABAergic neuron Glucose

Astroglia

Pyr/Lac

GAD67

Glucose

PAG P

GABA Glutamine

(cytoplasmic) Pyr/Lac

Pyr/Lac

GS Glutamate

a-KG

PAG GABAcyto P

GAD65

GABAecf

GABA

GABA

suc

(vesicular)

GABA

GABA GABA Figure 3 An idealized depiction of the relationships between cytoplasmic and vesicular g-aminobutyric acid (GABA) synthesis, GABA release pathways, and glutamic acid decarboxylase (GAD) isoforms. Cytoplasmic GABA (GABAcyto) synthesis is provided mainly by GAD67, and this pool/isoform is associated with the majority of neuronal GABA shunt flux. Forward and reverse movements of GABA through GATs lead to uptake or release of GABA to the extracellular fluid (GABAecf). Vesicular GABA synthesis is catalyzed by GAD65, and GABA not returned to the terminals by reuptake is metabolized in the astroglia to succinate (suc) and oxidized in the mitochondrial TCA. Transamination of GABA in the glia by GABA-T generates glutamate from a-ketoglutarate (a-KG), which is used in the synthesis of glutamine by glutamine synthetase (GS). Phosphate-activated glutaminase (PAG) provides glutamate precursors for GABA synthesis. In some GABA neurons not expressing PAG, a-ketoglutarate could serve as a GABA precursor following its uptake and transamination to glutamate in the terminals, possibly by leucine or alanine.

complex of vesicle-associated proteins, the vesicular GABA/Hþ antiporter and Hþ-ATPase, and Ca2þ/calmodulin protein kinase II, which together facilitate vesicular loading of GABA driven by the hydrogen ion gradient. An idealized depiction of the relationship between the two GAD isoforms and the synthesis of the vesicular and cytoplasmic GABA pools is shown in Figure 3. GAD65 is more strongly regulated by the cofactor, pyridoxal phosphate, than is GAD67. The majority of GAD in brain is present as inactive apoenzyme (GAD without bound cofactor), which serves as a reservoir of inactive GAD that can be activated when demand for GABA increases. GAD65 binds pyridoxal phosphate less tightly than does GAD67 and accounts for the majority of the apoGAD in rat brain. In contrast, pyridoxal phosphate is more tightly bound to GAD67, which explains the importance of this isoform in basal synthesis. The interconversion between pyridoxal phosphate bound (active) and free (inactive) GAD involves a complex cycle of reactions catalyzed by GAD. This cycle is strongly regulated by nucleoside triphosphates (ATP) and inorganic phosphate,

which could serve as important regulators to increase GABA synthesis in response to increased demand. The activities of both GAD isoforms also appear to be regulated in opposing directions by protein phosphorylation.

Inherited Disorders of GABA Metabolism Rare inherited disorders of GABA metabolism have been described involving deficiencies in enzymes in the pathways of GABA catabolism (GABA-T, SSADH, and serum carnosinase) and a cofactor (pyridoxine) required in GABA synthesis. High levels of GABA accumulate in brain and cerebrospinal fluid of individuals deficient in GABA-T and SSADH. In SSADH deficiency, the buildup of the product of GABA-T, succinic semialdehyde, promotes increased synthesis and accumulation of GHB. Homocarnosine levels are elevated in brain and cerebrospinal fluid of individuals with homocarnosinosis, a condition resulting from deficiency of serum carnosinase, a dipeptidase which hydrolyzes homocarnosine and carnosine. Pyridoxinedependent epilepsy, a rare autosomal recessive inherited

346 GABA Synthesis and Metabolism

trait characterized by intractable seizures, is treated by administration of pyridoxine (vitamin B6) at pharmacologic doses. Although pyridoxine is a required cofactor of GABA synthesis, and reduced pyridoxine interaction with GAD is suspected in this disorder, a definitive mechanism has not been established. See also: GABAA Receptor Synaptic Functions; GABAA

Receptors: Molecular Biology, Cell Biology and Pharmacology; GABAB Receptor Function; GABAB Receptors: Molecular Biology and Pharmacology.

Further Reading Asada H, Kawamura Y, Maruyama K, et al. (1996) Mice lacking the 65 kDa isoform of glutamic acid decarboxylase (GAD65) maintain normal levels of GAD67 and GABA in their brains but are susceptible to seizures. Biochemical and Biophysical Research Communications 229: 891–895. Asada H, Kawamura Y, Maruyama K, et al. (1997) Cleft palate and decreased brain gamma-aminobutyric acid in mice lacking the 67-kDa isoform of glutamic acid decarboxylase. Proceedings of the National Academy of Sciences of the United States of America 94: 6496–6499. Bak LK, Schousboe A, and Waagepetersen HS (2006) The glutamate/GABA-glutamine cycle: Aspects of transport, neurotransmitter homeostasis and ammonia transfer. Journal of Neurochemistry 98: 641–653. Battaglioli G, Liu H, and Martin DL (2003) Kinetic differences between the isoforms of glutamate decarboxylase: Implications for the regulation of GABA synthesis. Journal of Neurochemistry 86: 879–887. Beckman ML, Bernstein EM, and Quick MW (1999) Multiple G protein-coupled receptors initiate protein kinase C redistribution of GABA transporters in hippocampal neurons. Journal of Neuroscience 19: RC9 1–6. Belhage B, Hansen GH, and Schousboe A (1993) Depolarization by Kþ and glutamate activates different neurotransmitter release mechanisms in GABAergic neurons: Vesicular versus non-vesicular release of GABA. Neuroscience 54: 1019–1034. Berl S and Clarke DD (1969) Compartmentation of amino acid metabolism. In: Lajtha A (ed.) Handbook of Neurochemistry, vol. 2, pp. 447–472. New York: Plenum. Chaudhry FA, Schmitz D, Reimer RJ, et al. (2002) Glutamine uptake by neurons: Interaction of protons with system a transporters. Journal of Neuroscience 22: 62–72. Conti F, Minelli A, and Melone M (2004) GABA transporters in the mammalian cerebral cortex: Localization, development and pathological implications. Brain Research: Brain Research Reviews 45: 196–212. Fonnum F and Walberg F (1973) An estimation of the concentration of g-aminobutyric acid and glutamate decarboxylase in the inhibitory Purkinje axon terminals in the cat. Brain Research 54: 115–127. Hutson SM, Berkich D, Drown P, et al. (1998) Role of branched-chain aminotransferase isoenzymes and gabapentin

in neurotransmitter metabolism. Journal of Neurochemistry 71: 863–874. Jin H, Wu H, Osterhaus G, et al. (2003) Demonstration of functional coupling between gamma-aminobutyric acid (GABA) synthesis and vesicular GABA transport into synaptic vesicles. Proceedings of the National Academy of Sciences of the United States America 100: 4293–4298. Kanamori K, Ross BD, and Kondrat RW (1998) Rate of glutamate synthesis from leucine in rat brain measured in vivo by 15 N NMR. Journal of Neurochemistry 70: 1304–1315. Kaufman DL, Houser CR, and Tobin AJ (1991) Two forms of the gamma-aminobutyric acid synthetic enzyme glutamate decarboxylase have distinct intraneuronal distributions and cofactor interactions. Journal of Neurochemistry 56: 720–723. Lieth E, LaNoue KF, Berkich DA, et al. (2001) Nitrogen shuttling between neurons and glial cells during glutamate synthesis. Journal of Neurochemistry 76: 1712–1723. Martin DL and Tobin AJ (2000) Mechanisms controlling GABA synthesis and degradation in the brain. In: Martin DL and Olsen RW (eds.) GABA in the Nervous System: The View at Fifty Years, pp. 25–41. Philadelphia: Lippincott Williams & Wilkins. Patel AB, de Graaf RA, Martin DL, et al. (2005) Evidence that GAD65 mediates increased GABA synthesis during intense neuronal activity in vivo. Journal of Neurochemistry 97: 385–396. Patel AB, de Graaf RA, Mason GF, et al. (2005) The contribution of GABA to glutamate/glutamine cycling and energy metabolism in the rat cortex in vivo. Proceedings of the National Academy of Sciences of the United States of America 102: 5588–5593. Pearl PL and Gibson KM (2004) Clinical aspects of the disorders of GABA metabolism in children. Current Opinion in Neurology 17: 107–113. Richerson GB and Wu Y (2003) Dynamic equilibrium of neurotransmitter transporters: Not just for reuptake anymore. Journal of Neurophysiology 90: 1363–1374. Seiler N (1980) On the role of GABA in vertebrate polyamine metabolism. Physiological Chemistry and Physics 12: 411–429. Sepkuty JP, Cohen AS, Eccles C, et al. (2002) A neuronal glutamate transporter contributes to neurotransmitter GABA synthesis and epilepsy. Journal of Neuroscience 22: 6372–6379. Shank RP, Bennett GS, Freytag SO, et al. (1985) Pyruvate carboxylase: An astrocyte-specific enzyme implicated in the replenishment of amino acid neurotransmitter pools. Brain Research 329: 364–367. Soghomonian JJ and Martin DL (1998) Two isoforms of glutamate decarboxylase: Why? Trends in Pharmacological Science 19: 500–505. Waagepetersen HS, Sonnewald U, Gegelashvili G, et al. (2001) Metabolic distinction between vesicular and cytosolic GABA in cultured GABAergic neurons using 13C magnetic resonance spectroscopy. Journal of Neuroscience Research 63: 347–355. Wei J, Davis KM, Wu H, et al. (2004) Protein phosphorylation of human brain glutamic acid decarboxylase GAD65 and GAD67 and its physiological implications. Biochemistry 43: 6182–6189.

GABAA Receptors: Molecular Biology, Cell Biology, and Pharmacology T Goetz, P Wulff, and W Wisden, University of Aberdeen, Aberdeen, UK ã 2009 Elsevier Ltd. All rights reserved.

Introduction If nervous systems possessed only excitatory neurons, there would be no brake on any signal, excitation would generate further excitation, and a chain of neurons would produce amplifying cascades of excitation, making the system incapable of doing anything useful. By contrast, networks made from both excitatory and inhibitory neurons can self-organize and generate complex properties. Inhibition is essential for every operation performed by any neuronal circuit in any brain region. In vertebrate brains, the inhibitory agent used most often by neurons is g-aminobutyric acid (GABA), which after release from the presynaptic terminals diffuses across the synaptic cleft to bind to receptor molecules – the so-called GABAA receptors. As for all other ligandgated channels, GABAA receptors convert chemical messages into electrical signals. In less than a millisecond, the binding of two (tiny) molecules of GABA induces a conformational change in the (giant) receptor oligomer that opens the central chloride ion channel. This remarkable process is called ‘gating.’ Cl diffuses through the receptor pore, down the electrochemical gradient, entering the cell and hyperpolarizing it. This results in fast synaptic inhibition (on the millisecond timescale) of a domain of the postsynaptic cell (Figure 1). Strictly, GABAA receptors are actually GABA-gated anion-permeable channels, with a HCO3 /Cl permeability ratio of approximately 0.2–0.4. In adult cells, Cl ions usually move into the cell to produce strong inhibitory hyperpolarization, as the reversal potential for Cl is 15–20 mV more negative than the resting membrane potential. The Cl gradient is maintained by K-Cl cotransporters. HCO3 ions move out of the cells through the GABAA receptor channel (Figure 1); the HCO3 efflux is mildly depolarizing (HCO3 has a reversal potential of 12 mV), but this is normally offset by the Cl hyperpolarization. Depending on the local internal Cl concentration, Cl ions can also move out after GABAA receptor activation and depolarize the cell; this happens especially during embryonic and postnatal neuronal development.

Molecular Biology of GABAA Receptors In mammals, GABAA receptors form as heteropentameric assemblies from a family of 19 subunits encoded by distinct genes (a1–a6, b1–b3, g1–g3, d, e, y, p, and r1–r3). Depending on the subunit composition, GABAA receptors differ in their biophysical properties and affinity for GABA, their pharmacology, and their location on the cell. GABAA receptors were cloned by the combined efforts of the Eric Barnard and Peter Seeburg groups in 1987 by the classical method: Peptide sequences obtained from purified (bovine brain) receptors were used to construct synthetic DNA probes to screen brain complementary DNA (cDNA) libraries. This was the starting point. By 1990, this now historical technique of screening cDNA libraries had revealed most of the gene family, all the a1–6, b1–3, g1–3 subunits and one d subunit; over the remaining decade, a few more subunits, such as e, y, and p, were characterized. With the completion of the human genome database, an in silico hybridization method was used to screen for further mammalian GABAA receptor genes, but it found none. Most of the subunit gene family members are in clusters, suggesting gene and then cluster duplication during the evolutionary origin of vertebrates: b2, a6, a1, g2 form a cluster in that order on human chromosome 5q34; the b3, a5, g3 genes cluster in that order on human chromosome 15q13; the g1, a2, a4, b1 genes cluster in that order on chromosome 4p12; the e, a3, y genes in that order on X q28; the r1 and r2 genes, 40 kbp apart on 6 q15; and the p, r3, and d subunit genes are isolated on human chromosomes 5q35.1, 3q12.1, and 1p36.3, respectively. As determined by both in situ hybridization (mRNA localization) with gene-specific probes (Figure 2) and immunocytochemistry (protein localization) with subunit-specific antibodies, the expression of the individual subunit genes is age and region specific. Does the clustering of the GABAA receptor subunit genes imply they are coregulated? The a1 and b2 genes do indeed share identical transcription patterns; thus, these two genes may share regulatory elements. All the other subunit genes have sometimes common, sometimes divergent expression patterns, with no correlation with which gene is in which gene cluster.

GABAA Receptor Structure The GABAA receptor belongs to a superfamily of ligand-gated ion channels (‘Cys-loop receptors’) that

347

348 GABAA Receptors: Molecular Biology, Cell Biology, and Pharmacology 2 Na+ GABA

Presynaptic terminal

GAT-1

GABA

vGAT

Vesicle

vGAT

H+

T

Glutamate GAD

a1

DCN

GP

GABA

II H

Cl–

Synaptic cleft

a2 VI

Postsynaptic site

HCO3−

GABAA receptor

Figure 1 The GABAergic synapse in the mammalian brain. GABA is synthesized from glutamate by glutamic acid decarboxylase (GAD) and transported into vesicles by the vesicular GABA transporter (vGAT). When the presynaptic terminal is depolarized sufficiently, GABA vesicles are released into the cleft and diffuse rapidly to activate the postsynaptic GABAA receptors. GABA is salvaged from the synaptic cleft by, for example, the GAT-1 transporter on presynaptic terminals. The GABAA receptor is a GABA-gated Cl and HCO3 channel.

a3 dg CPu

T BS

a4 Ctx

Cbm CA3

in vertebrates includes the nicotinic acetylcholine receptors, the 5-hydroxytryptamine type 3 (5HT3) receptors, the zinc-activated ion channel, and the glycine receptors. No direct structural information is available for GABAA receptors. So in imagining how the GABAA receptor must look, we can do no better than quote Unwin for his empirical observations on the Torpedo nicotinic acetylcholine receptor: The receptor (a large 290 kDa glycoprotein) is composed of elongated subunits, which associate with their long axes approximately normal to the membrane, creating a continuous wall around the central ion-conducting path. The whole assembly presents a rounded, nearly 5-fold symmetric assembly when viewed from the synaptic cleft, but is wedge-shaped when viewed parallel with the membrane plane. The subunits of the receptor all have a similar size 30 A  40 A  160 A and the same three-dimensional fold. Each subunit is a three-domain protein and so portions the channel naturally into its ligand-binding, membrane-spanning and intracellular parts.

In GABAA receptors, the arrangement of subunits around the channel is probably gbaba or dbaba counterclockwise when viewed from the extracellular space (Figure 3(a)). For those cells in which they are expressed, e and p subunits probably replace

a5 Cbm

gr a6 Figure 2 Expression of the GABAA receptor a subunit genes in adult rat brain (sagittal sections) as detected by in situ hybridization. BS, brain stem; CA3, subdomain of hippocampus; Cbm, cerebellum; CPu, caudate-putamen; Ctx, neocortex; DCN, deep cerebellar nuclei; dg, dentate granule cells; GP, globus pallidus; gr, cerebellar granule cells; H, hippocampus; T, thalamus. Roman numerals mark neocortical layers.

the g and d subunit within the pentamer, whereas the y subunit might replace a b subunit. As for all members of the nicotinic receptor superfamily, all GABAA receptor subunits contain a large extracellular N-terminal domain of approximately 200 amino acids shaped by a cysteine disulfide bridge (the socalled Cys-loop, Figure 3(b)). For GABAA subunits, the amino acid consensus sequence of the Cys-loop is C******F/YP*D***C*****S (where * is a degenerate residue).

GABAA Receptors: Molecular Biology, Cell Biology, and Pharmacology 349

Subunit Assembly Rules for GABAA Receptors b

b a1

g2

b

a1

a6

d

a6

b

a N

GABAA receptor subunit combinations are partly governed by which cell types express which genes and partly by preferential partnering of subunits within a given cell; for example, the a4 and a6 subunits assemble preferentially with the d subunit (Figure 4(a)). The majority of mammalian brain GABAA receptors are probably abg2 combinations. The subunit ratio is probably 2a/2b/1g. Some receptors also contain different a and b subunits, such as a1a2b2g2. The a1bg2 combination is the most abundant GABAA receptor subtype in the brain (60% of total; Figure 4(b)).

C C Out TM4

TM3

TM2

TM1

COOH

In b Figure 3 (a) Subunit arrangement in g2-containing (synaptic) and d-containing (extrasynaptic) GABAA receptors; (b) schematic of an individual subunit topology. C, cysteine; N, N-terminus of the protein TM, transmembrane domain. The pore lining domain, TM2, is shown in blue. Pink triangles represent GABA; blue circle represents a BZ ligand.

Each subunit contains four predicted transmembrane spanning domains (TM1–TM4) of about 20 amino acids and a large intracellular loop between TM3 and TM4 (TM3–TM4 loop) (Figure 3(b)). Many GABAA receptor subunits have the amino acid sequence (TTVLTMTT) in the TM2 domain. Five of these eight amino acids probably line the ion channel. TM1, TM3, and TM4 segregate TM2 from membrane lipid. The selectivity filter and gate lies at the intracellular end of the TM2 domains and includes part of the TM1–TM2 loop. In the 5-HT3 and nicotinic receptors, residues in the TM3–TM4 loop region influence single channel conductance. The TM3–TM4 loop contributes key sites for attaching anchor and regulatory proteins involved in locating the receptor at synapses and in governing the activity of GABAA receptor. In the extracellular domain of a typical GABAA receptor consisting of two a, two b, and one g2 subunit, the binding pocket for GABA forms at the interface between the a and the b subunit, and the binding pocket for benzodiazepines (BZs) lies at the interface of the a and the g subunit (Figure 3).

Synaptic GABAA Receptors: abg Subunit Combinations and Anchoring Role of the g2 Subunit and Gephyrin Placing GABAA receptors at synapses requires specific proteins that interact directly or indirectly with the g subunits. For example, targeting some GABAA receptor subtypes to GABAergic terminals involves the widely expressed microtubule-binding protein gephyrin. Gephyrin either helps convey some GABAA receptor subtypes to the synapse or anchors them there – this requires the g2 subunit. Without the g2 subunit, no GABAA receptors are found in synapses in the developing or adult hippocampus, and without gephyrin, much-reduced numbers of some synaptic GABAA receptor subtypes, especially those containing a2, are found; some receptor clusters, especially those containing the a1 subunit, persist in hippocampal gephyrin knockout neurons. Other g subunits can replace synaptic targeting function of g2; in g2 knockout mice, GABAA receptors can be restored to hippocampal synapses by expressing the g3 subunit by transgenic rescue. Some conserved sequence identity in the large TM3–TM4 intracellular loops of the g subunits may indicate binding sites for parts of the synapse anchoring mechanism. Distributed cysteine residues are conserved in the g subunit large intracellular loops but are absent from the b and d subunits. Palmitoylation of these cysteine residues via a thioester bond plays some role in targeting g subunit-containing receptors to the synapse. A surprising finding is that the TM4 region of the g2 subunit is also involved in synaptic targeting, possibly by interacting with lipid rafts occurring in the synapse or by other membrane proteins. The g2 TM4 is necessary and sufficient for postsynaptic clustering of GABAA receptors, whereas the cytoplasmic g2 subunit domains are dispensable. In contrast, both the TM3–TM4 loop and the TM4 domain of the g2 subunit contribute to efficient recruitment of

350 GABAA Receptors: Molecular Biology, Cell Biology, and Pharmacology b2

a4

d b2 a4

d

b2

a4 T

a 4b 2d

a a1

OB

b2

g2 b2 a1

g2 H a1

Cb

b2

a 1b 2g 2

b Figure 4 Some GABAA receptor subunit combinations found in adult rat brain: (a) a4b2d; (b) a1b2g2. The gene expression is visualized by in situ hybridization (adult rat brain, horizontal sections). Cb, cerebellum; H, hippocampus; OB, olfactory bulb; T, thalamus. The a4b2d receptors are particularly prominent in the thalamus; the a1b2g2 subtype is widespread.

gephyrin to postsynaptic receptor clusters. Thus, the g2 subunit TM3–TM4 cytoplasmic loop might be needed for inserting receptors into the plasma membrane but is dispensable for delivery of receptors to subsynaptic dendritic sites. Gephyrin does not bind the g2 receptor subunit directly. The identity of the missing link(s) between gephyrin and GABAA receptor subunits is unknown.

GABAA Receptor Occupancy at Synapses Is Dynamic GABAA receptor expression on the surface of neurons is dynamic; receptors rapidly recycle and leave from or insert into the synapse by lateral diffusion and/or endo/exocytosis. GABAA receptors diffuse into a synaptic zone and are transiently ‘captured’ by the anchoring complex. However, for some inhibitory hippocampal synapses, a direct relationship exists between the number of synaptic GABAA receptors and the strength of the synapse, but it is not clear what mechanisms maintain fixed numbers of GABAA receptors long-term at specific synapses. As for glutamate receptors at excitatory synapses, neurons probably recycle GABAA receptors as a strategy for setting their degree of excitability. GABAA receptors constitutively internalize by clathrin-dependent endocytosis; this requires interactions between the b and g2 subunits and the AP2 adaptin complex.

Different Synaptic GABAA Receptor Subtypes Can Be Enriched on Different Domains of the Same Neuron As mentioned above, considerable GABAA receptor complexity arises from differential subunit gene expression and specific subunit assembly rules. In addition, there are undefined determinants specifying that particular abg2 subunit combinations are enriched at different postsynaptic locations on the same cell. An example of GABAA receptor complexity on a single cell type is provided by a hippocampal pyramidal cell (Figure 5). A pyramidal cell is covered with GABAergic terminals; a typical rat CA1 pyramidal cell receives around 1700 GABAergic synapses, with the highest density on the perisomatic region. Inhibition on different domains affects different aspects of pyramidal cell function. GABAA receptor-mediated inhibitory postsynaptic currents can be generated along the whole somatodendritic domain and on the axon-initial segment (AIS) of CA1 pyramidal cells. The apical dendritic trunk has a high density of GABAergic terminals relative to the rest of the dendrite; strong GABAergic stimulation onto this apical trunk region will isolate the dendritic compartment from the cell body. The inhibition on the more distal dendrites controls Ca2þ spike propagation. Inhibiting the AIS will powerfully clamp down the pyramidal cell’s activity as this is where action potentials initiate – a typical rat CA1 pyramidal cell has more

GABAA Receptors: Molecular Biology, Cell Biology, and Pharmacology 351

Extrasynaptic GABAA Receptors: a4bd and a6bd Subtypes Dendrite

Soma

Parv CCK Axon

GABA Parv Basket cell GABA Parv

Basket cell GABA CCK

Axon hillock

Axo-axonic cell

= a 1b g 2 = a 2b g 2 Figure 5 Enrichment of a1- and a2-subunit-containing abg2 GABAA receptors opposite different classes of GABAergic terminal (cholecystokinin (CCK)- or parvalbumin (Parv)-containing) on a hippocampal pyramidal cell. Note that for clarity, only a small subset of GABAergic interneurons that project onto any given pyramidal cell are shown.

than 25 GABAergic terminals per 50 mm of the AIS. Thus, GABAA receptors in the AIS deliver inhibition which controls the overall level of output activity of pyramidal cells (Figure 5). GABAA receptors show differential subunit distribution at different synapses of the same pyramidal cell. On hippocampal pyramidal cells, the a2 subunit is enriched in the AIS but is present at only a minority of cell body synapses and synapses on dendrites. The AIS synapse also contains the a1 subunit and possibly the a4 and a5 subunits. GABAA receptors at the AIS are positioned to exert strong inhibitory influence over action potential generation. Synapses formed by two types of presynaptic interneuron cells (GABAergic basket cells) on the soma of CA1 pyramidal cells contain distinct GABAA receptor subtypes: Only those synapses coming from parvalbumin-negative/CCK-positive basket cell interneurons contain the a2 subunit; synapses from parvalbumin-positive basket cells are often a2immunonegative. This is a preferential targeting or enrichment of the a2 in a particular type of synapse – not an absolute selectivity. Both synapse types contain the same level of immunoreactivity for the b2/3 subunits; both types of synapse made by basket cells onto pyramidal cell somata also contain the a1 and the g2 subunits. However, the a1 subunit is less abundant in the parvalbumin-negative pyramidal cell synapses than in the parvalbumin-positive ones (Figure 5).

Besides mediating precisely timed synaptic pointto-point inhibition (phasic inhibition) via g2 subunitcontaining receptors, GABAA receptors can convey less-time-locked signals. Low GABA concentrations in the extracellular space, resulting from GABA diffusing from the synapse, can tonically activate extrasynaptic GABAA receptors. This ‘tonic inhibition’ is temporally uncoupled from the fast synaptic events, causing a continually present background inhibitory conductance. Such conductances alter the input resistance of the cell and thus influence synaptic efficacy and integration; tonic extrasynaptic conductances, by increasing the electrical leakiness of the dendritic membrane, substantially and indiscriminately diminish the size of excitatory signals in dendrites (Figure 6). Receptors with the d subunit, a4bd in forebrain and a6bd in cerebellar granule cells, are extrasynaptic; d subunits are perisynaptic (annular), localized around the edge of synapses in hippocampal dentate granule cells, and totally extrasynaptic on cerebellar granule cells. In all regions so far tested (cerebellar granule cells, hippocampal dentate granule cells, thalamic relay nuclei), d subunits contribute to GABAA receptors that provide an extrasynaptic tonic conductance. For GABAA receptors containing a4bd and a6bd subunits, their key properties are high affinity for neurotransmitter and limited desensitization, enabling them to contribute to tonic background conductances (Figure 6).

GABAA Receptor Agonists, Antagonists, and Allosteric Modulators GABAA receptors display a rich pharmacology. Generic GABAA receptors are selectively activated by the GABA agonist muscimol and blocked competitively by the GABA antagonists bicuculline and SR95531 (receptors assembled with r subunits are bicuculline and barbiturate insensitive, having their own unique pharmacology). Picrotoxin blocks GABAA receptors noncompetitively, probably by binding to a site in the channel. Many drugs bind at sites on the GABAA receptor distinct from the GABA binding site; these drugs change the shape of the receptor oligomer so that the efficacy of GABA at opening the channel is either increased (positive allosteric agonists, e.g., BZs such as diazepam) or decreased (negative allosteric agonists, e.g., b-carbolines). A few allosteric modulators occur naturally in the brain (e.g., Zn2þ, neurosteroids). Generally, positive allosteric agonists are used widely in medicine (e.g., for the induction and maintenance of general anesthesia or to treat anxiety disorders, states of agitation, epilepsy, or

352 GABAA Receptors: Molecular Biology, Cell Biology, and Pharmacology

GABA

a 4b d

a 1b g 2 Phasic

Tonic Gaboxadol

BZ

pA

pA s

Gaboxadol

ms GABA GABA + BZ

Figure 6 Synaptic and extrasynaptic GABAA receptors conferring tonic (a4bd or a4bd) and phasic (g2) inhibition, respectively. Synaptic (g2-containing) GABAA receptors can be activated by BZs and BZ-like ligands (e.g., zolpidem). BZs act to increase the peak amplitude and slow the rate of decay of the Cl current (red trace). Synaptic transmission is on the millisecond timescale. Extrasynaptic (d-containing) receptors can be selectively activated by Gaboxadol (red). Note the longer timescale.

sleep disorders), and there is scope to develop these drugs further to produce receptor subtype-selective drugs with fewer side effects; however, negative allosteric agonists also have potential clinical applications; for example, the drug L-655 708 works selectively at a5bg2 receptors (a subtype mainly expressed in the hippocampus), and by decreasing GABA’s action there, it acts as a cognition enhancer. A feature of all allosteric modulators is that they usually work only when GABA is at submaximal activating concentrations (below 1 mmol l 1), and they do not work in the absence of GABA (with the exception of some intravenous anesthetics). Nevertheless, some modulators (e.g., BZs) also strongly influence the deactivation rate of the receptors, even at peak synaptic GABA concentrations, and this may be how some of their in vivo effects originate. In the following sections, the drugs that act on particular GABAA receptor subunit combinations are considered in more detail. GABAA Receptors: Allosteric Modulation by BZs and Related Ligands

The main effects of BZs are sedation, anxiolysis, suppression of seizures, and muscle relaxation. These drugs require abg2-type receptors (a1b2g2, a2b2g2, a3b3g2, and a5bg2) with the drug-binding site located between the a and g2 subunits (Figure 3(a)). Note that a4bg2- and a6bg2-type receptors are insensitive to most BZ drugs, as are any receptors that contain the

d subunit. The substances that act at the BZ binding site include the classical BZs like diazepam or flunitrazepam, as well as chemically different substances, such as the imidazopyridine zolpidem (relatively selective for a1bg2-type receptors), which is used as a sleeping pill. The BZ site is situated at the interface between the a and the g subunit. In the a1, a2, a3, and a5 subunits, a mutation from histidine to arginine at position 101 abolishes binding of classic agonists like diazepam. The diazepam-insensitive a4 (or a6) subunits naturally contain an arginine residue at the homologous position, and so a4bg2 or a6bg2 receptors are insensitive to most BZ ligands. In a beautiful series of studies, mice with specific mutations in the key H101 coding position affecting BZ sensitivity were generated in the a1, a2, a3, and a5 subunit genes, and the behavioral effects of diazepam were tested. These mice have normal GABAA receptors, but in a1H101R mice, for example, only the a2bg2-, a3bg2-, and a5bg2-type GABAA receptors are diazepam sensitive. Thus by a process of subtraction, it can be deduced how different a1bg2, a2bg2, a3bg2, and a5bg2 subtypes contribute to the diverse in vivo pharmacological effects of diazepam and other ligands requiring the H101 site. Thus, a1H101R mice no longer become sleepy when given diazepam, and so the a1bg2 receptors are required for the sedative effects of diazepam, whereas the a2 (and a3) mediates diazepam’s anxiolytic effects (under the influence of diazepam, a2H101R mice do not venture more into threatening areas, whereas their wild-type littermates do). A different set of studies using a3 selective agonists and inverse agonists also showed a significant contribution of the a3 subunit in anxiogenesis and anxiolysis. The muscle relaxant activity of diazepam is mediated by the a2 and a3 subunits, probably because these subunits are expressed in spinal motor neurons. New Subtype-Selective Drugs for GABAA Receptors That Work at the BZ Site

GABAA receptors have always been fertile ground for drug companies. BZs, although for many years the mainstay of clinical treatments for anxiety disorders, fell out of favor to selective serotonin reuptake inhibitors (SSRIs) because of side effects like sedation, cognitive impairment, and abuse liability. But SSRIs are too slow acting for some situations, requiring several weeks to work. Thus there is a medical need for fast-acting anxiolytics with few or no side effects. The knock-in mouse studies mentioned above suggest that a2-containing GABAA receptors would be a prominent target for developing drugs which would be specifically anxiolytic.

GABAA Receptors: Molecular Biology, Cell Biology, and Pharmacology 353 GABAA Receptors: Allosteric Modulation by Intravenous Anesthetics

At clinically relevant concentrations, general anesthetics modulate the activity of various ion channels. Whereas volatile anesthetics (e.g., halothane, enflurane, or isoflurane) are positive modulators of recombinant GABAA receptors, the intravenous anesthetics (e.g., barbiturates, steroidal anesthetics, propofol, and etomidate) can modulate GABA’s action at the receptor but can also activate the receptor directly in the absence of GABA at higher concentrations. Based on the analysis of knock-in mouse lines with propofoland etomidate-insensitive b subunits (see below), propofol and etomidate exert nearly all their anesthetic actions entirely through GABAA receptors, whereas volatile anesthetics produce their actions through many diverse ion channel targets. The action of etomidate and propofol requires residues in TM2 and TM3 in the b2 or b3 subunits. A mutation of asparagine to methionine at position 265 (N265M) in the second transmembrane domain of the b3 subunit abolishes the modulatory and direct effects of etomidate and propofol in recombinant receptors. A mutation of aspargine at the same position in the b2 subunit also abolishes the action of etomidate on the GABAA receptor. In b3(N265M) mice, propofol and etomidate do not suppress noxious-evoked movements and show a strongly decreased duration of the loss of righting reflex, two different endpoints of anesthesia. These results suggest that propofol and etomidate act mainly via the GABAA receptor and the b3 subunit in particular to induce deep anesthesia. The remaining effects of propofol and etomidate could be mediated by b2 subunit-containing receptors. Studies on b2(N265S) mice have suggested that the b2 subunit mediates the sedative effects of etomidate, whereas the b3 subunit is required for etomidate to induce a loss of consciousness. A highly interesting issue is the location in the brain where etomidate and propofol exert their anesthetic effects. Is the modulation of GABAA receptors in specific nuclei required to induce anesthesia or do these drugs produce global effects at many GABAA receptors in all brain circuits? GABAA Receptors: Allosteric Modulation by Neurosteroids

Neuroactive steroids modulate GABAA receptor function in many brain regions. Naturally occurring steroid metabolites form locally in the brain: 5a-reductase transforms progesterone to 5a-DPH, which in turn is reduced by 3a-hydroxysteroid oxidoreductase to allopregnanalone. Allopregnanalone potently activates

GABAA receptors. No absolute specificity of neurosteroids for particular GABAA receptor subunit combinations exists. Many GABAA receptors are sensitive to the steroid THDOC, but receptors with the d subunit are particularly sensitive. Thus, endogenous allopregnanolone may act on extrasynaptic a4bd and a6bd GABAA receptors to increase basal levels of inhibition. New Sleep-Promoting Ligands Acting as GABA Mimetics on Extrasynaptic a4bd Receptors

Modulating extrasynaptic GABAA receptors can produce profound effects on the nervous system (Figure 6). Many people suffer from insomnia. This is often treated with ligands, such as zolpidem, with a preferential affinity for the BZ site on synaptic a1bg2 receptors. By contrast, Gaboxadol is a selective extrasynaptic GABAA receptor agonist that has its greatest efficacy at a4bd and a6bd GABAA receptors, that is, BZ-insensitive receptors that contribute to tonic inhibitory conductances rather than synaptic inhibitory postsynaptic currents. It is important to emphasize that Gaboxadol, unlike BZs or related ligands, is an agonist at the GABA binding site itself – that is, it is a GABA mimetic, opening the receptor directly. Drugs which promote sleep acting through the BZ biding site of a1bg2-type receptors can cause various side effects such as disturbances in memory, rebound insomnia on drug withdrawal, and drug dependence. Gaboxadol acts on a4bd occurring exclusively extrasynaptically and enriched in particular neuronal pathways (e.g., thalamus – Figures 4(a) and 6), and thus has a different side effect profile. Sleep induced by increasing specifically the activity of extrasynaptic GABAA receptors may be of higher quality.

Conclusions Many subtypes of GABAA receptor are used by the nervous system to impart inhibition. These GABAA receptors differ in their affinity for neurotransmitter and allosteric modulators, activation rate, desensitization rate, channel conductance, and location on the cell. But exactly what selective advantages are conferred by so many receptor subtypes to brain function is unclear and remains a fascinating research issue. In particular, the large number of abg subunit-containing synaptic receptor subtypes, often co-expressed on a single cell type such as a cortical pyramidal cell (Figures 2 and 5), might enhance the computational potential of the cell. Receptors at different synapses might have properties (e.g., deactivation rates) fine-tuned for individual inputs. In principle, evolutionary processes selected these subunit combinations

354 GABAA Receptors: Molecular Biology, Cell Biology, and Pharmacology

to be optimal, or at least to provide a reasonable compromise, for the inhibitory function required at each type of location. In some cases, however, multiple genes may exist because of the need to have complex transcriptional regulation that would have been difficult to organize from one gene promoter, rather than from the need to have different receptor properties. If one could engineer it, the hippocampus, or even the entire brain, might work perfectly well with just one type of GABAA receptor, such as an a1b2g2 combination. In practice, the GABAA receptor system remains an important target for new therapeutic approaches. See also: GABA Synthesis and Metabolism; GABAA Receptor Synaptic Functions; GABAB Receptor Function; GABAB Receptors: Molecular Biology and Pharmacology.

Further Reading Belelli D and Lambert JJ (2005) Neurosteroids: Endogenous regulators of the GABAA receptor. Nature Reviews Neuroscience 6: 565–575. Buzsaki G (2006) Cycle 3: Diversity of cortical functions is provided by inhibition. In: Buzaki G (ed.) Rhythms of the Brain, pp. 61–79. New York: Oxford University Press. Darlison MG, Pahal I, and Thode C (2005) Consequences of the evolution of the GABAA receptor gene family. Cellular and Molecular Neurobiology 25: 607–624. Ernst M, Bruckner S, Boresch S, and Sieghart W (2005) Comparative models of GABAA receptor extracellular and transmembrane domains: Important insights in pharmacology and function. Molecular Pharmacology 68: 1291–1300. Fang C, Dent L, Keller CA, et al. (2006) GODZ-mediated palmitoylation of GABAA receptors is required for normal assembly and function of GABAergic inhibitory synapses. Journal of Neuroscience 26: 12758–12768. Farrant M and Nusser Z (2005) Variations on an inhibitory theme: Phasic and tonic activation of GABAA receptors. Nature Reviews Neuroscience 6: 215–229. Fritschy JM and Brunig I (2003) Formation and plasticity of GABAergic synapses: Physiological mechanisms and

pathophysiological implications. Pharmacology & Therapeutics 98: 299–323. Hammond C (2001) The Ionotropic GABAA receptor. In: Hammond C (ed.) Cellular and Molecular Neurobiology, 2nd edn., pp. 227–250. San Diego, CA: Academic Press. Korpi ER and Sinkkonen ST (2006) GABAA receptor subtypes as targets for neuropsychiatric drug development. Pharmacology & Therapeutics 109: 12–32. Olsen RW and Betz H (2006) GABA and glycine. In: Siegel GJ, Albers RW, Brady ST, and Price DL (eds.) Basic Neurochemistry: Molecular, Cellular and Medical Aspects, 7th edn., pp. 291–301. Amsterdam: Academic Press. Rudloph U and Mohler H (2004) Analysis of GABAA receptor function and dissection of the pharmacology of benzodiazepines and general anesthetics through mouse genetics. Annual Review of Pharmacology and Toxicology 44: 475–498. Schofield PR, Darlison MG, Fujita N, et al. (1987) Sequence and functional expression of the GABAA receptor shows a ligand-gated receptor super-family. Nature 328: 221–227. Seeburg PH, Wisden W, Verdoorn TA, et al. (1990) The GABAA receptor family: Molecular and functional diversity. Cold Spring Harbor Symposia on Quantitative Biology 55: 29–40. Siegel E, Baur R, Boulineau N, and Minier F (2006) Impact of subunit positioning on GABAA receptor function. Biochemical Society Transactions 34: 868–871. Simon J, Wakimoto H, Fujita N, Lalande M, and Barnard EA (2004) Analysis of the set of GABAA receptor genes in the human genome. Journal of Biological Chemistry 279: 41422–41435. Somogyi P and Klausberger T (2005) Defined types of cortical interneurone structure space and spike timing in the hippocampus. Journal of Physiology 562: 9–26. Unwin N (2005) Refined structure of the nicotinic acetylcholine receptor at 4A resolution. Journal of Molecular Biology 346: 967–989. Wafford KA and Ebert B (2006) Gaboxadol: A new awakening in sleep. Current Opinion in Pharmacology 6: 30–36. Whiting PJ (2006) GABAA receptors: A viable target for novel anxiolytics? Current Opinion in Pharmacology 6: 24–29.

Relevant Websites http://www.ionchannels.org – Ionchannels.org. http://www.ebi.ac.uk – Ligand-Gated Ion Channel Database. http://www.nature.com – Nature: Supplement: Insight: Ion Channels. http://en.wikipedia.org – Wikipedia articles on benzodiazepine and GABA.

GABAA Receptor Synaptic Functions I Mody, David Geffen School of Medicine at UCLA, Los Angeles, CA, USA ã 2009 Elsevier Ltd. All rights reserved.

Introduction The amino acid neurotransmitter g-aminobutyric acid (GABA) is responsible for mediating most of chemical inhibition in the central nervous system (CNS). The ionotropic GABA receptors (GABAARs) are members of the cysteine-loop ligand-gated ion channel family and form a Cl - and HCO3 -permeable ion pore assembled from five (heteropentameric) subunits selected from the following subunits: a1–6, b1–3, g1–3, d, E, y1–3, p, and r1–3. Hundreds of thousands of different combinations are possible, yet no more than a few dozen receptor combinations are likely to exist in the mammalian brain. This is thought to result in part from precise rules of subunit assembly (e.g., two a and two b subunits assemble with either one g or one d subunit) and in part from certain rules of specific subunit partnerships that also appear to define their subcellular localization inside and outside of synapses. Loss-of-function mutations in some of these subunits are associated with various rare genetic forms of epilepsy in humans, indicating their importance in subduing neuronal hyperactivity. Over the years GABAARs have become synonymous with inhibition, but we now know that depending on a variety of conditions, their activation may excite neurons or may synchronize them, resulting in physiological oscillations or pathological synchrony leading to seizures. In 1963, on behalf of the Nobel Committee upon awarding the Prize to Sir John C Eccles, Ragnar Granit stated: ‘‘If the arriving impulse is connected to excitatory synapses the response of the cell is yes, i.e., excitability increases; vice versa the inhibitory synapses make the cell respond with a no, a diminution of excitability.’’ Forty-three years later it is clear that in response to GABAergic ‘inhibition’ many cells have expanded their vocabulary to ‘maybe’ and even to ‘yes.’

GABAARs at Synapses In the mammalian brain GABA synapses constitute a minority. Only about every sixth synaptic terminal is GABAergic and only every fifth neuron utilizes GABA as a transmitter. Excitation clearly outnumbers inhibition. Yet, in spite of this overpowering excitatory drive, most of the time the GABAergic

system manages to keep excitability in check. It does this through a spatially and temporally selective activation of the GABAergic system. The pioneering Golgi impregnation studies of Ramo´n y Cajal, Lorente de No, and Ja´nos Szenta´gothai have distinguished a marked feature of cortical interneurons: they have a highly selective axonal arborization, innervating distinct spatial domains of the principal cells. GABA synapses can be found on cell bodies, dendritic shafts and branch points, spine necks, and axon initial segments, in far more diverse places than excitatory synapses. The strategic positioning of GABA synapses on the target cells can annihilate a specific input or the entire output of the cell. A good example of such selective control of a given input is the hippocampal oriens-lacunosum moleculare (OLM) cell that precisely innervates the most distal dendrites of pyramidal cells. This is precisely the region where the entorhinal cortical excitatory input lands on the principal cells. Since OLM cells are activated by massive intrahippocampal excitatory events, this spatial arrangement of their inhibitory output ensures that the entorhinal input does not have a say during this time. In contrast, the cortical chandelier or axo-axonic cells do not synapse onto a cellular region receiving a specific input. Instead, they control the output of pyramidal cells, by innervating the axon initial segment, where the ultimate action potential output of the target neuron is generated. According to the latest findings however, some of these axo-axonic inhibitory interneurons may elicit firing at the axon initial segment (see later), in stark contradiction to the function implied by their ‘inhibitory’ name. The location specificity of inhibitory synapses is complemented by a highly specific assembly of the GABAARs that depends on the type of inhibitory interneuron that forms the synapse. There is little known about the mechanisms that sustain this specificity at a given synapse over the lifetime of the neuron. Of the thousands of possible heteropentameric GABAAR combinations that can be formed through assembly from nearly 20 subunits, only a few dozen combinations occur in the brain. These combinations do not appear to be randomly inserted at synapses. The types of subunits present at a given synapse tend to be specified by the GABA neuron that provides the presynaptic bouton of the synapse. The axo-axonic chandelier cells, and the pyramidal cell soma innervating cholecystokinin (CCK)-positive basket cells, are known to form synapses on predominantly a2 subunit-containing receptors. The parvalbumincontaining basket cells synapse onto a1-containing

355

356 GABAA Receptor Synaptic Functions

GABAARs, while other combinations are found at synapses made by other interneuron types. The functional significance of this highly specific arrangement is only beginning to emerge. Different GABA receptor subunit assemblies have very different kinetic and desensitization properties when studied in expression systems. Accordingly, one would expect the various synaptic currents generated by diverse GABAARs also to be highly variable. However, paired recordings from two interneurons and their target cells usually show similar inhibitory postsynaptic current (IPSC) kinetics, in spite of known differences of GABAAR assemblies at the two synapses. Even if the kinetic differences observed in expression systems do not hold at GABAARs assembled at synapses in the brain, the different compositions and locations of GABAARs have undoubtedly evolved to participate in widely different control functions of neuronal excitability. Insight into these functions has come from cleverly engineered mice expressing various a subunits insensitive to benzodiazepines (BZs), a class of potent allosteric GABAAR modulators. Two a, two b, and a single g subunit assemble to form the most commonly found GABAARs in the brain. The function of these receptors is enhanced by BZs as long as the a subunits are either a1, a2, a3, or a5, but not a4 or a6. There is a single amino acid in a critical position (His) that differs between the BZsensitive and BZ-insensitive a subunits (Arg in the latter). Mice have been genetically engineered to have BZ-insensitive a1, a2, a3, or a5 subunit-containing GABAARs. Studies in these mice conclusively demonstrated that the a1 subunit-containing GABAARs are involved in the sedative–hypnotic actions of BZs, while the receptors containing a2 subunits mediate anxiolytic effects of BZs. Does this mean that the function of the synapses predominantly containing a2 subunits, such as those found at the axon initial segment or those apposed to the terminals of CCK-containing basket cells, are mainly related to the control of anxiety? At this time it is still too early to say, as there might be specific brain regions rich in a2 subunits that could play a significant role in the control of anxiety. Nevertheless, it is clear that specific GABAAR subunit assemblies innervated by specific interneurons play highly specialized roles in the development of neuronal circuitry and in mediating specific higher nervous system functions.

Activation of GABAARs at Synapses – Phasic Inhibition The GABAARs usually found at synapses are most likely to contain g2 subunits. GABA synapses are ‘symmetrical’ synapses; called such because on electron

micrographs they are devoid of a wide electron-dense postsynaptic density (PSD) which would make the appearance of the synapse ‘asymmetric.’ Therefore, compared to the PSDs of excitatory glutamatergic synapses containing a stable assembly of over 70 different receptor, scaffold, and signaling proteins, there is a lackluster showing of postsynaptic proteins at the poorly developed PSDs of GABA synapses. Glutamate receptors at synapses associate with the subsynaptic cytoskeleton. GABAARs do not. There are, however, some proteins that seem to play a role in the anchoring, insertion, and removal of GABAARs at synapses. Gephyrin links up with the subsynaptic microtubule and actin cytoskeleton. It interacts with several microfilament-regulating proteins, including collybistin, profilins, Mena/VASP proteins, and dynein light chain (Dlc)/myosin-Va. However, the role of these protein– protein interactions for the shaping of GABA synapses is largely a mystery. Some GABA synapses express the dystrophin–glycoprotein complex, better known as a connecting element between the extracellular and cytoskeletal matrices of muscle cells. Other proteins are implicated in the trafficking of GABAARs: AP2, BIG2, GABARAP, GODZ, Plic-1, and PRIP1/2. To make things more complicated, trafficking of GABAAR subunits seems to depend on the phosphorylation state of the subunit. Various kinases and phosphatases are at hand, as they bind directly to the subunits or are linked to specific GABAARs through highly specific adaptor proteins. As molecular biologists, biochemists, and anatomists continue to pursue the means of GABAAR anchoring and targeting to synapses, physiologists and biophysicist have focused on the means of activation of GABAARs at synaptic junctions. The kinetic profile of synaptic GABAARs consists of a moderate to low affinity for GABA, some desensitization, and a rapid activation/deactivation. The time course of GABA in the synaptic cleft has been estimated to be less than 500 ms, and the peak concentration reaching the receptors is probably in the low millimolar range. This means that at many synapses where usually no more than tens of GABAARs are present, the number of GABA molecules is in excess of the available binding sites, thus providing for a high occupancy of the available receptors. However, this may not be the case at all GABA synapses, as in cerebellar basket/stellate cells, where synapses with hundreds of receptors and synapses with tens of receptors coexist, only the latter appear to be saturated by the released transmitter. In general, the relatively low number of postsynaptic GABA receptors compared to the amount of GABA released into the cleft renders the process regulating the number of receptors an important aspect of GABAAR plasticity. The kinetic properties of the

GABAA Receptor Synaptic Functions 357

subunits assembled at the synapses should also affect the properties of inhibition, but thus far there are consistent reports only on the speedy decay of a1 subunit-containing synaptic receptors, while the variety of the biophysical receptor properties seen in expression systems has yet to be demonstrated at synapses. Desensitization is also an important mechanism to consider for regulating the decay of IPSCs, but prolonging the synaptic GABA transient by using uptake blockers does not always prolong the duration of synaptic events as it would be expected from a significant role of desensitization. It appears, however that the duration of the GABA transient in the cleft may itself regulate the decay of synaptic events, which in turn might make it difficult to discern differences that are simply due to underlying channel kinetics. The various brain-state-altering effects of BZs that act to prolong the duration of IPSCs predominantly affecting g2 subunit-containing synaptic receptors show the importance of regulating the decay phase of IPSCs. Like at any central synapse, presynaptic mechanisms can also effectively determine how GABA synapses are activated. Again, the most interesting presynaptic aspect of GABA synapses is their specificity with regard to the cell type supplying the presynaptic terminal. For example, there are clear differences in the way GABA is released from parvalbumin- and CCK-containing basket cells, and this may have important functional consequences for the operation of neuronal networks. Other GABA synapses are depressed most of the time by ambient levels of endocannabinoids, and do not become active unless the presynaptic cell fires at high frequencies. Different presynaptic Ca2þ channels are responsible for activating GABA release in different interneurons. Moreover, there is also accumulating evidence that the intricate vesicular release machinery may be different between glutamate- and GABA-containing terminals.

Activation of Peri- and Extrasynaptic GABA Receptors: Tonic Inhibition As is the case for many ligand-gated ion channels, synapses are not the only place GABAARs are likely to show up on a cell’s surface. But perhaps for more than any other ligand-gated channel, GABAARs found outside the synapses are quite special. Although many different combinations of GABAARs are scattered outside of synapses, there are only two types of combinations that have been shown to actually respond to the ambient GABA concentrations of less than a few micromoles per liter, as found in the interstitial space of neurons. These two combinations

are the d subunit-containing and the a5 subunitcontaining GABAARs. The d subunits combine with a4, a6, and most likely other a subunits to form, together with b2 or b3 subunits, a GABAAR combination that is well suited to be activated by low levels of ambient GABA. These receptors have a high affinity for GABA, they do not desensitize, and are preferentially found extra- or perisynaptically. Their modulation mainly occurs by increasing the efficacy of GABA rather than its potency, as the natural agonist is not very efficacious at these receptors. The other GABAAR combination shown to be activated by ambient GABA is the a5 subunit-containing receptors. The a5 subunits are not exclusively extrasynaptic; they probably combine with other a subunits in synaptic receptors, but in the cells where they are present they readily mediate tonic inhibition. The tonic inhibition, when present, is a significant force in controlling the excitability of the neurons. Charge movement through tonically active GABA channels outweighs 3:1 to 5:1 the charge going through the synaptic receptors, even when the frequency of the synaptic activity is quite high (40– 60 Hz). Such a tonically active conductance globally is well suited to regulate the input–output function of the cells. The d subunit-containing GABAARs mediating tonic inhibition are also an important site of action for various endogenous and exogenous allosteric modulators. Neurosteroids, brain-synthesized metabolites of various sex and stress steroid hormones, specifically act upon these receptors to enhance their activity. Ethanol, at concentrations relevant to impairing sobriety, also appears to target the d subunitcontaining GABAARs, which may justify them lately being referred to as the ‘ethanol receptor.’ The receptors mediating tonic inhibition show a remarkable plasticity, both during physiological (ovarian cycle) and pathological (epilepsy) alterations in neuronal excitability. Selectively eliminating tonic inhibition in hippocampal CA1 pyramidal cells facilitates certain forms of learning, presumably by increasing the sensitivity of pyramidal cells to incoming stimuli. These recent developments in the physiology and pharmacology of extrasynaptic tonically active GABAARs will lead to exciting insights into their complex roles in the control of neuronal excitability.

Excitation and Synchrony through ‘Inhibitory’ Synapses There is nothing mysterious about ‘inhibitory’ synapses extending their vocabulary to ‘maybe’ and ‘yes.’ The relationship between the time-averaged membrane potential of the neuron and the GABA

358 GABAA Receptor Synaptic Functions

equilibrium potential, which is a function of the equilibrium potentials of the two permeant ions, Cl and HCO3, determines whether activation of GABAARs will depolarize or hyperpolarize a cell. In fact, ‘yes’ is the first word uttered by an activated GABAAR. Early on during brain development, when cells are loaded with Cl , the activation of GABAAR channels leads to efflux of negatively charged Cl ions, thus depolarizing the cell. This makes GABA one of the first depolarizing signaling molecules during development. The depolarization helps bring the membrane potential to the threshold of activation of voltagegated Ca2þ entry, a cation that, once inside the cells, is critical for the activation of a myriad of important developmental signaling cascades. Eventually, later during development, Cl is extruded from the cells by various transport mechanisms. Inhibition can thus start to learn to say ‘no.’ Yet, depolarizing GABAARmediated events continue to play an important role in the adult nervous system. The survival of newly generated neurons in the mammalian brain depends largely on the activation of depolarizing, tonically active GABAARs. Local peaks in the GABA reversal potential can trigger depolarizing inhibitory postsynaptic potentials (IPSPs) that can passively propagate along the neuron to elevate excitability, or can outright trigger action potentials at the axon initial segment, as we have recently learned from the potentially ‘excitatory’ function of axo-axonic cells in the neocortex. Regulated by several exchangers and transporters, the GABA reversal potential can itself become a dynamic force in the control of neuronal excitability. One of the key players is a Kþ/Cl cotransporter called KCC2. This molecule is critical in extruding Cl during development, but reducing its function during spike timing-dependent plasticity of GABA synapses, high-frequency firing of the cells, or neuronal injury leads to an intracellular accumulation of Cl and thus the conversion of a hyperpolarizing event into one that will depolarize the affected neuron. An altered KCC2 function is most likely at work to cause the pathological alteration of the GABA reversal potential in subicular pyramidal neurons recorded in tissue samples obtained from temporal lobe epilepsy patients, and causes the conversion to a GABA-induced depolarization of spinal cord neurons during neuropathic pain. Even if depolarizing GABA-mediated events do not push the membrane potential all the way past the threshold of activation of voltagedependent Naþ or Ca2þ conductances, such events can passively propagate through the cell, thus increasing excitability and promoting neuronal synchrony.

Depolarizing neurons is not the only means by which activation of inhibitory synapses can promote synchrony. If the neurons are equipped with a hyperpolarization-activated excitatory conductance (e.g., Ih) or a low-threshold (T-type) Ca2þ current, as most neurons are, then a hyperpolarization induced in several neurons at the same time can lead to a rebound excitation following the cessation of the hyperpolarizing event. Hundreds or thousands of principal cells innervated by a single inhibitory interneuron, or by a few interneurons activated in unison, can thus synchronize their firing without requiring any synchronous excitatory drive. This type of synchronization may specifically spring into action when diffusely acting neurotransmitters, such as acetylcholine (ACh) or glutamate, enhance GABA release by acting at once on the terminals of several inhibitory cells through actions at nicotinic ACh or N-methyl-D-aspartate (NMDA) receptors. In summary, phasic and tonic GABAergic inhibitions exert a substantial control over neuronal excitability. The incredibly diverse GABAergic cells of the brain are the real puppet masters of the principal cells, by controlling their activity both at the level of different single neuronal compartments and at the level of intricately connected neuronal networks. For certain, the diversity and specificity of GABAARs, GABAergic neurons, and their functions will keep neuroscientists busy for a while. It may have taken some time, but now we know that it is much easier for the ‘inhibitory’ GABAergic system to say ‘maybe’ and ‘yes’ than it is for excitatory glutamate synapses to learn to say ‘no.’ See also: GABA Synthesis and Metabolism; GABAA

Receptors: Molecular Biology, Cell Biology and Pharmacology; GABAB Receptor Function; GABAB Receptors: Molecular Biology and Pharmacology.

Further Reading Ben-Ari Y (2002) Excitatory actions of GABA during development: The nature of the nurture. Nature Reviews in Neuroscience 3: 728–739. Buzsa´ki G (2006) Rhythms of the Brain. New York: Oxford University Press. Buzsa´ki G, Geisler C, Henze DA, et al. (2004) Interneuron diversity series: Circuit complexity and axon wiring economy of cortical interneurons. Trends in Neuroscience 27: 186–193. Chen ZW and Olsen RW (2007) GABA receptor associated proteins: A key factor regulating GABA receptor function. Journal of Neurochemistry 100(2): 279–294. Cobb SR, Buhl EH, Halasy K, et al. (1995) Synchronization of neuronal activity in hippocampus by individual GABAergic interneurons. Nature 378: 75–78.

GABAA Receptor Synaptic Functions 359 Cohen I, Navarro V, Clemenceau S, et al. (2002) On the origin of interictal activity in human temporal lobe epilepsy in vitro. Science 298: 1418–1421. Coull JA, Beggs S, Boudreau D, et al. (2005) BDNF from microglia causes the shift in neuronal anion gradient underlying neuropathic pain. Nature 438: 1017–1021. Farrant M and Nusser Z (2005) Variations on an inhibitory theme: Phasic and tonic activation of GABAA receptors. Nature Reviews Neuroscience 6: 215–229. Fiumelli H, Cancedda L, and Poo MM (2005) Modulation of GABAergic transmission by activity via postsynaptic Ca2þdependent regulation of KCC2 function. Neuron 48: 773–786. Freund TF and Buzsa´ki G (1996) Interneurons of the hippocampus. Hippocampus 6: 347–470. Gaiarsa JL, Caillard O, and Ben-Ari Y (2002) Long-term plasticity at GABAergic and glycinergic synapses: Mechanisms and functional significance. Trends in Neuroscience 25: 564–570. Hanchar HJ, Wallner M, and Olsen RW (2004) Alcohol effects on gamma-aminobutyric acid type A receptors: Are extrasynaptic receptors the answer? Life Sciences 76: 1–8. Hefft S and Jonas P (2005) Asynchronous GABA release generates long-lasting inhibition at a hippocampal interneuron–principal neuron synapse. Nature Neuroscience 8: 1319–1328. Hevers W and Lu¨ddens H (1998) The diversity of GABAA receptors. Pharmacological and electrophysiological properties of GABAA channel subtypes. Molecular Neurobiology 18: 35–86. Lien CC, Mu Y, Vargas-Caballero M, et al. (2006) Visual stimuliinduced LTD of GABAergic synapses mediated by presynaptic NMDA receptors. Nature Neuroscience 9: 372–380. Lu¨scher B and Fritschy JM (2001) Subcellular localization and regulation of GABAA receptors and associated proteins. International Review of Neurobiology 48: 31–64. Lu¨scher B and Keller CA (2004) Regulation of GABAA receptor trafficking, channel activity, and functional plasticity of inhibitory synapses. Pharmacology and Therapeutics 102: 195–221. Macdonald RL, Gallagher MJ, Feng HJ, et al. (2004) GABAA receptor epilepsy mutations. Biochemical Pharmacology 68: 1497–1506. Maguire JL, Stell BM, Rafizadeh M, et al. (2005) Ovarian cyclelinked changes in GABAA receptors mediating tonic inhibition alter seizure susceptibility and anxiety. Nature Neuroscience 8: 797–804.

Mody I and Pearce RA (2004) Diversity of inhibitory neurotransmission through GABA(A) receptors. Trends in Neuroscience 27(9): 569–575. Mozrzymas JW (2004) Dynamism of GABAA receptor activation shapes the ‘personality’ of inhibitory synapses. Neuropharmacology 47: 945–960. Rivera C, Voipio J, Thomas-Crusells J, et al. (2004) Mechanism of activity-dependent downregulation of the neuron-specific K–Cl cotransporter KCC2. Journal of Neuroscience 24: 4683–4691. Rudolph U and Mo¨hler H (2004) Analysis of GABAA receptor function and dissection of the pharmacology of benzodiazepines and general anesthetics through mouse genetics. Annual Reviews of Pharmacology and Toxicology 44: 475–498. Sieghart W and Sperk G (2002) Subunit composition, distribution and function of GABAA receptor subtypes. Current Topics in Medicinal Chemistry 2: 795–816. Somogyi P and Klausberger T (2005) Defined types of cortical interneurone structure space and spike timing in the hippocampus. Journal of Physiology 562(1): 9–26. Somogyi P, Tamas G, Lujan R, et al. (1998) Salient features of synaptic organisation in the cerebral cortex. Brain Research: Brain Research Reviews 26: 113–135. Soltesz I (2005) Diversity in the Neuronal Machine: Order and Variability in Interneuronal Microcircuits. New York: Oxford University Press. Stell BM, Brickley SG, Tang CY, et al. (2003) Neuroactive steroids reduce neuronal excitability by selectively enhancing tonic inhibition mediated by delta subunit-containing GABAA receptors. Proceeding of the National Academy of Sciences United States of America 100: 14439–14444. Szabadics J, Varga C, Molnar G, et al. (2006) Excitatory effect of GABAergic axo-axonic cells in cortical microcircuits. Science 311: 233–235. Watts J and Thomson AM (2005) Excitatory and inhibitory connections show selectivity in the neocortex. Journal of Physiology 562: 89–97. Whiting PJ (2003) The GABAA receptor gene family: New opportunities for drug development. Current Opinion in Drug Discovery and Development 6: 648–657.

GABAB Receptors: Molecular Biology and Pharmacology N G Bowery, GlaxoSmithKline, Verona, Italy ã 2009 Elsevier Ltd. All rights reserved.

Introduction g-Aminobutyric acid (GABA) is the single most important inhibitory neurotransmitter within the mammalian brain, in which it is estimated that 40% of all inhibitory synaptic activity is mediated through this neutral amino acid. GABA can activate both ionotropic (GABAA) and metabotropic (GABAB) receptors, and these are present inside and outside the brain of mammals and also evident in lower species. This article focuses on the metabotropic receptor, GABAB. The receptor was originally defined in the 1980s by pharmacological characterization in isolated tissue preparations. The structure of the receptor was identified approximately 18 years later. Unlike GABAA receptors, GABAB sites do not affect Cl membrane conductance in neurons but instead their activation increases Kþ conductance and decreases Ca2þ conductance. These effects tend, although not exclusively, to be site directed such that the decrease in Ca2þ is more associated with presynaptic receptors, whereas Kþ effects are predominantly postsynaptic. As a consequence of these effects, receptor activation can produce neuronal hyperpolarization (Kþ effect) or a decrease in evoked neurotransmitter release (Ca2þ effect). It is this latter effect that was the basis of the first observed characteristic of GABAB receptors. GABA dose dependently reduced the electrically evoked release of tritiated noradrenaline from rat isolated atria. This effect was neither blocked by recognized GABA antagonists nor mimicked by established GABA agonists. Conversely, baclofen (b-chlorophenyl GABA), which is inactive at chloride-dependent GABA receptors, was able to mimic the action of GABA in the atrial preparation and in other GABAB receptor systems. The subsequent development of a membrane receptor binding assay firmly established the existence of the metabotropic GABA receptor in mammalian brain tissue.

activating the G-protein-coupled signaling system. GABAB1 and GABAB2 must remain linked as a dimer after insertion into the cell membrane to fully maintain receptor function. GABAB1 subunits remain associated with the endoplasmic reticulum, and only when they link with GABAB2 is the dimer transported to the cell membrane. Each subunit has a seven transmembrane spanning domain and they are linked via their C-termini through coiled-coil structures. A number of splice variants of the GABAB1 subunit have been identified, namely GABAB1a–1g (Figure 1, inset), but not all have been shown to have a physiological function. Only GABAB1a and GABAB1b have been implicated in the functional receptor, and although GABAB1c has been reported to be functional in vitro, evidence for an in vivo role is equivocal. The 1a and 1b proteins differ significantly in their amino acid sequences within the N-terminal; 1a has a unique sequence of 162 amino acids, whereas 1b has a shorter unique sequence of 47 amino acids. The distribution patterns of these two subunits within the rat brain indicate that they have different roles, although these have yet to be established. For example, within the spinal cord, the level of GABAB1a is reported to be tenfold higher than that of GABAB1b within the primary afferent terminals, and the GABAB1b subunit appears to predominate at postsynaptic sites in the cerebellum. However, elsewhere in the brain, GABAB1b appears to predominate in presynaptic terminals, whereas GABAB1a is associated with postsynaptic sites, for example, within the thalamocortical system. There is no unequivocal evidence to support the existence of variants of the GABAB2 subunit, although two variants of the human GABAB2 subunit have been proposed. However, these were probably artifacts of the isolation process. The expression of GABAB1 and that of GABAB2 appear to be independent of each, even though there is a 1:1 stoichiometric relation between the two proteins. Invariably, GABAB2 is expressed more abundantly than GABAB1 under experimental pathological conditions.

Receptor Structure

Protein–Protein Interactions

GABAB receptors exist naturally as heterodimers and comprise subunits designated GABAB1 and GABAB2 (Figure 1). Although the subunits exhibit 35% homology, they have quite distinct characteristics. Whereas GABAB1 contains the binding domain for GABA in its extracellular N-terminal, the GABAB2 subunit appears to be responsible for engaging and

The possibility of a mismatch between GABAB1 and GABAB2 levels raises the possibility that other proteins might form a dimer with the GABAB subunit to form a functional receptor. Numerous proteins have been described which interact with either GABAB1 or GABAB2, and some of these have such high affinity that they are able to compete with one GABAB

360

GABAB Receptors: Molecular Biology and Pharmacology 361 GABAB1

GABAB2

GABA binding domain

Allosteric modulator site

N

N

Extracellular

Ca2+

g

g

7 6 5

3 4

2 1 a i/o

b g

cAMP +

Human 1c Rat/human 1a [62 aa deletion] [147 aa]

4 3

a i/o

g

K+

1 2

5 6 7

b

Intracellular

Rat/human 1b [18 aa]

3 4

b

AC

1 2 b

−

C

−

C

Rat 1c [31 aa addition]

5 6 7

Rat 1d [25 aa modification]

Figure 1 Diagram of the heterodimer that forms the GABAB receptor. The receptor comprises two subunits, GABAB1 and GABAB2, which are inserted into the plasma membrane after having been coupled intracellularly. GABAB1 contains the GABA binding domain in its extracellular N-terminal chain, whereas GABAB2 does not bind GABA but has a modulatory site within the transmembrane region which appears to be specific to this subunit. This subunit also seems to be responsible for coupling to the second messenger G-protein system. Receptor activation manifests as an increase in Kþ conductance, a decrease in Ca2þ conductance, and/or an inhibition of adenylate cyclase (AC). The inset indicates the primary isoforms of GABAB1. Courtesy of Stefano Tacconi and Rachel Ginham.

subunit for binding to the other subunit. However, no evidence of the formation of a functional receptor from any protein–protein interaction has been reported. Thus, if the interaction is not producing a receptor, what purpose(s) does this protein coupling serve? It might be to provide a directional mechanism to enable the correct site insertion of the dimer into the plasma membrane or merely to act as a scaffold or anchoring device to support the dimer when inserted. Alternatively, because of the high affinity of these proteins, they might regulate the formation of the dimer, limiting the level of functional receptor, or by interacting intracellularly with the formed dimer, they may reduce its cell surface expression. The exact role of these interacting proteins, including CREB2/ATF4, 14-3-3, fibulin-2, CHOP, and Marlin-1, has yet to be established.

Receptor Subtypes Despite the presence of structural variants of the GABAB1 subunits and the possibility of functional modifications of the receptor heterodimer by interacting proteins, there is still no unequivocal evidence for the existence of functional GABAB receptor subtypes. Pharmacological differences between autoreceptors and heteroreceptors as well as between the dual actions of GABAB agonists on adenylyl cyclase activity have been suggested by individual research

groups. However, both a lack of reproducibility and sufficient distinction between the effects and the failure to emulate observations made by other groups have cast doubt on the existence of functional receptor subtypes. In general, the available receptor agonists and antagonists do not distinguish between potential subtypes of the GABAB receptor. Nevertheless, it has been suggested that gabapentin is a GABAB agonist which selectively activates the form of the receptor comprising a combination of GABAB1a and GABAB2 and expressed in isolated Xenopus oocytes. However, this could not be confirmed by others, especially in in vivo systems. Thus, although this was an exciting and potentially important observation, the failure of others to reproduce the original findings has tempered enthusiasm not only for understanding the mechanism of action of this anticonvulsant/analgesic agent but also for the possibility of relating receptor structure to functional diversity. A further lack of support for receptor subtyping derives from studies performed on developmental knockout mice in which either GABAB1 or GABAB2 subunits have been deleted. No supporting evidence for any receptor subtyping was observed in these animals.

Receptor Distribution GABAB receptors are distributed throughout the mammalian system and are not confined to nervous tissue.

362 GABAB Receptors: Molecular Biology and Pharmacology

Although their function is probably of greatest physiological importance within the central nervous system, their presence in peripheral organs is well established. In fact, it was their presence in the heart atrium of the rat that led to their discovery. They are present on autonomic nerve terminals, and when activated by GABA, this decreases the evoked release of transmitter from the terminal. Of course, this is only of physiological importance when the receptor agonist is present, and in the periphery this is unlikely to occur at most sites. However, in the enteric nervous system, in which GABA-releasing neurons are present and impinge on autonomic cholinergic nerve fibers, GABAB receptors probably contribute to the control of intestinal movement and sphincters. Individual GABAB1 and GABAB2 subunits have been detected in peripheral tissue, but their distribution does not always appear to be coincident, for example, in the uterus and spleen. In general, there appears to be a paucity of GABAB2, suggesting that another protein may form a dimer with GABAB1 if the receptor is to function as described elsewhere. There is also a differential distribution of the GABAB1 isoforms such that GABAB1a is in, for example, the adrenals and prostate of the rat, whereas GABAB1b is the form present in the kidney. The distribution of GABAB sites in the mammalian brain is quite wide, although there are regional variations. Receptor autoradiography has shown that the highest densities of the receptor are in the thalamic nuclei, cerebellum (molecular layer), cerebral cortex, interpeduncular nucleus, and dorsal horn of the spinal cord. Moderate to low levels are present elsewhere, but this does not necessarily reflect the physiological importance of the receptor in those regions. For example, in the hippocampus the significance of GABAB receptors in neural transmission may not be matched by the overall density of receptors. The distribution patterns of the individual receptor subunits, as determined by immunocytochemistry, generally match each other, supporting the concept of heterodimer receptors. However, in the caudate putamen GABAB2 is not detectable even though GABAB1 and the native receptor are present, which could support the notion of another, as yet undefined, protein subunit.

Physiological Significance The physiological relevance of the GABAB receptor was first established within the hippocampus, where it was shown to be responsible for generating the late hyperpolarization associated with orthodromic synaptic transmission in Ammon’s horn. Subsequent studies provided evidence for the same pattern of activity within other brain regions, such as the lateral geniculate nucleus.

The mechanism underlying this neuronal hyperpolarization is an increase in membrane Kþ conductance mediated by activation of postsynaptic GABAB receptors. However, this is not the only mechanism through which GABAB activation occurs. The predominant location of the receptors appears to be presynaptic, where they function as autoreceptors and heteroreceptors to limit the release of a variety of neurotransmitters. A reduction in membrane Ca2þ conductance is produced by activation of GABAB sites, which reduces the cytosolic concentration of Ca2þ to inhibit transmitter release. Whereas the physiological significance of autoreceptors seems straightforward because of the local availability of neurotransmitter, the role of heteroreceptors on non-GABAergic terminals is not so obvious because evidence for axo-axonic contacts at nerve terminals is lacking. The only substantiated evidence is at primary afferent terminals in the spinal cord, where GABA interneurons innervate the terminal regions of the primary afferent fibers. However, despite the apparent lack of innervation, there is strong evidence for GABA acting in a ‘paracrine’ manner to activate GABAB heteroreceptors. GABA released from GABAergic terminals ‘washes over’ onto adjacent terminals where GABAB sites are present. The estimated concentration of GABA in the cleft, following synaptic release, is approximately 0.01 M, and because the affinity of GABA for heteroreceptors is approximately 0.000 000 01 M, sufficient GABA should be available to interact with sites in close proximity. Heteroreceptors are therefore of physiological as well as pharmacological importance.

Effector Mechanisms As indicated previously, more than one signaling system mediates the response to GABAB receptor activation. Either an increase in Kþ or a reduction in Ca2þ membrane conductance provides the neuronal response, and both of these events are mediated by G-proteins that are members of the pertussis toxinsensitive family Gia/Goa. However, presynaptically mediated events, which are generally associated with reduced Ca2þ conductance, appear to be less sensitive to pertussis toxin. The predominant calcium channel linked to GABAB sites appears to be the ‘N’ type, although ‘P’ and ‘Q’ type channels are also implicated. Multiple Kþ channel types seem to be associated with postsynaptic GABAB receptors. The third signaling system associated with GABAB sites is adenylyl cyclase, which is normally inhibited by receptor activation but if the enzyme has been activated by Gs-coupled receptor agonists such as isoprenaline, the b-adrenoceptor ligand, GABAB

GABAB Receptors: Molecular Biology and Pharmacology 363

receptor activation enhances the formation of cAMP above the level achieved with isoprenaline alone. This observation prompted a search for receptor subtypes by comparing the abilities of different GABAB receptor agonists to inhibit or enhance cAMP production. However, no clear separation has been established.

GABAB Receptor Ligands Agonists

One of the original criteria for establishing the presence of GABAB receptors was to show that b-[4-chlorophenyl] GABA (baclofen) was a stereospecific agonist. Subsequently, other agonists, such as 3-aminopropylphosphinic acid (3-APPA) and 3-aminopropyl-methylphosphinic acid (3-APMPA), emerged which were approximately tenfold more potent than R-( )-baclofen, the active isomer, at GABAB receptor binding sites (Figure 2). In vivo studies have also indicated that 3-APMPA has greater brain penetration than either baclofen or 3-APPA. The agonist CGP44532, which has an affinity comparable to that of ( ) baclofen for GABAB receptors (Figure 2), is reported to be selective for GABAB autoreceptors.

Structures and comparative activities of GABAB receptor ligands Comparative receptor binding in rat brain GABAB receptor agonists membranes (IC50s) R-(–)-baclofen

32 nM

Cl

O H2N

OH O

3-APPA H 2N

O

3-APMPA H2N OH

Receptor activation within the mammalian brain can produce a variety of effects as a consequence of

45 nM

O

H2N

P OH

GABAB receptor antagonists Phaclofen Cl

130 µM

O H2N

P OH OH

2-OH Saclofen

11 µM

Cl

O H2N

S OH O

HO CGP35348

Effects of GABAB Receptor Activation

16 nM

P OH

CGP 44532

27 µM

O

OEt P OH OEt

H2N

Antagonists

The first selective GABAB antagonists to be described were phaclofen, saclofen, and 2-hydroxy saclofen (Figure 2). These have only low affinity for the receptor, with pKi values of 4–5. However, phaclofen was used to great effect in obtaining the first evidence in 1988 for the physiological role of GABAB receptors in synaptic transmission within the rat hippocampus. Subsequent studies led to the introduction of a compound that was able to cross the blood–brain barrier (CGP 35348) and the first orally active agent (CGP 36742). However, both of these compounds have low affinities, as does the chemically distinct antagonist SCH50911 (Figure 2). A major breakthrough was obtained when 3,4-dichlorobenzyl or 3-carboxybenzyl substituents were attached to the existing molecules to produce a variety of compounds with affinities in the low nanomolar range, such as CGP 55845 and many others, which in all cases contain a phosphinic acid moiety. Ultimately, such antagonists, when radiolabeled with 125I, provided photoaffinity ligands suitable for the successful elucidation of the structure of the GABAB receptor.

5 nM

P H OH

38 µM

O

CGP36742 H2N

P OH O

SCH50911

3 µM

COOH

N H CGP 52432

Cl

55 nM

Cl

O

H N

OEt

P OH OEt

Cl

CGP 55845

6 nM

Cl H N

OH

O

Ph

P OH

Positive modulators tBu

CGP 7930

-

HO OH

But GS 39783

N

SMe N

N

-

NO2 N

Figure 2 Selection of GABAB receptor ligands and their comparative potencies at the native receptor in rat brain cerebral cortex membranes. Note the radical increase in potency of the examples of the dichlorobenzyl-substituted antagonists, CGP52432 and CGP55845.

364 GABAB Receptors: Molecular Biology and Pharmacology Table 1 GABAB receptor activation Effect

Site of action

Antinociception Antitussive action Drug addiction suppression Enhanced feeding Fat intake reduction Gastrin/gastric acid secretion altered Generation/exacerbation of absence epilepsy Inhibition of neurotransmitter release Inhibition of cognitive function Insulin/glucagon release Modulation of long-term potentiation Muscle relaxation Smooth muscle relaxation Smooth muscle contraction

Spinal cord, thalamus Cough center in medulla CNS–mesolimbic system Higher centers Higher centers Vagal center

Neuronal hyperpolarization Neutrophil chemotaxis enhanced Respiratory depression Suppression of CRH/MSH release Suppression of panic behavior Vasopressor action

Thalamus/somatosensory cortex CNS and peripheral nerve terminals Higher centers Pancreas CNS, hippocampus

allow for poor brain penetration. This problem has been addressed by the introduction of intrathecal administration using an indwelling pump inserted into the peritoneal cavity. Administration in this way, directly to the site of action within the spinal cord, requires only local concentrations that are too low to appear in the systemic circulation, thus avoiding the production of adverse effects. The reduced extracellular level of the agonist also decreases the possibility of receptor desensitization. The response to systemic administration of baclofen is reduced after chronic treatment, but this does not seem to be the case after chronic intrathecal infusion. Nociception

Spinal cord Lung, bladder, intestine Uterus, oviduct, gall bladder CNS (numerous locations) Leukocytes Brain stem Pituitary Dorsal periaqueductal gray Nucleus tractus solitarius

inhibition of transmitter release and/or postsynaptic neuronal hyperpolarization. Information derived from knockout mice, which exhibit hyperalgesia, seizures, hyperlocomotion, impaired learning, loss of responses to baclofen, and lack of GABAB binding sites throughout the brain, together with the actions of GABAB agonists and antagonists have provided the basis for defining the physiological role of the receptor as well as indicating the potential therapeutic benefits of receptor ligands. Some of the actions of baclofen (in vitro and in vivo) are shown in Table 1, and predominant among the in vivo effects are the muscle relaxant, antinociceptive, anti-drug-craving effects and reduction in cognitive behavior.

