Molecules in Time and Space: Bacterial Shape, Division, and Phylogeny 9780306485787, 0306485788

Citation preview

Molecules in Time and Space

This page intentionally left blank

Molecules in Time and Space Bacterial Shape, Division and Phylogeny Edited by

Miguel Vicente CSIC National Biotechnology Centre Madrid Spain

Javier Tamames BioAlma, S.L. Madrid Spain

Alfonso Valencia CSIC National Biotechnology Centre Madrid Spain

and

Jesús Mingorance CSIC National Biotechnology Centre Madrid Spain

KLUWER ACADEMIC PUBLISHERS NEW YORK, BOSTON, DORDRECHT, LONDON, MOSCOW

eBook ISBN: Print ISBN:

0-306-48579-6 0-306-48578-8

©2004 Springer Science + Business Media, Inc.

Print ©2004 Kluwer Academic/Plenum Publishers New York All rights reserved

No part of this eBook may be reproduced or transmitted in any form or by any means, electronic, mechanical, recording, or otherwise, without written consent from the Publisher

Created in the United States of America

Visit Springer's eBookstore at: and the Springer Global Website Online at:

http://www.ebooks.kluweronline.com http://www.springeronline.com

PREFACE

This book reviews the progress that recent developments in the way in which bacteria can be visualized, and in the sequencing of full bacterial genomes have brought about to the field of bacterial cell division and morphogenesis. Morphogenesis and cell division are two different but related processes. The shape of the cell is intimately connected to how and where division takes place. In its turn cell shape, or at least its aspect ratio, changes along the division cycle. Morphology and division are processes in which the bacterial cell becomes implicated as a whole, and therefore they are not easily amenable to the analytical in vitro procedures so commonly used by biochemists and molecular biologists. As a result the field has been dominated during long time by biophysical research and genetics that produced valuable images and models but failed to provide molecular mechanisms. Technical developments during the last decade are changing this, and, as described by Janet Siefert in Chapter 1, "Three decades later, for something as simple as prokaryotic morphology, the field has come of age". The application of fluorescence microscopy combined with immunolabelling techniques and GFP-fusions, the use of modern methods to crystallize proteins and solve their structures, and specially the sequencing of whole microbial genomes have changed in few years the way we look at bacterial cells. The old view of the cell as an enzyme bag is being replaced by that of an ordered cellular structure in which many proteins are precisely localized, or move through the cell following precise patterns, this order is reflected also in the organization and the evolution of the bacterial genomes, and it is more prominent in the processes of cell division and morphogenesis than in other cellular processes. The Editors

v

This page intentionally left blank

CONTENTS

Chapter 1. The Phylogeny of Bacterial Shape Janet Siefert

1

Chapter 2. Membranes and Prebiotic Evolution: Compartments, Spatial Isolation and the Origin Of Life Ervin Silva and Antonio Lazcano

13

Chapter 3. Topological Domains in the Cell Wall of Escherichia coli Miguel A. de Pedro

27

Chapter 4. Models for Pattern Formation in Bacteria Applied to Bacterial Morphogenesis Hans Meinhardt

59

Chapter 5. The Assembly of Proteins at the Cell Division Site William Margolin

79

Chapter 6. Regulation and Utilization of Cell Division for Bacterial Cell Differentiation Jeniffer Wagner and Yves Brun

103

Chapter 7. FtsZ Folding, Self-association, Activation and Assembly José M. Andreu, María A. Oliva and Sonia Huecas

133

Chapter 8. Sequence and Structural Alignments of Eukaryotic and Prokaryotic Cytoskeletal Proteins Eduardo López-Viñas and Paulino Gómez-Puertas

155

vii

viii

Contents

Chapter 9. Bacterial Morphogenes Jesús Mingorance, Anabel Rico and Paulino Gómez-Puertas

173

Chapter 10. Genome Structures, Operating Systems and the Image of the Machine Antoine Danchin and Stanislas Noria

195

Chapter 11. Gene Order in Prokaryotes: Conservation and Implications Manuel J. Gómez, Ildefonso Cases and Alfonso Valencia

209

Chapter 12. How Similar Cell Division Genes are Located and Behave in Different Bacteria Miguel Vicente, Javier Alvarez and Rocío Martínez- Arteaga

239

Chapter 13. The Bacterial dcw Gene Cluster: an Island in the Genome? Jesús Mingorance and Javier Tamames

249

Index

273

Chapter 1 The phylogeny of bacterial shape

JANET SIEFERT Rice University Dept. of Statistics MS 138 Houston, Texas 77251-1892 USA

1.

Introduction

To understand the evolution of bacterial shape, we must consider how cellular evolution came about initially. At the interface of prebiotic chemistry and life, the need to compartmentalize was a priority for life to evolve. How this compartmentalization first occurred has been the subject of much speculation and its shares of novel theories (Martin and Russell, 2003; Deamer et al., 2002; Wächtershäuser, 2000; Trevor, 2003). For life to diversify and disperse, the compartmentalization had to evolve into autonomous structures. It is obvious from earth's biota that whatever earlier solutions may have been investigated, cells bounded by a phospholipid bilayer became the minimal default condition. We know that further cellular evolution resulted in two basic strategies, the prokaryotic cell type and the eukaryotic cell type. All of extant life today can be divided into these two broad categories. Microscopically, prokaryotes appear 'simpler'. They are cells bounded by a membranous bag that separates the cell's biochemistry from its environment. In eukaryotes, the arrangement is a much more complex system of internal membranes, notably the nuclear membrane which houses the chromosomal complement. The eukaryotes, much to their evolutionary gain, evolved a well-developed cytoskeleton, allowing them special cellular complexity beyond what is seen in the prokaryotes. Prokaryotes, confined to the 'simple' life without this obvious cytoskeletal framework, explored other strategies for evolution. In

Molecules in Time and Space, Edited by Vicente et al. Kluwer Academic/Plenum Publishers, New York 2004

1

2

Siefert

the Bacterial super domain this strategy was the cell wall. Understanding how this cell wall conferred morphology to bacteria has influenced bacterial taxonomy and our perception of their evolution. However, recent investigations have led to the discovery, that indeed the Bacterial domain relies upon a cytoskeleton, with ancestral ties to the tubulin and actin protein families of eukaryotes (Errington, 2003; Shih et al., 2003) for continuity in shape. This information provides an exciting forum in which we can begin to postulate theories on what the first biological cell looked like, what the first bacterial cell looked like, and why each of these entities took the 'shape' they did. If we track the fossil record of shape in the molecular record we can obtain clues as to what environmental pressures shaped the morphology of bacteria.

2.

Crossing the Great Divide and The "First" Shape

The first cell was under tremendous selective pressure. It was necessary to get bigger and divide in a heritable way. It had to do so with the evolving protein toolbox available at that time and with the physicochemical structures dictated by the emerging phospholipid bilayer. It is reasonable to assume that the shape of this entity, once it was fully enclosed, was a coccus. This is not through any heritable system, but due solely to the mechanic forces of the bilayer. As the developing cell became more proficient in its biochemistry, it's needs for increased interaction with its environment became crucial. This connection involved transport within an increasingly bigger cell and transport of metabolites from the outside environment into the cell. Size would have mattered. At this point, it would have all been about surface to volume ratio (Koch, 1996). Getting bigger for a round cell meant losing valuable surface space in which to supply its increasing biochemical demands. Yet, 'growing' required sophisticated control of two things, size and division. It was a delicate dance for a cellular entity in its infancy. The membrane not only provides a containment system but also provides a selectively permeable barrier as well as a mechanism to create electron potential for the cell. As the cell developed from a most primitive state of basic containment, channels and pores that would communicate passively and actively through the membrane from the inside to the outside of the cell were evolved. Enzyme and metabolic system innovations that could make use of the energy potential across the membrane were discovered. Our understanding of how active and passive semi-permeable membrane systems evolved is just beginning to be resolved (Saier, 2003). However, this is critical information in our understanding of cellular

The phylogeny of bacterial shape

3

evolution and future genomic analyses should elucidate the evolutionary mechanisms in more detail. Amphiphilic molecules, such as the ones that are known to form the plasma membrane, spontaneously self-assemble into membrane vesicles. A very novel approach aimed at simulating initial events in cellular evolution has been put forward using fatty acids and clay montmorillonite (Hanczyc et al., 2003). Certainly these experimental conditions and what we know about spontaneous vesicle formation substantiate that the first containment vesicle were most likely spherical (or cocci). Its likely that some molecule like phospholipids were the primitive building block for membranes since all of life uses some combination or variation of the these highly polar molecules for membrane synthesis. A sphere would have been the most likely shape given that there was no information system directing the synthesis or deposition of simple amphiphilic molecules. As soon as a containment system is employed, the necessary next step must be growth and subsequently division. Hanczyc et al. (2003) recognize this necessity and report a method for doing just that by mechanical extrusion through a pore system resulting in cell division and reproduction in their fatty acid micelles. Complicating the challenge for containment was environment. Whether isolated or in multiple environments, the emerging cellular entities had to respond with direct feedback to their environmental conditions. What shape would allow cells the best interaction with their environment? Clearly, the response of these early cells produced more than one solution. How did this come about? Changing shape would have required new genetic instruction either by generation of novel proteins or co-opting existing ones and evolving them for new functions. What other morphological strategies were available that provided better solution? Certainly in any exploration of the possibilities, where the cell was evolving was decisive. It's at this critical juncture in cellular evolution that founding populations with certain physical identities emerged. If we restrict our consideration to the Archaea and Bacteria, the environmental pressures required that they develop a defense against osmotic pressures. The single surviving strategy for Bacteria was a rigid, single molecule outside the membrane. Behold, the cell wall made of murein.

3.

The Bacterial Cell Wall – beyond the sphere

What we need to understand is why or how did bacteria evolve a cell wall and did this solution dictate an original shape for the Bacterial domain that can be determined? Integral to the success of the cell wall was that it would provide some sort of environmental advantage while making use and

4

Siefert

increasing the sophistication of the already well-established cell division machinery. Bacteria could now explore a better solution to the problem of the surface to volume ratio. They could also explore a range of environments that hitherto may not have been accessible. It is important to understand what the protein toolbox included at this time of cellular evolution. We know of two very basic protein families that were most likely available to all three superdomains, one whose progeny today include tubulin and FtsZ and one including the actin superfamily, (MreB, Mbl, ParM, FtsA, Hsp70/DnaK, and hexokinases) (Jones et al., 2001; van den Ent et al., 2001, Doolittle and York, 2002). The course of fate for modern day tubulin and actin is well established. Cytoskeletons and internal compartmentalization not only paved the way for sex and greater size but a level of cellular differentiation that would eventually lead to morphologies that included segmented body parts, wings, and bones. The taxonomy of eukaryotes based on morphology, in at least a broad sense, recapitulated its ancestral history. For the prokaryotes and Bacteria specifically, exploitation of the protein toolbox did not evolve into this eukaryotic sophistication. Several gene duplications and recruitment of new functions of the primitive actin and tubulin protein families provided the genetic system in which to forge a cell wall in the emerging Bacterial lineage. Genes involved in peptidoglycan synthesis were part of the emerging Bacterial toolbox for morphology, most notably the penicillin binding proteins (Fukami-Kobayashi et al., 1999; Goffin and Ghuysen, 1998; Massova and Mobashery, 1998). The cell wall is a biological innovation that allows prokaryotes to survive in a hypotonic environment without bursting (Koch, 1998). In the case of some Archaea and almost all Bacteria a single, huge mesh like molecule (the sacculus or cell wall), covers the whole surface of the cell. Cell walls are present in all the major lineages with the exceptions of Planctomyces, Chlamydia, and Mycoplasmas. It is most likely that these exceptions are due to independent losses in these lineages, especially with regards to the mycoplasmas. Not only that, but murein (a peptidoglycan) is the ubiquitous macromolecule in bacterial cell walls. It is not found in any other super domain. A distantly related murein, pseudomurein, is found only in some methanogenic species of archaea that exhibit cell walls (Konig et al., 1989). It follows then, that very early in the history of the superdomain Bacteria, the synthesis of this sacculus and the ability of the bacteria to grow and divide were co-evolving. This mixing of physical attributes of the mechanical/physical replacement of cell wall constituency and any ribosome directed synthesis of division and growth allows us to begin to unravel the natural history of this process in bacteria.

The phylogeny of bacterial shape

5

We proceed with this line of thought by considering the pertinent facts about the cell wall. Within the Bacteria, there are two variations of cell walls. The so-called Gram+ bacteria have a thick, somewhat threedimensional sacculus composed of peptidoglycan and teichoic acids. Grambacteria have a much thinner cell wall of peptidoglycan and associated proteins as well as a second outer membrane composed of lipid, protein, and lipopolysaccharide. These differences between cell walls in bacteria provide a fundamental and generally phylogenetically cohesive attribute. It also confers a specific morphological identity to each bacterium. The emerging picture for the Bacteria is that selective advantage during their earliest evolution included a rigid cell wall composed of peptidoglycan. Understanding how this cell wall synthesis evolved to allow for division as well as growth in the Bacteria then leads to understanding the evolution of morphology in Bacteria (Koch, 2003). Bacteria produce peptidoglycan (PG) for the cell wall by a suite of enzymes that produce the PG precursors. These are a specific group of murein polymerases that include transglycosylases, transpeptidases and bifunctional transglycosylase/transpeptidases. They interact with murein hydrolases such as endopeptidases and lytic transglycosylases to insert the precursors into the existing wall. There is considerable information regarding the actions and genes involved in this process (Young, 2003). Bacteria exhibit a wide range of morphologies ranging from spherical (cocci) to variations of rods, including spirillum and filaments. As we mentioned, the ability to maintain a specific cell wall for a specific bacterial lineage required that the cell must have evolved strategies that would allow it to grow and divide and maintain the shape. A practical and generally accurate simple descriptor of bacterial morphology is that they are round (cocci) or non-round (which would include rods and all their variations). In this sense then we need to understand the evolution of two general modes of bacterial growth and division, those with spherical morphology (round) and those with rod morphology (non-round). There have been recent developments in our understanding of the process of growth, maintenance of shape and division of bacterial cells in this regard. Several investigations have been pursued that would give some indication as to the nature of the last common ancestor of bacteria. Gupta (2002) tackles the issue of which 'type' of bacterial cell wall was first, the Gram- or Gram+. Using indels (insertion deletion events) in specific proteins, he finds that a series of these events can be considered as gene signature on a timeline of emerging bacterial lineages. Gupta postulates that the Gram+ arrangement was the most ancient cell wall arrangement. His results are compelling but not because they indicate organismal evolution. It should be considered that what is actually being tracked is a metabolic response to environmental selection that has been fixed into certain

6

Siefert

populations. These various indels could be due to a combination of vertical and horizontal transmission of genes that provided a selective advantage in a sequence of environmental changes. In that sense, this study should be looked at in greater detail in light of what kind of environments in earth's history could produce the pattern that these investigators find. Interestingly, we have performed phylogenetic reconstructions on iron-sulfur proteins and various ferredoxins (unpublished results) and find firmicutes as the most basal lineages in our results as well. Both of these studies indicate that circumstantially we might consider that the initial condition for the earliest lineages of bacteria were Gram+, but more pertinent to this chapter, they indicate that it was also rod shaped. Siefert and Fox (1998) investigated the nature of the last common ancestor to Bacteria by "mapping" cell morphology upon the small subunit ribosomal RNA (16s rRNA) gene tree. This study found that rod shaped morphology was basal in the tree and that the most parsimonious explanation for coccus morphology was that it had evolved multiple times over the evolutionary history of Bacteria. Moreover, once the cocci morphology was evident in a lineage, subsequent diversification in that lineage never resulted in a return to the rod morphology. Examination of experimental data corroborates this finding. This line of thinking makes sense if one considers that the earliest 'bacteria' was a rod shape and that this evolved body plan was fate sealing for bacteria. The genetic control of cell division and peptidoglycan synthesis was geared toward maintenance of rod shaped morphology. Once that genetic system is interrupted, the coccus shape becomes an inescapable fate for the cell. Tamames et al. (2001) approached the problem in yet another way (Chapter 13). Postulating that genetic control over shape maintenance would be reflected in clustering or organization of genes, they looked at a welldescribed cluster of genes involved in division and cell-wall synthesis known as the dcw cluster. They deduced the phylogenetic relationship from the arrangement of genes in this cluster. Their tree structure revealed that cell shape was correlated with the order of genes and not just the presence or absence of genes. Their tree was close to the proposed phylogenies based on rRNA reconstructions, although there were exceptions. They observe that genomic organization was positively selected for in bacilli and that the cluster for this morphology is more compact and conserved than for other shapes. They corroborate the findings of Siefert and Fox, namely that reversion from cocci to rod morphology is unlikely. These investigations point to a rod shaped last common ancestor to bacteria. Until the last decade however, interpreting these reconstructions has been without a clear working knowledge of how growth and division was achieved within the various groups of Bacteria. Due to these new

The phylogeny of bacterial shape

7

developments, we can begin to investigate the evolutionary relationships of bacteria based upon their genetic mechanisms for heritable morphology.

4.

The Bacterial Cytoskeleton

For some time it was thought that the shape of bacteria was the sole property of the cell wall. In fact, if the sacculus is removed largely intact, the mesh like layer retains its shape (Höltje, 1998). This indicated that there were other determinants for cell shape. Experimental work from the labs of Jeff Errington and Lawrence Rothfield have been instrumental in elucidating exactly how bacteria divide and deliver peptidoglycan precursors to allow cell growth. In very clever experiments utilizing fluorescent microscopy, a bacterial cytoskeleton was revealed. The proteins involved were members of that ancient toolbox including MreB, Mbl, FtsZ, and the Min division proteins. Cell division begins by formation of a protein assembly nucleated by FtsZ, which under normal circumstances results in binary division with the appropriate genetic complement. Young (2003) presents a detailed review of how FtsZ not only directs cell division but also how it participates in cell shape. Concomitant with the septum site dictated by FtsZ proteins, de Pedro et al. (1997, 2003) shows that a special type of PG, inert peptidoglycan (iPG) is incorporated into the septal site. In non-dividing cells, PG is continually being replaced and recycled in growing cells. Careful regulation within the cell however, diverts this process to production of inert PG patches at the site of septation. Interestingly, these patches remain inert in most rod-shaped bacteria. Young proposes that these inert poles essentially function as overall shape determinants or girders for the cell, the inert patches at opposite ends of the rod dictating the girth and rod shape. This theory is substantially bolstered by a careful consideration of various mutants that appear to impose iPG patches in places other than the cell walls producing branching and bulging morphologies. Lutkenhaus and Sundaramoorthy (2003) report that the interaction of MinD and ATP binding with MinE is critical to the oscillatory mechanism involved in spatial regulation of division. Shih et al. (2003) further report that MinCDE proteins of Escherichia coli are required for correct placement of the division septum at midcell. They find that the Min proteins are organized into membrane-associated coiled structures (cytoskeletal elements) that wind around the cell between the two poles. They note that there is pole-to-pole oscillation of this system in agreement with Lutkenhaus and Sundarmoorthy.

8

Siefert

While FtsZ may represent the successful evolutionary innovation for division, another ancient protein superfamily has been indicted that provides internal or external scaffolding for PG synthesis in general. Recent findings indicate that there is a cytoskeleton in bacteria that is produced by a very ancient protein family homologous to actin (Carballido-López and Errington, 2003; Daniel & Errington, 2003; Jones et al. 2001; Pinho & Errington, 2003). This protein superfamily includes the MreB gene and various homologues such as FtsA (required for cell division) and genes not involved with cytoskeletal function such as the chaperones Hsp70/DnaK and hexokinases. In eukaryotes, actin monomers are polymerized into various filamentous structures that determine the myriad of cell shapes. Until recently, it was not known how these actin homologues participated in cytoskeletal components of bacteria or their role in bacterial shape. Errington and co-workers have shown exactly how the Mre systems of proteins contribute to rod and cocci shape in Gram- bacteria. Their results indicate that in the case of Bacillus subtilis, PG synthesis occurs in the cylindrical part of the cell in a helical pattern. Indeed it is likely that the Mbl cables undergo continuous increase in length, in parallel with length extension and sweep across the cell surface as it grows. They further suggest that stretching of the pre-existing PG layer provides an additional rotational torque that facilitates a complete sweep of the cell surface. This work provides evidence that the Mbl cables spatially direct PG synthesis. Work with mutants showed that Mbl alone is responsible for cylindrical growth in B. subtilis and raises the question for the role of MreB in PG synthesis. Daniel and Errington also report that in cocci (Streptococcus) where there is no MreB system, hence no cytoskeleton, all cell wall synthesis is directed by the division machinery. Division is protracted compared to Bacillus, and closure of the division site is accompanied by elongation from the division site. At the end of the division process, synthesis directed by the division machinery is shut down and the newly formed cell poles are inert. Growth is completely dependent on the formation of a new division site. There are some cases of rod morphology where no MreB cytoskeleton is evident. This is the case for two lineages of bacteria, the Corynebacterium/Mycobacteria and Agrobacterium/Rhizobium. Daniel and Errington (2003) find that the division machinery directs the Corynebacterium cell wall synthesis initially, but growth continues indefinitely at the poles. This is in contrast to the case in B. subtilis and Streptococcus. In the Corynebacterium it is the cylindrical part of the cell that becomes inert. The most salient point of these studies to the aboriginal shape of Bacteria is the origin and importance of the MreB and Min systems to Bacterial morphology. Daniel and Errington (2003) make use of the 16S

The phylogeny of bacterial shape

9

rRNA tree and map the presence of absence of the MreB system upon the tree. The most parsimonious explanation for their results is that the MreB system was in place as a morphology dictating system in the last common ancestor. We already know that the ancient protein family was available for recruitment. We are now left to reconstruct "why" this particular morphological choice became the default system in the super domain of Bacteria.

5.

Reconstructing the history of Bacterial morphology

In 1972, Konig and Kandler proposed a taxonomic scheme that classified lineages based on their peptidoglycans. While this was a difficult and insightful undertaking, it suffered due to the lack of understanding for the genetic control and maintenance of cell shape that could explain its complexity. Subsequently, Koch (2002) spent a great deal of effort in explaining cell morphology based upon classical mechanical force arguments. Several models were put forth that attempted to explain the way both coccus and bacillus organisms could achieve division and growth. Three decades later, for something as simple as prokaryotic morphology, the field has come of age. Experimental work that looks at primitive compartmentalization beyond Sidney Fox's work (1980) is forcing us to consider the initial first steps for cellular life (Woese 2002). Experiments using in vitro microscopic imaging have allowed us to see that bacteria do indeed contain cytoskeletons and these do indeed dictate the integration of peptidoglycan precursors in the cell wall. Whole genome analyses is allowing us to understand the divergence, conservation and occurrence of various genetic mechanisms in all three super domains for the maintenance of cell shape, but importantly, these analyses are providing substantial information on the molecular record of bacterial shape. Our present state of understanding for the evolution of bacterial morphology indicates that the environment in which the bacterial lineage emerged selected for a bacterial cell wall. More than that, the genetic system and primitive protein toolbox that was available selected for a system in which the rod shape was the default for the most recent last common ancestor to the present Bacterial super domain. Future work will help us to understand if this default was critical because of earth's early environment or just a frozen accident.

10

Siefert

References Carballido-López, R. and Errington, J. (2003). A dynamic bacterial cytoskeleton. Trends Cell Biol. 13, 577-583. Daniel, R. A. and Errington, J. (2003). Control of cell morphogenesis in bacteria: two distinct ways to make a rod-shaped cell. Cell 113, 767-776. De Pedro, M. A., Schwarz, H., and Koch, A. L. (2003). Patchiness of murein insertion into the sidewall of Escherichia coli. Microbiology 149, 1753-1761. Deamer, D., Dworkin, J. P., Sandford, S. A., Bernstein, M. P., and Allamandola, L. J. (2002). The first cell membranes. Astrobiology 2, 371-381. Doolittle, R. F. and York, A. L. (2002). Bacterial actins? An evolutionary perspective. Bioessays 24, 293-296. Fox, S. W. (1980). Metabolic microspheres: origins and evolution. Naturwissenschaften. 67, 378-383. Fukami-Kobayashi, K., Tateno, Y., and Nishikawa, K. (1999). Domain dislocation: a change of core structure in periplasmic binding proteins in their evolutionary history. J. Mol. Biol. 286, 279-290. Goffin, C. and Ghuysen, J. M. (1998). Multimodular Penicillin-Binding Proteins: An Enigmatic Family of Orthologs and Paralogs. Microbiol. Mol. Biol. Rev. 62, 1079-1093. Gupta, R. S. (2002). Phylogeny of Bacteria: Are we now close to understanding it? ASM News 68, 284-291. Hanczyc, M. M., Fujikawa, S. M., and Szostak, J. W. (2003). Experimental models of primitive cellular compartments: encapsulation, growth, and division. Science 302, 618622. Höltje, J. V. (1998) Growth of the stress-bearing and shape-maintaining murein sacculus of Escherichia coli. Microbiol. Mol. Biol. Rev. 62, 181-203. Jones, L. J., Carballido-López, R., and Errington, J. (2001). Control of cell shape in bacteria: helical, actin-like filaments in Bacillus subtilis. Cell 104, 913-922. Koch, A. L. (1996). What size should a bacterium be? A question of scale. Annu. Rev. Microbiol. 50, 317-348. Koch, A. L. (1998). How did bacteria come to be? Adv. Microb. Physiol. 40, 353-399. Koch, A. L. (2003). Were Gram-positive rods the first bacteria? Trends Microbiol. 11, 166170. Koch, A. L. (2002). Why are rod-shaped bacteria rod shaped? Trends Microbiol. 10, 452-455. Konig, H., Kandler, O., and Hammes, W. (1989). Biosynthesis of pseudomurein: isolation of putative precursors from Methanobacterium thermoautotrophicum. Can. J. Microbiol. 35, 176-181. Lutkenhaus, J. and Sundaramoorthy, M. (2003). MinD and role of the deviant Walker A motif, dimerization and membrane binding in oscillation. Mol. Microbiol. 48, 295-303. Martin, W. and Russell, M. J. (2003). On the origins of cells: a hypothesis for the evolutionary transitions from abiotic geochemistry to chemoautotrophic prokaryotes and from prokaryotes to nucleated cells. Philos. Trans. R. Soc. Lond. B Biol. Sci. 358, 59-85. Massova, I. and Mobashery, S. (1998). Kinship and diversification of bacterial penicillinbinding proteins and beta-lactamases. Antimicrob. Agents Chemother. 42, 1-17. Pinho, M. G. and Errington, J. (2003). Dispersed mode of Staphylococcus aureus cell wall synthesis in the absence of the division machinery. Mol. Microbiol. 50, 871-881. Saier, M. H., Jr. (1998). Molecular phylogeny as a basis for the classification of transport proteins from bacteria, archaea and eukarya. Adv. Microb. Physiol. 40, 81-136.

The phylogeny of bacterial shape

11

Shih, Y. L., Le, T., and Rothfield, L. (2003). Division site selection in Escherichia coli involves dynamic redistribution of Min proteins within coiled structures that extend between the two cell poles. Proc. Natl. Acad. Sci. U. S. A 100, 7865-7870. Siefert, J. L. and Fox, G. E. (1998). Phylogenetic mapping of bacterial morphology. Microbiology 144, 2803-2808. Tamames, J., González-Moreno,M., Mingorance, J., Valencia, A., and Vicente, M. (2001). Bringing gene order into bacterial shape. Trends Genet. 17, 124-126. Trevors, J. T. (2003). Possible origin of a membrane in the subsurface of the Earth. Cell Biol. Int. 27, 451-457. van den Ent, F., Amos, L., and Löwe, J. (2001). Bacterial ancestry of actin and tubulin. Curr. Opin. Microbiol. 4, 634-638. Wächtershäuser, G. (2000). Origin of life. Life as we don't know it. Science 289, 1307-1308. Woese, C. R. (2002). On the evolution of cells. Proc. Nat. Acad. Sci. U.S.A. 99, 8742-8747. Young, K. D. (2003). Bacterial shape. Mol. Microbiol. 49, 571-580.

This page intentionally left blank

Chapter 2 Membranes and prebiotic evolution: compartments, spatial isolation and the origin of life

ERVIN SILVA and ANTONIO LAZCANO* Facultad de Ciencias, UNAM Apdo. Postal 70-407 Cd. Universitaria, 04510 México D.F. México * corresponding author

1.

Introduction

Was compartmentalization essential for the appearance of life? While it can be argued that encapsulation within lipidic compartments of replicative and catalytic molecules of prebiotic origin would have favoured individuality and selection, it must be underlined that how the transition from the non-living to the living took place is still far from being understood. In addition, it must be underlined that the attributes of the first living organisms are unknown. They were probably simpler than any cell now alive, and may have lacked not only protein-based catalysis, but perhaps even the familiar genetic macromolecules, with their ribose-phosphate backbones. It is possible that the only property they shared with extant organisms was the structural complementarity between monomeric subunits of replicative informational polymers, e.g. the joining together of residues in a growing chain whose sequence is directed by preformed polymers. Such ancestral polymers may have not even involved nucleotides. Accordingly, the most basic questions pertaining to the origin of life relate to much simpler replicating entities predating by a long series of evolutionary events the oldest recognizable heat-loving prokaryotes represented in molecular phylogenies. As discussed elsewhere (Lazcano, 2001), the lack of an all-embracing, Molecules in Time and Space, Edited by Vicente et al. Kluwer Academic/Plenum Publishers, New York 2004

13

14

Silva and Lazcano

generally agreed definition of life sometimes gives the impression that what is meant by its origin is defined in somewhat imprecise terms, and that several entirely different questions are often confused. For instance, until a few years ago the origin of the genetic code and of protein synthesis were considered synonymous with the appearance of life itself. This is no longer a dominant point of view; the discovery and development of the catalytic activity of RNA molecules has given considerable support to the idea of the "RNA world" -a hypothetical stage before the development of proteins and DNA genomes during which alternative life forms based on ribozymes existed (Gesteland et al., 1999). This has led many to argue that the starting point for the history of life on Earth was the de novo emergence of the RNA world from a nucleotide-rich prebiotic soup. Others with a more skeptical view believe that it lies in the origin of cryptic and largely unknown preRNA worlds. There is even a third group that favours the possibility that life began with the appearance of chemoautotrophic autocatalytic metabolic networks, lacking genetic material. Despite the seemingly insurmountable obstacles surrounding the understanding of the origin of life (or perhaps because of them), there has been no shortage of discussion about how it took place. Not surprisingly, several alternative and even opposing suggestions have been made regarding how life emerged and whether membranes were essential or not for the appearance of living beings. As discussed here, although the classical version of the hypothesis of chemical evolution and primordial heterotrophy (Oparin 1924, 1938) needs to be updated, it still provides the most useful framework for addressing the issue of the emergence of life. Alternative theories, such as the autotrophic theory proposed by Wächtershäuser (1988) have been discussed elsewhere (Lazcano, 2001; Bada and Lazcano, 2002), and will not be considered here. The basic tenet of the heterotrophic theory is that the maintenance and reproduction of the first living systems depended primarily on prebiotically synthesized organic molecules, and as discussed below, lipidic molecules were very likely part of this inventory of compounds. However, while the role of membranes in the origin of life has been considered essential in some proposals like that of Oparin (1924, 1938), others who have followed the assumption of prebiotic synthesis and accumulation of organic compounds as a prerequisite for the origin of life, have different viewpoints.

2.

Prebiotic synthesis and the heterotrophic origin of life

It is unlikely that data on how life originated will be provided by the

Membranes and prebiotic evolution

15

palaeontological record. There is no geological evidence of the environmental conditions on the Earth at the time of the origin of life, nor any fossil register of the evolutionary processes that preceded the appearance of the first cells. Direct information is lacking not only on the composition of the terrestrial atmosphere during the period of the origin of life, but also on the temperature, ocean pH values, and other general and local environmental conditions which may or may not have been important for the emergence of living systems. The idea of life as an emergent feature of Nature was widespread during the last century, but it was not until Oparin (1938) proposed that first living systems were heterotrophic microorganisms that resulted from the evolution of abiotically synthesized organic compounds and the formation of selfsustaining supramolecular systems, that the study of the origin of life was transformed from a purely speculative discussion into a workable research program. Today scientific efforts in this field are not necessarily oriented towards the in vitro production of a living system, but rather towards the construction a coherent historical narrative by weaving together a large number of miscellaneous observational findings and experimental results. The hypothesis of chemical evolution is supported not only by a number of laboratory simulations, but also by a wide range of astronomical observations and the analysis of samples of extraterrestrial material. These include the existence of organic molecules of potential prebiotic significance in interstellar clouds and cometary nuclei, and of small molecules of considerable biochemical importance that are present in carbonaceous chondrites. The copious array of amino acids, carboxylic acids, purines, pyrimidines, hydrocarbons, and other molecules which have been found in the years-old Murchison meteorite and other carbonaceous chondrites gives considerable credibility to the idea that comparable syntheses took place in the primitive Earth (Oró et al., 1990; Ehrenfreund et al., 2002). There is also strong experimental support for the idea of prebiotic formation of organic molecules. The first successful synthesis of biochemical compounds under plausible primordial conditions was accomplished by the action of electric discharges acting for a week over a mixture of and yielding a racemic mixture which included several proteinic and non-proteinic amino acids, as well as hydroxy acids, urea, and other organic molecules (Miller, 1953). A few years later, Oró (1960) showed that adenine, a purinic compound that plays a central role in both genetic processes and biological energy utilization, was a major product of the non-enzymatic condensation of HCN, which in turn is an abundant constituent of interstellar clouds and cometary nuclei. The potential role of HCN as a precursor in prebiotic chemistry has been further

16

Silva and Lazcano

supported by experimental evidence showing that the hydrolytic products of its polymers include amino acids, purines, and orotic acid, an intermediate in the biosyntheses of uracil and cytosine, two major components of RNA, indicating that diverse biochemical compounds could had been formed simultaneously from simple reactants (Ferris et al., 1978). Laboratory syntheses under possible primitive conditions of other organic compounds of biochemical significance, such as tricarboxylic acids, alcohols, and a number of coenzymes has been reviewed elsewhere (Oró et al., 1990; Miller and Lazcano, 2002). There can be little doubt that amphiphilic molecules were also present in the prebiotic environment. Lipids are polar derivatives of hydrocarbons, and the abiotic synthesis of the later has been known since the late 19th century. Unfortunately, long-chain linear hydrocarbons and fatty acids are relatively difficult to synthesize under simulated prebiotic conditions (Deamer et al., 1994; Maynard Smith and Szathmary, 1995; Norris and Raine, 1998). Although high yields of several lipidic molecules, including phosphatidic acids, phosphatidylethanolamine, and phosphatidylcholine have been reported in abiotic synthesis starting from simple precursors and following a sequence of step-by-step reactions (Oró et al., 1978), an abiotic source of inorganic phosphate on the primitive Earth is not immediately apparent. However, certain components of carbonaceous meteorites can self-assemble into membranous structures, suggesting a prebiotic accumulation of lipidic compounds due to the infall of extraterrestrial material of meteoritic origin (Deamer et al., 1994). The above results suggest that the prebiotic soup must have been a bewildering organic chemical wonderland, but it could not include all the compounds or the molecular structures found today even in the most ancient extant forms of life -nor did the first cells spring completely assembled, like Frankestein's monster, from simple precursors present in the primitive oceans. The fact that a number of chemical constituents of contemporary forms of life can be synthesized non-enzymatically under laboratory conditions does not necessarily imply by itself that they were also essential for the origin of life, or that they were available in the primitive environment. Moreover, the lack of agreement on the chemical constituents of the primitive atmosphere has also led to major debates. Although it is generally accepted that free oxygen was absent, many planetologists favour the possibility that it consisted of much less-reduced gases such as and (Kastings, 1993), while prebiotic chemists prefer more reducing mixtures (Lazcano and Miller, 1996). The correlation between the compounds which are produced in prebiotic simulations and those found in carbonaceous meteorites (Becker et al., 2002) is too striking to be fortuitous, and strongly supports the contention that such

Membranes and prebiotic evolution

17

molecules were part of the chemical environment from which life evolved. However, the leap from biochemical monomers and small oligomers to membrane-bounded cells is enormous. There is a major gap between the current descriptions of the primitive soup and the appearance of nonenzymatic replication. Solving this issue is essential to our understanding of the origin of the biosphere: regardless of the chemical complexity of the prebiotic environment, life could not have evolved in the absence of a genetic replicating mechanism insuring the maintenance, stability, and diversification of its basic components.

3.

The search for the RNA world

How the ubiquitous nucleic acid-based genetic system of extant life originated is one of the major unsolved problems in contemporary biology. The discovery of catalytically active RNA molecules gave considerable credibility to prior suggestions that the first living organisms were largely based on ribozymes, an hypothetical stage called the RNA world (Gilbert, 1986; Joyce, 2002). This possibility is now widely accepted, but the chemical lability of RNA components suggests that this molecule was not a direct outcome of prebiotic evolution (Orgel, 2003). Moreover, many different (and sometimes even opposing) versions of the RNA world coexist, although its original formulation involved encapsulation of catalytic and replicative RNA molecules within liposomes (Alberts, 1986; Gilbert, 1986; Lazcano, 1986). It is unlikely that wriggling RNA molecules were floating in the primitive ocean, ready to be used as primordial genes. As reviewed elsewhere (Lazcano and Miller, 1996), from the chemical viewpoint the RNA world faces major problems, which include the origin of its ribose moiety, the rapid decomposition of this and other sugars under primitive conditions, and the availability of polyphosphates and phosphate esters, which are not prebiotic reagents. These difficulties have led to proposals of pre-RNA worlds, in which informational macromolecules with backbones different from those of extant nucleic acids may have also been endowed with catalytic activity, i.e., with phenotype and genotype also residing in the same molecules. The nature of the genetic polymers and the catalytic agents that may have preceded RNA is of course unknown (Orgel, 2003). One candidate are the so-called peptide nucleic acids, or PNAs, which are linear polymers in which the sugar-phosphate backbones of RNA and DNA are replaced by uncharged peptide-like backbones formed by achiral amino acid units linked by amide bonds, to which the bases are covalently attached

18

Silva and Lazcano

(Nielsen, 1993). However attractive PNAs or other candidates may be, the origin of non-enzymatic replication remains a major, unsolved problem. Nonetheless, experimental models provide interesting insights. Enhancement of monomer concentration in experimental systems simulating a drying lagoon has achieved a successful surface-bound template polymerization up to 53 nucleotides (Ferris et al., 1996), and the chiroselective self-assembly of long homochiral oligomers of nucleic acid analogues from racemic mixtures of smaller chains into oligomers has been reported (Bolli et al., 1997). There is evidence suggesting that replication may be a widespread phenomenon that includes chemical systems lacking the familiar nucleic acid structure (Orgel, 1992). This possibility is supported by (a) a 32-residue alpha-helical peptide which can template and catalyze its own synthesis from activated smaller fragments under aqueous conditions (Lee et al. , 1996); (b) a horseshoe-shaped product of the chemical reaction between an aminoadenosine and a complex aromatic ester, whose product enhances the formation of similar molecules in a non-aqueous solvent (Hong et al. , 1992); and (c) synthetic micelles containing lithium hydroxide and stabilized by an octanoid acid derivative, which swim in an organic solvent that acts as a substrate for the formation of additional micelles (Bachmann et al. , 1992). While it is unlikely that these non-informational autocatalytic systems are ancestral to our own DNA-based cellular reproduction, their diversity suggests that chemical replicative systems may be much more widespread in Nature than previously thought.