Potential Therapeutic Uses of GABAB Receptor Ligands Skeletal Muscle Relaxation

The centrally mediated muscle relaxant properties of baclofen make it the drug of choice for spasticity associated with cerebral palsy, multiple sclerosis, stiff-man syndrome, and tetanus. However, the side effects produced by this drug, which include seizures, nausea, drowsiness, dizziness, hypotension, muscle weakness, hallucinations, and mental confusion, are often poorly tolerated by patients. These are, in part, due to the need for high doses to be administered to

Baclofen and other GABAB receptor agonists have antinociceptive activity in acute pain models, such as the tail flick and hot plate tests in rodents. This occurs at doses below the threshold for muscle relaxation, enabling impairment of locomotor activity to be excluded as a confounding reason for the effect. This antinociceptive action stems, in part, from a reduction in the release of nociceptive transmitter from primary afferent fibers within the dorsal horn of the spinal cord. In addition, action within higher centers, particularly the thalamus, also contributes. In spinal cord slices from control rats, the application of GABAB receptor antagonists produces little or no increase in the evoked release of transmitter, but if the antagonist is applied to a cord slice from a monoarthritic rat, an evoked release of primary afferent transmitter occurs. If the antagonist is administered in vivo to rats with the same lesion, significant hyperalgesia occurs. This contrasts markedly with a lack of effect in control rats. These results suggest that an increase in GABAB innervation to primary afferent terminals occurs during chronic inflammation, and this acts as a pathological antinociceptive process to decrease the enhanced sensory input. The use of baclofen as an analgesic in humans has been very limited, presumably due, in part, to rapid tolerance and adverse effects following systemic administration. As mentioned previously, the spinal cord is not the only site where GABAB agonists exert their antinociceptive effect. Focal injections into the ventrobasal complex within the thalamus can suppress nociceptive processing in chronic inflammation. It has also been observed that the antinociceptive effect of the GABA uptake inhibitor tiagabine in rodents, which can be attributed to GABAB receptor activation, is associated with an increase in the extracellular concentration of GABA within the thalamus. In contrast to models of inflammatory pain, the induction of neuropathic pain in rodents does not

GABAB Receptors: Molecular Biology and Pharmacology 365

produce an increase in GABA levels within the spinal cord. However, baclofen can produce an antinociceptive effect in models of chronic neuropathy. Support for a role for GABAB receptors in pain mechanisms also derives from developmental knockout studies in mice. In these mice, functional GABAB receptors are not formed because the mice are deficient in either of the individual GABAB1 or GABAB2 subunits. In both forms of null mutant mice, hyperalgesia was exhibited in acute nociceptive tests, suggesting that functional heteromeric GABAB receptors are required to maintain pain thresholds. Cognitive Function

GABAB receptor agonists suppress cognitive behavior in animals, and this action is reversed by GABAB antagonists. Moreover, basal cognitive activity can even be enhanced by GABAB antagonists without prior suppression by an agonist. This raises the possibility that GABAB antagonists might provide a novel opportunity to treat cognitive impairment in humans, and this hypothesis is being tested in clinical trials. Evidence suggests that the site of action of GABAB antagonists in relation to cognition may be the hippocampus, where an increase in long-term potentiation (LTP) has been implicated but the nature of this modification appears to depend on the frequency of stimulation employed to produce LTP. Depression and Anxiety

A role for GABAB receptors in functional depression was first proposed more than 20 years ago. An upregulation in GABAB binding sites occurs in rat frontal cortex after chronic administration of a variety of antidepressant drugs, and although these findings were disputed, there is little doubt that GABAB mechanisms can be associated with depression. Antagonism of GABAB receptors produces a reversal of depressantlike behavior in animal models such as the rodent forced swim test and learned helplessness models. It has been observed that mice lacking GABAB1 or GABAB2 receptor subunits exhibit antidepressant-like behavior but they appear to be more anxious. Therefore, GABAB receptor activation appears to produce anxiolytic activity, whereas a loss or blockade of GABAB receptor function produces antidepressantlike effects. In support of this distinction, GABAB receptor activation in the dorsal periaqueductal gray of rats impairs one-way escape in the elevated T-maze test, which is consistent with an anxiolytic effect.

goal. A number of receptor systems appear to provide potential targets, including dopamine and glutamate receptors, but GABAB receptor activation may provide an additional approach. Baclofen was initially shown to reduce the reinforcing effects of cocaine in rats, but it soon became clear that other drugs of addiction, including nicotine, morphine-related agents, and ethanol, were also sensitive to GABAB agonists, whereas food reinforcement was unaffected. The finding that baclofen reduces craving for a variety of unrelated addictive substances, including heroin, alcohol, and nicotine, suggests that there may be an underlying common mechanism for the GABAB agonist in all cases. The reward center within the mesolimbic system, possibly the ventral tegmental area, would provide the focus for this action where control of the release/action of dopamine is implicated. Raising endogenous GABA levels within the mesolimbic system should have the same effect as administering a GABAB agonist. Thus, if vigabatrin, an inhibitor of GABA metabolism, or the GABA uptake inhibitor NO-711 are administered centrally in rats, they both attenuate heroin and cocaine selfadministration and prevent cocaine-induced increases in dopamine in this brain region. Clinical data indicate that baclofen is also effective against cocaine and alcohol craving in humans. However, the administration of baclofen can produce adverse effects as well as muscle relaxation, and these effects may detract from any potential benefits. Therefore, what might provide an alternative approach? The presence of a positive allosteric modulatory site on the GABAB2 subunit could provide the answer.

Allosteric Modulation of GABAB Receptors The location of the modulator allosteric site appears to be the heptahelical domain of the GABAB2 subunit. Although the GABAB1 subunit has the agonist binding domain, it does not have an allosteric site, whereas the converse appears to be true for GABAB2. Two compounds, CGP7930 and GS39783 (Figure 2), were originally reported to be positive modulators. Neither has any direct agonist activity, but both accentuate the effects of GABA and baclofen. These compounds have been examined in rat models of addiction and both modulators reduced the self-administration of cocaine.

Seizure Generation in Absence Epilepsy Drug Addiction

Therapeutic treatment of dependence on drugs of abuse is still inadequate and is therefore a major clinical

Absence seizures have a characteristic EEG waveform of a 3-Hz spike and wave which stem from discharges in the thalamic nuclei. It is believed that the thalamus

366 GABAB Receptors: Molecular Biology and Pharmacology

is the site from which these discharges originate, and although an intact thalamocortical network is necessary for generating spike and wave discharges, the origin of these discharges appears to lie outside the thalamus. The site of origin appears to be within the perioral region of the somatosensory cortex. This discharge spreads rapidly across the cortex and initiates a corticothalamic cascade. Injection of a GABAB agonist into the ventrobasal thalamus or reticular nucleus of a rat exacerbates this activity. By contrast, injection of a GABAB antagonist into the same regions suppresses the spike and wave discharges. If the GABAB antagonist is administered systemically, the same effect occurs. This might indicate that interference with the GABAergic innervation from the reticular nucleus disrupts the thalamocortical loop which generates the spike and wave activity. However, microinjection of the same selective antagonist into the somatosensory cortex can also suppress the seizure activity, indicating the involvement of GABABergic mechanisms also at this level. In patients with existing absence seizures, GABAB receptor agonists would likely enhance seizure activity, and any increase in GABA concentration in the vicinity of the thalamus would be expected to enhance seizure activity. Thus, for example, vigabatrin and tiagabine would be, and are, contraindicated in such individuals. The mechanism(s) underlying seizure exacerbation by GABAB agonists is unclear, but involvement of transient Ca2þ T currents seems possible. It has been suggested that g-hydroxybutyrate (GHB), which can produce absence-like seizures, is a weak GABAB receptor agonist. If GHB mimics the effect of the endogenous agonist, GABA, at GABAB receptors, then GABAB receptor antagonists would be expected to block the spike and wave discharges produced by GHB, which they do.

Conclusions The GABAB receptor system is of predominant, but not exclusive, importance within the central nervous system, and it may provide an important drug target in a variety of central nervous system disorders. It is a member of group 3 G-protein-coupled receptors and exists as a heterodimer with various isoforms comprising the GABAB1 subunit. However, lack of demonstrable functional subtyping of the receptor has limited exploitation of any pharmacological specificity. The only receptor agonist in current clinical use is baclofen, which has many limitations, including poor brain penetration. To date, little has emerged which is an improvement over baclofen. However, the discovery of allosteric modulators of the GABAB system may provide the possibility of producing

compounds that will readily gain access to the central nervous system while facilitating receptor function. See also: GABA Synthesis and Metabolism; GABAA Receptor Synaptic Functions; GABAA Receptors: Molecular Biology, Cell Biology and Pharmacology; GABAB Receptor Function.

Further Reading Bettler B, Kaupman K, Mosbacher J, and Gassmann M (2004) Molecular structure and physiological functions of GABAB receptors. Physiological Reviews 84: 835–867. Binet V, Brajon C, Le Corre L, Acher F, and Pin J-P (2004) The heptahelical domain of GABAB2 is activated directly by CGP7930, a positive allosteric modulator of the GABAB receptor. Journal of Biological Chemistry 279: 29085–29091. Bowery NG, Bettler B, Froestl W, et al. (2002) International Union of Pharmacology. XXXIII: Mammalian g-aminobutyric acidB receptors: Structure and function. Pharmacological Reviews 54(2): 247–264. Cryan JF and Kaupmann K (2005) Don’t worry ‘B’ happy! A role for GABAB receptors in anxiety and depression. Trends in Pharmacological Sciences 26(1): 36–43. Froestl W and Mickel SW (1997) Chemistry of GABAB modulators. In: Enna SJ and Bowery NG (eds.) The GABA Receptors, pp. 271–296. Totowa, NJ: Humana Press. Gassmann M, Shaban H, Vigot R, et al. (2004) Redistribution of GABAB(1) protein and atypical GABAB responses in GABAB(2)deficient mice. Journal of Neuroscience 24: 6086–6097. Ipponi A, Lamberti C, Medica A, Bartolini A, and MalmbergAiello P (1999) Tiagabine antinociception in rodents depends on GABAB receptor activation: Parallel antinociception testing and medial thalamus GABA microdialysis. European Journal of Pharmacology 368: 205–211. Kaupmann K, Malitschek B, Schuler V, et al. (1998) GABAB receptor subtypes assemble into functional heteromeric complexes. Nature 396: 683–687. Lehmann A, Antonsson M, Bremner-Danielsen M, Fla¨rdh M, Hansson-Branden L, and Ka¨rrberg L (1999) Activation of the GABAB receptor inhibits transient lower esophageal sphincter relaxations in dogs. Gastroenterology 117: 1147–1154. Lloyd KG, Thuret F, and Pilc A (1985) Upregulation of gammaaminobutyric acid (GABAB) binding sites in rat frontal cortex: A common action of repeated administration of different classes of antidepressant and electroshock. Journal of Pharmacology and Experimental Therapeutics 235: 191–199. Malcangio M and Bowery NG (1994) Spinal cord SP release and hyperalgesia in monoarthritic rats: Involvement of the GABAB receptor system. British Journal of Pharmacology 113: 1561–1566. Malcangio M and Bowery NG (1996) GABA and its receptors in the spinal cord. Trends in Pharmacological Sciences 17: 457–462. Manning J-P, Richards DA, and Bowery NG (2003) Pharmacology of absence epilepsy. Trends in Pharmacological Sciences 24: 542–549. Prosser HM, Gill CH, Hirst WD, et al. (2001) Epileptogenesis and enhanced prepulse inhibition in GABAB1-deficient mice. Molecular and Cellular Neuroscience 17: 1059–1070. Robbins MJ, Calver AR, Fillipov AK, Couve A, Moss SJ, and Pangalos MN (2001) The GABAB2 subunit is essential for G protein coupling of the GABAB receptor heterodimer. Journal of Neuroscience 21: 8043–8052.

GABAB Receptors: Molecular Biology and Pharmacology 367 Schuler V, Luscher C, Blanchet C, et al. (2001) Epilepsy, hyperalgesia, impaired memory, and loss of pre- and post-synaptic GABA B responses in mice lacking GABAB1. Neuron 31: 47–58. Slattery DA, Markou A, Froestl W, and Cryan JF (2005) The GABAB receptor-positive modulator GS39783 and the GABAB receptor agonist baclofen attenuate the reward-facilitating effects of cocaine: Intracranial self-stimulation studies in the rat. Neuropsychopharmacology 30: 2065–2072. Thuault SJ, Brown JT, Sheardown SA, et al. (2004) The GABAB2 subunit is critical for the trafficking and function of native GABAB receptors. Biochemical Pharmacology 68(8): 1655–1666.

Urwyler S, Mosbacher J, Lingenhoehl K, et al. (2001) Positive allosteric modulation of native and recombinant g-aminobutyric acidB receptors by 2,6-di-tert-butyl-4-(3-hydroxy-2,2dimethyl-propyl)-phenol (CGP7930) and its aldehyde analog CGP13501. Molecular Pharmacology 60: 963–971. Urwyler S, Pozza MF, Lingenhoehl K, et al. (2003) GS39783 (N,N0 dicyclopentyl-2-methylsulfanyl-5-nitro-pyrimidine-4,6-diamine) and structurally related compounds: Novel allosteric enhancers of g-aminobutyric acidB receptor function. Journal of Pharmacology and Experimental Therapeutics 307: 322–330. White JH, Wise A, Main M, et al. (1998) Heterodimerization is required for the formation of a functional GABAB receptor. Nature 396: 679–682.

GABAB Receptor Function A J Doherty, G L Collingridge, and S M Fitzjohn, University of Bristol, Bristol, UK ã 2009 Elsevier Ltd. All rights reserved.

Postsynaptic g-Aminobutyric Acid B-Mediated Inhibitory Postsynaptic Potential Activation of postsynaptic g-aminobutyric acid B (GABAB) receptors produces a hyperpolarization of neuronal membrane potential, accompanied by a decrease in cell input resistance. Synaptic activation of GABA receptors elicits a biphasic inhibitory postsynaptic potential (IPSP), of which the early component is sensitive to GABAA receptor blockade, and inhibitors of GABAB receptors block the slower, late component of the IPSP (Figure 1). This late component typically has a time to peak of 50–250 ms and decay time of 100–500 ms. The difference in time course of these two components is due to the different nature of the two GABA receptors. Thus, GABAA receptors consist of a ligand-gated Cl channel with fast kinetics, whereas the GABAB component is via G-protein activation of a Kþ conductance. The GABAB IPSP is mediated by bg G-protein subunits activating inwardly rectifying Kþ (Kir) channels, with efflux of Kþ causing membrane hyperpolarization (Figure 1). In the hippocampus, GABAB receptors couple to activation of the Kir3.2 subunit, and the GABAB receptor-mediated hyperpolarization of membrane potential is absent in Kir3.2 knockout mice. However, evidence suggests that a small component of the slow IPSP may, in some cases, also be mediated by activation of other Kþ channels, such as small-conductance Ca2þ-activated Kþ (SK) channels. In hippocampal pyramidal neurons, there is a strong co-localization of GABAB receptors and the Kir3.2 subunit in dendritic spines, with both proteins being enriched around glutamate synapses. In the main dendritic shaft, however, the two proteins show a much more segregated distribution. Kir channels are coupled to G-protein-coupled receptors by membrane diffusion of the bg subunit, with effective coupling occurring if the two proteins are located within a 500 nm distance. Thus, it is likely that GABABmediated activation of Kir and the associated membrane hyperpolarization are functionally much more important in dendritic spines than in the main dendritic shaft. GABAergic presynaptic fibers generally synapse onto the main dendritic shaft of hippocampal

368

pyramidal neurons, suggesting that the activation of Kir by postsynaptic GABAB receptors is initiated by GABA spillover from nearby GABA terminals. The slow time course of the GABAB IPSP and the localization of GABAB receptors in close vicinity to glutamate synapses have important implications for GABAB receptor function. Being closely located near the excitatory input to neurons, the GABAB-mediated response limits postsynaptic excitation. One function of the GABAB receptor-mediated increase in Kþ conductance and membrane hyperpolarization is to reduce activity of the N-methyl-D-aspartate (NMDA) subtype of ionotropic glutamate receptor (NMDAR). The NMDAR has slow kinetics and also displays a voltage-dependent block of the receptor channel by Mg2þ ions at membrane potentials more negative than 35 mV. As GABA is released from feed-forward interneurons, which receive their synaptic drive from the same glutamatergic fibers that form excitatory synapses onto hippocampal pyramidal neurons, the dual GABAA/GABAB IPSP is perfectly timed to prevent NMDAR activation during basal low-frequency activity. Activation of NMDARs is a key initiator of multiple forms of synaptic plasticity, which means that the GABAB receptor is an important factor in controlling the induction of synaptic plasticity.

Presynaptic Regulation of Neurotransmitter Release At presynaptic sites, GABAB receptors inhibit neurotransmitter release at both GABA synapses (acting as autoreceptors) and other neurotransmitter synapses (acting as heteroreceptors, e.g., at glutamate and noradrenaline synapses). This regulation is predominantly caused by inhibition of N- and P/Q-type voltage-gated Ca2þ channels. The inhibition is mediated by the bg subunits of GABAB receptorcoupled G-proteins. Inhibition of transmitter release brought about by activation of presynaptic Kþ channels may also occur. Presynaptic GABAB receptors produce a powerful effect on transmitter release, being capable of inhibiting release by more than 90%. Autoreceptors

GABAB receptors located on presynaptic GABA terminals act to decrease GABA release when two or more stimuli are delivered to presynaptic axons, leading to paired-pulse depression (PPD) of postsynaptic GABAA receptor-mediated responses (Figure 2). PPD occurs when pairs of stimuli are delivered in the frequency range of 0.1–50 Hz, with maximum PPD

GABAB Receptor Function 369

Excitatory terminal (glutamatergic) AMPAR NMDAR Inhibitory interneuron (GABAergic)

GABAAR

GABABR

Ca2+ K+

Cl−

EPSP

Cl−

Figure 1 Postsynaptic g-aminobutyric acid B (GABAB) receptors produce a slow postsynaptic membrane hyperpolarization. Schematic diagram showing membrane potential changes produced by synaptic stimulation. Release of glutamate from excitatory synapses activates postsynaptic a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptors (AMPARs), producing a fast excitatory postsynaptic potential (EPSP). Activation of GABAergic interneurons releases GABA that activates postsynaptic GABAA receptors, leading to Cl efflux and a fast inhibitory postsynaptic potential (IPSP). Spillover of GABA from the synapse activates GABAB receptors, where bg G-protein subunits activate nearby inwardly reactivating Kþ channels, producing a slow IPSP. This combined IPSP curtails the AMPAR-mediated EPSP and prevents activation of N-methyl-D-aspartate receptors (NMDARs). The traces show the change in membrane potential produced by activation of each receptor type. The color of the lines in these traces corresponds to the color of the receptor in the main diagram. The composite EPSP recorded in the postsynaptic cell is also shown.

seen at around 5–10 Hz. Depression of postsynaptic GABAA responses also occurs during trains of stimuli due to a depression of GABA release, although the effect of presynaptic GABAB inhibition decreases as the number of stimuli in the train increases or the frequency of the train increases (beyond around 10 Hz for a train of ten stimuli). With sufficient numbers of stimuli, the postsynaptic GABAA response undergoes activity-dependent facilitation and converts into a depolarizing response. Role of GABAB Autoreceptors in the Induction of NMDAR-Dependent Long-Term Potentiation

The effect of decreasing postsynaptic GABA-mediated inhibition during short trains of activity is important for the induction of long-term potentiation (LTP) of excitatory glutamatergic synaptic transmission, which has been particularly well studied in the CA1 region of the hippocampus. LTP is a well-studied form of synaptic plasticity that occurs in many brain regions and is considered a synaptic model of learning. Many, but not all, forms of LTP rely on the activation of NMDARs to act as a trigger for LTP induction. As discussed above, the postsynaptic GABAB receptormediated IPSP acts to inhibit the induction of LTP during basal, low-frequency synaptic activity

by preventing activation of the NMDAR. However, during short, high-frequency trains of activity, GABA release from presynaptic terminals is reduced by feedback of GABA onto presynaptic autoreceptors. The result is a decrease in postsynaptic GABA inhibition which, coupled to release from glutamate terminals, leads to a more persistent depolarization of the postsynaptic cell, activation of NMDARs, calcium influx, and the subsequent induction of LTP (Figure 2). Multiple protocols exist for inducing LTP in experimental systems, and the role of GABAB receptors in LTP induction varies depending on the protocol used. A typical induction protocol is the delivery of brief high-frequency bursts (e.g., two or more trains of four shocks at 100 Hz) with an interburst frequency of around 5 Hz (i.e., an interburst interval of 200 ms). This is termed a theta burst stimulus train as the 5 Hz pattern of activity mimics the physiological activity that occurs during the theta rhythm, a frequency of activation observed in the EEG when animals are exploring their environment. Such theta burst trains are thus considered potential physiological induction protocols for LTP induction. This frequency corresponds to the near-maximal effect of presynaptic GABAB autoreceptors, and activity at GABAB receptors is essential for induction of LTP. Thus,

370 GABAB Receptor Function

Excitatory terminal (glutamatergic)

AMPAR NMDAR

Inhibitory interneuron (GABAergic)

GABAAR

GABABR

Ca2+

EPSP

a

Excitatory terminal (glutamatergic) AMPAR NMDAR Inhibitory interneuron (GABAergic)

GABAAR GABABR

Ca2+

EPSP

b Figure 2 Activation of presynaptic g-aminobutyric acid B (GABAB) autoreceptors contributes to the induction of N-methyl-D-aspartate (NMDA) receptor-dependent LTP. (a) In response to a single stimulus, activation of postsynaptic GABA receptors produces a membrane hyperpolarization that curtails the excitatory postsynaptic potential (EPSP) and prevents NMDA receptor (NMDAR) activation, which remain blocked by Mg2þ ions (.). (b) In response to a train of stimuli, GABA activates presynaptic GABAB autoreceptors which decrease Ca2þ entry into the presynaptic bouton and thus decrease GABA release. This reduces the postsynaptic GABA-mediated inhibitory postsynaptic potential, which prolongs the AMPA receptor (AMPAR)-mediated membrane depolarization leading to expulsion of Mg2þ ions from the NMDAR and subsequent Ca2þ entry into the postsynaptic neuron. The inset traces show the membrane potential change produced by each constituent receptor and the composite EPSP recorded form the postsynaptic neuron.

GABAB Receptor Function 371

GABAB receptor antagonists block the induction of LTP by theta burst stimuli. Likewise, LTP induced by the delivery of a single priming stimulus followed approximately 200 ms later by a single burst (which could comprise as few as two stimuli, though typically four are used) is blocked by GABAB receptor antagonists. In this case the priming stimulus allows activation of the presynaptic GABAB autoreceptors such that during the subsequent burst of stimuli, GABA release is depressed sufficiently to allow NMDA receptor activation. LTP may also be induced by stronger stimulus trains, typically 100 stimuli delivered at a frequency of 100 Hz. Under these conditions, LTP induction is independent of GABAB receptor activity due to the conversion of the postsynaptic GABAA response into a depolarizing response. Depression of Vesicle Recruitment

At the excitatory calyx of Held synapse in the brain stem, activation of presynaptic GABAB receptors by GABA spillover from nearby inhibitory synapses elicits heterosynaptic depression of glutamate by depression of Ca2þ influx, as seen at other synapses in the brain. However, presynaptic GABAB receptor activation has an additional effect at this synapse in that it slows vesicle recruitment after strong presynaptic stimulation. This effect is mediated by G-protein inhibition of adenylyl cyclase and thus a reduction in cyclic adenosine monophosphate (cAMP) production. Vesicle priming at this synapse is enhanced by cAMP and Ca2þ/calmodulin, and thus a reduction in cAMP levels slows the recruitment of vesicles ready for release.

LTP of Postsynaptic GABAB Responses In addition to being involved in the induction of LTP of excitatory synaptic transmission, postsynaptic GABAB receptors may also undergo LTP. Thus in the hippocampus, NMDAR activation can lead to an enhancement of the slow, GABAB receptor-mediated IPSC in CA1 hippocampal pyramidal neurons. This form of LTP requires influx of Ca2þ through the NMDA receptor and the subsequent activation of Ca2þ/calmodulin dependent protein kinase II, although the precise mechanism of the enhancement of GABAB receptor function is unclear. It remains to be seen whether other forms of plasticity exist to modulate GABAB receptor function.

Coupling of GABAB and Metabotropic Glutamate Receptors In the cerebellar cortex, GABAB receptors are expressed in Purkinje cells (PCs) at the extrasynaptic site of excitatory input from parallel fibers. This extrasynaptic localization is also seen for the metabotropic glutamate

receptor subtype 1 (mGluR1), stimulation of which produces a slow excitatory postsynaptic current and mobilization of Ca2þ from intracellular stores. Activation of postsynaptic GABAB receptors by agonists or synaptic GABA release produces an enhancement of mGluR1-mediated synaptic currents and Ca2þ mobilization, an effect that is dependent on Gi/o G-protein signaling, activation of phospholipase C, and release of Ca2þ from intracellular stores. It is interesting that facilitation of mGluR1 in PCs by GABAB receptors can also occur independently of classical receptor activation. In the absence of GABAB receptor agonists, increases in extracellular Ca2þ elicit a constitutive increase in the sensitivity of mGluR1 to glutamate, facilitating mGluR1 effects. This effect of extracellular Ca2þ is independent of Gi/o activity and thus differs from the enhancement described above. GABAB and mGluR1 receptors appear to exist in a protein complex in PCs, and thus this enhancement may be due to changes in protein complex structure rather than to G-protein signaling.

Therapeutic Potential of GABAB Receptor Ligands Genetic deletion of either of the two GABAB receptor subunits or pharmacological blockade of native GABAB receptors produces spontaneous epileptic seizures, demonstrating the important role that GABAB receptors have in preventing overexcitation of neuronal networks and maintaining normal brain function. As such, it is possible that dysfunction of GABAB receptor activity may be implicated in human epilepsies, and polymorphisms of the GABAB1 receptor subunit are linked to increased susceptibility to temporal lobe epilepsy. In contrast, antagonists of GABAB receptors suppress absence seizures seen in animal models, whereas agonists promote seizures in these animals, suggesting that antagonists may be of potential therapeutic use in human absence seizures. However, in these models, high doses of GABAB receptor antagonists are prone to induce convulsions, highlighting the difficulty in targeting the GABAB receptor to treat this disease, in particular because of the widespread distribution of these receptors and the lack of receptor subtypes. The GABAB receptor agonist baclofen is currently used to treat spasticity and muscle rigidity in patients with multiple sclerosis and spinal cord injury. The effect of baclofen is thought to be due to activation of presynaptic heteroreceptors in the spinal cord, reducing the release of excitatory transmitters. Reduction of transmitter release in the spinal cord (particularly glutamate and substance P) probably also contributes to the antinociceptive effect of baclofen administration, although additional actions in the

372 GABAB Receptor Function

brain, such as in the rostral agranular insular cortex, likely also contribute to it. GABAB receptor subunit knockout animals display heightened sensitivity to painful stimuli such as heat, confirming the role of GABAB receptors in modulating pain pathways. A potential use of GABAB receptor agonists is in the treatment of drug addiction. Thus, agonists such as baclofen can reduce craving seen with drugs such as alcohol, cocaine, morphine, and nicotine. In addition, genetic studies suggest that there is a link between GABAB receptor variants and potential for addiction. This action of GABAB receptor activation in reducing addiction probably relates to activity in the mesolimbic dopamine system, which is part of the reward and reinforcing pathways of the brain. A further postulated use of activating GABAB receptors is to treat anxiety as GABAB receptor subunit knockout mice show increased anxiety in behavioral tests and baclofen has anxiolytic effects. See also: GABA Synthesis and Metabolism; GABAA

Receptor Synaptic Functions; GABAA Receptors: Molecular Biology, Cell Biology and Pharmacology; GABAB Receptors: Molecular Biology and Pharmacology; Long-Term Potentiation (LTP): NMDA Receptor Role; NMDA Receptor Function and Physiological Modulation.

Further Reading Bettler B, Kaupmann K, Mosbacher J, and Gasmann M (2004) Molecular structure and physiological functions of GABAB receptors. Physiological Reviews 84: 835–867. Bowery NG, Hill DR, Hudson AL, et al. (1980) ( )Baclofen decreases neurotransmitter release in the mammalian CNS by an action at a novel GABA receptor. Nature 283: 92–94.

Davies CH, Davies SN, and Collingridge GL (1990) Paired-pulse depression of monosynaptic GABA-mediated inhibitory postsynaptic responses in rat hippocampus. Journal of Physiology 424: 513–531. Davies CH, Starkey SJ, Pozza MF, and Collingridge GL (1991) GABA autoreceptors regulate the induction of LTP. Nature 349: 609–611. Dutar P and Nicoll RA (1988) A physiological role for GABAB receptors in the central nervous system. Nature 332: 156–158. Hirono M, Yoshioka T, and Konishi S (2001) GABA(B) receptor activation enhances mGluR-mediated responses at cerebellar excitatory synapses. Nature Neuroscience 4: 1207–1216. Huang CS, Shi SH, Ule J, et al. (2005) Common molecular pathways mediate long-term potentiation of synaptic excitation and slow synaptic inhibition. Cell 123: 105–118. Kornau HC (2006) GABA(B) receptors and synaptic modulation. Cell Tissue Research 326: 517–533. Kulik A, Vida I, Fukazawa Y, et al. (2006) Compartment-dependent colocalization of Kir3.2-containing Kþ channels and GABAB receptors in hippocampal pyramidal cells. Journal of Neuroscience 26: 4289–4297. Newberry NR and Nicoll RA (1984) Direct hyperpolarizing action of baclofen on hippocampal pyramidal cells. Nature 308: 450–452. Sakaba T and Neher E (2003) Direct modulation of synaptic vesicle priming by GABA(B) receptor activation at a glutamatergic synapse. Nature 424: 775–778. Soltesz I, Haby M, Leresche N, and Crunelli V (1988) The GABAB antagonist phaclofen inhibits the late Kþ-dependent IPSP in cat and rat thalamic and hippocampal neurones. Brain Research 448: 351–354. Tabata T, Araishi K, Hashimoto K, et al. (2004) Ca2þ activity at GABAB receptors constitutively promotes metabotropic glutamate signaling in the absence of GABA. Proceedings of the National Academy of Sciences of the United States of America 101: 16952–16957.

Glycine Receptors: Molecular and Cell Biology C Vannier and A Triller, INSERM U789, Ecole Normale Supe´rieure, Paris, France

mediated through gephyrin, but no direct link between GABARs and gephyrin has been demonstrated.

ã 2009 Elsevier Ltd. All rights reserved.

Molecular Biology and Topology of the Glycine Receptor Introduction The amino acid neurotransmitters glycine and g-aminobutyric acid (GABA) mediate inhibition in the vertebrate mature central nervous system by activating chloride conductance. Their ionotropic receptors belong to the canonical nicotinic acetylcholine receptor superfamily mediating fast synaptic transmission. Two features of this family, also known as the Cys-loop receptors, are its members’ heteropentameric structure delineating the channel pore, and their assembly from distinct gene products or subunit splice variants that all possess a strictly conserved disulfide loop in their extracytoplasmic domain. Upon transient neurotransmitter binding, a pore opens in the pentamer that is responsible for the selective movement of ions through the plasma membrane according to their electrochemical gradient. In fully mature neurons, glycine binding to its receptor mediates inhibition by eliciting an influx of chloride ions. Whereas GABAergic transmission is widespread in the nervous system, glycine-mediated inhibition and the glycine receptor (GlyR) are predominantly encountered in the brain stem and spinal cord. In the mid-1980s, GlyR was the first CNS receptor shown, by Triller and co-workers, to accumulate at postsynaptic differentiations (PSDs), in front of presynaptic release sites. Later on, variable subcellular organizations were demonstrated for many other receptors, including GABAA receptors (GABARs), which were analyzed by electron microscopy. Inhibitory synapses are intriguing, as both glycine and GABA open chloride channels and have a similar effect on the postsynaptic membrane. In the spinal cord, as well as in the brain stem and cerebellum, these synapses are either GABAergic, glycinergic, or mixed (i.e., respond to both neurotransmitters). The respective receptors accumulate in microdomains enriched with GABAR, GlyR, or a mixture of both, and it is the identity of the presynaptic element which imposes the composition of the postsynaptic membrane. Another protein, named gephyrin, has been copurified with GlyR on an affinity column. Antisense experiments and work with knockout mice demonstrate that gephyrin is required for the synaptic localization of both GlyR and GABAR. The GABAR g2 subunit is influential in synaptic insertion, since its absence in mouse results in a reduced number of GABA receptor clusters at synapses. The g2 subunit role is

GlyR was the first neurotransmitter receptor to be isolated from the mammalian brain more than two decades ago. The purification of the native molecule was achieved in H Betz’s laboratory, using affinity chromatography and taking advantage of the highaffinity interaction (KD  1–10 nM) of the receptor with its main antagonist, the well-known convulsant plant alkaloid, strychnine. The receptor appeared to be a complex of three proteins of 48 (a), 58 (b), and 93 kDa, in various vertebrate species. Several biochemical studies of purified (or the membrane-anchored) form of GlyR have established that the integral membrane glycoproteins a and b represent the constitutive subunits of the receptor. Cross-linking techniques showed that a and b assemble to build up the channel-containing transmembrane core of GlyR. The size of the complex (250 kDa) logically suggested a pentameric assembly of the subunits, a quaternary structure well established for other Cys-loop receptors (Figure 1). Unlike the a and b subunits, the 93 kDa co-purifying protein, gephyrin, is a nonglycosylated polypeptide able to interact reversibly with the a/b pentamer. Gephyrin is a cytoplasmic extrinsic membrane protein which binds tubulin dimers in vitro in a cooperative interaction of high affinity in the nanomolar range. It can also link brain microtubules to GlyR in a copolymerization assay. This finding underpinned the important notion that gephyrin could be a physical link between microtubules and GlyR in the postsynaptic membrane, as discussed in the following sections. The Glycine Receptor as a Member of the Ligand-Gated Ionotropic Receptor Family

Analysis of the primary structures of the a and b proteins revealed the high homology of GlyR subunits, with amino acid sequence identity close to 50%. The GlyR subunit domain organization is comparable to that of other subunits of the Cys-loop receptor family. Hydropathy analysis of the mature polypeptide predicted a conserved region of four hydrophobic segments (M1–M4) able to cross the membrane bilayer at positions identical to those of transmembrane segments of acetylcholine and GABAR subunits (Figure 1(a)). GlyR subunits are polytopic type I membrane proteins. In contrast to M4, the

373

374 Glycine Receptors: Molecular and Cell Biology NH2 a

A52 b CC

Ext

C C 212A

b

a

16GF159

K200 Y202 I111 T204 271R

a

M4 M1 M3

COOH

M2

M2

M2

b

M2 M2 M1

M1 M2 M3 M4 Int

a

R252

b

c

Figure 1 Structure of the glycine receptor. (a) Membrane topology of the a subunit, indicating the position of disulfide bridges, including the common Cys-loop delineated by Cys138 and Cys152 (red). Amino acid residues functionally important for agonist binding or modulation of the channel activity are shown. (b) Arrangement of a and b subunits in the mature a2b3 receptor. One a subunit is depicted in cross-section, and one b subunit is rendered transparent to show the four packed M1–M4 transmembrane regions (cylinders). The pore-forming domain M2 of the a subunit is dark brown. (c) The a/b complex viewed from the extracellular space along the fivefold symmetry axis. Shown are the N-terminal regions of the five protomers (dashed lines) the counterclockwise organization of the four transmembrane domains.

transmembrane domains M1–M3 are highly conserved among the various subunits of GlyR. All subunits possess an extended N-terminal domain, bearing potential glycosylation sites, which were detected early by immunocytochemistry in the synaptic cleft. Finally, a large region of significantly lower homology is represented by a hydrophilic loop between the M3 and M4 segments; this loop protrudes into the cytoplasm (M3–M4 loop). Thus GlyR exhibits the molecular design and transmembrane topology of the acetylcholine receptor superfamily, based on four membranespanning domains able to form a helices (M2) and/or b strands (M1, M3, M4), and is assumed to function as an allosteric protein. Our knowledge of this structure arose from extensive biochemical and electron microscopy data on the nicotinic acetylcholine receptor; these findings were recently refined by electron diffraction studies. The tertiary and quaternary structure of the soluble pentameric acetylcholine-binding protein (AChBP), which shares 20–24% and 17% sequence homology with nAChR and GlyR, respectively, has been useful in confirming or determining structure–function parameters for receptors of the family. Aside from an understanding of glycine and strychnine binding (see later), AChBP crystal structure provides a powerful model of the N-terminal domain of the Cys-loop receptor family. It shows that it is mainly a sandwich of antiparallel b sheets positioning conserved residues in order to stabilize the protomers, whereas variable residues are at their interface. The topology of the transmembrane domains (Figure 1) delineates the ion permeation pathway

away from the hydrophobic core of the phospholipid bilayer. For all Cys-loop receptors, the subunit’s a-helical M2 domain lines the central water-filled pore, while M1, M3, and M4 form the interface with the lipids and isolate M2 from a hydrophobic environment. Recent studies on the GlyR a1 subunit have challenged the four-helix model and provide evidence that whereas M2 and M4 are entirely helical, M1 and M3 also contain b strands. Assembly of GlyR

Initial studies of purified and membrane-anchored GlyR had suggested that a1 and b subunits assemble as a pentameric complex of a3b2 stoichiometry (Figures 1(b) and 1(c)). Short amino acid sequences (named assembly boxes), all located in the N-terminal extracytoplasmic domain of the b subunit and corresponding to three diverging motifs in a and b subunits, have been identified as having a role in precisely governing receptor assembly. Replacement of these motifs in the b subunit by the corresponding a1 motifs results in the loss of the subunit ratio in a/b oligomers, suggesting that different amino acid positions are determinants in the early step of subunit–subunit interaction. On the other hand, a1 and a2 co-assemble at variable subunit ratios. Importantly, these residues impose a mutually exclusive mode of assembly, either in complexes of invariant a/b stoichiometry or in homo-oligomers. A different stoichiometry (a2b3) has recently been established using affinity purification of expressed engineered tandem subunits. It confers on the b subunit a dominant role in the agonist binding properties of hetero-oligomeric a1/b GlyR, since ionic

Glycine Receptors: Molecular and Cell Biology 375

interactions at the a/b interface are required to stabilize glycine. Such a structure for GlyR is consistent with the concept that neurotransmitter binding in the Cys-loop family requires extracellular segments from two adjacent subunits. Molecular Diversity and Expression of Subunit Isoforms

The initial discovery of glycine receptor subtypes resulted from biochemical and immunochemical studies during rat spinal cord development. A comparison of GlyR immunoreactivities in neonatal and adult membrane extracts revealed a distinct form of GlyR predominating around birth (neonatal receptor, GlyRN) that displays a relatively low strychnine-binding affinity and only one major polypeptide band of 49 kDa. The discovery of functional a2 pentamers predominantly found during the fetal and neonatal stages supported the idea of a developmental switch from GlyRN to the adult form, GlyRA, corresponding to the formation of a1/b heteromers. This switch, taking place within 3 weeks after birth, is not as complete in other regions of the nervous system where mixtures of a1, a2, and b subunits remain throughout adulthood. One current view is that postsynaptic GlyR corresponds to a mixture of a2 pentamers and a2/b heteromers in immature neurons, while a1/b heteromers predominate in mature synapses. A greater diversity of GlyR subunits than anticipated was established and includes not only the original a1 and a2 but also the a3 and a2* subunits. A murine a4 subunit has also been identified which is absent in rat and human. These isoforms are highly homologous, with amino acid sequence identities on the order of 80%, and display a very low degree of interspecies variation (identity close to 99%). These isoforms can form glycine-gated chloride channels of comparable strychnine sensitivity, except a2*, which is more than 99% identical to human a2 but exhibits a 500-fold lower strychnine sensitivity. Further generation of diversity resides in alternative exon usage. For the a1 subunit, a variant mRNA encodes the form denoted ‘a1ins,’ a subunit containing eight additional amino acids. This sequence, inserted in the M3–M4 loop, bears a serine residue, representing a potential phosphorylation site and allowing a functional modulation of the a1 subunit. Alternative splicing was also found for the a2 and a3 subunits. The distribution of individual GlyR subunits was determined using in situ hybridization. However, transcription is not necessarily correlated with surface expression of functional receptors and the identification of the actual composition of GlyR in particular areas is not yet firmly established. The gene encoding

the a1 subunit and that encoding the a3 subunit, though to a lesser extent, are mainly transcribed in spinal cord and brain stem at later postnatal stages. The a3 subunit mRNA is present in the infralimbic system, the hippocampal complex, and the cerebellar granular layer. Expression levels of the a2 subunit, by contrast, are high in the embryonic and perinatal stages but barely detectable in the adult brain, with some expression in higher cortical regions. The a4 subunit, which is expressed at low levels in the adult but forms functional GlyRs, is restricted to the spinal cord and the sympathetic nervous system. The b subunit mRNA is more widely expressed throughout the embryonic and adult nervous system as compared to a subunit transcripts, and is found in brain loci devoid of strychnine-binding sites or GlyR immunoreactivity. The physiological significance of this widespread expression of the b subunit transcript is not understood, as so far there is no evidence that it assembles with a subunits of other receptors.

Gephyrin and Gephyrin-Binding Proteins The Gephyrin Polypeptide

Gephyrin is the best characterized GlyR-interacting protein. Biochemical characterization and heterologous expression have shown that gephyrin binds cooperatively to tubulin dimers with a high affinity, and to GlyR via an 18-amino-acid sequence in the b subunit M3–M4 cytoplasmic loop. It was therefore postulated that gephyrin is a linker of the receptor to microtubules, adapted to anchor GlyR at synapses via the cytoskeleton. Gephyrin exists as several splice variants, as indicated by the characterization of the mouse gephyrin single gene and of its exonic structure, and the cloning of several different full-length coding sequences. Gephyrin variants, differing by the presence of distinct nucleotide sequences termed cassettes, are differentially expressed in numerous tissues, including brain, heart, skeletal muscle, and kidney, liver, lung, spleen, or testis, illustrating the ubiquitous nature of gephyrin. Tertiary and Quaternary Structures of Gephyrin

The gephyrin polypeptide originated during evolution from the fusion of two genes of bacterial origin, respectively encoding the MogA and MoeA proteins involved in the biosynthesis of the molybdenum cofactor in Escherichia coli. These proteins are homologous to the gephyrin N- and C-terminal domains (G- and E-domains), respectively, flanking a 170residue unique linker region. Intriguingly, in gephyrin these domains are enzymatically functional, and

376 Glycine Receptors: Molecular and Cell Biology

disruption of the gephyrin gene leads to molybdoenzyme activity deficiency. These results indicate that gephyrin maintains MogA and MoeA activities in various tissues. Gephyrin tertiary and quaternary structures were determined following the crystallization of MogA and MoeA, and of gephyrin G- and E-domains. The G- and E-domains are compact structures forming stable trimers and dimers, respectively, which can be detected in crystals. If conserved in the gephyrin molecule, these oligomerization properties allow the construction of a two-dimensional hexagonal gephyrin lattice. This oligomerization pattern confers on gephyrin its function as a scaffold core protein and its ability to recruit GlyR at postsynaptic localizations. The GlyR-binding activity of gephyrin has been assigned in the E-domain to a pocket adjacent to the dimer interface.

Gephyrin-Interacting Proteins

Gephyrin is the core of the inhibitory postsynaptic scaffold (Figure 2). Binding partners for gephyrin other than GlyR have recently been identified, but their exact functions in receptor postsynaptic clustering have not yet been determined. Gephyrin may also be implicated in the modulation of signaling pathways. Gephyrin could thus participate in translational controls by binding to the RAFT1 kinase (a rapamycin and FKB12 target protein). The dynein light chains 1 and 2 (Dlc1/2) were recently identified as gephyrin-binding proteins. Further, gephyrin interaction with collybistin, a GDP/GTP exchange factor for GTPases of the Rho/Rac family, or with proteins such as profilin and Mena/VASP, could be a determinant for microfilament organization. This hypothesis

a

GlyR

b

PIP3

Drebrin Dlc2 Gephyrin Mena/ VASP RAFT1 Microtubules

Protein synthesis ?

Collybistin

MoCo synthesis

Profilin

CDC42 GDP

CDC42 GTP

F-actin

G-actin

Dystrophin

Figure 2 Molecular partners of the glycine receptor (GlyR)/gephyrin complex. Gephyrin is the only glycine receptor-interacting protein known so far. At synapses, it could form intermolecular trimers and dimers via its G- and E-domains, respectively. The localization of the receptor at synaptic loci relies on the ability of the former to interact directly with microtubules and indirectly with other proteins (collybistin and actin-binding proteins; yellow) able to modulate the dynamics of microfilaments. Until now the existence at synapses of a complex built up via simultaneous interactions involving all of these proteins has not been demonstrated. Double arrows indicate interactions disclosed by biochemical or two-hybrid strategies. A rapamycin and FKB12 target protein (RAFT1) is a translational activator which could mediate the synthesis of proteins required at synapses (Dlc2, dynein light chain 2; MoCo, molybdenum cofactor). Membrane binding sites for some partner proteins such as collybistin and profilin can be provided by phosphatidylinositol 3,4,5-trisphosphate (PIP3). Functions fulfilled by components of the proposed interaction network are shown by red arrows.

Glycine Receptors: Molecular and Cell Biology 377

is consistent with results supporting the role of microfilaments in the control of postsynaptic gephyrin– GlyR co-cluster density. A feature of gephyrin is the presence of sequence stretches predicted to be WW (class IV) domain interaction motifs (ELM database search). A WW domain which is encountered in dystrophin/utrophin polypeptides might mediate phosphorylation-dependent binding to gephyrin. Interestingly, dystrophin, which is co-localized with dystroglycans, is present in a subset of inhibitory GABAergic synapses in the hippocampus, where it could be involved in a link between the extracellular matrix and the cytoskeleton.

Glycine Receptor Dynamics Postsynaptic membrane domains and associated PSDs have long been viewed as stable multimolecular assemblies ensuring that a specific set of functions is performed locally. However, in recent years the notion has emerged that the postsynaptic domain is a highly dynamic protein complex adapted to the mechanisms underlying the plasticity of synaptic contacts. On the one hand, as for cellular multimolecular assemblies, postsynaptic complexes that contain GlyR correspond to the equilibrium between the exo- and endocytic delivery routes for membrane proteins. On the other hand, interactions involving cytoplasmic scaffolding proteins are pivotal in this steady-state balance that can be modified by cell differentiation or receptor activity. Activity and GlyR Stability at Synapses

Receptor stabilization and aggregation, mediated by interactions with scaffold elements, are coupled at synaptic loci. The formation of GlyR clusters retained in the postsynaptic membrane in relation with receptor activation has been studied using an activity blockade by strychnine of cultured spinal cord neurons. This blockade leads to a reversible intracellular redistribution of GlyR which disappears from synaptic loci without alteration of gephyrin. Therefore, aggregation of gephyrin and that of GlyR are governed by distinct mechanisms. GlyR redistribution results from a defect in biosynthetic routing. At the steady state, the spinal cord neurons incorporate only a minor fraction of newly synthesized GlyR in synapses. It turns out that the stability of glycinergic synapses is due to a tight control of the relative rates of receptor synthesis, synaptic anchoring, and degradation. In order to explain selective positioning of active GlyR in front of glycinergic terminals, it has been proposed that glycine, which depolarizes immature neurons, triggers a calcium influx promoting the assembly of the postsynaptic scaffold.

Ubiquitination has recently been proposed to account for the regulation of cell surface GlyR a1 subunit endocytosis and degradation, thus determining postsynaptic receptor density. This hypothesis is based on the demonstration that ubiquitination of homopentameric GlyR occurs almost exclusively at the plasma membrane rather than during intracellular transport of the receptor. Routing of Newly Synthesized GlyR to Synapses

The mechanisms of selective accumulation of receptors at synaptic sites are not fully understood. However, newly synthesized homomeric receptors spontaneously form microaggregates devoid of gephyrin in the neuronal plasma membrane. Although they can also in part remain in the extrasynaptic space, they associate with endogenous synaptic gephyrin according to a mechanism depending on the formation of the inhibitory innnervation. Interestingly, newly synthesized GlyR molecules are inserted in the plasma membrane as stable clusters which initially appear in the extrasynaptic space of the soma and dendritic proximal segments. Subsequently, they are redistributed over the dendritic tree. This occurs by diffusion of cell surface molecules, but not via exocytosis of vesicles routed to dendrites. Therefore, GlyR postsynaptic accumulation results from a diffusion/retention mechanism. This posits the crucial role of the retention signal, and in turn the anchoring role of gephyrin emerges. The complexity of the mechanisms of postsynaptic anchoring process, which relies on the exocytosis site, diffusion velocities, and on the concentration of synaptic binding sites, is illustrated by glutamate receptors of different compositions which accumulate at excitatory synapses with different apparent rates. Gephyrin associates with vesicular compartments of GlyR intracellular trafficking and is a mediator of the insertion of newly synthesized receptor molecules in the plasma membrane, a process dependent on microtubule integrity. It has been further proposed that a nonsynaptic gephyrin/GlyR complex can be recruited to microtubules by dynein for retrograde transport along neurites. The existence of intracellular gephyrin/GlyR complexes indicates a dual function in trafficking and in assembly of postsynaptic scaffolds, as for proteins of excitatory synapses. Local dendritic insertion of newly synthesized receptors is also a means of changing receptor numbers at synapses via mRNA localization. The discovery of the dendritic localization of GlyR a subunit mRNAs supported the view that local synthesis plays a role in synaptic plasticity. Remarkably, a microsecretory apparatus able to participate in the translation of membrane proteins is present in the subsynaptic cytoplasm.

378 Glycine Receptors: Molecular and Cell Biology Diffusion Properties of Cell Surface GlyR

In addition to membrane traffic, lateral diffusion within the plasma membrane has been demonstrated for GlyR. This concept has been extended to almost all excitatory and inhibitory receptors so far studied. The single-particle tracking (SPT) approach showed that receptor movements display interspersed periods of high and low diffusion rates in the extrasynaptic and synaptic membrane, respectively. Whereas periods of fast diffusion correspond to Brownian movement, those of slow mobility mainly result from transient association of GlyR with clusters of gephyrin. This association, which reduces diffusion coefficients from up to 0.5 mm2 s1 to less than 102 mm2 s1, imposes a restricted area of exploration (confinement) similar to that generated for other proteins by insertion into lipid rafts or by contact with protein fences or obstacles. The fact that the receptors are not irreversibly retained in gephyrin domains, but can also exit from them, is a key finding which sheds light on how receptor clusters are formed or modified during plastic processes. It favors the new notion that any postsynaptic cluster of gephyrin can behave as donor or/ and acceptor of GlyR. Therefore, the dynamic properties of GlyR–gephyrin interactions are also a means whereby the number of synaptic receptors can be regulated. It remains to be seen how the stabilizing scaffold molecules regulate these transitions, both in synaptogenesis and synaptic plasticity. Because of its interaction with scaffold proteins, the cytoskeleton is a component of the postsynaptic differentiation, at least during differentiation or plastic changes. Because gephyrin is thought to link GlyR to microtubules, depolymerization of the latter would disrupt GlyR clusters. Interestingly, microtubule depolymerization alters the number and density of synaptic GlyR/gephyrin clusters, but the same treatment has no effect on GABA receptor/gephyrin clusters. This discrepancy may reflect the structural heterogeneity of inhibitory synapses that likely relies on receptorspecific assemblies of cytoplasmic proteins.

Pathology – Molecular Biology of Spastic Syndromes Defects in glycinergic transmission are the cause of several neurological diseases. Some hereditary motor disorders in humans, mice, horses, and cattle result from the expression of mutated glycine receptor genes. The hallmark of these disorders is an exaggerated startle reflex in response to sudden stimuli, giving rise to rapid generalized responses in the form of hypertonia. Startle syndromes related to distinct

genetic defects in glycine-mediated neurotransmission correspond either to impairment of GlyR agonist binding function or to reduced expression of functional channels. Both mechanisms lead to inefficient glycinergic inhibition and increased muscle tone. Human Hyperexplexia

Hyperexplexia is a disorder in which unexpected sensory stimuli provoke exaggerated startle reflexes. Hereditary hyperexplexia, or Kok’s disease or familial startle hyperexplexia (STHE), is a rare autosomal dominant neurologic disorder characterized by marked and continuous, sometimes fatal, muscle rigidity in infancy which progressively evolves into transient massive muscle contractions, a reaction that persists throughout adulthood. Several missense mutations identified in the a1 subunit of GlyR can be responsible for higher glycine EC50 and/or lower channel conductance. The defective phenotype often results from mutations within, or in the vicinity of, the M2 transmembrane domain. Startle Syndromes in the Mouse

Studies of mutants in the mouse reveal that defects in two genes can cause recessive disorders with clinical and pharmacological features reminiscent of STHE. The inherited spasmodic and spastic phenotypes are autosomal diseases characterized by muscle tremor, hypertonia, and pronounced startle reactions. They are rarely lethal, with the exception of the oscillator phenotype, and during the first 2 weeks after birth homozygous mutant mice have a normal phenotype. The phenotypic similarity of spasmodic (spd) phenotype and STHE prompted the identification of the GlyR a subunit as the defective protein. A missense mutation, A52S, close to the protein N-terminus, decreases glycine sensitivity without changing the affinity for strychnine, in agreement with the normal binding of the antagonist in the spd spinal cord. The mechanism of action of this mutation is unknown, since the region of the mutation is not involved in ligand binding, receptor assembly, or gating. The phenotype of the mutant mouse ‘oscillator’ (spdot) is characterized by a fine motor tremor and muscle spasms. Oscillator is allelic with spasmodic but corresponds to a microdeletion coincident with the cytoplasmic extremity of the M3 transmembrane segment. A translational frameshift generates a truncated form of the a1 subunit lacking the M3–M4 loop and the M4 domain. The highly penetrant recessive ‘spastic’ mutation is associated with a dramatic reduction in the number of expressed GlyRs in the adult spinal cord and brain stem, but residual receptors exhibit normal subunit

Glycine Receptors: Molecular and Cell Biology 379

structure and function. The delayed onset of the spasticity coincides with the developmental switch of GlyR, suggesting a selective lack of the adult form, in the absence of transcriptional alteration of the a1 subunit. A mutation affects a b subunit gene allele, which results from the intronic insertion (intron 5) of a LINE-1 transposable element. As a consequence, fulllength b mRNAs are barely detectable, and truncated b subunits are generated by creation of premature stop codons. The general loss of surface receptors in ‘spastic’ animals is consistent with the role of the b subunit, and results from the lack of a gephyrin-binding site in (hetero-)oligomers. However, it should be noted that a reduction in transport to, and insertion in, specific membrane areas may alternatively arise from an altered quaternary structure that does not allow exit from the endoplasmic reticulum and leads to degradation of the misassembled receptor.

Concluding Remarks GlyR, which plays essential roles in inhibitory neurotransmission via a heterogeneous composition, has been the center of numerous studies during recent years. Better understanding has been gained at various levels, from structure–function relationships to the cell biology of synapses, including the associated postsynaptic density built up on gephyrin. Remarkably, pioneering work exploiting GlyR as a model molecule has helped develop new concepts concerning not only the biology of neurotransmitter receptors, but also the characterization of the dynamic equilibrium of GlyR-enriched postsynaptic membrane domains; such equilibrium accounts for the strength of synapses via control of their receptor number. GlyR and gephyrin are central to the assembly and activity of a nanomachine, which exploits the diffusion properties of the receptors in order to participate in the excitation–inhibition balance that is modified during plastic events within neuronal networks.

See also: GABA Synthesis and Metabolism.

Further Reading Aguayo LG, van Zundert B, Tapia JC, et al. (2004) Changes on the properties of glycine receptors during neuronal development. Brain Research: Brain Research Reviews 47: 33–45. Cascio M (2004) Structure and function of the glycine receptor and related nicotinicoid receptors. Journal of Biological Chemistry 279: 19383–19386. Choquet D and Triller A (2003) The role of receptor diffusion in the organization of the postsynaptic membrane. Nature Reviews Neuroscience 4: 251–265. Colquhoun D and Sivilotti LG (2004) Function and structure in glycine receptors and some of their relatives. Trends in Neuroscience 27: 337–344. Dahan M, Levi S, Luccardini C, et al. (2003) Diffusion dynamics of glycine receptors revealed by single-quantum dot tracking. Science 302: 442–445. Hanus C, Vannier C, and Triller A (2004) Intracellular association of glycine receptor with gephyrin increases its plasma membrane accumulation rate. Journal of Neuroscience 24: 1119–1128. Kim EY, Schrader N, Smolinsky B, et al. (2006) Deciphering the structural framework of glycine receptor anchoring by gephyrin. The EMBO Journal 25: 1385–1395. Kneussel M and Betz H (2000) Clustering of inhibitory neurotransmitter receptors at developing postsynaptic sites: The membrane activation model. Trends in Neuroscience 23: 429–435. Laube B, Maksay G, Schemm R, et al. (2002) Modulation of glycine receptor function: A novel approach for therapeutic intervention at inhibitory synapses? Trends in Pharmacological Sciences 23: 519–527. Legendre P (2001) The glycinergic inhibitory synapse. Cellular and Molecular Life Sciences 58: 760–793. Lynch JW (2004) Molecular structure and function of the glycine receptor chloride channel. Physiological Reviews 84: 1051–1095. Moss SJ and Smart TG (2001) Constructing inhibitory synapses. Nature Reviews Neuroscience 2: 240–250. Rosenberg M, Meier J, Triller A, et al. (2001) Dynamics of glycine receptor insertion in the neuronal plasma membrane. Journal of Neuroscience 21: 5036–5044. Triller A and Choquet D (2005) Surface trafficking of receptors between synaptic and extrasynaptic membranes: And yet they do move! Trends in Neuroscience 28: 133–139. Vannier C and Triller A (1997) Biology of the postsynaptic glycine receptor. International Review of Cytology 176: 201–244.

This page intentionally left blank

AMINES AND ACETYLCHOLINE

A. Dopamine and Norepinephrine

383

B. Serotonin (5-HT)

450

C. Acetylcholine

478

This page intentionally left blank

Dopamine J D Elsworth and R H Roth, Yale University School of Medicine, New Haven, CT, USA ã 2009 Elsevier Ltd. All rights reserved.

Anatomy Mesencephalon

The majority of central dopaminergic projections arise in the midbrain from the ventral tegmental area (VTA; A10), substantia nigra (SN; A9), and retrorubral area (A8) to innervate three principal sets of targets: the striatum (the caudate and putamen), the limbic cortex (medial prefrontal, cingulate, piriform, and entorhinal areas), and other limbic structures (regions of the septum, olfactory tubercle, nucleus accumbens, and amygdaloid complex). These systems are frequently referred to as the nigrostriatal, mesocortical, and mesolimbic dopamine (DA) projections, respectively. Although the A9 cells preferentially innervate the striatum, and the A10 cells mainly target the limbic and cortical regions, Figure 1 illustrates that the projections of the midbrain DA cells are not restricted to these regions. Many of the SN DA neurons give off collateral branches and ramify within more than one region. The projection fields of the A8, A9, and A10 cell groups are collectively known as the mesotelencephalic DA system. This term has been used to refer to the projections from the midbrain DA neurons (A8, A9, and A10) to the striatum and nucleus accumbens. The A8–A10 groups comprise approximately 20 000–30 000 neurons on each side of the rat brain. Also located in the mesencephalon, the rostral linear nucleus is sometimes considered part of the VTA. DA neurons of the rostral linear nucleus and periaqueductal gray region project locally or to different brain regions, such as the central amygdaloid nucleus, bed nucleus of the stria terminalis, sublenticular extended amygdala, and substantia innominata. It is now known that heterogeneities of DA innervation occur within a region. Thus, the nucleus accumbens can be divided into two cytoarchitectonically, physiologically, and pharmacologically distinct compartments, termed core and shell. The core region and the shell region receive preferential inputs from the A9 and A10 cell groups, respectively. Furthermore, two histochemically distinct compartments within the striatum have been defined, known as ‘patch’ (or ‘striosome’) and ‘matrix,’ which contain different densities of a number of markers for dopaminergic, cholinergic, and peptidergic neurons and

represent another level of processing in the basal ganglia. Evidence suggests that the mesostriatal DA input to these compartments arises from distinguishable subsets or ‘tiers’ of the DA-containing midbrain neurons. Diencephalon

These systems include the A11–A15 cell groups, which are located in the diencephalon, including regions such as the hypothalamus, caudal thalamus, zona incerta, and preoptic region. These neurons project principally within the diencephalon and have been divided into periventricular, incertohypothalamic, and tuberohypophyseal systems. There are approximately 2000–4000 DA neurons on each side of the brain in the rat diencephalic A11–A15 groups. An additional long-length system is the descending projections from the A11 group in the diencephalon to the spinal cord, where a dopaminergic innervation is apparent in the dorsal horn at all levels of the cord. Other Systems

Among the other DA systems are the short-length interplexiform amacrine-like neurons, which link inner and outer plexiform layers of the retina. There is also a population of periglomerular DA neurons in the olfactory bulb, which link mitral cell dendrites in separated adjacent glomeruli.

Function The different locations of the various DA systems in the brain dictate that they have different afferent and efferent connections, which in turn determine the roles they play in brain function. These are reviewed here. It should also be appreciated, however, that it is not possible to completely segregate the motor, motivational, and cognitive behaviors that are often attributed to the nigrostriatal, mesolimbic, and mesocortical DA systems, respectively. Nigrostriatal System

Dopaminergic innervation of the basal ganglia plays an essential role in many aspects of motor control, cognition, and emotion. The striatum is one of the principal input structures of the basal ganglia. The importance of the DA input to the striatum is evident from the motor abnormalities (e.g., bradykinesia, rigidity, and tremor) exhibited by patients with Parkinson’s disease, which is characterized by a marked loss of nigrostriatal DA neurons. DA innervations to limbic and cortical regions are also affected in Parkinson’s disease but to a lesser extent than the

383

384 Dopamine

A10 A9 Anterior cingulate cortex

Corpus callosum

A8 Hippocampus

Prefrontal cortex

Lateral septum Bed nucleus stria terminalis

Habenula

Striatum Perirhinal cortex Nucleus accumbens

A10 A9 A8

Olfactory tubercle Pyriform cortex

Central nucleus amygdala

Locus coeruleus Lateral parabrachial nucleus

Entorhinal cortex

Figure 1 Schematic diagram illustrating the distribution of the main neuronal pathways containing DA in the central nervous system. The cell groups are named according to the 1965 nomenclature of Dahlstrom and Fuxe. Different colors denote the terminal field projections of the A8, A9, and A10 cell groups. Recent investigations have established that collaterals of ascending DA neurons also provide innervation of the subthalamic nucleus, thalamus, and globus pallidus.

striatal input. DA release in the striatum modulates activity in two striatopallidal circuits (‘direct’ and ‘indirect’), which in turn exert control over thalamocortical circuits essential for voluntary control of movement. The signs of Parkinson’s disease do not appear until a large majority (approximately 80%) of nigrostriatal DA neurons have been lost, as surviving DA neurons efficiently increase their activity to compensate for the damage. It is now apparent that nigrostriatal DA neurons do not merely allow motor behavior to occur but that they also play an important role in the selection and initiation of actions and establishing motor skills and habits. Current treatments for Parkinson’s disease typically involve administration of the DA precursor, dihydroxyphenylalanine (L-dopa), or DA receptor agonists. Future treatments are likely to include restoration of dopaminergic tone by transplantation of fetal DA neurons or stem cells or by gene therapy. Mesolimbic System

The function of the mesolimbic dopaminergic system, particularly the projection from the VTA to the nucleus accumbens, has been strongly implicated in goal-oriented (motivated) behaviors, in addition to reward, attention, and pharmacologically induced locomotion. Enhancement of DA transmission in

this system has been linked with the addicting, reinforcing, and sensitizing effects of repeated exposure to psychostimulant drugs of abuse. Furthermore, the therapeutic effects of antipsychotic drugs used in the treatment of schizophrenia may depend on the inhibition of mesolimbic DA neuron activity that these drugs induce. The ability of antipsychotic drugs to block DA receptors and thereby reduce DA transmission is central to the 40-year-old DA hypothesis of schizophrenia, which posits that the disease is related to excessive central DA activity. Imaging studies in patients have provided support for the existence of disrupted DA transmission in schizophrenia. A revision to the hypothesis is the concept of underactivity in cortically projecting DA neurons together with overactivity in subcortical DA systems. Persuasive evidence implicates developmental, functional, and structural abnormalities in schizophrenia, involving other transmitter systems besides DA, such as glutamate and g-aminobutyric acid (GABA). Mesocortical System

The prefrontal cortex has rich connections with other neocortical regions, limbic regions, and other subcortical regions. The prefrontal cortex has been implicated in a wide variety of cognitive functions, and it particularly appears to be involved in directing

Dopamine 385

appropriate attention, prioritizing the significance of stimuli, monitoring the temporal sequence of stimuli, referencing stimuli to internal representations or cues, and devising abstract concepts. The well-defined DA projection to the prefrontal cortex has been suggested to be involved in short-term and ‘working’ memory. It has been hypothesized that the mesoprefrontal DA system is involved in the pathophysiology of schizophrenia and that defects in this system, or its associated neuronal connections, may be responsible for the negative and cognitive deficits that characterize the disorder. Diencephalon Systems

DA neurons in posterior dorsal hypothalamus and periventricular gray of central thalamus (A11) project to spinal cord and appear to have a role in sensory and nociceptive processing and sensory integration. Tuberoinfundibular DA neurons (A12) are located in the arcuate nucleus and periventricular nucleus of mediobasal hypothalamus and project to the external layer of median eminence. They play an important role in inhibiting the release of prolactin from the anterior lobe of the pituitary. Incertohypothalamic DA neurons (A13) reside in the rostral portion of medial zona inserta and terminate in the central nucleus of amygdala, horizontal diagonal band of Broca, and paraventricular nucleus of the hypothalamus. Their functions are not clear but probably involve integration of autonomic and neuroendocrine responses to sensory stimuli. Two subpopulations of A14 neurons have been distinguished. The periventricular– hypophysial (tuberohypophyseal) DA neurons terminate in the intermediate lobe of the posterior pituitary and inhibit secretion of a-melanocytestimulating hormone and pro-opiomelanocortinderived peptides such as b-endorphin. DA neurons also innervate the neural lobe, but their origin is controversial. The other defined group of A14 DA neurons is located in the periventricular hypothalamic nucleus. Fibers of these neurons extend into the medial preoptic nucleus and anterior hypothalamus. Females possess greater number of these neurons than males, and these neurons are believed to play a role in gonadotrophin secretion in females and reproductive behavior in males. DA neurons in the ventrolateral hypothalamic comprise the A15 cell group, and they extend processes to the lateral hypothalamus and caudal supraoptic nucleus, suggesting a role in oxytocin and/or vasopressin secretion. Other Systems

Retinal DA neurons in vertebrates appear to be particularly involved in contrast sensitivity and light adaptation. In 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine

(MPTP)-treated parkinsonian monkeys, a loss of retinal DA occurs, along with spatial frequency-dependent abnormalities in both the pattern electroretinogram and the visual evoked potential. In Parkinson’s disease, abnormalities occur in ocular motor control, visual evoked potentials, electroretinograms, color vision, and visual contrast sensitivity. Changes are also possible in the acquisition and perception of visual information. In addition to the DA systems in the brain and pituitary, DA appears to have a restricted role in the periphery. The best defined role for peripheral DA neurons is in renal regulation of sodium homeostasis, and impairment of this function may underlie development of hypertension.

Life Cycle Figure 2 is a schematic model of a dopaminergic nerve terminal illustrating the life cycle of DA and the regulatory mechanisms that modulate the synthesis, storage, and release of this important neurotransmitter. A brief description of this idealized nerve terminal and model synapse will help to summarize the present state of knowledge concerning the life cycle and regulatory control of DA in a typical DA synapse and possible sites at which drugs and chemicals can influence dopaminergic transmission. Synthesis

Blood-borne tyrosine, derived from dietary proteins and from phenylalanine metabolism, enters the brain by a low-affinity amino acid transport system. Tyrosine in brain extracellular fluid is taken up into DA neurons by high- and low-affinity amino acid transporters. The relative circulating levels of tyrosine and phenylalanine can affect central catecholamine metabolism because these amino acids compete for transport into the brain and for transport into the neuron. DA is synthesized from tyrosine in two enzymatic steps. The first reaction is catalyzed by the enzyme tyrosine hydroxylase in the cytosol. Because this step is rate limiting, it sets the pace for the conversion of tyrosine to DA and is the step most susceptible to physiologic regulation and pharmacologic manipulation. Short-term activation of tyrosine hydroxylase involves phosphorylation of the regulatory domain by protein kinases. The activated form of tyrosine hydroxylase is thought to have a lower Km for its pterin cofactor and a higher Ki for DA, which effectively reduces end product inhibition. The activity of the enzyme dihydropteridine reductase is indirectly linked to DA biosynthesis because this enzyme catalyzes the recycling of the

386 Dopamine

Nerve impulse

AT

VM

(+) TH cAMP (+) Tyrosine (+) Release modulating autoreceptor

Ca2+ DA

DA

Nonvesicular DA

DA DA

TH

(−) (−) DA TH* DOPA End product inhibition DA DA

DA

DA

Synthesis modulating autoreceptor DA

DA DA DA DA DA

DA

DAT

DA

DAT

Figure 2 Schematic model of a dopaminergic nerve terminal illustrating the life cycle of dopamine (DA) and the mechanisms that modulate its synthesis, release, and storage. Invasion of the terminal by a nerve impulse results in the Ca2þ-dependent release of DA. This release process is attenuated by release-modulating autoreceptors. Increased impulse flow also stimulates tyrosine hydroxylation. This appears to involve the phosphorylation of tyrosine hydroxylase (TH), resulting in conversion to an activated form with greater affinity for tetrahydrobiopterin cofactor and reduced affinity for the end product inhibitor DA. The rate of tyrosine hydroxylation can be attenuated by (1) activation of synthesis-modulating autoreceptors, which may function by reversing the kinetic activation of TH, and (2) end product inhibition by intraneuronal DA, which competes with cofactor for a binding site on TH. Release- and synthesis-modulating autoreceptors may represent distinct receptor sites. Alternatively, one site may regulate both functions through distinct transduction mechanisms. The plasma membrane DA transporter (DAT) serves an important physiological role in the inactivation and recycling of DA release into the synaptic cleft. The vesicular monoamine transporter (VMAT) transports cytoplasmic DA into storage vesicles, decreasing the cytoplasmic concentration of DA and preventing metabolism by MAO. VMAT modulates the concentration of free DA in the nerve terminals. Reproduced from Cooper JR, Bloom FE, and Roth RH (2003) The Biochemical Basis of Neuropharmacology, 8th edn. New York: Oxford University Press, with permission from Oxford Press.

quinonoid dihydrobiopterine to tetrahydrobiopterine, which is an essential cofactor of tyrosine hydroxylase. In addition, the synthesis of tetrahydrobiopterine is dependent on the activity of another enzyme, GTPcyclohydrolase-1. The second step is catalyzed by L-dopa decarboxylase in the cytosol. This enzyme is also referred to as aromatic amino acid decarboxylase because it catalyzes the decarboxylation of several endogenous aromatic amino acids. This enzyme so avidly decarboxylates L-dopa that the levels of this amino acid in brain are very low under normal conditions. Metabolism

DA can be metabolized in brain by several enzymatic reactions, which are summarized in Figure 3. The major mammalian enzymes of importance in the metabolic degradation of catecholamines are monoamine oxidase (MAO) and catechol O-methyltransferase (COMT). MAO converts catecholamines to their corresponding aldehydes. This aldehyde intermediate

is rapidly metabolized, usually by oxidation by the enzyme aldehyde dehydrogenase to the corresponding acid. In some circumstances, the aldehyde is reduced by aldehyde reductase. MAO is located on the outer membranes of mitochondria and thus, in brain, is present primarily in nerve terminals and glia. Separate genes encode two isoforms of MAO (types A and B), which can be distinguished by substrate specificity and sensitivity to the irreversible selective inhibitors. In brain, MAO-A is preferentially located in dopaminergic and noradrenergic neurons, whereas MAO-B appears to be the major form present in serotonergic neurons and glia. MAO is a particlebound protein localized largely in the outer membrane of mitochondria. Usually considered to be an interneuronal enzyme, it also occurs in abundance extraneuronally. The second enzyme of importance in catabolism of catecholamines is COMT. This is a relatively nonspecific enzyme that catalyzes the transfer of methyl groups from S-adenosylmethionine to the m-hydroxyl

Dopamine 387

CH2

HO 3

CH2

Dopamine

HO

CH2

CH3O

NH2

MAO

MAO (O2 ) CH2

HO

ALD-D

ALD-D (NAD) HO

ADH CH2

COOH

HO

CH3O

ALD-R

CH2

CH2OH

HO 3,4-Dihydroxyphenylethanol (DOPET)

COOH

Homovanillic acid (HVA) CH3O

COMT

CH2

HO

COMT

3,4-Dihydroxyphenylacetic acid (DOPAC) HO

CHO

HO

COMT

ALD-R (NADP)

CH2

CH3O

CHO

HO ADH (NAD)

NH2

3-Methoxytyramine (3MT)

HO

COMT (SAM)

CH2

CH2

CH2OH

HO 3-Methoxy-4-hydroxyphenylethanol (MOPET)

Figure 3 Main pathways of catabolism of DA. Enzymes involved (shown in italics) are monoamine oxidase (MAO), catechol-Omethyltransferase (COMT), aldehyde dehydrogenase (ALD-D), aldehyde reductase (ALD-R), and alcohol dehydrogenase (ADH). Cofactors for the enzymes are shown in parenthesis; SAM, S-adenosyl-L-methionine. In addition, dopamine (DA) and metabolites are substrates for phenolsulfotransferase; sulfation occurs on the 3-position, indicated on the structure for DA. In brain, DA is metabolized mainly to acidic (DOPAC and HVA) metabolites, with less formation of alcoholic metabolites (DOPET and MOPET); in the periphery, the alcohol metabolites may be metabolized further by alcohol dehydrogenase. In the periphery, glucoronyltransferase is able to form glucoronide conjugates of DA and metabolites.

group of catecholamines and various other catechol compounds. There are two isoforms of COMT, a membrane-bound form and a soluble form. Membrane-bound COMT is the major form found in the central nervous system (CNS), has a higher affinity for catecholamines, and is located principally in neurons. The soluble form has a lower affinity for catecholamines and is the major form expressed in the periphery, but it is present also in CNS glia. The membrane-bound isoform of COMT, having a high affinity for DA, is expressed at neuronal dendritic processes in cortex and striatum, but the expression and activity of COMT are higher in frontal cortex than in striatum. Relevant to this is evidence that in the prefrontal cortex DA transporter (DAT) is rarely expressed within synapses and exerts minimal influence on DA flux in the prefrontal cortex. Thus, COMT activity appears to have a more important function in regulating DA neurotransmission in the frontal cortex than in other regions. In fact, it has been estimated that the flux of DA through the COMT pathway exceeds 60% in the prefrontal cortex but only 15% in the striatum, where DAT is the chief mechanism for terminating the action of DA. The importance of COMT to the actions of DA in the prefrontal cortex is strongly supported by the finding

that compared with wild-type mice, COMT knockout mice have increased DA levels in prefrontal cortex, but not in striatum, and that such mice perform better on prefrontal cortex-dependent behavioral tasks. Another interesting facet to the role of COMT in frontal cortex is the finding that the COMT gene contains a polymorphism (Val158Met) that affects the in vivo activity of the enzyme. Met158 homozygotes have approximately one-third less COMT enzyme activity in prefrontal cortex than Val158 homozygotes. Consistent with its role in modulating prefrontal cortex DA levels, the Val158Met polymorphism is associated with performance on tests of working memory and executive function, which depend on prefrontal cortex function. Thus, the high-activity Val158 allele is linked with relatively poorer performance on such tasks, relative to the Met158 allele, presumably as a result of increased DA metabolism. The COMT polymorphism has been implicated in a number of neuropsychiatric phenotypes. In particular, there is evidence to support an association between COMT allele frequency and the genetic risk of schizophrenia. Working memory is highly dependent on DA function in the prefrontal cortex, and working memory dysfunction is a cardinal feature of schizophrenia.

388 Dopamine

The major DA metabolites found in brain differ, depending on the species under study. In general, however, acidic rather than neutral metabolites predominate. In rodents, the primary metabolites of DA found in the CNS are 3,4-dihydroxyphenylacetic acid (DOPAC) and homovanillic acid (HVA) and a small amount of 3-methoxytramine. DOPAC usually predominates. In addition, a considerable proportion of these metabolites in rodent brain are found in the form of conjugates. In both human and nonhuman primates, however, the major DA metabolite is HVA, and only a small amount is found in the conjugate form. It is well documented that in the rat, short-term fluctuations in DOPAC and HVA concentrations in the striatum, nucleus accumbens, and prefrontal cortex provide a useful index, respectively, of alterations in impulse flow in the nigrostriatal, mesolimbic, and mesocortical DA pathways.

Since the original studies on DA uptake, it has been known that DA is a good substrate for the NE transporter. However, not until relatively recently has the importance of this been appreciated. While in regions, such as striatum and nucleus accumbens, in which DA neurons express a high density of DAT and there is a relatively sparse NE innervation, DA uptake depends primarily on DAT. However, in regions such as the prefrontal cortex, in which DA neurons express low levels of DAT and there is a rich NE innervation, DA uptake depends strongly on the norepinephrine transporter. In fact, it has been suggested that DA can be co-released with norepinephrine by norepinephrine neurons as a result of nonspecific uptake of DA by the norepinephrine transporter and/or because DA is a precursor in the biosynthesis of NE.

Transporters

Much effort has been directed toward elucidating the regulatory mechanisms that control the function of the various DA systems because this may lead to an understanding of the vulnerability of certain DA systems to disease and may also enable drugs to be developed that target specific DA projections. Some aspects of regulation were discussed in relation to the life cycle of DA. Here, other important controls of DA neurotransmission are reviewed.

Presynaptic vesicles or granules present in the nerve terminals are specialized for the uptake and storage of DA, thus protecting it from degradation by the enzyme MAO found in mitochondria present in the nerve terminal. The storage granules contain a vesicular monoamine transporter (VMAT2) that has a pharmacology distinct from the plasma membrane DAT found on the nerve terminals. The Naþ- and Cl dependent DAT actively transfers the DA from the synaptic cleft back into the presynaptic cell, where it can be stored and reused. DA is concentrated to approximately 0.1 M in the storage vesicles, 10–1000 times higher than the level in the cytosol. The tuberoinfundibular, tuberohypophysial, olfactory bulb, mesoprefrontal, and mesoamygdala DA neurons appear to differ from the other DA systems described by the absence, or diminished expression, of a high-affinity DA uptake (transport) mechanism. High-affinity DA uptake serves to recapture released DA and limits the concentration and actions of DA in the synapse. Importantly, neurons equipped with a high density of DATs are especially susceptible to drugs that target this site, such as cocaine and the parkinsonian neurotoxins, 6-hydroxydopamine and MPTP. In fact, the toxicity of MPTP is dependent on the function of MAO-B, DAT, and VMAT2. Once inside the brain, MPTP is metabolized by MAO-B to the actual toxic species, MPPþ, which is a substrate for DAT. Whereas DAT is responsible for transport of the toxin into DA neurons, VMAT2 aids in sequestering the toxin and diminishing its detrimental impact on mitochondria in the neuron. Thus, the DAT-to-VMAT2 ratio should be a determinant of the susceptibility of a given brain region to MPTPinduced toxicity.