4.

Membranes and precellular evolution

It is frequently argued that a decisive step towards the emergence of the first living systems was the appearance of membrane-enclosed polymolecular systems, since semipermeable prebiotic membranes would have favoured (1) the cooperative interaction between different catalytic and replicative molecules, avoiding their dispersal, and opening the possibility of specific-surface chemistry processes; (2) the creation of internal microenvironments substantially different from the exterior milieu maintained by (at least partially) selective transmembrane transport; and (3) the preferential accumulation and, eventually, differential multiplication of self-sustaining replicating systems. Although many different models of precellular systems have been suggested the most significant may be in fact liposomes. Lipidic membranes are an essential component of cells, and amphiphilic molecules have been formed abiotically (Hargreaves and Deamer, 1978a,b; Epps et al., 1978; Rao et al., 1987). Moreover, the presence of lipids in the primitive environment

Membranes and prebiotic evolution

19

is further supported by the existence of membrane-forming non-polar molecules detected in samples of the Murchison meteorite (Deamer, 1985; 1998; Deamer and Pashley, 1989). Bilayered liposomic structures can easily auto-assemble from a wide variety of lipidic molecules under both physiological and putative prebiotic conditions (Deamer and Barchfeld; Cullis and Hope, 1985; Walde et al., 1994; Chakrabarti et al., 1994). Prebiotic liposomes could have formed from small, single-chain, ionic, linear fatty acids (Hargreaves and Deamer, 1978a,b). Encapsulation of replicative and catalytic molecules could have been driven by periodic environmental changes such as dehydrationhydration cycles (Deamer and Barchfeld, 1982), enhanced by metallic cations (Baeza et al., 1987), basic polypeptides (Jay and Gilbert, 1987), and polyamines derived from the non-enzymatic decarboxylation of basic amino acids. Neither the size or the functional properties of RNA molecules or polyribonucleotides appear to be altered by the encapsulation process, which can take place in the presence of histidine and prebiotic condensing agents such as cyanamide (Oró and Lazcano, 1990). It is reasonable to assume that once prebiotic liposomes were formed by the self-assembly of simple amphiphilic lipids in the presence of a large variety of biochemical monomers and oligomers, new physicochemical properties could result from the interactions between the components of these polymolecular systems. This is not purely speculative; that interactions between liposomes and different water-soluble polypeptides lead to major changes in the morphology and permeability of liposomes of phosphatidylL-serine, and to a transition of poly-L-lysine from a random coil into an that exhibits hydrophobic bonding with the lipidic phase has been documented in the laboratory (Hammes and Schullery, 1970). Comparable interactions may have taken place between lipids and different oligomers of prebiotic origin, leading to changes in the stability and catalytic properties of precellular systems. This hypothesis is consistent with the suggestion that an energy source for nutrient transport and chemical activation processes may have been provided by chemiosmotic proton gradients formed by simple pigments asymmetrically oriented in the lipidic layer (Deamer and Oró, 1980; Morowitz et al., 1988). The formation of liposomes raises the question of the uptake of small water-soluble molecules that cannot penetrate lipidic membranes (Lazcano et al., 1992; Wächtershäuser, 1992). It can be argued that nonselective pores may have existed in primitive membranes, since it is known that these structures are formed in mixed lipid bilayers (Robertson, 1983), when polyL-serine is added to liposomes (Hammes and Schulley, 1970), and when phosphatidate is present in the lipid mixture (Baeza et al., 1990). Complex transporters were of course absent during the pre-biological stages of

20

Silva and Lazcano

evolution, but carrier-mediated diffusion may have taken place. It has been suggested that a rudimentary transport mechanism may have existed, involving facilitated diffusion of complexes between aldehydes, amines, and metal ions with amino acids, sugars, and nucleotides, respectively (Stillwell, 1976, 1980; Stillwell and Rau, 1981). This possibility has gained support from experiments that have shown that neutral forms of amino acids go across lipidic membranes much more easily than their charged forms (Chakrabarti and Deamer, 1994).

5.

Some biological problems

It is not known how proton gradients originated and became coupled with ions and directionality. However, transporters must have appeared early in biological evolution. In order to study ion conduction across biological membranes, Lear et al. (1988) synthesized model oligopeptides that were large enough to span the hydrocarbon phase of the lipid bilayer. Quite significantly, they found that two small amphiphilic peptides with a simple repetitive structure behaved as more complex biological ion channels do. A 21-residue peptide with the sequence formed ion channels whose permeability and lifetime resembled that of the acetylcholine receptor, while an equally small peptide with the sequence produced protonselective channels. These experiments were not performed within an evolutionary context or under prebiotic conditions (Lear et al., 1988). However, they illustrate how small simple oligopeptides of prebiotic origin, or synthesized by primitive cells with limited coding abilities, could have been involved in selective ion transport across primitive membranes. The origin of transport mechanisms is related to the appearance of transduction systems and energy-producing mechanisms, i.e, to bioenergetic processes which lie at the very basis of metabolism (Holden, 1968; Maloney and Wilson, 1985). It is generally assumed that membranes provided the necessary separation between the internal microenvironment and the external surroundings which were to maintain higher reaction rates inside precellular systems. This has been clearly demonstrated by the enzyme-mediated synthesis of poly(A) within the boundaries of coacervate drops (Oparin, 1971) and, more recently, in phospholipid or oleic acid/oleate vesicles (Walde et al., 1994; Chakrabarti et al., 1994). However, it is not known how primordial energy transduction systems became coupled with polymerization reactions involving the components of primordial genetic polymers. The prebiotic synthesis of lipids implies that early membranes were formed, in a very literal sense, from the fat of the land (or of the oceans).

Membranes and prebiotic evolution

21

Hence, reproduction of the first cells would have been hindered by the exhaustion of this supply of lipidic material of prebiotic origin. A possible primitive synthesis of fatty acids from glycoaldehyde has been suggested (Weber, 1991), but very little is known of the evolutionary steps that led to the development of biosynthetic pathways of lipids. Since these molecules are a prerequisite to the origin and maintenance of cells, and since acetylcoA is mandatory in contemporary biosynthesis of fatty acids (Vance and Vance, 1985), one possibility is that the original activation of acetate may have resulted from its interaction with catalytic RNA molecules (Lazcano, 1986).

6.

Final remarks

There has been no shortage of discussion about how the prebiotic soup formed and the transition to the origin of life took place. However, it is likely that no single mechanism can account for the wide range of organic compounds that may have accumulated on the primitive Earth, and that the prebiotic soup was formed by contributions from endogenous syntheses in a reducing atmosphere, metal sulphide-mediated synthesis in deep-sea vents, and exogenous sources such as comets, meteorites and interplanetary dust (Miller and Lazcano, 2002). As summarized here, the existence of different abiotic mechanisms by which biochemical monomers can be synthesized under plausible prebiotic conditions is well-established. The wide range of experimental conditions under which organic compounds can be synthesized demonstrates that prebiotic syntheses of the building blocks of life are robust, i.e., the abiotic reactions leading to them do not take place under a narrow range defined by highly selective reaction conditions, but rather under a wide variety of experimental settings. The robustness of this type of chemistry is supported by the occurrence of most of these biochemical compounds in the Murchison meteorite (Miller and Lazcano, 2002), including membrane-forming lipidic molecules (Deamer et al., 1994). It is very attractive to assume that compartmentalization within liposomes formed by amphiphilic molecules of prebiotic origin was essential for the emergence of life. However, other alternatives include sequestering of catalytic and replicative molecules within.prebiotic compartments made of hydrophobic amino acid polymers (Lehmann and Kuhn, 1984), simple terpenoids (Ourisson and Nakatani, 1994), or alternating polypeptides (Brack and Orgel, 1975). As noted by Joyce (2002), alternative membrane-free systems such as the association of RNA (or its pre-RNA precursors) and other molecules to surfaces via transient covalent or noncovalent interactions (Gibson and Lamond, 1990) and passive

22

Silva and Lazcano

compartmentalization within aerosol drops (Dobson et al. , 2000) or in rock pores can also be envisioned. However, even if spatial isolation of the first replicative systems did not involve lipidic membranes, the transition from hypothetical precellular systems into the extant biological membranes and the biosynthesis of their lipidic components must have taken place very early in evolution. Testable descriptions of how such transition took place require coherent proposals involving ribozyme-mediated reactions or semienzymatic synthesis involving less-specific biological catalysts. Understanding how this took place remains a major, unsolved problem in our understanding of the emergence of life.

Acknowledgments Support from UNAM-DGAPA Proyecto PAPIIT IN 111003-3 to A. L. is gratefully acknowledged.

References Alberts, B. M. (1986) The function of the hereditary materials: biological catalyses reflect the cell's evolutionary history. Am. Zool. 26, 781-796. Bachmann, P. A., Luisi, P. L., and Lang, J (1992) Autocatalytic self-replicating micelles as models for prebiotic structures. Nature 357, 57-59. Bada, J. and Lazcano, A. (2002) Some like it hot, but not biomolecules. Science 296, 19821983. Baeza, I., Ibañez, M., Lazcano, A., Santiago, C., Argüello, C., Wong, C., and Oró, J. (1987) Liposomes with polyribonucleotides as models of precellular systems. Orig. Life Evol. Biosph. 17, 321-331. Baeza, I., Ibañez, M., Argüello, C., Wong, C., and Oró, J. (1990) Diffusion of ions into liposomes by phosphatidate and monitored by the activation of an encapsulated enzymatic system. J. Mol. Evol. 31, 453-461. Bolli, M., Micura, R., and Eschenmoser, A. (1997) Pyranosyl-RNA: chiroselective selfassembly of base sequences by ligative oligomerization of tetranucleotide-2',3'cyclophosphates (with a commentary concerning the origin of biomolecular homochirality). Chem. Biol. 4, 309-320. Brack, A. and Orgel, L. E. (1975) Structures of alternating polypeptides and their possible prebiotic significance. Nature 256, 383-387. Chakrabarti, A. C. and Deamer, D. W. (1994) Permeation of membranes by the neutral form of amino acids and peptides: relevance to the origin of peptide translocation. J. Mol. Evol. 39, 1-5. Chakrabarti, A. C., Breaker, C. C., Joyce, G. F., and Deamer, D. W. (1994). Production of RNA by a polymerase protein encapsulated within phospholipid vesicles. J. Mol. Evol. 39, 555-559. Cullis P. R. and Hope, M. J. (1985). Physical properties and functional roles of lipids in membranes. In D. E. Vance and J. E. Vance (eds), Biochemistry of Lipids and

Membranes and prebiotic evolution

23

Membranes (Benjamin/Cummings, Menlo Park). Deamer, D. W. (1985) Boundary structures are formed by organic components of the Murchison carbonaceous chondrite. Nature 317, 792-794. Deamer, D. W. (1998) Membrane compartments in prebiotic evolution. In Brack, A. (ed) The Molecular Origins of Life: assembling pieces of the puzzle (Cambridge University Press, Cambridge). Deamer, D. W. and Barchfeld, G. L. (1982) Encapsulation of macromolecules by lipid vesicles under simulated prebiotic conditions. J. Mol. Evol. 18, 203-206. Deamer, D. W. and Oró, J. (1980) Role of lipids in prebiotic structures. BioSystems 12, 167165. Deamer, D. W. and Pashley, R. M. (1989) Amphiphilic components of the Murchison carbonaceous chondrite: surface properties and membrane formation. Orig. Life Evol. Biosph. 19, 21-38. Deamer, D. W., Mahong, E. H., and Bosco, G. (1994) Self-assembly and function of primitive membrane structures. In Stefan Bengtson (ed) Early Life on Earth: Nobel Symposium No. 84 (Columbia University Press/Nobel Foundation, New York). Dobson, C. M., Ellison, G. B., Tuck, A. F., and Vaida, V. V. (2000) Atmospheric aerosols as prebiotic chemical reactors. Proc. Natl. Acad. Sci. USA 97, 11864-11868. Ehrenfreund, P., Irvine, W., Becker, L., Blank, J., Brucato, J., Colangeli, L., Derenne, S., Despois, D., Dutrey, A., Fraaije, H., Lazcano, A., Owen, T., Robert, F. (2002) Astrophysical and astrochemical insights into the origin of life. Reports Prog. Physi. 65, 1427-1487. Epps, D. E., Sherwood, E., Eichberg, J., and Oró, J. (1978) Cyanamide-mediated synthesis under plausible primitive Earth conditions. V. The synthesis of phosphatidic acids. J. Mol. Evol. 11, 279-292. Ferris, J. P., Joshi, P. D., Edelson, E. H., and Lawless, J. G. (1978) HCN: a plausible source of purines, pyrimidines, and amino acids on the primitive Earth. J. Mol. Evol. 11, 293311. Ferris, J. P., Hill, A. R., Liu, R., and Orgel, L. E. (1996) Synthesis of long prebiotic oligomers on mineral surfaces. Nature 381, 59-61. Gesteland, R. F., Cech, T., & Atkins, J. F. (eds) 1999. The RNA world II. (CSHL Press, Cold Spring Harbor). Gibson, T. J. and Lamond, A. I. Metabolic complexity in the RNA world and implications for the origin of protein synthesis. J. Mol. Evol. 30, 7-15. Gilbert, W. (1986) The RNA world. Nature 319, 618. Hammes, G. G. and Schullery, S. E. (1970) Structure of molecular aggregates. II. Construction of model membranes from phospholipids and polypeptides. Biochemistry 9: 2555-2563. Hargreaves, W. R. and Deamer, D. W. (1978a) Origin and early evolution of bilayer membranes. In D. W. Deamer (ed), Light-Transducing membranes: structure, function, and evolution (Academic Press, New York). Hargreaves, W. R. and Deamer, D. W. (1978b) Liposomes from ionic, single-chain amphiphiles. Biochemistry 17, 3759-3768. Holden, J. T. (1968). Evolution of transport mechanisms. J. Theoret. Biol. 21, 97-102. Hong, J. 1., Feng, Q., Rotello, V., and Rebek, J. Jr. (1992) Competition, cooperation, and mutation: improving a synthetic replicator by light irradiation. Science 255, 848-850. Jay, D. G. and Gilbert, W. (1987) Basic protein enhances the incorporation of DNA into lipid vesicles: model for the formation of primordial cells. Proc. Natl. Acad. Sci. USA 84, 1978-1980. Joyce, G. F. (2002) The antiquity of RNA-based evolution. Nature 418, 214-221.

24

Silva and Lazcano

Kasting, J. F. (1993) Earth's early atmosphere. Science 259, 920. Lazcano, A. (1986) Prebiotic evolution and the origin of cells. Treballs Soc. Cat. Biol. 39, 73. Lazcano, A. (2001) Origin of Life In Derek E. G. Briggs and Peter R. Crowther (eds) Palaeobiology II (Blackwell Science, London). Lazcano, A. & Miller, S. L. 1996. The origin and early evolution of life: prebiotic chemistry, the pre-RNA world, and time. Cell 85, 793. Lazcano, A., Fox, O.E. and Oró, J. (1992) Life before DNA: the origin and early evolution of early Archean cells. In R. P. Mortlock (ed) The Evolution of Metabolic Function (CRC Press, Boca Raton, FL). Lear J. D., Wasserman, Z. R., and DeGrado, W. F. (1988) Synthetic amphiphilic peptide models for protein ion channels. Science 240, 1177-1181. Lee, D. H., Granja, J. R., Martínez, J. A., Severin, K., and Ghadari, M. R. (1996) A selfreplicating peptide. Nature 382, 525-528. Lehmann, U. and Kuhn, H. (1984) Emergence of adaptable systems and evolution of a translation device. Adv. Space Res. 4, 153. Maloney, P. C. and Wilson, T. H. (1985) The evolution of ion pumps. BioScience 35, 43. Maynard Smith, J. and Szathmary, E. (1995) The Major Transitions in Evolution (W. H. Freeman, Oxford). Miller, S. L. (1953) A production of amino acids under possible primitive Earth conditions. Science 117, 528. Miller, S. L. and Lazcano, A. (2002) Formation of the building blocks of life. In J. W. Schopf (ed) Life's Origin: The beginnings of biological evolution (California University Press, Berkeley). Morowitz, H. J., Heinz, B. and Deamer, D. W. (1988) The chemical logic of a minimum protocell. Origins Life Evol. Biosph. 18, 281-287. Nielsen, P. E. 1993. Peptide nucleic acid (PNA): a model structure for the primordial genetic material? Origins Life Evol. Biosph. 23, 323-327. Norris, V. and Raine, D. J. (1998) A fission-fusion origin for life. Origins Life Evol. Biosph. 28, 523-537. Orgel, L. E. 1992. Molecular replication. Nature 358, 203-209. Orgel, L. E. (2003) Some consequences of the RNA world hypothesis. Origins Life Evol. Biosph. 33, 211-218. Oparin, A. I. (1924) Proiskhozhedenie Zhizni (Mosckovskii Rabochii, Moscow). Reprinted and translated in J. D. Bernal (1967), The Origin of Life (Weidenfeld and Nicolson, London). Oparin, A. I. (1938) The Origin of Life (MacMillan, New York). Oparin, A. I. (1971) Coacervate drops as models of prebiological systems. In A. P. Kimball and J. Oró (eds), Prebiotic and Biochemical Evolution (North-Holland, Amsterdam). Oró, J. (1960) Synthesis of adenine from ammonium cyanide. Biochem. Biophys. Res. Commun. 2, 407. Oró, J. and Lazcano, A. (1990) A holistic precellular organization model. In C. Ponnamperuma and F. Eirich (eds), Prebiological Self-Organization of Matter (A. Deepka, Hampton). Oró, J., Sherwood, E., Eichberg, J., and Epps, D. (1978) Formation of phospholipids under primitive Earth conditions and the role of membranes in prebiological evolution. In D. W. Deamer (ed), Light-Transducing membranes: structure, function, and evolution (Academic Press, New York). Oró, J., Miller, S. L., and Lazcano, A. (1990) The origin and early evolution of life on Earth. Annu. Rev. Earth Planet. Sci. 18, 317-356. Ourisson, G. and Nakatani, Y. (1994) The terpenoid theory of the origin of cellular life: the

Membranes and prebiotic evolution

25

evolution of terpenoids to cholesterol. Chem. Biol. 1, 11-23. Rao, M., Eichberg, J., and Oró, J. (1987) Synthesis of phosphatidyl ethanolamine under possible primitive Earth conditions. J. Mol. Evol. 25, 1-6. Robertson, R. N. (1983) The Lively Membranes (Cambridge University Press, Cambridge). Stillwell, W. (1976) Facilitated diffusion of amino acids across bimolecular lipid membranes as a model for selective accumulation of amino acids in a primordial protocell. BioSystem 8, 111-117. Stillwell, W. (1980) Facilitated diffusion as a method for selective accumulation of materials from the primordial ocean by a lipid-vesicle protocell. Origins of Life 10, 277-292. Stillwell, W. and Rau, A. (1981) Primordial transport of sugars and amino acids via Schiff bases. Origins of Life 10, 243-254. Vance, D. E. and Vance, J. E. (eds.) (1985) Biochemistry of Lipids and Membranes (Benjamin/Cummings, Menlo Park). Walde, P., Goto, A., Monnard, P.A., Wessicken, M., Luisi, P. L. (1994) Oparin's reaction revisited: enzymatic synthesis of poly(adenylic acid) in micelles and self-reproducing vesicles. J. Am. Chem. Soc. 116, 7541. Wächtershäuser, G. (1988) Before enzymes and templates: theory of surface metabolism. Microbiol. Rev. 52, 452-484. Wächtershäuser, G. (1992) Groundwork for an evolutionary biochemistry: the iron-sulphur world. Prog. Biophys. Molec. Biol. 58, 85-201. Weber, A. L. (1991) Origin of fatty acid synthesis: thermodynamics and kinetics of reaction pathways. J. Mol. Evol. 32, 93.

This page intentionally left blank

Chapter 3 Topological domains in the cell wall of Escherichia coli

MIGUEL A. DE PEDRO Centro de Biología Molecular "Severo Ochoa". Consejo Superior de Investigaciones Científicas-Universidad Autónoma de Madrid. Campus de Cantoblanco. 28049 Madrid. Spain.

1.

Introduction

Since the early days of research on the biology of the bacterial cell envelope it was clearly stated that the murein (peptidoglycan) layer or sacculus, was the element responsible for the high mechanical strength of the cell wall (Weidel and Pelzer, 1964). Furthermore it was soon realized that the net-like, covalently closed structure of the sacculus was particularly well suited to maintain cell shape whilst standing the high turgor pressure exerted by the cytosol against the cytoplasmic membrane of the cell. Because the sacculus fully wraps the cell body, variations in the physical dimensions of the cell are constrained by a concurrent variation of the sacculus proper (Höltje, 1998; Nanninga, 1988; Nanninga, 1991). Therefore it is the cytoplasmic material that has to adopt the shape of the rigid cell wall. Indeed, the ability of the isolated sacculus to accurately retain the shape of the cell further reinforced his role as a bacterial exo-skeleton (Höltje, 1998). This conception however had a psychological draw back. The idea of a resistant structure with a defined shape was inadvertently but unduly assimilated to the idea of inertness: Once made the sacculus would be a stable structure resilient to any further modification. In addition, the apparent morphological homogeneity of the bacterial cell envelope suggested a similar homogeneity at the functional level.

Molecules in Time and Space, Edited by Vicente et al. Kluwer Academic/Plenum Publishers, New York 2004

27

28

de Pedro

However the advancement of the last years in understanding cell wall metabolism, and in particular the introduction of new analytical techniques allowing for a high resolution analysis of the sacculus, changed the picture completely (de Pedro et al., 1997; Glauner et al., 1988; Ishidate et al., 1998; Kohlrausch et al., 1989). A new conception of the sacculus as a dynamic, adaptive structure made up of differentiated domains clearly emerges from the later studies.

2.

Structure of the sacculus

The sacculus can be envisaged as a macromolecular network in which long glycan strands are cross-linked by short peptide bridges (Höltje, 1998). The strength and elasticity of the sacculus are conditioned by three main parameters; the thickness, the extent of cross-linkage and the length of the glycan chains which define the number of chemical bonds opposing the cytoplasmic turgor pressure.

2.1

One or more murein layers?

The E. coli sacculus is often thought of as a single layer of murein. However experimental evidence suggests that the real structure is more complex. Evidence from small-angle neutron scattering analysis indicated that a substantial proportion of the sacculus surface is triple-layered (Labischinski et al., 1991). Calculation of the amount of murein per cell is also consistent with a basic monolayer structure with areas of multilayered murein (Wientjes et al., 1991), as are the observations that about 50% of the total murein is dispensable for survival and morphogenesis (Prats and de Pedro, 1989), and the existence of cross-linked trimers and tetramers discussed below (Glauner et al., 1988). These observations support the idea of a sacculus consisting of a single stress-bearing continuous layer with patches where additional layers, likely corresponding to intermediates in the insertion of precursors and/or in turnover, are apposed to either side.

2.2

The murein network: cross-linking of glycan strands

The essentially planar structure of the E. coli sacculus imposes a restriction on the degree of cross-linking amongst adjacent glycan strands. As consecutive disaccharide-peptide units are rotated by ca. 90° respect to each other, only half the peptides stem out in appropriate directions as to face each other and be cross-linked (Höltje, 1998; Labischinski and Maidhof, 1994). Indeed experimental determinations indicate that only

Topological domains in the cell wall of E. coli

29

between 25 and 35% of the peptide moieties are actually part of peptide bridges (Glauner et al., 1988; Glauner and Höltje, 1990; Pisabarro et al., 1985). The proportion of cross-linked muropeptides (cross-linking) is by no means constant. In fact it is strongly dependent on growth, with higher values found in murein from resting cells, and on murein age, lower in newly inserted material than in the older one (de Pedro and Schwarz, 1981; Glauner and Höltje, 1990; Pisabarro et al., 1985). An interesting point to consider is the finding of cross-linked trimers and tetramers in the murein of E. coli and other Gram negatives (Glauner et al., 1988; Quintela et al., 1995). The presence of these particular classes of muropeptides can only be explained if several glycan strands intersect at a particular point, or if areas of multilayered murein exist, in accordance with the neutron diffraction results (Labischinski et al., 1991). New precursors are cross-linked to pre-existing murein by the activity of DD-transpeptidases, an activity associated to the high molecular weight Penicillin-binding proteins (PBP's) (Höltje, 1998; Matsuhashi et al., 1990; Spratt, 1977a; Spratt and Cromie, 1988). Inhibition of DD-transpeptidation by is in fact the basis for the bactericidal effect of these compounds (Spratt, 1975; Spratt and Cromie, 1988). Until the introduction of high resolution methods for muropeptide analysis it was assumed that all cross-linked muropeptides were linked by D-D peptide bridges, the product of DD-transpeptidation (Höltje, 1998; Schleifer and Kandler, 1972). However, introduction of HPLC demonstrated the existence of L-D cross linked muropeptides in significant proportions (5-10 % of total cross-linked muropeptides) (Glauner et al., 1988; Glauner and Höltje, 1990; Höltje, 1998). The new family of muropeptides were cross-linked by an L-D peptide bond between the L-carboxyl group of the meso-diaminopimelic acid residue in the donor muropeptide, and the D-amino group of the mesodiaminopimelic acid residue in the acceptor muropeptide. The synthesis of LD-muropeptides requires specific LD-transpeptidases, and is penicillin insensitive (Caparrós et al., 1992; Höltje, 1998). L,D-transpeptidation seems important for the cell when the dominant DD-transpeptidation is restricted, conditions which lead to a drastic increase in the proportion of LDmuropeptides possibly playing a compensatory role (Blasco et al., 1988; Glauner et al., 1988; Kohlrausch and Höltje, 1991; Pisabarro et al., 1985). Nevertheless, the role of LD-cross-linking remains basically unknown. The possibility of LD-transpeptidation as a macromolecular-murein modifying (repair?) reaction has some data in favour. LD-Muropeptides increase in the presence of murein synthesis inhibitors that is under conditions when no new precursors are available, and the only possible substrate is macromolecular murein itself (Kohlrausch and Höltje, 1991).

30

de Pedro

2.3

The murein network: length and distribution of the glycan strands

The second key parameter is the length of the glycan chains. The average length of murein glycan strands is known since long. Because the glycan chain terminal residue of N-acetyl-muramic acid (NAM) is in the (1-6) anhydro form, the number of disaccharide units per glycan chain can be estimated as the inverse of the molar fraction of (l-6)anhydro-NAM muropeptides (Glauner et al., 1988; Höltje, 1998). An average glycan chain length of about 30 disaccharide units has been consistently found for E. coli (Glauner et al., 1988; Glauner and Höltje, 1990; Höltje, 1998; Pisabarro et al., 1985; Quintela et al., 1995). However, as for cross-linkage, a number of growth related variables strongly influence this parameter. In particular murein glycan strands are shorter in resting (ca. 15 units) than in actively growing cells (ca. 30 units) (Pisabarro et al., 1985). The introduction in the late 80's of high resolution methods for the separation of glycan chains according to their length, allowed a more precise approach to the size distribution of glycan strands (Harz et al., 1990). Unexpectedly the length distribution of glycan strands apparently follows a bimodal distribution. About 60 % of total material could be resolved by HPLC in a series of peaks separated by one disaccharide unit in length, from a single disaccharide up to 30 disaccharides. This fraction showed a modal value around 7-8 disaccharides with an average value about 9 disaccharide units. Unfortunately the remaining 40% can only be eluted from the HPLC columns as a single fraction by means of a stepwise solvent change. Analysis of this fraction showed that it consists of long glycan strands (>30 disaccharides) with an average chain length of 45 disaccharide units, but the shape of the distribution remains unknown (Harz et al., 1990; Höltje, 1998; Kohlrausch and Höltje, 1991). Interestingly recalculation of the glycan strand average length from the length distribution data gave a value of 21 disaccharide units, considerably shorter than estimations from the (1-6) anhydro-NAM content of murein. Technical considerations argue in favour of the smaller value as the closer to reality. Therefore the sacculus seems to be made of two subsets of peptidoglycan strands that could play specific roles in morphogenesis. Glycan strands in any case, are too short to cover a significant fraction of the cell circumference, or length, individually. As a disaccharide is about 1 nm long, and a normal E. coli is between 0.5 and in diameter, between 1500 and 3000 disaccharides would be needed to go around the circumference of the sacculus. At this moment the existence of a small subpopulation of such extremely long strands cannot be formally ruled out, but looks quite improbable.

Topological domains in the cell wall of E. coli

31

On the other hand, the structural role of short strands is puzzling. According to established cross-linkage values (25-30%), strands shorter than 6-7 disaccharide units (about 20-30 % of total murein) are expected to be linked to only one neighbouring strand. If this were the case those strands would be hanging out of the sacculus network, but could not be part of the net itself. This could represent recycling material, or even docking sites for murein-interacting molecules. An alternative is that cross-linking would not follow a simple statistical distribution. In fact, cross-linkage at the terminal muropeptides of the glycan strands is much higher than the average (Altmutter, 2002; Costa et al., 1999; Glauner et al., 1988). Therefore it is conceivable that glycan chains, irrespective of their length, have a very high probability to be cross-linked at both ends. Inevitably the higher than average cross-linkage for the shorter strands would only be possible at the expense of lower than average values for the longer strands. Similarly, strands between 7 and 11 disaccharide units (accounting for ca. 20% of total murein) are expected to be cross-linked in average at two positions only, with a high chance for both terminals. These arguments take us to a situation in which it is possible that a large fraction of total glycan strands are cross-linked at two (terminal) positions only. Such a disposition of strands in a turgor stressed structure would permit a much larger elasticity than the kind of regular net idealized in most models. Indeed, short glycan strands interconnecting longer strands by terminal peptide bridges would functionally be equivalent to long-range cross-links. Studies on the Gram negative bacteria Helicobacter pylori provided experimental evidence in favour of short glycan strands acting as long-range cross-links between long chains (Costa et al., 1999). 2.4

The murein network: Orientation of the glycan strands in the sacculus

Orientation of glycan strands in the sacculus is an important structural aspect under discussion. Determination of whether or not there is a preferential orientation of the glycan strands respect to the cell axis, how stringent it is, and of course which is the orientation, is of prime importance to understand growth and physico-chemical properties of the sacculus. X-ray diffraction studies showed that the E. coli sacculus is a non crystalline, relatively low order structure, discarding models based on a regular array of the glycan chains (Labischinski and Maidhof, 1994). Electron microscopy studies of purified sacculi subjected to partial digestion with DD-endopeptidases and lytic-transglycosylases suggested that glycan strands were preferentially oriented in a circumferential fashion, perpendicular to the cell axis (Verwer et al., 1978). Sacculi digested with the

32

de Pedro

peptide-bridge splitting endopeptidase apparently showed elliptic holes oriented perpendicular to the long cell axis. When the glycan backbone was degraded by a transglycosylase no orientation was evident. A similar conclusion was reached from data on the elastic properties of purified sacculi (Koch and Woeste, 1992; Yao et al., 1999). Sacculi were more elastic in the longitudinal than in the transversal axis. The glycan backbone is essentially inelastic in opposition to the peptide bridges which can stretch for about twice their length at the state of minimal energy (Koch and Woeste, 1992). Therefore, the bigger elasticity of sacculi in the longitudinal direction pointed also to a preferential orientation of the peptide-bridges parallel to the cell axis. However, recent experiments questioned the observations based on the enzymatic digestion of sacculi (de Pedro et al., 1997; Koch, 1998). A series of experiments showed that sacculi treated with a glycan backbone splitting muramidase did show the same kind of transversally oriented elliptic holes as described by Verwer et al.(1978). Data based on the elasticity of sacculi looks more convincing, but the length distribution of glycan strands was not considered in any of the original reports (Koch and Woeste, 1992; Yao et al., 1999). This is relevant because if the elastic behaviour of short glycan strands acting as long-range bridges were comparable to that of a regular peptide bridge, many of these strains could be longitudinally oriented and still allow for the observed anisotropy in the elastic properties of sacculi.

3.

Interactions of sacculi cytoplasmic membranes

with

the

outer

and

In E. coli the sacculus is closely associated to the outer membrane (OM) by means of a number of proteins able to link murein to components of the OM. One of them, Braun's lipoprotein (Lpp), is the only known protein that binds covalently to murein in E. coli (Braun and Rehn, 1969; Braun and Wu, 1994). Lpp is the most abundant protein in the cell envelope and coexists in free (60-70%) and murein bound (30-40%) forms (Braun, 1975; Inouye et al., 1972). Attachment of Lpp to the OM presumably occurs through the insertion of the acyl chains into the inner phospholipid leaflet of the OM (Braun, 1975; Mizushima, 1987; Yu et al., 1984). Free Lpp molecules form trimers which may contribute to the strength of the murein-OM complex by additional non-covalent interactions (Choi et al., 1986; Choi et al., 1987). The sacculus and the OM also interact through non-covalently bound proteins as the peptidoglycan associated lipoprotein (PAL) and OmpA. Binding of PAL to the sacculus is very strong and involves a murein-binding

Topological domains in the cell wall of E. coli

33

specific sequence (Lazzaroni and Portalier, 1992). PAL anchors into the OM by the acylated N-terminal and direct protein-protein interactions with other OM proteins (Abergel et al., 2001; Lazzaroni and Portalier, 1992; Mizuno et al., 1982). OmpA is a major integral OM protein with domains protruding from both the external and internal leaflets. External domains function as a phage receptor and the internal ones interact directly with murein (Hancock et al., 1994; Nikaido and Vaara, 1985). Interaction of OmpA with murein is strong enough to stabilize the cell envelope in the absence of Lpp (Sonntag et al., 1978; Yem and Wu, 1978). A number of other OM proteins as OmpF and OmpC also make minor contributions to the total interaction (Osborn and Wu, 1980). In spite of the crucial role played by the inner (cytoplasmic) membrane (IM) in the biosynthesis of the sacculus, very little is known about interactions between both structures at present. At first sight it seems reasonable that the turgor pressure of the cell should push the IM against the sacculus (Koch, 1983). Under this assumption the complete OM-sacculusIM system should be rather compact. However, very recent observation of E. coli by cryo-transmission electron microscopy of frozen-hydrated sections leave little doubt about the presence of a substantial space separating the sacculus and the IM (Matias et al., 2003), and add further support to the idea of a periplasmic gel (Kellenberger, 1990). Nevertheless, the fact that the Penicillin-binding proteins (PBP's), the key enzymes responsible for the polymerization of peptidoglycan strands and their insertion into the sacculus, are IM proteins demands sites of close contact between both structures (Höltje, 1998; Spratt, 1975; Spratt, 1977a; Spratt and Cromie, 1988). At present no specific attaching elements have been demonstrated, but quite interestingly the PAL/TOL complex could play such a role as three of its constituent proteins TolA, TolQ and TolR are IM proteins (Bouveret et al., 1999).

4.

Growth of the sacculus

Growth of E. coli occurs by the periodic succession of elongation and division events (Ayala et al., 1994; Cooper, 1991). Cells divide at their midpoint once the chromosome is replicated and initial mass (length) is doubled (Donachie, 1968; Donachie et al., 1976; Donachie, 1993). Cell elongation demands the concomitant enlargement of the sacculus, and cell division requires the formation of a transversal septum at the centre of the sacculus. As the sacculus supports the turgor pressure of the cell, its enlargement and formation of septa must proceed avoiding generation of discontinuities to prevent cell lysis (Höltje, 1998; Koch, 1988).