Regulation

Firing Pattern

Electrophysiological studies have shown that DA cells fire in two different modes, either single spiking or burst firing, which have a marked impact on the release of DA. Phasic levels of DA are mediated primarily by bursting events at the level of the cell body and lead to a much larger DA release (150–400 nM in striatum) than when these neurons fire in a slow, irregular single spike mode associated with tonic levels (5–10 nM in striatum). Because cells are able to switch between these levels, the transition in activity is a mechanism for altering the impact of DA neurotransmission on receptive cells. Bursting activity of DA neurons is thought to represent a key component of reward circuitry, signaling a reward, or indicating to what extent a reward occurs differently than predicted. Alterations in tonic levels of DA efflux occur on a much slower time scale than changes in phasic levels and enable function on a variety of motor, cognitive, and motivational processes, which are deficient in Parkinson’s disease and which can be restored by DA replacement therapy. Autoreceptors

A receptor sensitive to the transmitter secreted from the cell on which the receptor is located is termed an

Dopamine 389

‘autoreceptor.’ Autoreceptors exist on most portions of DA cells, including the soma, dendrites, and nerve terminals, so that autoreceptors are responsive to both terminal and dendritic DA release. Stimulation of DA autoreceptors in the somatodendritic region slows the firing rate of DA neurons, whereas stimulation of autoreceptors located on DA nerve terminals results in an inhibition of DA synthesis and release. Somatodendritic autoreceptors may also regulate DA release and synthesis by changing impulse flow. Thus, somatodendritic and nerve terminal autoreceptors work in concert to exert feedback regulatory effects on dopaminergic transmission. Three types of autoreceptors can be defined according to their functional effects: impulse-modulating, release-modulating, and synthesis-modulating autoreceptors. In general, all DA autoreceptors can be classified as D2-like DA receptors. The terminals of all midbrain DA neurons examined to date have been found to possess releasemodulating autoreceptors. This is not the case for synthesis- and impulse-modulating autoreceptors. Whereas most DA neurons located in the SN and many DA neurons in the VTA possess somatodendritic impulse-modulating and nerve terminal synthesismodulating autoreceptors, the DA neurons that project to the prefrontal and cingulate cortices and amygdala appear to have either a greatly diminished number, or none, of these receptors. Some of the differences in responsiveness among various midbrain DA neurons may be explained, in part, by differences in their autoreceptor function. Data suggest distinct differences in the autoreceptor regulation of hypothalamic DA neurons. Thus, in contrast to the incertohypothalamic DA neurons, tuberoinfundibular DA neurons appear to lack synthesis-modulating autoreceptors. The activity of tuberoinfundibular DA neurons is regulated by circulating levels of prolactin.

with nontransmitter proteins, including acetylcholinesterase, vesicular glutamate transporter, protein O-carboxymethyltransferase, cytochrome P450 reductase, and a vitamin D-dependent calcium-binding protein. The functional significance of such co-localization is not clear. DA neurons in which neuropeptides are co-localized may be regulated by peptide autoreceptors in a fashion analogous to DA autoreceptors. Furthermore, because it appears that release mechanisms for DA and co-stored peptides may be dissociated under certain conditions (e.g., dependent on the firing pattern and firing frequency of the neuron), co-localized peptides may be part of a cascade of regulatory features that exist in dopaminergic neurons. For example, nerve terminal autoreceptors in the prefrontal cortex have been shown to exert reciprocal effects on DA and neurotensin release. Stimulation of DA autoreceptors diminishes DA release and enhances neurotensin release, whereas blockade of DA receptors augments DA release and diminishes neurotensin release. Such differential release of DA and neurotensin could conceivably allow the prefrontal cortex DA neurons to differentially modulate the physiological activity of cortical postsynaptic follower cells. It appears likely that co-localization of peptides or nontransmitter proteins and DA will distinguish certain subpopulations of DA neurons. For example, cholecystokinin–DA co-localized neurons of the VTA project to the caudal, but not rostral, nucleus accumbens. Such distinctions may have important implications for the regionally specific function of DA in psychiatric and neurological disorders and also for the response of specific DA systems in these pathological conditions to pharmacological treatment.

Co-localized Peptides and Proteins

The different DA systems also vary in the nature of their interaction with other neurotransmitter systems that impinge on the DA neurons. At the cell body level, afferent inputs into the VTA and SN have been most extensively investigated. Many studies have suggested that these cells can be differentially modulated by afferent inputs. DA neurons are also regulated by neurotransmitter interactions that occur at the level of the nerve terminals. However, a detailed description of such interactions is outside the scope this article.

Another interesting intrinsic feature that may contribute to the variable susceptibility of DA neurons to toxicity is the presence of intracellular calcium-binding proteins, such as calretinin and calbindin. These proteins are preferentially expressed in DA neurons that are spared from degeneration in Parkinson’s disease and MPTP-treated animals, possibly by buffering calcium overload and protecting cells against excitotoxic damage. DA is co-localized in some neurons with neuropeptides and nontransmitter proteins. Certain midbrain DA neurons contain the peptide cholecystokinin, whereas another subpopulation of mesencephalic DA neurons contains the peptide neurotensin, and a third group of DA neurons in the VTA contains cholecystokinin, neurotensin, and DA. Similarly, DA is co-localized in certain mesencephalic neurons

Extrinsic

Postsynaptic Signaling There are known to be at least six different forms of the DA receptor. The D1 class of DA receptor has been divided into D1 and D5 receptor subtypes, and

390 Dopamine

the D2 class comprises D2short, D2long, D3, and D4 receptor subtypes. The pharmacological profile and regional distribution in brain of each subtype are different, with the exception of D2short and D2long, which appear to be identical in these respects. Although these different receptors are reviewed elsewhere, certain interesting or novel aspects of DA signaling are discussed here. Neuromodulation

It is now generally accepted that DA is not a classical, fast ionotropic neurotransmitter like glutamate acting at a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) or N-methyl-D-aspartate (NMDA) receptors or GABA acting at GABA-A receptors that elicit excitatory or inhibitory postsynaptic potentials that last a few milliseconds. All the DA receptors are G-protein-coupled receptors that are slow, metabotropic receptors with response times exceeding 100 ms and function to modulate other receptor systems and/or ion channels. Thus, DA is thought of as a neuromodulator and cannot be classified as either an excitatory or an inhibitory neurotransmitter, which may explain why DA does not yield identical effects under all experimental conditions. Volume Transmission

Neurons may operate via two modes of communication, the synaptic mode and the mode of communication involving short- to long-distance diffusion of chemical signals in the extracellular fluid pathways (volume transmission). The process of volume transmission has been defined as a functionally significant association of release and receptor sites via extrasynaptic diffusion. It is now clear that in certain regions DA functions by volume transmission. The retina provides the clearest example of DA-mediated volume transmission. In the SN, DA receptors are located extrasynaptically, indicating that volume transmission is the predominant form of intercellular communication mediated by DA in this region. In the striatum, volume transmission appears not to play a major role in DA transmission since most DA terminals establish synaptic contacts, and the high density of DA terminals express a high density of DAT. Volume transmission may be more important in maintaining dopaminergic signaling in the striatum after partial loss of nigrostriatal DA neurons, as occurs in Parkinson’s disease. With the degeneration of dopaminergic terminals and consequent loss of DAT, DA that is released from the remaining neurons may diffuse in the striatal extracellular fluid to reach supersensitive high-affinity DA receptors.

Somatodendritic Release

There is good evidence that DA can be synthesized and released from dendrites and cell bodies in the SN. The release is vesicular and mediated by exocytosis, as it is in axon terminals. Somatodendritic release is calcium and depolarization dependent in a variety of paradigms. In dendrites, DA appears to be stored both in classical vesicles and in saccules of smooth endoplasmic reticulum, consistent with the expression of VMAT2 in both of these organelles. The function of DA released in the SN in this manner is probably to interact with autoreceptors to control the firing rate of DA neurons, in addition to signaling via DA receptors located on non-DA neurons in the region. Long-Term Potentiation and Long-Term Depression

DA has a critical role in two distinct forms of plasticity, called long-term depression (LTD) and long-term potentiation (LTP). In regions such as the striatum and prefrontal cortex where DA modulates glutamate signaling, combined stimulation of both glutamate and DA receptors is able to induce either enhancement or weakening of synaptic transmission, depending on the parameters of the stimulation and the receptor subtypes that are activated. These mechanisms may lead to more permanent changes, such as the construction of new connections (synapses). Such plastic changes are thought to have an important role in learning and memory (hippocampus and prefrontal cortex) and acquisition of complex motor skills (striatum). Pyramidal Cell Dendritic Spines

DA also plays a critical role in the modulation of pyramidal cell dendritic spine synapses and spine density. Several studies have demonstrated that the loss of DA in the striatum results in decreased dendritic length and spine density. Employing multiphoton imaging, selective elimination of glutamatergic synapses on striatopallidal neurons has been observed in Parkinson disease models. It has clearly been shown that DA depletion leads to a rapid and profound loss of spines and glutamatergic synapses on striatopallidal medium spiny neurons but not on neighboring striatonigral medium spiny neurons. This loss of connectivity is triggered by an interesting mechanism – dysregulation of intraspine Cav1.3 L-type Ca2þ channels. The disconnection of striatopallidal neurons from motor command structures is likely to be a key step in the emergence of pathological activity that is responsible for symptoms in Parkinson’s disease.

Dopamine 391

DA modulation of pyramidal cell dendritic spines and synapses has been extended to the prefrontal cortex, where work has shown that selective lesion of the DA innervation of the prefrontal cortex caused a decrease in dendritic spine density in pyramidal cells in the prefrontal cortex. It is noteworthy that atypical antipsychotic drugs which cause a large increase in synaptic DA reverse the DA depletioninduced dendritic spine loss in prefrontal cortical pyramidal neurons. In schizophrenia, the prefrontal cortex DA innervation is compromised and postmortem studies have reported decreased dendritic spine density. Taken together, these data suggest that disruption of dopaminergic signaling may be the proximate cause of dendritic remodeling in the prefrontal cortex in schizophrenia. Excessive DA input may have the opposite effect because chronic cocaine or amphetamine use has been shown to increase spine density in nucleus accumbens and prefrontal cortex and increase synapse number in the prefrontal cortex. See also: Dopamine Receptors and Antipsychotic Drugs in Health and Disease; Dopamine: Cellular Actions.

Further Reading Akil M, Kolachana BS, Rothmond DA, et al. (2003) CatecholO-methyltransferase genotype and dopamine regulation in the human brain. Journal of Neuroscience 23: 2008–2013. Cooper JR, Bloom FE, and Roth RH (2003) The Biochemical Basis of Neuropharmacology, 8th edn. New York: Oxford University Press. Day M, Wang Z, Ding J, et al. (2006) Selective elimination of glutamatergic synapses on striatopallidal neurons in Parkinson disease models. Nature Neuroscience 9: 251–259.

Dunnett SB, Bentivoglio M, Bjorklund A, and Hokfelt T (2005) Handbook of Chemical Neuroanatomy, vol. 21. Elsevier: Amsterdam. Eisenhofer G, Kopin IJ, and Goldstein DS (2004) Catecholamine metabolism: A contemporary view with implications for physiology and medicine. Pharmacological Reviews 56: 331–349. Elsworth JD and Roth RH (1997) Dopamine autoreceptor pharmacology and function: Recent insights. In: Neve KA and Neve RL (eds.) The Dopamine Receptors, pp. 223–265. Humana: Totowa, NJ. Grace AA (2002) Dopamine. In: Davis KL, Charney D, Coyle JT, and Nemeroff C (eds.) Neuropsychopharmacology: The Fifth Generation of Progress, pp. 120–132. Philadelphia: Lippincott Williams & Wilkins. Jay TM (2003) Dopamine: A potential substrate for synaptic plasticity and memory mechanisms. Progress in Neurobiology 69: 375–390. Jentsch JD, Roth RH, and Taylor JR (2000) Role for dopamine in the behavioral functions of the prefrontal corticostriatal system: Implications for mental disorders and psychotropic drug action. Progress in Brain Research 126: 433–453. Joel D and Weiner I (2000) The connections of the dopaminergic system with the striatum in rats and primates: An analysis with respect to the functional and compartmental organization of the striatum. Neuroscience 96: 451–474. Lookingland KJ and Moore KE (2005) Functional neuroanatomy of hypothalamic dopaminergic neuroendocrine systems. Handbook of Chemical Neuroanatomy 21: 435–523. Miller GW, Gainetdinov RR, Levey AI, and Caron MG (1999) Dopamine transporters and neuronal injury. Trends in Pharmacological Sciences 20: 424–429. Rice ME (2000) Distinct regional differences in dopaminemediated volume transmission. Progress in Brain Research 125: 277–290. Seamans JK and Yang CR (2004) The principal features and mechanisms of dopamine modulation in the prefrontal cortex. Progress in Neurobiology 74: 1–58. Tzschentke TM (2001) Pharmacology and behavioral pharmacology of the mesocortical dopamine system. Progress in Neurobiology 63: 241–320.

Dopamine Receptors and Antipsychotic Drugs in Health and Disease P Seeman, University of Toronto,Toronto, ON, Canada ã 2009 Elsevier Ltd. All rights reserved.

Introduction The discovery of dopamine receptors is intertwined with the discovery and development of antipsychotic drugs. The research in this area started with the development of antihistamines after World War II, particularly with H Laborit using such compounds to enhance surgical analgesia. In patients receiving these medications, Laborit noticed a ‘euphoric quietude,’ and that the patients were ‘‘calm and somnolent, with a relaxed and detached expression.’’ Of this series of Rhoˆne-Poulenc compounds, RP4560, now known as chlorpromazine, was the most potent. Chlorpromazine was tested by many French physicians for use in various medical illnesses. Although Sigwald and Bouttier were the first to use it as the sole medication for a psychotic patient, their work was not reported until 1953, after a 1952 report by Delay and colleagues that chlorpromazine alleviated hallucinations and stopped internal ‘voices’ in eight patients. A significant aspect of the action of chlorpromazine was that it was effective within 3 days. This rapid improvement, especially during the first week of antipsychotic treatment, has been observed in many studies and summarized in reviews by S Kapur and O Agid. The clinically successful action of chlorpromazine stimulated the search to identify chlorpromazine’s mode of action. The assumption then, as now, was that the discovery of such a mode of action would open the avenue to uncovering the biochemical cause of psychosis and possibly of schizophrenia.

Early Days: Before Discovery of Dopamine Receptors In searching for the mechanism of chlorpromazine action in the 1960s and 1970s, many types of electrophysiological and biochemical experiments were done. Because high doses of chlorpromazine and other antipsychotics (or ‘neuroleptics,’ as they were then called) also elicited parkinsonism as a side effect, the basic science quickly focused on the action of antipsychotics on dopamine pathways in the brain. The rationale for examining brain dopamine regions was based on the finding by H Ehringer and O Horniekiewicz that the parkinsonism of Parkinson’s disease was associated

392

with a massive loss of brain dopamine. It was felt, therefore, that the unwanted side effect of chlorpromazine-induced parkinsonism, as well as the antipsychotic action itself, might arise by antipsychotics interfering with dopamine neurotransmission. The working assumption was that if the brain targets for antipsychotics could be found, then perhaps it could be determined whether these sites were overactive or underactive in psychosis or schizophrenia. A variety of mechanisms were explored for the mode of action of chlorpromazine, including its action on mitochondrial enzymes, sodium–potassiumATPase, and related enzymes, and its membranestabilizing action, such as its strong potency to inhibit membrane action potentials and to stabilize cellular and subcellular membranes from releasing their contents. It also became clear in 1963 that all antipsychotics were surface active, readily explaining their hydrophobic affinity for membranes. Some of these nonreceptor-related findings, such as the surface activities of the antipsychotics, showed an astonishingly excellent correlation with clinical antipsychotic potencies.

Therapeutic Concentrations of Antipsychotics All of the early experiments in the 1960s revealed that the in vitro active concentrations of the antipsychotics were generally between 20 and 1000 nM. These concentrations, however, were found in 1971 to be far in excess of the nanomolar concentrations (e.g., 1–2 nM for haloperidol) that exist in the spinal fluid in patients being successfully treated with these medications. In Vivo Experiments

In parallel with the in vitro experiments, there were many in vivo experiments by F Bloom, by G Aghajanian, and by B Bunney, showing that dopamine agonists can excite or inhibit neurons in the nigrostriatal dopamine pathway. Moreover, other workers (WD Heiss, J Hoyer) showed that direct application of dopamine on neurons also stimulated or inhibited snail neurons, and that haloperidol or fluphenazine could block these actions (H Struyker Boudier). These studies provided evidence for the existence of distinct dopamine receptors on neurons. Additional work in vivo showed that chlorpromazine and haloperidol increased the production of normetanephrine and methoxytyramine, metabolites of epinephrine and dopamine, respectively. To explain the increased production of these metabolites,

Dopamine Receptors and Antipsychotic Drugs in Health and Disease 393

Carlsson and Lindqvist suggested that ‘‘the most likely [mechanism] appears to be that chlorpromazine and haloperidol block monoaminergic receptors in brain; as is well known, they block the effects of accumulated 5-hydroxytryptamine.. . .’’ In other words, they proposed that antipsychotics might block all three types of receptors for noradrenaline, dopamine, and serotonin, but they did not identify which receptor was selectively blocked or how to identify or test any of these receptors directly in vitro. This study in 1963 by Carlsson and Lindqvist is often mistakenly cited as discovering ‘the dopamine receptor’ and that antipsychotics are selectively acting on this receptor. However, N-E Ande´n, a student of A Carlsson, had a different view, and proposed that ‘‘chlorpromazine and haloperidol delay the elimination of the (metabolites).. . .’’ Moreover, 7 years later Ande´n reported that antipsychotics increased the turnover of both dopamine and noradrenaline, but he could not show that the antipsychotics were selective in blocking dopamine; for example, chlorpromazine enhanced the turnover of noradrenaline and dopamine equally. Therefore, it remained for in vitro radio-receptor assays to detect the dopamine receptor directly and to demonstrate antipsychotic selectivity for the dopamine receptor. The Dopamine D1 Receptor

With the advent of assays for adenylate cyclase in the 1960s, J Kebabian found that dopamine stimulated adenylate cyclase in the superior cervical ganglion. This receptor was later named the dopamine D1 receptor, selectively labeled by [3H]SCH23390, and subsequently cloned by three research groups in 1990. The dissociation constants at D1 for dopamine agonists and antagonists of medical therapeutic interest are given in Tables 1 and 2. There is no correlation between the antipsychotic clinical doses and the dissociation constants of the antipsychotic antagonists at D1, as illustrated in Figure 1. These data suggested that D1 was not the major or common target for antipsychotics, in addition to the fact that the antipsychotic molarities at D1 are between 10 and 10 000 nM, far in excess of the therapeutic concentrations in the spinal fluid of treated patients. In addition to the lack of targeting D1 receptors by clinical doses of the common antipsychotics, D1selective compounds have not been found to be effective as antipsychotics (Figure 2). Discovery of the Antipsychotic Dopamine Receptor, or the Dopamine D2 Receptor

In 1974 and 1975, in order to detect and discover the dopamine receptors on which the antipsychotics

presumably acted, it was essential to label a receptor with a ligand, such as radioactive haloperidol, having an affinity (or dissociation constant) of 1 nM, because this was the haloperidol therapeutic concentration found in the spinal fluid or plasma water of treated patients. For this to occur, the specific activity of [3H]haloperidol would have to be at least 10 Ci mmol1. Although the [3H]haloperidol donated in 1971 by Janssen Pharmaceutica (J Heykants, J Brugmans) had a low specific activity of 32–71 mCi mmol1, I R E Belgique (M Winand) custom synthesized [3H]haloperidol at even lower specific activity (10.5 Ci mmol1) for Seeman’s laboratory by June 1974. Specific binding of this new [3H]haloperidol to brain striatal tissue was readily detected in 1975, and the following concentrations of compounds were found to inhibit the binding of [3H]haloperidol by 50%: 2 nM for haloperidol, 20 nM for chlorpromazine, 3 nM for (þ)butaclamol, and 10 000 nM for (–) butaclamol. The stereoselective action of butaclamol and the good correlation between the IC50% values and the clinical doses, as shown in Figure 3, indicated that the ‘antipsychotic receptor’ had finally been discovered. Equally important, of the endogenous compounds tested, dopamine was the most potent in inhibiting the binding of [3H]haloperidol, indicating that the antipsychotic receptor was actually a dopamine receptor (see Table 1). Using sequences related to the b-adrenoceptor, the antipsychotic/dopamine receptor, now named the dopamine D2 receptor, was finally cloned in 1988; its amino acid sequence is shown in Figure 4. Using the cloned D2 receptor, values for the dissociation constants for agonists and antagonists can be determined (Table 1). It is now known that therapeutic levels of antipsychotic drugs occupy 60–80% of brain D2 receptors, as shown by L Farde and colleagues. In fact, using the antagonist dissociation constants in Table 1, the concentrations necessary to occupy 75% of D2 receptors can be calculated and are essentially identical to the free molarities of the antipsychotics found in either the plasma water or the spinal fluid of treated patients, as shown in Figure 5. Because radiolabeled raclopride is less tightly bound to D2 receptors than radiolabeled spiperone, which in turn is less tightly bound to D2 receptors than radiolabeled nemonapride, the dissociation constant for any antagonist at D2 receptors is lowest when using radiolabeled raclopride, but rises when using radiolabeled spiperone, or rises higher still when using radiolabeled nemonapride (see Table 1). This dependence of the dissociation constant on the ligand is also seen in positron emission tomography studies, where it has been found, for example,

D1

D2

D3

D4

D5

Tissue [Ref.]:

Rat striatum or human D1 [f]

Human D2 Long clone [b]

Human D3 clone [b]

Human D4 clone

Human D5 clone

Radioligand:

[3H]SCH 23390

[3H]SCH 23390

[3H]SCH 23390

[3H]spiperone

[3H]SCH 23390

K High (nM )

K50 (nM )

K High (nM )

KLow (nM )

K High (nM )

K50 (nM )

K High (nM )

K50 (nM )

K50 (nM )

Neurotransmitters Dopamine

10.1

2340 [g]

25 I [a] 22 S n.i. n.i. n.i.

228 [g]

700 1700 9700 [g]

25 D 3.9 I [a] n.i. n.i. n.i.

148 [f]

No high No high No high

1400 D 729 R n.i. 805 D 4300 D

28 [f]

Noradrenaline Adrenaline Serotonin

2.4 D 1.4 R;1.3 A 9.8 D No high No high

n.i. n.i. n.i.

2200 [f] n.i. 4180 [f]

12 000 [g] n.i. 3000 [g]

1.2 39 No high No high 3.5 No high No high No high

1816 [g] n.i. 5400 672 [g] 360 [e] 330 [e] 2600 8200

0.14 D 0.4 D 0.5 D 0.8 D 0.75 D 1.3 D 1.7 D 12 D

54 D rat 62 D 1400 D 20 D 140 D 19 I [a] 202 R[d] 2600 D

0.25 D 0.6 D 3.2 D 1.3 D 2.6 D 0.9 D 0.45 R[d] 9D

0.8 n.i. n.i. 7.4 I [a] 73 I [a] 2.3 I [a] 256R [d];50 I [e] 11 I [e]

n.i. n.i. n.i. n.i. 4.1 [f] n.i. n.i. n.i.

4f 47 f n.i. 250 [e] 5 [e] 62 [e] 31 [e] 120 [e]

1136 [g] n.i. n.i. 720 [e] 13 [e] 39 [e] >1000 [g] >1000 [e]

1.1 [c] 10 [c]

380 [c] 4700 [c]

150 S[c] 1. 1 R [d]

8800 S[c] 42 R [d]

No high 0.06 R [d]

5000 [a] 0.2 R [d]

n.i. n.i.

1800 [f] 650 [a]

100 [g] n.i.

Dopamine agonists N-Propyl-norapomorphine-R-() (þ)PHNO ()Quinagolide.HCl, or CV205 502 Bromocriptine.base Apomorphine-R-().HCl Pergolide mesylate [LY 127,809] (–)Quinpirole, or ()-LY 171,555.HCl Pramipexole monohydrate SKF 38393 (þ)-7-OH-dipropylaminotetralin

[a], Sokoloff et al. (1992); [b], Seeman et al. (2005); [c], Seeman and Niznik (1988); [d], Malmberg and Mohell (1995); [e], Millan et al. (2002); [f] P. Seeman; [g], Sunahara et al. (1991); K50 (dissociation constant), C50% (conc. to inhibit 50% of binding)/(1 þ C*/Kd), where C* – concentration of radioligand in competition with agonist, and Kd – dissociation constant obtained by saturation with radioligand (Scatchard analysis); A, [3H]dopamine (Kd ¼ 1.3 nM); D, [3H]domperidone (Ref. b or P. Seeman, unpublished; Kd ¼ 0.43 nM); I, [125I]Iodosulpride (Kd ¼ 0.6 nM); R, [3H]raclopride (Kd ¼ 1.9 nM); S, [3H]spiperone (Kd ¼ 65 pM); rat, rat striatum; no high, no high-affinity state recognized by competing compound; n.i., no information available.

394 Dopamine Receptors and Antipsychotic Drugs in Health and Disease

Table 1 Agonist potencies at dopamine receptors (in 120 mM NaCl)

Dopamine Receptors and Antipsychotic Drugs in Health and Disease 395 Table 2 K values (dissociation constants) Human clone:

M1 nM

D1 nM

[3H]ligand used:

QNB

Sch.

[3H]Amisulpride Amoxapine Aripiprazole Butaclamol-(þ) Chlorpromazine [3H] Chlorpromazine Clozapine [3H]Clozapine Clozapine-iso Cyproheptadine Droperidol Epidepride Flupentixol-cis Flupentixol-trans Fluphenazine Haloperidol [3H]Haloperidol Iloperidone (HP873) Loxapine Loxapine-iso Melperone Molindone Norclozapine Olanzapine [3H]Olanzapine Perphenazine Pimozide Prochlorperazine Quetiapine [3H]Quetiapine Raclopride [3H]Raclopride Remoxipride Risperidone Risperidone-9-OH Sertindole [3H]Sertindole Spiperone [3H]Spiperone Sulpiride-S Trifluperazine Ziprasidone

D2 nM

1.7 49

3.4

16.5

— — — — 0.77

9.5 14.4

90 120

1.8

— 51 — — — — —

794

2.6 55

— —

109

5.6

—

117 203

18 16 148 1558 73 9.2

— — — — — —

0.4

17 2.1

2.7 4.5 470 135

>10 mM >10 mM 400

7.7 290

4900 42

— — — — 104 — 1.9 — —

22

3 1.2 — 0.065

265

2.9 9

— — —

D2 nM

D2 nM

D3 nM

Raclo.

Spip.

Raclo.

1.8 21 1.8 0.14 1.2 —

4.6 56

1.4

—

—

75 — 15 24 0.54 0.036 0.38 151 0.55 0.74 — 5.4

180 — 60

190 — 21

— 10

— 20

9.2 22 152 4.9 180 7.4 — 0.27 1.4 1.7 140 — 1.6 — 67 1.09 1.6 1.9 — 0.018 — 9.9 1.4 2.7

22.7 86 375 15 300 21 — 0.47 0.95 4 680 — 7.1 — 800 4

7.2 18 315 44

2.3 0.06 0.7

0.7

1.2 2.7

0.17 8.8

— — 1.2 — 2 — — — — — — — 0.85 —

5.5 70 9.6 — 22 — 21 2 15 30 2 — 9.6

14 — 0.23

— — — — 1.6 —

240 — 2.9 1.6 960 3.5

— — — — — — —

89 2000 — 2400 — 2400 4.4

3 —

0.06 8 3.8 6

Spip.

8 11 720 3900 120 15 — 32

6.5

—

D4 nM

— —

0.9 4.6

—

D4 nM

0.32 10 0.7 1.5

— 0.85 — 0.086 — — —

11 — — 1000 39 8

Blank cells indicate ‘not done’. Sch., Schering 23390; Raclo., Raclopride; Spip., spiperone.

when monitoring patients with [11C]raclopride, that therapeutic doses of clozapine occupy 50% of D2 receptors but that there is much less occupancy of D2 receptors when using [11C]methylspiperone or [18F]fluorethylspiperone, both of which bind more tightly to D2 receptors than raclopride (see Table 2). The correlations in Figures 3 and 5 remain a cornerstone of the dopamine hypothesis of psychosis or schizophrenia, and the dopamine hypothesis is still

the major contender for an explanatory theory of schizophrenia etiology. Two Classes of Dopamine Receptors

The D1 site and the [3H]haloperidol/dopamine receptor binding site were soon considered as distinct, because B Roufogalis found that the sulpiride antipsychotic did not block dopamine-stimulated adenylate cyclase. Two general classes of dopamine receptors were recognized,

396 Dopamine Receptors and Antipsychotic Drugs in Health and Disease

10−5

D1

K(mol l−1) on 3H-SCH23390 binding

Clebopride

Sulpiride Molindone

10−6

Chlorpromazine

Spiperone Clozapine 10−7

Haloperidol Thioridazine Fluphenazine Trifluperazine Flupenthixol

10−8

0.1

1

10 100 Range and average clinical dose for controlling schizophrenia (mg d−1)

1000

Figure 1 There is no correlation between the clinical antipsychotic doses and the antipsychotic dissociation constants (or concentrations) that inhibit the binding of a D1 ligand ([3H]SCH23390) at dopamine D1 receptors in homogenized striatal tissue. The high concentrations inhibiting the D1 receptor are far higher than those found clinically in the plasma water or spinal fluid. Adapted from Seeman P (1987) Dopamine receptors and the dopamine hypothesis of schizophrenia. Synapse 1: 133–152, with permission from John Wiley & Sons Inc.

Figure 2 Amino acid sequence of the dopamine D1 receptor. The two hydroxyls of dopamine are presumed to be associated with the two serine residues, while the tertiary nitrogen atom is presumed to be associated with the aspartic acid residue (D in transmembrane 3).

Dopamine Receptors and Antipsychotic Drugs in Health and Disease 397 10−7 Promazine Chlorpromazine Trazodone Clozapine Thioridazine Molindone Prochlorperazine Moperone Trifluperazine

IC50 (mol l−1)

10−8

Thiothixene Haloperidol Droperidol Fluphenazine Pimozide Trifluperidol

10−9

Benperidol 10−10 Spiroperidol

0.1 1 10 100 1000 Range and average clinical dose for controlling schizophrenia (mg d−1) Figure 3 The clinical antipsychotic doses correlate with the concentrations that inhibit by 50% the specific binding of [3H]haloperidol in homogenized caudate nucleus tissue (calf). These concentrations are similar to those found in the plasma water or spinal fluid in patients treated with antipsychotic drugs. Adapted from Seeman P, Lee T, Chau-Wong M, et al. (1976) Antipsychotic drug doses and neuroleptic/ dopamine receptors. Nature (London) 261: 717–719, with permission from Nature.

Figure 4 Amino acid sequence of the dopamine D2 receptor. The variants or polymorphisms of D2 include an alanine instead of a valine at position 96 (1% of population), an insertion of a 29-amino-acid polypeptide (D2Long) at the position shown by the lower left arrow, a serine instead of a proline at position 310, and a cysteine instead of a serine at position 311. Dopamine is presumed to be associated with the two serines (S) in transmembrane 5 and the aspartic acid (D) in transmembrane 3 (see Figure 2).

Line for identical values

S-Sulpiride Molindone Olanzapine

cu

pa

nc

y

100

Clozapine Remoxipride

%

oc

10 75

Concentration needed to occupy 75% of D2 (nM)

398 Dopamine Receptors and Antipsychotic Drugs in Health and Disease

1

Chlorpromazine Raclopride Thioridazine Haloperidol cis-Flupentixol Perphenazine

1 10 100 1000 Therapeutic free neuroleptic (nM) in spinal fluid or plasma water Figure 5 The therapeutic antipsychotic concentrations in the spinal fluid or in the plasma water in treated patients are essentially identical to the antipsychotic concentrations that occupy approximately 75% of the D2 receptors in vitro. The concentrations in the plasma water were obtained by correcting for the amount bound to the plasma proteins. The concentrations to occupy 75% of D2 were calculated as being three times higher than the dissociation constant at D2. Adapted from Seeman P (2002) Atypical antipsychotics: Mechanism of action. Canadian Journal of Psychiatry 47: 27–38.

D1 site (= dopamine-sensitive adyenylate cyclase) Dopamine: ∼3000 nM Spiperone: ∼2000 nM

D2 receptor

D3 site

Dopamine:∼5000 nM Spiperone:∼0.3 nM

Dopamine: 3 nM Spiperone: ∼1500 nM

D4 site Dopamine: 3 nM Haloperidol: ∼1 nM

Figure 6 Early version of dopamine receptors before clones of receptors became available. The D1 receptor (dopamine-stimulated adenylate cyclase) was stimulated by 3000 nM dopamine and inhibited by high concentrations of butyrophenones such as 2000 nM spiperone. The D2 receptor, or the [3H]haloperidol binding site, was highly sensitive to spiperone, but required 5000 nM dopamine to inhibit adenylate cyclase. The D3 site was a site labeled by [3H]dopamine and, therefore, very sensitive to dopamine at 3 nM, but required very high concentrations of antipsychotics to be inhibited. The D4 site was defined as being sensitive to both the agonists and the antipsychotic antagonists. While these definitions for the D3 and D4 sites are no longer used, the D1 and D2 properties are still valid for the cloned D1 and D2 receptors, with the additional point being that both D1 and D2 have high- and low-affinity states.

therefore, coupled or uncoupled to adenylate cyclase. These two classes were named D1 and D2 by J Kebabian and D Calne. Nomenclature of Dopamine Receptors

The data for the pattern of binding of [3H]haloperidol identifying the antipsychotic/dopamine D2 receptor

were very different from those for the pattern of [3H] dopamine binding described by studies in the mid1970s. For example, the binding of [3H]haloperidol was inhibited by 5000 nM dopamine, while that of [3H]dopamine was inhibited by 3 nM dopamine, as summarized in Figure 6. For several years, this latter [3H]dopamine binding site was termed the ‘D3 site,’ a

Dopamine Receptors and Antipsychotic Drugs in Health and Disease 399

term which is not to be confused with the later discovery of the dopamine D3 receptor.

Dopamine D2 Receptor Variants As noted in Figure 4, there are several variants of D2, the most important of which are the short form and the long form of D2, the latter having an additional 29 amino acids. There is also a D2Longer form where a dipeptide, valine–glutamine, is inserted into the intracellular loop, as shown in Figure 7. There are at least three polymorphisms in D2 (Figure 4): alanine replaces valine at position 96 in about 0.8–1% of some populations, serine replaces proline at position 310 in 0.4% of people, and cysteine replaces serine at position 311 in approximately 3–4% of the population. The variants at 310 and 311 are markedly less effective in inhibiting the synthesis of cyclic AMP than is the more common form of D2. Silent polymorphisms have also been found in the DNA code for D2, but there is no change in the amino acid (leucine at 141, histidine at 313, and proline at 319). There are also polymorphisms in the noncoding regions of D2, including an A-to-G at position –241, a C insert at position –141, an A-to-G before transmembrane 1 (Taq1B polymorphism), an A-to-G in the intron

within transmembrane 2 (Taq1D polymorphism), an A-to-C in the intron before transmembrane 4, a G-to-A in the intron before transmembrane 6, and an A-to-G situated  10 kb beyond transmembrane 7 (Taq1A polymorphism). A further mutation in D2 occurs in individuals with hereditary autosomal dominant myoclonus dystonia, with valine replaced by isoleucine at position 154 at the beginning of the fourth transmembrane segment of D2 (see Figures 4 and 7).

D2 Function and Distribution A wide variety of psychomotor functions, including biochemical, physiological, and pathological, have been attributed to D2. Specifically, D2 inhibits action potentials by eliciting a prolonged ‘inhibitory postsynaptic potential’ (IPSP), inhibits adenylate cyclase, and inhibits the entry of calcium ions into cells, thereby inhibiting many aspects of stimulus–response coupling in a variety of neurons and cells. D2 receptors are located on presynaptic terminals and, therefore, can readily inhibit the release of dopamine. These presynaptic receptors appear to be predominantly D2Short, while D2Long receptors are mostly postsynaptically located on dendritic spines. The genetic absence of D2 receptors leads to animals that are akinetic and Parkinson-like.

Figure 7 Amino acid sequence of D2Longer. Compared to D2Short and D2Long (Figure 4), D2Longer has an extra valine–glutamine dipeptide (as indicated by the arrow) that is usually spliced out in D2Short and D2Long.

400 Dopamine Receptors and Antipsychotic Drugs in Health and Disease

In alcoholics, it has been reported that the prevalence of the Taq1A polymorphism is about twofold higher than in control subjects. In schizophrenia, it has been found in 27 studies, comprising 3707 patients and 5363 controls, that the serine311cysteine polymorphism was significantly associated with schizophrenia. Furthermore, the number of D2 receptors in the caudate-putamen is elevated in schizophrenia, as measured in vivo (Corripio) or in postmortem samples in vitro (Figure 8). Although Figure 8 shows that the density of D2 receptors in postmortem human schizophrenia tissues is elevated, some of this elevation may have resulted from the antipsychotic administered during the lifetime of the patient. The postmortem tissues from half of the patients with schizophrenia revealed elevated densities of [3H]spiperone-labeled D2 receptors in the caudate-putamen tissue. The other half of the postmortem schizophrenia tissues were normal in D2 density, even though most of the patients were known to have been treated with antipsychotics during their lifetime. Such findings have long been controversial, because the D2 density is not elevated in schizophrenia when using [11C]raclopride. It should be noted, however, that the number of D2 receptors is significantly Control striata 12.9 + − 0.2 pmol g−1 20–88 years

Schizophrenia

elevated in healthy identical co-twins of individuals with schizophrenia, suggesting that an elevation of D2 may be a necessary but not sufficient requirement for schizophrenia. The distribution of D2 receptors within the various brain regions is reflected in the gene expression pattern of D2, as shown in Figure 9. An additional unique feature is that D2 receptors are organized in bands in the normal human temporal cortex, but these bands are not found in brains of patients with Alzheimer’s disease. Finally, the density of D2 is not constant over one’s lifetime, but slowly falls by about 2% per decade, as shown in the postmortem human tissues in Figure 10. There is a sharp transient rise during ages 1–3, but this is followed by a gradual pruning of neurons with D2 receptors.

The Dopamine D3 Receptor Using methods similar to those used for cloning the D2 receptor, the D3 receptor was cloned in 1990; its sequence is shown in Figure 11. D3 has several polymorphisms, including a serine replacing glycine at position 9 in 28% of the population (see Figure 11). In addition, there are nonfunctional forms of D3, where the amino acid chain stops after transmembrane 2, after transmembrane 3, or before transmembrane 6, the latter being found in, but not diagnostic for, Alzheimer’s disease and schizophrenia. Although BP897 is a partial agonist at D3, with a selectivity for D3 of about 100-fold higher than that for D2, this drug (at 10 mg day1) did not appear effective against schizophrenia symptoms. Other drugs moderately selective for D3, such as S33138 and A437,203, are currently being tested in schizophrenia patients. The highly D3-selective drug, FAUC 365, has not yet been tested in this disease. It is possible that the D3-selective compounds may be helpful in treating drug abuse.

The Dopamine D4 Receptor

0

4

8

12 16 20 24 28 D2 density (pmol g−1)

32

36

40

Figure 8 Bimodal pattern of D2 receptors in postmortem striata from individuals who had schizophrenia during life. The control individuals had died of nonneurological disorders. The bimodal pattern was found in the schizophrenia tissues, regardless of whether the individuals had received antipsychotics or not. Each square indicates a different postmortem human brain (caudate nucleus or putamen regions). Adapted from Seeman P and Niznik HB (1990) Dopamine receptors and transporters in Parkinson’s disease and schizophrenia. FASEB Journal 4: 2737–2744, with permission from Federation of American Societies for Experimental Biology.

The dopamine D4 receptor was cloned in 1999; its sequence is shown in Figure 12. The D4 receptor probably has more polymorphisms than any other protein in the body. For example, the intracellular loop contains repeat forms of a 16-amino acid polypeptide, the number of repeats varying from person to person. Most people have four such repeats, but up to ten repeats are known. Moreover, the precise sequence within each repeat usually varies from person to person, with at least 20 different types of repeat units known, thereby resulting in a massive number of polymorphisms in the human population. An unusual polymorphism in D4 occurs at position 194, where glycine replaces valine in 13% of Africans

Dopamine Receptors and Antipsychotic Drugs in Health and Disease 401

Figure 9 Anatomical location for gene expression of dopamine receptor genes in human brain. abbreviations: CX, cerebral cortex; L, lateral ventricles; C, caudate nucleus; P, putamen; G, globus pallidus; AC, nucleus accumbens; O, olfactory tubercle; H, hypothalamus; AM, amygdala; Hipp, hippocampus; VTA, ventral tegmental area; SN, substantia nigra; ICG, Islands of Calleja.

and Caribbeans (Figure 13), but not in Caucasians. This mutant form of D4 markedly reduces its sensitivity to dopamine. One young man in the Caribbean was found to be homozygous for this V194G polymorphism, but no medical abnormalities were found. Another D4 polymorphism occurs when the GASA sequence is missing at position 21. Interestingly, clozapine has a higher affinity at D4 than at D2, as shown in Table 2. Nevertheless, despite clozapine’s selectivity for D4, clozapine occupies the necessary 60–70% of brain D2 receptors at clinical doses (400 mg day1), compatible with the idea that D2 is the therapeutic target for clozapine, as with all the

other antipsychotics. It may be noted that isoclozapine causes catalepsy, in contrast to clozapine, which does not elicit catalepsy. Both drugs have dentical affinity for D4, but isoclozapine has higher affinity for D2 (see Table 2), and, therefore, causes catalepsy. Although the gene expression of D4 was found to be elevated in the frontal cortex of schizophrenia tissues, selective D4 antagonists, such as sonepiprazole and L-745,870, did not have any antipsychotic action. Some evidence suggests that the longer forms of D4, such as D4.7, with seven repeats, or D4.9, are found in hyperactive individuals or in those persons who take unusual risks, but this is controversial.

402 Dopamine Receptors and Antipsychotic Drugs in Health and Disease

D2 control striata (m and f) 2.2% loss per decade (p < 0.002)

D2 density, pmol g−1

30

Upper limit for adults 20

10

0

0 1 2 3 4 5 10

20

30 40 50 Age, years

60

70

80

90 100

Figure 10 Postmortem human D2 densities in the striatum. After an initial rapid growth period in the first 3 years of life, the D2 receptor density is rapidly pruned before age 10, and thereafter decays by 2.2% every 10 years. Adapted from Seeman P, Bzowei NH, Guan H-C et al. (1987) Human brain dopamine receptors in children and aging adults. Synapse 1: 399–404, with permission from John Wiliy & Sons Inc.

Figure 11 Amino acid sequence of the human cloned dopamine D3 receptor. A polymorphism occurs at position 9, where glycine replaces serine. When a frame shift occurs in the cytoplasmic loop, as shown, the receptor is nonfunctional.

The Dopamine D5 Receptor

Regulation of Dopamine Receptors

The dopamine D5 receptor was cloned in 1991 (Sunahara), and its sequence is shown in Figure 14. There are two pseudogenes of D5, where the amino acid sequence stops at position 154. Although D5 is essentially D1-like in sequence and function, the characteristic feature of D5 is that it is more sensitive to dopamine than D1 is, as indicated in Table 1 for the K50 values.

Each of the five dopamine receptors has a state of high affinity and a state of low affinity for dopamine, an example of which is shown in Figure 15 for the D2 receptor in anterior pituitary tissue. Dopamine receptors belong to a group of more than 1000 receptors known to be associated with G-proteins. The binding of an agonist to such a G-linked receptor occurs in two concentration ranges. Low nanomolar

Dopamine Receptors and Antipsychotic Drugs in Health and Disease 403

Figure 12 Amino acid sequence of the human dopamine D4 receptor and its polymorphisms. The cytoplasmic loop has repeat units of 16 amino acid polypeptides. Different humans have different numbers of repeats. Shown are four repeats (D4.4), the most common, and seven repeats (D4.7).

concentrations of the agonist binds to the high-affinity state of the receptor, while high micromolar concentrations bind to the low-affinity state of the receptor. Generally, it is the high-affinity state of the receptor that is the functionally active state of the receptor, because the agonist affinities for the high-affinity state are usually identical to the concentrations that elicit the physiological action of the agonists. This holds for many neurotransmitter receptors, including dopamine D2 receptors, cholinergic muscarinic receptors, a2-adrenoceptors, and b2-adrenoceptors. D2High is the functional state in the anterior pituitary, upon which dopamine and other dopamine-like drugs (bromocriptine) act to inhibit the release of prolactin. D2High is presumably also functional on the terminals of the dopamine-containing terminals, and these receptors are usually referred to as presynaptic receptors. Although it has been reported that 90% of the D2 receptors in brain slices are in the D2High state, the proportion of D2 receptors in the high-affinity

state in homogenized striatum in vitro is generally between 15% and 20%. The D2High state can be quickly converted into the D2Low state by guanine nucleotide, as illustrated in Figure 15 for anterior pituitary tissue and in Figure 16 for the striatum. Furthermore, as shown in Figure 16, the high-affinity state of D2 is most readily detected by dopamine competing with [3H]domperidone, but not [3H]raclopride or [3H]spiperone, which are less sensitive to the competitive action of dopamine. In fact, it is known that the physiological concentration of dopamine in the synaptic space (between neurons) is 2–4 nM, matching the known dissociation constant of 2 nM for dopamine at the D2High receptor. This latter value of 2 nM is obtained from the dopamine/[3H]domperidone competition curve in Figure 16, using the standard Cheng–Prusoff equation to correct for the ligand concentration and the [3H]domperidone Kd. There are at least two views of the physical existence of the high-affinity state. The traditional view is that

404 Dopamine Receptors and Antipsychotic Drugs in Health and Disease D4

D4 Gly194

Repeat ASAG (8%, Italy)

Valine G

T

C

G

G

C

Glycine

13% Africans and caribbeans 0% Caucasians

Figure 13 Additional polymorphisms of the human dopamine D4 receptor. Approximately 13% of Africans and Caribbeans have a glycine replacing valine at position 194. Approximately 8% of Italians have an additional ASAG peptide inserted at the position shown.

D5

Figure 14 Amino acid sequence of the human dopamine D5 receptor. Two polymorphisms of D5 exist where the sequence abruptly stops to create a sequence of 154 amino acids instead of the full-length sequence of 477 amino acids.

the high-affinity state of the receptor exists when the receptor, R, is associated with the G-protein, and the agonist, D, binds to this high-affinity state to form the ‘ternary complex,’ namely DRG. This view

of the receptor proposes that the low-affinity state occurs when the G-protein is not associated with the receptor. However, there are many significant shortcomings with this view of the high-affinity state of the

Dopamine Receptors and Antipsychotic Drugs in Health and Disease 405

D2 [Ant. Pit.]

80 h gu Wit

D2High

anin

60

e nu

l ro nt Co

cleo

40

tide

[3H]Spiperone bound, (% specific)

100

20 D2Low 0 10−11

10−10

10−9

10−8

10−7

10−6

Dopamine (mol

10−5

10−4

10−3

(+)BTC

l−1)

Figure 15 The high-affinity state of the dopamine D2 receptor, or D2High, occurs at low nanomolar concentrations of dopamine in inhibiting the binding of [3H]spiperone to anterior pituitary tissue (Ant. Pit.; porcine). The low-affinity state of the D2 receptor, or D2Low, occurs at high concentrations of dopamine, as shown. However, the presence of 200 mM guanilylimidodiphosphate converts all the D2High receptors into the D2Low form.

receptor in the ternary complex model. For example, the ternary complex suggests that RG should have a transient existence. This is not the case, however, because it has been found that the purified muscarinic RG is stable. Moreover, the purified muscarinic receptor, free of G and GDP, clearly shows high-affinity and low-affinity states. An alternate view of the high-affinity state of the receptor is the ‘cooperativity’ model, as worked out by J Wells and colleagues. The cooperative model proposes that the receptor cooperates with other receptors to form either a dimer, a tetramer, or a larger oligomer. The receptor is in the high-affinity state when it is vacant and unoccupied by the agonist. However, when the agonist binds to the vacant receptor, the occupied receptor interacts or ‘cooperates’ with the other receptors (within the tetramer) such that the affinity of the other receptors for the agonist is markedly reduced. This reduced affinity for the agonist is a result of ‘negative cooperativity’ between the receptors, and corresponds to the low-affinity state of the receptor. In other words, if there is very strong negative cooperativity, then the second, third, and fourth receptors (within the tetramer) would hardly bind the agonist, and only the high-affinity sites would be observed in the competition between, say, dopamine and [3H]domperidone, all taking place at the first receptor. These events are depicted in a diagram in Figure 17.

D2 Interactions with Other Receptors The D2 receptor is known to interact with the D1 receptor as well as with other receptors, such as the adenosine A2 receptor. Many D1/D2 interactions occur at all levels, including the molecular level, where D1 and D2 can form functional heterodimers with one another. Such interaction also occurs at the cellular level, where the block of D1 (by SCH23390) unmasks the high-affinity state of the D2 receptor, D2High, as shown in Figure 18. This experiment shows that D1 normally inhibits or suppresses the high-affinity state of D2.

Psychosis and the D2High Basis of Dopamine Supersensitivity The dopamine hypothesis of psychosis or schizophrenia was first outlined by J. Van Rossum in 1967: The hypothesis that neuroleptic drugs may act by blocking dopamine receptors in the brain has been substantiated by preliminary experiments with a few selective and potent neuroleptic drugs. There is an urgent need for a simple isolated tissue that selectively responds to dopamine so that less-specific neuroleptic drugs can also be studied and the hypothesis further tested. . . . When the hypothesis of dopamine blockade by neuroleptic agents can be further substantiated it may have fargoing consequences for the pathophysiology of schizophrenia. Overstimulation of dopamine receptors could then be part of the aetiology.

406 Dopamine Receptors and Antipsychotic Drugs in Health and Disease D2 clone: D2High detected by [3H]domperidone, but not by [3H]spiperone or [3H]raclopride in 120 mM NaCI +20 0 gua mM nine nuc leot ide

8000 6000 4000

[3H]raclopride a 00 +2

14 000

mM

12 000

e in an gu

[3H]spiperone

e tid eo cl nu

Bound DPM per filter

8000

10 000 b 8000 00

+2 mM

4000

gu ine

an

[3H]domperidone

3000

leo

c nu e

tid

2000 c 0.1

1

10

100

1000

Dopamine (nM)

100 00

100 000

10 µM S-sulpiride

Figure 16 Low nanomolar concentrations of dopamine readily inhibit the binding of [3H]domperidone at the high-affinity state of dopamine D2 receptors (between 1 and 100 nM), in contrast to [3H]raclopride and [3H]spiperone, which are less sensitive to competition by dopamine. The high-affinity state, D2High, is converted to the low-affinity state, D2Low, by guanine nucleotide. Adapted from Seeman P, Tallerico T, and Ko F (2003) Dopamine displaces [3H]domperidone from high-affinity sites of the dopamine D2 receptor, but not [3H] raclopride or [3H]spiperone in isotonic medium: Implications for human positron emission tomography. Synapse 49: 209–215, with permission from John Wiley & Sons Inc.

As noted earlier, it has not been clearly established that D2 receptors are elevated in psychosis or schizophrenia, although studies using brain imaging of healthy co-twins of schizophrenia individuals, as well as single-photon brain imaging of nonmedicated psychotic patients in at least one study, have shown significant elevations of D2. At present, the most promising direction in this field is to examine the molecular basis of dopamine supersensitivity, because up to 70% of patients are supersensitive to either methylphenidate or amphetamine at doses which do not affect controls. Moreover, a wide variety of brain alterations in animals (lesions, birth injury by C-section, amphetamine or phencyclidine treatment, knockouts of a variety of receptors) all lead to the final common finding of

behavioral dopamine supersensitivity and elevated proportions of D2 receptors in the D2High state in the striatum. For example, repeated administration of amphetamine to animals or humans leads to behavioral dopamine supersensitivity. While the density of D2 receptors in the striatum does not change in such studies, it is remarkable that the density of D2High receptors increases dramatically by several fold, as shown in Figure 19. A similar situation occurs in animals that receive hippocampal lesions neonatally. Such animals, as adults, reveal behavioral dopamine supersensitivity, and the striatum contains a marked increase in the proportion of D2 receptors in the high-affinity state, as shown in Figure 20. Therefore, the molecular control of the high-affinity state of D2 is emerging as a

Dopamine Receptors and Antipsychotic Drugs in Health and Disease 407

Figure 17 (a) The cooperativity model for the dopamine D2 receptor, according to J Wells and colleagues. The receptor is proposed to exist as an oligomer of D2 receptors, such as a tetramer of D2. Each of the vacant receptors is in the high-affinity state, D2High. When dopamine first attaches to one of the vacant D2 receptors in the tetramer, the occupied D2 then interacts with the other three receptors within the tetramer to reduce their affinity for dopamine (i.e., negative cooperativity). (b) One molecular explanation for dopamine supersensitivity is that an unknown factor may reduce the negative interaction between the D2 receptors, thereby allowing more dopamine to occupy more D2High receptors.

Block of D1 reveals high states for D2 in rat striatum Control

2500

Bound [3H]raclopride, DPM per filter

2000 With 100nM SCH 23390

1500 1000 a 500

With 500 nM ketamine

2000

Control

1500

With 30nM L745,870 With 30 nM ketanserin

1000 500

b 0.1

1

10 100 1000 Dopamine (nM)

10 000

10 µM S-sulpiride

Figure 18 Regulation of D2High by the D1 receptor. (a) Dopamine inhibited the binding of [3H]raclopride to D2 receptors at dopamine concentrations higher than 100 nM. However, in the presence of SCH23390 to block D1 receptors, the binding of [3H]raclopride was readily inhibited by 1–100 nM, corresponding to the presence of D2High. These data suggest that D1 actively suppresses the existence of the functional D2High state. (b) In comparison to 100 nM SCH23390 unmasking the D2High state (a), other drugs do not lead to such unmasking of D2High. L745,870 is a dopamine D4 receptor antagonist. Adapted from Seeman P and Tallerico T (2003) Link between dopamine D1 and D2 receptors in rat and human striatal tissues. Synapse 47: 250–254, with permission from John Wiley & Sons Inc.

408 Dopamine Receptors and Antipsychotic Drugs in Health and Disease

central problem in this field. At present, there is uncertainty as to whether this high-affinity state of D2 is controlled through Go or one of the Gi proteins, because this varies from cell to cell.

According to the negative cooperativity model (Figure 17), the increased number of D2 receptors in the high-affinity state, D2High, found in the striata of supersensitive animals, may be attributed to a reduction in the overall negative cooperativity between the receptors, as illustrated in Figure 17. Thus, in order to determine the molecular mechanism of dopamine supersensitivity, it will be essential to determine the factors that reduce negative cooperativity among the D2 receptors or that alter the association of the receptor with its G-protein. The role of guanine nucleotides in regulating the overall sensitivity of the dopamine D2 receptors would be to alter the extent of the receptor–receptor negative cooperativity.

Current Clinical and Basic Research on Dopamine Receptors

Figure 19 Repeated administration of amphetamine to rats leads to behavioral dopamine supersensitivity. While the total density of D2 receptors was normal in the striata of such supersensitive rats (about 26 pmol g1), the density of D2High receptors was markedly elevated by 355%, from a control value of 2.9 pmol g1 to the elevated level of 10.3 pmol g1. Nonspecific binding of [3H]raclopride was done in the presence of 10 mM S-sulpiride. In the presence of 200 mM guanilylimidodiphosphate (G.N.), all the D2High receptors were converted to D2Low. Adapted from Seeman P, Tallerico T, Ko F, et al. (2002) Amphetaminesensitized animals show a marked increase in dopamine D2High receptors occupied by endogenous dopamine – Even in the absence of acute challenges. Synapse 46: 235–239, with permission from John Wiley & Sons Inc.

D2High = 10% D2High = 37%

80

am

Sh

60

ion ion

es

sl

pu am

Neonatal hippocampus lesion increases D2High states

oc

20

pp

40

les

Hi

% Specific binding of [3H]domperidone

100

Of the five dopamine receptors and their many variants, the D2 receptor and its properties continue to be most actively investigated, because D2 is the main clinical target of antipsychotics and of dopamine agonist treatment of Parkinson’s disease. The D1 receptor, however, also has an important clinical role in treating Parkinson’s disease because the stimulation of D1 synergizes with the stimulation of D2, possibly via D1/D2 heterodimers or cell–cell interactions. A current active area of clinical research on dopamine receptors is to measure the occupancy of D2 receptors both in the striatum and outside the striatum in individuals taking antipsychotic medications. Some researchers find that the same D2 occupancy occurs in both striatal and limbic regions, while others find a lower occupancy in the limbic regions.

0 0.01

0.1

1

10 100 Dopamine (nM)

1000

10 000

Figure 20 Rats with lesions made neonatally reveal behavioral dopamine supersensitivity when they become adults. Such supersensitive animals reveal a marked increase of 3.7-fold in the proportion of D2 receptors in the high-affinity state, D2High. Nonspecific binding of [3H]domperidone was done in the presence of 10 mM S-sulpiride. Adapted from Seeman P, Weinshenker D, Quirion R, et al. (2005) Dopamine supersensitivity correlates with D2High states, implying many paths to psychosis. Proceedings of the National Academy of Sciences of the United States of America 102: 3513–3518.

Dopamine Receptors and Antipsychotic Drugs in Health and Disease 409

As previously noted, therapeutic doses of antipsychotics occupy 60–80% of the D2 receptors in psychotic patients, while D2 occupancies higher than 80% are associated with elevated serum prolactin and parkinsonism. The new aripiprazole antipsychotic tends to occupy more than 80% of D2 but does not seem to cause parkinsonism at these higher levels. This may be a result of the fact that aripiprazole, like clozapine and quetiapine, quickly desorbs from the D2 receptor (in under 60 s in vitro), minimizing prolactin elevation and parkinsonism. Rapid desorption permits the dopamine neurotransmission to proceed more normally. However, if the antipsychotic drug dose or antipsychotic concentration in the plasma remains high, then the antipsychotic drug will readsorb. In effect, fat-soluble antipsychotics remain adsorbed to D2 in patients for 1 or 2 days or even more, while the plasma concentration falls. Less fat-soluble antipsychotics such as clozapine or quetiapine remain attached to D2 in patients for only 6–12 h, immediately falling in D2 occupancy as the plasma concentration falls. Probably the most central question in determining the basis of psychosis or schizophrenia is to determine the molecular basis of dopamine supersensitivity and to determine which proteins or genes regulate the maintenance of D2 receptors in their high-affinity state. See also: Dopamine; Dopamine: Cellular Actions.

Further Reading Agid O, Kapur S, Arenovich T, et al. (2003) Delayed-onset hypothesis of antipsychotic action: A hypothesis tested and rejected. Archives of General Psychiatry 60: 1228–1235. Baumeister AA and Francis JL (2002) Historical development of the dopamine hypothesis of schizophrenia. Journal of the History of Neurosciences 11: 265–277. Bressan RA, Erlandsson K, Jones HM, et al. (2003) Is regionally selective D2/D3 dopamine occupancy sufficient for atypical antipsychotic effect? An in vivo quantitative [123I]epidepride SPET study of amisulpride-treated patients. American Journal of Psychiatry 160: 1413–1420. Farde L, Nordstrom AL, Wiesel FA, et al. (1992) Positron emission tomographic analysis of central D1 and D2 dopamine receptor occupancy in patients treated with classical neuroleptics and clozapine. Relation to extrapyramidal side effects. Archives of General Psychiatry 49: 538–544.

George SR, Watanabe M, Di Paolo T, et al. (1985) The functional state of the dopamine receptor in the anterior pituitary is in the high affinity form. Endocrinology 117: 690–697. Hagberg G, Gefvert O, Bergstro¨m M, et al. (1998) N-[11C]methylspiperone PET, in contrast to [11C]raclopride, fails to detect D2 receptor occupancy by an atypical neuroleptic. Psychiatry Research 82: 147–160. Hirvonen J, van Erp TG, Huttunen J, et al. (2005) Increased caudate dopamine D2 receptor availability as a genetic marker for schizophrenia. Archives of General Psychiatry 62: 371–378. Kapur S and Seeman P (2001) Does fast dissociation from the dopamine D2 receptor explain the action of atypical antipsychotics? – A new hypothesis. American Journal of Psychiatry 158: 360–369. Seeman P (1987) Dopamine receptors and the dopamine hypothesis of schizophrenia. Synapse 1: 133–152. Seeman P (2002) Atypical antipsychotics: Mechanism of action. Canadian Journal of Psychiatry 47: 27–38. Seeman P, Bzowej NH, Guan H-C, et al. (1987) Human brain dopamine receptors in children and aging adults. Synapse 1: 399–404. Seeman P, Chau-Wong M, Tedesco J, et al. (1975) Brain receptors for antipsychotic drugs and dopamine: Direct binding assays. Proceedings of the National Academy of Sciences of the United States of America 72: 4376–4380. Seeman P, Lee T, Chau-Wong M, et al. (1976) Antipsychotic drug doses and neuroleptic/dopamine receptors. Nature (London) 261: 717–719. Seeman P and Niznik HB (1990) Dopamine receptors and transporters in Parkinson’s disease and schizophrenia. FASEB Journal 4: 2737–2744. Seeman P and Tallerico T (1999) Rapid release of antipsychotic drugs from dopamine D2 receptors: An explanation for low receptor occupancy and early clinical relapse upon drug withdrawal of clozapine or quetiapine. American Journal of Psychiatry 156: 876–884. Seeman P and Tallerico T (2003) Link between dopamine D1 and D2 receptors in rat and human striatal tissues. Synapse 47: 250–254. Seeman P, Tallerico T, and Ko F (2003) Dopamine displaces [3H] domperidone from high-affinity sites of the dopamine D2 receptor, but not [3H]raclopride or [3H]spiperone in isotonic medium: Implications for human positron emission tomography. Synapse 49: 209–215. Seeman P, Tallerico T, Ko F, et al. (2002) Amphetamine-sensitized animals show a marked increase in dopamine D2High receptors occupied by endogenous dopamine – Even in the absence of acute challenges. Synapse 46: 235–239. Seeman P, Weinshenker D, Quirion R, et al. (2005) Dopamine supersensitivity correlates with D2High states, implying many paths to psychosis. Proceedings of the National Academy of Sciences of the United States of America 102: 3513–3518. Wilson AA, McCormick P, Kapur S, et al. (2005) Radiosynthesis and evaluation of [11C]-(þ)-PHNO as a potential radiotracer for in vivo imaging of the dopamine D2 high affinity state with positron emission tomography (PET). Journal of Medicinal Chemistry 48: 4153–4160.

Dopamine: Cellular Actions G Bernardi and N B Mercuri, Universita` di Roma ‘Tor Vergata’, and IRCCS Fondazione Santa Lucia, Rome, Italy ã 2009 Elsevier Ltd. All rights reserved.

There is a plethora of experimental and clinical evidence indicating that dopamine (DA) is a key element in promoting and regulating the reward and motivation processes occurring in the brain. Therefore, the state of pleasure and well-being that one might derive from everyday activities such as playing, eating palatable food, drinking, listening to a melody, smelling a perfume, and having sex certainly rely on the correct operation of this catecholamine at the neuronal level. On the other hand, DA plays a significant role when an altered behavior is instituted. Hence, drug abuse and craving, compulsive acts, and affective problems could depend on maladaptive changes in DA neurotransmission. This catecholamine is synthesized by discrete groups of cells (mesencephalic, ipothalamic, and retinal). It is noteworthy that the main source of DA for the central nervous system arises from the population of neurons originating in the ventral mesencephalon (ventral tegmental area, substantia nigra pars compacta) and mainly terminating in the cerebral cortex, accumbens, and striatum. The diffuse projections of the dopaminergic neurons suggest that DA is able to act in different brain areas to induce motivation and reward. Accordingly, behavioral, pharmacological, and clinical studies have demonstrated that the cerebral regions principally involved in incentive and behavioral activation are the prefrontal cortex, the ventral/dorsal striatum, and the ventral mesencephalon itself. Indeed, these regions are the richest in DA. Therefore, in order to better understand the reward processes, it is of crucial importance to understand the functions of DA at the cellular and synaptic level within the different areas indicated previously. Another important aspect to be considered for the clarification of DA action is that the catecholamine stimulates subtypes of specific receptors that cause distinct effects on neurons. There are two families of G-protein-linked DAergic receptors, D1 and D2. While the D1 family (constituting the D1 and D5 receptors), linked to Gas, stimulates adenylate cyclase, the D2 family (constituting the D2, D3, and D4 receptors), linked to Ga0/i, inhibits adenylate cyclase. Usually, the D1-mediated effects are cyclic adenosine monophosphate (cAMP)and protein kinase A-dependent, while the D2mediated effects depend on the inhibition of cAMP and protein kinase A (PKA), eventually the activation

410

of phospholipase C-inositol triphosphate (IP3), and changes in intracellular Ca2þ. The dichotomy in the formation of cAMP results in opposite consequences on neuronal excitability. In addition, by stimulating either similar or different receptors, DA might exert competing presynaptic and/or postsynaptic actions to modify neuronal responsiveness. In the present article we reexamine the electrophysiological actions of DA on neurons located in those brain areas that are linked to the reward processes.

Dopamine Actions in the Frontal Cortex DA in the frontal cortex has implications in reward, the psychomotor and hedonic effects of drug abuse, drug seeking, and certain forms of memory formation linked to salience. Electrophysiological data have extensively shown a variety of effects caused by DA and DAergic agonists in this structure. Although there is a consensus that DA inhibits cortical neurons, several studies have also reported DA-mediated neuronal excitation. DA might have different effects on the excitability of pyramidal neurons by reducing either persistent voltage-dependent sodium currents or voltage-dependent potassium currents. It has been also reported that the stimulation of D1 receptors can activate firing rate by modulating synaptic strength. In fact, some authors have shown that DA significantly enhances excitatory postsynaptic current (EPSC) amplitudes in the pyramidal cells. Recent data have also suggested that DA potentiates the late postsynaptic responses in an excitatory postsynaptic potential (EPSP) train evoked by a sustained presynaptic stimulation. However, a combined activation of D1 and D2 receptors could also determine a reduction of the efficacy of excitatory synaptic transmission at synapses onto pyramidal cells. Moreover, contrasting effects of DA on neuronal responsiveness to glutamate agonists have been reported. Thus, bath application of Nmethyl-D-aspartate (NMDA), a-amino-3-hydroxy-5methyl-4-isoxazole propionic acid (AMPA), and the D1 agonist SKF 38393 induces concentration-dependent excitability increases, whereas the D2 receptor agonist quinpirole causes a concentration-dependent excitability decrease. Furthermore, opposing D1- or D2mediated actions, depending on DA concentrations, have been shown to modify the inhibitory drive on cortical neurons. As a result, a nanomolar concentration of DA might enhance the inhibitory postsynaptic currents (IPSCs) via D1 receptors while a micromolar concentration could decrease the IPSCs via D2 receptors. In contrast, it has been reported that the stimulation of presynaptic D1-like receptors

Dopamine: Cellular Actions 411

decreases the IPSCs in layer II–III of neocortical pyramidal cells. In addition, DA could induce important changes in the local network connections of pyramidal cells with g-aminobutyric acid (GABA)ergic interneurons. Thus, the activation of D1 receptors reduces the amplitude of the EPSPs or causes a reversible membrane depolarization in the fast-spiking (FS) interneurons. On the other hand, the stimulation of D4 receptors, limiting interneuronal excitability, favors multiple spike discharge of pyramidal cells. Furthermore, DA has been shown to sustain enduring changes of the synaptic signals. Thus, the long-term potentiation (LTP) of the excitatory transmission on pyramidal neurons requires D1 receptor stimulation. This could activate PKA and increase the surface expression of GluR1 in prefrontal cortex neurons. On the other hand, a constant presence of DA in the extracellular space might facilitate long-term depression (LTD). It is supposed that these DAmediated persistent changes in synaptic responses (LTP-LTD) could be important in processes linked to reward.

Dopamine Actions in the Ventral and Dorsal Striatum A sizable body of evidence considers the ventral striatum (nucleus accumbens) as a fundamental station for the DA-regulated reward processes. In fact, accumbens DA modulates motivation and cognitive aspects related to incentives. In addition, DA effects in this structure are particularly important in mediating the activating and addictive effects of psychostimulants. As in the cerebral cortex, accumbal DA might change the activity of medium spiny neurons by involving multiple mechanisms. It has been shown that this catecholamine causes a membrane hyperpolarization that is due to the activation of D1 receptors and is associated with an increase in potassium conductance. However, DA also causes a membrane depolarization that is generated by the stimulation of D2 receptors and sustained by a decrease in potassium conductance. In addition, D2 receptors seem to reduce the direct excitability of medium spiny accumbal neurons by inhibiting evoked Naþ spikes through the involvement of slow A-type Kþ currents, while reported effects of DA on inward rectification and ‘leak’ Kþ currents could favor excitability. D2 receptors could also cause an enhancement of voltage-sensitive sodium currents mediated by the suppression of the cyclic AMP/PKA cascade and the facilitation of intracellular Ca2þ signaling. In addition, D1 receptor stimulation appears to suppress N- and P/Q-type Ca2þ currents by activating a cAMP/

protein kinase A/protein phosphatase signaling system, presumably leading to channel dephosphorylation. Moreover, a coactivation of D1 and D2 receptors and signaling through G-protein bg subunits and PKA could enhance spike firing in nucleus accumbens shell medium spiny neurons. Referring to the DA-induced changes of synaptic functions, there are data supporting an inhibitory role of this catecholamine on the EPSP-IPSP sequences The depression of the EPSP appears to be mainly mediated by a D2 receptor-induced decrease of AMPA currents, while the inhibition of the EPSCs progressively diminishes; the inhibition of the IPSCs seems to be persistent. The resultant effect could cause an increased excitability of these neurons. Activation of D3 receptors is also reported to suppress the efficacy of the inhibitory synaptic transmission in the nucleus accumbens by increasing the phospho-dependent endocytosis of GABA-A receptors. Interestingly, DA also attenuates the neuronal responses to repetitive activation of glutamatergic afferents and thereby could block LTP. On the other hand, the accumbal LTD requires, among other factors, dopamine D1 receptor stimulation and cAMPPKA activation. DA-mediated plastic adaptations of the neural synaptic functions in the nucleus accumbens caused by psychostimulants might control the development and long-term maintenance of sensitization to abused drugs. The increase of extracellular DA in the dorsal striatum of humans (caudoputamen) related to selfreported measures of liking and the feeling of a ‘high’ (euphoria) implies that this structure is an essential element of the circuitry responsible for the control of motivated behavior and reward. However, in spite of the recognized importance of DA in reward-related locomotor activation, the precise nature of the modulation that DA exerts on striatal neurons remains largely elusive. There are early electrophysiological studies demonstrating a predominant inhibitory effect of DA on low-firing medium spiny striatal cells. Thus, the most common responses produced by DA on these cells are the decrease of voltage-dependent inward conductances and the modulation of the corticostriatal synaptic transmission. The DA-mediated inhibition of firing activity seems to be mediated by the activation of D1 receptors. The stimulation of D1-like receptors initiates a cascade of intracellular events, including cAMP formation and activation of cAMP-dependent PKA that reduce excitability and possibly activate the neuronal changes linked to reward. An additional stimulation of D2 receptor reduces glutamate- and GABA-mediated currents. Some authors have also described a reduction of

412 Dopamine: Cellular Actions

AMPA-mediated postsynaptic current caused by D2 receptors in medium-sized striatal neurons. In contrast, the stimulation of D1 receptors could potentiate the NMDA-induced currents and reduce GABAevoked currents by activating a PKA/dopamine- and cAMP-regulated phosphoprotein (DARPP-32)/protein phosphatase 1 signaling cascade. It must also be considered that the cholinergic interneurons are affected by DA. While the stimulation of D1 receptors depolarizes and excites the cells, the stimulation of D2 receptors reduces the GABAergic and cholinergic inputs on these interneurons. With regard to the synaptic function, the two main subclasses of DA receptors contribute in the formation of the long-term changes (LTD/LTP) occurring at the corticostriatal terminals. Therefore, D1-like and D2-like receptors interact synergistically to allow LTD formation. A form of LTD in the striatum is also bidirectionally modulated by D2 receptors and requires the regulation of mGluR-dependent endocannabinoid release. The LTP is blocked by dopamine depletion but is not affected by a D2 antagonist. Furthermore, the use of knockout mice with the ablation of D1 receptors has demonstrated a disrupted corticostriatal LTP, whereas pharmacological blockade of D5 receptors prevented LTD in these animals. Interestingly, DA-dependent stimulation of DARPP-32, a small protein expressed in the spiny neurons which acts as a potent inhibitor of protein phosphatase-1, is necessary for the expression of both corticostriatal LTD and LTP and is required for the full expression of the behavioral response to cocaine and amphetamine. On this basis, a role for neostriatal DA in the formation of a pathological drug habit has been hypothesized.

Dopamine Actions in the Ventral Mesencephalon There is homogeneous experimental material demonstrating that DA, released from the dendrites, has an inhibitory action on the dopaminergic cells located in the ventral tegmental area and substantia nigra pars compacta. This inhibition is D2-mediated and is due to a membrane hyperpolarization depending on a G-protein (Go/i)-activated potassium current (GIRK). Therefore, by modifying firing discharge of the DAergic cells, the catecholamine might encode the salience of stimuli linked to natural and artificial rewards. Although there is evidence that the dopaminergic neurons fire during reward anticipation, encoding errors in reward prediction, the inhibition of firing mediated by locally released DA could play part of the reward processes. Therefore, a delicate balance of activity in different subsets of DA cells could regulate the timing of the various expressions of reward. In addition, DA, by modifying the strength

of the excitatory transmission in the ventral mesencephalon, might contribute to enhancing memory processes linked to salience. Thus, DA could block the induction of LTD in the ventral midbrain via activation of D2-like receptors. This DA-dependent modulation of the synaptic code could be a substrate of memory traces linked to reward. Of note is the fact that the DA-mediated long-term alterations in the strength of the excitatory transmission have been reported to occur after treatments with psychomotor stimulants. Since the development of addictive behaviors shares common features with traditional learning models, these DA-dependent modifications of synaptic plasticity could represent an important substrate for the acquisition of reward-related behaviors.

Conclusions It is believed that the multiple and, in some instances, opposite effects of DA on cerebral neurons could account for the whole range of behavioral modifications linked to reward processes. However, in spite of the large amount of experimental work so far available, we have to admit that the overall picture describing the actions of DA on single cells and groups of cells is rather complex. Undoubtedly, there is a certain unpredictability in the neuronal responses to DA that could likely depend on competing presynaptic, postsynaptic, and nonsynaptic mechanisms, the type of receptor involved, and the type of neuron. Notwithstanding these limitations, future effort in studying the function of DA should be concentrated on individuating and eventually correcting the processes that control reward at the cellular level. See also: Dopamine; Dopamine Receptors and Antipsychotic Drugs in Health and Disease.

Further Reading Bernardi G, Cherubini E, Marciani MG, et al. (1982) Responses of intracellularly recorded cortical neurons to the iontophoretic application of dopamine. Brain Research 245: 267–274. Beurrier C and Malenka RC (2002) Enhanced inhibition of synaptic transmission by dopamine in the nucleus accumbens during behavioral sensitization to cocaine. Journal of Neuroscience 22: 5817–5822. Bonci A, Bernardi G, Grillner P, et al. (2003) The dopaminecontaining neuron: Maestro or simple musician in the orchestra of addiction? Trends in Pharmacological Science 24: 172–177. Centonze D, Grande C, Saulle E, et al. (2003) Distinct roles of D1 and D5 dopamine receptors in motor activity and striatal synaptic plasticity. Journal of Neuroscience 23: 8506–8512. Cepeda C, Colwell CS, Itri JN, et al. (1998) Dopaminergic modulation of NMDA-induced whole cell currents in neostriatal neurons in slices: Contribution of calcium conductances. Journal of Neurophysiology 79: 82–94.

Dopamine: Cellular Actions 413 Chen G, Kittler JT, Moss SJ, et al. (2006) Dopamine D3 receptors regulate GABAA receptor function through a phospho-dependent endocytosis mechanism in nucleus accumbens. Journal of Neuroscience 26: 2513–2521. Kreitzer AC and Malenka RC (2005) Dopamine modulation of state-dependent endocannabinoid release and long-term depression in the striatum. Journal of Neuroscience 25: 10537–10545. Lacey MG, Mercuri NB, and North RA (1987) Dopamine acts on D2 receptors to increase potassium conductance in neurones of the rat substantia nigra zona compacta. Journal of Physiology 392: 397–416. Law-Tho D, Hirsch JC, and Crepel F (1994) Dopamine modulation of synaptic transmission in rat prefrontal cortex: An in vitro electrophysiological study. Neuroscience Research 21: 151–160. Perez MF, White FJ, and Hu XT (2006) Dopamine D(2) receptor modulation of K(þ) channel activity regulates excitability of

nucleus accumbens neurons at different membrane potentials. Journal of Neurophysiology 96: 2217–2228. Schultz W (2002) Getting formal with dopamine and reward. Neuron 36: 241–263. Seamans JK and Yang CR (2004) The principal features and mechanisms of dopamine modulation in the prefrontal cortex. Progress in Neurobiology 74: 1–58. Thomas MJ, Malenka RC, and Bonci A (2000) Modulation of long-term depression by dopamine in the mesolimbic system. Journal of Neuroscience 20: 5581–5586. Uchimura N, Higashi H, and Nishi S (1986) Hyperpolarizing and depolarizing actions of dopamine via D-1 and D-2 receptors on nucleus accumbens neurons. Brain Research 375: 368–372. Wise RA (2004) Dopamine, learning and motivation. Nature Reviews Neuroscience 5: 1–12.

Noradrenaline R D Wassall, University of Oxford, Oxford, UK N Teramoto, Kyushu University, Fukuoka, Japan T C Cunnane, University of Oxford, Oxford, UK ã 2009 Elsevier Ltd. All rights reserved.

Introduction From its humble beginnings, with biochemical analysis of adrenal glands and neurotransmitter release, we have come to realize that sympathetic neurotransmission is not quite as simple as first thought. We are taught that noradrenaline (norepinephrine) is the major neurotransmitter released from sympathetic nerves, and that with adrenaline (epinephrine), it is a key player in the ‘fight-or-flight’ response. Indeed, few people can escape the discussion of the ‘adrenaline-high’ associated with athletes and thrillseekers alike, and few of us will avoid the use of drugs to alter sympathetic signaling for cardiovascular problems, anxiety or depression as we age. We might think that such an important and familiar neurotransmitter system would be well characterized, but it has not always been so. In the past, noradrenaline proved to be an elusive agent to identify when the concepts of ‘neurochemical transmission’ were being formed, and even today, there are a myriad of questions surrounding its co-storage, release, and postjunctional receptor mechanisms. Here, we give a brief overview of some of the facts, and unravel some of the mysteries of this celebrated molecule.

A Brief History of the Discovery of Noradrenaline Galen (AD 130–200), arguably the greatest anatomist and physiologist of antiquity, first described the gross anatomical features of autonomic nerves. He suggested that, being hollow, they allowed the transfer of so-called ‘animal spirits’ between organs, producing the phenomena of ‘sympathy’, which was a vague term used to describe the coordination or cooperation of organs. However, the anatomical division of the parasympathetic and sympathetic nervous systems can be traced to Eustachius in the sixteenth century, who regarded the sympathetic and vagus nerves as separate, although describing incorrectly that the ‘sympathetic nerve’ was a continuation of the abducens (cranial nerve VI). Through the work of Winslow, Whytt, and Bichat two centuries later, an important outline of the sympathetic nervous system was laid out, with the visceral organs controlled by the

414

ganglia, whereas voluntary action was concerned with the brain. The first functional studies on true sympathetic nerves began with Bernard, who dilated vessels by sectioning the sympathetic nerves, and Brown-Se´quard, who contracted the vessels by stimulating the cut end of the nerve. Irritation of various neuronal structures had been shown to increase the heart rate and its contractility; however, modern work on the autonomic nervous system and its functional divisions has been based largely on the work of Gaskell and Langley, who defined the essentials of the macroscopic functions of the sympathetic nervous system in the early to mid-1900s. Langley’s student, Elliott, showed that adrenaline, discovered approximately 20 years earlier by Bates, had the same general effect as stimulation of the sympathetic nerves. The chemical nature of sympathetic neurotransmission had been assumed by Elliott and Dale, who showed the sympathetic-like effects of ergot alkaloids, and when Loewi monumentally demonstrated the chemical nature of neurotransmission in a frog’s heart, later found to be due to acetylcholine, it seemed that adrenaline would be an attractive candidate neurochemical in the sympathetic nervous system. The isolation of ‘sympathin’ from sympathetic nerves by Cannon revealed that adrenaline was similar to, but chemically distinct from, the neurotransmitter released, and its identity remained elusive until von Euler revealed that the true neurotransmitter was the nonmethylated derivative of adrenaline, noradrenaline.

Structure and Biosynthesis of Noradrenaline Noradrenaline, a catecholamine, is derived from L-tyrosine, an aromatic amino acid present in the body fluids and taken up by noradrenaline-producing cells. Through various intermediate steps (Figure 1), L-tyrosine is converted to noradrenaline and, finally, to its methylated form, adrenaline, in phenylethanolamine N-methyltransferase-containing cells. The first cytosolic enzyme, tyrosine hydroxylase, is inhibited by excess production of noradrenaline and is the ratelimiting step in the regulation of noradrenaline synthesis. Tyrosine hydroxylase is found only in the cytosol in catecholamine-containing cells and is used as a marker for detection of adrenergic neurons. Dopamine b-hydroxylase is also used as a selective marker for catecholamine-containing cells, but it is located in the secretory vesicles, usually membrane bound; however, a small amount is soluble and released with the vesicle contents upon exocytosis.