34

de Pedro

Growth of the sacculus requires the insertion of new material into the preexisting murein network. The immediate precursors for the growth of the sacculus are synthesized in the periplasmic side of the IM by the high molecular weight class A PBP's (Ghuysen, 1994; Höltje, 1998; Matsuhashi, 1994). These proteins polymerize the activated disaccharide-pentapeptide units into peptidoglycan strands by transglycosylation and concomitantly catalyze their insertion into the sacculus by transpeptidation. Cell elongation and septation both require "de novo" murein synthesis, but involve specific components and although coordinated, are largely independent from each other (Ayala et al., 1994; Höltje, 1998). Growth in length can proceed indefinitely in the absence of septation, and positioning of septation events can be altered without effect on longitudinal growth (Ayala et al., 1994; de Boer et al., 1989). Furthermore, as will be commented below in more detail, E. coli septal murein synthesis (SMS) can also proceed for longer than normal times in the absence of elongation (Ayala et al., 1994; de Boer et al., 1989; de Pedro et al., 2001). As the sacculus is a covalently closed structure, some covalent bonds have to be cleaved concomitantly with the incorporation of new material to permit the expansion of the pre-existing structure (Koch, 2000). The simple binding of the incoming strand to the sacculus does not promote its growth unless specific bonds are cleaved to permit the new material to move into, and expand, the stress bearing murein layer (Höltje, 1998). This action was traditionally ascribed to the murein hydrolases, a complex set of enzymes splitting the chemical bonds that keep murein units bound together (Shockman and Höltje, 1994). However, recent studies question this assumption at least in part. The key objection is the lack of growth-impaired phenotypes in (multiple) murein hydrolase mutants (Heidrich et al., 2002; Höltje and Heidrich, 2001). In most cases there is a serious impairment of septation but not of longitudinal growth. Such observations lead to a new hypothesis proposing different mechanisms for the incorporation of new precursors into the in-growing septum and into the elongating lateral wall. Whereas classical murein hydrolases and in particular the amidase AmiC, would be required for septal synthesis (Bernhardt and de Boer, 2003), precursors would be incorporated into the lateral wall by a murein transferase activity (Höltje and Heidrich, 2001). Transferases catalyze a two-step reaction; first, a bond in a donor substrate is cleaved concomitantly with the formation of a covalent enzymesubstrate intermediate. In the second step the donor substrate moiety is covalently bound (transferred) to the acceptor substrate and the enzyme is regenerated. Insertion of precursor strands into the sacculus could therefore be catalyzed by transferases coupling the opening of meshes (that is, cross

Topological domains in the cell wall of E. coli

35

bridges) in the sacculus with the covalent attachment of the new strands to the murein network (Höltje and Heidrich, 2001). Studies on the characteristics of murein synthesis throughout the cell cycle as well as on the properties of E. coli class A PBP's support this proposal (Charpentier et al., 2002; de Jonge et al., 1989). The comments above are consistent with the intervention of two independent biosynthetic systems in the metabolism of the sacculus, as originally proposed by Spratt and Satta (Lleo et al., 1990; Satta et al., 1985; Satta et al., 1994; Spratt, 1975). One system would be dedicated to the promotion of longitudinal growth, and the second to the synthesis, at a particular place and time, of the transversal septum. There is little, if any, doubt about the assembly of a large multiprotein complex specifically dedicated to septal synthesis (the septosome) at the precise time and place as to promote efficient cell division (Ayala et al., 1994; Cooper, 1991; Errington et al., 2003; Nanninga, 2001). Recent work centred on the analysis of molecular interactions among proteins involved in murein metabolism supports the real existence of multienzyme complexes dedicated to lateral wall murein synthesis (Höltje and Heidrich, 2001; Vollmer et al., 1999; Vollmer and Höltje, 2001). Perhaps the most attractive model at the moment is the "holoenzyme" model as postulated by Höltje (Höltje, 1996; Höltje, 1998; Höltje and Heidrich, 2001). This model proposes the existence of a basic multienzyme complex comprising a set of murein syntheses and hydrolases whose activity would be modulated by additional components to be directed towards lateral wall or septal murein synthesis.

5.

Topology of the precursor-insertion sites in the E. coli sacculus

A point of singular concern in the context of bacterial morphogenesis is how and where new precursors are incorporated into the pre-existing sacculus. That is the topography of cell wall growth. Obviously this is a most relevant aspect to understand how the final shape of the sacculus is generated, and to determine whether the sacculus as a growing component of the cell, is homogeneous (diffuse growth) or presents functionally differentiated areas (zonal growth). As the sacculus is "a priori" the only physically stable structure of the bacterial cell, the existence of domains could have important consequences for the general arrangement of the cell envelope.

36

de Pedro

The rest of this chapter will deal with the particular problem of the topology of cell wall growth and how it may condition the behaviour of those elements able to interact with the sacculus, in particular the OM.

5.1

Mapping the incorporation sites for radioactive murein precursors; the nuclear occlusion effect

Early attempts to directly map the growth sites in the sacculus of E. coli were based on the analysis of the distribution of newly inserted radioactive murein precursors by high resolution autoradiography (Verwer and Nanninga, 1980). This technique suffered of intrinsic technical limitations degrading the spatial resolution of the analysis. The main limitations came from the relatively low number of silver grains per cell in the autoradiographies imposed by the specific activity of available precursors, and the uncertainty generated by the free-flight trajectory of from the point of emission to the target film (up to for routinely used label). Nevertheless, the excellent experimental work and critical analysis of the results performed by Woldringh et al. (Koch and Woldringh, 1994; Mulder and Woldringh, 1991) showed that during elongation radioactivity was distributed essentially at random over the surface of the sacculus, irrespective of the labelling time, except in constricting cells where the central region of the cell showed an appreciably higher density of silver grains. The results gave support to the "Nuclear occlusion model" which concisely proposed that the bacterial nucleoid would have a short range inhibitory effect on murein synthesis (Woldringh et al., 1990; Woldringh et al., 1991). The nucleoid occupies most of the cell volume at any time of the cell cycle, except during the last stages of chromosome segregation when a nucleoid-free region develops between the segregated chromosomes, at the centre of the growing cell. Therefore, murein synthesis would proceed at a lower than optimum rate anywhere in the cell surface, except at the centre of the cells that would be nucleoid-free for the last stages of the cell cycle. The increase in the rate of synthesis at this time (when chromosomes are replicated and separated), and location (midcell) would then permit proper septation. Since its original proposal the nucleoid occlusion model has been further developed and at present the physical influence of the nucleoid is one of the many factors suspected to play a role in the regulation of the cell cycle (Woldringh, 2002). A most interesting aspect of these results is the idea of a diffuse insertion mode for new precursors active most of the cell cycle with periods of increased activity at the potential division sites, very much in accordance with the coexistence of two kinds of murein biosynthetic complexes discussed above.

Topological domains in the cell wall of E. coli 5.2

37

Mapping precursor insertion sites in D-amino acid modified sacculi

5.2.1 The D-amino acid murein-labelling method

5.2.1.1 Incorporation of D-amino acids into the bacterial sacculus An important advance for the studies on the topology of cell wall synthesis came unexpectedly from research on the bacteriolytic activity of D-methionine (D-met) when fed to bacteria at high (>5mg/ml) concentrations (Tsuruoka et al., 1984; Tsuruoka et al., 1985). Investigation of this phenomenon showed that D-met, was incorporated into the peptidoglycan of E. coli, and other Gram negative bacteria, by means of an enzymatic amino-acid exchange reaction in the periplasm (Caparrós et al., 1992; Caparrós et al., 1993b). The reaction, formally a LD-transpeptidation, resulted in the net exchange of the D-ala normally present at position 4 of the side peptides of murein for a D-met residue (Caparrós et al., 1992; Caparrós et al., 1993a). Substrate specificity was rather poor as several Damino acids could be accepted and subsequently incorporated into murein. Incorporation of D-met was not the direct cause for bacteriolysis, which was rather assigned to a direct inhibitory action on PBPs. After growth in media supplemented with D-met at relatively low concentrations (0.5-1 mg/ml) up to 40% of the peptide side chains in murein contained D-met instead of Dala, without any noticeable damage being inflicted to the cells (Caparrós et al., 1992; Caparrós et al., 1993b). The low substrate specificity and harmless nature of the D-amino acid exchange reaction could be exploited to label sacculi with an easily detectable molecule not naturally found in murein. Detection of the modifying residue in particular regions after a chase period would define the murein as "old". Regions without label would correspond to material inserted during the chase period. The distribution of the label would reflect the biosynthetic activity of different regions in the sacculus, and should therefore lead to the accurate cartography of putative cell wall domains. The compound of choice for labelling was D-Cys because: i) it is not normally present in the cell wall; ii) it is efficiently incorporated in a very homogeneous fashion all over the sacculus; iii) its distribution in sacculi is easy to image by optical or electron microscopy techniques using commercially available reagents to detect the –SH groups. The results obtained since the introduction of D-Cys labelling have demonstrated a good spatial resolution, and flexibility as to be of application in a large number of experimental set-ups (de Pedro et al., 1997).

38

de Pedro

Furthermore, the method is likely applicable to other bacteria as incorporation of D-amino acids has also been detected in Salmonella typhimurium, Enterobacter cloacae, and Pseudomonas aeruginosa indicating that it may be a rather general characteristic of Gram negative bacteria (unpublished results). 5.2.1.2 Experimental set-up Most experiments described below were based on a common experimental set-up, essentially a label-and-chase strategy. Cells were first incubated in the presence of D-Cys to label murein, and then transferred into D-Cys-free medium to perform the specific experimental treatment. Finally cell growth was stopped and sacculi were extracted in boiling 4% sodium dodecyl sulphate (SDS). Murein incorporated into the sacculus during the chase period would therefore be essentially free of D- cys. In the following steps, sacculi were purified, deproteinized, reduced with and biotinylated with a -SH-specific reagent. At the end of the treatment, the regions with old murein are therefore biotinylated whereas the regions of new murein are not. The distribution of biotin molecules in the sacculi was routinely visualized by immunofluorescence or immunoelectron microscopy (de Pedro et al., 1997). 5.2.2 Segregation of D-Cys labelled murein in rod-shaped E. coli

5.2.2.1 Segregation of murein in dividing cells The pattern of murein segregation in sacculi from growing cells was determined as described above. Control sacculi from non-chased cells showed a homogeneous distribution of label irrespective of their size and degree of constriction. However, in cells chased for one generation welldefined unlabelled areas were evident in many sacculi. The label-free areas evolved either as gaps at the central position in constricting sacculi or at one pole in the smaller cells. In the sample chased for two generations, all sacculi fell into one of the following categories: one unlabelled polar cap, one unlabelled cap and a central gap, or two unlabelled polar caps. The distribution of biotin in the labelled areas was denser in the poles than in the cylindrical wall, but essentially homogeneous in each area. The fact that after a very long chase time a certain proportion of sacculi still exhibited a heavily labelled pole was the first evidence indicating that polar murein was very long lived. Indeed in cultures chased for five generations 4.5% of the sacculi had one labelled polar cap, a value very close to the expected 6.2% assuming that the polar cap murein segregates in a conservative way through successive division cycles. Furthermore, the signal intensity in the labelled

Topological domains in the cell wall of E. coli

39

poles from sacculi chased for five generations was essentially (±10%) the same as in control non-chased sacculi. Interestingly, the boundaries between polar caps and cylindrical murein as well as between the later and the central gaps were remarkably sharp suggesting that each region behaved as an isolated domain. These results lead to the following statements to explain murein segregation: (i) murein in the polar caps is stable; (ii) at a particular moment in the cell cycle, a region of localized synthesis differentiates at the potential division site and generates a septum made of new murein; and (iii) longitudinal expansion of the cylindrical part of the cell wall occurs by diffuse insertion. The scheme in Fig. 4A illustrates the theoretical evolution of sacculi according to our postulates and how it correlates with the patterns of murein segregation actually observed. 5.2.2.2 Segregation of murein in division-blocked filament cells Analysis of murein segregation in cells inhibited for division during the chase period further confirmed the observations with dividing cells, and revealed unexpected features of cell wall topology (de Pedro et al., 1997). As a first approach chemical inhibitors were used to block division during the chase period with the consequent formation of filaments. Two inhibitors with different mechanisms of action were selected, the azthreonam and nalidixic acid. Azthreonam inhibits penicillin-binding protein 3 (PBP 3), the septum-making enzyme but a late recruit to the septosome (Ayala et al., 1994; Errington et al., 2003; Georgopapadakou et al., 1982), whilst nalidixic acid inhibits DNA replication and consequently the earlier steps of cell division via FtsZ (Ayala et al., 1994; Hooper, 2001; Lutkenhaus, 1983). After a chase period of one generation in azthreonam, most sacculi showed a neat unlabelled central gap, with the poles more intensely labelled than the cylindrical wall (Fig. 5A and B). In the larger sacculi two additional unlabelled gaps often developed at one-quarter and three-quarters of the sacculus. These longer sacculi correspond to cells close to a division event at the beginning of the chase, which were approaching a second round of division at harvest time. After longer chase times virtually all sacculi had multiple, regularly spaced, unlabeled gaps (Fig. 5C). In addition, the contrast between the polar caps and cylindrical wall was higher, as would be expected if polar murein were indeed stable. An unexpected finding in the azthreonam induced filaments was that the gaps at the center (oldest) and at the "fourths" (younger) of sacculi had equal sizes. The expectation was for the oldest gap (central) to be significantly larger than younger ones because gap and cell lengths should increase in proportion. However, proportionality between both parameters holds only for a reduced range of cell lengths (from 2.5 to for the strain used in

40

de Pedro

the experiments); below no gaps were found, and above the gap width remained constant (ca. irrespective of cell length. This observation convincingly indicates that the gap-generating mechanism is active only for a limited period of the cell cycle. A noteworthy fact was the absence of unlabelled gaps in sacculi from cells chased in the presence of nalidixic acid. The polar regions showed a stronger labelling that the lateral wall, but no gaps nor any other kind of label-free area was observed on the surface of treated sacculi, irrespective of chase time (de Pedro et al., 1997). 5.2.2.3 Stability of polar cap murein A most remarkable result from the murein segregation studies above was the demonstration of the stable nature of polar cap murein in E. coli, as previously proposed on (mostly) theoretical grounds (Koch, 1983; Koch, 2000; Koch and Woldringh, 1994). The observations in sacculi from filament cells confirmed this point beyond doubt. Persistence of polar murein implies that the rates of synthesis and turnover must be essentially null at the poles, which in turn requires their functional isolation from enzymes and/or substrates. Studies on the minicells produced by certain E. coli mutants provided important complementary information. Minicells are produced when a septation event takes place at a polar "potential division site", because of a failure to properly locate the active septosome (de Boer et al., 1989; Jaffe et al., 1988) . As a consequence small, DNA-less spherical cells (minicells), and multinucleate filaments are formed concomitantly. Therefore, minicells can be considered as made up of two cell poles put together. A series of experiments by Markiewicz and Höltje (Markiewicz and Höltje, 1992), showed that E. coli minicells are absolutely recalcitrant to treatments able to induce autolysis in normal cells, a result strongly in support of the functional segregation of polar murein from murein-metabolizing enzymes proposed above. Chemical modification of polar murein would be a likely means to make polar murein refractory to enzymatic activities. However, comparative analysis of murein from minicells and normal cells revealed only minor quantitative differences (Obermann and Höltje, 1994), and the analysis of murein in synchronized cultures demonstrated a constant composition throughout the cell cycle (de Jonge et al., 1989; Obermann and Höltje, 1994). Moreover, enzymatic treatment of purified sacculi with different murein hydrolases showed that polar and lateral wall murein were equally susceptible to the enzymes (de Pedro et al., 1997; Verwer et al., 1978), Besides, the enzyme responsible for the incorporation of D-amino acids did not show any discrimination between polar and lateral wall murein (de Pedro

Topological domains in the cell wall of E. coli

41

et al., 1997). All these observations argue against a chemical modification of polar murein. A likely alternative is the existence of some kind of compartmentalization that would restrict enzymatic access to murein or specific substrates. This possibility finds some support from fluorescence recovery after photobleaching experiments showing a restriction to the free diffusion of molecules associated to the generation of division sites (Foley et al., 1989), and in the existence of favoured locations for the development of plasmolysis bays, with highest probability at the new poles and lower at the constriction sites (Mulder and Woldringh, 1993). The inert character of polar caps was first demonstrated in the Grampositive rod Bacillus subtilis, therefore it could be a general property of rodlike bacteria. The metabolic inertness of polar regions was a prediction of the so called "Surface Stress Theory" (SST) proposed by A.L. Koch as a theoretical framework to understand bacterial morphogenesis on physicochemical terms (Koch, 1983; Koch, 1990; Koch and Woldringh, 1994). Insertion of precursors in the semi-spherical polar regions would lead to a progressive rounding of the cells due to the effect of the internal turgor pressure. As polar caps are the product of division, it is immediate to propose that the mechanism responsible for the inertness of polar murein becomes operative in coordination with septation itself. This idea is supported by the behaviour of the "septal gaps" in filament cells. The fact that gap growth stops at a particular moment whilst cell elongation keeps on going can only be explained if murein in the gap regions becomes refractory to the insertion of new precursors once it is laid down, irrespective of its position in the cell. As formation of gaps apparently depends on the machinery that normally makes septa, the result supports that the inert status of polar murein is acquired at the very moment of the septation event that gives birth to the new poles. 5.2.2.4 Generation of septal gaps in cell division mutants The differential effect of azthreonam and nalidixic acid in the pattern of murein segregation, and in particular the absence of septal gaps in sacculi from D-Cys labelled cells chased in the presence of the later antibiotic, suggested that gap formation might require previous assembly of an FtsZ ring, which in turn represents the earlier discernible event in cell commitment to septation (Errington et al., 2003; Lutkenhaus and Addinall, 1997). Analysis of murein segregation in cell division and DNA replication mutants supported this idea. Gaps developed normally in sacculi of ftsA, ftsQ, and ftsI mutants chased at restrictive temperature, but neither in ftsZ nor in dnaX mutants. Hence, gap differentiation has a strict requirement for

42

de Pedro

active FtsZ protein but not for the products of ftsI (PBP3), ftsA, and ftsQ, whose activities might become essential at the later stages of septation as suggested for PBP3 (Nanninga, 1991). These results clearly indicate that septal gaps represent abortive cycles of septal murein synthesis (SMS) and raise the question: Which murein synthases are responsible for the early stages of SMS, and how is the process timed? At present there is no answer, but might be worthwhile to speculate on how SMS is regulated. The only known element of the septosome endowed with murein biosynthetic activity is PBP3 (Ayala et al., 1994; Errington et al., 2003), but in spite of its DD-transpeptidase activity PBP3 by itself is unable to make new murein because it lacks the transglycosylase activity required to polymerize monomeric precursors (Adam et al., 1997). Therefore it looks as if the role of PBP3 were the cross-linking of peptidoglycan strands made by a "helper" enzyme in such a way as to promote the inwards growth of a region of the cell wall and produce the septum. The nature of the "helper" activity involved is unknown, but is tempting to consider the class A-PBPs (PBP1A, 1B or 1C in the case of E. coli) as the "helper" proteins. Experimental results support a participation of PBP1B in cell division in E. coli (García et al., 1991; García and de Pedro, 1990). However as PBP1B mutants do indeed divide, the task of PBP1B would have to be either dispensable or taken on by a functionally redundant protein (PBP1A or 1C?). In such a scenario the class-A PBP would provide new peptidoglycan strands because of its transglycosylase activity, but the DD-transpeptidase activity would be modified or superseded by that of PBP3 in order to achieve the required change from longitudinal to transversal growth in the vicinity of the newly assembled septosomal complexes. Lack of PBP3 would still permit activation of SMS at the septosome assembly site by the class-A PBP, but new peptidoglycan strands would be inserted with the "elongation geometry" simply promoting localized longitudinal enlargement of the sacculus at the potential division site; what is detected as a "gap" in the D-Cys label-andchase experiments. A critical implication of this proposal is that assembly of the septosomal complex would trigger the change from a random distribution of biosynthetic sites in the elongation phase of cell growth, to a local accumulation of such sites. This event does not require a redistribution of sites, but rather activation (Assembly?) of new ones in the neighbourhood of the septosome. A local accumulation of active sites would explain why cells treated with lyse preferentially at active division sites (de Pedro et al., 2002; García et al., 1989; Schwarz et al., 1969). Interference with closely packed synthetic sites during SMS may favour local accumulation of murein lesions and lysis, whilst the cell may stand better the disperse lesions caused by inhibition of elongation committed biosynthetic complexes.