Noradrenaline 415 COOH L-tyrosine

NH2

HO

Tyrosine hydroxylase HO

COOH

L-3,4-dihydroxyphenylalanine

NH2

HO

(L-DOPA) DOPA Decarboxylase

HO NH2

HO

OH

Dopamine Dopamine b-Hydroxylase

HO Noradrenaline HO

NH2 OH

HO HO

Phenylethanolamine N-methyltransferase Adrenaline

NH CH3

Figure 1 Biosynthetic pathway of noradrenaline and adrenaline.

Where Is Noradrenaline Found? Noradrenaline is released with adrenaline into the blood from the medullae of the adrenal glands where it can act as a circulating hormone, but it is classically thought of as the main neurotransmitter released from postganglionic sympathetic neurons. Sympathetic fibers project to the heart and along blood vessels, controlling cardiovascular responses to maintain blood pressure, and, with the aid of circulating adrenaline, regulate bronchodilation, glycogen and fat metabolism, thermogenesis, and secretions of hormones and mucus membranes (Table 1). There are also large networks of noradrenergic and adrenergic neurons originating in or close to the locus coeruleus and reticular formation in the brain stem, extensively projecting to the cortex, hippocampus, and cerebellum, and these neurons are involved in baroreceptor and blood pressure reflexes, and also in complex behaviors such as arousal and mood. Many of these tissues are being used to study noradrenaline release and reuptake, but the vas deferens offers the ideal system in which to study sympathetic neuronal release due to its ease of isolation, dense sympathetic innervation, high noradrenaline content, and exquisite sensitivity to pharmacological manipulation. It is perhaps useful at this stage to remind the nonspecialist reader of the anatomy of the autonomic nervous system. As shown by Langley, autonomic nerves typically have both a preganglionic nerve traveling

from the central nervous system and a postganglionic nerve traveling to the effector organ. Parasympathetic fibers either arise from specific cranial nuclei and travel in the occulomotor (III), facial (VII), glossopharyngeal (IX), and vagus (X) cranial nerves or leave the spinal cord at sacral levels S2–S4 before synapsing close to the tissues in collections of cell bodies known as ganglia, and continuing by way of short postganglionic fibers to the effector cells. Sympathetic neurons, however, are generally characterized by short preganglionic fibers, exiting the spinal cord at thoracolumbar levels T1–L3, synapsing in the paravertebral chains, and continuing to the effector organs as long postganglionic fibers (possibly more than 1 m long). The innervation of the rodent vas deferens, a wellcharacterized model system, is unusual in that it is supplied by short postganglionic sympathetic neurons whose cell bodies lie within the hypogastric ganglia situated close to the prostatic end. These postganglionic nerves are mainly nonmyelinated axons (0.2–2 mm in diameter) embedded in Schwann cells, together with a small number of fine myelinated fibers (1 or 2 mm in diameter), which travel through the connective tissue and divide into numerous branches as they enter the prostatic end. Traversing the connective sheath, branches pass into the smooth muscle layers, splitting into smaller bundles of two to eight axons, and become varicose. The varicosities, which are approximately 1 or 2 mm in length and 1–1.5 mm in cross section,

416 Noradrenaline Table 1 Summary of major adrenoceptor signaling and some drugs that can affect it Adrenoceptor subtype

a1

a2

b1

b2

b3

Primary signal transduction

Stimulation of phospholipase C (Gq/11) "IP3, "DAG, and "Ca2þ Stimulation of phospholipase D  Vasoconstriction  Relaxation of gastrointestinal tract smooth muscle  Contraction of seminal tract and uterus  Contraction of iris radial muscle  Glycogenolysis in liver  Kþ release from salivary glands

Inhibition of adenylyl cyclase (Gi/0)

Stimulation of adenylyl cyclase (Gs) "cAMP Stimulation of guanylyl cyclase (Actions mainly on heart)  Positive inotropy  Positive chronotropy  Positive dromotropy  Amylase secretion from salivary glands  Renin release from juxtaglomerular cells of the kidney

Stimulation of adenylyl cyclase (Gs) "cAMP Stimulation of guanylyl cyclase  Vasodilation  Bronchodilation  Relaxation of gastrointestinal tract smooth muscle  Relaxation of seminal tract and uterus  Relaxation of ciliary muscle  Glycogenolysis in liver and skeletal muscle  Increased release of noradrenaline from nerve terminals Salbutamol Salmeterol Butoxamine ICI 118,551

Stimulation of adenylyl cyclase (Gs) "cAMP Stimulation of guanylyl cyclase  Lipolysis and thermogenesis in adipose tissue  Thermogenesis in skeletal muscle

Secondary signal transduction Major physiological effects

Selective agonists Selective antagonists

Phenylephrine Cirazoline Prazosin Doxazosin

#cAMP Stimulation of adenylyl cyclasea  Vasoconstriction  Relaxation of gastrointestinal tract smooth muscle  Decreased release of acetylcholine and noradrenaline from autonomic nerve terminals  Decreased insulin release  Platelet aggregation  Inhibition of sympathetic outflow in brain stem Clonidine Medetomidine Yohimbine Idazoxan

Dobutamine Xamoterol Atenolol Metoprolol

BRL 37344 ZD 7114 SR 59230A

a The role of stimulating adenylyl cyclase in addition to its inhibition by a2-adrenoceptor activation in some systems remains unknown. For simplicity, subtypes of receptors are not shown. IP3, inositol-1,4,5-trisphosphate; DAG, diacylglycerol; cAMP, 30 -50 -cyclic adenosine monophosphate.

are packed with vesicles and mitochondria and are separated by intervaricose regions approximately 3–5 mm in length. The diameter of the axon spanning the intervaricose regions is only 0.1 mm, and each axon gives rise to between 10 000 and 30 000 varicosities (Figure 2). The conduction velocity in these nerves is less than 1 m s 1. In the vas deferens, the varicosities traverse the smooth muscle cells in complex patterns, such that at least one varicosity makes close contact (approximately 20–50 nm) with each smooth muscle cell, with the varicosities being the sites where noradrenaline is synthesized and released.

How Can Noradrenaline Release Be Measured? Early electrophysiological investigations into sympathetic nerves relied on intracellular recording. Basing the technique on the Nobel prize-winning work of Hodgkin, Huxley, and Katz, the nerve cell bodies in ganglia were impaled using glass microelectrodes and electrical changes across the membrane were

measured. As at the neuromuscular junction, electrical signals could be detected in the absence of stimulation (excitatory postsynaptic potentials in nerves and end plate potentials in the motor end plate), as could the resulting action potentials on stimulation. Although this method gave fascinating insights into the release of acetylcholine, the primary neurotransmitter at these synapses, it could not reveal anything about the terminal release of noradrenaline from the postganglionic nerve. Electrophysiological measurements from nerve terminal varicosities proved difficult due to their small size and the large microelectrode diameters; progress remained slow. In an effort to study and localize noradrenaline, a number of methods have been employed, including high-performance liquid chromatography, overflow, radiolabeling, and voltametry techniques. These approaches all indicated a neuronal release of noradrenaline, the most abundant neurotransmitter in sympathetic neurons, and that noradrenaline was stored in the vesicles in preparation for exocytosis. However, fractional release studies showed that

Noradrenaline 417

~0.1 µm 2 µm ~1.5 µm

30 mm

1 cm–1 m

~1 µm

a

Soma

b

Axon bundles

c

Varicosities

Figure 2 (a) Schematic diagram of a postganglionic sympathetic neuron showing the difference in size between cell body, axon, and varicosity. (b) A set of axon bundles loaded with the Ca2þ indicator Oregon Green 488 BAPTA-1 dextran. Note that only a single axon is represented in the corresponding box in (a). (c) Image of part of a sympathetic terminal loaded with the Ca2þ indicator Oregon Green 488 BAPTA-1 dextran. Reproduced from Jackson VM and Cunnane TC (2001) Neurotransmitter release mechanisms in sympathetic neurons: Past, present, and future perspectives. Neurochemical Research 26: 875–889, with permission from Springer Science and Business Media.

there was a disparity between the amount of noradrenaline released and the noradrenaline content of a vesicle. Although these methods did not have the temporal or spatial resolution at the time to investigate the release characteristics of noradrenaline from a single varicosity, biochemical analysis of these studies provided two radically different hypotheses to explain the mechanism of noradrenaline release from the ‘average’ varicosity. First, the ‘fractional release’ model suggested that the action potential only released a small fraction of the contents of one vesicle (1–3%) from each varicosity every time the nerve was stimulated, similar to the modern ‘kiss-and-run’ mechanism observed at some synapses in the central nervous system. In this scenario, on average 30–100 action potentials are required to release the equivalent of the neurotransmitter content of one vesicle from each varicosity, and this would mean that there is a constant release of neurotransmitter during neuronal activity providing

an average ‘biophase’ concentration which bathes the smooth muscle and maintains a tonic contraction. With the assumption that the average varicosity has 1000 vesicles, this would mean that only approximately 300–400 molecules of transmitter would be released with each action potential, which should nonetheless be sufficient to reach maximal concentrations for receptor activation in a close contact neuroeffector junction. Today, many textbooks still favor this explanation (as did the early investigators) because it appears to be a more efficient and economical way of transmitter turnover to achieve a given effector response while at the same time reusing the vesicles and allowing for efficient noradrenaline reuptake (approximately 90%). However, this would mean that sympathetic nerves would not behave like all other nerves, where quantal transmission was accepted. An equally radical hypothesis is the ‘intermittent release’ model, which suggested that each action potential activates only 1–3% of all varicosities in the

418 Noradrenaline

tissue. When the release machinery was activated, an individual varicosity discharged a single packet of neurotransmitter equivalent to the neurotransmitter content of a single vesicle. In this case, neurotransmitter release would be highly intermittent at the level of the individual varicosity. This would cause approximately 50-fold higher noradrenaline concentrations in the junction, giving a larger ‘margin of safety’ for transmission, but would demand extreme efficiency of the local noradrenaline reuptake mechanisms to clear noradrenaline and terminate the response before it diffused away. If this theory were true, it would raise more questions about why and how a varicosity would ‘decide’ to release neurotransmitter or not. It was understandable that the investigators favored the previous explanation. If this controversy was to be settled, then techniques with better resolution needed to be developed.

What Can We Learn About Sympathetic Neurotransmitter Release from Electrophysiology? Burnstock and Holman, working in Melbourne, Australia, used one of the most densely innervated sympathetic tissues in the body, the vas deferens, to study neurotransmitter release. Sympathetic nerves in the body typically have low-frequency firing rates, and unlike the overflow studies, which require high frequencies to cause measurable overflow of transmitter, electrophysiology allowed the measurement of release on an impulse-to-impulse basis. Using microelectrode techniques similar to those used in ganglia, they impaled smooth muscle cells of the guinea pig vas deferens and recorded resting membrane potentials. In response to sympathetic nerve stimulation, transient depolarizations were recorded, termed excitatory junction potentials (EJPs). EJPs had a fairly slow time course (0.5–1 s), were graded with stimulus intensity, and increased in amplitude in response to the first few stimuli in a train (i.e., they exhibited facilitation). Perhaps more important, in the absence of stimulation, spontaneous EJPs (SEJPs) were recorded, which were reminiscent of miniature end plate potentials (MEPPs) at the skeletal neuromuscular junction that had famously been used to describe the phenomena surrounding quantal neurotransmission. SEJPs implied that neurotransmitter was released in multimolecular packets; however, unlike MEPPs and end plate potentials in skeletal muscle, EJPs and SEJPs had different time courses and could not be compared directly. The time course differences were probably a reflection of the tight electrical coupling between smooth muscle cells of the vas deferens, which caused rapid dissipation of charge from the point

source, and also that the EJP represented the sum of electrical activity at several simultaneous release sites being activated throughout the muscle. Taken together with the overflow studies, and the difficulty in interpreting the results for a single release site, the electrical signal seemed to support the fractional release model.

High-Resolution Studies of EJPs As the fractional release model enjoyed prime status, new methods were devised to break down EJPs into their component parts. Blakeley and Cunnane noted discontinuities between the rising phases of individual EJPs in the mouse and guinea pig vas deferens, and, on electronically differentiating the rising phases, these discontinuities were extracted as transient peaks in the rate of depolarization of the EJP, termed ‘discrete events’ (Figure 3). On comparing discrete events generated from SEJPs and EJPs, it was found that they matched almost perfectly, suggesting that release was intermittent and monoquantal. Various groups quickly verified the data, and so the fractional release model was moved aside for the intermittent model. Perhaps, in hindsight, it should not have come as a surprise that sympathetic nerves behave like all other nerves in the body and exhibit quantal or packeted neurotransmission.

0.5 V s–1

10 mV 20 ms Figure 3 Intracellular recordings of the membrane potential and its time derivative in single smooth muscle cells of the guinea pig vas deferens. The preparation was stimulated at 0.91 Hz and the transient peaks in the rate of depolarization (the ‘discrete event’) occur intermittently, showing that action potential evoked neurotransmitter release is intermittent. Reproduced from Blakeley AG and Cunnane TC (1979) The packeted release of transmitter from the sympathetic nerves of the guinea-pig vas deferens: An electrophysiological study: Journal of Physiology 296: 85–96; and Jackson VM and Cunnane TC (2001) Neurotransmitter release mechanisms in sympathetic neurons: Past, present, and future perspectives. Neurochemical Research 26: 875–889, with permission from Springer Science and Business Media.

Noradrenaline 419

Mechanisms of Intermittence Sympathetic nerves proved to be difficult to comprehend. If they only released once in every 100 stimuli or so, then how could we use them to study release? We need to understand some of the mechanisms behind intermittence before we can investigate noradrenaline release further. There were two suggestions put forward to explain intermittence: . Intermittent failure of the action potential to propagate throughout the complex nerve terminal network of branching axons and varicosities . A very low probability of neurotransmitter release in the invaded varicosities To resolve this conflict, focal extracellular measurement was developed by Brock and Cunnane. By moving an electrode close to the tissue and applying gentle suction, a small area could be ‘isolated’ from the tissue and drugs applied either inside or outside the electrode, which was perfused with physiological solution. It allowed the simultaneous measurement of the nerve terminal impulse (NTI) and also neurotransmitter release – the excitatory junction current (EJC). In the absence of stimulation, spontaneous EJCs were recorded, and following electrical nerve stimulation it was found that the action potential (NTI) arrived each time, whereas neurotransmitter release was highly intermittent (Figure 4). Application of tetrodotoxin, an inhibitor of voltage-gated Naþ channels, abolished the NTI and evoked neurotransmitter release, but did not abolish spontaneous EJCs. Intermittence of neurotransmitter release was concluded to be due to some failure in depolarization–secretion coupling after the sympathetic nerve terminal is invaded by the action potential rather than action potential propagation failure in the complex terminal arborization of varicosities. One of the more interesting observations occurred when pharmacological agents were used to manipulate neurotransmission. Following blockade of a-adrenoceptors, the major type of receptors for noradrenaline in the vas deferens, there was an increase in the overflow of noradrenaline and also an increase in the electrical signals recorded with microelectrodes in the smooth muscle cells. It transpired that there were receptors located on nerve terminals, a2-adrenoceptors, that were distinct from postjunctional adrenoceptors (mainly a1-adrenoceptors) and limited the release of noradrenaline. This led to the novel concept of a2-adrenoceptor-mediated inhibition of neurotransmitter release in the great majority of sympathetically innervated tissues and added a new dimension to our understanding of the control of sympathetic nerve function. Previously, the centrally

1 Hz 1 Hz 2 Hz

2 Hz

4 Hz 4 Hz 100 µV

NTI

EJCs

10 ms Figure 4 Frequency-dependent facilitation of neurotransmitter release in the guinea pig vas deferens. Simultaneous measurement of the nerve terminal impulse (NTI) and evoked excitatory junctional currents (EJCs) using focal extracellular recordings showed that although the impulses travel through the varicosities every time, EJCs are intermittent. Each panel shows 25 stimuli at 1, 2, and 4 Hz in the top, middle, and bottom panels, respectively. Reproduced from Jackson VM and Cunnane TC (2001) Neurotransmitter release mechanisms in sympathetic neurons: Past, present, and future perspectives. Neurochemical Research 26: 875–889, with permission from Springer Science and Business Media.

determined patterns of nerve firing were primarily thought to control neurotransmitter release, but now it was shown that neurotransmitter release could be controlled locally due to feedback inhibition of the action of released noradrenaline at distinct receptors on nerve terminals producing a brake on further neurotransmitter release. There was no doubt intermittence could be regulated by this local inhibition at the level of the varicosity, and although this idea was largely born in the peripheral nervous system, it is generally accepted that presynaptic receptors exist at every synapse. Indeed, we now know that many nerves have both inhibitory and excitatory receptors located on their nerve terminals, such as the prejunctional excitatory nicotinic receptors and the prejunctional inhibitory muscarinic receptors at the neuromuscular junction, which can increase or decrease acetylcholine release, respectively. The importance of increases in intracellular Ca2þ concentrations as the trigger for exocytosis is wellknown, and so perhaps intermittence was due to the failure of most varicosities to undergo appropriate changes in Ca2þ levels. Developments in confocal microscopy and Ca2þ indicators mean that we can now finally resolve Ca2þ dynamics in intact tissue preparations. By loading the nerve terminals through the cut end of the vas deferens, it has been shown that Ca2þ enters all varicosities when the action potential arrives.

420 Noradrenaline

Figure 5 Discreet and intermittent smooth muscle Ca2þ transients following nerve stimulation. The first six frames show selected images of the same smooth muscle cell taken during 2 Hz stimulation. There is no response to most stimuli (frame 1), whereas some stimuli evoke focal Ca2þ transients in the smooth muscle cell (frames 2–6). The final frame was obtained from a confocal section 3 mm above that of the preceding images, and it shows an overlying nerve terminal varicosity. The white dots denote the location (epicenter) of a single smooth muscle Ca2þ transient occurring at some time during eight sets of recordings. Reproduced from Brain KL, Jackson VM, Trout SJ, and Cunnane TC (2002) Intermittent ATP release from nerve terminals elicits focal smooth muscle Ca2þ transients in mouse vas deferens. Journal of Physiology 541: 849–862.

Through Ca2þ studies on neurons, it was questioned whether it would be possible to resolve neurotransmitter release optically. Occasionally, for indeterminate reasons, some smooth muscle cells were loaded with the Ca2þ indicator, and distinct changes in Ca2þ were noted following nerve stimulation both in the varicosity and directly below in the smooth muscle cell. The focal changes in Ca2þ recorded in the smooth muscle cells were termed neuroeffector Ca2þ transients (NCTs) (Figure 5). By simultaneously monitoring both pre- and postjunctional responses evoked by low-frequency nerve stimulation with cell-permeant Ca2þ indicators, Brain, Jackson, and Cunnane found that these NCTs correlated well with action potentialevoked Ca2þ transients in varicosities and with the release probability of individual varicosities on the same axon. Surprisingly, they found that some varicosities had an apparent release probability as high as 0.09, but there was a population of varicosities that never released and remained ‘silent.’ Finally, there is an optical method that provides a unique approach to detect neurotransmitter release at the level of the individual sympathetic neuroeffector junction (i.e., single varicosities on the same nerve terminal branch) at an unparalleled resolution. Currently, we still do not know the reason for intermittence. However, the ability to demonstrate neurotransmitter release at high resolution surely holds the key to unraveling the mechanisms underlying release modulation at individual varicosities on an impulseto-impulse basis in an intact organ. Autoinhibition, variation in secretory proteins, and discrete changes in Ca2þ in different varicosities are among some of the current hypotheses undergoing evaluation, but we are now in a better position to study neurotransmitter

release from these nerves. However, what is the neurotransmitter? Surely it is noradrenaline.

The Identity Revealed Although there was broad agreement across all techniques that noradrenaline is released on an impulseto-impulse basis from sympathetic nerves, there were a number of observations that did not fit the noradrenergic theory. For example, when the rodent vas deferens was depleted of noradrenaline using reserpine, which irreversibly interferes with the Mg2þand adenosine-50 -triphosphate (ATP)-dependent uptake of biogenic amines into vesicles and thereby depletes noradrenaline from nerves, electrophysiological methods showed that the EJPs were still present and often larger in amplitude. This was due to a prejunctional action rather than a postjunctional sensitization because the size of the SEJPs remained unchanged. Similarly, when yohimbine was applied to block prejunctional autoinhibitory a2-adrenoceptors, EJPs were no longer potentiated. In other words, we have a tissue devoid of noradrenaline, but EJPs, the electrical sign of neurotransmitter release, are still present – a strange paradox. Further investigation using selective antagonists led to the rather revolutionary concept of noradrenaline– ATP co-transmission. Imagine the controversy when Burnstock, Westfall, and others suggested that ATP, Krooll’s ‘universal currency of bioenergy,’ could also be released from sympathetic nerves and that it was ATP, and not noradrenaline, that was responsible for the generation of the EJP. This was not so surprising in retrospect because Douglas and Poisner had shown that ATP was co-stored and co-released

Noradrenaline 421

with adrenaline and noradrenaline in the adrenal medulla approximately 20 years previously. Some people called purinergic transmission the ‘pure imagination’ hypothesis on account of its radical and unbelievable concepts, but it is now generally accepted that ATP is released from almost all nerves and acts as a separate neurotransmitter. Indeed, moving away from Dale’s principle, which states that each neuron can only secrete one type of neurotransmitter, we now know that neurotransmitter release can differ at different synapses of the same cell, both in the class of neurotransmitter and also in the quantities of the cocktails of neurotransmitters released. Yet again, an idea born in the peripheral nervous system is now accepted widely throughout the central nervous system. Work on purinergic transmission spawned the discovery of G-protein-coupled P2Y purinoceptors and of a unique class of ligand-gated ion channels, P2X receptors, which do not follow the structure of any other family of ligand-gated channels. After activation by extracellular ATP, through nonselective cation entry, P2X receptors are responsible for the EJP and SEJP because P2X receptor antagonists and desensitizers abolish both EJPs and SEJPs. NCTs are a measure of the Ca2þ flowing through the nonselective cation pore together with amplification from the release of Ca2þ from intracellular stores, and similarly, these drugs abolish NCTs in the vas deferens, blood vessels, and bladder. In the vas deferens, the purinoceptor subtype P2X1 is particularly important because knockout mice that do not have P2X1 receptor expression lack EJPs, SEJPs, NCTs, and spontaneous NCTs.

It would seem that the electrophysiological and confocal investigations have not been monitoring the effects of released noradrenaline but, rather, measuring how sympathetic nerves release ATP. However, if noradrenaline and ATP are co-stored in sympathetic nerve vesicles, as the biochemical analysis would suggest, then ATP release will also reflect how noradrenaline is being released assuming exocytosis of the entire neurotransmitter content of a vesicle. Indeed, research using amperometry, a direct measurement of the oxidation current of noradrenaline, has shown that noradrenaline can be released intermittently on an impulseto-impulse basis (Figure 6). Finally, overflow studies, optical studies, and electrophysiological methods appear to be converging in order to explain sympathetic neurotransmitter release, be it ATP or noradrenaline.

The Elusive Nature of Noradrenaline – The Paradox Noradrenaline produces its effects through G-proteincoupled receptors (Table 1) and its action is terminated by reuptake into the prejunctional neuronal (uptake 1; inhibited by cocaine, desipramine, amphetamines, and tricyclic antidepressants) and postjunctional nonneuronal (uptake 2; inhibited by corticosterone) cells. Reuptake results in either the metabolism of noradrenaline by monoamine oxidases and catechol-O-methyl transferases within a reaction pathway that finally produces vanillylmandelic acid or 3-methoxy,4-hydroxyphenylglycol, which are excreted in the urine, or the noradrenaline can be recycled and released again from the prejunctional nerve terminal. In the smooth muscle

0.5 pA

1.5 s Figure 6 Intermittent release of noradrenaline from rat tail artery revealed by continuous amperometry. Five stimuli per panel at 0.1 Hz. Reproduced from Jackson VM and Cunnane TC (2001) Neurotransmitter release mechanisms in sympathetic neurons: Past, present, and future perspectives. Neurochemical Research 26: 875–889, with permission from Springer Science and Business Media; and Reproduced from Msghina M, Gonon F, and Stja¨rne L (1993) Intermittent release of noradrenaline by single pulses and release during short trains at high frequencies from sympathetic nerves in rat tail artery. Neuroscience 57: 887–890, with permission from Elsevier.

422 Noradrenaline

of the vas deferens, noradrenaline mediates most of its effects through a1-adrenoceptors, which activate phospholipase C, and with increases in inositol-1,4,5trisphosphate (IP3) and diacylglycerol (DAG), respectively, lead to increases in intracellular Ca2þ and contraction and should be independent of an electrical change. Of course, there are exceptions to this rule, the most notable example being the rat anococcygeus, in which increases in intracellular Ca2þ result in an electrical depolarization by opening a Cl channel, but generally under low-frequency stimulations, most tissues show little or no electrical changes on a1-adrenoceptor activation. It would be expected that IP3-generated Ca2þ waves would be detected using Ca2þ-imaging confocal microscopy. Indeed, slow propagating waves in small arteries following nerve stimulation can be attributed to a1-adrenoceptor stimulation because they are blocked by prazosin, an a1-adrenoceptor antagonist. Similarly, cholinergic stimulation in the bladder results in waves that are blocked by cyclopentolate, a muscarinic receptor antagonist, which are thought to be generated secondary to muscarinic receptor-induced IP3 synthesis. Such waves rarely occur within the longitudinal layer of the vas deferens upon nerve stimulation, but they can be induced by applying exogenous noradrenaline. We therefore have a paradox: exogenously applied noradrenaline causes Ca2þ waves, whereas neuronally released noradrenaline does not readily produce waves, and yet both cause a contraction that is mediated by a1-adrenoceptors. Once again, the remarkable sympathetic nervous system confronts us with a paradox. The knockout mouse for P2X1 purinoceptors provided the ideal opportunity to study noradrenergic transmission in isolation because P2X1 receptor effectswerelost,andoneofthecompensatorymechanisms for the purinergic ‘shortfall’ was the supersensitivity of the tissue to noradrenaline. Using higher frequency nerve stimulation, which is known to favor the noradrenergic component of release, even in the presence of both yohimbine, to antagonize autoinhibitory a2-adrenoceptors to maximize noradrenaline release, and desipramine, to inhibit the reuptake of noradrenaline and maximize its concentration within the neuroeffector cleft, there were still few Ca2þ transientsduetoneuronalnoradrenalinerelease. The preparations still responded well to exogenously added noradrenaline and appeared to have a larger contraction mediated by a1-adrenoceptors to both neuronal and exogenous noradrenaline. Perhaps neuronally released noradrenaline activates a different population of receptors compared to exogenously applied noradrenaline and these receptors have different pathways. Indeed, prazosin

is a non-subtype selective a1-adrenoceptor antagonist, with a1A-adrenoceptors favoring the IP3-mediated increase in intracellular Ca2þ, whereas a1B- and a1D-adrenoceptors may favor the DAG and protein kinase C pathways of contraction (which do not require such large increases in intracellular Ca2þ); this might explain the apparent differences due to prazosin’s nonselectivity. Of course, there are crossovers between the phospholipase C (IP3 and DAG) and D (DAG only) pathways because the receptors are promiscuous and activate many different intracellular signaling cascades, including the Rho-kinase pathway, which allows contraction without any concomitant increase in Ca2þ within smooth muscle cells. How these results can be explained and how the various pathways and cotransmitters interact remain under investigation. Noradrenaline and its effects remain elusive despite the current sophisticated investigative techniques.

Conclusion We began with noradrenaline being the only neurotransmitter released from postganglionic sympathetic nerves, but investigations into the nature of neurotransmitter release have revealed major new insights into the mechanism by which sympathetic nerves function – intermittent neurotransmitter release, autoinhibition through prejunctional/presynaptic receptors, facilitation of release through prejunctional nicotinic receptors, cotransmission, and perhaps the most startling finding of all, that some varicosities do not release neurotransmitter at all. There is a rich diversity of pre- and postjunctional receptors and interactions that remain to be investigated, and we look forward to the advances that studying the sympathetic nervous system will bring. It is fair to say that in the twenty-first century we remain surprisingly ignorant of how nerves release neurotransmitters and the effects that noradrenaline and its cotransmitters have on the innervated organs. See also: Adenosine Triphosphate (ATP).

Further Reading Blakeley AG and Cunnane TC (1979) The packeted release of transmitter from the sympathetic nerves of the guinea-pig vas deferens: An electrophysiological study. Journal of Physiology 296: 85–96. Brain KL, Jackson VM, Trout SJ, and Cunnane TC (2002) Intermittent ATP release from nerve terminals elicits focal smooth muscle Ca2þ transients in mouse vas deferens. Journal of Physiology 541: 849–862. Brock JA and Cunnane TC (1987) Relationship between the nerve action potential and transmitter release from sympathetic postganglionic nerve terminals. Nature 326: 605–607.

Noradrenaline 423 Brock JA and Cunnane TC (1992) Electrophysiology of neuroeffector transmission in smooth muscle. In: Burnstock G and Hoyle CHV (eds.) The Autonomic Nervous System: Autonomic Neuroeffector Mechanisms, pp. 121–213. Chur, Switzerland: Harwood Academic. Goodman LS, Gilman A, Hardman JG, and Limbird LE (2001) Goodman & Gilman’s the Pharmacological Basis of Therapeutics, 10th edn., New York: McGraw-Hill. Hague C, Chen Z, Uberti M, and Minneman KP (2003) a1Adrenergic receptor subtypes: Non-identical triplets with different dancing partners. Life Science 74: 411–418. Jackson VM and Cunnane TC (2001) Neurotransmitter release mechanisms in sympathetic neurons: Past, present, and future perspectives. Neurochemical Research 26: 875–889. Li F, De Godoy M, and Rattan S (2004) Role of adenylate and guanylate cyclases in b1-, b2-, and b3-adrenoceptor-mediated relaxation of internal anal sphincter smooth muscle. Journal

of Pharmacology and Experimental Therapeutics 308: 1111–1120. Msghina M, Gonon F, and Stja¨rne L (1993) Intermittent release of noradrenaline by single pulses and release during short trains at high frequencies from sympathetic nerves in rat tail artery. Neuroscience 57: 887–890. Rang HP, Dale MM, Ritter JM, and Moore PK (2003) Pharmacology, 5th edn., London: Churchill Livingstone. Wenzel-Seifert K, Liu HY, and Seifert R (2002) Similarities and differences in the coupling of human b1- and b2-adrenoceptors to Gsa splice variants. Biochemical Pharmacology 64: 9–20.

Relevant Website http://www.iuphar.org – International Union of Pharmacology.

Norepinephrine: Adrenergic Receptors D B Bylund, University of Nebraska Medical Center, Omaha, NE, USA ã 2009 Elsevier Ltd. All rights reserved.

Introduction Norepinephrine (also called noradrenaline) is a neurotransmitter in both the peripheral and central nervous systems. Norepinephrine produces many effects in the body, the most notable being those associated with the ‘fight-or-flight’ response to perceived danger. The effects of norepinephrine and a related catecholamine, epinephrine (also called adrenaline), are mediated by the family of adrenergic receptors. The chemical structure of norepinephrine, as shown in Figure 1, indicates that it is a catecholamine, because it has both the catechol moiety consisting of a benzene ring (green) with two hydroxyl groups (red) and an amine (blue). Adrenergic receptors (also called adrenoceptors) are widely distributed throughout the body. There are three major adrenergic receptor types: alpha-1, alpha-2, and beta. Each of these three receptor types is further divided into three subtypes. Adrenergic receptors are seven transmembrane receptors which consist of a single polypeptide chain with seven hydrophobic regions that form alpha-helical structures that span or transverse the membrane. Because the mechanism of action of adrenergic receptors includes the activation of guanine nucleotide regulatory binding proteins (G-proteins), they are also called G-protein-coupled receptors. Stimulation of adrenergic receptors by endogenous catecholamines released in response to activation of the sympathetic autonomic nervous system (as well as by exogenous drugs called adrenergic agonists) results in a variety of effects such as increased heart rate, regulation of vascular tone, and bronchodilatation. In the central nervous system (CNS), adrenergic receptors are involved in many functions including memory, learning, alertness, and the response to stress.

History In the late 1940s, Ulf von Euler in Sweden and Holtz in Germany identified norepinephrine as the neurotransmitter of the mammalian sympathetic nerves and soon thereafter also found it to be a normal constituent of mammalian brain. Norepinephrine was subsequently shown to also be a central neurotransmitter and was visualized in the brain by fluorescent and immunohistochemical techniques.

424

Adrenergic receptors were originally divided into two major types, alpha and beta, based on their pharmacological characteristics (i.e., rank order potency of agonists). Subsequently, the beta adrenergic receptors were subdivided into beta-1 and beta-2 subtypes, and more recently a beta-3 subtype was defined. The alpha adrenergic receptors were first subdivided into postsynaptic (alpha-1) and presynaptic (alpha-2) subtypes. After it was realized that not all alpha receptors with alpha-2 pharmacological characteristics were presynaptic, the pharmacological definition was used. The current classification scheme is based on three major types: alpha-1, alpha-2, and beta. Each of these three receptor types is further divided into three subtypes as shown in Figure 2: alpha-1A, alpha-1B, alpha-1D; alpha-2A, alpha-2B, alpha-2C; and beta-1, beta-2, beta-3. Although adrenergic receptors are found throughout the body in tissues innervated by both the peripheral and CNSs, a few locations are of special interest. Alpha-1 adrenergic receptors are found peripherally in some blood vessels, whereas alpha-2 receptors are located on platelets and on nerve terminals in both the peripheral and CNSs. In the heart, beta-1 receptors predominate, where as in the smooth muscles of the lungs, some blood vessels, and the uterus, beta-2 receptors are relatively abundant.

Neurochemistry of Norepinephrine Biosynthesis of Norepinephrine

Norepinephrine is synthesized in neurons starting with the amino acid tyrosine, which is obtained from the diet and can also be synthesized from phenylalanine. Tyrosine is converted to dihydroxyphenylalanine (DOPA) by the enzyme tyrosine hydroxylase, which is the rate-limiting enzyme for norepinephrine biosynthesis. DOPA, in turn, is converted to dopamine in the cytoplasm. Dopamine, which is also a neurotransmitter, is taken up into vesicles and converted to norepinephrine by the enzyme dopamine beta-hydroxylase. In the adrenal medulla and in a few brain regions, norepinephrine is converted to epinephrine by the enzyme phenylethanolamine N-methyltransferase. Storage, Release, Reuptake, and Metabolism of Norepinephrine

Norepinephrine is stored in vesicles (also called storage granules) in the nerve terminals, which concentrates it and protects it from metabolism, until it is released following nerve stimulation. The major mechanism by which the effects of norepinephrine

Norepinephrine: Adrenergic Receptors 425

are terminated is reuptake back into the nerve terminal by a high-affinity transporter. Norepinephrine can also be metabolized to inactive products. Inhibition of either of these processes results in an increase in the synaptic level of norepinephrine and a prolongation of its effects. The reuptake of norepinephrine is mediated by the norepinephrine transporter (NET). Approximately 80% to 90% of the norepinephrine released into many synapses is cleared by this mechanism. NET belongs to the protein superfamily defined as Naþ/Cl -dependent transporters and has 12 putative transmembrane domains. Some inhibitors of NET are antidepressants and are used in the treatment of clinical depression and other affective disorders. They are also sometimes used to treat anxiety disorders, obsessive–compulsive disorder, attentiondeficit/hyperactivity disorder (ADHD), and chronic neuropathic pain. Several classes of antidepressant drugs inhibit NET, including the tricyclic antidepressants (e.g., imipramine, desipramine, doxepin), serotonin-norepinephrine reuptake inhibitors (SNRIs; e.g., venlafaxine, duloxetine), and the norepinephrine reuptake inhibitors (e.g., atomoxetine, reboxetine). Norepinephrine is metabolized by the enzymes monoamine oxidase and catechol-O-methyltransferase

to 3-methoxy-4-hydroxymandelic acid and 3-methoxy4-hydroxyphenylglycol. These inactive degradation products can be quantitated in tissues, blood, and urine as a measure of norepinephrine turnover.

Ontogeny of Norepinephrine in the Brain Cell bodies expressing tyrosine hydroxylase can be detected by 4 weeks of gestation in humans, and norepinephrine itself is detectable by 5–6 weeks. Norepinephrine levels in humans increase throughout the first trimester, especially from about 2 months of gestation forward. Following the initial increase in spinal norepinephrine, a decrease of 30–40% in concentration occurs between 6 months of gestation and early childhood. Noradrenergic neurons differentiate in rats between gestational day 10 and 13. From this point forward there is a steady differentiation and a nearly linear development of markers for noradrenergic neurons in the CNS, increasing approximately 100- to 1000-fold by adulthood. Thus, although the development of norepinephrine occurs fairly early and is reasonably rapid, it is slower than some of the other neurotransmitter systems, particularly the serotonin system.

Effects of Norepinephrine Effects Mediated by the Autonomic Nervous System

OH (R)

NH2

The autonomic nervous system (also called the involuntary nervous system) is divided into two components, the sympathetic nervous system and the parasympathetic nervous system. The final nerves (postganglionic) in the sympathetic system are

HO OH Figure 1 The structure of norepinephrine.

Adrenergic receptors

Alpha-1

Alpha-1A

Alpha-2

Alpha-1B

Alpha-1D

Alpha-2A Figure 2 The classification of the adrenergic receptors.

Alpha-2B

Beta

Beta-1

Alpha-2C

Beta-2

Beta-3

426 Norepinephrine: Adrenergic Receptors

adrenergic and thus release norepinephrine in the various tissues (end organs). The adrenal medulla, which is part of the sympathetic system, releases epinephrine into the circulation. The activation of the sympathetic system, in response to perceived danger, results in the release of large quantities of norepinephrine and epinephrine. Norepinephrine acting at alpha-1 receptors causes vasoconstriction (contraction) of cutaneous blood vessels, whereas epinephrine acting at beta-2 receptors in the blood vessels of the skeletal muscles causes vasodilatation (relaxation), resulting in blood flow being increased in the muscles. Norepinephrine and epinephrine acting at beta-1 receptors increase the force and rate of contraction of the heart, whereas epinephrine acting at beta-2receptors causes bronchodilatation in the lungs and relaxation of the smooth muscle in the uterus. Effects Mediated by the CNS

In the CNS, the cell bodies of noradrenergic neurons are found primarily in the locus coeruleus in the brain stem. These neurons, however, project widely throughout the brain and spinal cord. The locus coeruleus, and hence norepinephrine, is an important regulator of a variety of physiologic functions, including sleep/wake cycles, attention, orientation, mood, memory, and cardiovascular as well as other autonomic and endocrine functions. Although adrenergic receptors are found throughout the brain, the alpha-2 receptors in the CNS are of particular importance because they regulate the release of norepinephrine, as well as many other neurotransmitters.

Alpha-1 Adrenergic Receptors Three genetic and four pharmacological alpha-1 adrenergic receptor subtypes have been defined. The alpha-1A and alpha-1B subtypes were initially defined based on their differential affinity for adrenergic agents such as WB4101 and phentolamine and their differential sensitivities to the site-directed alkylating agent chloroethylclonidine. The alpha-1B subtype was subsequently cloned from the hamster, and the alpha-1A was cloned from bovine brain, although it was originally called the alpha-1c adrenergic receptor. A third subtype, the alpha-1D adrenergic receptor, was subsequently cloned from the rat cerebral cortex, although this clone was originally called the alpha-1a subtype by some investigators. A fourth pharmacological subtype, the alpha-1L, has been identified in vascular tissues from several species but may represent a conformational state of the alpha-1A receptor. The current classification scheme includes

the alpha-1A, the alpha-1B, and the alpha-1D, but there is no alpha-1C (Figure 2). Pharmacological and Molecular Characteristics of Alpha-1 Adrenergic Receptors

In addition to norepinephrine and epinephrine, alpha-1 receptors are activated by agonists as phenylephrine and methoxamine. These agonists are relatively selective for alpha-1 receptors and have lower affinity for alpha-2 and beta receptors. By contrast, they have similar affinities for the three alpha-1 subtypes and are thus non-subtype-selective agonists. Similarly, antagonists including prazosin and tamsulosin are relatively selective for alpha-1 receptors and block alpha-2 and beta receptors only at high concentrations. Several other antagonists such as phentolamine and phenoxybenzamine block both alpha-1 and alpha-2 adrenergic receptors with similar affinities. Alpha-1A-selective antagonists include 5-methylurapidil and niguldipine, whereas cirazoline appears to be a selective alpha-1A agonist. The alpha-1 adrenergic receptors are single polypeptide chains of 446–572-amino-acid residues that span the membrane seven times, with the N-terminus being extracellular and the C-terminus intracellular. Thus, there are three intracellular loops and three extracellular loops. In contrast to the alpha-2 receptors, but similar to the beta receptors, the alpha-1 receptors have a long C-terminal tail (137–179amino-acid residues) and a short third intracellular loop (68–73-amino-acid residues) The C-terminal tails also have multiple sites of phosphorylation which are thought to be important in the desensitization, recycling, and downregulation of the receptor. The human alpha-1 adrenergic receptor genes consist of two exons and a single large intron of at least 20 kbp in the region corresponding to the sixth transmembrane domain. No splice variants are known for the alpha-1B and alpha-1D subtypes. By contrast, at least ten splice variants of human alpha-1A subtype have been reported, but only four produce full-length receptors.

Alpha-2 Adrenergic Receptors Three genetic and four pharmacological alpha-2 adrenergic receptor subtypes have also been defined (Figure 2). The alpha-2A and alpha-2B subtypes were initially defined based on their differential affinity for adrenergic agents such as prazosin and oxymetazoline. These subtypes were subsequently cloned from human, rat, and mouse. A third subtype, alpha-2C, was originally identified in an opossum kidney cell line and

Norepinephrine: Adrenergic Receptors 427

has also been cloned from several species. A fourth pharmacological subtype, the alpha-2D, has been identified in the rat, mouse, and cow. This pharmacological subtype is a species ortholog of the human alpha-2A subtype and thus is not considered to be a separate genetic subtype. Pharmacological and Molecular Characteristics of Alpha-2 Adrenergic Receptors

In addition to norepinephrine and epinephrine, alpha-2 receptors are activated by clonidine and brimonidine. These agonists are relatively selective for alpha-2 receptors and have lower affinity at alpha-1 and beta receptors. Similarly, the antagonist yohimibine is relatively selective for alpha-2 receptors and blocks alpha-1 and beta receptors only at higher concentrations. Antagonists that are at least somewhat selective for one of the alpha-2 subtypes include BRL44408 for the alpha-2A, prazosin and ARC-239 for the alpha-2B (note, however, that these two agents have much higher affinities for alpha-1 receptors), and rauwolscine for the alpha-2C subtype. Ozymetazoline is a partial agonist that has a higher affinity for the alpha-2A subtype as compared to the alpha-2B and alpha-2C subtypes. The alpha-2 adrenergic receptors are single polypeptide chains of 450–462-amino-acid residues. In contrast to the alpha-1 and beta receptors, the alpha-2 receptors tend to have long third intracellular loops (148–179-amino-acid residues) and a short C-terminal tail (20–21-amino-acid residues). The third intracellular loops have multiple sites of phosphorylation which are thought to be important in the desensitization, recycling, and downregulation of the receptor. The alpha-2 adrenergic receptor genes do not contain introns, and thus there are no splice variants.

Pharmacological and Molecular Characteristics of Beta Adrenergic Receptors

Isoproterenol is the prototypic non-subtype-selective beta agonist which has little effect at alpha-1 and alpha-2 receptors. Epinephrine is 10- to 100-fold more potent at the beta-2 receptor as compared to the beta-1 subtype, whereas norepinephrine is more potent than epinephrine at the beta-3 subtype. Many beta-2selective agonists, such as terbutaline and salmeterol, have been developed for the treatment of asthma. Due to their subtype selectivity they have a lower incidence of side effects mediated by the beta-1 receptor. Propranolol is the prototypic non-subtype-selective beta antagonist which has equal affinities at the beta-1 and beta-2 subtypes. Other nonselective beta adrenergic antagonists include timolol, pindolol (which is actually a weak partial agonist), and carvedilol, which is also an alpha-1 antagonist. Several beta-1-selective antagonists have been developed, such as metoprolol and esmolol. The beta adrenergic receptors are single polypeptide chains of 408–477-amino-acid residues. In contrast to the alpha-2 receptors, but similar to the alpha-1 receptors, the beta receptors tend to have longer C-terminal tails (61–97-amino-acid residues) and shorter third intracellular loops (54–80-amino-acid residues). The C-terminal tails have multiple sites of phosphorylation which are thought to be important in the desensitization, recycling, and downregulation of the receptor. The beta-1 and beta-2 adrenergic receptor genes do not contain introns, and thus they have no splice variants. By contrast, the beta-3 receptor has one intron, resulting in two splice variants. However, no functional differences have been found between the two splice variants.

Regulation of Adrenergic Receptors Beta Adrenergic Receptors Three beta adrenergic receptor subtypes have been identified. The beta-1 adrenergic receptor, the dominant receptor in heart and adipose tissue, is equally sensitive to epinephrine and norepinephrine, whereas the beta-2 adrenergic receptor, responsible for relaxation of vascular, uterine, and airway smooth muscle, is less sensitive to norepinephrine than to epinephrine. The beta-3 receptor is insensitive to the commonly used beta adrenergic receptor antagonists and was previously referred to as the ‘atypical’ beta adrenergic receptor. A beta-4 receptor has been postulated; however, definitive evidence of its existence is lacking, and it is now thought to be a ‘state’ of the beta-1 adrenergic receptor.

The processes involved in desensitization and downregulation have been extensively investigated for the beta-2 adrenergic receptor. The other adrenergic receptors, as well as many other G-protein-coupled receptors, appear to behave in a similar, although not identical, manner. Initial uncoupling of the beta-2 receptor from the G-protein after agonist binding is mediated by phosphorylation of specific residues in the carboxyl tail of the receptor. The phosphorylated beta-2 receptor serves as a substrate for the binding of b-arrestin, which not only uncouples the receptor from the signal transduction process but also serves as an adapter protein that mediates the binding of additional signaling proteins and entry into the internalization pathway. The mechanisms of beta-2 adrenergic receptor downregulation appear to involve both

428 Norepinephrine: Adrenergic Receptors

an increase in the rate of degradation of the receptor and a decrease in the levels of beta receptor mRNA.

Adrenergic Receptor Signal Transduction Pathways The alpha-1 adrenergic receptors activate the Gq/11 family of G-proteins, leading to the dissociation of the a and bg subunits and the subsequent stimulation of the enzyme phospholipase C. This enzyme hydrolyzes phosphatidylinositol in the membrane, producing inositol trisphosphate (IP3) and diacylglycerol. These molecules act as second messengers mediating intracellular Ca2þ release via the IP3 receptor and activating protein kinase C. Other signaling pathways that have also been shown to be activated by alpha-1 receptors include Ca2þ influx via voltage-dependent and independent calcium channels, arachidonic acid release, and activation of phospholipase A2, phospholipase D activation, and mitogen-activated protein kinase. The alpha-2 adrenergic receptors activate the Gi/o family of G-proteins and alter (classically inhibit) the activity of the enzyme adenylate cyclase, which in turn, decreases the concentration of the second messenger cyclic AMP. In addition, the stimulation of alpha-2receptors can regulate several other effector systems including the activation of Kþ channels, inhibition or activation of Ca2þ channels, and activation of phospholipase A2, phospholipase C, and Naþ/Hþ exchange. The beta adrenergic receptors activate the Gs family of G-proteins and activate adenylate cyclase, thus increasing in cyclic adenosine monophosphate (AMP) concentrations. Beta adrenergic receptors interact with many other signaling proteins, including the phosphoprotein EBP50 (ezrinradixin-moesin-binding phosphoprotein-50), the Naþ/Hþ exchanger regulatory factor, and cyclic nucleotide ras guanine-nucleotide exchange factor (CNrasGEF). Traditionally it has been assumed that intrinsic efficacy of a ligand for the adrenergic receptors is a property of the ligand. However, recent evidence has shown that certain functionally selective ligands may act as either agonists or antagonists for different functions mediated by the same receptor. This functional selectivity likely involves differences in ligand-induced intermediate conformational states, the diversity of G-proteins, scaffolding and signaling partners, and receptor oligomers.

Adrenergic Receptor Polymorphisms Polymorphisms have been identified in some of the alpha-2 and beta adrenergic receptor subtypes, which may have important clinical implications. A common polymorphism has been identified in the third

intracellular loop of the alpha-2B receptor, which consists of a deletion of three glutamate residues (301–303) and is a risk factor for acute coronary events, but not hypertension. This deletion results in a loss of short-term agonist-induced desensitization. The most common polymorphism associated with the human alpha-2C receptor is the alpha-2C-del (residues 322 to 325 are missing), which occurs in 40% of Blacks and 4% of those from other ethnic backgrounds. As the alpha-2C-del receptor appears to be less efficiently coupled to G-proteins, it is less able to mediate a response to agonists. The consequence of this in humans is an impaired feedback inhibition of catecholamine release that is associated with elevated blood pressure and an increased risk of heart failure. The gene encoding the human beta-1 adrenergic receptor is quite polymorphic, with 18 single nucleotide polymorphisms (SNPs), seven of which cause amino acid substitutions. A total of 13 polymorphisms in the beta-2 adrenergic receptor gene and its transcriptional regulator upstream peptide have been identified. Three closely linked polymorphisms, two coding regions at amino acid positions 16 and 27 and one in the upstream peptide, are common in the general Caucasian population. The glycine-16 receptor exhibits enhanced downregulation in vitro after agonist exposure. In contrast, arginine-16 receptors are more resistant to downregulation. Some studies have suggested a relationship among these polymorphisms, airway responsiveness (e.g., asthma), and the responsiveness to beta adrenergic agonists. A tryptophan-64 to arginine polymorphism has been identified in the beta-3 adrenergic receptor. The allele frequency is approximately 30% in the Japanese population, higher in Pima Indians, and lower in Caucasians. Type 2 diabetic patients with this mutation showed a significantly younger age of onset of diabetes and an increased tendency to obesity, hyperinsulinemia, and hypertension.

Drugs Which Mimic or Block the Effects of Norepinephrine Norepinephrine itself is rarely used as a drug due to its rapid metabolism and many sites of action. Drugs which evoke responses similar to sympathetic nerve stimulation are called sympathomimetic drugs. They produce their effects either directly by stimulating adrenergic receptors (adrenergic receptor agonists) or indirectly by promoting the release of norepinephrine or by blocking its reuptake. Table 1 gives examples of various sympathomimetic drugs with their mechanism of action, the effects they produce, and their therapeutic indications.

Norepinephrine: Adrenergic Receptors 429 Table 1 Drugs which mimic the effects of norepinephrine Receptor

Examples of drugs

Effect

Therapeutic indication

Alpha-1 Alpha-2 Beta-1 Beta-2 (Indirect acting) (Indirect acting) (Indirect acting)

Phenylephrine Clonidine, brimonidine Dobutamine Albuterol, terbutaline, ritodrine Amphetamine Desipramine, tomoxetine Phenelzine

Vasoconstriction Lowers fluid pressure Increases heart rate Relaxes smooth muscle in lung and uterus CNS stimulant Blocks norepinephrine reuptake Inhibits monoamine oxidase inhibitor (MAO)

Hypotension, nasal congestion Hypertension, glaucoma Cardiovascular shock Asthma, premature labor Narcolepsy, hyperactivity Depression Depression

Table 2 Drugs which block the effects of norepinephrine Receptor

Examples of drugs

Effect

Therapeutic indication

Alpha-1 Alpha-2 Beta Beta-1

Prazosin, terazosin Mirtazapine Propranolol, timolol Atenolol, metoprolol

Vasodilatation Antidepressant Decreases heart rate Decreases heart rate

Hypertension, benign prostatic hypertrophy Depression Hypertension, angina, glaucoma Hypertension

Drugs which block the responses to sympathetic nerve stimulation and thus block the effects of norepinephrine are called adrenergic receptor antagonists. They have no direct effects of their own, but act by blocking the effects of either released norepinephrine or an administered adrenergic receptor agonist. Table 2 gives examples of adrenergic receptor antagonists (also called adrenergic blockers) with their receptor selectivity, the effects they produce, and their therapeutic indication. See also: Monoamines: Release Studies; Norepinephrine: CNS Pathways and Neurophysiology.

Further Reading Bylund DB (1988) Subtypes of alpha-2 adrenoceptors: Pharmacological and molecular biological evidence converge. Trends in Pharmacological Science 9: 356–361. Bylund DB, Eikenberg DC, Hieble JP, et al. (1994) International Union of Pharmacology nomenclature of adrenoceptors. Pharmacology Review 46: 121–136. Cooper JR, Bloom FE, and Roth RH (2003) Norepinephrine and epinephrine. In: The Biochemical Basis of Neuropharmacology, 8th edn., pp. 181–223. New York: Oxford University Press. Eisenhofer G, Kopin IJ, and Goldstein DS (2004) Catecholamine metabolism: A contemporary view with implications for physiology and medicine. Pharmacology Reviews 5: 331–349. Ferguson SSG (2001) Evolving concepts in G protein-coupled receptor endocytosis: The role in receptor desensitization and signaling. Pharmacology Reviews 53: 1–24. Frazer A (2000) Norepinephrine involvement in antidepressant action. Journal of Clinical Psychiatry 61(supplement 10): 25–30. Gordon M Autonomic (ANS) pharmacology: Introduction. In: Medical Pharmacology and Disease-Based Integrated,

Instruction, Ch. 4. http://www.pharmacology2000.com/Autono mics/Introduction/Introobj1.htm(accessed April 2007). Hieble JP, Bylund DB, Clarke DE, et al. (1995) International Union of Pharmacology. X: Recommendation for nomenclature of alpha-1 adrenoceptors: Consensus update. Pharmacology Reviews 47: 267–270. Hokfelt T, Johansson O, and Goldstein M (1984) Central catecholamine neurons as revealed by immunohistochemistry with special reference to adrenaline neurons. In: Bjo¨rklund A and Hokfelt T (eds.) Handbook of Chemical Neuroanatomy, Vol. 2: Part 1: Classical Transmitters in the CNS. Amsterdam: Elsevier. Murrin LC, Sanders JD, and Bylund DB (2007) Comparison of the maturation of the adrenergic and serotonergic neurotransmitter systems in the brain: Implications for differential drug effects on juveniles and adults. Biochemical Pharmacology 73: 1225–1236. Neumeister A, Charney DS, Belfer I, et al. (2005) Sympathoneural and adrenomedullary functional effects of alpha-2C adrenoreceptor gene polymorphism in healthy humans. Pharmacogenetics and Genomics 15: 143–149. Rockman HA, Koch WJ, and Lefkowitz RJ (2002) Seventransmembrane-spanning receptors and heart function. Nature 415: 206–212. Small KM, McGraw DW, and Liggett SB (2003) Pharmacology and physiology of human adrenergic receptor polymorphisms. Annual Review of Pharmacology and Toxicology 43: 381–411. Urban JD, Clarke WP, von Zastrow M, et al. (2007) Functional selectivity and classical concepts of quantitative pharmacology. Journal of Pharmacology and Experimental Therapeutics 320: 1–13.

Relevant Websites http://www.adrenoceptor.com – Adrenoceptor Online. http://pubchem.ncbi.nlm.nih.gov – PubChem Compound Summary: PubChem Compound 5814.

Norepinephrine: CNS Pathways and Neurophysiology G Aston-Jones, Medical University of South Carolina, Charleston, SC, USA C A Meijas-Aponte, National Institutes of Health, Baltimore, MD, USA B Waterhouse, Drexel University College of Medicine, Philadelphia, PA, USA ã 2009 Elsevier Ltd. All rights reserved.

Norepinephrine Cell Groups and Pathways Locus Coeruleus

The nucleus locus coeruleus (LC) is a dense cluster of noradrenergic neurons in the dorsorostral pons. This nucleus gained prominence in the 1960s when new anatomical approaches revealed it to be the major source of norepinephrine (NE) in brain with projections throughout most central nervous system (CNS) regions. These findings stimulated a great deal of research into this system, resulting in a wealth of knowledge at the cellular, systems, and behavioral levels. Considerable evidence indicates that these neurons are important in a variety of cognitive, affective, arousal, and other behavioral functions, as well as associated clinical dysfunctions. Locus coeruleus efferent projections As summarized in Table 1, LC NE neurons project throughout the neuraxis, including the cerebral cortex, thalamus, cerebellum, midbrain, and spinal cord. In this regard, it is different from other ‘typical’ brain systems that have very limited projection targets (e.g., thalamocortical projections). Despite this high degree of divergence, LC neurons exhibit substantial regional and laminar specificity in their efferent projections. Notably, brain areas in primates that are associated with attentional processing (e.g., parietal cortex, pulvinar nucleus, and superior colliculus) have particularly dense LC projections and NE receptors. Notable areas that receive little or no input from the LC include the caudate putamen and most of the hypothalamus. Ultrastructural evidence indicates that these neurons often make conventional synaptic appositions with postsynaptic targets, although the frequency of such synaptic contacts is controversial. Extrasynaptic (paracrine) release of NE from LC terminals also appears to be possible. Locus coeruleus afferents A variety of evidence indicates that the LC receives a wealth of different neurotransmitter inputs. This nucleus is densely innervated by fibers that contain opiates, glutamate,

430

g-aminobutyric acid (GABA), serotonin, epinephrine, histamine, and the peptide orexin/hypocretin. The sources of these various inputs have not been fully elucidated, though some major inputs have been identified. The nucleus paragigantocellularis lateralis (PGi) in the ventrolateral rostral medulla, a major input that strongly excites LC neurons, is a source for glutamate, GABA, enkephalin, corticotropin-releasing hormone (CRH), and epinephrine. A strongly inhibitory GABA and enkephalin input originates from the dorsomedial rostral medulla. Excitatory orexin and histamine inputs originate in the hypothalamus. The shell of dendrites surrounding the LC nucleus offers additional extensive targets for afferent termination, and indeed it appears that several areas target these extranuclear dendrites more so than the LC nucleus proper. Thus, projections from the periaqueductal gray matter, parabrachial region, preoptic area, amygdala, and medial prefrontal cortex, among other sites, project primarily to the peri-LC region, with relatively little input to the LC proper. Although some of these projections have been shown to contact LC dendrites, additional ultrastructural studies are needed to examine others. A recent study has extended our understanding of LC afferents to the circuit level. Using transynaptic tracing, lesions, and electrophysiology, Aston-Jones and colleagues determined that the suprachiasmatic nucleus (SCN) indirectly innervates the LC, using the dorsomedial hypothalamus as a relay. Based upon the role of the SCN as the brain’s major circadian pacemaker and the role of the LC in arousal and performance (see later), these findings reveal a possible circuit for circadian regulation of arousal and performance. Recent work by Iba, Aston-Jones, and colleagues reveal that in monkey strong projections to LC originate in cingulate and orbital cortices (Figure 1). Given the roles of these prefrontal areas in conflict monitoring, decision making, and other cognitive functions, these projections may be responsible at least in part for the different modes of activity that occur in LC neurons during cognitive performance (described later). Interestingly, these prefrontal projections to LC appear to be stronger in primates than in rats, in concert with the more highly evolved prefrontal cortex and more prominent role of this structure in regulating behavior in primates. Taken together, these observations indicate that LC is not a relay nucleus for primary sensory or motor information and that it receives highly processed information concerning internal and external sensory stimuli as well as behavioral and affective states. This suggests a highly integrative function for this system.

Norepinephrine: CNS Pathways and Neurophysiology 431 Table 1 Organization of the adrenergic system Fiber density

Forebrain Cortex Hippocampal formation Amygdala Medial Central Basal Basal forebrain/septum Medial septum Medial preoptic area Substatia innominata Lateral septum Bed nucleus stria terminalis Ventral pallidum Zona incerta Subthalamic nucleus Nucleus accumbens Thalamus Sensory Lateral geniculate nu Medial geniculate nu Ventroposterior medial nu Ventroposterior lateral mu Ventrobasal nu Ventromedial nu Reticular nu Motor Ventrolateral nu Ventralanterior nu

Locus coeruleus A4 þ A6

2þ 2–3þ

a a

1þ 3þ 3þ

a a

2þ

3þ 5þ

a a a a a

Dorsal medulla cluster

A1

C1

A5

A2

C2

C3

a a

a

a

a a

a

a

a a a a a a a

a a a a a a

a

a a a a

1–3þ

3–4þ 1–2þ

A7

a a a a a

a a a a

a

3–4þ 3þ 3–4þ

a a a a a a a

3þ 3þ

a a

Viscero-limbic Anteriormedial nu Anteriodorsal nu Anteriovental Mediodorsal nu Central medial nu Central lateral nu Reuniens nu Parafascicular nu Paraventrivular nu Rhomboid nu

4þ 3þ 2þ 1–2þ 1þ 1–2þ 1þ 1þ2 5þ 2–3þ

a a a a a a a a *

Epithalamus Medial habenulla Lateral habenulla

5þ 4–5þ

* *

Hypothalamus Periventricular nucleus Suprachiamatic nucleus Retrochiasmatic Paraventricular nuclues Supraoptic nuclues Dorsomedial Arcute nuclues Medial eminence Perifornical area Lateral area

5þ 1–2þ 4þ 3–5þ 4þ 3–5þ 3þ 2þ 4þ 3þ

a *

2þ

a

Brain stem Sensory Main sensory trigeminal nu

Ventral pons and medulla cluster

a a a a

a a a a a a a

a

a

a

a

a a a

a a a a a

a

Continued

432 Norepinephrine: CNS Pathways and Neurophysiology Table 1 Continued Fiber density

Spinal trigeminal nu Mesencephalic trigeminal nu Superior olivary complex Cochlear nu Vestibular nu

2þ 4þ 1þ 2þ 0–1þ

Motor Oculomotor and trochlear nu Motor trigeminal nu Abducens nu Facial nu Hypoglossal nu Nucleus ambiguus Inferior olive Red nucleus Cerebellum

0–1þ 4þ 0–1þ 3–4þ 2–3þ 5þ 5þ 1þ

Sensorimotor Superior colliculus Superficial layers Deep layers Inferior colliculus Visceral Nucleus of the solitary tract including A2 and C2 areas Dorsal motor nucleus of vagus Caudal ventrolateral medulla including A1 area Rostral ventrolateral medulla including C1 area Periaqueductal gray Parabrachial nucleus Area postrema Modulatory systems Substantia nigra, pars compacta Ventral tegmental area Retrorubral field Raphe Central linear Dorsal Magnun Pallidus Obscurus Locus coeruleus A5 area Spinal cord Intermediolateral cell column Dorsal horn Ventral horn

Locus coeruleus A4 þ A6

a a a a a

a a a a

Ventral pons and medulla cluster

Dorsal medulla cluster

A1

A2

C1

A5

A7

C2

C3

a a

a a

a

a

a a

a a

a a

a a

2þ

a

a a a

2þ 5þ

a

5þ 3þ

a

a

a a

a a

3þ

a

2þ5 4þ 1þ

a a a

a a a

a a a

a a

0–1þ 1þ 1þ 2þ 2þ 3þ 5þ 1þ 2þ

a

a

a a

a a a a

a a a

a a

a a a a

a

a a a **

a

a

a

a

This table is based on the organization of LC versus non-LC systems as originally reviewed by Moore and Card in 1984 and published in the Handbook of Chemical Neuroanatomy, volume 2. The descriptions of fiber density were obtained from Moore and Card, whereas the information regarding efferent projections was updated based upon a comprehensive survey of research reports available in the literature. The information presented is meant to reflect major efferents of the adrenergic cell groups. Single asterisks indicate structures known to not receive LC afferents but for which the source of adrenergic innervation is unknown. The double asterisk indicates ventral horn innervation by LC, which is dependent upon rat strain.

Norepinephrine release Microdialysis methods have recently provided in vivo confirmation that certain drugs and behavioral conditions increase or decrease the release of norepinephrine from locus

coeruleus axons. For example, reuptake blockers and stress both enhance extracellular NE levels in areas (such as the neocortex) that receive their only NE innervation from the LC. In addition, the

Norepinephrine: CNS Pathways and Neurophysiology 433 A24

A25

CC

A26

A27

CC

a

A22

CC

A23

CC

A24

CC

b Figure 1 Plots of retrogradely labeled neurons in the nucleus accumbens and orbitofrontal cortex after injections of cholera toxin B into monkey locus coeruleus. (a) Accumbens neurons labeled from the monkey locus ceruleus. Lower sections are high-power views containing plotted cells; upper sections are low-power views to give orientation. (b) Orbitofrontal cortex neurons labeled from the monkey locus coeruleus. Sections on the right are high-power views containing plotted cells; sections on the left are low-power views to give orientation. For both panels, labels A22–A27 refer to distances (in millimeters) from the interaural line. CC, corpus callosum. From Aston-Jones G and Cohen JD (2005) Adaptive gain and the role of the locus coeruleus–norepinephrine system in optimal performance. Journal of Comparative Neurology 493: 99–110.

amount of NE release in cortex increases in proportion to activity within LC, even at low frequencies (2–4 Hz). Notably, burst stimulation elicits more NE release per impulse than evenly do spaced stimuli. This indicates that phasic LC responses (see later) may be particularly effective in releasing NE in target areas. Locus coeruleus impulse activity Sleep and waking. Spontaneous LC activity in rats, cats, or monkeys varies with the stage of the sleep–wake cycle, with firing being most rapid during wakefulness, slower during slow-wave sleep, and virtually absent during paradoxical sleep. LC neurons are phasically

activated by conspicuous unconditioned stimuli. Notably, stimuli that elicit large LC responses in either rats or monkeys also typically disrupt sleep or ongoing behavior, and evoke waking or a behavioralorienting response. The same stimuli do not disrupt behavior if they elicit small LC responses. Stress. Other studies have revealed strong activation of LC neurons by stressors. Stimuli such as sciatic nerve activation or other painful events strongly activate LC cells. Other stressors, such as a puff of air in the awake monkey or a variety of environmental or physiological stressors, also activate LC neurons. In addition, LC neurons are activated by corticotropin-releasing hormone, which mediates the response of LC cells to certain physiological stressors such as hypotension. Locus coeruleus activity during cognitive performance. The preceding results suggest that, in addition to a role in sleep and waking, LC neurons may be involved in behavioral responses to sensory stimuli. Aston-Jones and colleagues tested this by recording impulse activity of LC neurons in monkeys performing a target detection task. These studies indicated that the LC exhibits two modes of impulse activity: a tonic mode in which tonic activity is high but phasic responses to targets are low or absent, and a phasic mode in which LC neurons fire tonically more slowly but exhibit consistent activation shortly following target stimuli. Behaviorally, the tonic mode corresponds to poor task performance with many false alarm errors, whereas the phasic mode is associated with nearly error-free performance. These results indicate that these different modes of LC activity may facilitate either selective, focused attention (phasic mode) or flexible behavior and scanning (tonic mode). Each of these LC and behavioral modes is adaptive, depending on the context. LC activation at decision completion. Recent studies recording LC neural activity in monkeys performing a forced-choice task revealed that the aforementioned phasic responses described are neither purely sensory nor motor/premotor in nature. As shown in Figure 2, these LC responses are more closely linked to behavioral responses (lever releases) than to sensory stimuli that trigger the behavioral responses, but do not occur for lever releases elicited outside of the task context. Finally, the onset of these LC responses occurs at about 200 ms prior to the lever releases, at a time that the decision concerning lever release is being made. Aston-Jones and colleagues concluded from these results that these phasic LC responses reflected completion of a decision process, and that the modulatory gain-enhancing effects of the associated NE release (see later) facilitate brain processes and behaviors called upon by the decision just

434 Norepinephrine: CNS Pathways and Neurophysiology

9

Correct trials Incorrect trials

8

Omissions

7

Spikes s−1

6 5 4 3 2 1

Lever release

0 −0.4

0.0

−0.2

a

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Stimulus Correct trials 9

Incorrect trials

8

Bar releases not assoc. with stimuli

7

Spikes s−1

6 5 4 3 Stimulus onset

2 1 0 −0.6

b

−0.4

−0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Response Time ( s )

Figure 2 Stimulus- and response-locked population perievent time histograms (PETHs) showing LC responses to cues for trials yielding correct and incorrect behavioral responses. (a) Stimulus-locked population PETHs showing LC response to cues (presented at time 0) for trials yielding correct or incorrect behavioral responses. Note that the LC response peaks sooner and is less prolonged on correct, compared to incorrect, trials in this analysis (17 533 and 1362 trials, respectively). No fluctuation in LC activity was detected on omission trials (dashed line, 1128 trials). Vertical lines indicate the mean behavioral RTs. Curves below PETHs represent the normalized RT distributions for correct and incorrect trials. (b) The difference in the phasic LC response between correct and incorrect trials is no longer evident on response-locked population PETHs. In addition, no LC response occurred prior to or following lever releases not associated with stimulus presentation (dashed line, 3381 trials). Vertical lines indicate the mean stimulus onset times. From Clayton EC, Rajkowski J, Cohen JD, et al. (2004) Phasic activation of monkey locus coeruleus neurons by simple decisions in a forced choice task. Journal of Neuroscience 24: 9914–9920.

made. In this way, the LC phasic response supports the current task/goal and facilitates focused performance in that task. The projections to LC from orbital and cingulate cortices (described previously) may mediate these phasic LC responses associated with decision processes.

Non-LC Norepinephrine and Epinephrine Cell Groups

Neurons in other brain nuclei aside from those in the LC constitute 55% of all CNS NE and epinephrine (E) cells. These neurons are clustered in two

Norepinephrine: CNS Pathways and Neurophysiology 435

rostrocaudally aligned columns within the brain stem. One column is located primarily in the ventral aspect of the pons and medulla and contains 44% of all NE/E neurons, whereas the second is situated in the dorsomedial medulla. The individual clusters of non-LC NE/E neurons within these cell columns are further subdivided on the basis of their specific anatomical location and the ability to synthesize E (Figure 3). Non-LC NE/E projections An analysis of projection patterns of LC versus non-LC NE/E neurons reveals similarities and differences (Table 1). Non-LC NE neurons project exclusively subcortically and make a preponderance of connections onto motor nuclei in the medulla and spinal cord, and onto structures associated with visceral function. Within the brain stem, targets of non-LC NE/E neurons include cranial motor nuclei such as the dorsal motor vagus, motor trigeminal, motor facial, and the hypoglossal nucleus, as well as visceral–sensory nuclei such as the nucleus of the solitary tract, parabrachial complex, and periaqueductal gray. In the forebrain, non-LC cell clusters provide dense projections to visceral-related structures such as the bed nucleus of the stria terminalis (BNST), central nucleus of the amygdala, and paraventricular hypothalamus. Thalamic and epithalamic targets include the paraventricular nucleus, rhomboid nucleus, and habenula. Within the different non-LC adrenergic nuclei, there is a preferential innervation of motor systems by A5 and A7 groups, whereas visceral–sensory areas are primarily innervated by medullary A1 and A2 nuclei. The E-containing fibers from C1 neurons

AQ

Cer

ebe

llum

Q

V4

A7 Po n

LC s

A5

Hind brai n C3 C2

A2

CC

C1 M e d u l l a A1

Figure 3 Localization of norepinephrine (red) and epinephrine (green) nuclei in the brain stem. The cell column in the ventral pons and medulla includes A1, C1, A5, and A7 cell groups, whereas the dorsal medulla column includes A2, C2, and C3 cell groups.

innervate both motor and visceral–sensory structures. C1 neurons send collaterals that innervate both spinal cord and brain stem targets. However, most C1 neurons that project to hypothalamus do not project to spinal cord or brain stem structures. In clear contrast to this pattern of organization, LC projections to the brain stem and spinal cord are directed almost exclusively to relay sites along ascending sensory pathways, including those associated with somatosensory, auditory, gustatory, and visual systems. This sensory-predominant bias of the LC efferent projection continues to the level of the thalamus and also includes the olfactory system. An overlap between LC and non-LC NE efferent projections to motor structures becomes evident in the cerebellar system, and the LC is the sole source of NE fibers to thalamic motor relay nuclei and motor (as well as sensory) processing regions of the cerebral cortex. Non-LC impulse activity LC and non-LC neurons have similar electrophysiological signatures but exhibit differing responses to afferents and neurotransmitter systems. They are slow-firing neurons (60-fold selective over 5-HT1D and other receptors; Table 1), to increase 5-HT release in the frontal cortex. NAS-181 has been reported to be the only selective antagonist of the rodent 5-HT1B receptor and it may add to our knowledge of the operational characteristics of the rat receptor subtype.

Serotonin (5-Hydroxytryptamine; 5-HT): Receptors 461

Complementary studies in 5-HT1B null mutant (‘knockout’) mice have proved unhelpful in understanding the pharmacology of the 5-HT1B receptor, as it has been suggested that a compensatory adaptation to the constitutive loss of 5-HT1B receptors becomes an important determinant of the altered response of 5-HT1B knockout mice to a variety of pharmacological challenges. Altered expression and functions of serotonin 5-HT1A and 5-HT1B receptors in knockout mice lacking the 5-HT transporter (SERT knockout mice) have also been reported. Although the effects on 5-HT release are unclear from 5-HT1B knockout mice, in vivo studies have shown that stimulation of central postsynaptic 5-HT1B receptors in mice, but not rats, causes hyperlocomotion, and penile erection and hypophagia in rats are also reportedly 5-HT1B receptor mediated. Postsynaptic 5-HT1B receptor activation in another species, the guinea pig, is reported to induce hypothermia, while the hypothermic response to 5-HT1B agonists in the rat remains to be fully characterized. The putative 5-HT1B receptor agonist, anpirtoline, has analgesic, cognitive, and antidepressant-like properties in rodents and it is of interest that mutant mice lacking the 5-HT1B receptor are reported to be both highly aggressive and have an increased preference for alcohol. Great interest in 5-HT1B receptor agonists has been largely generated by the highly successful antimigraine drug sumatriptan, a nonselective 5-HT1D and 5-HT1B receptor agonist with low selectivity for other receptors in functional studies (Table 1). This compound may act either via constriction-mediating 5-HT1B receptors on cerebral arteries or by blocking neurogenic inflammation and nociceptive activity with trigeminovascular afferents. This latter action has been argued to be 5HT1B receptor mediated, as protein extravasation induced by trigeminal ganglion stimulation is blocked by sumatriptan, the nonselective 5-HT1B receptor agonist, by CP-93129 (>100-fold selectivity over 5-HT1A, 5-HT1B, and 5-HT2 receptors; Tables 1 and 2), and by the close structural analog of sumatriptan, CP-122288, in wild-type, but not mutant, mice lacking the 5-HT1B receptor. However, while 5-HT1B receptor mRNA has been detected in rats, only 5-HT1D mRNA has been detected in the guinea pig and human trigeminal ganglia. Thus, in humans, the antimigraine properties of sumatriptan may be either 5-HT1B or 5-HT1D receptor mediated. It is therefore abundantly clear that our understanding of the operational and physiological consequences of the 5-HT1B activation across species is still largely in its infancy. The 5-HT1D(Gi/Go) Receptor

The 5-HT1D receptor has 63% overall structural homology with the 5-HT1B receptor (formerly

5-HT1Db) and a 77% amino acid sequence homology in the seven-transmembrane domains. The receptor is located on human chromosome HR1D/1p34.3–p36.3 and contains 377 and 374 amino acids for the human and rodent gene, respectively. Low levels of the 5-HT1D receptor mRNA are found in the rat brain, predominantly in the caudate putamen, nucleus accumbens, hippocampus, and cortex, but also in the dorsal raphe and locus coeruleus. It has been proposed that neurogenic inflammation and nociceptive activity within the trigeminovascular afferents may be 5-HT1D receptor mediated due to the presence of 5-HT1D, but not 5-HT1B, receptor mRNA in the guinea pig and human trigeminal ganglia. Thus antagonism of plasma extravasation, induced by electrical stimulation of the trigeminal nerves, is observed with the antimigraine 5-HT1B/1D receptor agonists sumatriptan, naratriptan, rizatriptan, and zolmitriptan. Some advocates, however, claim that this is a 5-HT1D receptor-mediated response. Several other pharmacological nonselective tool 5-HT1B/1D ligands have been used extensively to characterize the 5-HT1D receptor. These include sumatriptan, PU109291, L694247 (pKD 10.0), CP-122288, MK 464, and BW311C90, and the more brain penetrant agonists SKF 99101H, GR46611, and GR125743. The location of 5-HT1D receptor mRNA in the raphe suggests that, like the 5-HT1B receptor, it too may function as a 5-HT autoreceptor. Indeed, there is electrophysiological, release, and voltammetric evidence to this effect. These data are further substantiated by the use of ketanserin and ritanserin, which, in addition their high affinity for the 5-HT2A site, also have high affinity for the 5-HT1D, but not for the 5-HT1B, site (Table 1). The antagonist SB714786 (pKD 9.1) and BRL-15572 (pKD 7.9) 5-HT1B/D ligands mediate their effects at autoreceptors involved in the local inhibitory control of 5-HT release and may play a role in the pathogenesis of major depressive disorder (MDD) and in the antidepressant effects of the SSRIs in patients. A number of compounds in this class – including GR125743, SB224289, and L694247; AZD8129, a 5-HT1B receptor antagonist; and CP-448187, a 5-HT1B/1D receptor antagonist – are in phase II studies for depression, and the outcome of these studies is eagerly awaited. The 5-HT1E(Gi/Go) Receptor

The 5-HT1E receptor was first characterized in humans as a [3H]-5-HT binding site in the presence of 5-carboxyamidotryptamine (5-CT) that blocked binding to the 5-HT1A and 5-HT1D receptors. Human brain binding studies have reported that 5-HT1E receptors (representing up to 60% of 5-HT1 receptor binding) are concentrated in the caudate putamen,

462 Serotonin (5-Hydroxytryptamine; 5-HT): Receptors

with lower levels in the amygdala, frontal cortex, and globus pallidus. This is consistent with the observed distribution of 5-HT1E mRNA. The receptor has been mapped to human chromosome HTR1E/6q14–q15, and the gene encodes a 365-amino-acid protein (Table 1). There are no reported selective or highaffinity ligands for this receptor (except for 5-HT) and its function is currently unknown. However, based on its distribution, one could argue that the amygdala is regarded as the ‘emotional engine,’ and dysfunction of the amygdala (and caudate putamen) has been associated with numerous psychiatric and neurological disorders, ranging from epilepsy to anxiety disorders, attention-deficit/hyperactivity disorder (ADHD), and social phobia, to Alzheimer’s disease. Some schools of thought have even proposed that the amygdala is involved in sociopathic and criminal behaviors. Only time will tell, with the development of selective agents for the 5-HT1E receptor. The 5-HT1F(Gi/Go) Receptor

This receptor subtype is most closely related to the 5-HT1E receptor, with 70% sequence homology across the 7TM domains. The gene is located on chromosome HTR1F/3p11, encoding 366 amino acids across human and rodent species. The mRNA coding for the receptor is concentrated in the dorsal raphe, hippocampus, and cortex of the rat and also in the striatum, thalamus, and hypothalamus of the mouse. 5-HT1F receptor mRNA has been detected in human brain and is also present in the mesentery and uterus. Sumatriptan has almost equal affinity for the 5-HT1F (pKi 7.6) and 5-HT1B/1D receptors (pKi 8.4 and 8.1, respectively; Table 1). Thus, it has been hypothesized that the 5-HT1F receptor might be a target for drugs with antimigraine properties. 5-HT1F mRNA has been detected in the trigeminal ganglia, the stimulation of which leads to plasma extravasation in the dura, a component of neurogenic inflammation that is thought to be a possible cause of migraine. The first 5-HT1F receptor selective agonist, LY334370 (pKD 9.4; see Table 2), with >100-fold separation over the 5-HT1B/ 1D receptors, has been claimed to block the effects of trigeminal nerve stimulation, as does sumatriptan, naratriptan, rizatriptan, and zolmitriptan . LY334370 has also been used successfully as a radioligand (KD 0.5 nM) and demonstrated a reasonable correlation between the receptor protein and mRNA distribution, with the highest binding in the cortical areas, striatum, hippocampus, and olfactory bulb. The 5-HT2(Gq11) Receptor Family

5-HT2 receptors are characterized by having a relatively lower affinity for indolealkylamines, including 5-HT, and are linked to the Gq/phospholipase

C pathway of signal transduction. The structural organization of the 5-HT2 receptor has largely been determined by mutagenesis analyses that have identified a number of primary binding residues located mainly on TMs 3, 5, and 6, which are believed to constitute the purported binding pocket. Other residues identified on TMs 2, 3, 6, and 7 have been implicated in receptor activation and G-protein coupling. The 5-HT2 receptor family mediates a large array of physiological and behavioral functions in humans via three distinct subtypes: 5-HT2A, 5-HT2B, and 5-HT2C. While selective 5-HT2A receptor antagonists have been known for some time, knowledge of the precise effects of 5-HT2C and 5-HT2B receptor antagonists has been hampered by the existence of only mixed receptor antagonists for 5-HT2A, 5-HT2B, and 5-HT2C. However, selective 5-HT2A, 5-HT2C, and 5-HT2B receptor antagonists have emerged over the past 5–10 years. Indeed, several structural classes belonging to the various pharmacophores have been reported. The 5-HT2A Receptor Subtype

The 5-HT2A receptor subtype refers to the classical ‘D’ 5-HT2 receptor as defined by the neuroleptic [3H]spiperone in rat brain frontal cortex. Much of the early pharmacology of the 5-HT2A receptor was pioneered using the selective ligand, [3H]ketanserin, which made an important contribution to the characterization and localization of 5-HT2A receptors. Since the early ligand binding studies, only a few compounds have shown over all selectivity for the 5-HT2A site. Of these compounds, MDL100097 possesses subnanomolar affinity for the 5-HT2A receptor (pKB 9.4), with a minimum 300-fold selectivity for the 5-HT2A site (pKB 8.5–9.5) over other receptors examined, including the 5-HT2C receptor. Other less selective compounds include ICI 169369, ICI 170809, RP62203 (fananserin), and SR46349B (20- to 30-fold selective). The recent introduction of more selective 5-HT2A receptor antagonists (e.g., APD-125) will greatly accelerate the clinic potential of this receptor subtype. However, as yet, no selective agonists exist to provide a more precise classification of this receptor. Only a-methyl-5-HT has been shown to be a useful selective agonist for identifying 5-HT2A/2C/2B receptors (Table 2) Other (nonselective) compounds with agonist properties at 5-HT2 receptors include 1-(2,5-dimethoxy-4-methylphenyl)-2-amino propane (DOM) and its 4-bromophenyl (DOB) and 4-iodophenyl (DOI) congeners. DOI is the preferred radioligand, albeit that it is relatively nonselective. The actions elicited via 5-HT2A receptors comprise mostly those previously considered to be mediated by postsynaptic ‘D’ receptor and include vascular

Serotonin (5-Hydroxytryptamine; 5-HT): Receptors 463

smooth muscle contraction in several preparations (e.g., rabbit aorta, rat caudal artery, dog gastrosplenic vein); contraction of bronchial, uterine, bladder, and some gastrointestinal smooth muscle; platelet aggregation; increased capillary permeability; and some behavioral syndromes in rodents, such as head twitch, forepaw treading, and coarse body tremors induced by 5-HT agonists. In humans, hallucinations and vivid dreams have been reported to be induced by 5-HT2A receptor stimulation. The contribution of 5-HT neurotoxins in understanding the role of 5-HT2A receptors and the mechanisms of action of hallucinogenic amphetamine analogs such as 3,4-methylenedioxymethamphetamine (MDMA) have implicated the 5-HT2A/2C receptors in psychiatric disorders. Several nonselective 5-HT2A/5-HT2C antagonists under investigation have failed to show any clinical utility for the treatment of several psychiatric disorders, including schizophrenia, anxiety, neurotic depression, migraine, sleep disorders, and drug abuse. However, a number of compounds have been shown to have therapeutic potential in cardiovascular disorders and migraine prophylaxis. As considered earlier, the striking similarities between 5-HT2A and 5-HT2C receptors with respect to amino acid sequence (70% in 7TM domains) holds true for the signal transduction mechanism, as both receptors stimulate phosphatidylinositol (PI) turnover. However, the receptors have distinct chromosomal localizations (HTR2A/13q14–q21 and HTR2C/Xq24). The characterization of the 5-HT2A receptor-coupled signal transduction system was first studied in human platelets. It is likely that the biochemical events mediated by the 5-HT2A receptor are important for both the shape change and the aggregation of human platelets. Controversy still exists, but a few studies have reported 5-HT-induced platelet aggregation in patients with peripheral cerebrovascular disease. Peripheral arterial vasospasms subsequent to platelet activation appear to be mediated in part by 5-HT and can be effectively reduced by ketanserin. In animal studies, postthrombotic peripheral collateral circulation can be significantly restored by treatment with ketanserin (pKD 8.5–9.5) and MDL100907 (pKD 9.4), and 5-HT-induced reduction of blood supply through the collateral system can be effectively counteracted by pretreatment with ketanserin. In ex vivo human platelet studies, the 5-HT2A/5-HT2C/5-HT2B receptor antagonist ICI 170809 significantly inhibited 5-HT-induced aggregation (minimum effective dose, 0.1 mg kg 1 per os), thus lending support for the potential therapeutic role of 5-HT2A receptor antagonists in modulating the vasospastic effects of 5-HT released from platelets during vascular trauma. The clinical significance of selective 5-HT2A (see also

discussions of 5-HT2B and 5-HT2C) receptor antagonists in this vascular disorder and various psychiatric disorders still awaits further conclusive evidence of clinical efficacy with selective agents. In psychiatric disorder studies, two recent 5-HT2A receptor antagonists have shown significant clinical efficacy in schizophrenia. Pimavanserin (ACP-103) significantly improved efficacy scores (positive and negative syndrome scale; PANSS), showing a faster onset and improved side-effect profile, following co-therapy with risperidone. Fananserin, which binds with high affinity to dopamine D4, in addition to 5-HT2A receptors, also showed good clinically efficacy in schizophrenia, with an improved side-effect profile. This therapeutic opportunity is currently been actively pursued by several pharmaceutical companies, and several compounds acting at 5-HT2A and dopamine D2 receptors, are at various stages of clinical development (e.g., asenapine, blonanserin, eplivanserin, pruvanserin, paliperidone, uoperidone, and vabicaserin). Another selective 5-HT2A receptor antagonist, APD-125, is currently being assessed in the clinic for insomnia and sleep maintenance. The 5-HT2B(Gq11) Receptor

The 5-HT2B mRNA transcript has been now been identified in human and nonvertebrates and the receptor has been mapped to human chromosome HTR2B/2q36.3–2q37.1, with the human gene encoding 481 amino acids (Table 2). It was first identified in the stomach of the rat and in the small intestine, kidney, heart, and cerebellum in the mouse. The presence of the transcript in the rat stomach and the pharmacology of the cloned 5-HT2B receptor have led ultimately to defining the rat stomach fundus receptor. In Gaddum and Picarelli’s first classical attempt at subdivision of 5-HT receptors, the D subtype (5-HT2A) preceded the first descriptions of the rat stomach fundus 5-HT receptor by only a few pages in the same issue of the British Journal of Pharmacology, whereas the 5-HT2C receptor was not identified until some 28 years later. The 5-HT2B receptor represents the second member of the 5-HT2 family, but it is still unclear how the clone relates pharmacologically to that 5-HT receptor which mediates contraction of the rat stomach fundus. Evidence indicates that the 5-HT2B contractile receptor in the rat stomach fundus is coupled to calcium influx through voltage-dependent calcium channels, intracellular calcium release, and activation of protein kinase C (PKC). These actions may reflect a novel coupling mechanism unrelated to increases in PI hydrolysis, unlike the other 5-HT2 receptors. In the 1990s a series of potent, selective 5-HT2B receptor antagonists were identified, of which

464 Serotonin (5-Hydroxytryptamine; 5-HT): Receptors

SB204741 (pKi 7.8) has proved to be an important tool for characterizing this receptor, along with the less selective 5-HT2B receptor agonists BW723C86 and a-Me-5-HT. Recent additions to the 5-HT2B tool-kit have been the introduction of two highly selective 5-HT2B receptor antagonists, RS127445 (pKi 9.5) and EGIS-7625 (pKi 9.0). The clinical potential of compounds selective for the 5-HT2B receptor site has pointed to an alternative target for initiation of migraine and may account for the continued use of methysergide, pizotifen, and cyproheptadine for migraine prophylaxis. More selective compounds have been developed for this indication, but to date no selective 5-HT2B compound has shown clinical efficacy for migraine. Moreover, the only 5-HT2B compound in current clinical development for the treatment of pulmonary hypertension is PRX-08066, which is surprising, considering the critical importance of the 5-HT2B receptor in mammalian heart development. The 5-HT2C(Gq11) Receptor

The human 5-HT2C receptor gene HTR2C/Xq24 encodes 377 amino acid residues with seven hydrophobic domains of 20–30 residues (Table 2). The rat and mouse sequences suggest the presence of an eighth transmembrane region not seen in the human receptor. The rat and human sequences are otherwise very similar, with a 90% overall homology. They are also positively coupled to phosphoinositide hydrolysis, providing a common secondary messenger system for 5-HT2C and 5-HT2A receptors. It is well established that 5-HT2C receptors display a high degree of constitutive activity, and consequently inverse agonists have a large effect when they bind to the 5-HT2C receptor. Furthermore, the mRNA for the 5-HT2C receptor undergoes mRNA editing, which results in the expression of multiple 5-HT2C receptor isoforms. Correspondingly, INI (unedited) and VSV (a fully edited version) isoforms are abundant in rat brain. The VSV isoform lacks the highaffinity recognition site for 5-HT, which may be caused by low-efficiency coupling to G-proteins. These edited isoforms appear to have reduced constitutive activity and reduced capacity to couple to Gq11 signaling proteins. This suggests that the INI isoform of the 5-HT2C receptor is pharmacologically similar to the VSV form of the 5-HT2C receptor, but that it couples more efficiently to G-proteins. Importantly, the capacity of some cells in the brain to edit the 5-HT2C mRNA is reduced. It has been suggested that this may well be the underlying pathophysiology in schizophrenic patients and that this reduced capacity is evident in brains of patients who had committed suicide. Changes in the RNA-editing capacity of the

brain will result in altered patterns of expression of the edited 5-HT2C isoforms and perhaps in altered responses to drugs. Consequently, it is important to understand the signaling characteristics and mechanism of the edited 5-HT2C receptor isoforms in various psychiatric and neurological conditions, and selective agents are allowing this work to progress. The pharmacological characteristics of the 5-HT2A and 5-HT2C (formally named 5-HT1C) receptors are remarkably similar, with many of the ‘classical’ 5-HT receptor antagonists showing high affinity for both sites. Furthermore, some compounds may have different affinities for human as opposed to rat receptors. The first 5-HT2C selective antagonist SB200646A (50-fold) was reported in the early 1990s; however, few truly selective 5-HT2C receptor agonists have been identified. MK 212 (30-fold) is the most selective CNS active agonist, but is also a 5-HT2A agonist, while a-methyl-5-HT has equal affinity for all 5-HT2 receptor subtypes. Of the various 5-HT2 receptor agonists, (þ)-DOB and m-chlorophenylpiperazine (mCPP) possess approximately tenfold selectivity over other binding sites in the rat but not in humans. However, mCPP is of particular interest because of its use as a clinical probe in the 1990s to study human 5-HT2C receptor function. mCPP is commonly observed to induce anxiety and panic attacks in humans. This mCPP-induced anxiogenesis is inhibited by the 5-HT2C and 5-HT2A receptor antagonist ritanserin, which is consistent with 5-HT2C receptor mediation. If 5-HT2C receptor activation is anxiogenic, the blockade of this receptor might induce anxiolysis, provided the receptors are tonically innervated. This argument is supported by the anxiolytic properties of SB243213, a selective 5-HT2C antagonist, in rat models of anxiety with quite different motivational and aversive bases (the rat social interaction and Geller–Seifter tests). To date this hypothesis awaits clinical testing. In support of this hypothesis, the selective 5-HT reuptake inhibitors (SSRIs) paroxetine, fluoxetine, clomipramine, sertraline, and citalopram, on chronic administration, have all been observed to desensitize behavioral responses to mCPP. The monoamine oxidase inhibitors phenelzine and nialamide, which also raise extraneuronal 5-HT levels, have a similar effect after chronic treatment. These results suggest that SSRIs and MAO inhibition antidepressant treatments may desensitize the 5-HT2C receptor, although there is no direct evidence from receptor binding, mRNA, or secondary messenger systems as yet. Furthermore, unlike other SSRIs, fluoxetine and its metabolite norfluoxetine have weak affinity for the 5-HT2C receptor and are therefore likely to accumulate and directly block 5-HT2C sites.

Serotonin (5-Hydroxytryptamine; 5-HT): Receptors 465

Paradoxically, there is an emerging rationale and literature for the use of selective 5-HT2C receptor agonists in psychiatric and eating disorders that has been driven by recently reported compounds that have shown greater selectivity for the 5-HT2C receptor than has mCPP. Early studies with mCPP pointed the way to far more selective agonists in eating and psychiatric disorders. Several reports suggest that the 5-HT2C receptor agonists VER-2692 and WAY16909 are potent and selective agonists, and these are currently being evaluated as antiobesity agents. Early clinical studies with mCPP in bulimic patients induced migraine-like headaches 8–12 h later. This response was correlated with plasma levels of mCPP and was more pronounced in patients with a personal or family history of migraine. Conversely, the nonselective 5-HT2C/5-HT2A receptor antagonists pizotifen, cyproheptadine, and methysergide are all clinically effective as migraine prophylactics. ICI 169369 and sergolexole, which have high affinity for 5-HT2C and 5-HT2A receptors, but little affinity for other sites, have subsequently been found to have modest antimigraine efficacy in an early trial. As these drugs only share a high affinity for the 5-HT2C and 5-HT2A receptors, and as the selective 5-HT2A receptor antagonist ketanserin is ineffective as a migraine prophylactic, it was proposed that blockade of 5-HT2C receptors mediated this effect (although this has recently been challenged on the basis of their affinity for the 5-HT2B receptor). Furthermore, downregulation of 5-HT2C receptors may mediate the antimigraine properties of the antidepressants amitriptyline and fluoxetine; alternatively, the latter compounds may act by direct antagonism of vascular 5-HT2C receptor sites. Other possible therapeutic targets for drugs acting at 5-HT2C receptors include eating disorders, sexual dysfunction, and head trauma, based on the actions of nonselective 5-HT2C/2A agonists and antagonists. It is highly likely that the 5-HT2 family will continue to grow. There is already considerable evidence for a number of other ‘orphan’ receptors which may be adopted into the family. Several orphans are thought to exist, one of which is the endothelial 5-HT receptor, which is present in rabbit and rat jugular vein and pig pulmonary artery and vena cava. The rat jugular vein is now known to represent a 5-HT2B receptor. More recently, a site identified in choroid plexus has been shown to bind the 5-HT2-selective ligand RP62203 with high affinity, but does not possess identity with any of the three established 5-HT2 receptors. However, it would appear that the endothelial receptors do not represent a pharmacologically homogeneous class. While possessing many of the

characteristics of 5-HT2C and 5-HT2B receptors, subtle yet robust differences render any explicit affiliation impossible. These problems are confounded by the need, in many instances, to make comparisons across species. The potential for substantial species differences in pharmacology is highlighted on pharmacological comparison of mouse and rat homologs of the cloned 5-HT2B receptor. Thus the eventual classification of the orphan endothelial 5-HT receptor awaits robust interspecies comparison of 5-HT2A, 5-HT2B, and 5-HT2C receptors. In this regard, in rat jugular vein, the endothelial receptor, which was previously classified as 5-HT2C-like, has now been shown to be more like 5-HT2B receptors, using ligands which discriminate the rodent (rat) 5-HT2 receptor subtypes. Another therapeutic area that has gained prominence recently arose from observations in humans that mCPP reduces total sleep time, sleep efficiency, slow-wave sleep (SWS), and rapid eye movement sleep (REMS). The actions of 5-HT2A/2C receptor antagonists on sleep architecture have since been widely studied. Thus ritanserin has been reported to increase SWS, reduce sleep onset latency, and improve subjective sleep quality in volunteers, while REMS is reduced in some, but not all, studies. Ritanserin has also been observed to be of therapeutic use in insomniacs and in patients suffering from dysthymia. These effects were maintained with chronic treatment. Other nonselective 5-HT2A/5-HT2C receptor antagonists such as mianserin, cyproheptadine, and pizotifen have similar effects. In rats, ritanserin generally increases SWS and reduces REMS, although some studies report that the deepest phase of SWS (SWS2) is increased but total SWS is unaffected. Studies with three other 5-HT2A/5-HT2C receptor antagonists, ICI 169369 and ICI 170809 and SR46349B, report reduced REMS, but little effect on undifferentiated SWS. These results, therefore, suggest that a 5-HT2C receptor antagonist may improve sleep quality (see also agomelatine). Of great interest to pharmaceutical companies is the role of the 5-HT2C receptor in anxiety and depression. Two 5-HT2C receptor inverse agonists, SB243213 and WAY163909, have been reported to be in development for the treatment of depression. The fact that both compounds show a similar pharmacology in several animal models of depression reflects the ability of the 5-HT2C receptor agonists and antagonists/inverse agonists to desensitize the 5-HT2C receptor, as observed following chronic SSRI treatment. Another drug that has recently gained European regulatory approval is agomelatine, which is a melatonergic, MT1 and MT2 receptor agonist/5-HT2C antagonist with

466 Serotonin (5-Hydroxytryptamine; 5-HT): Receptors

antidepressant properties and which may be acting through a similar mechanism. It is reported to improve disrupted sleep patterns in depressed patients (which also is likely, due to its 5-HT2C receptor antagonist properties), without affecting daytime vigilance. The reported lack of effect on sexual function, tolerability problems, or discontinuation symptoms offers several advantages over current therapy for the treatment of depression. It is clearly evident that therapeutic opportunities are still to be realized in the 5-HT2 receptor family and will undoubtedly follow with clinical evaluation of selective, nonselective, and combination agents. The Ligand-Gated 5-HT3 Receptor

The original naming of the ‘M’ receptor by Gaddum and Picarelli was based on the ability of morphine to antagonize the serotonin receptor which mediates depolarization of cholinergic nerves in the guinea pig ileum. Morphine was the first pharmacological tool to help characterize this receptor, now referred to as the 5-HT3 receptor. Since the time of these early observations, 5-HT3 receptors have been reported to be present on postganglionic autonomic neurons in the peripheral sympathetic and parasympathetic nervous systems, on enteric neurons, and on sensory neurons in various tissues. The 5-HT3 receptor is a member of the Cys-Cys loop ligand-gated ion channel superfamily and as such is composed of multiple subunits, containing 478 amino acids. Five human genes have been identified on chromosome HTR3/11, which encodes the 5-HT3 subunit (5-HT3A–E) (Table 3). To date, conflicting reports as to the existence of central 5-HT3B subunits have been reported. However, immunohistochemical studies using sections of human temporal lobe demonstrated both 5-HT3A and 5-HT3B immunoreactivity associated with pyramidal neurons within all CA fields of the hippocampus, as well as with large neurons within the hilus. The expression of both h5-HT3A and h5-HT3B subunit mRNAs in human hippocampus would therefore support the presence of both homomeric h5-HT3A and heteromeric h-5-HT3A/3B receptors in human hippocampus. The pathophysiological relevance of central h5-HT3A and h5-HT3B receptors remains to be fully evaluated with selective receptor antagonists in preclinical and clinical studies. The pioneering animal studies in the 1980s with the nonselective antagonist ondansetron clearly point to a role for 5-HT3 receptors in anxiety, depression, and schizophrenia. However, this was not substantiated in human studies with several 5-HT3 receptor antagonists, including, ondansetron, granisetron, BRL46470, and dolasetron.

5-HT is found in both the brain and the gut, but it is now widely understood that 95% of the serotonin in the body resides in the gut. The two leading 5-HT3 receptor antagonists, ondansetron and granisetron, have proved to be highly successful antiemetic agents, and they are widely used in the prophylactic treatment of chemotherapy-induced nausea and vomiting. This discovery arose from carefully reasoned experimental research. It was shown in animal studies that chemotherapeutic agents such as cisplatin released 5-HT in the gut in vast quantities, activating 5-HT3 receptors in this region. Activation of 5-HT3 receptors resulted in the firing of a vagally mediated response resulting in emesis. The discovery of the antiemetic properties of these 5-HT3 receptor antagonists, along with several others that followed, opened the way for this class of compounds to become the gold standard in the clinical treatment of chemotherapy-induced nausea and emesis. A second non-CNS therapeutic target that has gained prominence in recent years has been irritable bowel syndrome (IBS) and its association with the 5-HT3 receptor and 5-HT4 receptor. 5-HT3 receptor antagonists were predicted to be effective in diarrheapredominant IBS based on animal studies. Clinical studies with ondansetron and granisetron showed a reversal in rectal hypersensitivity in IBS, and the longer acting agent alosetron showed significant clinical improvement in IBS symptoms. However, alosetron has had a checkered clinical existence following early withdrawal after reports of rare serious gastrointestinal adverse events among patients taking the drug. Similar drugs (i.e., cilansetron, ramosetron, and DDP-733) are also in clinical studies for IBS, with the hope that these side effects are not class related. The latter compound is a partial agonist that has recently been reported to be active in a phase II IBS study. The complex symptomatology in IBS may be attributed to activation of other 5-HT receptor subtypes in the gastrointestinal tract. It was the identification of 5-HT4 receptors that led many pharmaceutical companies to believe that agents acting as 5-HT4 receptors would be useful in the treatment of this condition (see discussion of the 5-HT4 receptor and Table 4). The 5-HT4(Gs) Receptor

The 5-HT4 receptor was originally identified in primary cell cultures of mouse embryo colliculli, where later it was found to have a broad tissue distribution and to be positively coupled to adenylate cyclase. In early functional assays studies the 5-HT4 receptor was identified in the rat esophageal muscularis mucosae and guinea pig colon, where activation resulted in

Serotonin (5-Hydroxytryptamine; 5-HT): Receptors 467

relaxation and contraction, respectively. The cloned human and rodent receptor was shown to consist of 388 and 406 amino acids, respectively. Various isoforms of the 5-HT4 receptor have now been identified (5-HT4b,c and 5-HT4d), encoded by a gene located on chromosome HTR4/5q31–q33 (Table 4). In the brain, 5-HT4 receptors appear to be located on neurons, whereas in the peripheral nervous system they are reported to facilitate the release of acetylcholine in smooth muscle of the intestinal tract. It was these observations on the gastrointestinal tract that led to the notion that 5-HT4 receptors may have a role in the pathophysiology of irritable bowel syndrome and gastroesophageal reflux disease (GERD). The complex symptomatology and pharmacology of IBS led to the development of several highly selective 5-HT4 receptor agonists and antagonists that are now established in clinical practice for this indication. Tegaserod (HTF-919) and purcalopride (R93877) are 5-HT4 receptor partial agonists, and piboserod (SB207266) is an antagonist. All of these medications are reported to relieve the symptoms of constipation, the notion being that the 5-HT3/4 agonists are intended to have prokinetic activity, like the mixed 5-HT4 receptor agonist/5-HT3 receptor agonist, cisapride. Such agents would therefore be beneficial for constipation-predominant IBS. Conversely, the antagonist piboserod would be indicated for diarrhea-predominant IBS. However, the therapeutic potential of piboserod was noted to be limited to a decrease in rectal sensitivity, and the compound was not advanced in further clinical studies. The disappointment with this class of drugs has been their lack of efficacy in IBS-related pain. However, other 5-HT3 agonists (pumosetrag) and 5HT4 agonists (TD-2749 and TD-5108) are also being evaluated in IBS, which may address this important, unmet medical need. The potential CNS efficacy of 5-HT4 receptor antagonists is less well developed, though the 5-HT4 partial agonist SL65.0155 (5-(8-amino-7-chloro2,3dihydro-1,4-benzodioxin-5yl)-3-[1-(2-phenylethyl)4-piperidinyl]1,3,4-oxadiazol-2(3H)-one-monohydro chloride) has been reported to show cognitive enhancing effects in animal models. The high density of 5-HT4 receptors in the nucleus accumbens has suggested that these receptors may be involved in reward mechanisms and may influence self-administration. Recent studies with GR113808 and its effects on reducing alcohol intake in rats would concur with this suggestion. The clinical benefit of selective agents for the 5-HT4 receptors in medical practice and the promise of more developments clearly represent one of many success stories of 5-HT drug development.

The 5-ht5 Receptor (G-Protein Coupling, None Identified)

The 5-HT5 receptor is highly localized in the rodent CNS; its distribution in the hippocampus (CA1, CA3, dentate gyrus), cortex, cerebellum (granular layer), olfactory bulb, habenula, and spinal cord would suggest a role in the pathophysiology of psychiatric and neurological disorders. A human 5-HT5A, but not a 5-ht5B, receptor has been identified; the human 5-HT5A receptor gene, like the rodent gene, contains two codon exons separated by a single large intron. In the case of the human and rodent 5-HT5A receptor, the gene contains the same number of amino acids (357); similarly, the rodent 5ht5b gene contains 387. To elucidate the pathophysiology of the 5-HT5 gene, chromosomal localization studies have identified the receptor on HTR5A/7q36, whereas the 5ht5b gene is localized on htr5b/2q11–q13. The signal transduction mechanism for the 5-HT5 receptor is still a matter for conjecture, as some suggest it may utilize a novel (perhaps ion channel) second-messenger system (Table 4). The physiological function of the 5-HT5A receptor is poorly understood. 5-HT5A null mutant mouse studies suggest a role in exploratory behavior, but the lack of selective pharmacological tools has hampered progress. Localization studies have revealed that 5-HT5A receptors have widespread distribution in the CNS. It has been speculated that, on the basis of their localization, the receptors may be involved in motor control, learning and memory consolidation, and anxiety and depression. Recently a selective 5-HT5A receptor antagonist, SB699551A, has been identified. This compound has been reported to attenuate 5-carboxyamidotryptamine-induced inhibition of raphe neuronal cell firing in vitro and to increase 5-HT levels in prefrontal cortex in vivo, suggesting a role for the 5-HT5A receptor in the modulation of raphe 5-HT neuronal activity. This is another pharmacological opportunity whereby ‘tool’ compounds will expose the receptor subtype operational and functional characteristics in native tissue, as it has been suggested that many of the operational characteristics of the 5-HT1D receptor may be attributed to the 5-HT5A receptor. The 5-HT6(Gs) Receptor

5-HT6 receptors have been identified in areas of the rat and human brain associated with learning and memory: hippocampus, CA1, CA3, dentate gyrus, olfactory tubercles, cerebral cortex, nucleus accumbens, and striatum. The 5-HT6 gene is localized on chromosome HTR6/1p35–p36, and studies to

468 Serotonin (5-Hydroxytryptamine; 5-HT): Receptors

determine its linkage to several CNS disorders (memory dysfunction and schizophrenia) are actively being pursued (Table 4). However, the potential of 5-HT6 receptor antagonists in the treatment of cognitive disorders has been an actively researched topic in recent years, and the availability of numerous tool compounds for this receptor, some of which are currently in clinical trial, will advance our understanding of the role of this receptor in cognitive process. The notion that several atypical antipsychotic agents such as clozapine, quetiapine, and olanzapine possess high affinity for the 5-HT6 receptor adds translational creditability to this approach. The leading pharmacophores are indole-related compounds, of which several have emerged, including the recently described aryl sulfonamide indole chemotype, SB271046 (pKi 8.6), with excellent CNS bioavailability and in vivo efficacy, compared to several earlier compounds. More recently, 5-HT6 receptors have been demonstrated to regulate central cholinergic neurotransmission. In this regard, administration of the 5-HT6 receptor-selective antagonist Ro 04–6790 reversed scopolamine-induced rotation in 6-hydroxydopaminelesioned rats. Additionally, 5-HT6 receptor antisense oligonucleotides or 5-HT6 receptor-selective inhibitors enhanced retention by rats of the learned platform position in the Morris water maze. These data suggest that 5-HT6 receptor antagonists might boost cholinergic neurotransmission and reduce the cognitive impairments experienced by patients with dementia or schizophrenia. Intriguingly, it has recently been determined that the 267C allele of the 5-HT6 receptor is a significant risk factor for Alzheimer’s disease. Taken together, these findings indicate that 5-HT6 antagonists might prove useful in treating a number of common illnesses, including dementia and schizophrenia. A major concern in drug discovery is that the cloned mouse receptor is significantly different in rat and human 5-HT6 receptors. In addition to species differences in the binding of drugs to 5-HT6 receptors, differences in the regional expression of 5-HT6 receptors are also apparent. Furthermore, quantitative polymerase chain reaction studies have demonstrated that the mouse 5-HT6 receptor mRNA is at least tenfold less abundant than are the rat and human 5-HT6 receptor mRNAs, in every brain region examined. Surprisingly, whereas 5-HT6 receptor mRNA and radioligand binding activity is enriched in the basal ganglia of rat and human brain, there is no such enrichment in the mouse brain. Finally, using a combination of site-directed mutagenesis and molecular modeling studies, it has been

shown that the peculiar mouse 5-HT6 pharmacology is attributable to two amino acids – Tyr188 (in helix 5, which is Phe188 in rats and humans) and Ser290 (in helix 6, which is Asn290 in rats and humans) – and these account for the bulk of the differences in pharmacology. These findings are a cautionary note but may have important implications for drug discovery. The 5-HT7(Gs) Receptor

Since the identification of the 5-HT7 receptor in the early 1990s, evidence has accumulated supporting a role for 5-HT7 receptors in various physiological functions, including sleep disorders, circadian rhythms, cognition, mood, and sleep (see Table 5). A number of splice variants of the 5-HT7 receptor have been identified and exhibit overlapping mRNA distributions, and all couple to Gs, though no differences in operational characteristics have been shown, which may have led to differentiation of behavioral phenotypes. Work is still ongoing to elucidate the pathophysiology of these splice variants, and chromosomal localization studies have identified the receptor on HTR5A/7q36, encoding 445 and 488 amino acids for the human and rodent receptor, respectively (Table 5). The finding that several drugs used in clinical psychiatric practice had high affinity for 5-HT7 receptor binding sites in the brain initiated an ongoing interest in evaluating the involvement of neuropsychiatric disorders. The distribution of 5-HT7 receptors in areas of the brain associated with neuropsychiatric disorders – hippocampus (CA1, CA2), hypothalamus, thalamus, raphe nuclei, and superior colliculus – supported this notion. In concordance with its distribution in the hippocampus, the selective 5-HT7 receptor antagonist SB269700-Z was shown to attenuate phencyclidine (PCP)-induced cognition dysfunction associated with schizophrenia. The strongest case has been made for the importance of the 5-HT7 receptor in depression. Inactivation or blockade of the receptor leads to an ‘antidepressant’ state in behavioral models of depression. Furthermore, 5-HT7 receptor downregulation following chronic administration of antidepressants in animal studies points to a possible molecular mechanism of action. This hypothesis is supported by recent animal experiments using 5-HT7 receptor knockout mice that clearly show an antidepressantlike profile in the rat forced swim test and mouse tail suspension test. Selective antagonist tool compounds SB258717, SB269970, and SB656104-A have all been shown to exhibit antidepressant-like properties in animal studies and to modulate REM sleep in rats, results that concur with this line of investigation. Several lines

Serotonin (5-Hydroxytryptamine; 5-HT): Receptors 469

of evidence have shown that 5-HT-induced hypothermia is mediated by the 5-HT7 receptor, as defined with selective antagonists and in 5-HT7 knockout mice studies. What is interesting, however, is recent work with the endogenous fatty acid oleamide that suggests it may act through an independent mechanism as well as at an allosteric 5-HT7 receptor site to regulate body temperature and perhaps circadian rhythms. Clinical progress in this area has been limited by suboptimal clinical candidates, and clinical efficacy data for selective agents are still awaited.

The Future of 5-HT Research It is now over 50 years since Gaddum first suggested ‘‘that a drug with a specific antagonistic action to 5-HT – might be used in therapeutics.’’ Clearly, drugs acting at the 5-HT receptor subtypes have revolutionized therapeutics during the intervening years. However, it is only now that selective agents are appearing to differentiate many operational and physiological characteristics of 5-HT actions on the various receptor subtypes currently identified. The future therapeutic potential of drugs acting at these 5-HT subtypes and isoforms or in combination with other neurotransmitter receptors will open up next generation of novel serotonergic therapeutic agents. See also: Serotonin (5-Hydroxytryptamine; 5-HT): CNS Pathways and Neurophysiology; Serotonin (5-Hydroxytryptamine; 5-HT): Neurotransmission and Neuromodulation.

Further Reading Alexander SPH, Mathie A, and Peters JA (2008) Guide to receptors and channels. British Journal of Pharmacology 153 http://www. nature.com/bjp/journal/vgrac/ncurrent/index.html (accessed Jul. 2008). Barnes NM and Sharpe T (1999) A review of central 5-HT receptors and their function. Neuropharmacology 38: 1083–1152. Bonasera SJ and Tecott LH (2000) Mouse models of serotonin receptor function: Toward a genetic dissection of serotonin systems. Pharmacology & Therapeutics 88: 133–142. Branchek TA and Blackburn TP (2000) 5-HT6 receptors as emerging targets for drug discovery. Annual Review of Pharmacology and Toxicology 40: 319–334.

Brockaert J, Claeysen S, Compan V, et al. (2004) 5-HT4 receptors. Current Drug Targets – CNS & Neurological Disorders 3: 39–51. Glennon RA (2003) Higher-end serotonin receptors: 5-HT5, 5-HT6, and 5-HT7. Journal of Medical Chemistry 46: 2795–2812. Hartig PR, Hoyer D, Humphrey PP, et al. (1996) Alignment of receptor nomenclature with the human genome: Classification of 5-HT1B and 5-HT1D receptor subtypes. Trends in Pharmacological Sciences 17: 103–105. Hoyer D, Clarke DE, Fozard JR, et al. (1994) International Union of Pharmacology classification of receptors for 5-hydroxytryptamine (serotonin). Pharmacological Reviews 46: 157–203. Hoyer D and Martin G (1997) 5-HT receptor classification and nomenclature: Towards a harmonization with the human genome. Neuropharmacology 36: 419–428. Lanfumey L and Hamon M (2004) 5-HT1 receptors. Current Drug Targets – CNS & Neurological Disorders 3: 1–10. Jones JJ and Blackburn TP (2002) The medical benefit of 5-HT. Pharmacology, Biochemistry and Behavior 71: 555–568. Leysen JE (2004) 5-HT2 receptors. Current Drug Targets – CNS & Neurological Disorders 3: 11–26. Lui M and Gershon MD (2005) Slow excitatory (‘‘5-HT1P’’-like) responses of mouse myenteric neurons to 5-HT: Mediation by heterodimers of 5-HT1B/1D and Drd2 receptors. Gastroenterology 128(4, supplement 2): A87. McLean PG, Borman RA, and Kee K (2007) 5-HT in the enteric nervous system: Gut function and neuropharmacology. Trends in Neurosciences 1: 9–13. Nelson DL (2004) 5-HT5 receptors. Current Drug Targets – CNS & Neurological Disorders 3: 53–58. Pauwels PL (2000) Diverse signalling by 5-hydroxytryptamine (5-HT) receptors. Biochemical Pharmacology 60: 1743–1750. Peters JA, Hales TG, and Lambert JJ (2005) Molecular determinants of single channel conductance and ion selectivity in the Cys-loop transmitter-gated ion channels: Insights from the 5-HT3 receptor. Trends Pharmacological Sciences 26: 587–594. Sanders-Bush E, Fentress H, and Hazelwood L (2003) Serotonin 5-HT2 receptors: Molecular and genomic diversity. Molecular Interventions 3: 319–330. Steward LJ, Ge J, Bentley KR, et al. (2007) Guide to receptors and channels (GRAC), 2nd edition (2007 revision). British Journal of Pharmacology 150(supplement 1): S1–S168. Thomas DR and Hagen JJ (2004) 5-HT7 receptors. Current Drug Targets – CNS & Neurological Disorders 3: 81–90. Woolley ML, Marsden CA, and Fone KC (2004) 5-ht6 receptors. Current Drug Targets – CNS & Neurological Disorders 3: 59–79.

Relevant Website http://www.iuphar-db.org – IUPHAR Receptor Database.

Serotonin (5-Hydroxytryptamine; 5-HT): CNS Pathways and Neurophysiology G Aghajanian and R-J Liu, Yale School of Medicine, New Haven, CT, USA ã 2009 Elsevier Ltd. All rights reserved.

Introduction Serotonin (5-hydroxytryptamine; 5-HT), a biogenic indoleamine derived metabolically from the dietary amino acid tryptophan, was first discovered in peripheral tissues and subsequently found in brain. Shortly afterward, in the early 1950s, a structural relationship was noted between serotonin and d-lysergic acid diethylamide (LSD), the most potent of all known hallucinogenic drugs. Subsequently, it was recognized that a wide range of psychedelic hallucinogens, both indoleamines such as N,N-dimethyltryptame (DMT) and phenethyamines such as mescaline, had structural similarities to LSD (Figure 1). This relationship to powerful psychotropic drugs sparked interest in the idea that serotonin might function as a central ‘neurohumoral’ substance with a profound influence on behavior. However, it was not obvious how to investigate this idea directly until the mid-1960s, when, through the use of a new histochemical method (the Falck– Hillarp method), it was discovered that serotonin was highly concentrated in specific sets of neurons located almost exclusively in the raphe nuclei of the brain stem. Histochemical maps showed that serotonergic projections emanating from the raphe nuclei innervate, to varying degrees, virtually all other parts of the central nervous system, ranging from the cerebral cortex to the spinal cord. Based on these maps, it became possible to study the properties of identified serotonergic neurons and their postsynaptic targets. Although it was its link to hallucinogenic drugs that first brought it to prominence, serotonin has now been implicated in almost every conceivable physiological or behavioral function – affect, aggression, appetite, cognition, emesis, endocrine function, gastrointestinal function, motor function, neurotrophism, perception, sensory function, sex, sleep, respiration, and vascular function. Moreover, many drugs that are currently used for the treatment of psychiatric disorders (e.g., depression, mania, schizophrenia, autism, obsessive–compulsive disorder, anxiety disorders) are thought to act, at least partially, through serotonergic mechanisms. How is it possible for serotonin to be involved in so many different processes? One answer lies in the anatomy of the serotonergic system, where small

470

clusters of serotoninergic cell bodies located in the brain stem raphe nuclei are positioned through their vast projections to influence all regions of the neuraxis. Another answer lies in the molecular diversity and differential cellular distribution of the many serotonin receptor subtypes that are expressed differentially in brain and other tissues. For example, the effects of hallucinogenics are mediated via the 5-HT2A subtype of serotonin receptor, which is expressed extensively in the cerebral cortex. In contrast, antidepressant effects are thought to involve members of the 5-HT1 family of receptors located in the brain stem raphe as well as in certain postsynaptic regions. Given this diversity, it is useful to approach serotonin neurophysiology within the context of the neuroanatomical and cellular distribution of serotonin receptor subtypes and their associated transduction pathways.

Serotonergic Neuronal Pathways Serotonergic neurons are clustered in a series of midline nuclei, labeled B1–B9, extending from the midbrain to the medulla oblongata. The dorsal and median raphe nuclei (B7 and B8) comprise the largest of these groupings of brain serotonergic neurons, and together they provide the major serotonergic input to the forebrain; the lower brain stem raphe nuclei supply the main serotonergic projections to the spinal cord. A single serotonergic neuron projects to large numbers of postsynaptic cells and does so in an overlapping fashion with other serotonergic neurons. While the density of this input varies widely, both regionally and between different cell types within a given region, virtually no corner of the CNS is lacking a serotonergic input despite the fact that serotonergic neurons represent far less than 1% of all the neurons in the brain. Given these anatomical considerations, the serotonergic system is well positioned to provide a broad modulatory influence upon diverse functions rather than the precise point-to-point transmission characteristic of traditional sensory, motor, or associational pathways. In view of its widespread projections arising from a relative handful of neurons, is it possible to assign an overarching function to the serotonergic system? An answer to this question may come from the fact that serotonergic neurons display a pattern of firing that is dependent upon behavioral state: they fire most rapidly during alert waking, decelerate considerably during slow-wave sleep, and become almost entirely silent during dream – or rapid-eye-movement

Serotonin (5-Hydroxytryptamine; 5-HT): CNS Pathways and Neurophysiology 471

H5C2 NH2

H5C2

O NC

N CH3

HO N H

N H 5-HT (5-hydroxytryptamine) H3C N

LSD

CH3

N H N,N-dimethyltryptamine

NH2

H3CO

OCH3 OCH3 Mescaline

Figure 1 The structural formulas of serotonin and three related hallucinogenic drugs: LSD, DMT, and mescaline. Note that serotonin, LSD, and DMT have in common an indolethylamine moiety. While it is not an indolethylamine, mescaline shares with LSD a phenethylamine moiety. The structures are drawn so as to emphasize these shared structural features. Provided by GK Aghajanian.

(REM) – sleep. Thus, in a broad sense, one can view serotonergic neurons as constituting a ‘wake-on’ system. In this context, many of the electrophysiological effects of serotonin in postsynaptic regions can be seen as facilitating motor, sensory, and cognitive functions characteristic of the waking state. All of this would be reversed when serotonergic neurons become quiescent during sleep states, when there is an emergence of sleep- and dream-related systems. Such a broad, state-related function for serotonin has been likened to the peripheral sympathetic system where small clusters of noradrenergic neurons project to a myriad of organs including the heart, gut, and skin to influence their function in a concerted manner. Interestingly, a central counterpart to the sympathetic system exists in the brain stem where small clusters of noradrenergic neurons (e.g., in the locus coeruleus) parallel the broad projections of the serotonergic system and similarly have a broad influence on a wide range of wake-related functions.

Regulation of Serotonergic Neuronal Activity Ultimately, the electrophysiological effects of serotonin depend upon the activity of the serotonergic cells of origin in the raphe nuclei. In various mammalian species, serotonergic neurons in waking states have been found to have a slow, tonic pattern of firing

Figure 2 Schematic diagram showing norepinephrine (NE), hypocretin/orexin (hcrt/orexin), and g-aminobutyric acid (GABA) inputs to a 5-HT cell in the brain stem raphe. The NE input originates from nearby sites in the brain stem while the hcrt/orexin input is hypothalamic in origin; both of these are excitatory (þ). Also depicted is a 5-HT collateral input to the 5-HT cell; this input is inhibitory () via somatodendritic 5-HT1A autoreceptors. Finally, an inhibitory input () from a local GABA interneuron is shown; the GABA cell itself receives 5-HT collaterals, which can be excitatory (via 5-HT2A receptors) or inhibitory (via 5-HT1A) receptors. Provided by GK Aghajanian.

(1–4 spikes s1). During slow-wave sleep, there is a deceleration of firing rate, and during REM (or dream) sleep, nearly a total cessation of activity. Thus, serotonergic neurons can be seen as being ‘wake-on’ cells. The maintenance of rhythmic firing during the waking state has suggested that serotonergic neurons possess regulated tonic pacemaker mechanisms. Intracellular and whole cell recordings from dorsal raphe neurons show that spikes arise from gradual depolarizing ramps (pacemaker potentials) rather than synaptic potentials. The regular pacemaker rhythm of serotonergic neurons is shaped by a complex interplay of intrinsic ionic currents; these include a voltage-dependent transient outward potassium current, a low-threshold inward calcium current, and a large calcium-activated outward potassium current. However, the drive behind the pacemaker activity mainly comes from extrinsic excitatory inputs, including brain stem noradrenergic inputs and hypothalamic hypocretin (orexin) inputs (Figure 2). The hypocretin system is essential for maintaining the normal waking state and loss of this system leads to the condition of narcolepsy, characterized by excessive somnolence. Both norepinephrine (via a1-adrenoceptors) and the peptide hypocretin/orexin

472 Serotonin (5-Hydroxytryptamine; 5-HT): CNS Pathways and Neurophysiology

Figure 3 Activation of pacemaker activity in a serotonergic neuron by norepinephrine (NE) or hypocretin in brain slice. Whole cell recording from a serotonergic neuron located in the dorsal raphe nucleus of a midbrain slice. As is typical of serotonergic neurons in brain slice where excitatory afferents are disconnected, this cell shows no basal spiking activity. When NE (top traces) or hypocretin/orexin (hcrt; bottom traces) is applied, there is a slow depolarization leading to tonic pacemaker activity. The expanded traces to the right show the gradual depolarizing ramps leading up to the onset of spiking activity. NE and hcrt/orexin can be seen as restoring the tonic pacemaker activity that is typical of serotonergic neurons in vivo. Provided by R-J Liu and GK Aghajanian.

(via hypocretin receptors 1 and 2), by closing potassium channels and opening nonselective cation channels, accelerate pacemaker activity of serotonergic neurons by inducing a slow depolarization (Figure 3). In addition to neurotransmitter inputs, elevated levels of CO2 also have a strong activating effect on serotonergic as well as noradrenergic neurons. This would provide an alerting signal in response to states of respiratory insufficiency. Serotonergic pacemaker activity is negatively modulated by the autoinhibitory action of serotonin, acting via somatodendritic 5-HT1A autoreceptors. The maintenance of tonic 5-HT1A autoinhibition depends upon the rate of serotonins synthesis, which in turn depends on availability of the initial serotonin precursor L-tryptophan. The activity of tryptophan hydroxylase, the limiting step in serotonin synthesis, is highly regulated by the tryptophan hydroxylase-activating kinases calcium/calmodulin protein kinase II (CaMKII) and protein kinase A (PKA). Increased calcium influx at higher firing rates, by activating tryptophan hydroxylase via CaMKII and PKA, can work together with tryptophan to enhance negative feedback control of the output of the serotonergic system.

Electrophysiology of Serotonin Receptor Subtypes The electrophysiological actions of serotonin are a function of the receptor subtype(s) expressed by a given cell. The Gi/Go-coupled 5-HT1 receptors generally mediate inhibitory effects on neuronal firing through an opening of potassium channels or a closing

of calcium channels. Inhibitions mediated by 5-HT1A and other 5-HT1 receptor subtypes have been observed in postsynaptic neurons located in many regions of the central nervous system. Presynaptic 5-HT1A receptors, located at the somatodendritic region of serotonergic neurons, have been termed ‘autoreceptors,’ mediating negative feedback inhibition via the cell’s own transmitter. The 5-HT2A receptors mediate slow excitatory effects through a decrease in potassium conductance or an increase in nonselective cation conductance. These are merely a few examples of how serotonin, with its widespread projections and large array of receptor subtypes/transduction pathways, is able to modulate neuronal excitability in a complex fashion throughout the neuraxis. While the electrophysiological actions of serotonin may seem quite varied, there is considerable uniformity within each of the major receptor families. Because of their common G-protein coupling, all members of the 5-HT1 receptor family tend to have inhibitory actions either pre- or postsynaptically. Similarly, all members of the 5-HT2 family tend to have excitatory actions. Therefore, the discussion of serotonin electrophysiology in selected regions of brain will be organized according to receptor families and their transduction pathways (Figure 4). 5-HT1 Receptors

Members of the 5-HT1 receptor family are coupled via Gi/Go G proteins and produce their effects through opening of inwardly rectifying Kþ channels or closing of Ca2þ channels. Particularly high levels of 5-HT1A receptors are found in certain regions,

Serotonin (5-Hydroxytryptamine; 5-HT): CNS Pathways and Neurophysiology 473

The somatodendritic autoreceptors of serotonergic neurons in both the dorsal raphe and other raphe nuclei appear to be predominantly of the 5-HT1A subtype, as a variety of drugs with 5-HT1A selectivity (e.g., 8-OH-DPAT) share the ability to potently inhibit raphe cell firing in a dose-dependent manner. Highly selective 5-HT1A antagonists (e.g., WAY 100635) have been found which potently block the direct inhibition of dorsal raphe serotonergic neurons both by serotonin and selective 5-HT1A agonists. As described above, the somatodendritic 5-HT1A autoreceptor is importantly involved in the negative feedback regulation of sertonergic cell activity through the release of endogenous serotonin.

Figure 4 The four major groupings of serotonin receptor families and their transduction pathways. Three of the four groups are coupled through G-proteins: 5-HT1-Gi/Go, 5-HT2-Gq, and 5-HT4,6,7-Gs. The remaining receptor is 5-HT3, which is a ligand-gated channel and is not coupled through a G protein. cAMP, adenosine 30 ,50 -cyclic monophosphate; DAG, diacylglycerol; PKC, protein kinase C; sAHP, slow after hyperpolarization; PLC, phospholipase C. Provided by GK Aghajanian.

including the dorsal raphe nucleus, hippocampus, lateral septum, and certain layers of the cerebral cortex. Studies in these regions and other regions have been useful in delineating the physiological role of this receptor. Raphe nuclei Serotonergic neurons of the raphe nuclei are inhibited by the local (microiontophoretic) application of serotonin to their cell body region. Thus, the receptor mediating this effect has been termed a somatodendritic autoreceptor (as opposed to the prejunctional autoreceptor). Functionally, the somatodendritic serotonin autoreceptor has been shown to mediate collateral inhibition. The ionic basis for the autoreceptor-mediated inhibition, either by serotonin or LSD, is an opening of Kþ channels to produce a hyperpolarization; these channels are characterized by their inwardly rectifying properties. Patch clamp recordings in the cell-attached and outside-out configuration from such acutely isolated dorsal raphe neurons show that the increase in Kþ current results from a greater probability of opening of unitary Kþ channel activity.

Other subcortical regions Inhibitory or hyperpolarizing responses to serotonin have been reported in a wide variety of neurons in the spinal cord, brain stem, and diencephalon. In general, such responses have been attributed to mediation by 5-HT1 receptors. In the nucleus prepositus hypoglossi, focal electrical stimulation evokes inhibitory postsynaptic potentials (IPSPs) that are mediated by 5-HT1A receptors to activate an inwardly rectifying Kþ conductance and a novel outwardly rectifying Kþ conductance. In the midbrain periaqueductal gray, a region known to be involved in pain modulation and fear responses, approximately half the cells are inhibited/hyperpolarized by 8-OH-DPAT, suggesting mediation by 5-HT1A receptors. In the ventromedial hypothalamus and lateral septum, serotonin and 5-HT1A agonists produce inhibitory effects also by activating a potassium conductance. In addition to these postsynaptic effects, serotonin has been shown to suppress glutamatergic synaptic transmission via presynaptic 5-HT1B receptors in various regions including the hypoglossal nucleus and the nucleus accumbens. In the rat laterodorsal tegmental nucleus (LDT), bursting cholinergic neurons are hyperpolarized by serotonin via 5-HT1 receptors. In freely behaving rats, the direct injecton of serotonin into the LDT was found to suppress REM sleep. In unanesthetized cats, a corresponding population of neurons that are active selectively during REM states (REM-on neurons) in the LDT are inhibited by direct application of the 5-HT1A agonist 8-OH-DPAT. It has been proposed that during REM sleep, the removal of a tonic inhibitory serotonin influence from these cholinergic neurons may be responsible for the emergence of an activated EEG during this behavioral state (see later). Hippocampus Pyramidal cells of the CA1 region express high levels of 5-HT1A receptor mRNA and 5-HT1A receptor binding. Serotonin-induced inhibition in both CA1 and CA3 pyramidal cells is mediated

474 Serotonin (5-Hydroxytryptamine; 5-HT): CNS Pathways and Neurophysiology

by the activation of receptors of the 5-HT1A subtype coupled to an opening of Kþ channels. In addition to the above direct effects on pyramidal cells, serotonin has been shown to depress both excitatory and inhibitory synaptic potentials in the hippocampus. Relatively high concentrations of serotonin cause a reduction in electrically evoked excitatory postsynaptic potentials (EPSPs) in CA1 pyramidal cells, an effect that is mimicked by 8-OH-DPAT, suggesting mediation by 5-HT1A receptors. Indirect measures indicate that serotonin acts presynaptically to reduce Ca2þ entry and thereby gluamatergic synaptic transmission. In addition, there is a 5-HT1A-mediated inhibitory effect on putative inhibitory interneurons of the hippocampus. Functionally, the 5-HT1A-mediated inhibition of GABAergic interneurons in the hippocampus leads to a disinhibition of pyramidal cells in CA1. Clearly, the effects of serotonin in the hippocampus are highly complex, involving both pre- and postsynaptic actions which may, to varying degrees, be inhibitory or disinhibitory, facilitatory or disfacilitatory. Cerebral cortex 5-HT1A-induced hyperpolarizing/ inhibitory responses in pyramidal cells of the cerebral cortex have been described in a number of studies. In entorhinal cortex, where there is an especially high density of 5-HT1A receptors (and a low density of 5-HT2A receptors), unopposed 5-HT1A receptormediated hyperpolarizing responses are seen. However, cortical neurons in most other regions typically display mixed inhibitory and excitatory responses to serotonin due to the expression by the same pyramidal cells of multiple serotonin receptor subtypes (e.g., 5-HT1A, and 5-HT2/2C). Hyperpolarizing responses mediated by 5-HT1A receptors are often unmasked or enhanced in the presence of 5-HT2 antagonists, consistent with the idea that there is an interaction between 5-HT1A and 5-HT2A receptors at an individual neuronal level. In addition to these postsynaptic effects, there are various presynaptic effects mediated by 5-HT1 receptors in the cerebral cortex. In cingulate cortex, serotonin, acting upon presynaptic 5-HT1B receptors, reduces the amplitude of electrically evoked EPSPs, including both N-methyl-D-aspartate (NMDA) and non-NMDA components. Similar modulations of EPSPs, mediated by 5-HT1A or 5-HT1B receptors, have been reported for several cortical regions including medial prefrontal and entorhinal cortex. 5-HT2 Receptors

Members of the 5-HT2 family of receptors are coupled through Gq-type G proteins and have excitatory effects through closing of Kþ channels or opening

nonselective cation channels. High concentrations of 5-HT2 are expressed in certain regions of the forebrain such as the neocortex, piriform cortex, claustrum, and olfactory tubercle. With few notable exceptions (e.g., motor nuclei and the nucleus tractus solitarius), relatively low concentrations of 5-HT2 receptors or mRNA expression are found in the brain stem and spinal cord. Studies aimed at examining the physiological role of 5-HT2 receptors in several of these regions are described in the following sections. Motoneurons In spinal cord and brain stem motor nuclei, motoneurons have a high density of 5-HT2 receptor binding sites. Early studies in vivo showed that serotonin applied microiontophoretically does not by itself induce firing in the normally quiescent facial motoneurons but does facilitate the subthreshold and threshold excitatory effects of glutamate. Intracellular recordings from facial motoneurons in vivo or in brain slices in vitro show that serotonin induces a slow, subthreshold depolarization associated with an increase in input resistance, indicating a decrease in a resting Kþ conductance. Selective 5-HT2 antagonists are able to selectively block the excitatory effects of serotonin in facial motoneurons. The facilitation of motoneuron excitability contributes to the role of serotonin as a ‘wake-on’ system to promote activity during the waking state (see later). Other subcortical regions In brain slices of the medial pontine reticular formation, serotonin induces depolarizing responses that have a 5-HT2 pharmacology and are associated with a decrease in membrane conductance resulting from a decrease in an outward Kþ current. In brain slices of the substantia nigra pars reticulata, a majority of neurons are excited by serotonin via 5-HT2 receptors, possibly of the 5-HT2C rather than 5-HT2A subtype. Neurons in the inferior olivary nucleus are excited by serotonin via 5-HT2A receptors, thereby altering the oscillatory frequency of input to cerebellar Purkinje cells. In the nucleus accumbens, the great majority of neurons are depolarized by serotonin, inducing them to fire. This depolarization is associated with an increase in input resistance due to a reduction in an inward rectifier Kþ conductance. In addition, GABAergic neurons within a number of regions (e.g., dorsal raphe nucleus, medial septal nucleus, hippocampus, cerebral cortex) are also excited by serotonin via 5-HT2 receptors, suggesting that in multiple locations within the CNS there are subpopulations of interneurons that are excited by serotonin via 5-HT2 receptors, giving rise to indirect inhibitory effects.

Serotonin (5-Hydroxytryptamine; 5-HT): CNS Pathways and Neurophysiology 475

Cerebral cortex The electrophysiological effects of serotonin have been studied in several corticalregions. In vitro studies in the brain slice preparation have shown that pyramidal cells in various cortical regions respond to serotonin by either a small hyperpolarization, depolarization, or no change in potential. Depending on the region of cortex under study, as described below, the depolarizations appear to be mediated by 5-HT2A or 5-HT2C receptors. In addition to these postsynaptic effects, serotonin induces an increase in ‘spontaneous’ (nonelectrically evoked) postsynaptic potentials or currents (PSPs/PSCs) recorded in brain slices from various cortical regions. These may originate from the activation of intracortical pathways or through interactions with subcortical inputs such as the thalamus. Activation of 5-HT2A receptors also enhances late components of evoked responses in cerebral cortex (Figure 5). These late responses (also termed UP states) are generated by sustained recurrent network activity. It has recently been shown that UP states rely on glutamate spillover onto extrasynaptic NMDA receptors. Interestingly, these UP states are characteristic of alert waking and are enhanced by psychedelic hallucinogens which are

known to be 5-HT2A/2C partial agonists. It has been proposed that the hallucinogenic drugs produce their characteristic aberrations in perception, cognition, and affect by driving cortical UP states beyond normal limits. It is significant that the facilitation of UP states by 5-HT2A receptors is opposed by 5-HT1A receptors which tend to suppress UP states. This damping effect of 5-HT1A receptors on UP states may explain why selective 5-HT2A agonists are hallucinogenic, whereas serotonin, the natural transmitter, is not. This example further illustrates how serotonin, rather than having a monolithic action, exercises numerous checks and balances at a cellular and systems level via its diverse receptor subtypes.

Figure 5 Enhancement of electrically evoked recurrent network activity (prolonged late excitatory postsynaptic current (EPSC) or UP state) in prefrontal brain slice by the hallucinogen LSD. Under basal conditions (top traces), note the fast EPSC, which occurs within a few milliseconds of a local electrical stimulus applied to the brain slice; following the fast EPSC, a low-amplitude late EPSC can be seen. After the application of a low concentration of LSD in the bath, the late EPSC or UP state is greatly enhanced both in amplitude and duration. Whole cell recording from a layer V pyramidal cell in rat prefrontal cortex. Provided by GK Aghajanian.

5-HT4, 5-HT6, 5-HT7

5-HT3 Receptors

Excitatory responses to serotonin have been found in various central neurons that have a rapid onset of action and rapid desensitization, features that are typical of ligand-gated ion channels rather than G-protein-coupled receptor responses. The cloned 5-HT3 receptor homologous with the nicotinic acetylcholine receptor and the b1 subunit of the GABAA receptor indicate that it is a member of the ligandgated ion channel superfamily. Typically members of this superfamily are comprised of multiple subunits; however, only one 5-HT3 receptor subunit and an alternatively spliced variant has been cloned to date. In hippocampal slices, serotonin has been reported to increase spontaneous GABAergic IPSPs, most likely through a 5-HT3 receptor-mediated excitation of inhibitory interneurons; these responses also show fading with time. A similar induction of 5-HT3 receptor-mediated induction of inhibitory postsynaptic currents (IPSCs) has been reported in the neocortex. While fast, rapidly inactivating excitation has generally become accepted as characteristic of 5-HT3 receptors, non-desensitizing responses have also been reported. In dorsal root ganglion cells, a relatively rapid but non-inactivating depolarizing response has been described that has a 5-HT3 pharmacological profile. In neurons of the nucleus tractus solitarius brain slices, there is a postsynaptic depolarizing response to 5-HT3 agonists that is not rapidly desensitizing.

There are three known Gs-coupled serotonin receptors: 5-HT4, 5-HT6, and 5-HT7, all mediating positive coupling of serotonin responses to adenylyl cyclase. The resulting increase in cyclic AMP can produce its electrophysiological effects through interacting directly with ion channels or through the activation of the PKA signal transduction pathway, which can interact

476 Serotonin (5-Hydroxytryptamine; 5-HT): CNS Pathways and Neurophysiology

with a myriad of ion channels, transporters, and other kinases. At this time, electrophysiological studies are available only for the 5-HT4 and 5-HT7 receptors, as described below. 5-HT4 receptors Binding studies using a selective 5-HT4 ligand indicate that 5-HT4 receptors are present in several discrete regions of the mammalian brain including the striatum, substantia nigra, olfactory tubercle, and hippocampus. As these regions also express 5-HT4 receptor mRNA, it appears likely that these receptors function postsynaptically to mediate certain actions of serotonin. The best studied of these regions is the hippocampus where both biochemical and electrophysiological studies have provided detailed picture of the actions of serotonin at 5-HT4 receptors. Electrophysiological studies show that 5-HT4 receptors mediate an inhibition of a calcium-activated potassium current, which is responsible for the generation of a slow afterhyperpolarization (AHP) in hippocampal pyramidal cells of the CA1 region. A suppression of the AHP would enhance the ability of these cells to respond to excitatory inputs with robust spike activity. 5-HT7 receptors The 5-HT7 receptor is positively coupled to adenylate cyclase and can mediate electrophysiological actions through the cyclic AMP pathway. One such effect is through a cyclic nucleotide enhancement of the hyperpolarizing-activated nonselective cationic current Ih. An increase in Ih tends to prevent excessive hyperpolarization and increase neuronal excitability. 5-HT7 receptors can also influence neuronal excitability through cyclic AMPactivated PKAs, leading to suppression of slow AHPs. This occurs in the midline/intralaminar nuclei of the thalamus where the highest expression of 5-HT7 receptors in the brain occurs. As the thalamocortical pathway projecting from the midline/ intralaminar nuclei is the final link in the ascending arousal pathway, this is one way that serotonin exerts its effects as a wake-on system to promote alertness and attention (see later).

Behavioral and Clinical Correlations The diversity of receptors and transduction pathways underlying the varied electrophysiological actions of serotonin, together with the differential expression of these receptors in different neuronal populations, helps explain how one transmitter can be involved in such a large array of behaviors, clinical conditions, and drug actions. Alterations in serotonin function have been linked to affective disorders, anxiety states,

schizophrenia, obsessive–compulsive disorder, eating disorders, migraine, and sleep disorders, including sleep apnea. Drugs that modify serotonergic transmission include antidepressants, atypical antipsychotics, antiemetics, hallucinogens, antimigraine drugs, and appetite suppressants. For example, the most commonly used antidepressant drugs act by blocking serotonin reuptake, which results in increased synaptic availability of this transmitter. Antagonism of 5HT2A receptors has been proposed as contributing to the favorable clinical profile of atypical antipsychotic drugs. Conversely, an agonist action of psychedelic hallucinogens at 5-HT2A receptors is thought to mediate the psychotomimetic effects of these drugs. Because of its ubiquitous distribution and diverse actions in the central nervous system, it is likely that future research will continue to uncover an involvement of serotonin in many pathological as well as normal behavioral processes. It has been hypothesized that activation of serotonin neurons works in concert with other components of the ascending arousal system projecting to the forebrain. For example, noradrenergic neurons, which also constitute a wake-on system, directly activate serotonergic neurons. A further linkage is seen by the fact that both noradrenergic and serotonergic neurons are activated by inputs from hcrt/orexin neurons, which are essential for normal wakefulness. A concerted action is also seen upstream, by the excitatory/facilitatory effects of both serotonin and hcrt upon the relay neurons of the midline/intralaminar nuclei which project to prefrontal cortex. This prefrontal input, which represents the final synapse in the ascending arousal system, is essential for awareness and attention. Within the cortex itself, 5-HT2A receptors enhance UP states, which are closely associated with the alert waking state. However, excessive stimulation of 5-HT2A receptors by hallucinogenic drugs can be pathological by leading to uncontrolled UP states. In conclusion, through its effects on neuronal excitability in diverse regions of the brain and spinal cord, the serotonergic system is positioned to coordinate complex sensory and motor patterns during different behavioral states. More specifically, since the activity of serotonergic neurons is greatest during periods of active waking, reduced during slow-wave sleep, and almost completely absent during REM (dream) sleep, the serotonin system can be seen as promoting neuronal activity in support of the waking state while concomitantly suppressing neuronal activity underlying sleep and dream states. See also: Serotonin (5-Hydroxytryptamine; 5-HT): Neurotransmission and Neuromodulation; Serotonin (5-Hydroxytryptamine; 5-HT): Receptors.

Serotonin (5-Hydroxytryptamine; 5-HT): CNS Pathways and Neurophysiology 477

Further Reading Aghajanian GK and Sanders-Bush E (2002) Serotonin. In: Davis KL, Charney D, Coyle JT, et al. (eds.) Neuropsychopharmacology: The Fifth Generation of Progress, pp. 15–46. New York: Raven Press, Ltd. Jacobs BL and Azmitia EC (1992) Structure and function of the brain serotonin system. Physiological Reviews 72: 165–229. Lambe EK and Aghajanian GK (2006) Electrophysiology of 5HT2A receptors and relevance for hallucinogen and atypical drug actions. In: Roth BL (ed.) The Serotonin Receptors: From Molecular Pharmacology to Human Therapeutics. Totowa, NJ: Humana Press. Lambe EK and Aghajanian GK (2006) Hallucinogen-induced UP states in the brain slice of rat prefrontal cortex: Role of

glutamate spillover and NR2B-NMDA receptors. Neuropsychopharmacology 31: 1682–1689. Liu R-J, Lambe EK, and Aghajanian GK (2005) Somatodendritic autoreceptor regulation of serotonergic neurons: Dependence on L-tryptophan and tryptophan hydroxylase-activating kinases. European Journal of Neuroscience 21: 945–958. Liu R-J, van den Pol AN, and Aghajanian GK (2002) Hypocretins (orexins) regulate serotonin neurons in the dorsal raphe nucleus by excitatory direct and inhibitory indirect actions. Journal of Neuroscience 22: 9453–9464. Richerson GB (2004) Serotonergic neurons as carbon dioxide sensors that maintain pH homeostasis. Nature Reviews Neuroscience 5: 449–461. Whitaker-Azmitia PM (1999) The discovery of serotonin and its role in neuroscience. Neuropsychopharmacology 21: 2S–8S.

Cholinergic Neurotransmission in the Autonomic and Somatic Motor Nervous System T C Westfall, St. Louis University School of Medicine, St. Louis, MO, USA ã 2009 Published by Elsevier Ltd.

Introduction Acetylcholine is the principal neurotransmitter at all autonomic preganglionic nerve terminals (sympathetic and parasympathetic), all postganglionic parasympathetic neuroeffector junctions, a few postganglionic sympathetic neuroeffector junctions (e.g., eccrine sweat glands), the somatic motor nerve neuromuscular junction, and cholinergic synapses in the central nervous system (CNS). The synthesis, storage, release, and inactivation of acetylcholine are similar at all these structures. This is probably also true for nonneuronal cholinergic systems in various parts of the body (e.g., human airway). The neurochemical events involved in the life cycle of acetylcholine are depicted in Figure 1.

Synthesis of Acetylcholine Choline and Choline Transport

Critical to the synthesis of acetylcholine is the availability of choline. There is little de novo synthesis of choline in cholinergic neurons; therefore, choline is provided mostly from the diet. Choline is a dietary component essential for normal function of all cells, not just cholinergic neurons. Choline or its metabolites ensure the structural integrity and signaling functions of all cell membranes; it is the major source of methyl groups in the diet, and it obviously directly affects cholinergic neurotransmission. Cholinergic neurons have especially large requirements for choline because they use choline not only for membrane synthesis but also for acetylcholine synthesis. Choline is taken up from the extracellular space by two transport systems. First is the ubiquitous low-affinity, sodium-independent transport system that is inhibited by hemicholinium 3 with a Ki of about 50 mmol l 1. Second is the high-affinity, sodium- and chloride-dependent, hemicholinium 3-sensitive (Ki ¼ 10–100 nmol l 1) choline transport system. The latter form is found predominantly in cholinergic neurons and is responsible for providing choline for acetylcholine synthesis. Once acetylcholine is released from cholinergic neurons following the arrival of action potentials, acetylcholine is hydrolyzed by acetylcholinesterase (AChE; see the section titled

478

‘Enzymatic hydrolysis: AChE’) to free acetate and choline. Choline is recycled after reuptake into the nerve terminal of cholinergic cells and utilized for acetylcholine synthesis. Under many circumstances, this reuptake and availability of choline appear to serve as the rate-limiting step in acetylcholine synthesis. Acetylcholine synthesis and release are sustained during persistent and prolonged neuronal stimulation, as long as choline is available and the transporter is functional. The low-affinity choline uptake transporter is used to synthesize membrane phospholipids. These lipids are reservoirs from which choline can also be used by choline neurons for acetylcholine synthesis. Despite the fact that the high-affinity choline uptake system was first described in the1960s, the gene for the high-affinity choline transporter (CHT1) has only recently been cloned from a variety of species: Caenorhabditis elegans, mouse, Torpedo, Limulus, and humans. It is interesting that CHT1 is not homologous to neurotransmitter transporters but is homologous to the Naþ-dependent glucose transporter family. With the molecular identification of the CHT1, it was discovered that it does not predominantly reside at the nerve terminal plasma membrane but rather in intracellular vesicular structures. It has been observed that CHT1 co-localizes with synaptic vesicle markers such as vesicle-associated membrane protein 2 and vesicular acetylcholine transporter (VAChT), at least when expressed in a cholinergic cell line. Although information is still incomplete, it has been hypothesized that arrival of action potentiates in nerve terminals results in exocytosis of synaptic vesicles, release of acetylcholine into the synaptic cleft, and increased trafficking of CHT1 to the plasma membrane, where it functions to take up choline after hydrolysis of acetylcholine. Because choline transport is rate-limiting for acetylcholine synthesis, increased availability of choline via its transport by CHT1 would favor an increase in acetylcholine stores to maintain high levels of transmitter release during neuronal stimulation. This also suggests that the availability of CHT1 at the cell surface is dynamically regulated in a manner very similar to the regulation of the exocytosis of synaptic vesicles. Much recent information has been obtained on how trafficking of CHT1 between the cell surface and subcellular compartments can regulate choline uptake activity. Nevertheless, the precise mechanisms involved in maintaining the distribution of CHT1 predominantly in intracellular vesicles rather than at the terminal surface like other neurotransmitter transporters is unclear.

Cholinergic Neurotransmission in the Autonomic and Somatic Motor Nervous System 479

Hemicholinium Choline

AcCoA + choline +

−

ChAT

Na+

Mitochondrion ACh −

3Na+

Vesamicol ADP ATP

Ach ACh Co-T Co-T

Cholinergic varicosity

2K+ nAChR

+

Ca2+

Na+, K+ - ATPase

mAChR VAMPS Ca2+

Choline

Co-T ACh

Acetate

− Botulinium toxin

SNAPS ACh +

Effector cell membrane

+ AChE

mAChR (M1 − M5)

nAChR (NN − NM)

Figure 1 Schematic representation of a cholinergic neuroeffector junction showing features of the synthesis, storage, and release of acetylcholine (ACh) and receptors on which ACh acts. The synthesis of ACh in the varicosity depends on the uptake of choline via a sodium-dependent carrier. This uptake can be blocked by hemicholinium. Choline and the acetyl moiety of acetyl coenzyme A (AcCoA), derived from mitochrondria, form ACh, a process catalyzed by the enzyme choline acetyltransferase (ChAT). ACh is transported into the storage vesicle by another carrier that can be inhibited by vesamicol. ACh is stored in vesicles along with other potential cotransmitters (Co-T) such as adenosine triphosphate (ATP) and vasoactive intestinal polypeptide at certain neuroeffector junctions. Release of ACh and the Co-T occurs on depolarization of the varicosity, which allows the entry of Ca2þ through voltage-dependent Ca2þ channels. Elevated [Ca2þ]in promotes fusion of the vesicular membrane with the cell membrane, and exocytosis of the transmitters occurs. This fusion process involves the interaction of specialized proteins associated with the vesicular membrane, called vesicle-associated membrane proteins (VAMPs) and the membrane of the varicosity, called synaptosome-associated proteins (SNAPs). The exocytotic release of ACh can be blocked by botulinum toxin. Once released, ACh can interact with the muscarinic receptors (mAChRs), which are G-protein-coupled receptors, or nicotinic receptors (nAChRs), which are ligand-gated ion channels, to produce the characteristic response of the effector. ACh can also act on presynaptic mAChR or nAChR to modify its own release. The action of ACh is terminated by metabolism to choline and acetate by the enzyme acetylcholinesterase (AChE), which is associated with synaptic membranes. ADP, adenosine diphosphate; ATPase, adenosine triphosphatase. From Westfall TC and Westfall DP (2006) Neurotransmission: The autonomic and somatic motor nervous systems. In: Brunton LL, Lazo JS, and Parker KL (eds.) Goodman and Gilman’s the Pharmacological Basis of Therapeutics, 11th edn., pp. 137–182. New York: McGraw-Hill.

Choline Acetyltransferase

The synthesis of acetylcholine occurs within the cytoplasm of cholinergic nerve terminals by a reaction in which choline is acetylated with acetyl coenzyme A (CoA) by the enzyme choline acetyltransferase. Acetyl CoA is derived mainly from glycolysis and is ultimately produced by the enzyme pyruvate

dehydrogenase. The synthesis of acetyl CoA occurs at the inner membrane of mitochondria, and it is transported by citrate to the cytoplasm, where citrate is freed by citrate lyase. Choline acetyltransferase is synthesized in the perkaryon and transported along the length of the axon to the terminal by axoplasmic flow.

480 Cholinergic Neurotransmission in the Autonomic and Somatic Motor Nervous System

Storage of Acetylcholine Following its synthesis in the cytoplasm of the nerve terminal, acetylcholine is transported into synaptic vesicles by VAChT, which uses a proton electrochemical gradient to move acetylcholine to the inside of the organelle. Hydrophobicity algorithm prediction of secondary structure of VAChT suggests a protein comprising 12 transmembrane domains, with hydrophilic N- and C-termini in the cytoplasm. Sequence homology places VAChT into a family of transporter proteins that includes two vesicular monoamine transporters. An adenosine triphosphatase that pumps protons into the vesicle provides the energy necessary for the transport. Transport of protons out of the vesicle is coupled to uptake of acetylcholine into the vesicle and against a concentration gradient via an acetylcholine antiporter. Two types of vesicles appear to exist in cholinergic terminals: an electron-luscent vesicle (40–50 nm in diameter) and dense core vesicles (80–150 nm). The vesicle core contains both acetylcholine and adenosine triphosphate (ATP), which are dissolved in the fluid phase with metal ions (Ca2þ and Mg2þ) and a protoglycan vesiculin. In some cholinergic terminals, there are also peptides such as vasoactive intestinal polypeptide (VIP) that act together with ATP and acetylcholine as cotransmitters. The peptides are thought to reside in the larger dense-core vesicles. Vesicular membranes are rich in lipids (cholesterol and phospholipids) as well as proteins. The VAChT allows for the transport of acetylcholine agonist, which has a considerable concentration gradient and is saturable and adenosine triphosphatasedependent. Vesamicol is a drug which inhibits the transport of acetylcholine via targeting the acetylcholine Hþ antiporter channel. Vesamicol leads to a reduction in acetylcholine storage and interferes with neurosecretion. Although not precisely known, the content of synaptic vesicles is estimated to range from 1000 to more than 50 000 molecules per vesicle. There may also be large amounts of acetylcholine in the extravesicular cytoplasm.

Release of Acetylcholine Release of acetylcholine and cotransmitters (e.g., ATP and VIP) occurs on depolarization of the nerve terminals and takes place by exocytosis. Depolarization of the terminals allows the entry of Ca2þ through voltage-gated Ca2þ channels. Elevated Ca2þ concentration promotes fusion of the vesicular membrane with the plasma membrane, allowing exocytosis to occur. The molecular mechanisms involved in the release and regulation of release have been the subject of intense study but are still not completely

understood. As mentioned, acetylcholine, like other neurotransmitters, is stored in vesicles located at special release sites, close to presynaptic membranes and ready for release on the appropriate stimulus. The vesicles initially dock and are primed for release. A multiprotein complex appears to form and attach the vesicle to the plasma membrane close to other signaling elements. The complex involves proteins from the vesicular membrane and the presynaptic neuronal membrane, as well as other components that help link them together. It has been established that these various synaptic proteins, including the plasma membrane protein syntaxin and synaptosomal protein 25 kDa (SNAP-25), and the vesicular membrane protein, or synaptobrevin, form a complex. This complex interacts in an ATP-dependent manner with soluble N-ethylmalemide-sensitive fusion protein and soluble SNAPs. The ability of synaptobrevin, syntaxin, and SNAP-25 to bind SNAPs has led to their designation as SNAP regulators (SNARES). It has been hypothesized that most, if not all, intracellular fusion events are mediated by SNARE interactions. Important evidence supporting the involvement of SNARE proteins in transmitter release comes from the fact that botulinum neurotoxins and tetanus toxin, which block neurotransmitter release, proteolyze these three proteins. Two pools of acetylcholine appear to exist. One pool, the ‘depot’ or ‘readily releasable’ pool, consists of vesicles located near the plasma membrane of the nerve terminals and is filled with newly synthesized transmitter. Depolarization of the terminals causes these vesicles to release acetylcholine rapidly or readily. The other pool, the ‘reserve’ pool, seems to replenish the readily releasable pool and may be required to sustain acetylcholine release during periods of sustained or intense nerve stimulation.

Regulation of Acetylcholine Neurotransmission It is now well accepted that receptors are located on perkaryon, dendrites, and axons of neurons when they may respond to neurotransmitters or neuromodulators released from the same neurons (autoreceptors) or adjacent neurons or cells (heteroreceptors). Cholinergic nerve terminals also contain autoreceptors and heteroreceptors. Acetylcholine neurotransmission is therefore subject to complex regulation by mediators including acetylcholine itself acting on M2 and M4 autoinhibitory receptors and nicotinic acetylcholine excitatory receptors. Acetylcholine-mediated inhibition of acetylcholine release via muscarinic cholinergic receptors M2 and M4 is thought to represent a physiological

Cholinergic Neurotransmission in the Autonomic and Somatic Motor Nervous System 481

negative feedback control mechanism. Acetylcholine released from cholinergic neurons may also alter the release of other neurotransmitters. For instance, at the neuroeffector junction in the myenteric plexus in the gastrointestinal tract or sinoatrial node of the heart, postganglionic sympathetic and parasympathetic terminals lie in juxtaposition to each other. The release of acetylcholine not only produces autoinhibition of its own release but may attenuate the release of norepinephrine via an inhibitory action of muscarinic heteroreceptors located on sympathetic terminals. The reverse is also true: after its release of sympathetic nerve terminals, norepinephrine can inhibit the release of acetylcholine by an action of a2A and a2C heteroreceptors on parasympathetic nerve terminals. Muscarinic and nicotinic auto- and heteroreceptors also represent drug targets, or targets for locally formed autacoids, hormones, or transmitters.

Inactivation of Acetylcholine Enzymatic Hydrolysis: AChE

Subsequent to its release from cholinergic neurons, acetylcholine is subjected to a number of inactivation processes, including (1) diffusion from the site of release, (2) dilution in extracellular fluids, (3) binding to nonspecific sites, and (4) enzymatic destruction. The most important of these by far is the enzymatic destruction of acetylcholine to its hydrolysis products, acetic acid and choline. The reaction is catalyzed by AChE. While AChE is found in cholinergic neurons (dendrites, perikarya, and axons), it is distributed more widely than cholinergic synapses. It is highly concentrated at postsynaptic sites, especially at the neuromuscular junction. AChE exists in red blood cells and the placenta as well as nervous tissue. A second enzyme is butyrylcholinesterase (BuChE). BuChE exists in intestine and skin and abundantly in plasma. AChE can be distinguished from BuChE by selectivity of acetylcholine over butyrylcholine hydrolysis. BuChE shows a wider substrate capacity than AChE does. Almost all the pharmacological effects of AChE drugs are due to the inhibition of AChE, with a consequent accumulation of endogenous acetylcholine in the vicinity of the nerve terminal. Distinct but single genes encode AChE and BuChE in mammals, and the diversity of molecular structure of AChE arises from alternate messenger RNA processing. Complementary DNAs for several species, including Torpedo, Drosophila, nematode, rat, mouse, and human AChE, as well as mouse, rabbit, and human BuAChE, are known. The contact between acetylcholine and the surface of AChE occurs at two places on the protein, One is

the anionic site, which is negatively charged and stereospecific. It attracts the positively charged nitrogen atom of acetylcholine and binds the attracted methyl groups by Van der Waals forces. The other site is the ‘esteratic’ site, which combines the electrophilic carboxy carbon atom of the acetyl group and the serine hydroxyl group of the esteratic site (the nucleophilicity of which is enhanced by the histadine imidazole group). The three stages for the hydrolysis are as follows: (1) equilibrium is established between the enzyme and acetylcholine, forming an enzyme substrate complex; (2) the complex reacts to release choline and leaves an acetylated enzyme; and (3) the acetylated enzyme reacts with water to give acetic acid and regenerated enzyme with a t½ of approximately 40 ms. Although the principal function of AChE is to inactivate acetylcholine at cholinergic neuronal junctions, it has been reported to have multiple biological functions that are not obvious. These include neritogenesis, cell adhesion, synaptogenesis, amyloid fiber assembly, and activation of dopamine receptors, hematopoiesis, and thrombopoiesis, to name but a few.