Topological domains in the cell wall of E. coli

43

5.2.2.5 Growth of the sacculus as a cyclic alternation between diffuse and zonal modes of synthesis; a model The experiments and interpretations presented above can be summarized in a model whose mainlines are as follows: Throughout the elongation period of the cell cycle murein precursors are randomly incorporated over the cylindrical part of the sacculus. At the initiation of septation FtsZ-ring assembly triggers a change in the mode of murein synthesis leading to localized insertion of precursors into the septal area, which under normal conditions develops into the transversal septum. Concomitantly with the in-growth of septal structures forced by the septosomal complex, murein becomes metabolically inert because of functional isolation from the metabolic enzymes. At the end of the process murein synthesis stops, the septum is split and segregates as the new poles of the offspring cells. When normal development of the septum is impeded at a stage later than the Z-ring dependent activation of SMS, the septosomal complex remains able to promote localized murein synthesis in the longitudinal direction for a fixed period of time generating a murein annulus with the same properties than polar murein. The outcome of this series of events is the generation of domains in the cell wall that can be differentiated on the basis of murein half-life, and mode of insertion. Three domains can be defined: the polar domain made up of inert, long-lived murein, the lateral wall domain where diffuse insertion of precursors and turn-over (Park, 1993; Park, 2001), make for a fast mixing of new and pre-existing murein, and the septal region where activation of zonal synthesis for limited periods of time leads to a region exclusively made up of new murein. What limits the SMS activity period? Under normal conditions one may argue that disassembly of the septosome once the septum is finished (Sun and Margolin, 1998), may in turn switch-off (Disassemble?) the SMS complexes. However in division-impaired cells growing as filaments the time-regulation of SMS is somewhat more difficult to understand. A study dedicated to define the half-life of Z-rings in cells induced to filament is missing. If the dynamics of Z-ring are maintained in filament cells, the same argument as above could be of application (Lutkenhaus and Addinall, 1997; Sun and Margolin, 1998). Some published pictures of Z-rings in filament cells from mutants in other fts genes, apparently support disassembly of Zring in the older sites (Addinall et al., 1996; Addinall et al., 1997), and some observations suggest that stability of Z-rings is negatively affected when normal progression to division is blocked (Addinall et al., 1997; Lutkenhaus and Addinall, 1997). Therefore, it could well be that the natural cycle of the Z-ring would be the "timing" device for SMS in both dividing and filament cells.

44

de Pedro

5.2.3 Segregation of D-Cys labelled murein in spherical E. coli cells Research on murein synthesis in spherical E. coli cells provided relevant information on the characteristics of SMS. A brief introductory remark may be helpful. 5.2.3.1 Penicillin-binding protein 2 and round cell shape In E. coli, the class B penicillin-binding protein 2 (PBP2) has a welldefined morphogenetic role. Impairment of PBP2 activity leads to the generation of pleiomorphic, spherical cells (Spratt, 1975; Spratt, 1977a; Spratt, 1977b). Like PBP3, PBP2 has DD-transpeptidase, but not transglycosylase, activity (Ishino et al., 1982; Ishino et al., 1986). To be functional PBP2 requires an active RodA protein (Ishino et al., 1986). The activity of PBP2 (+RodA) is thought to be involved in side-wall murein synthesis during longitudinal growth (Begg and Donachie, 1985; Canepari et al., 1984; Canepari et al., 1997; García et al., 1989; García and de Pedro, 1991). In the appropriate genetic backgrounds, mutational impairment of PBP2 and RodA leads to round cells that keep the ability to grow and divide in a stable way under appropriate conditions (Begg and Donachie, 1985; Begg and Donachie, 1998; Spratt et al., 1980; Stoker et al., 1983). An important tool to induce round cell morphology is the specific PBP2 inhibitor amdinocillin (Mecillinam, FL-1060), this blocks PBP2 and allows for a detailed study of the morphological transition from rod to sphere (James et al., 1975; Spratt, 1977b). Since PBP3 and PBP2 share some properties and are involved in discrete morphological events, it has been proposed that each could be part of murein biosynthetic complexes active at alternating periods of the cell cycle (Matsuhashi et al., 1990): PBP3 would complex to the septosome and be active exclusively at septation, and PBP2-complexes would be active at cell elongation or both elongation and division (Begg and Donachie, 1985; Botta and Park, 1981; Burman et al., 1983; Canepari et al., 1997; de Pedro et al., 1997). An implication of such models is that murein synthesized upon PBP2 impairment should be made by complexes that are normally committed to septation and therefore should have the properties of septal murein. 5.2.3.2 Murein segregation in cells with impaired PBP2-RodA activity Investigation of murein segregation by the D-Cys labelling method in cells of E. coli made spherical by interference with the activity of PBP2 produced quite amazing results. Inhibition of PBP2 results in a drastic modification of the pattern of murein synthesis. In spherical cells incorporation of new precursors occurs exclusively as a zonal process, localized at the potential division sites for as long as the cell keeps growing

Topological domains in the cell wall of E. coli

45

in mass (de Pedro et al., 2001). Consequently, there is no mixing of new and old murein. Old murein is preserved for long periods of time in large, welldefined areas and new murein is continuously inserted in a zonal manner at places corresponding to division sites which may, or may not, be proficient for actual division. From these results it has been proposed that impairment of PBP2 would result in the failure of cells to switch off SMS at the end of septation events, and revert to PBP2-dependent lateral wall murein synthesis (de Pedro et al., 2001). Round cells would be generated by suppressing lateral wall murein synthesis and keeping SMS permanently on. In the case of spherical cells, each round of SMS would end with the completion of two new hemispherical sections and, in the absence of the system for lateral wall synthesis, would be succeeded by the immediate initiation of a new round of SMS. Thus, growth and division of the sacculus would continue essentially in the same way as in natural gram-negative cocci such as Neisseria spp. (Westling-Haggstrom et al., 1977). Septation in round E. coli cells, as in Neisseria, is an asymmetric process (Begg and Donachie, 1985; Begg and Donachie, 1998; Westling-Haggstrom et al., 1977; Zaritsky et al., 1999b; Zaritsky et al., 1999a). Septal invagination starts as a lateral furrow that produces a two-lobed sacculus. The lateral furrow then progresses inwards, making the two lobes deeper until eventually two new spherical cells are released. An essentially identical series of events has been proposed on the basis of FtsZ ring assembly in spherical cells (Addinall and Lutkenhaus, 1996; Zaritsky et al., 1999b). 5.2.3.3 Murein in sacculi of PBP2-RodA impaired cells is stable A prediction from the previous considerations is that if round cells are made up exclusively by the SMS system, then their murein should be equally stable all over the surface. The experimental results apparently support this point, showing that old murein was neither mixed with new one nor subjected to turnover. Indeed, density of labelling was kept essentially constant during the chase period, and was very uniform. Sacculi from round cells had large areas of uniformly labelled murein and large areas of unlabelled material, but regions with intermediate densities were never found (de Pedro et al., 2001). Therefore it seems that the sacculus of spherical cells is essentially an inert structure as regards murein synthesis except at the SMS active sites. 5.2.3.4 PBP2 and the diffuse-zonal-diffuse transitions in the mode of murein synthesis The fact that PBP2 impaired cells are locked into the SMS mode suggests that lack of PBP2 impedes activation or assembly of elongation proficient

46

de Pedro

synthetic complexes. Continuous activity of SMS may represent an emergency bypass for the cell, perhaps triggered by some growth-monitoring device. When the cell senses unbalance between mass growth and murein biosynthetic capacity, it might overrun the inhibitory signals for termination of SMS. The dispensability of PBP2 under poor growth conditions (minimal medium, high internal ppGpp levels) supports the idea that SMS may be enough for survival in tough times (Begg and Donachie, 1998; D'Ari et al., 1988; D'Ari, 1997; Joseleau-Petit et al., 1994). However, information about the molecular basis for the shifts from diffuse to zonal and back is still missing. The requirement of Z-ring assembly for SMS activation as well as the apparent coincidence between disassembly of septosomal complexes and inhibition of SMS suggests a causal relation as indicated above. A possibility is that assemblage of the Z-ring would dictate assembly of murein biosynthetic complexes in its environs activated in a non-committed status, i.e. without associated PBP2 or PBP3. Soon after, sequential integration of other septosomal elements (FtsA, ZipA, FtsK, FtsQ, FtsL, FtsW) would make the differentiating septosome proficient to bind "mureinconnected" proteins PBP3, FtsN, and AmiC (Bernhardt and de Boer, 2003; Errington et al., 2003). This event would endow the murein synthetic complexes with the ability to make the inwards growing septum. Failure to incorporate the late proteins would prevent commitment of murein synthetic complexes that would remain active but unable to force inwards growth, therefore promoting the annular insertion of new precursors, for a defined period of time. Total or partial disassembly of the septosome, or interaction with other division regulatory elements (Min?) would be responsible for the switching off. Elongation of the sacculus could be promoted by PBP2-driven synthetic complexes that would polymerize new murein strands and insert them into the sacculus promoting elongation in an apparently diffuse and homogeneous way. However, on the assumption of a 1:1 stoichiometry for PBP2, the number of biosynthetic complexes has to be limited, in the order of a few tens, the estimated number of PBP2 molecules per cell (Spratt, 1977a). Therefore, at any particular time incorporation of precursors would take place in a small number of discrete patches of new material with a size conditioned by the stability and mobility "of the individual complexes. Computer analysis of murein segregation images captured by the D-Cys labelling technique, supports the idea of a "patchy" incorporation of new precursors (de Pedro et al., 2003). Some proposals on the elongation/septation two complex model postulated that both processes were mutually exclusive; only one kind of site would by active at a time (Canepari et al., 1997). Therefore, at the initiation

Topological domains in the cell wall of E. coli

47

of septation, elongation complexes would have to be inhibited, and reactivated at the end. However, measurements of the rate of murein synthesis throughout the cell cycle did not evidenced any abrupt variation in the rate compatible with a sudden inhibition of cell wall elongations systems, but rather supported activation of additional sites by the time of septation (de Jonge et al., 1989; Mulder and Woldringh, 1991; Wientjes and Nanninga, 1989). Thus it looks as if the activity period of the elongation complexes would be governed by the half-live of the individual complexes rather than by a periodic and general inhibition/activation cycle. 5.2.4 Segregation of D-Cys labelled murein in branching cells of E. coli

The demonstration of inert polar caps as important morphogenetic domains received important support from a series of studies about murein synthesis in branching strains of E. coli (de Pedro et al., 2003b). 5.2.4.1 The branching phenotype Certain mutant strains of E. coli generate branched cells under certain growth conditions (Denome et al., 1999; Gullbrand et al., 1999; Nelson and Young, 2000). Branching is characterized by the presence of cells with Y or X shapes and cells with kinks or buds of irregular diameter. The mechanism of branching remains unclear but is apparently unrelated to the division process: Branching points do not correlate with division sites, and neither of the essential division proteins FtsI or FtsZ is required for branch formation (Gullbrand et al., 1999). Work on the role of the low-molecular-weight PBP's led to the construction of mutant strains with a strong branching phenotype (Denome et al., 1999; Nelson and Young, 2000). Generation of morphological anomalies in these strains requires multiple mutations but is ultimately conditioned by the activity of PBP 5 a DD-carboxypeptidase I coded for by the dacA gene (Broome-Smith et al., 1988; Nicholas and Strominger, 1988). The branching phenotype is normally expressed as such in cells blocked for division. Dividing cells also develop morphological anomalies, but as they keep dividing the small cell size complicates observation.

5.2.4.2 Inert murein is associated to malformations in branching cells The most relevant observation from studies of the murein segregation patterns, in branching cells was that every single morphological anomaly was associated to a patch of inert murein (de Pedro et al., 2003b). Areas of inert murein were found at anomalous positions, often in-between the pole and the active division site of the cell. Aberrant shapes apparently developed

48

de Pedro

from these areas. Furthermore detailed electron microscopy of the polar regions in labelled-and-chased cells suggested that old poles were often split, and with certain frequency more than once. Apparently split poles were also a source for the bizarre shapes of branching cells (de Pedro et al., 2003b). The critical deduction to be made from these observations is that the branching phenotype is a consequence of the inability of the mutant cells to properly control placement and stability of those areas of the cell wall that must be made inert. Failure to do so ends up in the materialization of inert areas that could be considered as ectopic poles. Those areas could in turn alter the geometry and activity of the nearby murein biosynthetic complexes leading to the emergence of branches, bends, kinks and other morphological aberrations. Therefore, these experiments support on the one hand, the formerly claimed morphological relevance of PBP5 (Denome et al., 1999; Nelson et al., 2002; Nelson and Young, 2000; Nelson and Young, 2001) as a protein involved in the differentiation and stability of polar murein, and on the other hand show that metabolic status of the poles has a direct effect on the maintenance of the regular rod shape of E. coli.

6.

Domains of restricted protein mobility in the outer membrane of E. coli

The difference in the half-live of murein in different regions of the sacculus may have important repercussions on other structural elements of the cell wall, in particular regarding the OM. The complex and strong connections between the OM and the sacculus commented on the initial paragraphs of this chapter, suggests that mobility of OM elements could be restrained in the polar areas by interaction with the underlying inert murein. Interaction amongst OM elements and the sacculus could be homogeneous all over the cell. However, as in the lateral (cylindrical) area old and new murein are continuously blended, due to biosynthesis and recycling (Höltje, 1998; Jacobs et al., 1994), pre-existing OM molecules interacting with the sacculus would be rapidly redistributed and mixed with new ones, exhibiting an apparent freedom of movement. In contrast, molecules interacting with the inert poles would be "anchored" to those regions and would exhibit a reduced mobility. This proposal found support in a series of experiments which followed the fate of fluorescently labelled surface molecules of E. coli (de Pedro et al., 2003b; de Pedro et al., 2003a). Essentially the surface of growing cells was covalently labelled with a fluorescent reagent and cells were then allowed to grow in reagent-free medium, following the distribution of label after appropriate periods of time.

49

Topological domains in the cell wall of E. coli

The results clearly showed that mobility and turnover of labelled proteins in the poles was virtually null, whereas the rate of dilution in the rest of the cell surface matched the values expected from growth parameters. Furthermore, experiments in round and branching cells left little doubt about a very close association between regions of inert murein in the sacculus and regions of restricted protein mobility in the outer membrane (de Pedro et al., 2003b; de Pedro et al., 2003a). Murein bound Braun's lipoprotein (Lpp) was an evident candidate to mediate immobilization of OM proteins by direct protein-protein interactions (Braun, 1975; Braun and Rehn, 1969). However, analysis of Lpp deficient strains showed that lack of Lpp had no apparent influence on the behaviour of OM proteins. This observation indicates that immobilization does not depend on a single element, but rather on the joint action of all the different components able to link OM and sacculus.

7.

Domains of specific lipid composition cytoplasmic membrane of E. coli

in

the

Investigation of the possible influence of murein domains on cytoplasmic membrane (IM) elements has not been feasible yet. However, there is serious evidence showing that polar regions of the IM are also domains differentiated from the cylindrical region. Indeed, polar caps are enriched in cardiolipin (Koppelman et al., 2001; Mileykovskaya and Dowhan, 2000), a lipid which might be involved in the modulation of DNA replication through its inhibitory effect on the binding of DnaA to oriC (Ichihashi et al., 2003; Makise et al., 2002). In principle, cardiolipin rich areas would be more refractory to DNA initiation events, which would fit with DNA replication starting at a central position in the cell. The existence of additional domains in the IM heterogeneous for lipid composition has found support in the demonstration of patches of high and low affinity for lipid-specific fluorescent dyes in the cylindrical part of the IM, and in experiments indicating the presence of segregated phosphatidyl-ethanolamine and phosphatidyl-glycerol enriched domains (Fishov and Woldringh, 1999; Vanounou et al., 2003).

8.

Conclusions

The general conclusion of all the previous work is that the E. coli cell envelope cannot be considered as a functionally homogeneous structure, but rather as a mosaic of domains endowed with specific functions in bacterial physiology and morphogenesis. Murein itself is not likely to be the active

50

de Pedro

element in domains differentiation, but rather the substrate. Nevertheless, it may contribute a physically stable scaffold on which specific systems may anchor either directly or indirectly, and in this way may play a positive role in the differentiation process. The generation of topologically differentiated domains in the polar caps of the envelope might in turn be essential for the determination of cell polarity (Lybarger and Maddock, 2001; Maddock et al., 1993). The stable nature of poles may in addition represent a structural memory for the cell, possibly required for proper maintenance of cell shape through successive generations. Much has to be done yet to reach a clear understanding of the molecular mechanisms underlying shape generation in bacteria, but to our satisfaction, it is also true that the effort of the many researchers working in the field is producing such a wealth of knowledge that, for the first time, is realistic to thing such a goal is at reach.

References Abergel, C., Walburger, A., Chenivesse, S., and Lazdunski, C. (2001). Crystallization and preliminary crystallographic study of the peptidoglycan-associated lipoprotein from Escherichia coli. Acta Crystallogr. D Biol. Cryst. 57, 317-319. Adam, M., Fraipont, C., Rhazi, N., Nguyen-Disteche, M., Lakaye, B., Frere, J.M., Devreese, B., Van Beeumen, J., van Heijenoort, Y., van Heijenoort, J., and Ghuysen, J.M. (1997). The bimodular G57-V577 polypeptide chain of the class B penicillin- binding protein 3 of Escherichia coli catalyzes peptide bond formation from thiolesters and does not catalyze glycan chain polymerization from the lipid II intermediate. J. Bacteriol. 179, 6005-6009. Addinall, S.G., Bi, E., and Lutkenhaus, J. (1996). FtsZ ring formation in fts mutants. J. Bacteriol. 178, 3877-3884. Addinall, S.G., Cao, C., and Lutkenhaus, J. (1997). Temperature shift experiments with an ftsZ84(Ts) strain reveal rapid dynamics of FtsZ localization and indicate that the Z ring is required throughout septation and cannot reoccupy division sites once constriction has initiated. J. Bacteriol. 179, 4277-4284. Addinall, S.G. and Lutkenhaus, J. (1996). FtsZ-spirals and -arcs determine the shape of the invaginating septa in some mutants of Escherichia coli. Mol. Microbiol. 22, 231-237. Altmutter, C. Mass spectrometric characterization of therminal muropeptides isolated from the cell wall of Escherichia coli. 2002. Universität Wien. Master Thesis. Ayala, J.A., Garrido, T., de Pedro, M.A., and Vicente, M. (1994). Molecular biology of bacterial septation. In Bacterial cell wall. (J.M. Ghuysen and R. Hakenbeck, eds.), Elsevier, Amsterdam, 73-101. Begg, K.J. and Donachie, W.D. (1985). Cell shape and division in Escherichia coli: experiments with shape and division mutants. J. Bacteriol. 163, 615-622. Begg, K.J. and Donachie, W.D. (1998). Division planes alternate in spherical cells of Escherichia coli. J. Bacteriol. 180, 2564-2567. Bernhardt, T.G. and de Boer, P.A. (2003). The Escherichia coli amidase AmiC is a periplasmic septal ring component exported via the twin-arginine transport pathway. Mol. Microbiol. 48, 1171-1182.