Cholinergic Receptors Once released from cholinergic neurons, acetylcholine produces its physiological or pharmacological effects by first interacting with specific receptors. The various responses are mediated by the interaction of the transmitter with different types of receptors located in different organs, tissues, or cells. There are two general types of cholinergic receptors: muscarinic and nicotinic. The original classification was based on acetylcholine’s mimicking the effect of the alkaloid muscarine (muscarinic) or nicotine (nicotinic). Multiple subtypes are now known to exist. Originally the subtypes of receptors were defined by the rank order of potency of agonists and the specificity of antagonists (e.g., atropine as an antagonist of muscarinic receptors and hexamethonium or curare as antagonists for nicotinic acetylcholine receptors (nAChRs)). More recently, the various subtypes have been identified by molecular cloning of the genes. nAChRs

The nAChRs are members of a superfamily of ligandgated ion channels (ionotropic receptors). They primarily exist at the neuromuscular junction, autonomic ganglia and adrenal medulla, parasympathetic and sympathetic nerve terminals, and central nervous system. These receptors are the physiologic target of the neurotransmitter acetylcholine as well as naturally occurring alkaloids and synthetic drugs. The nAChR

Receptor (primary receptor subtype)a

Main synaptic location

Membrane response

Molecular mechanism

Agonists

Antagonists

Skeletal muscle (Nm) (a1)2 b1 e d Adult (a1)2 b1 g d Fetal

Skeletal neuromuscular junction (postjunctional)

Excitatory; endplate depolarization; skeletal muscle contraction

Increased cation permeability (Naþ; Kþ)

ACh Nicotine Succinylcholine

Peripheral neuronal (NN) (a3)2 (b4)3

Autonomic ganglia; adrenal medulla

Increased cation permeability (Naþ; Kþ)

ACh Nicotine Epibatidine Dimethylphenylpiperazinum

Central neuronal (CNS) (a4)2 (b4)3 (a-btox-insensitive)

CNS; pre- and postjunctional

Excitatory; depolarization firing of postganglion neuron; depolarization and secretion of catecholamines Pre- and postsynaptic excitation Prejunctional control of transmitter release Pre- and postsynaptic excitation Prejunctional control of transmitter release

Atracurium Vecuronium d-Tubocurarine Pancuronum a-conotoxin a-bungarotoxin Trimethaphan Mecamylamine

Increased cation permeability (Naþ; Kþ)

Cytisine, epibatidine Anatoxin A

Increased permeability (Ca2þ)

Anatoxin A

(a7)5 (a-btox-sensitive)

CNS; pre- and postsynaptic

Mecamylamine Dihydro-b-erythrodine Erysodine Lophotoxin Methllycaconitine a-Bungarotoxin a-Conotoxin IMI

a Nine individual subunits have been identified and cloned in human brain which combine in various confirmations to form individual receptor subtypes. The structure of individual receptors and the subtype composition is incompletely understood, however. Only a finite number of naturally occurring functional nAChRs constructs have been identified to date. a-btox, a-bungarotoxin. Adapted from Westfall TC and Westfall DP (2006) Adrenergic agonists and antagonists. In: Brunton LL, Lazo JS, and Parker KL (eds.) Goodman and Gilman’s the Pharmacological Basis of Therapeutics, 11th edn., pp. 237–296. New York: McGraw-Hill.

482 Cholinergic Neurotransmission in the Autonomic and Somatic Motor Nervous System

Table 1 Characteristics of subtypes of nicotinic acetylcholine receptors

Table 2 Characteristics of muscarinic acetylcholine receptor subtypes Size: chromosome location

Cellular and tissue locationa

Cellular responseb

Functional responsec

M1

460 aa 11q 12–13

CNS; most abundant in cerebral cortex, hippocampus, and striatum Autonomic ganglia Glands (gastric and salivary) Enteric nerves

Increased cognitive function (learning and memory) Increased seizure activity Decreased dopamine release and locomotion Increased depolarization of autonomic ganglia Increased secretions

M2

466 aa 7q 35–36

Widely expressed in CNS, heart, smooth muscle, autonomic nerve terminals

Activation of PLC; " IP3 and " DAG ! " Ca2+ and PKC Depolarization and excitation (" sEPSP) Activation of PLD2 " AA Couples via Gq/11 Inhibition of adenylyl cyclase # cAMP Activation of inwardly rectifying K+ channels Inhibition of voltage-gated Ca2þ channels Hyperpolarization and inhibition Couples via Gi/Go (PTX sensitive)

M3

590 aa Iq 43–44

Widely expressed in CNS (< than other mAChRs) Abundant in smooth muscle and glands Heart

Activation of PLC; " IP3 and " DAG ! " Ca2+ and PKC Depolarization and excitation (" sEPSP) Activation of PLD2 ; PLA2; " AA Couples via Gq/11

M4

479 aa 11p 12–11.2

Preferentially expressed in CNS, particularly forebrain

M5

532 aa 15q 26

Expressed in low levels in CNS and periphery. Predominant mAchR in dopamine neurons in VTA and substantia nigra

Inhibition of adenylyl cyclase #cAMP Activation of inwardly rectifying Kþ channels Inhibition of voltage-gated Ca2þ channels Hyperpolarization and inhibition Couples via Gi/Go (PTX sensitive) Activation of PLC; " IP3 and " DAG ! " Ca2+ and PKC Depolarization and excitation (" sEPSP) Activation of PLD2, PLA2; " AA Couples via Gq/11

a

Heart: SA node: slowed spontaneous depolarization; hyperpolarization #HR AV node: decrease in conduction velocity Atrium: # refractory period, # contraction Ventricle: slight # contraction Smooth muscle: "Contraction Peripheral nerves: Neural inhibition via autoreceptors and heteroreceptor) # Ganglionic transmission. CNS: Neural inhibition " Tremors; hypothermia; analgesia Smooth muscle: " contraction (predominant in some, e.g., bladder) Glands: " secretion (predominant in salivary gland) Increases food intake, body weight fat deposits. Inhibits dopamine release Synthesis of NO Autoreceptor and heteroreceptor mediated inhibition of transmitter release in CNS and periphery. Analgesia; cataleptic activity Facilitation of dopamine release Mediator of dilation in cerebral arteries and arterioles (?) Facilitates dopamine release Augmentation of drug-seeking behavior and reward (e.g., opiates, cocaine)

Most organs, tissues, and cells express multiple mAChRs. M1, M3, and M5 mAChRs appear to couple to the same G-proteins and signal through similar pathways. Likewise, M2 and M4 mAChRs couple through similar G-proteins and signal through similar pathways. c Although multiple subtypes of mAChRs coexist in many tissues, organs, and cells, one subtype may predominate in producing a particular function, or there may be equal predominance. mAChR, muscarinic acetylcholine receptor; PLC, phospholipase C; IP3, inositol-1,4,5-triphosphate; DAG, diacylglycerol; PKC, protein kinase C; EPSP, excitatory postsynaptic potential; PLD2, phospholipase D; AA, arachidonic acid; CNS, central nervous system; PLA, phospholipase A; cAMP, cyclic adenosine monophosphate; SA node, sinoatrial node; AV node, atrioventricular node; HR, heart rate; PTX, pertussis toxin; VTA, ventral tegmental area. Adapted from Westfall TC and Westfall DP (2006) Adrenergic agonists and antagonists. In: Brunton LL, Lazo JS, and Parker KL (eds.) Goodman and Gilman’s the Pharmacological Basis of Therapeutics, 11th edn., pp. 237–296. New York: McGraw-Hill. b

Cholinergic Neurotransmission in the Autonomic and Somatic Motor Nervous System 483

Receptor

484 Cholinergic Neurotransmission in the Autonomic and Somatic Motor Nervous System Table 3 Summary of peripheral cholinergic receptors and responses Organ Eye Sphincter (circular) muscle of iris Iris Ciliary muscle Heart SA node Atria AV node Ventricle Blood vesselsa Arteries Veins Lung Bronchial muscle Bronchial glands Gastrointestinal tract Motility Sphincter Secretion Urinary bladder Detrusor muscle Trigone and sphincter Glands Sweat, salivary, lacrimal, bronchial, etc. Nasopharyngeal Nerves Parasympathetic terminal Sympathetic terminal Autonomic ganglia Skeletal muscle (Neuromuscular junction)

Response

Receptor

Contraction (miosis) Contraction (accommodation)

M3, M2 M3, M2

Decrease in rate (negative chronotropic) Decrease in contractile strength (negative inotropic) Decrease in refractive period Decrease in conduction velocity (negative dromotropic effect) Increase in refractory period Small decrease in contractile strength

M2M3 M2M3 M2M3 M2M3 M2M3

Dilation (via nitric oxide from endothelial cells) Dilation (via nitric oxide from endothelial cells)

M3 M3

Contraction (bronchoconstriction) Stimulation

M2 ¼ M3 M3, M2

Increase Relaxation Stimulation

M3 ( M2) M, M2 M3, M2

Contraction Relaxation

M3 > M2 M3 > M2

Secretion

M3

Decrease in release of acetylcholine Decrease in release of norepinephrine Stimulation

M2, M4 M2, M4 NN or N1

Contraction

NM or N2

a

Most blood vessels are not innervated by parasympathetic nerves but have M3 receptors on endothelial cells.

is a pentamer consisting of five subunits around a central ion channel pore. Several subunits have been identified: a, b, g, d, and e. All these subunits share 35–50% homology with one another. Depending on the location of the receptor, it has a different combination of subunits. At the neuromuscular junction (N1 or Nm), the AChR is composed of two a (a1) subunits; one b (b1); one d; and in embryonic or denervated muscle, one g subunit. In the adult innervated muscle, the g is replaced with one e subunit. Functional neuronal nAChRs (CNS Nn) also exist as pentamers composed of two a and three b subunits. Eight human a subunits (a2–a7, a9, and a10) and three b subunits (b2–b4) have been cloned. Both the muscle and the neuronal nAChR share structural and functional properties with other ionotropic receptors, such as g-aminobutyric acid A receptor, 5-hydroxytryptamine type 3, and glycine receptors. Autonomic ganglia (N2)

and the adrenal medulla form homomeric a7 and heteromeric a3/b4, with (a3)2 (b4)3 being the most prevalent. The a subunits are responsible for binding acetylcholine, and the conformational change in the a subunits is responsible for permitting ion flow through the central pore of the receptor. The pentameric structure of the CNS–neuronal receptor and the large molecular diversity of its subunits (a2–a8, b2–b4) suggest there could be a large number of nAChRs with different physiological properties. They may subserve discrete functions and obviously represent multiple drug targets. The stoichiometry of the nAChR in brain is the subject of intense study. Selective ligands are becoming more and more prominent, but it is not yet possible to make a pharmacological classification based on subtypes. The characteristics of subtypes of nAChR are described in Table 1.

Cholinergic Neurotransmission in the Autonomic and Somatic Motor Nervous System 485 Muscarinic Acetylcholine Receptors

There are five distinct subtypes of muscarinic acetylcholine receptors (mAChRs), M1–M5, each produced by a distinct gene and all members of the family of G-protein-coupled receptors (metabotropic receptors). Like the different nAChRs, M1–M5 have distinct anatomical locations in the periphery and CNS and different chemical specificities. The mAChRs have been shown to be present in virtually all organs, tissues, and cell types, but various subtypes predominate in various tissues and organs (e.g., M2 in the heart, M3 in the detrusor muscle of the bladder). The mAChRs are located in smooth muscle, exocrine and endocrine glands, and the myocardium. They mediate the actions of acetylcholine in organs and tissues innervated by postganglionic parasympathetic nerves and postganglionic sympathetic nerves innervating eccrine sweat glands. They are also present at sites that lack parasympathetic nerves (e.g., most blood vessels). Table 2 summarizes the effects produced by activation of mAChRs. In the CNS, mAChRs are involved in regulating a large number of cognitive, behavioral, motor, and autonomic functions although some of these are also influenced by nAChRs. Effects produced by activating mAChR are coupled through G-protein-induced changes in membrane-bound effectors and production of second-messenger molecules. The M1, M2, and M3 receptor subtypes couple through the pertussis toxininsensitive G11 and G13, resulting in stimulation of phospholipase C activity. This results in the hydrolysis of membrane phosphatidylinositol 4,5 diphosphate to form inositol triphosphate (IP3) and diaceylglycerol. IP3 causes release of intracellular Ca2þ from the endoplasmic reticulum, which results in smooth muscle contraction as well as secretion from the appropriate cell. Diaceylglycerol activates protein kinase C, resulting in phosphorylation of numerous proteins and various physiological responses. Activation of M1, M2, and M3 receptors also results in the activation of phospholipase A2, leading to the release of arachidonic acid and consequent eicosanoid synthesis. Stimulation of the M2/M4 receptors leads to interaction with Gi and Go proteins, which results in inhibition of adenylyl cyclase and the resultant decrease in cyclic adenosine monophosphate, activation of inwardly rectifying Kþ channels, and inhibition of

voltage-gated Ca2þ channels. Functional consequences are hyperpolarization and inhibition of excitable membranes.

Physiological Responses Produced Following Increases in Cholinergic Neurotransmission Table 3 depicts the physiological responses due to increases in cholinergic neurotransmission or stimulation of parasympathetic neurons. See also: Cholinergic Pathways in CNS; Muscarinic Receptors: Autonomic Neurons; Nicotinic Acetylcholine Receptors.

Further Reading DeBiasi M (2002) Nicotinic mechanisms in the autonomic control of organ systems. Journal of Neurobiology 53: 568–579. Dhein S, van Koppen CJ, and Brodde OE (2001) Muscarinic receptors in the mammalian heart. Pharmacological Reviews 44: 161–182. Ferguson SM and Blakely RD (2004) The choline transporter resurfaces: New roles for synaptic vesicles? Molecular Interventions 4: 22–37. Jahn R, Lang T, and Su¨dhof T (2003) Membrane fusion. Cell 112: 519–533. Lindstrom JM (2000) Acetylcholine receptors and myasthenia. Muscle & Nerve 23: 453–477. Ribeiro FM, Black SA, Prado VF, Rylett RJ, Ferguson SS, and Prado MA (2006) The “ins” and “outs” of the highaffinity choline transporter CHT1. Journal of Neurochemistry 97: 1–12. Soreq H and Seidman S (2001) Acetylcholinesterase: New roles for an old actor. Nature Reviews Neuroscience 2: 294–302. Taylor P (2006) Anticholinesterase agents. In: Brunton LL, Lazo JS, and Parker KL (eds.) Goodman and Gilman’s The Pharmacological Basis of Therapeutics, 11th edn., pp. 201–216. New York: McGraw-Hill. van Koppen CJ and Kaiser B (2003) Regulation of muscarinic acetylcholine receptor signaling. Pharmacology & Therapeutics 98: 197–220. Wang H and Sun X (2005) Desensitized nicotinic receptors in brain. Brain Research. Brain Research Reviews 48: 420–437. Wess J (2004) Muscarinic acetylcholine receptor knockout mice: Novel phenotypes and clinical implications. Annual Review of Pharmacology and Toxicology 44: 423–450. Westfall TC and Westfall DP (2006) Neurotransmission: The autonomic and somatic motor nervous systems. In: Brunton LL, Lazo JS, and Parker KL (eds.) Goodman and Gilman’s The Pharmacological Basis of Therapeutics, 11th edn., pp. 137–182. New York: McGraw-Hill.

Cholinergic Pathways in CNS A C Cuello, McGill University, Montreal, QC, Canada ã 2009 Elsevier Ltd. All rights reserved.

Acetylcholine as a Central Nervous System Transmitter Historically, acetylcholine (ACh) was considered to be the most important central nervous system (CNS) transmitter. As early as 1936, Quastel demonstrated that ACh was synthesized in the brain. Indeed, ACh was the first substance to be proposed to act as a CNS transmitter in the pioneering work of Marthe Vogt and Feldberg in the 1940s and 1950s, largely based on prior lessons obtained with this substance in the peripheral nervous system by a number of outstanding pharmacologists, ranging from Otto Loewi to Sir Henry Dale. Later, Fatt, Katz, and collaborators provided evidence for a quantal release of ACh in the neuromuscular junction, laying the foundations for the idea that this transmitter substance was somehow ‘packaged’ in nerve terminals. The realization that this was the case came about with the ultrastructural finding in 1955 of synaptic vesicles in CNS nerve terminals by De Robertis and Bennett, and also Palay and Palade, and the compelling demonstration that isolated synaptic vesicles obtained from subcellular fractionation were enriched in ACh content, as shown by De Robertis in Buenos Aires and Whitaker in Cambridge. These studies not only were important for the realization that ACh was a bona fide CNS transmitter but they also provided strong foundations to the chemical synaptic transmission hypothesis, a concept currently applied to a wide variety of so-called classical (e.g., amines, aminoacids) and nonclassical transmitters (e.g., peptides).

CNS Cholinergic Synapses Contrary to historical expectations, current methods have demonstrated that cholinergic synapses do not account for the majority of transmitter-specific synapses in the CNS. In the cerebral cortex, for example, cholinergic synapses account for about 5–7% of the total synaptic population. However, considerable variations in the number of cholinergic synapses across the CNS should be expected, given the degree of differential concentrations of the ACh-synthesizing enzyme (choline acetyltransferase, ChAT) in diverse brain areas. ChAT is localized mainly in the cytosolic compartment. The CNS cholinergic synapses are endowed with the expected set of organelles (e.g.,

486

mitochondria, synaptic vesicles) and molecules (e.g., classical synaptic proteins), but in addition they contain ‘signature’ proteins such as the high-affinity choline transporter (CHT1) and the vesicular acetylcholine transporter (VAChT). These two proteins endow cholinergic synapses with their physiological specifications. CHT1 plays an essential role in maintaining the required level of ACh synthesis, as it is the availability of choline rather than the ChAT enzymatic activity that is the rate-limiting factor for the synthesis of ACh. As expected, CHT1 is localized in synaptic membranes to facilitate the incorporation of the choline precursor. Dietary levels of choline have been shown to have an important impact on the replenishment of ACh stores in the CNS. Besides its predicted synaptic membrane localization, CHT1 is abundantly present in synaptic vesicles. This is somewhat perplexing, as ChAT is primarily a cytosolic enzyme and the bulk of the ACh synthesis is known to occur in the cytoplasm. This location might be a consequence of the retrieval of synapses after exocytosis, but the vesicular CHT1 might also play a role in mobilizing choline from vesicles to cytoplasm and vice versa. The VAChT is heavily localized in membranes of synaptic vesicles of cholinergic presynaptic endings. Its function is the straightforward uptake of newly synthesized cytoplasmatic ACh into the vesicular stores. The ACh-enriched vesicles are thus ready for release on demand. The newly synthesized ACh appears to be the first to be released, probably from vesicles closer to the synaptic membranes, while more distant vesicles containing ‘older’ ACh pools are seemingly located in the cholinergic presynaptic terminal further away from the point of synaptic contact. The cholinergic synaptic vesicles, from the early De Robertis descriptions, are recognized as being small, clear, and roundish. Figure 1 represents the main features of CNS cholinergic synapses. After the identification of the CNS cholinergic neurons and their projections (see later), it was proposed that their terminations at target sites, particularly in the cerebral cortex, were nonsynaptic in nature. In other words, it was assumed that CNS cholinergic presynaptic boutons release ACh diffusely in a ‘cloudlike’ modality. The idea of ‘volume’ or ‘extrasynaptic’ transmission of acetylcholine in the CNS was born out of the prevalent notion that CNS terminations containing monoamines are apparently nonsynaptic, in a manner reminiscent of the peripheral autonomic axonal terminations. The cholinergic cortical terminations were, therefore, included in the category of nonsynaptic neurons due to the lack of evidence for

Cholinergic Pathways in CNS 487 Acetate Choline ChAT CHT1 VAChT Acetate Choline ACh

AChE

ChAT

AChE Nicotinic receptor

Muscarinic receptors (M1−M5)

Figure 1 Schematic representation of typical CNS cholinergic synapses, as described in the text. ACh, acetylcholine; ChAT, choline acetyltransferase; CHT1, choline transporter 1; VAChT, vesicular acetylcholine transporter; AChE, acetylcholinesterase; M1–M5, muscarinic receptors; M2, presynaptic muscarinic receptor 2; N, nicotinic receptors.

synaptic specialization in ChAT-immunoreactive (IR) boutons when observed by electron microscopy. However, Mesulam and collaborators, using ChAT antibodies, noted that cholinergic boutons in the cerebral cortex of the human brain mainly displayed symmetric synaptic contacts. A more comprehensive ultrastructural study in the rat neocortex, applying improved protocols for tissue preservation and utilizing antibodies against VAChT, revealed that the cortical cholinergic presynaptic boutons in their majority establish classical synaptic contacts with symmetric membrane specializations (parallel membranes of equal thickness in the pre- and postsynaptic sides of contacts). These classical synaptic contacts occur preferentially in dendritic shafts and differ in their termination pattern from the bulk of ‘noncholinergic’ presynaptic boutons. Cholinergic presynaptic boutons are seldom observed contacting neuronal cell bodies and only occasionally contacting dendritic spines. These latter synapses are exclusively of an asymmetric nature (i.e., the postsynaptic membrane is notably thicker than the opposing presynaptic membrane). Figure 2 illustrates the typical ultrastructural characteristics of cholinergic (VAChT-IR) presynaptic sites (boutons) in layer V of the rat parietal cortex.

Main CNS Cholinergic Pathways Although the existence of CNS cholinergic neurons was well accepted, it took nearly half a century for

their unequivocal identification. Early attempts to define CNS systems operating with ACh applied immunohistochemical methods to reveal the presence (activity) of acetylcholinesterase (AChE), the AChinactivating enzyme, a procedure which was pioneered by Shute and Lewis at Cambridge University. This method revealed cell groups that later proved to be indeed cholinergic in nature, but it had the intrinsic problem that it demonstrated both ‘cholinergic’ and ‘cholinoceptive’ neurons. This approach was modified later by Butcher and collaborators when they introduced the so-called pharmaco-histochemical technique. This technique involved the inhibition of AChE, followed, after few hours, by the AChE histochemical reaction, to demonstrate newly generated AChE in cell body groups, thus avoiding the confounding images of AChE-positive neurites in the CNS tissue. While this approach was of some value, the problem remained that it also revealed several AChE-positive cells which were not cholinergic in nature. We now know that there is not a 100% correlation between ChAT-IR and AChEpositive histochemical reactions in CNS neurons. Notable examples are the conspicuous presence of AChE-positive material in dopaminergic substantia nigra and noradrenergic locus coeruleus neurons in the absence of ChAT immunoreaction. The chemical isolation of ChAT allowed the generation of reliable polyclonal and monoclonal antibodies, which opened the door for the reliable demonstration of the main groups of CNS cholinernergic neurons. Thus, a number

488 Cholinergic Pathways in CNS

Figure 2 Electron micrograph images of VAChT-IR boutons displaying synaptic specializations in layer V of the rat parietal cortex. (A) and (B) VAChT-IR boutons (asterisks) establishing symmetric synaptic contacts with dendritic branches (d). Note the aggregation of synaptic vesicles adjacent to the presynaptic membranes. (C) Illustration of a small VAChT-IR varicosity which is occasionally observed in a synaptic contact with a dendritic spine (s); note the synapse is of the asymmetric type. (D) Micrograph of one of the few VAChT-IR boutons establishing a synaptic contact with a cell body (cb), probably belonging to a nonpyramidal neuron. Synaptic contacts are indicated by two arrowheads; g, Golgi complex. Scale bar (for all micrographs) ¼ 0.25 mm. Reprinted from Turrini P, Casu MA, Wong TP, et al. (2001) Cholinergic nerve terminals establish classical synapses in the rat cerebral cortex: Synaptic pattern and age-related atrophy. Neuroscience 105(2): 277–285, with permission from Elsevier.

of reports arose during the early 1980s, applying immunohistochemistry to depict the localization of CNS ChAT-IR cell bodies and fiber tracts. At the time, the Swedish histochemists Fuxe and Dahlstrom introduced a letter and number nomenclature to define the catecholaminergic pathway. This proved popular and was widely accepted. Mesulam and collaborators, following the same principles, applied a Ch and number nomenclature to define CNS cholinergic nuclear organizations, as revealed with anti-ChAT monoclonal antibodies. These studies were the most exacting and comprehensive and provided a useful guide for further investigations, although the Ch classification is not universally applied. Following these developments, a number of lesion and track-tracing experiments revealed the existence of a variety of CNS pathways and local circuit cholinergic neurons. The main cholinergic cell groups and pathways are represented in Figure 3 in a parasaggital view of the rat CNS, where the prominent and

extensively studied basalocortical and septo-hippocampal pathways are highlighted in gray. In summary, these immunohistochemical investigations demonstrated that the ventral telencephalon displays a continuous stream of cholinergic cell bodies, notably in the olfactory tubercle, islands of Calleja, all components of the diagonal band, medial septum, nucleus preopticus magnocellularis, and the innermost region of the globus pallidus. These neurons are usually referred to (as an ensemble) as the basal forebrain cholinergic system. In rodents, the ChAT immunoreactive neurons distributed loosely in the innermost third of the globus pallidus represent the nucleus basalis, which is responsible for the large majority of cholinergic input to the cerebral cortex. The rodent cerebral cortex contains in addition a few homogenously dispersed, local circuit, fusiform cholinergic cell bodies. The organization of this ‘basalocortical’ pathway is very different from that of the

Cholinergic Pathways in CNS 489

C

H OB CP

CG

MH BN

TH

LDT

S AON DB IP

PPT

OT Ar

A Figure 3 Schematic sagittal view of the rat CNS, representing the main cholinergic pathways, as discussed in the text. OB, olfactory bulb; AON, anterior olfactory nucleus; DB, diagonal band; S, septum (cholinergic neurons restricted to medial division); CG, cingulated cortex; CP, large interneurons in the caudate putamen; H, hippocampus; BN, nucleus basalis; A, amygdala; TH, thalamus; Ar, arcuate nucleus; PPT, peduculo-pontine tegmental nucleus; LDT, lateral doral tegmental nucleus; C, cortex; IP, nucleus interpenducularis; MH, medial habenula; OT, olfactory tubercle. For clarity, cholinergic motor neurons and cholinergic preganglionar neurons are not represented. The gray bands represent the major forebrain basalocortical and septal–hippocampal cholinergic pathways. Adapted from Cuello AC and Sofroniew MV (1984) The anatomy of the CNS cholinergic neurons. Trends in Neurosciences 7: 74–78.

human brain. In the human brain, the equivalent cholinergic group is composed of large (ergo ‘magnocellularis’) multipolar neurons located in a flattened nuclear organization lying below the anterior commissure. This organization would correspond to the neuronal group, which is nowadays referred to as the nucleus basalis magnocellularis of Meynert, or simply the nucleus of Meynert. However, it has to be stressed that this nomenclature is only applicable to the human brain and to subhuman primate brains. The rodent structure should be referred to simply as the nucleus basalis. These nuclei, both in rodents and humans, correspond to Ch4 in Mesulam’s classification. A possible explanation of this differential organization of the nucleus basalis in primates and rodents might reside in the fact that in rodents the globus pallidus is split where the medial division of the globus pallidus becomes the entopenducular nucleus (also containing cholinergic neurons), while the external division becomes the globus pallidus. In consequence, in rodents, this component of the basal forebrain cholinergic neurons does not find the obstacle of the anterior commissure barrier, invading the innermost third of the globus pallidus in a diffuse manner. In the rodent, the most caudal portions of the nucleus basalis become a flattened neuronal group located between the optic tract and the last portion of the corpus striatum. Much attention has

been paid to this nuclear organization, as it has been reported to be compromised in Alzheimer’s disease (AD), a disease that, in advanced states, is accompanied by the loss of cortical cholinergic markers (see the section entitled ‘CNS cholinergic neurons and AD’). It is well documented that the nucleus basalis innervates all areas of the neocortex in a topographicspecific manner, that is, more rostral regions of the nucleus basalis will supply cholinergic fibers to the most rostral cortical regions, and the most caudal to the most caudal. While the ‘basalocortical cholinergic pathway’ supplies fibers to the entire neocortex, the cingulate and entorhinal cortex most probably is innervated by cholinergic septal and diagonal band neurons. Figure 4 illustrates the appearance and relative density of cortical cholinergic presynaptic boutons as compared to the overall density of presynaptic boutons as revealed by synaptophysin. The other prominent basal forebrain cholinergic nuclear group is located in two arching bands of neurons meeting at the midline at dorsal aspects of the septal nucleus. This group of neurons is part of a continuum of cholinergic neurons from more rostrally located cholinergic neurons of the diagonal band of Broca. These septal cholinergic neurons project heavily to all regions of the hippocampal formation via the fimbria-fornix; this is referred to as the ‘septal–hippocampal’ cholinergic pathway.

490 Cholinergic Pathways in CNS

a

b

Figure 4 Immunocytochemical demonstration of cholinergic presynaptic boutons in the cerebral cortex of the rat (parietal cortex, lamina (b) as revealed with anti-VAChT antibodies. The pattern and density of the total population of cortical presynaptic boutons, as revealed with antisynaptophysin antibodies, are illustrated in (a). Scale bar ¼ 20 mm.

These two cholinergic nuclei (basalis and medial septum) are frequently referred to in the literature as ‘the’ basal forebrain cholinergic neurons; however, the basal forebrain cholinergic system includes other nuclei forming a continuum in that region of the brain (as previously described). These two cholinergic neuronal groups (basolocortical and septo-hippocampal) are heavily involved in higher CNS functions such as attention, learning, and memory, and are compromised in aging and in AD (see the section entitled ‘CNS cholinergic neurons and AD’). The largest CNS concentration of cholinergic fibers occurs, without any doubt, in the nucleus caudatus in humans and in the caudoputamen complex in rodents. These fibers and nerve terminals originate in very large local circuit neurons distributed throughout this region of the basal ganglia. Indeed, the highest CNS concentration of ACh and cholinergic markers in general occurs in the basal ganglia. The cholinergic system here plays an important role in mechanisms implicated in movement control and there is good physiological and pharmacological evidence for a dopaminergic/ cholinergic balance in this region. Other less conspicuous cholinergic cell groups are found in the thalamus (midline, intralaminar), epithalamus (medial habenula), and hypothalamus (posterior region and arcuate nucleus), with less defined functions than those of the forebrain cholinergic neurons. The medial habenula nuclei project heavily to the nucleus interpeduncularis via the so-called fasciculus retroflexus. In the brain stem, the motor and autonomic preganglionar neurons of cranial nerves are all of cholinergic nature, as expected from classical physiological studies preceding the advent of immunocytochemistry. In addition to these motor neurons, there is a collection of ChAT-IR neurons distributed in various regions of the brain stem. Thus, there is a diffuse reticular system (lateral reticular nucleus, parvicellular) and several nuclei,

including the pedunculo-pontine tegmental (PPT), lateral dorsotegmental (LDT), parabrachial, nucleus trapezoid, superior olive, and the raphe system (nucleus magnus and obscurus). PPT and LDT assembly of cholinergic neurons projects mainly to the thalamus, epithalamus, and tectum, but also sends axons more ventrally to the basal forebrain, lateral hypothalamus, and substantia nigra. Descending branches from the PPT and LDT cholinergic nuclei reach diverse nuclei in the lower brain stem, and long projections from these neurons can reach the spinal cord. In the spinal cord, occasional small cholinergic neurons can be observed in the dorsal horn in addition to the preganglionar nuclei of the lateral horn and the large motor neurons of the ventral horn. Although a number of cholinergic neuronal pathways have been described, there is still an incomplete account of the physiological roles for the diverse CNS cholinergic projections. The PPT–LDT ascending cholinergic pathways, along with forebrain neurons, appear to play an important role in sleep control mechanisms, whereas there is scant information regarding the possible functions of the brain stem– spinal cord descending cholinergic projections. The latter projections are suspected to participate in paincontrol mechanisms.

The Functions and Neurobiology of Basal Forebrain Cholinergic Neurons For several decades, the CNS cholinergic system was suspected to play an important role in memory. The most influential early studies pointing out such CNS cholinergic functions were carried out by Drachman and collaborators in young human subjects in which the muscarinic antagonist scopolamine, in low doses, brought about memory deficits similar to those observed in aged individuals. On the other hand, the application of the muscarinic agonists in young

Cholinergic Pathways in CNS 491

volunteers improved the recall of learned verbal material. These observations on the participation of the cholinergic system in memory tasks were corroborated in subhuman primates by Bartus and collaborators, who elaborated ‘‘the cholinergic hypothesis of geriatry memory dysfunction.’’ Numerous further investigations in rodents supported the concept of a cholinergic involvement in memory and learning. These functions, attributed initially to the so-called ascending reticular pathway, elaborated in the 1940s by Moruzzi and Magoun, had an important impact on developing physiological–anatomical theories for higher CNS functions. This anatomically ill-defined system was described as responsible for the control of wakefulness, sleep patterns, alertness, and memory. The anatomical–neurochemical identification in the brain stem and basal forebrain of neuronal cell groups and ascending pathways of catecholaminergic, serotoninergic, and cholinergic nature in the 1970s and 1980s allowed a more precise and specific understanding of the contribution of these transmitter systems to these functions. Today, the basal forebrain cholinergic system can be considered to be the rostralmost representation of such a system. There is solid experimental evidence that the basal forebrain cholinergic neurons do participate importantly in attentional, learning, and memory functions. These investigations constitute a very rich assembly of pharmacological, anatomical, and selective lesion studies combined with a wide variety of behavioral tasks. In more recent years, some emphasis has been placed on the attentional functions of cortical cholinergic inputs, thus conditioning learning and memory functions. Nevertheless, a dramatic argument in favor of the idea that ACh is necessary for learning and memory has been provided by the restoration of these functions with the cortical grafting of genetically modified cells producing ACh in rats with learning and memory impairments as a consequence of nucleus basalis lesions. The forebrain cholinergic input to cortex and hippocampus might have important modulatory roles in defining activity-dependent synaptic maps. The most provoking evidence in that direction was provided by Kilgard and Merzenich in 1998. This study demonstrated that episodic electrical stimulation of the nucleus basalis, paired with an auditory stimulus, resulted in a progressive reorganization of the primary auditory cortex in the adult rat, similar to that with behavioral training. While these findings suggest that the basal forebrain plays a role in modifying cortical synaptic sensory representation, additional pharmacological data would be necessary to firmly relate this to a cholinergic function.

Trophic Factor Dependency of Forebrain Cholinergic Neurons The embryonic development of the basal forebrain cholinergic system is highly dependent on the expression of nerve growth factor (NGF) and the highaffinity NGF receptor TrkA and, to a lesser extent, on the expression of low-affinity p75LNTR receptors. The low- and high-affinity p75LNTR neurotrophin receptors appear to function, depending on the physiological and/or pathological circumstances, in a cooperative or competitive manner. In early postnatal stages, relatively high levels of NGF are expressed, but this expression substantially decreases after birth. In early ontogenic stages, forebrain cholinergic neurons remain highly sensitive to exogenously administered NGF. The expression and release of NGF are very low after neuronal differentiation in adulthood. However, the NGF dependency of cholinergic neurons in mature and fully differentiated CNS remains. The experimental evidence indicates that in the adult CNS, NGF maintains cholinergic neuronal phenotype. The main supply of NGF is thought to be ‘target derived,’ that is, from the cerebral cortex for cholinergic nucleus basalis neurons, and from the hippocampus for medial septum neurons. Thus, target ablation of the neocortex or hippocampus provokes cholinergic neuronal cell shrinkage but not death. However, transection of the fimbria-fornix can provoke a mixed situation of neuronal atrophy and cell death of cholinergic neurons in the medial septum. The application of exogenous NGF can restore the cholinergic phenotype in both lesion models. Interestingly, NGF application after limited cortical strokelike lesions is sufficient to restore the size and biochemistry of nucleus basalis cholinergic neurons as well as preventing behavioral memory deficits, even when applying NGF after lesions (see Figure 5). Exogenous application of NGF in cortically lesioned adult rats also induces cholinergic synaptogenesis in the remaining, intact neocortex. On the other hand, the blocking of endogenously produced NGF by applying NGF-immunoneutralizing monoclonal antibodies in the cerebral cortex or synthetic TrkA receptor antagonists was shown to remove preexisting cortical cholinergic synapses in vivo. In other words, these observations would indicate that the production of small, baseline endogenous NGF in the cerebral cortex regulates the steady-state number of cholinergic synaptic boutons in the adult CNS. As trophic factors have been shown to be produced and liberated in an activity-dependent fashion, it would be reasonable to assume that the number of cortical

492 Cholinergic Pathways in CNS

cs c o o

a

d

b

e

c

f

n

c

sm CPu

o o o

RF

o o

o

o

Rt

o

sm

31

o oo

Pir ACo LOT

SCh

Io

Figure 5 The schematic drawing on the left represents the distribution of abundant forebrain neurons intensively immunoreactive to p75 low-affinity NGF receptor, represented as filled circles. The cells in the inner portion of the globus pallidus correspond to cholinergic neurons of the nucleus basalis. Open circles indicate moderately p75 immunoreactive large neurons in the caudate putamen (cholinergic interneurons). The micrographs on the right (a) indicate cholinergic neurons of the nucleus basalis, as revealed by ChAT immunocytochemistry. Note the rich dendritic branching of the cholinergic neurons of the nucleus basalis. The extreme right column illustrates the computer-assisted definition of the cell soma cross-sectional area. The micrographs illustrate the efficacy of NGF to rescue retrogradely shrunken neurons following cortical lesions: (a) and (d) show ChAT-IR nucleus basalis neurons from naive control adult rats; (b) and (e) show gross cell shrinkage of the same neurons after cortical lesions; and (c) and (f) show prevention of cell shrinkage and maintenance of normal phenotype with the administration of microgram amounts of NGF. ACo, anterior cortical olfactory nucleus; c, corpus callosum; CPu, caudate putamen; cs, cingulate striae; lo, lateral olfactory tract; LOT, nucleus of the olfactory tract; n, fimbria fornix; Pir, piriform cortex; RF, rhinal fissure; Rt, reticular nucleus of the thalamus; SCh, nucleus suprachiasmaticus; sm, stria medullaris; 3, third ventricle. Reproduced from Cuello AC (2006) Cholinergic synaptic terminations in the cerebral cortex, trophic factor dependency, and vulnerability to aging and Alzheimer’s pathology. In: Giacobini E and Penney JB (eds.) The Brain Cholinergic System in Health and Disease, figure 3.2, pp. 33–46. Oxon, UK: Informa Healthcare, with permission from Taylor & Francis.

cholinergic synapses changes constantly in the human brain, based on experience and brain activity. Such a concept would be in line with the classical tenet of Hebb that the strength of synaptic connections was linked to a growth process which takes place with synaptic efficacy.

CNS Cholinergic Neurons and AD It is possible that CNS cholinergic neurons are somehow involved in a variety of neurodegenerative disorders. However, up to the present, none of these conditions appears as striking as in the case of

Alzheimer’s disease. In the 1970s, Davis and Maloney and Bowen et al. reported evidence of marked deficits in cortical cholinergic markers (particularly loss of ChAT levels) in AD brain. After the identification of the main CNS cholinergic groups, Whitehouse and collaborators reported losses of ‘magnocellular’ neurons in the the nucleus basalis of Meynert. These findings were coincidental with ‘‘the cholinergic hypothesis of geriatry memory dysfunction’’ (as discussed previously). At the time, evidence that the loss of substantia nigra dopaminergic neurons was mainly responsible for Parkinsonian symptoms, along with the evidence that the administration of dopamine

Cholinergic Pathways in CNS 493

precursors could be an effective transmitter-based therapy, brought about by analogy ‘the cholinergic hypothesis of Alzheimer’s disease.’ This proposal provoked a great deal of experimental and clinical research. We now know that the cholinergic involvement in AD is a component of the neuropathology, most likely secondary to the Ab burden in the cerebral cortex and hippocampus. It is now also well established that the cholinergic depletion occurs at advanced stages of the AD pathology, while at the AD-prodromic stage of mild cognitive impairment (MCI) an upregulation of cholinergic markers takes place instead. Furthermore, transgenic animal models of the amyloid pathology can replicate features of the cholinergic involvement, suggesting a secondary rather than a primary involvement in the disease process. However, as discussed previously, the forebrain cholinergic system is so intimately related to memory processes that it was perceived as a viable therapeutic target. Early attempts at using muscarinic agonists and choline supplements – in the latter case to stimulate ACh synthesis, given (as discussed previously) that the high-affinity choline uptake system is the de facto rate-limiting factor – were without substantive effect on memory or other cognitive parameters. Instead, the application of cholinesterase inhibitors (capable of blocking acetylcholinesterases or butyrilcholinesterases in diverse degrees) to block ACh degradation has been used widely as symptomatic therapy in AD. This approach is by no means capable of halting the pathological process (disease modifying), but has proved effective in delaying cognitive decay in a good number of patients, although the efficacy wanes after a year or two of treatment. The cholinergic system is, however, involved in the disease process. The activation of M1 and M3 receptors has been shown to shift amyloid precursor protein (APP) metabolism to a nonamyloidogenic modality, an aspect that has been confirmed repeatedly in experimental models and in the clinic with the application of anticholinesterases. These observations have provided renewed interest in the development of newer, safer, and more efficaceous muscarinic agonists. Knowing the great trophic dependency of forebrain cholinergic neurons on NGF support, clinical attempts have involved applying mouse NGF in the brain of Alzheimer’s patients, but with disapponting results. In these patients, the large doses applied and the diffusion of NGF to undesirable targets (e.g., sensory nociceptive fibers and hypothalamic nuclei) produced pain and marked weight loss. As NGF can act on the nucleus basalis also in the somatodendritic region in a paracrine fashion (in contrast with target derived), there is current interest in applying grafts of genetically modified NGF-producing cells, implanted

in nucleus basalis neighboring areas of Alzheimer’s patients, and, long range, of regulable lentoviral vectors, allowing the in situ expression of NGF. New therapeutic approaches to correct the cholinergic deficit might emerge from the realization that this is most probably due to the dysregulation of the protease cascade, which is responsible for the maturation of the NGF-precursor protein (ProNGF) into mature and biologically active NGF and for its degradation. See also: Cholinergic Neurotransmission in the Autonomic and Somatic Motor Nervous System; Nicotinic Acetylcholine Receptors.

Further Reading Bartus RT, Dean RL, Beer B, et al. (1982) The cholinergic hypothesis of geriatric memory dysfunction. Science 217: 408–417. Butcher LL and Woolf NJ (2004) Cholinergic neurons and networks revisited. In: Paxinos G (ed.) The Rat Nervous System, pp. 1257–1268. London: Elsevier. Cuello AC (1996) Effects of trophic factors on the CNS cholinergic phenotype. Progress in Brain Research 109: 347–358. Cuello AC (2006) Cholinergic synaptic terminations in the cerebral cortex, trophic factor dependency, and vulnerability to aging and Alzheimer’s pathology. In: Giacobini E and Penney JB (eds.) The Brain Cholinergic System in Health and Disease, pp. 33–46. Oxon, UK: Informa Healthcare. Cuello AC and Sofroniew MV (1984) The anatomy of the CNS cholinergic neurons. Trends in Neurosciences 7: 74–78. Debeir T, Saragovi HU, and Cuello AC (1999) A nerve growth factor mimetic TrkA antagonist causes withdrawal of cortical cholinergic boutons in the adult rat. Proceedings of the National Academy of Sciences of the United States of America 96(7): 4067–4072. Kilgard MP and Merzenich MM (1998) Cortical map reorganization enabled by nucleus basalis activity. Science 279(5357): 1714–1718. Mesulam MM (1990) Human brain cholinergic pathways. Progress in Brain Research 84: 231–241. Mesulam MM, Mufson EJ, Wainer BH, et al. (1983) Central cholinergic pathways in the rat: An overview based on an alternative nomenclature (Ch1–Ch6). Neuroscience 10(4): 1185–1201. Ribeiro FM, Black SA, Prado VF, et al. (2006) The ‘ins’ and ‘outs’ of the high-affinity choline transporter CHT1. Journal of Neurochemistry 97(1): 1–12. Sofroniew MV, Eckenstein F, Thoenen H, et al. (1982) Topography of choline acetyltransferase containing neurons in the forebrain of the rat. Neuroscience Letters 33: 7–12. Sofroniew MV, Galletly NP, Isacson O, et al. (1990) Survival of adult basal forebrain cholinergic neurons after loss of target neurons. Science 247: 338–342. Turrini P, Casu MA, Wong TP, et al. (2001) Cholinergic nerve terminals establish classical synapses in the rat cerebral cortex: Synaptic pattern and age-related atrophy. Neuroscience 105(2): 277–285. Whitehouse PJ, Price DL, Struble RG, et al. (1982) Alzheimer’s disease and senile dementia: Loss of neurons in the basal forebrain. Science 215: 1237–1239. Winkler J, Suhr ST, Gage FH, et al. (1995) Essential role of neocortical acetylcholine in spatial memory. Nature 375(6531): 484–487.

Muscarinic Receptors: Autonomic Neurons R S Aronstam and P Patil, University of Missouri – Rolla, Rolla, MO, USA ã 2009 Elsevier Ltd. All rights reserved.

Introduction Muscarinic receptors recognize the neurotransmitter acetylcholine, translating this recognition into electrical transients and altered cell behavior by activating and suppressing an assortment of signaling pathways. Muscarinic receptors comprise one of the two classes of receptors for the neurotransmitter acetylcholine, with nicotinic receptors comprising the other class. Muscarinic receptors are selectively activated by the alkaloid muscarine from the mushroom Amanita muscaria and are blocked by belladonna alkaloids, such as atropine and scopolamine (Figure 1). Muscarinic receptors are involved in the transduction of cholinergic signals in the central nervous system, autonomic ganglia, smooth muscles, and other parasympathetic end organs. The history of muscarinic systems is intimately associated with the development of receptor theory, pharmacology, and the discovery of neurotransmitter transmission. Muscarinic receptors are members of the superfamily of G-protein-coupled receptors, specifically class A (rhodopsin-like) receptors. Muscarinic receptors are related to the ionotropic nicotinic acetylcholine receptors only insofar as their physiological activator is acetylcholine; muscarinic and nicotinic receptors share little similarity in their structure, physiological functions, or pharmacology (except for a few close analogues of acetylcholine).

Muscarinic Receptor Subtypes The genes for five subtypes of muscarinic receptors, M1–M5, were identified, cloned, and sequenced between 1986 and 1990. These receptor subtypes differ in their primary structure, distribution, pharmacology, and signal transduction activity (Table 1). This heterogeneity presents the possibility of selectively affecting specific muscarinic functions in the brain and other organs; accordingly, the pharmacology of muscarinic receptor subtypes has been the subject of intensive investigation. Unfortunately, the selectivity of antagonists for receptor subtypes rarely exceeds tenfold. In the absence of pharmacological uniqueness, knockout mice are proving exceptionally useful in delineating functions associated with specific receptor subtypes.

494

Our understanding of muscarinic receptor biology is largely dependent on (1) physiological and behavioral analysis of the actions of muscarinic agonists and antagonists (notably, smooth muscle contraction, cardiac function, vascular tone, glandular secretion, arousal, attention, and memory); (2) ligand binding studies employing radiolabeled probes (notably, the antagonists [3H]N-methylscopolamine and [3H]quinuclidinyl benzilate); (3) biochemical characterization of receptor proteins (including primary, secondary, and tertiary protein structure, and patterns of phosphorylation, glycosylation, phosphorylation, and lipid modification, dimerization, nitrosylation, and internalization); (4) measurement of receptor influence on intracellular second messengers (notably, inositol triphosphate (IP3), diacylglycerol, Ca2þ, nitric oxide, cAMP, cGMP, and arachidonic acid); (5) analysis of receptor gene sequence and control; (6) genetic manipulation of receptor expression; (7) genetic manipulation of receptor protein structure; and (8) elimination of specific receptor subtypes using gene knockout technology. The properties of the different subtypes are summarized in Table 1. A schematic diagram of the human M2 receptor is presented in Figure 2, a possible arrangement of the seven transmembrane domains is schematically represented in Figure 3, the amino acid sequence alignment of the five subtypes is presented in Figure 4, and an alignment tree of receptor amino acid sequences suggesting evolutionary relations is shown in Figure 5. The receptors were named on the basis of the order of their discovery, which reflects their relative abundance. Muscarinic receptors have been traditionally divided into two groups based on both their pharmacology and signaling properties: M2 and M4 comprise one group, and M1, M3, and M5 comprise the other (Figure 6).

Muscarinic Receptor Structure Muscarinic receptor proteins are single polypeptides of 460–590 amino acids with an extracellular N-terminus and an intracellular C-terminus (Figures 2 and 3). The N-termini do not possess consensus sites for protein cleavage, and epitope tags applied to the N-terminus provide convenient markers to monitor receptor expression and transport. The coding regions of muscarinic receptor genes are contained within a single exon. Hydropathic analyses of the amino acid sequences reveal seven regions of 20–24 amino acids that are likely to form membrane spanning a-helical structures. The amino acid composition of the membrane-spanning regions is

Muscarinic Receptors: Autonomic Neurons 495 Muscarinic ligands H O

CH3

HO

CH3

H2 C

H3C C

H3C

OCH2CH2

N

CH3

CH3 O

H

Acetylcholine endogenous agonist

+ N

H

CH3

CH3 Muscarine prototypical nonselective agonist

CH3 N + OCH2CH2N(C2H5)3 + OCH2CH2N(C2H5)3

O

O

CH2OH

C

CH

+ OCH2CH2NH(C2H5)3

Atropine nonselective antagonist

Gallamine allosteric ligand

O H

O N

CH C

C

CH3 O

+

N CH3

4-DAMP M3 selective antagonist

N O

H N N

O N

C

N AF-DX 116 M2 selective antagonist

Pirenzepine M1 selective antagonist

O N

N CH3 Figure 1 Structures of muscarinic ligands. Acetylcholine is the physiological agonist. Muscarine and atropine are the prototypical agonist and antagonist which define the receptor class. Gallamine is an allosteric receptor antagonist. Pirenzepine, AF-DX 116, and 4-DAMP are antagonists with a degree of selectivity for the M1, M2, and M3 receptor subtypes, respectively.

highly conserved (90% sequence similarity) among the five subtypes, as it is among the larger family of G-protein-coupled receptors. Between the fifth and sixth membrane-spanning units is a large intracellular loop that is highly variable in composition and size. Several consensus sites for phosphorylation are located on the third intracellular

loop, as well as on the C-terminal chain. A disulfide bridge is formed between a conserved cysteine adjacent to the third transmembrane segment and a cysteine in the middle of the second extracellular loop in all five subtypes. Chemical modification of these cysteine groups decreases agonist binding affinity as well as the ability of the receptor to couple to transducer

Property

Molecular weighta Amino acids Genomic location G-protein coupling Second messengersb Tissuesc

Brain regions

Subtypeselective antagonistsd Functions revealed by analysis of knockout mice

a

Subtype M1

M2

M3

M4

M5

51 240

51 715

66 127

53 058

60 186

460 11q13

466 7q31–q35

590 1q43

479 11p12–p11.2

532 15q26

Gq, G11

Gi, Go

Gq, G11

Gi, Go

Gq, G11

IP3/DAG, NO

cAMP (#)

IP3/DAG, NO

cAMP (#)

IP3/DAG, NO

Brain, autonomic ganglia, secretory glands, vas deferens, sympathetic ganglia Cerebral cortex, hippocampus and dentate gyrus, striatum, olfactory bulb and tubercle, amygdala Pirenzepine, telenzepine, mamba venom toxins

Brain, heart, sympathetic ganglia, lung, ileum, uterus and other smooth muscles Cerebellum, medulla, pons, basal forebrain, olfactory bulb, diencephalon

Brain, secretory glands, smooth muscles

Brain, lung

Cerebral cortex, thalamus, piriform cortex, olfactory bulb, brain stem nuclei

Midbrain

AF-DX-116, methoctramine, gallamine, himbacine

4-DAMP, hexahydro-sila-difenidol

Occipital cortex, caudate and putamen, visual nuclei, olfactory tubercle, hippocampus Tropicamide, himbacine

Agonist-induced seizures, learning, control of locomotor activity; indirect inhibition of N- and L-type calcium channels; M current inhibition in sympathetic ganglion neurons; MAPK pathway activation

Agonist-induced akinesia and central tremor; corticosterone release; analgesia; hypothermia; smooth muscle contraction; bradycardia; inhibition of N- and P/Q-type calcium channels

Papillary and urinary bladder constriction; salivary secretion; smooth muscle contraction; control of dopamine release; maintenance of body mass and food intake

Dopamine release; acetylcholine release (hippocampus); depression of motor control; analgesia

Dilation of cerebral arteries and arterioles; dopamine release

Molecular weights are calculated for the polypeptide contribution only. IP3/DAG, stimulation of phospholipase C to release the second messenger inositol triphosphate (IP3) and diacylglycerol (DAG) from phosphatidylinositol; NO, increase in nitric oxide production. cAMP(#): inhibition of adenylate cyclase, thereby lowering cellular content of the second messenger cAMP. c Only low levels of M5 receptor expression have been detected in tissues. d Antagonist selectivity of the subtypes is relative; antagonists generally differ in their interactions with the different subtypes by a factor of less than 20. b

496 Muscarinic Receptors: Autonomic Neurons

Table 1 Properties of muscarinic receptor subtypes

A

L

N

S

L

S

S

N

N

S

T

N

N

M

G

E C

V

Y

G

T S

D

E

V

T

R

P

Y K

T

F

F V

A

I

L

Y

T

S

M

S I

L

I

S

V

D

N

L

M

F

L

L A L

L V F

I C F

N R

L

Q

T

L

Y

D

I

V

S F

N

S

I

I

D R

N

Y

A

M T

G W

K

K

F C V

T

K

P

Y

T

F

I

L

I

L

R

E

S

A T

V

F

K

K

K

K

A

P Q K

T M

K

I

V

T S

V

G

S

S

G

Q

N

G

E

V

T

T

N

T

P

T

C

S

D S

L

G

H

S

K

D

E N

S

K

Q

T

D

Q

T

I

E

D

D

R

E

K

E

S

S

N

D

E

S V

T N

E

T

E N

C

V Q

G

V

P

D

R

P

A

K

G

N

Q

I

K N

D

P

V

S

P

S

L

V

Q

G

R

Q

K

H L

R

M

K

Q

N

I

A

V

K

T K

P

C

C

E

H

T

R

I

M N

S

A

V

D S

T

L G

S D

H

K

G

I

E

Y N I

V D

G A

A

T

R

S

S

A N

F

K

K P

T

K

S P

I

V

K

P

N

N

N

N

M

Figure 2 Schematic diagram of a human M2 muscarinic receptor showing the primary amino acid sequence. Four consensus sites of glycosylation at asparagine residues are indicated on the extracellular N-terminal domain. The aspartate and tyrosine and threonine residues in the membrane-spanning regions that are thought to be involved in agonist binding are highlighted in yellow. Regions of the third intracellular loop involved in G-protein recognition and coupling are indicated by blue shading. Possible sites of threonine phosphorylation in the C-terminal domain are also indicated by highlighting.

Muscarinic Receptors: Autonomic Neurons 497

V

K

E

A

L

C

V

I

D

P

A

C

Y

L P

R

K

T

I

N

K

K

Y

I

N

S

N

K

K

W

L

W

A

T

I

Y G C

T

I

T

V

W

P

S

Y

T

L

S

P

A

P

P

T

R

I N

R

S V

P

V

I

R

A

A

H

T

M

L

L

Y

V

I

M

T

N

I

L

P

V

I

A

I

M

M

W

A

I

A

F

Y

L

T

A

I

A

F

S

L

V A

L

L

V S

W

V T

F

G

L

I S

I

C

N

N

A

A

V A

P

A

S

W

F

P

T

F

Q

A

C

F

F F

L

V

M A

N

V H

A

N

A

Y

K V

D

L

N V

G

L

W Y

Q

I

V

C

G

I

V

G

I

P

Y T

L

V G

V

V V

S

L T

V

I S-S

V

G

W E

I

L

P

498 Muscarinic Receptors: Autonomic Neurons

IV

II

V I

III VI

VII

Figure 3 Possible arrangement of muscarinic receptor transmembrane domains, as suggested by structural analysis of rhodopsin. Amino acid moieties involved in the binding of acetylcholine are located approximately one-fourth of the way into the plane of the membrane.

G-proteins. Moreover, nitrosylation of receptor cysteines after reduction of the disulfide bond severely diminishes ligand binding to the receptor. Elimination of either of these cysteines, however, precludes expression of the receptor. A conserved cysteine in the N-terminus may be a site for palmitoylation (thereby creating another intracellular loop) and an N-terminal polybasic region in four of the subtypes (all except the M2 receptor) may be involved in recruitment of adapter proteins in certain signaling pathways. M3 receptors (at least) have been found to form disulfide-linked dimers on the cell surface. Two to four consensus sites for N-glycosylation are present in the extracellular N-terminal domain of each receptor, and 25% of the receptor mass may be contributed by these oligosaccharides. Although this carbohydrate component may affect the processing and orientation of the receptor, it plays no discernible role in ligand binding or signal transduction: enzymatic removal of the carbohydrate has little effect on the binding and signal transduction potential of mature receptors. Chimeric receptors and mutational analysis have revealed sites on the receptor proteins that are specifically involved in ligand binding and coupling to transducer G-proteins. Acetylcholine binds to a site within a pocket formed by the roughly circularly arranged transmembrane domains (Figure 3). An aspartate moiety in the third transmembrane

region participates in an ionic interaction with the quaternary nitrogen of acetylcholine, whereas a series of tyrosine and threonine residues located in the membrane-spanning segments approximately onefourth of the distance through the membrane likely form hydrogen bonds with muscarinic ligands. In agreement with classical pharmacological analysis, the binding site for competitive antagonists is thought to overlap the acetylcholine recognition site but additionally involves contiguous hydrophobic areas of the receptor protein and membrane. Muscarinic receptors also possess a site(s) or region that participates in allosteric regulation by a variety of compounds, including gallamine (Figure 1). These ligands compete with classical muscarinic ligands but slow the dissociation of previously bound probes, indicating the existence of a ternary complex (receptor plus allosteric and orthosteric ligands). Gallamine binding is sensitive to modification of the sixth transmembrane domain as well as the third extracellular loop. A number of regions have been identified that participate in the interactions of muscarinic receptors with transducer G-proteins. In particular, the second intracellular loop and the N- and C-terminal portions of the third intracellular loop are involved in coupling to G-proteins. Desensitization of muscarinic receptors may entail phosphorylation of conserved threonine residues on the C-terminal portion of the receptor, as well as other sites on the third intracellular loop. Prolonged stimulation of muscarinic receptors induces a rapid phosphorylation that is associated with a decrease in agonist affinity and an uncoupling from transducer G-proteins. Agonist-induced phosphorylation involves a G-protein receptor kinase (GIRK) and not kinases activated by Ca2þ, cAMP, cGMP, or phorbol esters. Phosphorylation-dependent internalization of muscarinic receptors appears to be mediated by arrestins. The evolutionary relationship of muscarinic receptor subtypes is suggested by amino acid sequence (Figure 5). M1, M3, and M5 comprise one cluster, with M1 and M3 being somewhat more closely related. M2 and M4 comprise a second cluster. Analysis of nucleotide sequence of the coding regions reveals an essentially similar pattern. Sequence analyses reveal more genomic divergence within the two clusters than between the two groups. The five muscarinic receptor genes are scattered throughout the genome; the only two on the same chromosome (M1 and M4 on chromosome 11) are only distantly related.

Muscarinic Receptor Gene Structure Whereas the coding sequences of the gene for each subtype are uninterrupted or intronless, the genes are

Muscarinic Receptors: Autonomic Neurons 499

M1 M2 M3 M4 M5

10

20

30

40

50 60 70 80 MNTSAPPAVSPNITVLAPGKGPWQVAFIGITTGLLSL MNNSTNSSNNSLALTSPYKTFEVVFIVLVAGSLSL MTLHNNSTTSPLFPNISSSWIHSPSDAGLPPGTVTHFGSYNVSRAAGNFSSPDGTTDDPLGGHTVWQVVFIAFLTGILAL MANFTPVNGSSGNQSVRLVTSSSHNRYETVEMVFIATVTGSLSL MEGDSYHNATTVNGTPVNHQPLERHRLWEVITIAAVTAVVSL .. * . ..*

M1 M2 M3 M4 M5

90 100 110 120 130 140 150 160 ATVTGNLLVLISFKVNTELKTVNNYFLLSLACADLIIGTFSMNLYTTYLLMGHWALGTLACDLWLALDYVASNASVMNLL VTIIGNILVMVSIKVNRHLQTVNNYFLFSLACADLIIGVFSMNLYTLYTVIGYWPLGPVVCDLWLALDYVVSNASVMNLL VTIIGNILVIVSFKVNKQLKTVNNYFLLSLACADLIIGVISMNLFTTYIIMNRWALGNLACDLWLAIDYVASNASVMNLL VTVVGNILVMLSIKVNRQLQTVNNYFLFSLACADLIIGAFSMNLYTVYIIKGYWPLGAVVCDLWLALDYVVSNASVMNLL ITIVGNVLVMISFKVNSQLKTVNNYYLLSLACADLIIGIFSMNLYTTYILMGRWALGSLACDLWLALDYVASNASVMNLL *. **.**..* *** *.*****.* ********** **** .* * . . * ** . ******.*** *********

M1 M2 M3 M4 M5

170 180 190 200 210 220 230 240 LISFDRYFSVTRPLSYRAKRTPRRAALMIGLAWLVSFVLWAPAILFWQYLVGERTVLAGQCYIQFLSQPIITFGTAMAAF IISFDRYFCVTKPLTYPVKRTTKMAGMMIAAAWVLSFILWAPAILFWQFIVGVRTVEDGECYIQFFSNAAVTFGTAIAAF VISFDRYFSITRPLTYRAKRTTKRAGVMIGLAWVISFVLWAPAILFWQYFVGKRTVPPGECFIQFLSEPTITFGTAIAAF IISFDRYFCVTKPLTYPARRTTKMAGLMIAAAWVLSFVLWAPAILFWQFVVGKRTVPDNQCFIQFLSNPAVTFGTAIAAF VISFDRYFSITRPLTYRAKRTPKRAGIMIGLAWLISFILWAPAILCWQYLVGKRTVPLDECQIQFLSEPTITFGTAIAAF .*******..*.**.* .** . * .** **..**.******* **. ** *** .* *** * .*****.***

M1 M2 M3 M4 M5

250 260 270 280 290 300 310 320 YLPVTVMCTLYWRIYRETENRARELAALQ--GSETPGKGGGS----------------SSSS--ERSQPGAE-------YLPVIIMTVLYWHISRASKSRIKKDKKEPVANQDPVSPSLVQG---------------RIVKPNNNNMPSSD-------YMPVTIMTILYWRIYKETEKRTKELAGLQASGTEAETENFVH----------------PTGS--SRSCSSYE-------YLPVVIMTVLYIHISLASRSRVHKHRPEGPKEKKAKTLAFLKS---------------PLMKQSVKKPPPGE-------YIPVSVMTILYCRIYRETEKRTKDLADLQ--GSDSVTKAEKRKPAHRALFRSCLRCPRPTLAQRERNQASWSSSRRSTST *.** .*. ** .* . * .

M1 M2 M3 M4 M5

330 340 350 360 370 380 390 400 GSPETPP-GRCCRCCRAPRLLQAYSWK---EEEEEDEGS------------MESLTSSEG-EEPGSE------VVIKMPDGLEHNKIQNG-KAPRDPVTENCVQGE---EKESSNDSTS-----------VSAVASNMRDDEITQD------------LQQQSMKRSNRRKYGRCHFWFTTKSWKPSSEQMDQDHSSSDSWNNNDAAASLENSASSDE-EDIGSETRAIYSIVLKLPG AAREELR--NG-KLEEAPPPALPPPPR---PVADKDTSNESS----------SGSATQNTKERPATE----------LST TGKPSQATGPSANWAKAEQLTTCSSYP---SSEDEDKPATDP--------VLQVVYKSQGKESPGEE----------FSA . .

M1 M2 M3 M4 M5

410 420 430 440 450 460 470 480 --MVDPEAQAPTK---QPPRSSPNTVKRPTKKGRDRAGK-----GQKPR------------------------------ENTVSTSLGHSKD---ENSKQTCIRIGTKTPKSDSCTPT-----NTTVEVVGSSGQNGD--------------------HSTILNSTKLPSSDNLQVPEEELGMVDLERKADKLQAQKSVDDGGSFPKSFSKLPIQLESAVDTAKTSDVNSSVGKSTAT TEATTPAMPAPPL---QPRALNPASRWSKIQIVTKQTGN---ECVTAIEIVPATPAGMR--------------------EETEETFVKAETE---KSDYDTPNYLLSPAAAHRPKSQK---CVAYKFRLVVKADGNQE---------TNNGCHKVKIMP .

M1 M2 M3 M4 M5

490 500 510 520 530 540 550 560 -----------------GKEQLAKRKTFSLVKEKKAARTLSAILLAFILTWTPYNIMVLVSTFCKDCVPETLWELGYWLC -----EKQNIVARKIVKMTKQPAKKKPPPSR-EKKVTRTILAILLAFIITWAPYNVMVLINTFCAPCIPNTVWTIGYWLC LPLSFKEATLAKRFALKTRSQITKRKRMSLVKEKKAAQTLSAILLAFIITWTPYNIMVLVNTFCDSCIPKTFWNLGYWLC -----PAANVARKFASIARNQVRKKRQMAAR-ERKVTRTIFAILLAFILTWTPYNVMVLVNTFCQSCIPDTVWSIGYWLC CPFPVAKEPSTKGLNPNPSHQMTKRKRVVLVKERKAAQTLSAILLAFIITWTPYNIMVLVSTFCDKCVPVTLWHLGYWLC * *.. *.* ..*. *******.**.***.***. *** *.* * * .*****

M1 M2 M3 M4 M5

570 580 590 600 610 YVNSTINPMCYALCNKAFRDTFRLLLLCRWDKRRWRKIPKRPGSVHRTPSRQC* YINSTINPACYALCNATFKKTFKHLLM* YINSTVNPVCYALCNKTFRTTFKMLLLCQCDKKKRRKQQYQQRQSVIFHKRAPEQAL* YVNSTINPACYALCNATFKKTFRHLLLCQYRNIGTAR* YVNSTVNPICYALCNRTFRKTFKMLLLCRWKKKKVEEKLYWQGNSKLP* *.***.** ****** .*. **. **.

Palmitoylation site

G-Protein binding region ACh binding site Intramolecular disulfide bond Possible site of glycosylation Transmembrane regions in boxes

Figure 4 Amino acid sequence alignment of the five human muscarinic acetylcholine receptor subtypes. Conserved and semiconserved residues are indicated by the asterisks and dots below the sequences; possible N-terminal glycosylation sites are indicated by yellow shading; the conserved cysteines involved in disulfide bond formation are indicated by blue shading; the N-terminal site of palmitoylation is highlighted; sequences believed to be involved in receptor interactions with G-proteins are indicated by red bars; transmembrane regions are identified by the green shading. Alignment was performed using the ClustalW procedure provided in MacVector.

500 Muscarinic Receptors: Autonomic Neurons 0.198

0.071

0.198 0.236

0.033

0.225 0.225

M4 M2 M5 M3 M1

0.05 Figure 5 UPGMA distance tree illustrating the evolutionary relationship of the muscarinic receptor subtypes revealed by amino acid sequence. M1, M3, and M5 comprise a cluster; M2 and M4 comprise a second cluster. Analysis of nucleotide sequence of the coding regions reveals an essentially similar pattern. The 0.05 scale marker is approximately equivalent to a 5% difference in sequence.

Receptor

M2, M4

M1, M3, M5

Gi

Gq/11

Transducer

− Effector

2nd Messenger

Kinase

+ + Adenylate cyclase

ATP

+ PLC-beta

cAMP

IP3

DAG

PKA

Ca2+

PKC

Figure 6 Muscarinic receptor signaling pathways. The initial stages of signal transduction include receptor, transducer (G-protein), effector (typically an enzyme or ion channel), second messenger production (e.g., Ca2þ, cAMP, cGMP, DAG, and IP3), and protein kinase activation. It has long been appreciated that M2 and M4 receptors inhibit adenylate cyclase through activation of Gi, whereas M1, M3, and M5 receptors stimulate phospholipase C-beta through activation of Gq/11. In fact, muscarinic signaling involves an impressive variety of other pathways. For instance, under the right conditions all receptors can stimulate adenylate cyclase through the activation of Gs. Other pathways are activated directly or secondarily. PKA, protein kinase A; PKC, protein kinase C; IP3, inositol triphosphate; DAG, diacylglycerol.

complex, spanning large distances, and often contain multiple exons that result in multiple splice variants that are under the control of tissue-specific promoters. TATA and CAAT boxes are not involved in the transcriptional control of muscarinic receptor gene expression. The promoter structures of several muscarinic acetylcholine receptor subtypes have been analyzed. Consensus regulatory sequences, including binding sites for SP1, AP-1, and AP-2, as well as tissuespecific silencer elements, have been identified. All muscarinic receptor genes contain a large (10–26 kb) intron close to the 50 end of the translational start site. The relatively simple M1 gene contains an upstream 657 bp exon with a flanking 900 bp promoter region

that contains consensus sites for SP1, NZF-1, AP-1, AP-2, E-box, NF-kB, and Oct-1. The M2 gene spans over 126 kb on chromosome 7, and it contains four exons that result in at least eight different splice variants. The M2 gene has at least three distinct promoters that result in distinct tissue-specific expression, and these promoters are highly conserved between mammalian species. The M3 gene spans at least 285 kb encompassing eight exons (the coding sequence is in exon 8). Multiple AP-2 regulatory sequences are located in the 1.2-bp 50 -flanking promoter region, and the transcription factor AP-2a increases M3 expression. Muscarinic gene sequences tend to be highly conserved; relatively few nonsynonymous polymorphisms have been reported.

Physiological Functions Muscarinic receptors subserve a number of physiological functions. Muscarinic receptors are present in autonomic ganglia and on the target organs of postganglionic parasympathetic fibers. Thus, muscarinic receptors mediate such parasympathetic effects as smooth muscle contraction, blood vessel dilation, decreases in heart rate and cardiac contractility, and glandular secretion. Muscarinic agonists have relatively little use in therapeutics, being limited to stimulating muscles of the gastrointestinal tract and bladder and to lowering intraocular pressure. A great deal of work has focused on developing M1 subtype-specific agonists that might be useful in alleviating the cholinergic losses that characterize Alzheimer’s disease. It has been sporadically proposed that subtype-selective cholinomimetics would be useful in the treatment of psychiatric diseases. Arecoline, an alkaloid found in the betel nuts that are chewed by a significant portion of the world’s population, is undoubtedly the most widely consumed muscarinic agonist. Arecoline causes parasympathetic stimulation (notably increased salivation) and produces euphoria and enhanced cognitive function.

Muscarinic Receptors: Autonomic Neurons 501

Muscarinic antagonists are used clinically to inhibit parasympathetic activity in a variety of situations. For example, antimuscarinic drugs have been used in anesthesia (to reduce secretions), parkinsonism, overactive bladder, bradycardia, and irritable bowel syndrome. In the central nervous system, cholinergic fibers arise in nuclei in the midbrain and pons to diffusely innervate muscarinic receptors on neurons in the thalamus and other diencephalic structures, as well as the reticular formation and other brain stem and cranial nerve nuclei. A second major projection system arises from the basal forebrain cholinergic system (which includes the medial sepal nucleus, diagonal band nuclei, and nucleus basalis) and innervates muscarinic receptors on neurons throughout the telencephalon. Cholinergic interneurons innervating muscarinic receptors are located in several brain structures, including the cerebral cortex, nucleus accumbens, and striatum. Muscarinic receptor density is particularly high in the striatum, cerebral cortex, hippocampus, and other specific nuclei. Muscarinic receptor subtypes display striking regional and laminar heterogeneity (Table 1). Central muscarinic systems are prominently involved in arousal, attention, learning, and memory. Other functions involving central muscarinic actions include movement, analgesia, and temperature regulation. Subsets of M2 and M4 receptors have a presynaptic localization and play a role in the control of neurotransmitter release; the other receptors are predominantly postsynaptic. Analysis of knockout mice lacking expression of specific receptor subtypes has shed considerable light on the physiological functions of receptor subtypes (Table 1). Muscarinic knockout mice are both healthy and fertile; recognition of deficits frequently requires specific focused analysis. These studies reveal that M1 receptors are involved in M current (potassium channel) regulation, agonist-induced seizures mitogen-activated protein kinase (MAPK) activation, certain learning paradigms, and inhibition of a slow, voltage-independent calcium current. M2 receptors are involved in the production of bradycardia; smooth muscle contraction (stomach, urinary bladder, and trachea); and central tremor, analgesia, and hypothermia. M3 receptors are seen to be involved in salivary secretion, pupillary constriction, and bladder detrusor contraction. M4 receptors are seen to be involved in limiting locomotor activity (including modulation of motor responses to dopamine). There is an interesting relationship between M5 receptors and dopamine neurotransmission: M5 receptors facilitate dopamine release, thereby affecting brain reward systems. M5 receptors also appear to play a specific role in the dilation of cerebral arteries and arterioles.

Muscarinic Signal Transduction Muscarinic receptors alter the activity of receptive cells by a number of different signal transduction pathways (Figure 6). The biochemical pathways activated depend on the nature and quantity of the receptor subtype, transducer proteins, effector molecules, and protein kinase substrates expressed in the tissue, as well as the potential for cross-talk between the various transduction pathways. Conventional wisdom holds that the odd-numbered receptor subtypes – M1, M3, and M5 – efficiently couple through pertussis toxin-insensitive G-proteinsofthe Gq/11 family to a stimulation of phospholipase C (b subtype). Phospholipase C releases the second messengers diacylglycerol and IP3 from the membrane phospholipid, phosphatidylinositol-4, 5-bisphosphate. Diacylglycerol activates protein kinase C, whereas IP3 releases Ca2þ from intracellular stores by acting on specific receptors on the endoplasmic reticulum. M1, M3, and M5 also activate the MAP kinase pathway as well as phospholipase D (releasing arachidonic acid). Cross-talk between signaling pathways makes the distinction between proximal and indirect signaling events problematic. M2 and M4, inhibit adenylate cyclase through actions involving Gi (any of the three Gi proteins may be involved). A few, but by no means all, isoforms of mammalian adenylate cyclase are inhibited by Gia subunits (e.g., AC5 and AC6). An even larger number of adenylate cyclase subunits are inhibited by Gbg dimers. Thus, the molecular basis for inhibition of the enzyme is not always obvious. Inactivation of Gi by ADP-ribosylation catalyzed by pertussis toxin uncovers activation of adenylate cyclase by all receptor subtypes. In the case of M3 at least, this activation seems to be mediated by a muscarinic receptor–Gs interaction. Thus, our initial binary classification of muscarinic signaling paradigms (i.e., receptor:transducer:effector ¼ M1/M3/M5:Gs:phospholipase Cb or M2/M4: Gi:adenylate cyclase) has been extended in response to the identification of pathways that involve additional transducer proteins as well as an appreciation of secondary effects and cross-talk between the pathways. Other well-documented muscarinic actions include stimulation of adenylate cyclase, inhibition of phosphodiesterase, stimulation of phospholipase A2 (M1, M3, and M5), and stimulation of phospholipase D (M1 and M3). Muscarinic receptor stimulation activates a number of depolarizing and hyperpolarizing currents through both direct and indirect mechanisms. Prominent effects include (1) stimulation of inwardly rectifying potassium conductances by M2 and M4 receptors; (2) activation of calcium-dependent potassium, chloride,

502 Muscarinic Receptors: Autonomic Neurons

and cation conductances by M1, M3, and M5 receptors; (3) inhibition of the M current, a voltage- and time-dependent potassium conductance, by M1 and M3 receptors; and (4) inhibition of a calcium conductance by M2 and M4 receptors, directly via Go and indirectly via reductions in cAMP. Other novel biochemical effects of muscarinic receptor activation have been described. This signaling heterogeneity may reflect (1) multiple isoforms of signal transduction effectors (e.g., at least 11 isoforms of adenylate cyclase), (2) cross-talk between transduction pathways due to a lack of specificity or the presence of indirect and secondary effects, or (3) cell type-specific differences due to expression of different assortments of signal transduction molecules. Muscarinic receptormediated activation of both phosphodiesterase and nitric oxide synthetase has been reported. Muscarinic receptors have substantial mitogenic potency. The molecular details of muscarinic receptors of the pathways that mediate these effects are incompletely understood. The activities of multiple kinases and phosphatases are affected by muscarinic receptors under specific conditions, including Raf, tyrosine kinases, MAPK, and phosphatidyl inositol-3-kinase. The activities of multiple small GTPases, including Ras and Rho, are also affected by muscarinic receptor activation. Muscarinic signaling is subject to regulation by agonist-induced phosphorylation and downregulation (internalization). Mice deficient in G-protein-coupled receptor kinase-5 display muscarinic supersensitivity. Although domains located in the third intracellular loop are prominently involved in receptor phosphorylation and downregulation, a number of subtypedependent variations in arrestin association and internalization pathways have been described. Muscarinic receptors regulate the expression of a variety of genes, notably a variety of transcription factors and signaling factors. An appreciation of this regulation promises to increase our understanding of muscarinic physiology.

Beltran B, Orsi A, Clementi E, and Moncada S (2000) Oxidative stress and S-nitrosylation of proteins in cells. British Journal of Pharmacology 29: 953–960. Birdsall NJ and Lazareno S (2005) Allosterism at muscarinic receptors: Ligands and mechanisms. Mini-Reviews in Medicinal Chemistry 5: 523–543. Brann MR, Klimkowski VJ, and Ellis J (1993) Structure/function relationships of muscarinic acetylcholine receptors. Life Sciences 52: 405–412. Buckley NJ, Bachfischer U, Canut M, et al. (1999) Repression and activation of muscarinic receptor genes. Life Sciences 64: 495–499. Eglen RM (2005) Muscarinic receptor subtype pharmacology and physiology. Progress in Medicinal Chemistry 43: 105–136. Felder CC (1995) Muscarinic acetylcholine receptors: Signal transduction through multiple effectors. FASEB Journal 9: 619–625. Hulme EC, Birdsall NJM, and Buckley NJ (1990) Muscarinic receptor subtypes. Annual Review of Pharmacology and Toxicology 30: 633–673. Ma W, Li BS, Zhang L, and Pant HC (2004) Signaling cascades implicated in muscarinic regulation of proliferation of neural stem and progenitor cells. Drug News and Perspectives 17: 258–266. Nadler LS, Rosoff ML, Hamilton SE, et al. (1999) Molecular analysis of the regulation of muscarinic receptor expression and function. Life Sciences 64: 375–379. Nathanson NM (2000) A multiplicity of muscarinic mechanisms: Enough signaling pathways to take your breath away. Proceedings of the National Academy of Sciences of the United States of America 97: 6245–6247. Saffen D, Mieda M, Okamura M, and Haga T (1999) Control elements of muscarinic receptor gene expression. Life Sciences 64: 479–486. Tobin AB and Budd DC (2003) The anti-apoptotic response of the Gq/11-coupled muscarinic receptor family. Biochemical Society Transactions 31: 1182–1185. von der Kammer H, Demiralay C, Andresen B, et al. (2001) Regulation of gene expression by muscarinic acetylcholine receptors. Biochemical Society Transactions 67: 131–140. Watson S and Arkinstall S (1994) The G-Protein Linked Receptor Facts Book. London: Academic Press. Wess J (1993) Molecular basis of muscarinic acetylcholine receptor function. Trends in Pharmacological Sciences 14: 308–313. Wess J (2004) Muscarinic acetylcholine receptor knockout mice: Novel phenotypes and clinical implications. Annual Review of Pharmacology and Toxicology 44: 423–450.

Relevant Websites See also: Cholinergic Neurotransmission in the

Autonomic and Somatic Motor Nervous System.

Further Reading Ashkenazi A and Peralta EG (1994) Muscarinic acetylcholine receptors. In: Peroutka SJ (ed.) Handbook of Receptors and Channels. G Protein-Coupled Receptors, pp. 1–27. Boca Raton, FL: CRC Press.

http://www.alomone.com – Alomone Labs. http://www.calbiochem.com – Calbiochem. http://www.gpcr.org – GPCRDB: Information system for G-proteincoupled receptors. http://www.scbt.com – Santa Cruz Biotechnology. http://www.sigma-aldrich.com – Sigma–Aldrich. http://stke.sciencemag.org – Signal transduction knowledge environment. http://www.cdna.org – University of Missouri – Rolla cDNA Resource Center.

Nicotinic Acetylcholine Receptors J-P Changeux, Institut Pasteur, Paris, France Y Paas, Bar-Ilan University, Ramat-Gan, Israel ã 2009 Elsevier Ltd. All rights reserved.

Following Claude Bernard’s investigations on the effect of curare, Langley proposed in 1905 that a ‘receptive substance,’ presently designated as a receptor, ‘‘receives the stimulus from the nerve and transmits it to the effector cell.’’ On the basis of the resemblance of acetylcholine (ACh) effects to those of naturally occurring plant alkaloids, Sir Henry Dale distinguished in 1914 between two main classes of receptors for ACh, the muscarinic and the nicotinic receptors. Typically, muscarinic receptors are activated by muscarine and blocked by atropine, whereas nicotinic receptors are activated by nicotine and blocked by curare. Subclasses have also been described. Only nicotinic ACh receptors (nAChRs) are present at the motor endplate of vertebrate skeletal muscles. Muscarinic receptors are found on smooth muscles and gland cells and, together with nicotinic receptors, on autonomic ganglion cells and neurons from the central nervous system.

The nAChR Has an Intrinsic Cationic Channel Binding of ACh to the nAChR causes the opening of a cationic channel that is physically linked to the receptor site (Figure 1). This intrinsic channel is permeable to Naþ, Kþ, and, in many nAChR-subunit combinations, Ca2þ ions. By contrast, muscarinic ACh receptors transduce extracellular signals into the cell by interacting with a G-protein to regulate indirectly effector systems such as adenylate cyclase and ion channels. In 1970, the nAChR became the first neurotransmitter receptor to be isolated and purified and, in 1983, it was the first to be chemically defined by molecular genetic methods. Two factors played a decisive role in this fundamental progress: (1) the presence of very high concentrations of nAChR in the electric organs of the electric fish Electrophorus electricus and Torpedo, and (2) the availability of small polypeptide a-toxins from snake venoms that bind with very high affinity and selectivity to the nAChR from electric organs. The nAChR is an oligomeric glycoprotein of about 300 000 Da. In side views obtained by cryo-electron microscopy, it appears as a transmembrane cylinder of 8 nm diameter and 16 nm length; the long axis being

perpendicular to the membrane plane (Figure 1(a)). In a top view, it looks like a rosette of five subunits organized around a symmetry axis (Figure 1(b)).

Combinatorial Assembly of Subunits Encoded by Homologous Genes Results in a Wide Diversity of nAChRs In the electric organ and muscle, the nAChR oligomer is composed of four different polypeptide chains associated into a pentamer with the a1, g, a1, d, b1 subunits organized in a counterclockwise manner (Figure 1(b)). In nerve cells, combinatorial assembly of up to nine different a subunits (a2 to a10) and three b subunits (b2 to b4) contributes to a wide diversity of pentamers, which may include one, two, or three different types of subunits and possess distinct pharmacological, electrophysiological, and kinetic properties. Each subunit traverses the cell membrane four times and all subunits are in contact with the lipid bilayer. The complete amino acid sequences of the electric organ, muscle, and neuronal subunits in vertebrates and invertebrates have been deduced based on the nucleotide sequence of their cloned cDNA. The primary structure of the nAChR subunits is highly conserved throughout the vertebrate phylum from Torpedo to humans. For instance, the a1 subunit of Torpedo shares 77% sequence identity with the human a1 subunit. Likewise, the human a1 subunit shares 36%, 33%, 30%, and 34% identity with the human b1, g, e, and d subunits, respectively (in mammals, an e subunit replaces the fetal g subunit early in postnatal life). These striking homologies suggest that the various subunits evolved by gene duplication from a common ancestral gene. Tissue-specific expression accounts for the muscle and neuronal subunit subfamilies.

Three Major Domains Arise from the Transmembrane Topology of the nAChR Subunits Each subunit is a polypeptide of 450–700 amino acids that consists of a long extracellular N-terminal segment, four transmembrane hydrophobic segments of about 20 amino acids each (termed M1–M4), a long cytoplasmic segment, and a short extracellular C-terminal tail. The five long extracellular segments fold and assemble into a large extracellular hydrophilic ACh-binding domain (Figure 1). The transmembrane segments form a membrane-embedded domain whose elements are closely organized around

503

504 Nicotinic Acetylcholine Receptors

Figure 1 Structural model of the nACHR from the Torpedo electric ray. Ribbon representations were prepared using coordinates deposited in the research collaboratory for structural bioinformatics (RCSB) protein data bank (PDB) under ID# 2BG9. The ribbon representations are shown on the background of the gray-colored molecular surface. The two facing a1 subunits are colored in orange while the b1, g, and d subunits are shown in cyan, yellow, and green, respectively. (a) Side view, from within the membrane. The thick gray lines delineate the membrane borders. (b) Top view, from the extracellular side. The M2 segments are five a helices closely organized around the axis of ion conduction that is at the center of the molecule and is perpendicular to the viewer. In both panels, the extrcellular end of M2 of the right a1 subunit is labeled with a black asterisk. Note that most of the cytoplasmic domain is missing due to the lack of 3-D structural information.

an axis perpendicular to the membrane plane, delineating thereby an aqueous pore. The five cytoplasmic segments fold into a hydrophilic domain that is thought to associate with components that anchor the receptor to the cytoskeleton, for example, the 43-kDa-Rapsyn protein. The cytoplasmic domain also contains structural elements that modulate channel activity. Notably, the nAChR-subunit topology is common for all other pentameric neurotransmitter receptor channels that are gated by serotonin, glycine, g-aminobutyric acid (GABA), glutamate, or histamine. These ionic channels constitute the Cys-loop receptor superfamily whose members share a conserved disulfide bridge in their extracellular ligandbinding domain. The binding pockets for ACh are formed by the interface between each a subunit and the adjacent subunit. The a subunit contributes, to the ACh-binding pocket, the so-called principal component, which consists of a series of loops carrying highly conserved ACh-binding residues, mainly aromatic but also a cysteine pair. The neighboring subunit contributes a series of complementary ACh-binding residues. Affinity labeling, binding studies, and electrophysiological recordings performed with wild-type (WT) and mutated nAChRs identified ACh-binding pockets at the a/g and a/d interfaces in the case of the electric organ and

muscle nAChRs (Figure 1(b)), a/b interfaces in the case of neuronal heteromeric nAChRs, and a/a interfaces in the case of homomeric neuronal nAChRs. The acetylcholine-binding protein (AChBP) is a glial pentameric water-soluble protein sharing structural homology with the extracellular neurotransmitterbinding domain of the nAChRs’ a subunits. It does not have the transmembrane channel and the cytoplasmic domains (Figure 2). The AChBP is secreted to the synaptic cleft between neurons of the mollusk Lymnaea stagnalis and of Aplysia, where it apparently sequesters ACh, regulating thereby cholinergic transmission. The high-resolution X-ray crystal structure of the AChBP also supports the concept of intersubunit ligand-binding pockets (Figure 2). The five M2 helical segments of the full-length nAChRs line the ion-channel pore (Figure 1), which was initially identified with noncompetitive pore blockers such as chlorpromazine or triphenylmethylphosphonium. Residues belonging to the M2 segments define the activation gate, which acts as a barrier obstructing the flow of ions when the receptor is at rest (i.e., displaying a closed pore). The location of the activation gate is still controversial. On the one hand, three-dimensional (3-D) structural models built on the basis of cryo-electron microscopy suggested that the activation gate is a ‘hydrophobic girdle’

Nicotinic Acetylcholine Receptors 505

Figure 2 X-ray crystal structure of the AChBP from the mollusk Lymnaea stagnalis. Ribbon representations were prepared using coordinates deposited in the PDB under ID# 1UW6. Nicotine molecules are shown at the interface between each two neighboring subunits, as green and blue spheres (carbon and nitrogen atoms, respectively). Note that the five subunits share identical primary, secondary, and tertiary structures. (a) Side view; (b) top view.

located midway between the extracellular and intracellular pore’s vestibules. It was further suggested that the ‘hydrophobic girdle’ breaks apart when the M2 segments rotate, around their own longitudinal axis, upon activation (i.e., upon channel opening). On the other hand, state-dependent accessibility of methanethiosulfonates or zinc ions to cysteines or histidines introduced along the pore revealed that the activation gate is a constriction located close to the intracellular end of the M2 segments, that is, close to the bottom of the pore. These and other structure–function relation studies also indicate that (1) the resting and active states share similar patterns of amino acid accessibility to the pore’s solvent and (2) channel gating predominantly involves rigid tilting motions of the M2 segments, which widen (open) or narrow (close) the bottom-pore constriction. Amino acids located within the M1–M2 connecting segment, close to the intracellular end of the M2 segments, contribute to the cationic (versus anionic) selectivity of the channel. As these residues are also part of the bottom-pore constriction, it is concluded that components of the activation gate act as a selectivity filter upon channel opening.

The nAChR is an Allosteric Protein Cryo-electron microscopy, affinity labeling, binding kinetics of various nicotinic ligands, and rapid kinetic measurements of ion-channel opening and closing events showed that the ACh receptor exhibits properties typical of allosteric proteins. That is, first, the ACh-binding site and the ion channel are spaced far apart (35 A˚). Second, the nAChR can undergo reversible transitions between distinct allosteric conformations even in the absence of an agonist, including rare but detectable spontaneous channel openings. The various allosteric states, which preexist in

reversible equilibrium already before ligand binding, are stabilized not only by agonists or competitive antagonists that bind to the ACh-binding sites, but also by ‘allosteric’ effectors (e.g., Ca2þ ions) that bind to topographically distinct sites. The ion channel of cloned nAChRs expressed in heterologous cell systems opens at relatively high concentration of ACh (KD in the 50–100 mM range), which is close to the concentration of ACh found in the synaptic cleft during transmission. On the other hand, prolonged exposure of the nAChR to the neurotransmitter leads to stabilization of states that display (1) higher affinity for the agonists (KD  1 mM or 5 nM for ACh in case of neuronal or muscle nAChR, respectively) and (2) a closed pore conformation that does not conduct ionic currents. These states are referred to as ‘desensitized,’ and they are stabilized by ACh at equilibrium. Hence, the high-affinity binding states do not take part in the transduction of the physiological signal at the synapse; rather, they play a role in the regulation of the efficacy of signal transmission.

Compartmentalized Expression of Muscle nAChRs Depends on Electric Activity and Release of Neurotrophic Factors In the adult neuromuscular junction, nAChR is localized almost exclusively in the postsynaptic (muscle) membrane of the motor endplate, with a density of 10 000 to 25 000 molecules mm–2 at the crest of the folded membrane. In the embryonic muscle fiber, nAChR molecules are evenly distributed extrasynaptically on the surface of the muscle cell, freely mobile and metabolically labile (half-life of about 18 h). During the formation of the endplate, nAChR molecules aggregate on postsynaptic membrane regions,

506 Nicotinic Acetylcholine Receptors

opposite to the motor-nerve ending, where they become immobile and metabolically stable (half-life of about 11 days). Around birth, in some species, an e subunit replaces the fetal g subunit, giving rise to an adult form with changed channel properties. Meanwhile, the extrasynaptic receptors disappear owing to repression of their synthesis, which takes place as soon as the muscle electric activity starts. In the adult, preventing stimulation of the muscle by sectioning of the motor nerve reactivates the biosynthesis of extrasynaptic nAChR, leading to the so-called phenomenon of ‘denervation hypersensitivity.’ Electrical stimulation of the muscle or its reinnervation by neurons restores the original postsynaptic distribution. Such compartmentalized expression of nAChR genes results, in part, from the regulation of transcription, which is active in most sarcoplasmic nuclei of developing myotubes. In the adult, this process becomes restricted to the subsynaptic ‘fundamental’ nuclei. In the promoter of the receptor genes, distinct elements (N Box vs. E Box) control subsynaptic transcription versus activitydependent extrasynaptic repression. Posttranslational mechanisms include the conformational maturation of the receptor protein, its transit via a specialized Golgi apparatus (in the mature endplate), and its targeting, aggregation, metabolic stabilization, and immobilization in the postsynaptic membrane. Factors of neural origin that are involved in compartmentalizing nAChR gene expression include acetylcholine receptor-inducing activity (ARIA), a factor homologous to human heregulin, and glial growth factor, which binds to tyrosine kinase receptors of the erbB family. Another important component, the acetylcholine receptor-aggregating factor, referred to as AGRIN, elicits formation of nAChR clusters. A cytoskeletal protein 43-kDa-Rapsyn, which is considered to be associated with the nAChR, contributes to AGRIN immobilization and stabilization in the postsynaptic membrane. In the nematode Caenorhabditis elegans, a transmembrane protein, LEV-10, was found to be required for clustering of nAChR molecules in cholinergic neuromuscular junctions. Another transmembrane protein of C. elegans, RIC-3, is involved in enhancing the maturation (subunit folding and assembly) of nAChRs.

nAChRs Play Decisive Physiological and Pathological Roles Activation of nAChRs on the muscle fibers initiates a cascade leading to muscle contraction. Hence, elimination of nAChR molecules from the neuromuscular junction, as occurs in myasthenia gravis, results in severe muscle weakness. The latter is an autoimmune

disease induced by anti-nAChR autoantibodies and aggravated by activation of autoimmune T lymphocytes. Congenital myasthenic syndromes are human disorders resulting from a range of mutations in the muscle nAChR that impair channel activity and may lead to degeneration of the endplate. In the brain, nicotinic receptor genes are expressed exclusively in neurons and display different patterns of expression, from highly restricted to a few nerve cells (e.g., the a2 subunit) to widespread ones (e.g., the b2 subunit). Inactivation of the b2-subunit gene by homologous recombination in mice interferes with their passive avoidance learning and alters a cognitive behavior referred to as ‘exploratory,’ while a more automatic ‘navigatory’ behavior is preserved. nAChRs are responsible for autosomal dominant nocturnal frontal lobe epilepsy and modulate pain transmission, and are also considered to be involved in autism and schizophrenia. Correlations have been reported between cigarette smoking and protection against ulcerative colitis and Parkinson’s disease. Addiction to nicotine involves high-affinity nicotinic receptors associated with midbrain dopaminergic neurons, as demonstrated in laboratory mice lacking the nAChR b2 or a4 subunit. These mice are also useful animal models of attention-deficit activity disorders and sudden infant death syndrome. Nicotinic receptor-binding drugs are considered as potential therapeutic agents in Alzheimer’s disease, Tourette’s syndrome, and anxiety disorders. See also: Cholinergic Pathways in CNS; Cholinergic Neurotransmission in the Autonomic and Somatic Motor Nervous System.

Further Reading Brejc K, van Dijk WJ, Klaassen RV, et al. (2001) Crystal structure of an ACh-binding protein reveals the ligand-binding domain of nicotinic receptors. Nature 411: 269–276. Bourne Y, Talley TT, Hansen SB, et al. (2005) Crystal structure of a Cbtx–AChBP complex reveals essential interactions between snake alpha-neurotoxins and nicotinic receptors. EMBO Journal 24: 1512–1522. Celie PH, van Rossum-Fikkert SE, van Dijk WJ, et al. (2004) Nicotine and carbamylcholine binding to nicotinic acetylcholine receptors as studied in AChBP crystal structures. Neuron 41: 907–914. Changeux JP and Edelstein SJ (2005) Nicotinic Acetylcholine Receptors. New York: Odile Jacob Publishing Corporation. Cymes GD, Ni Y, and Grosman C (2005) Probing ion-channel pores one proton at a time. Nature 438: 975–980. Engel AG, Ohno K, and Sine SM (2003) Neurological diseases: Sleuthing molecular targets for neurological diseases at the neuromuscular junction. Nature Reviews Neuroscience 4: 339–352.

Nicotinic Acetylcholine Receptors 507 Gally C, Eimer S, Richmond JE, et al. (2004) A transmembrane protein required for acetylcholine receptor clustering in Caenorhabditis elegans. Nature 431: 578–582. Halevi S, McKay J, Palfreyman M, et al. (2002) The C. elegans ric-3 gene is required for maturation of nicotinic acetylcholine receptors. EMBO Journal 21: 1012–1020. Hansen SB, Sulzenbacher G, Huxford T, et al. (2005) Structures of Aplysia AChBP complexes with nicotinic agonists and antagonists reveal distinctive binding interfaces and conformations. EMBO Journal 24: 3635–3646. Karlin A (2002) Emerging structure of the nicotinic acetylcholine receptors. Nature Reviews Neuroscience 3: 102–114. Keramidas A, Moorhouse AJ, Schofield PR, et al. (2004) Ligand-gated ion channels: Mechanisms underlying ion selectivity. Progress in Biophysics & Molecular Biology 86: 161–204. Lester HA, Dibas MI, Dahan DS, et al. (2004) Cys-loop receptors: New twists and turns. Trends in Neuroscience 27: 329–336. Lindstrom J (1997) Nicotinic acetylcholine receptors in health and disease. Molecular Neurobiology 15: 193–222. Maskos U, Molles BE, Pons S, et al. (2005) Nicotine reinforcement and cognition restored by targeted expression of nicotinic receptors. Nature 436: 103–107.

Paas Y, Gibor G, Grailhe R, et al. (2005) Pore conformations and gating mechanism of a Cys-loop receptor. Proceedings of the National Academy of Sciences of the United States of America 102: 15877–15882. Sine SM and Engel AG (2006) Recent advances in Cys-loop receptor. Nature 440: 448–455. Sunesen M and Changeux JP (2003) Transcription in neuromuscular junction formation: Who turns on who? Journal of Neurocytology 32: 677–684. Tapper AR, McKinney SL, Nashmi R, et al. (2004) Nicotine activation of alpha4 receptors: Sufficient for reward, tolerance, and sensitization. Science 306: 1029–1032. Taylor P, Osaka H, Molles B, et al. (2000) Contributions of studies of the nicotinic receptor from muscle to defining structural and functional properties of ligand-gated ion channels. In: Clemanti F, Fornasari D, and Gotti C (eds.) Handbook of Experimental Pharmacology: Neuronal Nicotinic Receptors, vol. 144. Berlin: Springer. Unwin N (2005) Refined structure of the nicotinic acetylcholine receptor at 4 A˚ resolution. Journal of Molecular Biology 346: 967–989. Wilson GG and Karlin A (1998) The location of the gate in the acetylcholine receptor channel. Neuron 20: 1269–1281.

This page intentionally left blank

NEUROPEPTIDES AND NEUROTROPHIC FACTORS

A. Neuropeptides

511

B. Neurophic Factors

570

This page intentionally left blank

Neuropeptide Synthesis and Storage J A Sobota, B A Eipper, and R E Mains, University of Connecticut, Farmington CT, USA ã 2009 Elsevier Ltd. All rights reserved.

Introduction Small, biologically active peptides are crucial to intercellular communication in most multicellular organisms. Both neuropeptides and classical neurotransmitters are secreted at most nerve terminals. Neurons and endocrine cells have the ability to synthesize bioactive peptides, store them for long periods, and secrete them upon command; many other cells in the body also use peptides for communication, although storage in nonneuroendocrine cells is usually minimal. Peptides are used as neurotransmitters, neuromodulators, neurotoxins, autocrine and paracrine agents acting locally, and growth factors and hormones acting long distance. These peptides range in size from three amino acids (TRH) to more complex, disulfide-linked peptides such as insulin, with its A (21 amino acids) and B (30 amino acids) chains. All bioactive peptides share important common features in their biosynthesis, and these features affect the way in which peptides are stored for secretion in a regulated manner.

Steps in Peptide Biosynthesis Peptides are synthesized as larger inactive precursors with an NH2-terminal signal sequence that guides the insertion of the nascent chain into the lumen of the endoplasmic reticulum (ER), after which the signal sequence is removed (Figure 1). Peptide synthesis always begins in the cell body and requires the ER and the Golgi complex. This is a critical difference from the biosynthesis of small-molecule neurotransmitters such as glutamate, g-aminobutyric acid (GABA), acetylcholine, and norepinephrine, which are synthesized at the nerve terminal and can be replenished locally after depletion during neurotransmission. Some peptide precursors are glycosylated; N-linked oligosaccharide chains are added co-translationally (Figure 1, ER). The NH2-terminal regions of many promolecules contain Cys residues that form intramolecular disulfide bonds; this process occurs in the ER. As intact promolecules proceed through the Golgi complex, N-linked oligosaccharides are modified, O-linked sugar chains can be added, and other modifications such as sulfation and phosphorylation

can occur (Figure 1, Golgi). The same enzymes that carry out these early modifications of peptide precursors are also responsible for modifying integral membrane proteins and other secreted proteins; they are not unique to neuropeptide biosynthesis. Depending on the promolecule and the cell type, the first enzymatic step unique to neuropeptide biosynthesis, endoproteolytic cleavage, may begin to occur at the distal, or exit, side of the Golgi complex in the trans-Golgi network (TGN) (Figure 1). In neurons and endocrine cells, the promolecules and a number of peptide-processing enzymes enter immature secretory granules, where a series of cell-type specific endoproteolytic cleavages proceed more rapidly. As the secretory granules mature, endoproteolytic cleavages continue, along with additional enzymatic modifications of the smaller peptides: COOH-terminal and NH2-terminal exoproteolytic removal of basic amino acids, a-amidation at exposed COOH-terminal Gly residues, conversion of NH2-terminal Gln residues into pyroglutamic acid, and NH2-terminal acetylation. Many of these seemingly minor modifications are essential for biological activity. For many neuropeptides, closely related precursors produce families of related product peptides; examples include the oxytocin and vasopressin precursors; the three precursors for opiate peptides, b-endorphin, methionine-enkephalin, and dynorphin; substance P; and related tachykinins. The information content encoded by neuropeptides is magnified by several factors. Nearly every peptide precursor is expressed in more than one tissue (e.g., procholecystokinin and proenkephalin in the gut and brain), and different sets of product peptides are often found in each tissue. For example, cholecystokinin (CCK-4) is produced in the brain, whereas larger molecules such as CCK-8 and CCK-33 are more prevalent in the gut; methionineenkephalin is produced in the brain, whereas larger molecules such as peptides E and F are found in adrenal chromaffin cells. To add further complexity, several bioactive peptides can be produced from a single precursor. The most widely studied example is proopiomelanocortin (POMC), which can give rise to corticotropin (which stimulates adrenal glucocorticoid production); melanotropins (which cause skin darkening and affect appetite); and b-endorphin, an opiate-active peptide. The anterior pituitary POMC-producing cells (corticotropes) produce corticotropin, but do not make the endoproteolytic cleavages that yield melanotropins or opiate-active peptides. The intermediate pituitary POMC-producing cells cleave corticotropin to

511

512 Neuropeptide Synthesis and Storage

Ribosomes

mRNA KR

Signal sequence

Q

Nascent polypeptide

Promolecule

ER

Q

Synthesis

cis Oligosaccharide modification

Q

Regulated pathway

Trans-Golgi network

Immature granule

PO3−

Q Maturation

Mature granule storage

G

PC1, PC2 QC 37 glutamines) in huntingtin (htt). In a recent study, Rong et al. observed that htt-associated protein-1, a contributing factor to HD pathology, interacts with the dynein motor in sympathetic neurons. Mutant htt appears to influence this interaction and in so doing reduces the stability of intracellular TrkA. As a result, retrograde NGF signaling may well be compromised in the diseased neurons. Given that retrograde neurotrophic signaling may share mechanisms and protein machinery with other retrograde processes, it is conceivable that disruption or alteration in any components of the machinery will also inevitably affect retrograde NT signaling. Disruption of retrograde axonal transport has been suggested to contribute to the pathogenesis of a number of progressive disorders of motor neurons. In one study, transgenic overexpression of p50 was shown to result in inhibition of axonal transport and cause lateonset progressive degeneration in motor neurons. In addition, a family with chronic motor neuron disease has been shown to harbor a mutation in p150 glued, a protein that plays a critical role in the function of the dynein complex. Genetic mutations that affect the stability of the microtubules have been suggested to cause progressive motor neuronopathy in mice. A Trp524Gly substitution at the last residue of the tubulin-specific chaperone e protein (gene locus: chromosome 13) leads to a reduced number of microtubules in the sciatic and phrenic nerves. The affected mice develop progressive caudocranial degeneration of their motor

Retrograde Neurotrophic Signaling 589

axons from the age of 2 weeks and die 4–6 weeks after birth. Axonal degeneration apparently starts at the endplates and is prominent in the sciatic nerve and its branches and the phrenic nerve. Since intact microtubules are required for retrograde NGF transport, it is conceivable that retrograde NGF–pTrkA signaling is compromised in these mice as well, which may well in turn contribute to the disease. In another example, alsin, a putative guanine nucleotide exchange factor for Rab5, plays a critical role in regulating early endosomal fusion. Loss of alsin-2 function is believed to be the cause for a subset of amyotrophic lateral sclerosis (ALS) in humans, juvenileonset primary lateral sclerosis. In Als2(/) mice, endosomal transport of insulin-like growth factor 1 and BDNF receptors was selectively disturbed. In another mouse model of ALS, a missense mutation L967Q in Vps54 has been identified in the spontaneous autosomal recessive wobbler mutation. Axonal transport is impaired in wobbler mice, and Vps54 might be critical for retrograde vesicular transport and in particular for axonal transport in motor neurons. Since internalized NGF was found to localize in early endosomes in cholinergic neurons, retrograde NGF signaling is likely affected in these neurons as well. Recent studies have revealed that late endosomes that are marked by Rab7 may also play a role in modulating retrograde NGF–pTrkA signaling. In genetic studies, the locus for Rab7 has been implicated as the disease gene for Charcot-Marie-Tooth disease type 2B (CMT2B). CMT2B is a rare autosomal dominant genetic disorder that leads to peripheral sensory neuropathy with prominent axonal degeneration. Missense mutations in the highly conserved C-terminus of Rab7 (Leu129Phe, Val162Met, Lys157Asn) have been identified in CMT2B patients. It will be of great interest to examine whether retrograde NT trafficking or signaling is affected in neurons that harbor Rab7 mutations.

Summary Retrograde NT signaling plays a critical role in regulating survival and differentiation of specific populations of neurons in both the central nervous system and the peripheral nervous system. There is increasing evidence that NGF and other NTs bind to and activate their receptors at the axonal terminus to elicit the recruitment of an array of signaling molecules. Endocytosis of the NT–pTrk signaling complex is then sorted into a subpopulation of early endosomes or other cellular organelles to give rise to signaling endosomes. The signaling endosome driven by the dynein motor complex is retrogradely transported

via the MT to the cell body to effect gene expression and other somal events. Retrograde NT signaling thus appears to be a very complex process that involves numerous cellular components and events. Disruption of retrograde NGF and NT signaling may contribute or lead to many types of neurodegenerative diseases. See also: Neurotrophins: Physiology and Pharmacology.

Further Reading Campenot RB and MacInnis BL (2004) Retrograde transport of neurotrophins: Fact and function. Journal of Neurobiology 58: 217–229. Chao MV (2003) Neurotrophins and their receptors: A convergence point for many signalling pathways. Nature Reviews Neuroscience 4: 299–309. Cooper JD, Salehi A, Delcroix JD, et al. (2001) Failed retrograde transport of NGF in a mouse model of Down’s syndrome: Reversal of cholinergic neurodegenerative phenotypes following NGF infusion. Proceedings of the National Academy of Sciences of the United States of America 98: 10439–10444. Cui B, Wu C, Chen L, et al. (2007) One at a time, live tracking of NGF axonal transport using quantum dots. Proceedings of the National Academy of Sciences of the United States of America 104(34): 13666–13671. Deinhardt K, Salinas S, Verastegui C, et al. (2006) Rab5 and Rab7 control endocytic sorting along the axonal retrograde transport pathway. Neuron 52: 293–305. Delcroix JD, Valletta JS, Wu C, Hunt SJ, Kowal AS, and Mobley WC (2003) NGF signaling in sensory neurons: Evidence that early endosomes carry NGF retrograde signals. Neuron 39: 69–84. Heerssen HM, Pazyra MF, and Segal RA (2004) Dynein motors transport activated Trks to promote survival of targetdependent neurons. Nature Neuroscience 7: 596–604. Howe CL and Mobley WC (2005) Long-distance retrograde neurotrophic signaling. Current Opinion in Neurobiology 15: 40. Huang EJ and Reichardt LF (2001) Neurotrophins: Roles in neuronal development and function. Annual Review of Neuroscience 24: 677–736. LaMonte BH, Wallace KE, Holloway BA, et al. (2002) Disruption of dynein/dynactin inhibits axonal transport in motor neurons causing late-onset progressive degeneration. Neuron 34: 715–727. Levi-Montalcini R (1987) The nerve growth factor 35 years later. Science 237: 1154–1162. Salehi A, Delcroix JD, Belichenko PV, et al. (2006) Increased App expression in a mouse model of Down’s syndrome disrupts NGF transport and causes cholinergic neuron degeneration. Neuron 51: 29–42. Taylor AM, Blurton-Jones M, Rhee SW, Cribbs DH, Cotman CW, and Jeon NL (2005) A microfluidic culture platform for CNS axonal injury, regeneration and transport. Nature Methods 2: 599–605. Wu C, Lai CF, and Mobley WC (2001) Nerve growth factor activates persistent Rap1 signaling in endosomes. Journal of Neuroscience 21: 5406–5416. Zweifel LS, Kuruvilla R, and Ginty DD (2005) Functions and mechanisms of retrograde neurotrophin signalling. Nature Reviews Neuroscience 6: 615–625.

BDNF in Synaptic Plasticity and Memory N H Woo and B Lu, National Institutes of Health, Bethesda, MD, USA Published by Elsevier Ltd.

Introduction Among members of the neurotrophin family, brainderived neurotrophic factor (BDNF) stands out for its ability to regulate synaptic plasticity and various cognitive functions of the brain. A Medline search with the terms ‘BDNF’ and ‘synaptic’ yields more than 700 research articles, mostly published in the last 7 years. Given that neurotrophins were initially defined as secretory factors that promote neuronal survival and differentiation during development, the role of BDNF in synaptic modulation was not recognized until late 1990s. A number of observations have aided in the realization that the primary function of BDNF is to regulate synaptic transmission and plasticity, rather than neuronal survival. One is that BDNF is widely distributed in many regions of the adult brain, with levels much higher than any other neurotrophins. The other is that the expression of BDNF can be rapidly enhanced by neuronal activity under conditions relevant to synaptic plasticity. Because neuronal activity is known to be crucial for synaptic plasticity, it was hypothesized that activity-dependent synaptic modulation is mediated by BDNF. Indeed, early studies demonstrated that BDNF mimics neuronal activity in altering the number and/or strength of synaptic connections. Subsequent studies revealed a much more complex and interesting picture. On the one hand, BDNF regulates various forms of synaptic plasticity, leading to changes in neuronal circuitry subserving complex behaviors. On the other hand, many aspects of BDNF cell biology, such as transcription and secretion, are tightly controlled by neuronal activity. Complex interactions between BDNF and neuronal activity may offer a plethora of means to control sophisticated cognitive functions of the mammalian brain.

Cell Biology of BDNF All neurotrophins arise from their precursors as a result of proteolytic cleavage of the prodomain. Proneurotrophins have long been thought to be inactive precursors. However, this view was challenged a few years ago when proneurotrophins were shown to promote apoptosis via p75 neurotrophin receptor (p75NTR). This is opposite to the cell survival effect by mature neurotrophins, which act via their preferred

590

tropomyocin-related receptor tyrosine kinase (Trk) receptors. Similarly, proBDNF and mature BDNF (mBDNF) have been shown to elicit opposite effects on synaptic plasticity (Figure 1). In recognition that proneurotrophins are biologically active, cleavage of proneurotrophins becomes an important regulatory mechanism that controls the direction of neurotrophin regulation. Signal Transduction

The biological functions of BDNF are mediated by two receptor systems: TrkB and p75NTR. It is well established that mBDNF binds TrkB with high affinity. Upon binding, BDNF triggers TrkB dimerization resulting in tyrosine phosphorylation in its cytoplasmic domain. These autophosphorylation events recruit a series of intracellular proteins that primes subsequent activation of several signaling pathways. Three classical signaling pathways have been identified: phosphatidylinositol-3-kinase (PI3K) pathway, phospholipase C-g (PLC-g) pathway, and Map-Erk Kinase (MEK)-MAPK pathway (Figure 1). The majority of BDNF actions described thus far are attributed to signaling cascades activated by TrkB. In addition to cell surface signaling, BDNF induces the endocytosis of TrkB. Rather than simply inactivating TrkB, the endocytosis of BDNF-TrkB complex results in the formation of BDNF-TrkB signaling endosomes, triggering signaling cascades different from those initiated from cell surface TrkB. This process is required for translation-dependent long-term functions, and is involved in retrograde propagation of BDNF signal from axonal terminals to cell body. All neurotrophins including proBDNF bind p75NTR, which triggers signaling events distinct from those by Trk receptors. The cytoplasmic domain of p75NTR lacks intrinsic catalytic activity. Upon ligand binding, several intracellular signal transduction cascades are activated, including nuclear factor kappa B (NFkB), Jun kinase, and sphingomyelin hydrolysis (Figure 1). Notably, p75NTR activation is associated with the initiation of apoptosis. For many years, p75NTR was considered a ‘low affinity’ neurotrophin receptor. Recent studies indicate that preferred ligands for p75NTR are proneurotrophins, with binding affinities just as high as that between mature neurotrophins and Trk receptors. Current data support a model that pro- and mature neurotrophins induce very different functions through two distinct receptor-signaling systems. An added complexity is the newly discovered coreceptor sortilin. The prodomain of proneurotrophins bind sortilin, whereas the mature domain binds p75NTR. The formation of

BDNF in Synaptic Plasticity and Memory 591 Synthesis of unprocessed BDNF protein in rough ER

Intracellular cleavage by furin or PC1

TransGolgi

Sorting of BDNF

Constitutive secretory pathway

Regulated secretory pathway

tPA

Plasminogen

Extracellular cleavage by tPA/plasmin or MMP

mBDNF Plasmin proBDNF

TrkB receptor

p75NTR

PI3K pathway NF-kB

AKT

P

PI3

RhoA-GTP

P

P

P

P

P

P

P

PLCg

JNK

P

P

RAS

ERK/MAPK

ERK/MAPK pathway

PLCg pathway IP3 PKC

Cell death LTD

DAG PIP2

Ca2+ i

Cell survival LTP

Figure 1 Synthesis, trafficking, and receptor-signaling of BDNF. Initially synthesized in the endoplasmic reticulum (ER) as a precursor protein, proBDNF is properly folded in the ER and Golgi network and packaged into secretory vesicles. Subsequently, BDNF is sorted into either the constitutive or regulated secretory pathway, and transported to the appropriate site of release. The prodomain can be cleaved intracellularly by furin or protein convertases, resulting in the secretion of mature BDNF (mBDNF). Alternatively, proBDNF can be secreted and cleaved extracellularly by the tPA/plasmin cascade or metalloproteinases to yield mBDNF. Once secreted, proBDNF and mBDNF elicit diverse and often opposing biological actions via two distinct receptor-signaling systems. mBDNF binds TrkB, leading to the autophosphorylation of tyrosine residues in the tyrosine kinase domain. Consequently, three major signaling cascades can be activated by mBDNF-TrkB, including the PI3K pathway, ERK/MAPK pathway, and PLCg pathway. In contrast, proBDNF binds p75NTR, resulting in the activation of several signaling molecules, including NF-kB, JNK and RhoA.

592 BDNF in Synaptic Plasticity and Memory

the sortilin–proneurotrophin–p75NTR triplex may be necessary for p75NTR signaling. Activity-Dependent Controls

A cardinal feature of BDNF is that its expression is regulated by neuronal activity. It is now recognized that a multitude of physiological stimuli can alter BDNF expression. For example, visual input and sensory stimulation of the whiskers control BDNF expression in the visual cortex and barrel cortex, respectively. In the superchiasmatic nucleus and amygdala, expression of BDNF is regulated by circadian rhythm and fear emotion. Remarkably, learning or exercise can also enhance BDNF expression in the hippocampus. Moreover, BDNF levels are also affected in a variety of pathological conditions associated with altered neuronal activity in the brain including seizure, Alzheimer’s, depression, and stress. In addition to regulation of BDNF gene expression, a new theme emerging from recent studies is that neuronal activity also controls many cellular processes of BDNF, including intracellular trafficking, secretion of BDNF, and perhaps cleavage of proBDNF. Transcription The genomic structure of BDNF is quite complex. In rats, there are at least four promoters controlling four short 50 exons. Each 50 exon is alternatively spliced onto a common 30 exon (exon V) encoding the pre-proBDNF protein. In humans, the latest study reported seven promoters and eight exons, with exon VIII being the common exon coding for preproBDNF. It has long been a puzzle why nature has designed multiple BDNF transcripts that encode exactly the same protein. Cumulative evidence now indicates that these transcripts are distributed in different brain regions, different cell types, and even different parts of the cells (soma vs. dendrites). Importantly, their expression can be altered in response to different physiological stimuli. For instance, exon III transcript is detected only in cell bodies, whereas exon IV transcript is present both in cell bodies and dendritic processes of neurons in the visual cortex. During cerebellar development, thyroid hormone treatment selectively primarily enhances the expression of exon II mRNA. Emerging evidence indicates that BDNF promoters are differentially involved in various neurological and psychiatric disorders. Promoter II-drive transcription can be suppressed by a neuronal silencer, and such suppression is removed by huntingtin, which binds and sequesters the silencer in the cytosol of cortical neurons. This is important for the survival of cortical neurons that projects to the striatum. In Huntington’s disease, the mutant huntingtin can no longer bind the

silencer, resulting in the translocation of the silencer into the nucleus and suppression of BDNF promoter II. Another striking example of BDNF promoter specific regulation involves MeCP2, which is a methyl-CpG-dependent transcriptional repressor that binds methylated DNA BDNF promoter III. Neuronal depolarization dissociates MeCP2 from promoter III, leading to the expression of exon III transcript in hippocampal neurons. Mutation in MeCP2, which occurs in 80% of Rett syndrome patients, abolishes this activity-dependent form of regulation. Promoter IV has been implicated in stress and is the major target of glucocortical and mineralocortical receptors. In clinical studies, a human single nucleotide polymorphism (SNP) in the exon IV has been associated with epilepsy and late-onset Alzheimer’s. Among all the promoters, promoter III has received much attention because it is by far the most effectively regulated by neuronal activity in the amygdala, hippocampus, and cortex. An increase in promoter III-driven transcription has been associated with long-term potentiation (LTP) and memory. Early work showed that BDNF gene expression was dependent on a rise in intracellular calcium and that application of high Kþ to cultured cortical neurons selectively enhanced exon III expression. Based on these observations, three elements in promoter III were characterized to be involved in Ca2þ-dependent expression of BDNF: the Ca2þ responsive sequence 1 (CaRE1) that binds Ca2þ responsive transcription factor (CaRF), the E-Box that binds upstream stimulatory factor (USF), and the classic cAMP responsive element (CRE) that binds cAMP responsive element binding (CREB). In addition, the transcription through promoter III is regulated by NF-kB and MeCP2. Taken together, transcription of BDNF exon III is tightly regulated by several mechanisms that couple neuronal activity with gene transcription. Processing and trafficking Like all neurotrophins, BDNF mRNA is translated into a precursor protein, pre-proBDNF, which enters into the endoplasmic reticulum (ER) lumen through its N-terminal ‘pre’ sequence (signal peptide). After the removal of the pre-sequence by signal peptidases in the rough ER, the protein is folded in the trans-Golgi network and then packaged into secretory vesicles. Once folded correctly, BDNF is sorted into one of two principal pathways, the constitutive (i.e., spontaneous release) or regulated (i.e., release in response to stimuli) secretory pathway (Figure 2). The BDNF-containing vesicles are trafficked to the appropriate subcellular compartment. In neuronal dendrites and spines, BDNF appears to be stored in a special type of

BDNF in Synaptic Plasticity and Memory 593

175

E-LTP

Control

EPSP slope (%)

BDNF 150 BDNF

125 100

Facilitates vesicle docking

Readily releasable pool (RRP)

75 −20

0

20 40 Time (min)

a

60

80 b

WT (12 slices, 6 mice) BDNF+/− (8 slices, 5 mice) fEPSP slope (% of baseline)

L-LTP 250

TrkB

200 Plasminogen mBDNF

150 100

tPA

Plasmin

Translation

50 TBS

0 −20 0

20 40 60 80 100 120 140 160 180 Time (min)

c

Vehicle control (7 slices, 5 mice)

d

Transcription BDNF promoter III proBDNF

proBDNF (7 slices, 4 mice)

fEPSP slope (% of baseline)

LTD

e

160 a

140

b

b

a p75NTR

p75NTR−/− (4 slices, 3 mice)

120 100

?

a

80

NR2B

60 40

b 2 ng/ml−1 proBDNF

20 0

0

10

20

30 40 50 Time (min)

LFS 60

70

proBDNF Postsynaptic

80 f

Figure 2 Regulation of distinct forms of hippocampal synaptic plasticity by BDNF. (a) mBDNF facilitates E-LTP in neonatal hippocampus in which BDNF level is low. Application of tetanic stimulation to neonatal slices (p12-p13) induces only short-term potentiation (STP), which can be converted to E-LTP by exposure to exogenous BDNF. (b) mBDNF facilitates E-LTP by promoting vesicle docking. (c) mBDNF is also involved in L-LTP. Strong theta-burst stimulation (TBS) induces robust L-LTP in wild-type (WT), but not in BDNF þ/ mice hippocampal slices. (d) ProBDNF ! mBDNF conversion by tPA/plasmin is required for L-LTP. Strong TBS triggers the secretion of tPA, which cleaves plasminogen to form plasmin. Plasmin subsequently cleaves proBDNF to yield mBDNF, which binds TrkB and permits L-LTP expression. (e) proBDNF promotes hippocampal LTD. Slices from p75NTR / mice fail to exhibit NMDA receptor-dependent LTD, whereas treatment with cleavage-resistant proBDNF enhances LTD in wild-type slice. (f) proBDNF binds to p75NTR to facilitate LTD, possibly through the regulation of NR2B, a distinct NMDA receptor subunit implicated in hippocampal LTD. EPSP, excitatory postsynaptic potential; fEPSP, field excitatory postsynaptic potential; LFS, low-frequency stimulation. (a) Reproduced from Figurov A, Pozzo-Miller L, Olafsson P, Wang T, and Lu B (1996) Regulation of synaptic responses to high-frequency stimulation and LTP by neurotrophins in the hippocampus. Nature 381: 706–709, with permission. (c) Reprinted with permission from Pang PT, Teng HK, Zaitsev E, et al. (2004) Cleavage of proBDNF by tPA/plasmin is essential for long-term hippocampal plasticity. Science 306: 487–491. Copyright 2004 AAAS. (e) Reproduced from NH Woo, Teng HK, Ciao C, et al. (2005) Activation of p75NTR by proBNF facilitates hippocampal long-term depression. Nature Neuroscience 8: 1069–1077, with permission.

594 BDNF in Synaptic Plasticity and Memory

secretory granules that lack chromogranin A (CGA), a marker for large dense core vesicles (LDCV). In contrast, conventional BDNF-containing LDCVs have been found in axons and terminals, possibly through anterograde axonal transport. Nonneuronal cells such as fibroblasts and Schwann cell secrete neurotrophins constitutively, whereas principal neurons and neuroendocrine cells secrete neurotrophins in response to depolarization and a rise in intracellular calcium. The dendritic trafficking and synaptic localization of BDNF appear to be influenced by its own prodomain. This was implicated in a study that examined a SNP in the pro-region of the human BDNF gene. This SNP, located at nucleotide 196, produces a valine-to-methionine substitution at amino acid 66 (Val66Met). In cultured hippocampal neurons, fluorescence-tagged val-BDNF is distributed in cell body, as well as dendrites. A fraction of valBDNF is also located at synapses, as revealed by their co-localization with synaptic markers. In marked contrast, met-BDNF is largely located in cell body and proximal dendrites. Met-BDNF was rarely localized at distal dendrites and was absent at synapses. These results suggest that the prodomain, particularly the region containing Val66, is critical for dendritic trafficking and synaptic localization of BDNF. A long-held view is that proneurotrophins, particularly proNGF and proBDNF, are processed by intracellular proteases including the serine protease furin in the trans-Golgi network and the prohormone convertases (PC1/3) in the secretory granules. Recent studies have shown that a large fraction of BDNF in the brain is secreted in the proform, which is converted to mBDNF by extracellular proteases including plasmin or metalloproteinases (MMP3 or MMP7). Because proBDNF and mBDNF elicit distinct and often opposing biological actions through different receptors, proteolytic cleavage has now emerged as a new mechanism that determines the function of BDNF. Of particular interest is tissue plasminogen activator (tPA), an extracellular protease that converts the inactive zymogen plasminogen to plasmin. tPA is secreted from axonal terminals in response to neuronal activity. It is conceivable that neuronal activity could control proBDNF ! mBDNF conversion by triggering tPA secretion. Secretion BDNF is perhaps the only neurotrophin indisputably secreted in response to neuronal activity. In fact, majority of BDNF is sorted into the regulated, rather than the constitutive, secretory pathway. Experiments using green fluorescent protein (GFP)tagged BDNF revealed that BDNF can be secreted

from either pre- or postsynaptic sites. The amount of BDNF secretion depends on the pattern of neuronal activity. Generally, tetanus such as those used to induce LTP is more effective in inducing BDNF secretion than low-frequency stimulation. This has been demonstrated in cultured neurons as well as in slices that underwent different forms of plasticity. Studies of Val66Met SNP have drawn attention to the role of prodomain in activity-dependent BDNF secretion. In neurons transfected with met-BDNF, depolarization-induced secretion was selectively impaired while constitutive secretion remained normal. Subsequent studies demonstrate that proBDNF is co-localized with the neurotrophin coreceptor sortilin intracellularly in secretory granules, and that sortilin interacts specifically with the prodomain in a region encompassing Val66Met. Remarkably, inhibition of the interaction between the prodomain and intracellular sortilin attenuates secretion of BDNF induced by depolarization, suggesting that this interaction is critical for regulated secretion. However, sortilin could interact with the prodomain of other neurotrophins incapable of regulated secretion, making it less likely to be a specific mechanism for sorting. On the other hand, a sorting motif was recently identified in the mature domain of BDNF, but not nerve growth factor (NGF), that interacts with a well-known sorting receptor, carboxypeptidase E (CPE). Such an interaction was deemed essential for sorting proBDNF into regulated pathway vesicles for activity-dependent secretion. Given that the prodomain promotes proper folding of neurotrophins, it is conceivable that interaction between the prodomain and sortilin may hold proBDNF in a correct configuration, exposing the mature domain to the sorting receptor CPE, which sorts proBDNF into the regulated secretory pathway.

Roles of BDNF in Synaptic Plasticity The ability of the mammalian brain to adapt or modify itself in response to experience and/or environment depends on the plasticity of synaptic connections. Substantial evidence indicates that the number and strength of synapses is readily altered by neuronal activity. This process, known as synaptic plasticity, displays several physiological properties that substantiate its role as a cellular correlate for multiple cognitive processes, including learning and memory. These include the activity dependence and associative nature of induction as well as the input specificity of expression, all of which endow the vast storage and processing capacity of the mammalian brain. Remarkably, BDNF is involved in many of these features. An emerging theme is that BDNF plays a

BDNF in Synaptic Plasticity and Memory 595

critical role in regulating several forms of synaptic plasticity in distinct regions of the brain, including the hippocampus and visual cortex. Input Specificity

Most activity-dependent forms of synaptic plasticity expressed in the brain are input specific, namely modifications at one synapse are not spread to synapses nearby. Experiments have shown that BDNF elicits its actions in a local and synapse-specific manner, with particular preference to active synapses. This unique property may be attributed to several key characteristics of BDNF signaling. First, secretion of BDNF is activity dependent and is likely to occur at synaptic sites. The control of BDNF secretion occurs relatively fast, usually in a timescale of seconds. Imaging studies demonstrate that BDNF is often co-localized with pre- and postsynaptic markers in hippocampal neurons, suggesting the synaptic localization of BDNF. Application of high-frequency stimulation (HFS) induces a rapid decay of GFP-tagged BDNF, indicating BDNF is secreted in an activity-dependent manner at synapses. Due to its negative charge, BDNF is thought to have a limited capacity for diffusion, and therefore constrains the actions of BDNF at or near its site of secretion. Second, there is good evidence that BDNF exon II and IV transcripts can be targeted into the dendrites of hippocampal neurons, and neuronal activity enhances such targeting. Recent studies have shown that dendritic BDNF mRNA can be translated locally into BDNF protein. Taken together, a possible scenario is that local synaptic activity triggers dendritic translation of BDNF and may serve as an alternative mechanism to ensure synapse-specific modulation by BDNF. Finally, local synaptic activity may ensure a better response of target synapses to BDNF by regulating TrkB trafficking. High-frequency neuronal activity has been shown to promote the insertion of TrkB into the surface membrane of hippocampal neurons. This process appears to be ligand independent and requires calcium influx and activation of Ca2þ/ calmodulin-dependent kinase II (CaMKII). BDNF secreted from active synapses/neurons recruits TrkB from extrasynaptic sites into lipid rafts, microdomains of membrane enriched at synapses. This lateral movement requires TrkB tyrosine kinase activity. Synaptic activity often induces a rise of postsynaptic cAMP. This local increase in cAMP concentration at active synapses facilitates translocation of TrkB into the postsynaptic density, and functions to gate synapse-specific effects of BDNF by controlling TrkB tyrosine phosphorylation locally. Finally, neuronal

activity promotes BDNF-induced TrkB endocytosis, a signaling event important for many long-term BDNF functions. All of these could contribute to a more efficient response to BDNF at active synapses. Early Phase Long-Term Potentiation

It is well established that BDNF plays a key role in LTP, a persistent enhancement of synaptic strength. LTP can be divided into an early phase (E-LTP) and a later phase (L-LTP). E-LTP is relatively short-lasting (1 h) and depends on protein phosphorylation, while L-LTP lasts many hours and requires new protein synthesis. Early work has focused on BDNF regulation of E-LTP in the hippocampus. In neonatal hippocampus in which BDNF levels are low, exogenous BDNF facilitates E-LTP induced by HFS (Figure 2(a)). In addition, exogenous BDNF facilitates LTP induced by subthreshold tetanus that normally induces weak potentiation. Conversely, in the adult hippocampus, a stage where endogenous levels of BDNF are relatively high, inhibition of BDNF activity either by functionblocking BDNF antibody or BDNF scavengers, TrkB immunoglobulin G (IgG), attenuates the expression of E-LTP. In agreement with pharmacological studies, genetically modified mice with mutation of either BDNF or TrkB gene exhibit severe impairments in E-LTP. Interestingly, heterozygous mice (BDNFþ/ ) with only half of the BDNF gene dosage show a similar degree of impairment as homozygous mice (BDNF / ), arguing that a critical level of BDNF is important for hippocampal LTP. The impairment of LTP in BDNF-mutant mice is reversed by acute application of recombinant BDNF or by virus-mediated BDNF gene transfer. Deletion of BDNF gene selectively in the adult forebrain by inducible knockout approaches confirms that the effects of BDNF on E-LTP are not due to developmental abnormalities. The effects of BDNF on hippocampal E-LTP result primarily from alterations of presynaptic function (Figure 2(b)). Exogenous BDNF enhances synaptic response to HFS and paired pulse facilitation (PPF), two indicators of presynaptic function. In mice lacking BDNF, posttetanic potentiation (PTP) and PPF are significantly reduced. Electron microscopy reveals a reduction in the number of vesicles docked at presynaptic active zones in these mutant mice. Moreover, biochemical experiments using hippocampal synaptosomes indicate that BDNF modulates the levels or phosphorylation of synaptic proteins involved in vesicle docking and fusion, such as synapsin, synaptophysin, and synaptobrevin. Taken together, the presynaptic role of BDNF for the mobilization and/or docking of synaptic vesicles to presynaptic active

596 BDNF in Synaptic Plasticity and Memory

zones may allow hippocampal synapses to follow tetanic stimulation more effectively, resulting in the facilitation of hippocampal LTP. It is important to note that the biological effects of BDNF are not exclusively the result of its presynaptic actions. In the dentate gyrus, the induction of LTP requires postsynaptic BDNF signaling. A series of studies have demonstrated that BDNF can exert postsynaptic modulatory effects by modulating a-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA)-type and N-methyl-D-aspartate (NMDA)-type glutamate receptors in neuronal cultures, and some potassium channels in hippocampal slices. In this respect, TrkB receptors have been observed to localize in the postsynaptic density of isolated synaptosomes prepared from cortical neurons. However, whether these postsynaptic modulatory effects of BDNF directly participate in LTP remains to be established. Late Phase Long-Term Potentiation

Several early studies suggest that BDNF may also play a role in late phase long-term potentiation (L-LTP). L-LTP-inducing tetanic stimulation selectively enhances the expression of BDNF and TrkB mRNAs in the hippocampus. BDNF transcription is regulated in part by CREB, a transcription factor required for L-LTP expression. The delayed and sustained enhancement of BDNF synthesis correlates well with the time course of L-LTP, which increases 2–4 h after L-LTP induction. More direct evidence comes from electrophysiology experiments showing a significant reduction in L-LTP recorded from slices treated with TrkBblocking antibody, or those from BDNF þ/ mice (Figure 2(c)). However, BDNF only regulates L-LTP induced by theta-burst stimulation (TBS) or application of the adenylate cyclase activator forskolin, but not by four spaced trains of HFS, a more standard protocol used to induce L-LTP. These results suggest that strong tetanic stimulation may induce signaling downstream of BDNF, bypassing the requirement of BDNF in L-LTP. Recent experiments have provided more in-depth insights as to the role of BDNF in L-LTP. In the presence of BDNF, weak tetanus that normally induces only E-LTP resulted in robust L-LTP. Moreover, perfusion of BDNF to hippocampal slices rescued L-LTP that is normally absent when protein synthesis is inhibited. It appears that BDNF is secreted largely in its precursor form (proBDNF), which is converted to mBDNF by extracellular proteases. Biochemical and genetic experiments indicate that this conversion is mediated by an enzymatic cascade that involves tPA and plasmin, two secreted

proteases found at hippocampal synapses. tPA has long been implicated in L-LTP, but the precise downstream effector(s) of tPA was not established. The current data support the notion that tPA cleaves the inactive zymogen plasminogen to form plasmin, which in turn cleaves proBDNF to generate mBDNF. Application of mBDNF, but not cleavageresistant proBDNF, completely reversed the L-LTP deficit observed in tPA and plasmin knockout mice. Thus, conversion of proBDNF to mBDNF by the tPA/ plasmin system is critical for L-LTP expression (Figure 2(d)). Two key cellular events associated with long-term synaptic plasticity are synaptic growth and de novo protein synthesis, both of which are regulated by BDNF. In addition to stimulating axonal growth, chronic application of BDNF to hippocampal slices increases the dendritic spine density of CA1 pyramidal neurons. This is particularly relevant since dendritic spines and protrusions are enhanced during L-LTP. In cell cultures, BDNF has been shown to increase mammalian target of rapamycin (mTOR)dependent translation of a panel of synaptically expressed transcripts including GluR1 and homer2 mRNAs in the dendrites of hippocampal neurons. However, exogenous BDNF applied immediately after strong TBS rescues L-LTP in hippocampal slices in which protein synthesis was blocked for the entire course of the experiments. This provocative result, together with the finding that gene expression of BDNF is stimulated by L-LTP-inducing tetanic stimulation, implies BDNF is a key protein synthesis product required for long-term modifications necessary for L-LTP expression. Long-Term Depression

In addition to its role in LTP, BDNF regulates longterm depression (LTD), a persistent reduction in synaptic strength induced by prolonged low-frequency stimulation. Expression of LTD is developmentally regulated and exists in several forms mediated by different glutamate receptors. The best-known form of LTD is the NMDA receptor (NMDAR)-dependent form, which is robustly expressed in young animals. mBDNF inhibits LTD in the visual cortex and hippocampus. Collectively, these observations point to a general theme that BDNF facilitates synaptic strengthening, but attenuates synaptic depression. Analogous to the survival and apoptosis effects of mBDNF and proBDNF, respectively, in the periphery, an important advance is that proBDNF, if uncleaved, enhances NMDAR-dependent LTD via p75NTR in the CNS. Compared to TrkB, the role of p75NTR in synaptic plasticity has not been studied until

BDNF in Synaptic Plasticity and Memory 597

recently. In p75NTR mutant mice (p75NTR / ), NMDAR-dependent LTD was completely absent while other forms of plasticity including NMDARdependent LTP and NMDAR-independent LTD were intact. Biochemical experiments indicate that NR2B, a specific NMDAR subunit uniquely implicated in LTD, was significantly reduced in the mutant hippocampus. Whole-cell recordings revealed a severe reduction in NR2B-mediated synaptic currents in CA1 neurons of p75NTR / mice. More importantly, application of cleavage-resistant proBDNF increased NR2B-mediated synaptic currents and enhanced LTD in hippocampal slices derived from wild-type mice but not in p75NTR / mice (Figure 2(e)). These findings suggest that proBDNF is an endogenous ligand of p75NTR during development, which acts to enhance LTD via modulating NR2B function (Figure 2(f)). Learning, Memory, and Other Cognitive Functions

Given its central role in synaptic plasticity, numerous studies have examined how BDNF regulates the acquisition (learning) and retention (memory) of new information. Thus far, the strongest correlation is observed between BDNF and hippocampal-dependent forms of memory, which include declarative or episodic and spatial memory. During contextual learning, BDNF expression is rapidly and selectively upregulated in the hippocampus. When BDNF signaling is disrupted either by inhibitors or by genetic knockout, spatial learning is significantly impaired, as reflected by poor performance in the Morris water maze. In many cases

Val

R

Pro Val-BDNF

R a

3.5 3.0 2.5 2.0 1.5 1.0 0.5 0

impairments in memory were also mirrored with LTP deficits. For instance, deletion of BDNF or TrkB gene in the adult forebrain results in a significant attenuation of contextual fear or spatial memories, as well as hippocampal LTP. A major advance came from a study on a SNP, which converts a valine (val) to a methionine (met) in the prodomain of the human BDNF gene. This SNP occurs with a frequency of approximately 19–25% in the Caucasian. Human subjects with the met allele exhibit lower hippocampal N-acetylaspartate (NAA), a putative measure of neuronal integrity and synaptic abundance. Functional imaging reveals an association of the met allele with abnormal hippocampal activation (Figure 3(a)). Most remarkably, subjects with the met-BDNF allele performed poorer in a hippocampal-dependent episodic memory task, but not in hippocampal-independent working memory and semantic memory tasks. In cultured neurons derived from rodent hippocampus, BDNF (valBDNF) is packaged in secretory granules that are distributed as puncta throughout cell body and dendrites, with some localized at synapses. In contrast, significantly less met-BDNF-containing granules are localized to dendrites and synapses. Moreover, regulated secretion of met-BDNF induced by neuronal depolarization, but not constitutive secretion, is significantly reduced. Thus, impairments in trafficking, synaptic targeting, and/or regulated secretion may explain the specific memory deficits seen in human subjects with the met allele (Figure 3(b)). These results represent the first demonstration of a role for

Met BDNF Met-BDNF

R Impairments in:

b

− dendritic trafficking − synaptic targeting − regulated secretion

Figure 3 Impact of a SNP in the BDNF gene on cognitive brain function and intracellular trafficking of BDNF. The SNP converts a valine to a methionine in amino acid 66 located in the prodomain of BDNF (val66met). (a) Differences in fMRI responses between val/val and val/ met subjects during a memory task. Subjects with val/met genotype exhibit abnormal hippocampal activation (shown in red). The met/met subjects also exhibit deficits in hippocampus-dependent episodic memory. (b) Cellular phenotypes associated with the BDNF val66met SNP. Val-BDNF is distributed throughout a typical hippocampal neuron including distal dendrites and synapses. In contrast, met-BDNF is rarely localized at distal dendrites or synapses and fails to undergo depolarization-induced secretion. Failure of intracellular trafficking and activity-dependent secretion of BDNF may underlie the cognitive deficits observed in subjects with the met allele. (a) Reproduced from Egan MF, Kojima M, Callicott JH, et al. (2003) The BDNF val66met polymorphism affects activity-dependent secretion of BDNF and human memory and hippocampal function. Cell 112: 257–269, with permission from Elsevier.

598 BDNF in Synaptic Plasticity and Memory

BDNF in human hippocampal function and of a single gene affecting human episodic memory. Genetic analyses in mice show that genes affecting memory performance often impact other cognitive functions. It is now recognized that deficits in BDNF may contribute to neurological and psychiatric disorders. In studies of drug addiction, LTP and LTD have emerged as candidate mechanisms for drug-induced alterations in the nucleus accumbens and ventral tegmental area. Several reports have demonstrated that BDNF modulates behavioral sensitization to cocaine. Substantial evidence also points to its role in depression. There is reduced BDNF expression in the hippocampus of animal models for depression; chronic treatment with antidepressants increases its levels. However, it is unclear whether antidepressants achieve their clinical effects on depression by upregulation of BDNF.

Conclusion Stemming from the multidisciplinary approaches used in present-day research ranging from cellular systems to behavior, BDNF is now recognized as a key regulator for synaptic circuits underlying many cognitive functions. New and unidentified role(s) of BDNF in plasticity and cognition will undoubtedly continue to surface and will provide abundant intellectual stimulation to drive future advances. Ultimately, understanding the actions of BDNF will aid in the development of therapeutic interventions that will alleviate a wide spectrum of neurological and psychiatric disorders derived from BDNF dysfunction. See also: Long-Term Depression (LTD): Metabotropic Glutamate Receptor (mGluR) and NMDAR-Dependent Forms; Long-Term Potentiation (LTP): NMDA Receptor Role.

Further Reading Barker PA (2004) p75NTR is positively promiscuous: Novel partners and new insights. Neuron 42: 529–533. Chao MV (2003) Neurotrophins and their receptors: A convergence point for many signaling pathways. Nature Reviews Neuroscience 4: 299–309. Chen ZY, Ieraci A, Teng H, et al. (2005) Sortilin controls intracellular sorting of brain-derived neurotrophic factor to the regulated secretory pathway. Journal of Neuroscience 25: 6156–6166. Egan MF, Kojima M, Callicott JH, et al. (2003) The BDNF val66met polymorphism affects activity-dependent secretion of BDNF and human memory and hippocampal function. Cell 112: 257–269. Figurov A, Pozzo-Miller L, Olafsson P, Wang T, and Lu B (1996) Regulation of synaptic responses to high-frequency stimulation and LTP by neurotrophins in the hippocampus. Nature 381: 706–709. Lee R, Kermani P, Teng KK, and Hempstead BL (2001) Regulation of cell survival by secreted proneurotrophins. Science 294: 1945–1948. Lou H, Kim SK, Zaitsev E, et al. (2005) Sorting and activitydependent secretion of BDNF require interaction of a specific motif with the sorting receptor carboxypeptidase E. Neuron 45: 245–255. Lu B (2003) BDNF and activity-dependent synaptic modulation. Learning and Memory 10: 86–98. Lu B, Pang PT, and Woo NH (2005) The yin and yang of neurotrophin action. Nature Reviews Neuroscience 6: 603–614. Nagappan G and Lu B (2005) Activity-dependent modulation of the BDNF receptor TrkB: Mechanisms and implications. Trends in Neuroscience 28: 464–471. Pang PT, Teng HK, Zaitsev E, et al. (2004) Cleavage of proBDNF by tPA/plasmin is essential for long-term hippocampal plasticity. Science 306: 487–491. Patterson SL, Abel T, Deuel TA, et al. (1996) Recombinant BDNF rescues deficits in basal synaptic transmission and hippocampal LTP in BDNF knockout mice. Neuron 16: 1137–1145. Poo MM (2001) Neurotrophins as synaptic modulators. Nature Reviews Neuroscience 2: 24–32. Sun YE and Wu H (2006) The ups and downs of BDNF in Rett syndrome. Neuron 49: 321–323. Woo NH, Teng HK, Siao CJ, et al. (2005) Activation of p75NTR by proBDNF facilitates hippocampal long-term depression. Nature Neuroscience 8: 1069–1077.

GFL Neurotrophic Factors: Physiology and Pharmacology M Saarma, University of Helsinki, Helsinki, Finland ã 2009 Elsevier Ltd. All rights reserved.

Introduction Growth factors are secretory proteins that bind to their cognate receptors at the plasma membrane of the cells and activate receptors that in turn trigger the activation of cellular biochemical signaling processes and lead to the cell proliferation, differentiation, migration, morphological changes, or induce the death of cells. Currently, about 100 growth factors that affect neurons have been described, but only those that mainly regulate the development and physiology of neurons are called neurotrophic factors. Neurotrophic factors are growth factors that regulate the number of neurons in a given population, neurite branching and synaptogenesis, adult synaptic plasticity, and maturation of neuronal phenotype. Neurotrophic factors include, for example, neurotrophins (NGF, BDNF, NT-3, NT-4), neurokines (CNTF, LIF, IL-6, CT-1, etc.), and glial cell line-derived neurotrophic factor (GDNF) family ligands (GFLs).

GFLs: Structure, Secretion, and Processing GDNF and the related factors artemin (ARTN), neurturin (NRTN), and persephin (PSPN) form the GDNF family of neurotrophic factors (Figure 1(a)). GFLs are synthesized by the cells as preproGFL proteins that are in the endoplasmic reticulum (ER) processed to proGFLs and secreted into the extracelluar space as mature GFLs or as proGFLs. The 211 amino-acidlong preproGDNF protein is finally processed into 134 amino acid mature GDNF monomer (Figure 1(b)) that is inactive and must homodimerize to generate a biologically active factor. All GFLs have seven conserved cysteines with similar relative spacing. Six of these cysteines are involved in the intramolecular S–S bonding whereas the seventh cysteine forms an intermolecular S–S bond stabilizing the homodimers of GFLs. Based on the conserved location of the seven cysteines, GFLs form a distant subfamily in the large transforming growth factor-beta (TGF-b) superfamily of growth factors. Orthologs of all four GFLs are found in all mammals, and also in teleost fishes. Interestingly, chicken genome lacks PSPN gene, but has a functional PSPN coreceptor GDNF family receptor a4 (GFRa4). Clawed frog Xenopus tropicalis genome surprisingly lacks the ortholog of NRTN.

Even more surprisingly insects have a receptor tyrosine kinase (RTK) rearranged during transfection (RET), a putative GFL coreceptor homolog but no obvious GFL homologs. The atomic structure of the major part of GDNF has been determined and GDNF together with all other TGF-b superfamily growth factors, including all GFLs, belong to the cystineknot family of proteins. The main structural elements of the GDNF molecule are helical regions and two fingers. Fingers 1 and 2 are involved in the binding of GDNF to its receptor GFRa1 (Figure 1(c)). In the GDNF structure the first 38 residues are disordered. Sequence analysis and experimental data indicate that this region binds heparin stabilizing the structure of GDNF and mediating GFL interactions with extracellular matrix (ECM). Recently, two laboratories have solved the crystal structure of ARTN. Although the overall fold of the ARTN covalent homodimer is similar to GDNF, the shape and the flexibility of elongated homodimer differs. ARTN also differs from GDNF in its overall charge and electrostatic distribution. Arginines in ARTN at positions 48, 49, and 51, as well as its amino terminus play a role in heparin binding. Although the three-dimensional (3-D) structures of NRTN and PSPN are not known, it is likely that their general spatial structure is very similar to GDNF and ARTN (Figure 1(c)). There are also heparin-binding regions in NRTN, but not in PSPN. The detailed mechanism of the secretion, processing, and activation of GFLs is poorly studied. Several cells and tissues can also secrete proGDNF in addition to mature GDNF (Figure 1(b)), suggesting that processing of proGDNF (and possibly all GFLs) can occur also in the ECM. Whether proGFLs can bind GFL receptors and trigger cellular signaling is currently not known.

GFL Receptors GFL receptor system is unique in several ways. First, GFLs can signal through three different receptor systems. Second, all three signaling receptors, RET, neural cell adhesion molecule (NCAM), and syndecans, are shared by four ligands, with the exception of PSPN that is not binding to syndecans. Third, the main GFL signaling receptor is the RTK RET, which is the only known RTK that does not bind the ligands directly but via coreceptors of GFRa family. Finally, RET activity is strictly regulated by Ca2þ ions, and in the intracellular part of RET in addition to tyrosine residues serines are also phosphorylated. The RET gene was identified in 1985 as a novel oncogene. Normal RET is the transmembrane RTK.

599

600 GFL Neurotrophic Factors: Physiology and Pharmacology

Figure 1 (a) Schematic structure of GFL showing the relative lengths (and number of amino acids) of the mature GDNF (blue), NRTN (pink), ARTN (brown), and PSPN (yellow) molecules and their pre-(brown) and pro-(light green) domains, as well as relative positions of the seven conserved cysteine residues (black vertical lines). GDNF has two putative N-glycosylation sites (N). (b) The preregion of GFLs is cleaved off inside the cells during secretion. Cells can secrete mature GDNF and proGDNF, indicating that the processing of proGDNF can also occur in the ECM. (c) The 3-D structure of dimeric GDNF shows that it belongs to the cystine-knot family of proteins. The main structural elements of the GDNF molecule are helical regions and fingers 1 and 2 that are involved in its binding to GFRa1 coreceptor. The 3-D structure of ARTN is similar to that of GDNF. The 3-D structures of NRTN and PSPN are not known, but the modeling of NRTN shows that its spatial structure is very similar to GDNF. The binding of heparan sulfate to GDNF is shown (left). NRTN can also bind heparan sulfate. Putative binding site is shown (green).

In human cancer, especially in papillary thyroid carcinoma, the RET tyrosine kinase domain is fused with N-terminal regions of the variety of unrelated genes. In addition to these somatic rearrangements germ line mutations in RET are responsible for two inherited disorders: activating mutations cause multiple endocrine neoplasia 2 (MEN2) and inactivating mutations Hirschsprung’s disease (HSCR) or congenital central hypoventilation syndrome. The alternative splicing of the human RET gene gives rise to at least three splice isoforms. The larger of these encodes a protein of 1114 amino acids that contains 51 amino acids at the C-terminus (RET51) that are replaced by nine amino acids in the RET9 isoforms. RET51 has two additional tyrosines: Tyr1090 and 1096. Orthologs of RET have been identified in higher and lower vertebrates, as well as in Drosophila (D-RET).

The tyrosine kinase domain of D-RET is conserved as it is functional in vitro and activates similar intracellular signaling pathways as mammalian RET. The extracellular domain of D-RET is less conserved and fails to interact with mammalian GDNF/GFRa1 complex. Interestingly, the Drosophila genome does not encode putative GFLs, but has a GFRa-like gene, called munin. RET is the only RTK among the receptors of TGF-b superfamily factors, all others being the receptor serine–threonine kinases. Although RET is similar to other RTKs, it also has several special features. First, the extracellular domain of RET consists of four cadherin-like domains (CLD1–4) and a single cysteine-rich domain (CRD). RET binds one molecule of Ca2þ that is required for its correct folding and function as the GFL receptor. Second, RET is the only

GFL Neurotrophic Factors: Physiology and Pharmacology 601 RET CLD1 GDNF

NRTN

CLD2

ARTN

PSPN

Ca2+ CLD3 CYSrich domains GFRa1

CLD4

GFRa2

GFRa3

GFRa4

CYS-rich domain

Signaling

Figure 2 GDNF-family ligands and receptor interactions. Homodimeric GFLs activate RET RTK by first binding to their cognate GFRa coreceptors. GDNF binds preferentially to GFRa1, NRTN to GFRa2, ARTN to GFRa3, and PSPN to GFRa4. Solid arrows indicate the preferred functional ligand–coreceptor interactions, whereas dotted arrows indicate putative cross-talk. GFRa proteins are attached to the plasma membrane through a GPI-anchor and consist of three (GFRa4 has only two) globular cysteine-rich domains joined together by adapter sequences. The extracellular domain of RET, that consists of CLD1–4 and one CRD interacts with all four GFL–GFRa complexes. Binding of Ca2þ ions to one CLD of RET is required for its activation by GFLs. RET intracellular part contains tyrosine kinase domain and tyrosine residues that become phosphorylated upon RET activation.

RTK that does not directly bind the ligands, that is, GFLs (Figure 2). Instead, RET is activated by binding to a complex formed by a GFL bound to its cognate glycosylphosphatidylinositol (GPI)-anchored coreceptor GFRa. Thus, GFLs interact with RET, but only in the presence of specific GFRa coreceptors. Four different GFL-GFRa pairs exist in mammals: GDNF-GFRa1, NRTN-GFRa2, ARTN-GFRa3, and PSPN-GFRa4 (Figure 2). With significantly lower-affinity GDNF can interact with GFRa2 and GFRa3, and NRTN and ARTN with GFRa1, but the physiological significance of these interactions is unclear. The GFL and GFRa homodimeric complex brings two RET molecules together, triggering transphosphorylation of specific tyrosine residues in their tyrosine kinase domains and activating intracellular signaling (Figure 3). GFRa Coreceptors

The GFL binding specificity to RET is determined by GFRa proteins that have unique binding affinities for each GFL. GFRas are cysteine-rich GPI-anchored proteins. Mammalian GFRa1–3 and most of the vertebrate GFRa consist of three CRDs, whereas

mammalian PSPN receptor GFRa4 lacks CRD1 and thus has only two CRDs (Figure 2). The structure of GFRa1 domain 3 revealed a novel protein fold: an all-a five-helix bundle stabilized by five disulfide bridges. The model for domain 2 was constructed, using its homology to domain 3. Recently, the crystal structure of ARTN in the complex with domains 2 and 3 of the GFRa3 was reported. These data reveal crucial amino acids in domain 2 of the GFRa3 receptor involved in ARTN binding and indicate that domains 2 and 3 define the potential RET binding site. Structural data on the ARTN-GFRa3 complex, on GFRa1 domain 3, and the conserved location of cysteine residues in GFRa1–4 CRDs predict a very similar general structural fold for all GFRas. Phylogenetic analysis indicates that all vertebrate classes from teleost fishes to mammals have orthologs for all four GFRa receptors. For several neuronal populations the full biological activity of GDNF is exerted only in the presence of TGF-b. It appears that TGF-b is required for the action of GDNF via GFRa1, but not for the action of NRTN, as TGF-b is increasing the surface location of GFRa1 but not of GFRa2.

602 GFL Neurotrophic Factors: Physiology and Pharmacology

a

b

c

d

Figure 3 GDNF interacts with coreceptor GFRa1 and activates RET. (a) Biologically active GDNF is a homodimer. In the absence of GDNF, GFRa1 and RET interaction is very weak. (b) A dimer of GDNF binds to the second CRD of GFRa1. (c) GDNF brings together two molecules of GFRa1. (d) GDNF-GFRa1 complex induces dimerization of two molecules of RET leading to transphosphorylation and activation of their tyrosine kinase domains.

Two proteins that are structurally similar to GFRa proteins – GAS1 and GFRAL – are described, but their relatedness to the GFL system is not yet clear. NCAM

NCAM plays an important role in neural development as the homophilic cell adhesion molecule. P140NCAM can also function as the alternative signaling receptor for all four GDNF family ligands. GFLs can directly bind to NCAM with low affinity, but do not trigger intracellular signaling. In the presence of cognate GFRa coreceptors GFLs can bind to p140NCAM with higher affinity and activate Src-like kinase Fyn and focal adhesion kinase FAK (Figure 4(a)). GFRa1 interaction with NCAM even in the absence of GDNF can inhibit NCAM-mediated cell adhesion. Thus GDNF–GFRa1– NCAM and GFRa1–NCAM interactions may have different biological readouts and physiological consequences. Current experiments indicate that GDNF by binding to GFRa1 and activating NCAM stimulates Schwann cell migration, hippocampal neurite outgrowth, and development of olfactory neurons in RET-independent manner. Further experiments are needed to unravel the in vivo roles of GDNF–GFRa1– NCAM and GFRa1–NCAM signaling and possible cross-talk of RET and NCAM in GFL signaling. Syndecans

Although the vast majority of GFL signaling utilizes GFRa/RET or GFRa/NCAM receptor systems, novel data indicate that additional GFL receptors exist. GDNF was originally purified by heparin-affinity chromatography, and later its interaction with heparan sulfates was documented. Recent experimental data indicate that heparan sulfates are required for GDNF signaling via GFRa1 and RET. GDNF,

ARTN, and NRTN, but not PSPN, can with high affinity bind to heparan sulfate proteoglycan (HSPG) syndecans (Figure 4(b)). Syndecans that bind GDNF can function as coreceptors and deliver GDNF to GFRa1/RET complex or activate directly Src signaling pathway triggering cell spreading and inducing neurite outgrowth of hippocampal neurons. GDNF interaction with syndecan also stimulates the migration of cortical neurons. Thus, GDNF can signal in two different ways – as a soluble protein and as a matrix-bound protein. In a soluble form it binds to GFRa1 and activates intracellular signaling via RET or NCAM. When bound to ECM (both transmembrane and soluble HSPGs), GDNF triggers qualitatively different signaling that is currently poorly defined.

Signaling Pathways The interactions of the intracellular part of RET with various signaling molecules are mainly triggered by the cognate ligands (GFLs), but in some cases also GFL-independently. GFL–GFRa homodimeric complex brings two molecules of RET together, triggering transphosphorylation of specific tyrosine residues of RET, followed by the activation of intracellular signaling cascades (Figures 3 and 5). Several cascades can be activated, which regulate cell survival, differentiation, proliferation, migration, chemotaxis, branching morphogenesis, neurite outgrowth, synaptic plasticity, etc. The mitogen activated protein (MAP) kinase pathway is involved in neurite outgrowth of the neurons. The phosphoinositide 3-kinase (PI3K)-Akt pathway is responsible for both neuronal survival and neurite outgrowth. GFL signaling also activates Src-family kinases, which elicit, for example, neurite outgrowth, neuronal survival, and kidney ureteric branching. In most cases tyrosine residues

GFL Neurotrophic Factors: Physiology and Pharmacology 603 NCAM

NCAM

IgG-like domain GFRa1

Fibronectinlike domain

GDNF

FYN

FYN Signaling

a Syndecan

Signaling b Figure 4 RET-independent signaling by GFLs. (a) p140NCAM is an alternative signaling receptor for GFLs. It interacts with a GDNF– GFRa1 dimer leading to the activation of Fyn, a Src-like kinase. It is unclear, whether GDNF triggers NCAM dimerization. NCAM has five IgG-like domains and two fibronectin-like domains. (b) HSPG syndecans carry heparan sulfate side chains that bind with high affinity GDNF, ARTN, and NRTN. Unlike GFRa1, syndecans can bind many GFL molecules simultaneously. GDNF binding to syndecans can activate Src-family kinases or modulate signaling through GFRa-RET.