Topological domains in the cell wall of E. coli

51

Blasco, B., Pisabarro, A.G., and de Pedro, M.A. (1988). Peptidoglycan biosynthesis in stationary-phase cells of Escherichia coli. J. Bacteriol. 170, 5224-5228. Botta, G.A. and Park, J.T. (1981). Evidence for involvement of penicillin-binding protein 3 in murein synthesis during septation but not during cell elongation. J. Bacteriol. 145, 333340. Bouveret, E., Benedetti, H., Rigal, A., Loret, E., and Lazdunski, C. (1999). In vitro characterization of peptidoglycan-associated lipoprotein (PAL)-peptidoglycan and PALTolB interactions. J. Bacteriol. 181, 6306-6311. Braun, V. (1975). Covalent lipoprotein from the outer membrane of Escherichia coli. Biochim. Biophys. Acta 415, 335-377. Braun, V. and Rehn, K. (1969). Chemical characterization, spatial distribution and function of a lipoprotein (murein-lipoprotein) of the E. coli cell wall. The specific effect of trypsin on the membrane structure. Eur. J. Biochem. 10, 426-438. Braun, V. and Wu, H.C. (1994). Lipoproteins, structure, function, biosynthesis and model for protein export. In Bacterial cell wall. (J.M. Ghuysen and R. Hakenbeck, eds.), Elsevier Science Publisher, Amsterdam, 319-341. Broome-Smith, J.K., Ioannidis, I., Edelman, A., and Spratt, B.G. (1988). Nucleotide sequences of the penicillin-binding protein 5 and 6 genes of Escherichia coli. Nucleic Acids Res. 16, 1617. Burman, L.G., Raichler, J., and Park, J.T. (1983). Evidence for diffuse growth of the cylindrical portion of the Escherichia coli murein sacculus. J. Bacteriol. 155, 983-988. Canepari, P., Botta, G., and Satta, G. (1984). Inhibition of lateral wall elongation by mecillinam stimulates cell division in certain cell division conditional mutants of Escherichia coli. J. Bacteriol. 157, 130-133. Canepari, P., Signoretto, C., Boaretti, M., and Del Mar, L. (1997). Cell elongation and septation are two mutually exclusive processes in Escherichia coli. Arch. Microbiol. 168, 152-159. Caparrós, M., Pisabarro, A.G., and de Pedro, M.A. (1992). Effect of D-amino acids on structure and synthesis of peptidoglycan in Escherichia coli. J. Bacteriol. 174, 5549-5559. Caparrós, M., Pittenauer, E., Schmid, E.R., de Pedro, M.A., and Allmaier, G. (1993a). Molecular weight-determination of biosynthetically modified monomeric and oligomeric muropeptides from Escherichia coli by plasma desorption-mass spectrometry. FEBS Lett. 316, 181-185. Caparrós, M., Quintela, J.C., Leguina, J.I., and de Pedro, M.A. (1993b). Ammo acids as useful tools in the study of murein metabolism in Escherichia coli. In Bacterial growth and lysis. (M.A. de Pedro, J.V. Höltje, and W. Loffelhardt, eds.), Plenum, London, 167175. Charpentier, X., Chalut, C., Remy, M. H., and Masson, J. M. (2002). Penicillin-binding proteins 1a and 1b form independent dimers in Escherichia coli. J. Bacteriol. 184, 37493752. Choi, D.S., Yamada, H., Mizuno, T., and Mizushima, S. (1986). Trimeric structure and localization of the major lipoprotein in the cell surface of Escherichia coli. J. Biol. Chem. 261, 8953-8957. Choi, D.S., Yamada, H., Mizuno, T., and Mizushima, S. (1987). Molecular assembly of the lipoprotein trimer on the peptidoglycan layer of Escherichia coli. J. Biochem. (Tokyo) 102, 975-983. Cooper, S. (1991). Bacterial growth and division. Academic Press, Inc., San Diego, California. Costa, K., Bacher, G., Allmaier, G., Dominguez-Bello, M. G., Engstrand, L., Falk, P., dePedro, M. A., and Garcia del Portillo, F. (1999). The morphological transition of

52

de Pedro

Helicobacter pylori cells from spiral to coccoid is preceded by a substantial modification of the cell wall. J. Bacteriol. 181, 3710-3715. D'Ari, R. (1997). The Escherichia coli cell cycle, cell division and ppGpp: regulation and mechanisms. Folia Microbiol. (Praha), 42, 161-164. D'Ari, R., Jaffe, A., Bouloc, P., and Robin, A. (1988). Cyclic AMP and cell division in Escherichia coli. J. Bacteriol. 170, 65-70. de Boer, P.A., Crossley, R.E., and Rothfield, L.I. (1989). A division inhibitor and a topological specificity factor coded for by the minicell locus determine proper placement of the division septum in E. coli. Cell 56, 641-649. de Jonge, B.L., Wientjes, F.B., Jurida, I., Driehuis, F., Wouters, J.T., and Nanninga,N. (1989). Peptidoglycan synthesis during the cell cycle of Escherichia coli: composition and mode of insertion. J. Bacteriol. 171, 5783-5794. de Pedro, M.A., Donachie, W.D., Höltje, J.V., and Schwarz, H. (2001). Constitutive septal murein synthesis in Escherichia coli with impaired activity of the morphogenetic proteins RodA and Penicillin-Binding Protein 2. J. Bacteriol. 183, 4115-4126. de Pedro, M. A., Grünfelder, D. G., and Schwarz, H. (2003a). Restricted mobility of cell surface proteins in the polar regions of Escherichia coli. J. Bacteriol. (Submitted). de Pedro, M.A., Höltje, J.V., and Schwarz, H., (2002). Fast lysis of Escherichia coli filament cells requires differentiation of potential division sites. Microbiology 148, 79-86. de Pedro, M.A., Quintela, J.C., Höltje, J.V., and Schwarz, H. (1997). Murein segregation in Escherichia coli. J. Bacteriol. 179, 2823-2834. de Pedro, M.A. and Schwarz, U. (1981). Heterogeneity of newly inserted and preexisting murein in the sacculus of Escherichia coli. Proc. Natl. Acad. Sci. USA 78, 5856-5860. de Pedro, M.A., Young, K.D., Höltje, J.V., and Schwarz, H. (2003b). Branching of Escherichia coli cells arises from multiple sites of inert peptidoglycan. J. Bacteriol. 185, 1147-1152. de Pedro, M.A., Schwarz, H., and Koch, A.L. (2003). Patchiness of murein insertion into the sidewall of Escherichia coli. Microbiology 149, 1753. Denome, S.A., Elf, P.K., Henderson, T.A., Nelson, D.E., and Young, K.D. (1999). Escherichia coli mutants lacking all possible combinations of eight penicillin-binding proteins: Viability, characteristics, and implications for peptidoglycan synthesis. J. Bacteriol. 181, 3981-3993. Donachie, W.D. (1968). Relationship between cell size and time of initiation of DNA replication. Nature 219, 1077-1079. Donachie, W.D. (1993). The cell cycle of Escherichia coli. Annu. Rev. Microbiol. 47, 199230. Donachie, W.D., Begg, K.J., and Vicente, M. (1976). Cell length, cell growth and cell division. Nature 264, 328-333. Errington, J., Daniel, R.A., and Scheffers, D.J. (2003). Cytokinesis in bacteria. Microbiol. Mol. Biol. Rev. 67, 52-65. Fishov, I. and Woldringh, C.L. (1999). Visualization of membrane domains in Escherichia coli. Mol. Microbiol. 32, 1166-1172. Foley, M., Brass, J.M., Birmingham, J., Cook, W.R., Garland, P.B., Higgins, C.F., and Rothfield, L.I. (1989). Compartmentalization of the periplasm at cell division sites in Escherichia coli as shown by fluorescence photobleaching experiments. Mol. Microbiol. 3, 1329-1336. Garcia del Portillo, F. and de Pedro, M.A. (1990). Differential effect of mutational impairment of penicillin-binding proteins 1A and 1B on Escherichia coli strains harboring thermosensitive mutations in the cell division genes ftsA, ftsQ, ftsZ, and pbpB. J. Bacteriol. 172, 5863-5870.

Topological domains in the cell wall of E. coli

53

García del Portillo, F. and de Pedro, M.A. (1991). Penicillin-binding protein 2 is essential for the integrity of growing cells of Escherichia coli ponB strains. J. Bacteriol. 173, 45304532. García del Portillo, F., de Pedro, M.A., and Ayala, J.A. (1991). Identification of a new mutation in Escherichia coli that suppresses a pbpB (Ts) phenotype in the presence of penicillin-binding protein 1B. FEMS Microbiol. Lett. 68, 7-13. García del Portillo, F., de Pedro, M.A., Joseleau-Petit, D., and D'Ari, R. (1989). Lytic response of Escherichia coli cells to inhibitors of penicillin- binding proteins la and 1b as a timed event related to cell division. J. Bacteriol. 171, 4217-4221. Georgopapadakou, N.H., Smith, S.A., and Sykes, R.B. (1982). Mode of action of azthreonam. Antimicrob. Agents Chemother. 21, 950-956. Ghuysen, J.M. (1994). Molecular structures of penicillin-binding proteins and betalactamases. Trends Microbiol. 2, 372-380. Glauner, B. and Höltje, J.V. (1990). Growth pattern of the murein sacculus of Escherichia coli. J. Biol. Chem. 265, 18988-18996. Glauner, B., Höltje, J.V., and Schwarz, U. (1988). The composition of the murein of Escherichia coli. J. Biol. Chem. 263, 10088-10095. Gullbrand, B., Akerlund, T., and Nordstrom, K. (1999). On the origin of branches in Escherichia coli. J. Bacteriol. 181, 6607-6614. Hancock, R.E.W., Nedra Kanmaratne, D., and Bernegger-Egli, C. (1994). Molecular organization and structural role of outer membrane macromolecules. In Bacterial cell wall. (J.M. Ghuysen and R. Hakenbeck, eds.), Elsevier, Amsterdam, 263-279. Harz, H., Burgdorf, K., and Höltje, J.V. (1990). Isolation and separation of the glycan strands from murein of Escherichia coli by reversed-phase high-performance liquid chromatography. Anal. Biochem. 190, 120-128. Heidrich, C., Ursinus, A., Berger, J., Schwarz, H., and Höltje, J.V. (2002). Effects of multiple deletions of murein hydrolases on viability, septum cleavage, and sensitivity to large toxic molecules in Escherichia coli. J. Bacteriol. 184, 6093-6099. Höltje, J.V. (1996). A hypothetical holoenzyme involved in the replication of the murein sacculus of Escherichia coli. Microbiology 142, 1911-1918. Höltje, J.V. (1998). Growth of the stress-bearing and shape-maintaining murein sacculus of Escherichia coli. Microbiol. Mol. Biol. Rev. 62, 203. Höltje, J.V. and Heidrich, C. (2001). Enzymology of elongation and constriction of the murein sacculus of Escherichia coli. Biochimie 83, 103-108. Hooper, D. C. (2001). Mechanisms of action of antimicrobials: Focus on fluoroquinolones. Clin. Infect. Dis. 32, S9-S15. Ichihashi, N., Kurokawa, K., Matsuo, M., Kaito, C., and Sekimizu, K. (2003). Inhibitory effects of basic or neutral phospholipid on acidic phospholipid-mediated dissociation of adenine nucleotide bound to DnaA protein, the initiator of chromosomal DNA replication. J. Biol. Chem. 278, 28778-28786. Inouye, M., Shaw, J., and Shen, C. (1972). The assembly of a structural lipoprotein in the envelope of Escherichia coli. J. Biol. Chem. 247, 8154-8159. Ishidate, K., Ursinus, A., Höltje, J.V., and Rothfield, L. (1998). Analysis of the length distribution of murein glycan strands in ftsZ and ftsI mutants of E. coli. FEMS Microbiol. Lett. 168, 71-75. Ishino, F., Park, W., Tomioka, S., Tamaki, S., Takase, I., Kunugita, K., Matsuzawa, H., Asoh, S., Ohta, T., and Spratt, B.G. (1986). Peptidoglycan synthetic activities in membranes of Escherichia coli caused by overproduction of Penicillin-binding Protein 2 and RodA protein. J. Biol. Chem. 261, 7024-7031.

54

de Pedro

Ishino, F., Tamaki, S., Spratt, B.G., and Matsuhashi, M. (1982). A mecillinam-sensitive peptidoglycan crosslinking reaction in Escherichia coli. Biochem. Biophys. Res. Commun. 109, 689-696. Jacobs, C., Huang, L.J., Bartowsky, E., Normark, S., and Park, J.T. (1994). Bacterial cell wall recycling provides cytosolic muropeptides as effectors for beta-lactamase induction. EMBO J. 13, 4684-4694. Jaffe, A., D'Ari, R., and Hiraga, S. (1988). Minicell-forming mutants of Escherichia coli: production of minicells and anucleate rods. J. Bacteriol. 170, 3094-3101. James, R., Haga, J., and Pardee, A.B. (1975). Inhibition of an early event in the cell division cycle of Escherichia coli by FL1060, an amidinopenicillanic acid. J. Bacteriol. 122, 12831292. Joseleau-Petit, D., Thevenet, D., and D'Ari, R. (1994). ppGpp concentration, growth without PBP2 activity, and growth-rate control in Escherichia coli. Mol. Microbiol. 13, 911-917. Kellenberger, E. (1990). The 'Bayer bridges' confronted with results from improved electron microscopy methods. Mol. Microbiol. 4, 697-705. Koch, A.L. (1983). The surface stress theory of microbial morphogenesis. Adv. Microb. Physiol. 24, 301-366. Koch, A.L. (1988). Biophysics of bacterial walls viewed as stress-bearing fabric. Microbiol. Rev. 52, 337-353. Koch, A.L. (1990). The surface stress theory for the case of Escherichia coli: the paradoxes of gram-negative growth. Res. Microbiol. 141, 119-130. Koch, A.L. (1998). Orientation of the peptidoglycan chains in the sacculus of Escherichia coli. Res. Microbiol. 149, 689-701. Koch, A. L. (2000). The exoskeleton of bacterial cells (The sacculus): Still a highly attractive target for antibacterial agents that will last for a long time. Crit. Rev. Microbiol. 26, 1-35. Koch, A.L. and Woeste, S. (1992). Elasticity of the sacculus of Escherichia coli. J. Bacteriol. 174, 4811-4819. Koch, A.L. and Woldringh, C.L. (1994). The metabolic inertness of the pole wall of a Gram negative rod. J. Theor. Biol. 171, 415-425. Kohlrausch, U. and Höltje, J.V. (1991). Analysis of murein and murein precursors during antibiotic-induced lysis of Escherichia coli. J. Bacteriol. 173, 3425-3431. Kohlrausch, U., Wientjes, F.B., and Höltje, J.V. (1989). Determination of murein precursors during the cell cycle of Escherichia coli. J. Gen. Microbiol. 135, 1499-1506. Koppelman, C. M., Den Blaauwen, T., Duursma, M. C., Heeren, R. M. A., and Nanninga, N. (2001). Escherichia coli minicell membranes are enriched in cardiolipin. J. Bacteriol. 183, 6144-6147. Labischinski, H. and Maidhof, H. (1994). Bacterial peptidoglycan: overview and evolving concepts. In Bacterial cell wall. (J.M. Ghuysen and R. Hakenbeck, eds.), Elsevier, Amsterdam, 23-38. Labischinski, H., Goodell, E.W., Goodell, A., and Hochberg, M.L. (1991). Direct proof of a "more-than-single-layered" peptidoglycan architecture of Escherichia coli W7: a neutron small-angle scattering study. J. Bacteriol. 173, 751-756. Lazzaroni, J.C. and Portalier, R. (1992). The excC gene of Escherichia coli K-12 required for cell envelope integrity encodes the peptidoglycan-associated lipoprotein (PAL). Mol. Microbiol. 6, 735-742. Lleo, M.M., Canepari, P., and Satta, G. (1990). Bacterial cell shape regulation: testing of additional predictions unique to the two-competing-sites model for peptidoglycan assembly and isolation of conditional rod-shaped mutants from some wild-type cocci. J. Bacteriol. 172, 3758-3771.

Topological domains in the cell wall of E. coli

55

Lutkenhaus, J. and Addinall, S.G. (1997). Bacterial cell division and the Z ring. Annu. Rev. Biochem. 66, 93-116. Lutkenhaus, J.F. (1983). Coupling of DNA replication and cell division: sulB is an allele of ftsZ. J. Bacteriol. 154, 1339-1346. Lybarger, S. R. and Maddock, J. R. (2001). Polarity in action: Asymmetric protein localization in bacteria. J. Bacteriol. 183, 3261-3267. Maddock, J.R., Alley, M.R., and Shapiro, L. (1993). Polarized cells, polar actions. J. Bacteriol. 175, 7125-7129. Makise, M., Mima, S., Katsu, T., Tsuchiya, T, and Mizushima, T. (2002). Acidic phospholipids inhibit the DNA-binding activity of DnaA protein, the initiator of chromosomal DNA replication in Escherichia coli. Mol. Microbiol. 46, 245-256. Markiewicz, Z. and Höltje, J.V. (1992). Failure to trigger the autolytic enzymes in minicells of Escherichia coli. FEMS Microbiol. Lett. 70, 119-123. Matias, V.R., Al Amoudi, A., Dubochet, J., and Beveridge, T.J. (2003). Cryo-transmission electron microscopy of frozen-hydrated sections of Escherichia coli and Pseudomonas aeruginosa. J. Bacteriol. 185, 6112-6118. Matsuhashi, M. (1994). Utilization of lipid-linked precursors and the formation of peptidoglycan in the process of cell growth and division: membrane enzymes involved in the final steps of peptidoglycan synthesis and the mechanism of their regulation. In Bacterial cell wall. (J.M. Ghuysen and R. Hakenbeck, eds.), Elsevier, Amsterdam, 55102. Matsuhashi, M., Wachi, M., and Ishino, F. (1990). Machinery for cell growth and division: penicillin-binding proteins and other proteins. Res. Microbiol. 141, 89-103. Mileykovskaya, E. and Dowhan, W. (2000). Visualization of phospholipid domains in Escherichia coli by using the cardiolipin-specific fluorescent dye 10-N-nonyl acridine orange.J. Bacteriol. 182, 1172-1175. Mizuno, T., Kagiyama, R., and Kageyama, M. (1982). The peptidoglycan-associated lipoprotein (PAL) of the Proteus mirabilis outer membrane: characterization of the peptidoglycan-associated region of PAL. J. Biochem. (Tokyo) 91, 19-24. Mizushima, S. (1987). Assembly of membrane proteins. In Bacterial outer membranes as model systems. (M. Inouye, ed.), John Wiley & Sons, New York, 163-185. Mulder, E. and Woldringh, C.L. (1991). Autoradiographic analysis of diaminopimelic acid incorporation in filamentous cells of Escherichia coli: repression of peptidoglycan synthesis around the nucleoid. J. Bacteriol. 173, 4751-4756. Mulder, E. and Woldringh, C.L. (1993). Plasmolysis bays in Escherichia coli: are they related to development and positioning of division sites? J. Bacteriol. 175, 2241-2247. Nanninga, N. (1988). Growth and form in microorganisms: morphogenesis of Escherichia coli. Can. J. Microbiol. 34, 381-389. Nanninga, N. (1991). Cell division and peptidoglycan assembly in Escherichia coli. Mol. Microbiol. 5, 791-795. Nanninga, N. (2001). Cytokinesis in prokaryotes and eukaryotes: Common principles and different solutions. Microbiol. Mol. Biol. Rev. 65, 319. Nelson, D. E., Ghosh, A. S., Paulson, A. L., and Young, K. D. (2002). Contribution of membrane-binding and enzymatic domains of penicillin binding protein 5 to maintenance of uniform cellular morphology of Escherichia coli. J. Bacteriol. 184, 3630-3639. Nelson, D. E. and Young, K. D. (2000). Penicillin binding protein 5 affects cell diameter, contour, and morphology of Escherichia coli. J. Bacteriol. 182, 1714-1721. Nelson, D. E. and Young, K. D. (2001). Contributions of PBP 5 and DD-carboxypeptidase penicillin binding proteins to maintenance of cell shape in Escherichia coli. J. Bacteriol. 183, 3055-3064.