Tyr905, Tyr981, Tyr1015, Tyr1062, and Tyr1096 of RET are phosphorylated and docking proteins and intracellular proteins are bound and activated through interaction with these and possibly other RET phosphotyrosines (Figure 5). Interestingly, RET activation affects different downstream targets inside and outside lipid rafts that are the dynamic assemblies of cholesterol and sphingolipids scattered within the disordered phase of the lipid bilayer. Differences in GDNF

signaling through RET within and outside the rafts could lead to significantly different cellular responses. RET is also activated by several GFL-independent pathways. First, increase in the intracellular concentration of cyclic AMP level triggers protein kinase A (PKA)-dependent Ser696 phosphorylation of RET independently of GFL that regulates the migration of enteric neural crest cells in the developing gut. Second, in the mature sympathetic neurons, RET51

604 GFL Neurotrophic Factors: Physiology and Pharmacology

RET intracellular domain

S696 Y752

JNK

STAT3 Grb7/10

Y905

Src

Y981

Grb2 AKT ERK

Y928 Y1015

PLCg

PKC

Y1062

Shc FRS2

Y1096 Y1096

Rac

JNK

Grb2 Gab1

P13K

AKT

Grb2 SOS

RAS

ERK

Nck JNK RasGAP ERK Dok4/5/6 IRS1/2 Enigma p38MAPK ERK5 Dok1

Figure 5 GDNF-induced signaling pathways. The model shows the intracellular domain of RET and highlights the phosphorylated tyrosines and different intracellular RET-binding proteins and activated downstream signaling pathways. The phosphorylated tyrosine residues (Y752, Y905, Y928, Y981, Y1015, Y1062, Y1096) activate multiple signaling pathways. Serine 696 (orange) is phosphorylated GFL-independently by PKA, and Y1096 (orange) can be phosphorylated GFL-independently by NGF.

isoform becomes phosphorylated and functionally activated by nerve growth factor that signals via unrelated receptor TrkA. The physiological importance of this signaling is not yet known. Third, there is in vitro evidence that RET is also a dependence receptor that triggers apoptotic cell death when not bound by the ligands. However, the in vivo relevance of this RET-triggered apoptosis is completely unclear.

Biology and Physiology Mice that lack GDNF, GFRa1, or RET die soon after birth, whereas mice lacking other GFLs or GFRa coreceptors or alternative receptors NCAM or syndecans are viable and fertile. The neuronal phenotypes of the different GFLs and their receptor-knockout mice are summarized in Table 1. The strongly overlapping phenotypes of ligand and coreceptor knockouts demonstrate, under physiological conditions, a specific pairing of each GFL and corresponding GFRas. The vast majority of cells and tissues that are affected in GFL- or GFRa-knockout mice also express RET, indicating that in vivo this is the main signaling receptor for GFLs. GDNF-, GFRa1-,

or RET-deficient mice share a phenotype of kidney agenesis, and an absence of many parasympathetic and enteric neurons. Mice that lack NRTN or GFRa2 have similar deficits in enteric and parasympathetic innervation. Characterization of ARTN- and GFRa3-deficient mice revealed similar abnormalities in the migration and axonal projection pattern of the entire sympathetic nervous system. Ablation of GFRa4 or its ligand PSPN impairs thyroid calcitonin production in young mice. Both GDNF- and NCAM-deficient mice have impaired migration of the rostral migratory stream-derived neuronal precursors. Likewise, GDNF- and syndecan-deficient mice have an impaired migration of GABAergic cortical neurons. In some cases the ligand- and its receptor-deficient mice phenotypes differ. Mice lacking the long RET51 isoform seem to be normal, whereas mice lacking the short RET9 isoform were similar to mice lacking all RET isoforms. Only the short isoform can rescue the phenotype of the RET-null mutation in the kidney and enteric nervous system. In another study, however, homozygous RET9 and RET51 mice were viable and show normally developed kidneys. Knock-in mice with mutated Tyr1062 of RET had defects in the enteric nervous system similar to those

GFL Neurotrophic Factors: Physiology and Pharmacology 605 Table 1 Phenotypes of mice lacking GFLs and their receptors Gene knockout

RET

GDNF/GFRa1

NRTN/GFRa2

ARTN/GFRa3

PSPN/GFRa4

Gross phenotype

P0 lethal

P0 lethal

Viable, fertile. Pseudoptosis growth retardationb

Viable, fertile ptosis

Viable, fertile Gfra4 /

Breathing defect

10 mM at the A2B and A3 ARs. A2A AR antagonists, such as the xanthine KW6002 and the nonxanthines SCH58261, SCH442416, VER6947, and VER7835, are of interest for use in treating Parkinson’s disease.

There is a marked species dependence of antagonist affinity at the A3 AR. Therefore, commonly used antagonists must be treated with caution in species other than humans. In general, one must be cognizant of potential species differences for both AR agonists and antagonists. Radioligands

Radioligands commonly used for the ARs are A1 agonist [3H]CCPA, antagonist [3H]DPCPX, A2A agonist [3H]CGS21680, antagonist [3H]ZM241,385 or [3H]SCH58261, A3 agonist [125I]I-AB-MECA, and antagonist [3H]PSB-11. Ligands for in vivo positron emission tomographic (PET) imaging of A1 and A2A ARs have been developed. For example, the xanthine [18F]CPFPX and the nonxanthine [11C]FR194921 have been developed as centrally active PET tracers

634 Adenosine

for imaging the A1 AR in the brain. Potent fluorescent ligands have been reported for A1, A2A, and A3 ARs.

under consideration for disease treatment. Such modulators, either positive enhancers or negative allosteric inhibitors, might have advantages over the directly acting (orthosteric) receptor ligands. The action of the allosteric compounds would depend on the presence of a high local concentration of

Allosteric Modulation

In addition to directly acting AR agonists and antagonists, allosteric modulators of agonist action are

A1 antagonists O

H

CH3(CH2)2 N

N

R=

HO

R N

N

O

HN N

N

(CH2)2CH3

N

KW3902 0.72

(CH2)2

CO2H

WRC-0571 1.7

O

O

H

CH3(CH2)2 N

N N

N

(CH2)3F

N

CPFPX 4.4

NH2

CH3 N

N

X

N

N HO

(CH2)2NH

N

N

O

N

FR194921 2.9

A2A antagonists O

CH3

N

N N

O

BG 9928(BIO-9002) 29

CH(CH3)2

CH3

N-0861 700

BG 9719 (CVT-124) 0.43

CH3(CH2)n

N

N

CH3

O

CH3

N

N

N

N

DPCPX 3.9

HN

CH3(CH2)n

N

O

N

N

ZM241, 385 1.6

X = 3,4-(OMe)2, n = 1, KW6002 39 X = 3-Cl, n = 2, CSC 54 (r)

RO

NH2 N

O N

N H2N

a Figure 3 Continued

N

N O

R = phenyl, VER 6947 1.1 R = 2-thienyl, VER 7835 1.7

C NHCH2-R

N

N

O

N

(CH2)n N N

R = H, n = 2, SCH 58261 5.0 R = CH3, n = 3, SCH 442416 0.048

Adenosine 635 A2B antagonists O CH3(CH2)2 N

H3CCONH O

N

CONH

N

NH

N N

O

(CH2)3

H

N

N

R

N H

N

(CH2)2CH3 R = CN, MRS1754 2.0

OSI P 339391 0.5

R = COCH3, MRS1706 1.4 O

H

CH3(CH2)2 N

N

N

CF3

N-CH2

H3C

N

N

N

N

N

O

N

N

O

O

H

CH3(CH2)2 N

O

O

CO

NH

(CH2)2CH3

(CH2)2CH3

O

CVT-6883 8.3

N SO3H N

N H

A3 antagonists

O MRE 2029-F20 3.0

H

CH3(CH2)2 N O

PSB-1115 53

I

N N

NHCH2 N

N

C O

N

N

CH3CH2

O HN

N

O

C

O

O HO

O

H

O

CH3

H

R

N

H R = H, MRS1191 31

OH

MRS1292 29

VUF5574 4.0 CH3CH2

F3C

N

HN

O

H N

N

N O

N

NH

N

N

MRS3777 47 CH3(CH2)3

N

N

N CH3

N

N

N

CH3O

R = NO2,MRS1334 2.7

N

O

N

I

PSB-11 3.5

N H

NHCONH N

N

O

CH3CH2

N

N

O

CH3CH2S

N

NHCH2

CH3(CH2)2 O

N

N CH3(CH2)2

N

OCH3

OT-7999 0.61

O(CH2)2CH3

S

N HO

MRE 3008-F20 0.82

N

N

MRS 1523 19

OH

LJ1251 4.2

b

Figure 3 Structures of selected (a) A1 and A2A and (b) A2B and A3 adenosine receptor antagonist probes used as pharmacological tools, and in some cases as clinical candidates. The Ki values (nM) in binding to the appropriate human adenosine receptor are indicated after the name (all are human adenosine receptors, unless indicated ‘R’ for rat).

adenosine, which often occurs in response to a pathological condition. In some cases (dependent on tissue, receptor subtype, and other conditions), one would wish to boost the adenosine effect, and therefore an allosteric enhancer would be useful; in other

cases, the elevated adenosine may be detrimental, in which case one would want to apply a negative modulator. Positive allosteric modulators have been explored for the A1 (benzoylthiophene derivatives) and A3 (imidazoquinoline derivatives) AR subtypes.

636 Adenosine

Role of ARs in Autonomic Nervous System Disorders Distribution

ARs are widely distributed in the autonomic and enteric nervous systems. Distribution of neural ARs in the human intestine has been investigated. Messenger RNAs (mRNAs) of subtypes of AR are differentially expressed in neural and nonneural layers of the jejunum, ileum, colon, and cecum. The A1 AR is expressed in jejunal myenteric neurons and colonic submucosal neurons. The A2A AR is also found in other neurons, but A2B AR immunoreactivity is more prominent than that of the A2A AR in myenteric neurons, nerve fibers, and glia. The A3 AR largely occurs in substance P-positive jejunal submucosal neurons and less in vasoactive intestinal peptide (VIP) neurons. The AR that mediates the relaxation of bladder strips induced by AR agonists such as CGS21680 and NECA has been classified as A2A; however, the highest mRNA levels are found for A2B transcripts in the bladder. However, reports of tissue distribution based on mRNA levels might not correspond to protein levels. Thus, quantitative determination with a potent and selective A2B AR radioligand may be necessary. Western blot analysis shows that all four ARs are expressed in the uroepithelium. A1 ARs are prominently localized to the apical membrane of the umbrella cell layer, whereas A2A, A2B, and A3 ARs are localized intracellularly or on the basolateral membrane of umbrella cells and the plasma membrane of the underlying cell layers. Various subtypes of ARs have also been detected in the heart, blood vessels, kidney, and other organs throughout the body. A1, A2A, and A2B ARs are all expressed on normal human airway smooth muscle cells, and both A1 and A2A ARs are expressed in vagal pulmonary C fibers. A1 and A2A ARs are highly expressed in gastric mucosa. In the brain, the A1 AR is widely expressed in almost all areas. Pre- and postsynaptic activation of the A1 AR inhibits synaptic transmission, in part by suppressing the release of excitatory transmitters. The A2A AR is less widely expressed, with greatest density in the striatum, nucleus accumbens, and olfactory tubercle. The A2B and A3 ARs are expressed at low density in most brain regions, and are implicated in purinergic signaling in neuronal–glial interactions.

Functions of ARs in the Autonomic and Enteric Systems The enteric nervous system contains several hundred million neurons located in the myenteric plexuses

between muscle coats and submucous plexus. ARs in the enteric nervous system are critical for the control of motor and secretomotor functions. Adenosine is known to suppress intestinal motility by activating putative neural A1 ARs in the small intestine. A2A and A2B ARs in the myenteric neurons were also suggested to contribute to effects of adenosine on motility. ARs in circular muscle may contribute to the postjunctional actions of adenosine on motility. Adenosine directly modulates intestinal tone in the rat by causing relaxation via A2B ARs or contraction via A1 ARs in longitudinal muscle cells. Electrophysiological studies in rodents provided evidence for pre- and postsynaptic A1 AR-mediated inhibition of slow synaptic transmission and presynaptic inhibition of fast synaptic transmission. A2B AR gene expression products are widely expressed in the mucosa of the human intestinal tract, where they are postulated to be involved in the pathophysiology of diarrheal diseases. Both A2A and A3 ARs have been found in mucosal tissues, suggesting that they influence secretion and/or absorption in the human intestine. The ARs are involved in neuroplastic changes occurring in inflamed gut. The A2A AR modulates the activity of colonic excitatory cholinergic nerves via facilitatory control on inhibitory nitrergic pathways, and such a regulatory function is enhanced in the presence of bowel inflammation. Activation of the A2A AR inhibits stress-induced gastric inflammation and damage. Thus, selective A2A AR agonists may be useful for preventing ulcers and gastric inflammation. Both A1 and A2A ARs are involved in gastrin release. The modification of AR expression by changes in intraluminal acidity may represent a novel regulatory feedback mechanism to control gastric acid secretion. The discharge of urine from bladder is controlled by the nervous system. The causes of bladder dysfunction related to the nervous system include multiple sclerosis, traumatic or developmental brain or spinal cord injury, or Parkinson’s disease. Although it is generally agreed that ACh acting on smooth muscle muscarinic receptors is the primary neurologic mechanism controlling bladder emptying, neural stimulation of the bladder is only partially inhibited in many cases by the muscarinic receptor antagonist atropine. The atropine-resistant component of parasympathetic contraction was later found to be ATP sensitive. There is plenty of evidence to suggest that ARs also play an important role in bladder function. It has been shown that adenosine-evoked membrane hyperpolarization and relaxation of bladder smooth muscle is mediated by A2A AR-mediated activation of KATP channels via adenylate cyclase and elevation

Adenosine 637

of cAMP. High mRNA levels of A2B transcript are also found in the bladder. Adenosine was found to be released from the uroepithelium, which was potentiated tenfold by stretching the tissue. It is generally accepted that the sensation of bladder fullness is relayed through the mucosal layer by afferent nerves, which are activated by the release of neurotransmitters such as ATP from the urothelium when it is stretched as the bladder fills. It is suggested that adenosine reduces the force of nerve-mediated contractions by acting predominantly at the presynaptic sites at the nerve muscle junction via the A1 AR. The autonomic nervous system, a complex and self-organized entity, plays a key role in regulating cardiovascular function. In the heart, a number of intrinsic nerves in the atrial and intra-atrial septum have been shown to release ATP, ACh, 5-hydroxytryptamine, and other neurotransmitters. Adenosine may regulate both the release and the interaction with their receptors of these neurotransmitters. Adenosinemediated myocardial protection has been suggested to be through a neurogenic pathway. The infarctreducing effect of intravenous adenosine in intact rats was blocked with both the ganglionic blocker

hexamethonium and the nitric oxide synthase inhibitor No-nitro-L-arginine. The A1 AR was determined to be the primary subtype involved in the modulation of norepinephrine release from cardiac nerve terminals using isolated rat hearts. Vascular tone in blood vessels is controlled by perivascular nerves and the endothelial cells. Adenosine acts directly on smooth muscles as well as modulates the release of neurotransmitters, such as ACh and ATP. A study of neurotransmitter release following electrical depolarization of nerve endings from the rat mesenteric artery suggests that activation of the presynaptic A2A AR and A3 AR modulates neurotransmission by inhibiting the release of norepinephrine but not neuropeptide Y. The A1 ARs, not P2 receptors, inhibit prejunctionally sympathetic neurotransmission in the hamster mesenteric arterial bed. Activation of both A1 AR and A2A AR is required to attenuate neurogenic coronary constriction due to sympathetic stimulation. In the respiratory system, adenosine has been shown to stimulate vagal pulmonary C fiber terminals through activation of the A1 AR, and subsequently cause bronchoconstriction, which was significantly attenuated by A1 AR antagonists. Recent evidence suggests

Table 1 Exploration of the role of adenosine receptors in disorders of the nervous system Condition/system

Model a

Subtypes implicated

Relationship

Aggression

A1 KO A2A KO b-Amyloid A2A KO Various Swim A3 KO Various 3-NP 3-NP A1 KO Ischemia Ischemia Ischemia A1 KO A2A KO A3 KO Constriction injury Formalin A1 KO A2A KO 6-OHDA MPTP Haloperidol ADA KO mice A2A KO

A1 A2A A2A A2A A1 A2A A3 A1 A1 A2A A1 A1 A2A A3 A1 A2A A3 A1 A2A A1 A2A A2A A2A A2A A2B A2A

Increased aggressiveness Increased aggressiveness Reduced neurotoxicity by antagonists Increased anxiety Agonist protects Antagonist improves Increased despair-like behavior Agonist protects in some models Agonist protects striatal damage Antagonist protects striatal damage No change Agonist protects Antagonist protects Chronic agonist protects No effect in adults, benefits newborns Detrimental in newborns, benefits adults Detrimental in adults Agonist protects Agonist protects Increased thermal nociception Lowers thermal nociception Antagonist protects Antagonist protects Decreased catalepsy by antagonists Antagonist protects Agonist lost effect

Alzheimers’s disease Anxiety Cardiac arrhythmias Depression Epilepsy Huntington’s disease Memory Neurodegeneration

Pain

Parkinson’s disease

Pulmonary inflammation Sleep disorders a

ADA, adenosine deaminase; KO, knockout; 6-OHDA, 6-hydroxydopamine; MPTP, 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine; 3-NP, 3-nitropropionic acid.

638 Adenosine

that both A1 AR and A2A AR are involved in the activation of vagal sensory C fibers in the lung. The A1 AR has also been suggested to be directly involved in the mobilization of calcium in human bronchial smooth muscle cells. The A2B AR is known to mediate mast cell degranulation in large animals and humans, and A2B AR antagonists are of potential clinical applications for asthma. A1 AR agonists reduce pain signaling in the spinal cord, where the receptors are highly expressed. In humans, infusion of adenosine in the spinal cord is effective in decreasing postoperative pain. Recent studies suggest that A1 ARs might be more important in chronic pain than in acute pain. The A1 AR agonists are being evaluated in phase II clinical trials for the treatment of pain and migraine. An A1 ARselective allosteric enhancer is also used in clinical trials as a treatment for neuropathic pain. Peripherally administered A2A AR agonists have an antinociceptive effect. In summary, in the peripheral nervous systems, ARs are related to gut inflammation, rheumatoid arthritis, ischemia, Crohn’s disease, constipation, bladder disorders, and a variety of other conditions. In the brain, ARs appear to regulate important functions in cerebral ischemia, dementia, Parkinson’s disease and other neurodegenerative diseases, pain, sleep disorders, anxiety, and schizophrenia (Table 1). Genetic Deletion of ARs

Deletion of each of the four subtypes has been carried out, and the resulting single-AR knockout (KO) mice are viable and not highly impaired in function. The pharmacological profile indicates that the analgesic effect of adenosine is mediated by the A1 AR, and analgesia is lost in mice in which the A1 AR has been genetically eliminated. Genetic KO of the A1 AR in mice removes the discriminative-stimulus effects, but not the arousal effect, of caffeine, and increases anxiety and hyperalgesia. Study of A2A AR KO mice reveals functional interaction between the spinal opioid receptors and peripheral ARs. A1 AR KO mice demonstrate a decreased thermal pain threshold, whereas A2A AR null mice demonstrate an increased threshold to noxious heat stimulation, supporting an A1 AR-mediated inhibitory and an A2A AR-mediated excitatory effect on pain transduction pathways. KO of the A2A AR eliminates the arousal effect of caffeine. Genetic KO of the A2A AR also suggests a link to increased anxiety and protects against damaging effects of ischemia and the striatal toxin 3nitropropionic acid. Genetic KO of the A3 AR leads to increased neuronal damage in a model of carbon monoxide-induced brain injury. Neutrophils lacking A3 ARs show correct directionality but diminished

speed of chemotaxis. Although studies on A2B ARs KO mice have been reported, the importance of A2B ARs in the brain still awaits future investigation. In conclusion, adenosine is a ubiquitous neuromodulator of many functions, by activating one or more of the widely distributed AR subtypes. Selective AR agonists and antagonists have been developed, some of which are in advance stages of clinical trials for therapeutic applications. The use of both knockout animals and selective drugs has contributed toward elucidation of the physiological role of adenosine and the signaling pathways involved.

Further Reading Burnstock G (2007) Physiology and pathophysiology of purinergic neurotransmission. Physiological Reviews 87: 659–797. Christofi FL, Zhang H, Yu JG, et al. (2001) Differential gene expression of adenosine A1, A2a, A2b, and A3 receptors in the human enteric nervous system. Journal of Comparative Neurology 439: 46–64. Costanzi S, Ivanov AA, Tikhonova IG, et al. (2007) Structure and function of G protein-coupled receptors studied using sequence analysis, molecular modelling, and receptor engineering: Adenosine receptors. In: Frontiers in Drug Design and Discovery, vol. 3. Hilversum, The Netherlands: Bentham Science Publishers, Inc. Fields RD and Burnstock G (2006) Purinergic signalling in neuronglia interactions. Nature Reviews Neuroscience 7: 423–436. Franco R, Casado V, Mallol J, et al. (2006) The two-state dimer receptor model: A general model for receptor dimers. Molecular Pharmacology 69: 1905–1912. Fredholm BB, IJzerman AP, Jacobson KA, et al. (2001) International Union of Pharmacology. XXV. Nomenclature and classification of adenosine receptors. Pharmacological Reviews 53: 527–552. Fredholm BB, Chen JF, Masino SA, et al. (2005) Actions of adenosine at its receptors in the CNS: Insights from knockouts and drugs. Annual Review of Pharmacology and Toxicology 45: 385–412. Jacobson KA and Gao ZG (2006) Adenosine receptors as therapeutic targets. Nature Reviews Drug Discovery 5: 247–264. Linden J (2005) Adenosine in tissue protection and tissue regeneration. Molecular Pharmacology 67: 1385–1387. McGaraughty S, Cowart M, Jarvis MF, et al. (2005) Anticonvulsant and antinociceptive actions of novel adenosine kinase inhibitors. Current Topics in Medicinal Chemistry 5: 43–58. Moro S, Gao ZG, Jacobson KA, et al. (2006) Progress in pursuit of therapeutic adenosine receptor antagonists. Medicinal Research Reviews 26: 131–159. Ruggieri MR Sr. (2006) Mechanisms of disease: role of purinergic signaling in the pathophysiology of bladder dysfunction. Nature Clinical Practice Urology 3: 206–215. Schulte G and Fredholm BB (2003) Signalling from adenosine receptors to mitogen-activated protein kinases. Cell Signaling 15: 813–827. Yan L, Burbiel JC, Maass A, et al. (2003) Adenosine receptor agonists: From basic medicinal chemistry to clinical development. Expert Opinion in Emerging Drugs 8: 537–576. Yu L, Huang ZL, Mariani J, et al. (2004) Selective inactivation or reconstitution of adenosine A2A receptors in bone marrow cells reveals their significant contribution to the development of ischemic brain injury. Nature Medicine 10: 1081–1087.

Adenosine Triphosphate (ATP) G Burnstock, Royal Free and University College School of Medicine, London, UK ã 2009 Elsevier Ltd. All rights reserved.

Early History Drury and Szent-Gyo¨rgyi, in 1929, were the first to demonstrate the potent extracellular actions of adenosine 50 -triphosphate (ATP) and adenosine on the heart and coronary blood vessels. In 1948, Emmelin and Feldberg demonstrated that intravenous injection of ATP into cats caused complex effects that affected both peripheral and central mechanisms. Injection of ATP into the lateral ventricle produced muscular weakness, ataxia, and a tendency of the cat to sleep. Application of ATP to various regions of the brain produced biochemical or electrophysiological changes. Holton presented in 1959 the first hint of a transmitter role for ATP in the nervous system by demonstrating the release of ATP during antidromic stimulation of sensory nerves supplying the ear artery. Buchthal and Folkow recognized a physiological role for ATP at the neuromuscular junction in 1948, finding that acetylcholine (ACh)-evoked contraction of skeletal muscle fibers was potentiated by exposure to ATP. Presynaptic modulation of ACh release from the neuromuscular junction by purines in the rat was also reported. The existence of nonadrenergic, noncholinergic (NANC) neurotransmission in the gut and bladder was established in the mid-1960s. Several years later, after many experiments, Burnstock and his colleagues published a study that suggested that the NANC transmitter in the guinea pig taenia coli and stomach, rabbit ileum, frog stomach, and turkey gizzard was ATP. The experimental evidence included mimicry of the NANC nerve-mediated response by ATP; measurement of release of ATP during stimulation of NANC nerves with luciferin-luciferase luminometry; histochemical labeling of subpopulations of neurons in the gut with quinacrine, a fluorescent dye known to selectively label high levels of ATP bound to peptides; the later demonstration that the slowly degradable analogue of ATP, a,b-methylene ATP (a,b-meATP), which produces selective desensitization of the ATP receptor, blocked the responses to NANC nerve stimulation. Soon after, evidence was presented for ATP as the neurotransmitter for NANC excitatory nerves in the urinary bladder. The term ‘purinergic’ was proposed in a short letter to Nature in 1971, and the evidence for purinergic transmission in a wide variety of systems was presented in Pharmacological Reviews in 1972 (Figure 1). This

concept met with considerable resistance for many years. This was partly because it was felt that ATP was established as an intracellular energy source involved in various metabolic cycles and that such a ubiquitous molecule was unlikely to be involved in extracellular signaling. However, ATP was one of the biological molecules to first appear, and therefore it is not really surprising that it should have been utilized for extracellular, as well as intracellular, purposes early in evolution. The fact that potent ectoATPases were described in most tissues in the early literature was also a strong indication for the extracellular actions of ATP. Purinergic neurotransmission is now generally accepted, and a volume of Seminars in Neuroscience was devoted to purinergic neurotransmission in 1996.

Purinergic Cotransmission Another concept that has had a significant influence on our understanding of purinergic transmission was that of cotransmission. Burnstock wrote a commentary in Neuroscience in 1976 titled, ‘‘Do some nerves release more than one transmitter?’’ This position challenged the single-neurotransmitter concept, which became known as ‘Dale’s Principle,’ even though Dale himself never defined it as such. The commentary was based on hints about cotransmission in the early literature describing both vertebrate and invertebrate neurotransmission and more specifically, with respect to purinergic cotransmission, on the surprising discovery in 1971 that ATP was released from sympathetic nerves supplying the taenia coli as well as from NANC inhibitory nerves. The excitatory junction potentials (EJPs) recorded in the vas deferens were blocked by a,b-meATP, a selective desensitizer of P2X receptors (Figures 2(a) and 2(b)). This finding clearly supported the earlier demonstration of sympathetic cotransmission in the vas deferens in the laboratory of Dave Westfall, following an earlier report of sympathetic cotransmission in the cat nictitating membrane. Purinergic cotransmission was later described in the rat tail artery and in the rabbit saphenous artery. Noradrenaline (NA) and ATP are now well established as cotransmitters in sympathetic nerves (see Figure 3 (a)), although the proportions vary in different tissues and species, during development and aging, and in different pathophysiological conditions. ACh and ATP are cotransmitters in parasympathetic nerves supplying the urinary bladder. Subpopulations of sensory nerves have been shown to utilize ATP in addition to substance P and calcitonin generelated peptide; it seems likely that ATP cooperates

639

640 Adenosine Triphosphate (ATP)

ATP

Large opaque vesicle containing ATP

ATP resynthesis Adenosine

P1 purinoceptor Adenosine uptake

ATP ADP

ATPase

AMP 5⬘-nucleo- tidase Adenosine A.deaminase P2 Inosine Purinoceptor

Circulation

Smooth muscle receptors

Figure 1 Purinergic neuromuscular transmission depicting the synthesis, storage, release, and inactivation of adenosine 50 -triphosphate (ATP). ATP, stored in vesicles in nerve varicosities, is released by exocytosis to act on postjunctional P2 purinoceptors on smooth muscle. ATP is broken down extracellularly by ATPases and 50 -nucleotidase to adenosine, which is taken up by varicosities to be resynthesized and restored in vesicles. Adenosine acts prejunctionally on P1 purinoceptors to modulate transmitter release. If adenosine is broken down further by adenosine deaminase to inosine, it is removed by the circulation. Adapted from Burnstock G (1972) Purinergic nerves. Pharmacological Reviews 24: 509–581, with permission from the American Society for Pharmacology and Experimental Therapeutics.

with these peptides in axon reflex activity. ATP, vasoactive intestinal polypeptide and nitric oxide (NO) are cotransmitters in NANC inhibitory nerves, but that they vary considerably in proportion in different regions of the gut. More recently, ATP has been shown to be a cotransmitter with NA, 5-hydroxytryptamine, glutamate, dopamine and g-aminobutyric acid (GABA) in the central nervous system (CNS) (see Figure 3(b)). ATP and NA act synergistically to release vasopressin and oxytocin, which is consistent with ATP cotransmission in the hypothalamus. ATP, in addition to glutamate, is involved in long-term potentiation in hippocampal CA1 neurons associated with learning and memory.

Receptors for Purines and Pyrimidines Implicit in the purinergic neurotransmission hypothesis was the presence of purinoceptors. A basis for distinguishing two types of purinoceptor, identified as P1 and P2 for adenosine and ATP/adenosine diphosphate (ADP), respectively, was recognized in 1978. This helped resolve some of the ambiguities in earlier reports, which were complicated by the breakdown

of ATP to adenosine by ectoenzymes so that some of the actions of ATP were directly on P2 receptors, whereas others were due to indirect action via P1 receptors. At about the same time, two subtypes of P1 (adenosine) receptor were recognized, but it was not until 1985 that a pharmacological basis for distinguishing two types of P2 receptors (P2X and P2Y) was proposed. A year later, two further P2 receptor subtypes were named, a P2T receptor selective for ADP on platelets and a P2Z receptor on macrophages. Further subtypes followed, perhaps the most important being the P2U receptor, which could recognize pyrimidines such as uridine 50 triphosphate (UTP) in addition to ATP. However, to provide a more manageable framework for newly identified nucleotide receptors, Abbracchio and Burnstock proposed in 1994 that purinoceptors should belong to two major families: a P2X family of ligand-gated ion channel receptors and a P2Y family of G-protein-coupled receptors. This was based on studies of transduction mechanisms and the cloning of nucleotide receptors: P2Y receptors were cloned first, in 1993, and a year later P2X receptors were cloned. This nomenclature has been widely adopted, and currently seven P2X subtypes and eight P2Y

Adenosine Triphosphate (ATP) 641

1s

2 g

a 3 ⫻ 10−8 mol l−1

Control

a,b -methylene ATP 10 mV 3 ⫻ 10−6 mol l−1 1s

b Figure 2 (a) EJPs in response to repetitive stimulation of adrenergic nerves (white dots) in the guinea pig vas deferens. The upper trace records the tension, the lower trace the electrical activity of the muscle recorded extracellularly by the sucrose gap method. Note both summation and facilitation of successive junction potentials. At a critical depolarization threshold, an action potential is initiated which results in contraction. (b) The effect of various concentrations of a,b-methylene ATP (a,b-meATP) on EJPs recorded from guinea pig vas deferens (intracellular recordings). The control responses to stimulation of the motor nerves at 0.5 Hz are shown on the left. After at least 10 min in the continuous presence of the indicated concentration of a,b-meATP, EJPs were recorded using the same stimulation parameters. (a) Reproduced from Burnstock G and Costa M (eds.) (1975) Adrenergic neuroeffector transmission. In Adrenergic Neurones: Their Organisation, Function and Development in the Peripheral Nervous System, pp. 51–106. London: Chapman and Hall, with permission from Springer Science and Business Media. (b) Reproduced from Sneddon P and Burnstock G (1984) Inhibition of excitatory junction potentials in guinea-pig vas deferens by a,bmethylene-ATP: Further evidence for ATP and noradrenaline as cotransmitters. European Journal of Pharmacology 100: 85–90, with permission from Elsevier.

receptor subtypes are recognized. Four subtypes of P1 receptor have been cloned and characterized. P2X receptors in general mediate fast neurotransmission but are sometimes located prejunctionally to mediate increase in release of cotransmitters, for example, glutamate in terminals of primary afferent neurons in the spinal cord. P2X3 and P2X2/3 receptors are prominent in sensory neurons and are involved in nociception. P2X7 receptors are involved in cell death. P2Y receptors are particularly involved in prejunctional inhibitory modulation of transmitter release, as well as long-term (trophic) events such as cell pro liferation. P2Y1 receptors are widespread in many regions of the brain, while the P2Y2 receptors have been

localized on pyramidal neurons in the hippocampus and prefrontal cortex, on supraoptic magnocellular neurosecretory neurons in the hypothalamus, and on neurons in the dorsal horn of the spinal cord. In addition, mRNA but not protein has been reported for P2Y4 and P2Y6 receptor subtypes in the cerebellum and hippocampus, while P2Y12 receptor mRNA has also been described in the cerebellum and P2Y13 in the cortex. In the periphery, P2Y1,2,4,6 receptors have been described on subpopulations of sympathetic neurons, P2Y2 and P2Y4 receptors in intracardiac ganglia, P2Y1 and P2Y2 receptors on sensory neurons (although P2Y4 and P2Y6 mRNA have also been reported) while P2Y1 receptors appear to be the dominant subtype on enteric neurons. P2Y1,2,4,6 functional receptors have been described on astrocytes in the CNS and also on microglia, where functional P2Y12 receptors have also been identified. P2Y1 and P2Y2 receptors have been located in Schwann cells and oligodendrocytes, where functional P2Y12 receptors also appear to be present. P2Y2 (and/or P2Y4) receptors are expressed on enteric glial cells. There is also emerging evidence for P2Y receptors on stem cells.

ATP Release and Degradation There is clear evidence for exocytotic vesicular release of ATP from nerves, and the concentration of nucleotides in vesicles is claimed to be up to 1000 mmol l 1. It was generally assumed that the main source of ATP acting on purinoceptors was damaged or dying cells. However, it is now recognized that ATP release from many cells is a physiological or pathophysiological response to mechanical stress, hypoxia, inflammation, and some agonists. There is debate, however, about the ATP transport mechanisms involved. There is compelling evidence for exocytotic release from endothelial and urothelial cells, osteoblasts, astrocytes, mast, and chromaffin cells, but other transport mechanisms have also been proposed, including ATP binding cassette transporters, connexin hemichannels, and plasmalemmal voltage-dependent anion channels. Much is now known about the ectonucleotidases that break down ATP released from neurons and nonneuronal cells. Several enzyme families are involved: ecto-nucleoside triphosphate diphosphohydrolases (E-NTPDases), of which NTPDase1, 2, 3, and 8 are extracellular; ectonucleotide pyrophosphatase of 3 subtypes; alkaline phosphatases; ecto-50 -nucleotidase; and ecto-nucleoside diphosphokinase. NTPDase1 hydrolyzes ATP directly to adenosine monophosphate (AMP) and UTP to uridine diphosphate (UDP), while NTPDase2 hydrolyzes ATP to ADP and 50 -nucleotidase AMP to adenosine.

642 Adenosine Triphosphate (ATP)

Cotransmission − sympathetic nerves

Sympathetic nerve

ATP NA

ATP

NA

P2X

Smooth muscle a1

IP3

EJP

Ca2+ influx

Ca2+ release

Contraction

a (i) Sympathetic

(ii) Parasympathetic

NA ATP NPY

ACh VIP ATP

(iii) Enteric (NANC inhibitory)

(iv)Sensory motor

ATP NO VIP

(v) CNS ACh+ATP: cortex/cuneate nucleus NA+ATP: locus coerulreus/hypothalamus GABA+ATP: hypothalamus/dorsal horn/retina Glutamate+ATP: hippocampus DA+ATP: mesolimbic systems

CGRP SP ATP

b Figure 3 (a) Cotransmission in sympathetic nerves. Adenosine 50 -triphosphate (ATP) and noradrenaline (NA) from terminal varicosities of sympathetic nerves can be released together. With NA acting via the postjunctional a1-adrenoceptor to release cytosolic Ca2+, and with ATP acting via P2X1-gated ion channels to elicit Ca2þ influx, both contribute to the subsequent response (contraction). IP3 is inositol triphosphate, EJP is excitatory junction potential. (b) Schematic diagram of the principal cotransmitters with ATP in the nervous system. Nerve terminal varicosities of (i) sympathetic, (ii) parasympathetic, (iii) enteric (NANC inhibitory), (iv) sensory-motor neurons, and (v) central nervous system (CNS). (a) Adapted from Kennedy C, McLaren GJ, Westfall TD, and Sneddon P (1996) ATP as a cotransmitter with noradrenaline in sympathetic nerves – Function and fate. In: Chadwick DJ and Goode J (eds.) P2 Purinoceptors: Localization, Function and Transduction Mechanisms, pp. 223–235. Chichester: John Wiley and Sons, with permission from John Wiley & Sons. (b) Reproduced from Burnstock G (2007) Physiology and pathophysiology of purinergic neurotransmission. Physiological Reviews 87: 659–797, with permission from The American Physiological Society.

Physiology of Purinergic Neurotransmission Purinergic signaling appears to be a primitive system that is involved in many nonneuronal and neuronal mechanisms, in both short-term and long-term (trophic) events, including exocrine and endocrine secretion, immune responses, inflammation, mechanosensory transduction, platelet aggregation, endothelial-mediated vasodilatation and in cell proliferation,

differentiation, migration, and death in development and regeneration. The first clear evidence for nerve–nerve purinergic synaptic transmission was published in 1992. Synaptic potentials in the coeliac ganglion and in the medial habenula in the brain were reversibly antagonized by the antitrypanosomal agent suramin. Since then, many articles have described either the distribution of various P2 receptor subtypes in the brain and

Adenosine Triphosphate (ATP) 643

spinal cord or electrophysiological studies of the effects of purines in brain slices, isolated neurons, and glial cells. Synaptic transmission has also been demonstrated in the myenteric plexus and in various sensory, sympathetic, parasympathetic, and pelvic ganglia. Adenosine produced by the ectoenzymatic breakdown of ATP acts through presynaptic P1 receptors to inhibit the release of excitatory neurotransmitters in both the peripheral and the central nervous systems. Purinergic signaling is also implicated in higher order cognitive functions, including learning and memory in the prefrontal cortex.

CNS Control of Autonomic Function Functional interactions seem likely to occur between purinergic and nitrergic neurotransmitter systems; these interactions might be important for the regulation of hormone secretion and body temperature at the hypothalamic level and for cardiovascular and respiratory control at the level of the brain stem. The nucleus tractus solitarius (NTS) is a major integrative center of the brain stem involved in reflex control of the cardiovascular system; stimulation of P2X receptors in the NTS evokes hypotension. P2X receptors expressed in neurons in the trigeminal mesencephalic nucleus might be involved in the processing of proprioceptive information.

Neuron–Glia Interactions ATP is an extracellular signaling molecule between neurons and glial cells. ATP released from astrocytes might be important in triggering cellular responses to trauma and ischemia by initiating and maintaining reactive astrogliosis, which involves striking changes in the proliferation and morphology of astrocytes and microglia. Some of the responses to ATP released during brain injury are neuroprotective, but at higher concentrations, ATP contributes to the pathophysiology initiated after trauma. Multiple P2X and P2Y receptor subtypes are expressed by astrocytes, oligodendrocytes, and microglia. ATP and basic fibroblast growth factor (bFGF) signals merge at the mitogenactivated protein kinase cascade, which underlies the synergistic interactions of ATP and bFGF in astrocytes. ATP can activate P2X7 receptors in astrocytes to release glutamate, GABA, and ATP, which regulate the excitability of neurons. Microglia, immune cells of the CNS, are also activated by purines and pyrimidines to release inflammatory cytokines such as interleukins 1b (IL-1b) and IL-6 and tumor necrosis factor a. Thus, although microglia

might play an important role against infection in the CNS, overstimulation of this immune reaction might accelerate the neuronal damage caused by ischemia, trauma, or neurodegenerative diseases. P2X4 receptors induced in spinal microglia gate tactile allodynia after nerve injury. P2X7 receptors mediate superoxide production in primary microglia and are upregulated in a transgenic mouse model of Alzheimer’s disease, particularly around b-amyloid plaques.

Purine Transmitter and Receptor Plasticity The purinergic neurotransmission field is expanding rapidly; there is increasing interest in the physiology and pathophysiology of this neurosignaling system, and therapeutic interventions are being explored. The autonomic nervous system shows marked plasticity: that is, the expression of cotransmitters and receptors shows dramatic changes during development and aging, in nerves that remain after trauma or surgery, and in disease conditions. There are several examples where the purinergic component of cotransmission is increased in pathological conditions. The parasympathetic purinergic nerve-mediated component of contraction of the human bladder is increased to 40% in pathophysiological conditions such as interstitial cystitis, outflow obstruction, idiopathic instability, and also some types of neurogenic bladder. ATP also has a significantly greater cotransmitter role in sympathetic nerves supplying hypertensive compared to normotensive blood vessels. Upregulation of P2X1 and P2Y2 receptor mRNA in hearts of rats with congestive heart failure has been reported, and there is a dramatic increase in expression of P2X7 receptors in the kidney glomerulus in diabetes and hypertension.

Neuroprotection In the brain, purinergic signaling is involved in nervous tissue remodeling following trauma, stroke, ischemia, or neurodegenerative disorders. The hippocampus of chronic epileptic rats shows abnormal responses to ATP associated with increased expression of P2X7 receptors. Neuronal injury releases fibroblast growth factor, epidermal growth factor, and platelet-derived growth factor. In combination with these growth factors, ATP can stimulate astrocyte proliferation, contributing to the process of reactive astrogliosis and to hypertrophic and hyperplasic responses. P2Y receptor antagonists have been proposed as potential neuroprotective agents in the cortex, hippocampus, and cerebellum. Blockade of

644 Adenosine Triphosphate (ATP)

A2A (P1) receptors antagonizes tremor in Parkinson’s disease. ATP–MgCl2 is being explored for the treatment of spinal cord injuries.

Dual Purinergic Neural and Endothelial Control of Vascular Tone and Angiogenesis ATP and adenosine are much involved in the mechanisms underlying local control of vessel tone in addition to cell migration, proliferation, and death during angiogenesis, atherosclerosis, and restenosis following angioplasty. ATP, released as a cotransmitter from sympathetic nerves, constricts vascular smooth muscle via P2X receptors, whereas ATP released from sensory-motor nerves during ‘axon reflex’ activity dilates vessels via P2Y receptors. Furthermore, ATP released from endothelial cells during changes in flow (shear stress) or hypoxia acts on P2Y receptors in endothelial cells to release NO, resulting in relaxation (Figure 4). Adenosine, following breakdown of extracellular ATP, produces vasodilatation via smooth muscle P1 receptors.

Pain and Purinergic Mechanosensory Transduction The involvement of ATP in the initiation of pain was recognized first in 1966 and later in 1977 using human skin blisters. A major advance was made when the P2X3 ionotropic receptor was cloned in 1995 and shown later to be predominantly localized in the subpopulation of small nociceptive sensory nerves that label with isolectin B4 in dorsal root ganglia whose central projections terminate in inner lamina II of the dorsal horn. A unifying purinergic hypothesis for the initiation of pain was proposed in 1996 with ATP acting via P2X3 and P2X2/3 receptors associated with causalgia, reflex sympathetic dystrophy, angina, migraine, and pelvic and cancer pain. This has been followed by an increasing number of published reports expanding on this concept for acute, inflammatory, neuropathic, and visceral pain. P2Y1 receptors have also been demonstrated in a subpopulation of sensory neurons that colocalized with P2X3 receptors. A hypothesis was proposed that purinergic mechanosensory transduction occurred in visceral tubes and sacs, including ureter, bladder, and gut, where ATP, released from epithelial cells during distension, acted on P2X3 homomultimeric and P2X2/3 heteromultimeric receptors on subepithelial sensory nerves, initiating impulses in sensory pathways to pain centers in the CNS (Figure 5(a)). Subsequent studies of bladder, ureter, gut, tongue, and tooth pulp have produced

NA ATPP1

NA ATP

− ?

Nerve Advent.

a1 + P2X +

Muscle −

ATP

−

P1

Endoth.

EDRF P2Y

Hypoxia shear stress

ADO ATP ADP

Platelets Figure 4 A schematic representation of the interactions of adenosine 50 -triphosphate (ATP) released from perivascular nerves and from the endothelium (Endoth.). ATP is released from endothelial cells during hypoxia to act on endothelial P2Y receptors, leading to the production of endothelium-derived relaxing factor (EDRF), nitric oxide (NO), and subsequent vasodilation (–). In contrast, ATP released as a cotransmitter with noradrenaline (NA) from perivascular sympathetic nerves at the adventitia (Advent.)–muscle border produces vasoconstriction (+) via P2X receptors on the muscle cells. Adenosine (ADO), resulting from rapid breakdown of ATP by ectoenzymes, produces vasodilation by direct action on the muscle via P1 receptors and acts on the perivascular nerve terminal varicosities to inhibit transmitter release. Reproduced from Burnstock G (1987) Local control of blood pressure by purines. Blood Vessels 24: 156–160, with permission from S. Karger AG, Basel.

evidence in support of this hypothesis. P2X3 knockout mice were used to show that ATP released from urothelial cells during distension of the bladder act on P2X3 receptors on subepithelial sensory nerves to initiate both nociceptive and bladder voiding reflex activities. In the distal colon, ATP released during moderate distension acts on P2X3 receptors on lowthreshold intrinsic subepithelial sensory neurons to influence peristalsis, whereas high-threshold extrinsic subepithelial sensory fibers respond to severe distension to initiate pain (see Figure 5(b)). ATP is also a neurotransmitter released from the spinal cord terminals of primary afferent sensory nerves to act at synapses in the central pain pathway. Using transverse spinal cord slices from postnatal rats, excitatory postsynaptic currents have been shown to be

Adenosine Triphosphate (ATP) 645

ATP

ATP

Smooth muscle Subepithelial sensory nerves

ATP ATP

ATP

ATP

ATP

ATP ATP ATP ATP ATP ATP ATP

Epithelial cells

ATP

ATP

P2X2/3 receptors

ATP ATP ATP ATP ATP ATP ATP A

Distension

ATP ATP ATP ATP ATP ATP ATP ATP

ATP ATP

ATP

ATP ATP

ATP

ATP ATP ATP ATP ATP ATP ATP A

ATP

ATP

ATP

P2X2/3

a Spinal cord and brain pain centers

Hypothesis: Purinergic sensory pathways in gut

Moderate distension ATP release stimulation of low threshold intrinsic sensory fibers peristaltic reflexes

DRG

Powerful distension ATP release stimulation of high threshold extrinsic sensory nociceptive fibers pain Intrinsic sensory nerves

Extrinsic sensory nerves P2X3 IB4

P2X3 calbindin

ATP

ATP

ATP ATP

ATP

ATP

ATP

ATP

ATP

ATP

ATP

ATP

ATP

Subepithelial sensory nerve plexus (with P2X3 receptors)

ATP ATP ATP ATP

ATP ATP

ATP

ATP

ATP

ATP ATP

b Figure 5 (a) Schematic representation of the hypothesis for purinergic mechanosensory transduction in tubes (e.g., ureter, vagina, salivary, and bile ducts and gut) and sacs (e.g., urinary and gall bladders and lung). It is proposed that distension leads to the release of adenosine 50 triphosphate (ATP) from the epithelium lining the tube or sac, which then acts on P2X2/3 receptors on subepithelial sensory nerves to convey sensory (nociceptive) information to the central nervous system (CNS). (b) Schematic of a novel hypothesis about purinergic mechanosensory transduction in the gut. It is proposed that ATP released from mucosal epithelial cells during moderate distension acts preferentially on P2X3 receptors on low-threshold subepithelial intrinsic sensory nerve fibers (labeled with calbindin), contributing to peristaltic reflexes. ATP released during extreme distension also acts on P2X3 receptors on high-threshold extrinsic sensory nerve fibers (labeled with isolectin B4 (IB4)) that send messages via the dorsal root ganglia (DRG) to pain centers in the CNS. (a) Adapted from Burnstock G (1999) Release of vasoactive substances from endothelial cells by shear stress and purinergic mechanosensory transduction. Journal of Anatomy 194: 335–342, with permission from Blackwell Publishing. (b) Adapted from Burnstock G (2001) Expanding field of purinergic signaling. Drug Development Research 52: 1–10, with permission of Wiley-Liss, Inc.

mediated by P2X receptors activated by synaptically released ATP, in a subpopulation of less than 5% of the neurons in lamina II, a region known to receive major input from nociceptive primary afferents. There is an urgent need for selective P2X3 and P2X2/3 receptor antagonists that do not degrade

in vivo. Pyridoxal-phosphate-6-azophenyl-20 , 40 -disulphonic acid is a nonselective P2 receptor antagonist but has the advantage that it dissociates about 100 to 10 000 times more slowly than other known antagonists. The trinitrophenyl-substituted nucleotide TNPATP is a selective and very potent antagonist at both

646 Adenosine Triphosphate (ATP)

P2X3 and P2X2/3 receptors. 5-((3-Phenoxybenzyl) [(1S)-1,2,3,4-tetrahydro-1-naphthalenyl]amino carbonyl)-1,2,4-benzenetricarboxylic acid (A-317491) is a potent and selective nonnucleotide antagonist of P2X3 and P2X2/3 receptors, and it reduces chronic inflammatory and neuropathic pain in the rat. Antisense oligonucleotides have been used to downregulate the P2X3 receptor, and in models of neuropathic (partial sciatic nerve ligation) and inflammatory (complete Freund’s adjuvant) pain, inhibition of the development of mechanical hyperalgesia was observed within 2 days of treatment. P2X3 double-stranded-short interfering RNA also relieves chronic neuropathic pain and opens up new avenues for therapeutic pain strategies in humans. Tetramethylpyrazine, a traditional Chinese medicine used as an analgesic for dysmenorrhoea, is claimed to be a P2X receptor antagonist, and it inhibited significantly the first phase of nociceptive behavior induced by 5% formalin and attenuated slightly the second phase in the rat hindpaw pain model. Antagonists to P2 receptors are also beginning to be explored in relation to cancer pain.

development and thus might be useful in the treatment of Me´nie`res disease, tinnitus, and sensorineural deafness. ATP, acting via P2Y receptors, depresses sound-evoked gross compound action potentials in the auditory nerve and the distortion product otoacoustic emission, the latter being a measure of the active process of the outer hair cells. P2X splice variants are found on the endolymphatic surface of the cochlear endothelium, an area associated with sound transduction. Sustained loud noise produces an upregulation of P2X2 receptors in the cochlear, particularly at the site of outer hair cell sound transduction. P2X2 receptor expression is also increased in spiral ganglion neurons, indicating that extracellular ATP acts as a modulator of auditory neurotransmission that is adaptive and dependent on the noise level. Excessive noise can irreversibly damage hair cell stereocilia, leading to deafness. Data have been presented that release of ATP from damaged hair cells is required for Ca2þ wave propagation through the support cells of organ of Corti, involving P2Y receptors, and this might constitute the fundamental mechanism to signal the occurrence of hair cell damage.

Special Senses Eye

P2X2 and P2X3 receptor mRNA is present in the retina and receptor protein expressed in retinal ganglion cells. P2X3 receptors are also present on Mu¨ller cells, which release ATP during Ca2þ wave propagation. ATP, acting via both P2X and P2Y receptors, modulates retinal neurotransmission, affecting retinal blood flow and intraocular pressure. Topical application of diadenosine tetraphosphate has been proposed for the lowering of intraocular pressure in glaucoma. The formation of P2X7 receptor pores and apoptosis is enhanced in retinal microvessels early in the course of experimental diabetes, suggesting that purinergic vasotoxicity might have a role in microvascular cell death, a feature of diabetic retinopathy. The possibility has been raised that alterations in sympathetic nerves might underlie some of the complications observed in diabetic retinopathy; ATP is well established as a cotransmitter in sympathetic nerves, raising the potential for P2 receptor antagonists in glaucoma. P2Y2 receptor activation increases salt, water, and mucus secretion and thus represents a potential treatment for dry eye conditions. Ear

Both P2X and P2Y receptors have been identified in the vestibular system. ATP regulates fluid homeostasis, cochlear blood flow, hearing sensitivity, and

Nasal Organs

The olfactory epithelium and vomeronasal organs contain olfactory receptor neurons that express P2X2, P2X3, and P2X2/3 receptors. It is suggested that the neighboring epithelial supporting cells or the olfactory neurons themselves can release ATP in response to noxious stimuli, acting on P2X receptors as an endogenous modulator of odor sensitivity. Enhanced sensitivity to odors was observed in the presence of P2 antagonists, suggesting that low-level endogenous ATP normally reduces odor responsiveness. It has been suggested that the predominantly suppressive effect of ATP on odor sensitivity could be involved in reduced odor sensitivity that occurs during acute exposure to noxious fumes and might be a novel neuroprotective mechanism. See also: Adenosine; Purines and Purinoceptors: Molecular Biology Overview.

Further Reading Abbracchio MP and Burnstock G (1994) Purinoceptors: Are there families of P2X and P2Y purinoceptors? Pharmacology and Therapeutics 64: 445–475. Abbracchio MP and Burnstock G (1998) Purinergic signalling: Pathophysiological roles. Japanese Journal of Pharmacology 78: 113–145. Bodin P and Burnstock G (2001) Purinergic signalling: ATP release. Neurochemical Research 26: 959–969.

Adenosine Triphosphate (ATP) 647 Burnstock G (1972) Purinergic nerves. Pharmacological Reviews 24: 509–581. Burnstock G (1987) Local control of blood pressure by purines. Blood Vessels 24: 156–160. Burnstock G (1999) Release of vasoactive substances from endothelial cells by shear stress and purinergic mechanosensory transduction. Journal of Anatomy 194: 335–342. Burnstock G (2001) Expanding field of purinergic signaling. Drug Development Research 52: 1–10. Burnstock G (2001) Purine-mediated signalling in pain and visceral perception. Trends in Pharmacological Sciences 22: 182–188. Burnstock G (2002) Purinergic signalling and vascular cell proliferation and death. Arteriosclerosis, Thrombosis and Vascular Biology 22: 364–373. Burnstock G (2001) Purinergic signalling in gut. In: Abbracchio MP and Williams M (eds.) Handbook of Experimental Pharmacology, Vol. 151: Purinergic and Pyrimidinergic Signalling II: Cardiovascular, Respiratory, Immune, Metabolic and Gastrointestinal Tract Function, pp. 141–238. Berlin: Springer. Burnstock G (2001) Purinergic signalling in lower urinary tract. In: Abbracchio MP and Williams M (eds.) Handbook of Experimental Pharmacology, Vol. 151: Purinergic and Pyrimidinergic Signalling I: Molecular, Nervous and Urinogenitary System Function, pp. 423–515. Berlin: Springer. Burnstock G (2003) Purinergic receptors in the nervous system. In: Schwiebert EM (ed.) Current Topics in Membranes, Vol. 54: Purinergic Receptors and Signalling, pp. 307–368. San Diego, CA: Academic Press. Burnstock G (2004) Cotransmission. Current Opinion in Pharmacology 4: 47–52.

Burnstock G (2006) Pathophysiology and therapeutic potential of purinergic signalling. Pharmacological Reviews 58: 58–86. Burnstock G (2007) Physiology and pathophysiology of purinergic neurotransmission. Physiological Reviews 87: 659–797. Burnstock G and Costa M (eds.) (1975) Adrenergic neuroeffector transmission. In Adrenergic Neurones: Their Organisation, Function and Development in the Peripheral Nervous System, pp. 51–106. London: Chapman and Hall. Burnstock G and Knight G (2004) Cellular distribution and functions of P2 receptor subtypes in different systems. International Review of Cytology 240: 31–304. Dunn PM, Zhong Y, and Burnstock G (2001) P2X receptors in peripheral neurones. Progress in Neurobiology 65: 107–134. James G and Butt AM (2002) P2Y and P2X purinoceptor mediated Ca2þ signalling in glial cell pathology in the central nervous system. European Journal of Pharmacology 447: 247–260. Kennedy C, McLaren GJ, Westfall TD, and Sneddon P (1996) ATP as a co-transmitter with noradrenaline in sympathetic nerves – Function and fate. In: Chadwick DJ and Goode J (eds.) P2 Purinoceptors: Localization, Function and Transduction Mechanisms, pp. 223–235. Sneddon P and Burnstock G (1984) Inhibition of excitatory junction potentials in guinea-pig vas deferens by a, b-methylene-ATP: Further evidence for ATP and noradrenaline as cotransmitters. European Journal of Pharmacology 100: 85–90. Zimmermann H (2001) Ectonucleotidases: Some recent developments and a note on nomenclature. Drug Development Research 52: 44–56.

Purines and Purinoceptors: Molecular Biology Overview G Burnstock, Royal Free and University College School of Medicine, London, UK

death that occur in development and regeneration are also mediated by purinergic receptors.

ã 2009 Elsevier Ltd. All rights reserved.

P1 Receptors Early Studies A seminal paper describing the potent actions of adenine compounds was published by Drury and Szent-Gyo¨rgyi in 1929. Many years later, ATP was proposed as the transmitter responsible for nonadrenergic, noncholinergic transmission in the gut and bladder, and the term ‘purinergic’ was introduced by Burnstock in 1972. Early resistance to this concept appeared to stem from the fact that ATP was recognized first for its important intracellular roles in many biochemical processes, and the intuitive feeling was that such a ubiquitous and simple compound was unlikely to be utilized as an extracellular messenger, although powerful extracellular enzymes involved in its breakdown were known to be present. Implicit in the concept of purinergic neurotransmission was the existence of postjunctional purinergic receptors, and the potent actions of extracellular ATP on many different cell types also implicated membrane receptors. Purinergic receptors were first defined in 1976, and 2 years later a basis for distinguishing two types of purinoceptor, identified as P1 and P2 (for adenosine and ATP/ADP, respectively), was proposed. At about the same time, two subtypes of the P1 (adenosine) receptor were recognized, but it was not until 1985 that a pharmacological basis for distinguishing two types of P2 receptor (P2X and P2Y) was made. In 1993, the first G-protein-coupled P2 receptors were cloned and a year later two iongated receptors were cloned, and in 1994 Abbracchio and Burnstock, on the basis of molecular structure and transduction mechanisms, proposed that purinoceptors should belong to two major families: a P2X family of ligand-gated ion channel receptors and a P2Y family of G-protein-coupled purinoceptors. This nomenclature has been widely adopted and currently seven P2X subtypes and eight P2Y receptor subtypes are recognized, including receptors that are sensitive to pyrimidines as well as purines. It is widely recognized that purinergic signaling is a primitive system involved in many nonneuronal as well as neuronal mechanisms, including exocrine and endocrine secretion, immune responses, inflammation, pain, platelet aggregation, and endothelial-mediated vasodilatation. Cell proliferation, differentiation, and

648

Four different P1 receptor subtypes, A1, A2A, A2B, and A3, have been cloned. All are G-protein-coupled receptors (GPCRs). At 318 amino acids in length, the A3 subtype is the shortest, while A2A is the longest (412 residues). Their N-termini are relatively short (7–13 residues in length), as are their C-termini (32–120 residues). In the transmembrane domains (TMI–TMVII), human adenosine receptors share 39–61% sequence identity with each other and 11–18% identity with P2Y receptors. Each of the four human P1 receptor genes contains an intron within the coding region, located immediately after the end of the third transmembrane domain (see Figure 1(a)). P1 receptors couple principally to adenylate cyclase. A1 and A3 are negatively coupled to adenylate cyclase through the Gi/o protein a subunits, whereas A2A and A2B are positively coupled to adenylate cyclase through Gs. The human A2B receptor has also been observed to couple through Gq/11 to regulate phospholipase C activity, and the A3 receptor may interact directly with Gs. A number of P1 subtype-selective agonists and antagonists have been identified (see Table 1). Alteration or opening of the ribose ring drastically reduces affinity. The hydroxyl group at the 20 position is needed for both affinity and activity. The most selective agonist for the A1 subtype is 2-chloro-N6cyclopentyladenosine (CCPA). CGS 21680 is the most selective A2A agonist; NECA is the most potent A2B receptor agonist. 2-Cl-IB-MECA is 11-fold selective for the human A3 receptor and about 1400-fold selective for the rat A3 receptor. In general, methylxanthines such as caffeine and theophylline are weak P1 receptor antagonists. DPCPX (8-cyclopentyl-1,3dipropylxanthine) is an A1 receptor antagonist with subnanomolar affinity. The most selective A2B receptor antagonist is MRS1754. MRE3008-F20 is the most selective human A3 receptor antagonist. The diverse physiological effects mediated by the different P1 receptor subtypes, particularly modulation of the cardiovascular, immune, and central nervous systems, have been confirmed by transgenic knockout mice for A1, A2A, and A3 receptors. In contrast to knockout studies, overexpression of either A1 or A3 subtypes in transgenic mice has a cardioprotective effect.

Purines and Purinoceptors: Molecular Biology Overview

649

Agonist & antagonist recognition site S

S HN2

Binding of 5⬘-substituted agonists

S V

VI S

VII

H

Extracellular

H I

IV

II

III

Intracellular

a

COOH

S-S Extracellular surface

S-S H5

M1

NH2

s Plasma membrane

M2

COOH b

s

NH2

c

Intracellular surface COOH

Figure 1 Membrane receptors for extracellular ATP and adenosine. The P1 family of receptors for extracellular adenosine comprises G-protein-coupled receptors signaling by inhibiting or activating adenylate cyclase (a). The P2 family of receptors binds extracellular ATP or ADP, and comprises two types of receptors (P2X and P2Y). The P2X family receptors are ligand-gated ion channels (b), and the P2Y family members are GPCRs (c). (a) Reproduced from Ralevic V and Burnstock G (1998) Receptors for purines and pyrimidines. Pharmacological Reviews 50: 413–492, with permission from the American Society for Pharmacology and Experimental Therapeutics. (b) Reprinted by permission from Macmillan Publishers Ltd: [Nature] (Brake AJ, Wagenbach MJ, and Julius D (1994) New structural motif for ligand-gated ion channels defined by an ionotropic ATP receptor. Nature 371: 519–523), copyright (1994). (c) Adapted from Barnard EA, Burnstock G, and Webb TE (1994) G protein-coupled receptors for ATP and other nucleotides: A new receptor family. Trends in Pharmacological Sciences 15: 67–70, with permission from Elsevier.

P2X Receptors Molecular Structure

The first cDNAs encoding P2X receptor subunits were isolated in 1994. Members of the family of ionotropic P2X1–7 receptors show a subunit topology of intracellular N- and C-termini possessing consensus binding motifs for protein kinases; two

transmembrane-spanning regions (TM1 and TM2), the first involved with channel gating and the second lining the ion pore; a large extracellular loop, with ten conserved cysteine residues forming a series of disulfide bridges; a hydrophobic H5 region close to the pore vestibule, for possible receptor/channel modulation by cations; and an ATP-binding site, which may involve regions of the extracellular loop adjacent

Receptor

Main distribution

Agonists

Antagonists

Transduction mechanisms

CCPA, CPA, S-ENBA

DPCPX, N-0840, MRS1754

Gi/o #cAMP

A2A A2B

Brain, spinal cord, testis, heart, autonomic nerve terminals Brain, heart, lungs, spleen Large intestine, bladder

CGS 21680, HENECA NECA (nonselective)

GS "cAMP GS "cAMP

A3

Lung, liver, brain, testis, heart

IB-MECA, 2-Cl-IB-MECA, DBXRM, VT160

KF17837, SCH58261, ZM241385 Enprofylline, MRE2029-F20, MRS1754, MRS1706 MRS1220, L-268605, MRS1191, MRS1523, VUF8504

Smooth muscle, platelets, cerebellum, dorsal horn spinal neurons

a,b-meATP ¼ATP ¼ 2-MeSATP (rapid desensitization), L-b,g-meATP ATP  ATPgS  2-MeSATP  a,b-meATP (pH þ zinc sensitive) 2-MeSATP  ATP  a,b-meATP  Ap4A (rapid desensitization) ATP  a,b-meATP, CTP, ivermectin ATP  a,b-meATP, ATPgS

TNP-ATP, IP5I, NF023, NF449

Intrinsic cation channel (Ca2þ and Naþ)

Suramin, isoPPADS, RB2, NF770 TNP-ATP, PPADS, A317491, NF110

Intrinsic ion channel (particularly Ca2þ) Intrinsic cation channel

TNP-ATP (weak), BBG (weak) Suramin, PPADS, BBG

Intrinsic ion channel (especially Ca2þ) Intrinsic ion channel

P1 (adenosine) A1

P2X P2X1

P2X2 P2X3 P2X4 P2X5 P2X6 P2X7 P2Y P2Y1

P2Y2 P2Y4 P2Y6

Smooth muscle, CNS, retina, chromaffin cells, autonomic and sensory ganglia Sensory neurons, NTS, some sympathetic neurons CNS, testis, colon Proliferating cells in skin, gut, bladder, thymus, spinal cord CNS, motor neurons in spinal cord Apoptotic cells in, for example, immune cells, pancreas, skin Epithelial and endothelial cells, platelets, immune cells, osteoclasts Immune cells, epithelial and endothelial cells, kidney tubules, osteoblasts Endothelial cells

P2Y11

Some epithelial cells, placenta, T cells, thymus Spleen, intestine, granulocytes

P2Y12

Platelets, glial cells

P2Y13

Spleen, brain, lymph nodes, bone marrow

P2Y14

Placenta, adipose tissue, stomach, intestine, discrete brain regions

(Does not function as homomultimer) BzATP > ATP  2-MeSATP  a,bmeATP

Gi/o Gq/11 #cAMP "IP3

Intrinsic ion channel KN62, KN04, MRS2427, Coomassie brilliant blue G

Intrinsic cation channel and a large pore with prolonged activation

2-MeSADP ¼ADPbS > 2MeSATP ¼ADP > ATP, MRS2365 UTP ¼ATP, UTPgS, INS37217

MRS2179, MRS2500, MRS2279, PIT

Gq/G11; PLC-b activation

Suramin > RB2, AR-C126313

UTP  ATP, UTPgS

RB2 > suramin

UDP > UTP  ATP, UDPbS

MRS2578

Gq/G11 and possibly Gi; PLC-b activation Gq/G11 and possibly Gi; PLC-b activation Gq/G11; PLC-b activation

AR-C67085MX > BzATP  ATPgS > ATP 2-MeSADP  ADP  ATP

Suramin > RB2, NF157, 50 -AMPS

Gq/G11 and GS; PLC-b activation

CT50547, AR-C69931MX, INS49266, AZD6140, PSB0413, ARL66096, 2-MeSAMP MRS2211, 2-MeSAMP

Gi/o; inhibition of adenylate cyclase

ADP ¼ 2-MeSADP  ATP and 2-MeSATP UDP glucose ¼ UDP-galactose

Gi/o Gq/G11

BBG, brilliant blue green; BzATP, 20 - and 30 -O-(4-benzoyl-benzoyl)-ATP; cAMP, cyclic AMP; CCPA, chlorocyclopentyl adenosine; CPA, cyclopentyl adenosine; CTP, cytosine triphosphate; IP3, inosine triphosphate; IP5I, diinosine pentaphosphate; 2-MeSADP, 2-methylthio-ADP; 2-MeSATP, 2-methylthio-ATP; NECA, 5’-N-ethylcarboxamido adenosine; NTS, nucleus tractus solitarius; PLC, phospholipase C; RB2, reactive blue 2. Adapted and reproduced from Burnstock G (2003) Introduction: ATP and its metabolites as potent extracellular agonists. Current Topics in Membranes 54: 1–27, with permission from Elsevier.

650 Purines and Purinoceptors: Molecular Biology Overview

Table 1 Characteristics of purine-mediated receptors

Purines and Purinoceptors: Molecular Biology Overview Table 2 Potential coassembly of P2X receptor subunitsa

P2X1 P2X2 P2X3 P2X4 P2X5 P2X6 P2X7

P2X1

P2X2

P2X3

þ

þ þ

þ þ þ

P2X4

þ

P2X5

P2X6

þ þ þ þ þ

þ þ

651

Table 3 Chromosomal localization of human P2X receptorsa P2X7

Subunit

Chromosome

Accession number

17p13.2

þ

P2X1 P2X2 P2X3 P2X4 P2X5 P2X6 P2X7

X83688 AF190826 Y07683 Y07684 AF016709 AB002059 Y09561

þ þ

11q12 12q24.31 17p13.3 22q11 12q24.31

a

P2X receptor subunits carrying either one of two epitope tag units were expressed in pairs of HEK293 cells. þ, Subunits immunoprecipitated with antibody to one epitope could be detected with an antibody to the second epitope. Reproduced from Torres GE, Egan TM, and Voigt MM (1999) Hetero-oligomeric assembly of P2X receptor subunits. Specificities exist with regard to possible partners. Journal of Biological Chemistry 274: 6653–6659, with permission from the American Society for Biochemistry and Molecular Biology.

a

Accession numbers are those for the original submission of cDNA sequences. Chromosomal localizations are from human genome databases. P2X2 chromosomal location is not yet determined. The mouse gene is located on chromosome 5, in a region that is syntenic with the extreme end of the long arm of human chromosome 12 (some 6 MB from the P2X4 and P2X7 genes). Reproduced from North RA (2002) Molecular physiology of P2X receptors. Physiological Reviews 82: 1013–1067, with permission from the American Physiological Society.

to TM1 and TM2 (see Figure 1(b)). The P2X1–7 receptors show 30–50% sequence identity at the peptide level. The stoichiometry of P2X1–7 receptors is thought to involve three subunits, which form a stretched trimer or a hexamer of conjoined trimers. All of the P2X receptor subunits have consensus sequences for N-linked glycosylation. The pharmacology of the recombinant P2X receptor subtypes expressed in oocytes or other cell types is often different from the pharmacology of P2Xmediated responses in naturally occurring sites. Several contributing factors may account for these differences. First, heteromultimers as well as homomultimers are involved in forming the trimer ion pore. P2X2/3, P2X1/2, P2X1/5, P2X2/6, P2X4/6, and P2X1/4 receptor heteromultimers have been identified (Table 2). P2X7 does not form heteromultimers, and P2X6 will not form a functional homomultimer. Second, spliced variants of P2X receptor subtypes might play a part. There are seven genes for P2X receptor subunits. P2X4 and P2X7 subunit genes are located close to the tip of the long arm of chromosome 12. P2X4 and P2X7 subunits are among the most closely related pairs in amino acid sequences. P2X1 and P2X5 genes are also very close together on the short arm of chromosome 13. The remaining genes are on different chromosomes (Table 3). The genes vary considerably in size (e.g., mP2X3 ¼ 40 kb; hP2X6 ¼ 12 kb). The fulllength forms have 11–13 exons, and all share a common structure, with well-conserved intron/exon boundaries. Many spliced forms of the receptor subunits (or fragments) have been described. Several fulllength nonmammalian vertebrate sequences have been identified. There are no reports of homologous sequences from invertebrate species, although there is considerable functional evidence that extracellular

ATP and other nucleotides can directly gate ion channels in invertebrates. Recent advances have been made by the preparation of knockout mice for P2X1, P2X2, P2X3, P2X4, and P2X7 receptors, and transgenic mice that overexpress the P2X1 receptor. P2X Receptor Subtypes

P2X1 receptors A cDNA encoding the P2X1 receptor was isolated by direct expression in Xenopus oocytes, beginning with a cDNA library made from rat vas deferens. Human and mouse cDNAs have also been cloned and expressed. The homomeric P2X1 receptor is a cation-selective channel that shows little selectivity for sodium over potassium. It has a relatively high permeability to calcium. A major property of the P2X1 receptor is the mimicry of the agonist actions of ATP by a,b-methylene ATP (a,b-meATP), which distinguishes P2X1 and P2X3 receptors from the other homomeric forms. 20 ,30 -O-(benzoyl-4-benzoyl)-ATP (BzATP) is also an effective agonist. P2X1 receptors are blocked by suramin and pyridoxal-phosphate-6-azophenyl-20 ,40 disulfonic acid (PPADS), but there are newer antagonists that are more P2X1-selective (see Table 1). A valuable antagonist at P2X1 receptors is 20 ,30 -O(2,4,6-trinitrophenyl)-ATP (TNP-ATP), which has an IC50 of about 1 nM. Desensitization means the decline in the current elicited by ATP during the continued presence of ATP. In some P2X receptors this decline occurs in milliseconds (fast desensitization: P2X1, P2X3), and in others it occurs 100–1000 times more slowly (slow desensitization: P2X2, P2X4). Recovery from desensitization is extremely slow. Treatment with

652 Purines and Purinoceptors: Molecular Biology Overview

apyrase allows P2X1 receptors to recover from desensitization. Adenoviral expression of a P2X1 receptor– green fluorescent protein construct in vas deferens shows the receptor to be localized in clusters, with larger ones apposing nerve varicosities. P2X2 receptors The rat P2X2 receptor cDNA was isolated from a library constructed from nerve growth factor (NGF)-differentiated PC12 cells by testing pools for functional expression in Xenopus oocytes. The human receptor cDNA was amplified from pituitary gland. There are no agonists currently known that are selective for P2X2 receptors. However, P2X2 receptors are potentiated by protons and by low concentrations of zinc and copper. There are no selective antagonists for P2X2 receptors. The P2X2 receptor is generally described as nondesensitizing, compared with the P2X1 and P2X3 receptors. When oocytes are injected with RNAs encoding P2X2 receptors, and also the a3 and b4 subunits of nicotinic receptors, they show responses to both ATP and acetylcholine; these can be selectively antagonized with appropriate receptor blockers. However, with concomitant application of both agonists, the resultant current is less than the expected sum of the two independent currents, indicating an interaction between the two receptors. Heteromeric P2X1/2 receptors P2X1 and P2X2 receptor subunits have been co-expressed in defolliculated Xenopus oocytes and the resultant receptors were studied under voltage clamp conditions. Coexpression yielded a mixed population of homomeric and heteromeric receptors, with a subpopulation of novel pH-sensitive P2X receptors showing identifiably unique properties that indicate the formation of heteromeric P2X1/2 ion channels. It has been claimed that trimeric P2X1/2 receptors incorporate one P2X1 and two P2X2 subunits. P2X3 receptors P2X3 receptor subunit cDNAs were isolated from rat dorsal root ganglion cDNA libraries, from a human heart cDNA library, and from a zebra fish library. The mimicry of ATP by a,b-meATP makes these receptors similar to P2X1 and distinct from the other homomeric forms. 2-MethylthioATP is as potent as or more potent than ATP at P2X3 receptors. The antagonists suramin, PPADS, and TNP-ATP do not readily distinguish between P2X1 and P2X3 receptors, but NF023 is about 20 times less effective at P2X3 than at P2X1 receptors. Similar to P2X1 receptors, desensitization is fast and recovery is very slow. P2X3 receptors are prominently expressed on nociceptive sensory neurons.

Heteromeric P2X2/3 receptors Direct association between P2X2 and P2X3 receptor subunits has been shown by co-immunoprecipitation. P2X2/3 heteromeric channels can be defined on the basis of a sustained current elicited by a,b-meATP. P2X2/3 receptor channels and, like homomeric P2X2 receptors, are potentiated by low pH, and do not desensitize rapidly. The P2X2/3 heteromer, like the homomeric P2X3 receptor, is blocked by TNP-ATP, as well as PPADS and suramin. IP5I is much more potent for blocking P2X1 and P2X3 homomers than for blocking the P2X2/3 heteromers and is therefore useful to distinguish between P2X3 and P2X2/3 receptors. P2X2/3 receptors have been identified in subpopulations of sensory neurons, sympathetic ganglion cells, and brain neurons. P2X4 receptors cDNAs for the rat P2X4 receptor were isolated independently from superior cervical ganglion, brain, hippocampus, and pancreatic islet cells. Human, mouse, chick, and Xenopus cDNAs have also been isolated. Homomeric P2X4 receptors are activated by ATP, but not by a,b-meATP. The most useful distinguishing feature of ATP-evoked currents at P2X4 receptors is their potentiation by ivermectin. When the application of ATP is of short duration, P2X4 receptors operate as cation-selective channels; the calcium permeability is relatively high. When the application of ATP is continued for several seconds, the P2X4 receptor channel becomes increasingly permeable to larger organic cations such as N-methyl-D-glucamine (NMDG). Desensitization at P2X4 receptors is intermediate between that observed at P2X1 and P2X2. The rat P2X4 receptor is unusual among the P2X receptors in its relative insensitivity to blockade by the conventional antagonists suramin and PPADS. Currents evoked by ATP at the mouse P2X4 receptor are actually increased by PPADS and suramin, probably because of their ectonucleotidase inhibitory activity. Heteromeric P2X1/4 receptors Co-injection of P2X1 and P2X4 subunits into Xenopus oocytes showed that both subunits were present in trimeric complexes of the same size. Voltage clamp experiments revealed functional P2X receptors with kinetic properties resembling those of homomeric P2X4 receptors and a pharmacological profile similar to that of homomeric P2X1 receptors. Preliminary results show that the P2X1 receptor from the vas deferens and the P2X4 receptor from salivary gland form complexes of the same size as the recombinant trimeric complexes expressed in oocytes.

Purines and Purinoceptors: Molecular Biology Overview

P2X5 receptors The P2X5 receptor cDNA was first isolated from cDNA libraries constructed from rat celiac ganglion and heart. A P2X receptor was also cloned from embryonic chick skeletal muscle. The only human cDNAs reported are missing exon 10 (hP2X5a) or exons 3 and 10 (hP2X5b). A feature of the currents elicited by ATP in cells expressing the rat P2X5 receptor is their small amplitude, compared with the currents observed with P2X1, P2X2, P2X3, or P2X4 receptors expressed under similar conditions. The currents otherwise resemble those seen at P2X2 receptors: they show little desensitization, are not activated by a,b-meATP, and are blocked by suramin and PPADS. P2X5 mRNA is highly expressed in developing skeletal muscle. Heteromeric P2X1/5 receptors P2X1 and P2X5 subunits can be co-immunoprecipitated and the defining phenotype of this heteromer is a sustained current evoked by a,b-meATP, which is not seen for either of the homomers when expressed separately. Cells expressing the heteromeric receptor are very sensitive to ATP, concentrations as low as 3 or 10 nM evoking measurable currents. The sensitivity to the antagonist TNPATP is intermediate between the sensitive homomeric P2X1 receptor and the insensitive homomeric P2X5 receptor. P2X6 receptors The rat P2X6 receptor was cloned from superior cervical ganglion cDNA and from rat brain. The human equivalent was isolated from peripheral lymphocytes as a p53-inducible gene. This was originally designated P2XM to reflect its abundance in human and mouse skeletal muscle. The P2X6 receptor appears to be a ‘silent’ subunit, in the sense that no currents are evoked by ATP when it is expressed as a homomultimer in oocytes or HEK293 cells. It appears that the P2X6 subunit is only functionally expressed as a heteromultimer. Heteromeric P2X2/6 receptors P2X2 and P2X6 receptors have been found to co-immunoprecipitate after expression in HEK293 cells. Oocytes expressing this combination have subtly different responses to ATP as compared to oocytes expressing only P2X2 receptors. The most convincing of these differences is the fact that at pH 6.5 the inhibition of the current by suramin is clearly biphasic; one component has the high sensitivity of homomeric P2X2 receptors, whereas the other component is less sensitive. P2X2/6 receptors are prominently expressed by respiratory neurons in the brain stem. Heteromeric P2X4/6 receptors P2X4 and P2X6 receptors form a heteromeric channel when co-expressed

653

in oocytes. The subunits can be co-immunoprecipitated from oocytes and HEK293 cells. The principal functional evidence for co-expression is that currents elicited by ATP are larger in oocytes 5 days after injection of mRNAs for P2X4 and P2X6 than after injection of P2X4 alone. However, the phenotype of the heteromer differs only in minor respects from that of P2X4 homomers. P2X4/6 receptors are prominent in adult trigeminal mesencephalic nucleus and in hippocampal CA1 neurons. P2X7 receptors A chimeric cDNA encoding the rat P2X7 receptor was first constructed from overlapping fragments isolated from superior cervical ganglion and medial habenula; full-length cDNAs were subsequently isolated from a rat brain cDNA library. Human and mouse cDNAs were cloned from monocyte and microglial cells, respectively. The main feature of the P2X7 receptor is that in addition to the usual rapid opening of the cation-selective ion channel, with prolonged exposure to high concentrations of ATP it becomes permeable to larger cations (e.g., NMDG) and later undergoes a channel-to-pore conversion to allow the passage of large dye molecules such as ethidium and YO-PRO-1, and this usually leads to cell death. Evidence for P2X7 receptor activation includes inward currents and increase in [Ca2þ]i; other end points involve uptake of YO-PRO-1 or similar fluorescent dyes which bind to nucleic acid, and structural changes in the cell, such as membrane blebbing. BzATP is a potent agonist at the P2X7 receptor. There are five main types of blockers (see Table 1): ions (calcium, magnesium, zinc, copper, and protons), the suramin analog NF279, Coomassie brilliant blue G (which is most effective at rat P2X7 receptors), oxidized ATP, and KN62, which is selective for the human P2X7 receptor. ATP or BzATP induces remarkable changes in the appearance of HEK293 cells transfected with the rat P2X7 receptor. After continuous application of BzATP (30 mM) for about 30 s, the plasma membrane begins to develop large blebs, and after 1 or 2 min, these become multiple and sometimes coalesce. Blebs are usually preceded by the appearance of smaller vesicles (