56

de Pedro

Nicholas, R.A. and Strominger, J.L. (1988). Site-directed mutants of a soluble form of penicillin-binding protein 5 from Escherichia coli and their catalytic properties. J. Biol. Chem. 263, 2034-2040. Nikaido, H. and Vaara, M (1985). Molecular basis of bacterial outer membrane permeability. Microbiol. Rev. 49, 1-32. Obermann, W. and Höltje, J.V. (1994). Alterations of murein structure and of penicillinbinding proteins in minicells from Escherichia coli. Microbiology 140, 79-87. Osborn, M.J. and Wu, H.C. (1980). Proteins of the outer membrane of gram-negative bacteria. Annu. Rev. Microbiol. 34, 369-422. Park, J.T. (1993). Turnover and recycling of the murein sacculus in oligopeptide permeasenegative strains of Escherichia coli: indirect evidence for an alternative permease system and for a monolayered sacculus. J. Bacteriol. 175, 7-11. Park, J. T. (2001). Identification of a dedicated recycling pathway for anhydro-Nacetylmuramic acid and N-acetylglucosamine derived from Escherichia coli cell wall murein. J. Bacteriol. 183, 3842-3847. Pisabarro, A.G., de Pedro, M.A., and Vazquez, D. (1985). Structural modifications in the peptidoglycan of Escherichia coli associated with changes in the state of growth of the culture. J. Bacteriol. 161, 238-242. Prats, R. and de Pedro, M.A. (1989). Normal growth and division of Escherichia coli with a reduced amount of murein. J. Bacteriol. 171, 3740-3745. Quintela, J.C., Caparrós, M., and de Pedro, M.A. (1995). Variability of peptidoglycan structural parameters in gram-negative bacteria. FEMS Microbiol. Lett. 125, 95-100. Satta, G., Canepari, P., Maurici, R., and Marcialis, M.A. (1985). Interactions between lateral wall elongation and septum formation during cell cycle in Klebsiella pneumoniae. Ann. Inst. Pasteur Microbiol. 136A, 85-89. Satta, G., Fontana, R., and Canepari, P. (1994). The two-competing site (TCS) model for cell shape regulation in bacteria: the envelope as an integration point for the regulatory circuits of essential physiological events. Adv. Microb. Physiol. 36, 181-245. Schleifer, K.H. and Kandler, O. (1972). Peptidoglycan types of bacterial cell walls and their taxonomic implications. Bacteriol. Rev. 36, 407-477. Schwarz, U., Asmus, A., and Frank, H. (1969). Autolytic enzymes and cell division of Escherichia coli. J. Mol. Biol. 41, 419-429. Shockman, G.D. and Höltje, J.V. (1994). Microbial peptidoglycan (murein) hydrolases. In Bacterial cell wall. (J.M. Ghuysen and R. Hakenbeck, eds.), Elsevier, Amsterdam, 131166. Sonntag, I., Schwarz, H., Hirota, Y., and Henning, U. (1978). Cell envelope and shape of Escherichia coli: multiple mutants missing the outer membrane lipoprotein and other major outer membrane proteins. J. Bacteriol. 136, 280-285. Spratt, B.G. (1975). Distinct penicillin binding proteins involved in the division, elongation, and shape of Escherichia coli K12. Proc. Natl. Acad. Sci. USA, 72, 2999-3003. Spratt, B.G. (1977a). Properties of the penicillin-binding proteins of Escherichia coli K12. Eur. J. Biochem. 72, 341-352. Spratt, B.G. (1977b). The mechanism of action of mecillinam. J. Antimicrob. Chemother. 3 Suppl B, 13-19. Spratt, B.G., Boyd, A., and Stoker, N. (1980). Defective and plaque-forming lambda transducing bacteriophage carrying penicillin-binding protein-cell shape genes: genetic and physical mapping and identification of gene products from the lip-dacA-rodA-pbpAleuS region of the Escherichia coli chromosome. J. Bacteriol. 143, 569-581. Spratt, B.G. and Cromie, K.D. (1988). Penicillin-binding proteins of gram-negative bacteria. Rev. Infect. Dis. 10, 699-711.

Topological domains in the cell wall of E. coli

57

Stoker, N.G., Broome-Smith, J.K., Edelman, A., and Spratt, B.G. (1983). Organization and subcloning of the dacA-rodA-pbpA cluster of cell shape genes in Escherichia coli. J. Bacteriol. 155, 847-853. Sun, Q. and Margolin, W. (1998). FtsZ dynamics during the division cycle of live Escherichia coli cells. J. Bacteriol. 180, 2050-2056. Tsuruoka, T., Tamura, A., Miyata, A., Takei, T., Inouye, S., and Matsuhashi, M. (1985). Second lytic target of beta-lactam compounds that have a terminal D-amino acid residue. Eur. J. Biochem. 151, 209-216. Tsuruoka, T., Tamura, A., Miyata, A., Takei, T., Iwamatsu, K., Inouye, S., and Matsuhashi, M. (1984). Penicillin-insensitive incorporation of D-amino acids into cell wall peptidoglycan influences the amount of bound lipoprotein in Escherichia coli. J. Bacteriol. 160, 889-894. Vanounou, S., Parola, A.H., and Fishov, I. (2003). Phosphatidylethanolamine and phosphatidylglycerol are segregated into different domains in bacterial membrane. A study with pyrene-labelled phospholipids. Mol. Microbiol. 49, 1067-1079. Verwer, R.W. and Nanninga, N. (1980). Pattern of meso-dl-2,6-diaminopimelic acid incorporation during the division cycle of Escherichia coli. J. Bacteriol. 144, 327-336. Verwer, R.W., Nanninga, N., Keck, W., and Schwarz, U. (1978). Arrangement of glycan chains in the sacculus of Escherichia coli. J. Bacteriol. 136, 723-729. Vollmer, W. and Höltje, J. V. (2001). Morphogenesis of Escherichia coli. Curr. Opin. Microbiol. 4, 625-633. Vollmer, W., von Rechenberg, M., and Höltje, J.V. (1999). Demonstration of molecular interactions between the murein polymerase PBP1B, the lytic transglycosylase MltA, and the scaffolding protein MipA of Escherichia coli. J. Biol. Chem. 274, 6726-6734. Weidel, W. and Pelzer, H. (1964). Bagshaped macromolecules - a new outlook on bacterial cell walls. Adv. Enzymol. 26, 193-232. Westling-Haggstrom, B., Elmros, T., Normark, S., and Winblad, B. (1977). Growth pattern and cell division in Neisseria gonorrhoeae. J. Bacteriol. 129, 333-342. Wientjes, F.B. and Nanninga, N. (1989). Rate and topography of peptidoglycan synthesis during cell division in Escherichia coli: concept of a leading edge. J. Bacteriol. 171, 34123419. Wientjes, F.B., Woldringh, C.L., and Nanninga, N. (1991). Amount of peptidoglycan in cell walls of gram-negative bacteria. J. Bacteriol. 173, 7684-7691. Woldringh, C.L. (2002). The role of co-transcriptional translation and protein translocation (transertion) in bacterial chromosome segregation. Mol. Microbiol. 45, 17-29. Woldringh, C.L., Mulder, E., Huls, P.G., and Vischer, N. (1991). Toporegulation of bacterial division according to the nucleoid occlusion model. Res. Microbiol. 142, 309-320. Woldringh, C.L., Mulder, E., Valkenburg, J.A., Wientjes, F.B., Zaritsky, A., and Nanninga, N. (1990). Role of the nucleoid in the toporegulation of division. Res. Microbiol. 141, 3949. Yao, X., Jericho, M., Pink, D., and Beveridge, T. (1999). Thickness and elasticity of gramnegative murein sacculi measured by atomic force microscopy. J. Bacteriol. 181, 68656875. Yem, D.W. and Wu, H.C. 1978. Physiological characterization of an Escherichia coli mutant altered in the structure of murein lipoprotein. J. Bacteriol. 133, 1419-1426. Yu, F., Furukawa, H., Nakamura, K., and Mizushima, S. (1984). Mechanism of localization of major outer membrane lipoprotein in Escherichia coli. Studies with the OmpF-lipoprotein hybrid protein. J. Biol. Chem. 259, 6013-6018.

58

de Pedro

Zaritsky, A., Van Geel, A., Fishov, I., Pas, E., Einav, M., and Woldringh, C.L. (1999a). Visualizing multiple constrictions in spheroidal Escherichia coli cells. Biochimie, 81, 897-900. Zaritsky, A., Woldringh, C.L., Fishov, I., Vischer, N.O.E., and Einav, M. (1999b). Varying division planes of secondary constrictions in spheroidal Escherichia coli cells. Microbiology 145, 1015-1022.

Chapter 4 Models for pattern formation in bacteria applied to bacterial morphogenesis

HANS MEINHARDT Max-Planck-Institut für Entwicklungsbiologie Spemannstrasse 35, D-72076 Tübingen Germany

1.

Introduction

In bacterial morphogenesis many components become reliably localized at particular positions of the cell (for review see Shapiro and Losick, 2000). How is such localization accomplished? The assumption that spatial determinants exist that direct this localization leads to a circular argument. Recent observations have shown that pattern formation within bacterial cells can be a highly dynamic process. A very impressive example is the rapid pole-to-pole oscillation of MinD, one of the key players for the determination of the division plane in Escherichia coli (Raskin and de Boer, 1999b, Hu and Lutkenhaus, 1999). Although most if not all components for positioning of the division plane were known, it was unclear how the components work together to accomplish the task. The properties of a complex system cannot be directly deduced from the components. By modelling integrating properties of the known components we have shown what type of interaction could in principle account for the observed phenomena (Meinhardt and de Boer, 2001). This modelling provides a bottom-up approach that supplements the experimental observation. The models can be checked, modified if necessary, and checked again. In this chapter, the molecular interactions that account for the formation of stable patterns without pre-localized determinants will be discussed first.

Molecules in Time and Space, Edited by Vicente et al. Kluwer Academic/Plenum Publishers, New York 2004

59

60

Meinhardt

Figure 1. Pattern formation by an activator-inhibitor interaction. (A) Reaction scheme. Assumed is a substance, the activator, which feeds back to increase its own concentration. The local autocatalyis is balanced by an inhibitor whose production is under activator control. (B) In fields of the size of the activator range, i.e., in fields much smaller than the range of the inhibitor, a high concentration can appear only at one marginal position. (C) Upon growth new maxima become inserted whenever the inhibitor concentration becomes too low. The polar pattern changes into a symmetric pattern with peaks at each terminal position. Upon further growth, periodic patterns can emerge. Calculations are done in a linear arrangement of spatial elements ('cells'). Minute fluctuations in the local ability to perform the self-enhancing reaction or in the initial distribution of the self-enhancing component are sufficient to initiate pattern formation. PC-programs for such simulations are available (Meinhardt, 2003).

These models were originally developed to explain pattern formation in higher organisms. In the second part it is shown that local destabilization of once generated maxima by additional antagonistic components leads to highly dynamic patterns. The pole-to-pole oscillation in E. coli mentioned above can be described in this way. In a third part, it will be shown that the formation of a narrow belt-like structure such as the FtsZ-ring is a difficult patterning process and possible solutions are elaborated.

2.

The driving force for stable patterning: short-range self-enhancement and long-range inhibition

The possibility of generating patterns by the interaction of two substances with different diffusion rates was discovered by Turing (1952). However, a different spread of two interacting substances does not guarantee the capability to form a pattern. In fact, only a very restricted class of interactions allows pattern formation. The crucial condition is that a local self-enhancing reaction is coupled with an antagonistic reaction of longer range (Gierer and Meinhardt, 1972, Meinhardt, 1982, Meinhardt and Gierer,

Models for bacterial pattern formation

61

2000). The range is the average distance a molecule can move between its birth and decay. For an intuitive understanding of the underlying mechanism it is helpful to realize that pattern formation from almost uniform initial situations is by no means unique to living systems. In the desert, high sand dunes are formed. Naively one would expect that the permanent redistribution of the sand by the wind would lead to a uniform distribution. However, the opposite is true: it is the wind that gives rise to the sand dunes. A uniform distribution of the sand represents an unstable situation. A small elevation creates a wind shelter behind which more sand is deposited, the incipient wind shelter grows further, and so on. A similar argument can be made for other inorganic patterning processes, like the formation of valleys by erosion, of clouds, galaxies and lightning. A common point in all these processes is that small deviations from a homogeneous distribution have a strong positive feedback such that the deviations grow further while longranging antagonistic effects lead to a local restriction and stabilization of the resulting patterns. A simple molecular realization of a self-enhancing (autocatalytic) process would consist of a single molecular species we called the "activator" that has a positive feedback on its own production. Of course, the term "activator" may also subsume a more complex chain of interactions; essential is only the self-enhancement. Since for bacterial morphogenesis patterning has to take place within a single cell, the spread of the self-enhancing component must be small compared with the extension of the cell, otherwise any pattern would be smoothed out. Therefore, it is assumed that the self-enhancing process takes place at the cell membrane. To make sure that this self-enhancement remains locally restricted, two types of antagonistic reactions can be envisaged: (i) The activator controls the production or the release of a rapidly diffusing "inhibitor" which, in turn, suppresses the activator in a larger region (Fig. 1). (ii) The antagonistic effect results from the depletion of a substrate, precursor or co-factor that is required for the autocatalysis (Fig. 2). Both reactions have somewhat different properties. In higher organisms several activator-inhibitor systems are known. An example is the Nodal-lefty2 system (Cheng et al., 2000; Meno et al., 2000). The components involved in bacterial morphogenesis found so far suggest, however, that the reactions are of the activator depleted-substrate type. The following discussion is, therefore, mainly focused on the later mechanism. It is assumed that the autoregulatory process consist in the accretion of precursor molecules to the membrane and that this is facilitated by molecules of the same type that are already attached. Necessarily, any attachment of precursor molecules to the membrane leads to a lowering of the pool of precursor molecules. The

62

Meinhardt

Figure 2. Pattern formation by an activator-depleted substance (precursor) model. In contrast to the reaction simulated in Fig. 1, the antagonistic reaction results from the depletion of a substance required for the autocatalytic reaction. (A) At a small field size only a marginal activation is possible. The pattern is self-regulating. After removal of the activated region (arrow), the substrate concentration rises steeply in the remaining field, causing a new activator maximum to be formed. (B) In a growing field, the marginal activation becomes unstable. The activation shifts to the centre since there the precursors can be obtained from both sides. Upon further growth, the maximum splits into two, and so on (Meinhardt, 1982; 2003).

depletion of the non-attached precursor molecules that move freely in the cytoplasm provides a natural long-ranging antagonistic reaction. In such a scheme the crucial condition for pattern formation - short-ranging selfenhancement coupled to a long-ranging antagonistic reaction is satisfied. Globally such distributions are stable. A general elevation of the activator is regulated back by the action of the antagonist. However a small local perturbation can initiate the patterning. A minute local elevation of the activator concentration will increase further due to the self-enhancement while the spreading antagonistic reaction causes a de-activation in the surrounding of this incipient maximum. Random fluctuations are sufficient for the initiation. In the activator - depleted substrate reaction, the elevated rate of local attachment of the precursors leads to a decrease of the precursor concentration. Due to their high diffusion rate, this affects precursor concentration in a larger surrounding (Fig. 2). A new stable steady state is reached when the local maximum is in a dynamic equilibrium with the precursors derived from a more extended region.

Models for bacterial pattern formation

63

This pattern formation requires that the field (i.e., the region in which the pattern formation can take place) is larger than the range of the activator. The term "range" denotes the mean distance a molecule can diffuse between its synthesis and disappearance; it depends on its diffusion rate and the halflife. The resulting pattern depends on the size of the field. When the field size is of the order of the activator range, only a polar pattern can be generated. High and low activator concentrations emerge at the maximum distance from each other, i.e., at opposite sides of the field (Figs. 1, 2). Therefore, there is no need to assume any pre-localized factors to accomplish an asymmetric accumulation of the activator at one pole. The resulting pattern is in a wide range independent of the initial conditions. The pattern is self-regulating. After removal of an activator maximum the substrate concentration rises steeply until the autocatalysis is triggered again from a low-level constitutive activator production (Fig. 2). Polar patterns are essential for the generation of embryonic axes in multicellular organisms and their self-regulating properties are important for the robustness of development (Meinhardt and Gierer, 2000; Meinhardt, 2001). When the field size is about twice that of the activator range, symmetrical patterns are favoured. In a field of this size, and with an elongated or ovoid geometry, only two patterns are possible: Either a high concentration forms at each pole and a minimum in the centre, or a single maximum forms in the centre and both poles remain non-activated. At even larger sizes, several maxima can form at more or less regular, distances. The activator-inhibitor and the activator-depleted substrate model differ in the transition to periodic pattern during growth: while in an activator-inhibitor system new maxima appear at a certain distance from existing maxima (Fig. 1), in the depleted substrate scheme maxima preferentially split and shift (Fig. 2).

3.

If the antagonistic reaction has a longer time constant: periodic pattern in time

The same type of interactions that leads to patterns in space can also lead to patterns in time, i.e., to oscillations. This occurs if the antagonistic reaction has a longer time constant (Fig. 3). If the inhibitor reacts too slowly, first a burst-like activation occurs. The slowly accumulating inhibitor leads to a sudden breakdown of the activator synthesis. After the decay of the inhibitor, the next cycle of activation can occur from a low-level activator synthesis. Likewise oscillation occurs in an activator - depleted substrate scheme if the rate of substrate production is insufficient to maintain a steady state activation. The accumulation of substrate, the pulse-wise activation

64

Meinhardt

Figure 3. Oscillations. The same interactions that lead to pattern in space can generate patterns in time. Oscillations occur if the antagonistic reaction has a longer time constant when compared with the self-enhancing reaction (Meinhardt and Gierer, 1974). (A) Activatorinhibitor system (black-grey). The burst-like increase of the activator is followed by an increase of the inhibitor, which causes eventually a breakdown of the activator synthesis. After decay of the inhibitor surplus, a new activation can be triggered from a baseline activator production. (B) activator-substrate model: Whenever the substrate (precursor) concentration reaches a certain level, activation can be triggered that leads to a breakdown of the substrate and therewith of the activator production. A next burst has to wait until the substrate concentration recovered.

leading to the collapse of the substrate concentration and the refractory period required for its recovery is clearly visible in Fig. 3B.

4.

Generation of a polar pattern that oscillates between the two poles

As mentioned above, to generate a stable pattern in space a long-ranging but rapidly reacting antagonist is required. In contrast, an oscillating pattern results if a long-lasting antagonist is involved (it can be of short range). The pole-to-pole oscillation of MinD mentioned above requires obviously both a pattern in space and a pattern in time. This can be achieved by two antagonists (Fig. 4). A long-ranging antagonist is responsible for a pattern in space. The accumulation of long-lasting but local antagonistic component leads to a local down-regulation of the self-enhancing process once a pattern is formed. The newly formed maximum breaks down. This, however, also leads to a relaxation of the long-ranging and rapidly reacting antagonist. One possibility is that the maximum becomes permanently shifted into adjacent regions, leading to travelling waves that move back and forth (Fig. 4A, B). Alternatively the maximum breaks down and a new one appears at distance,

Models for bacterial pattern formation

65

Figure 4. Travelling waves or pole-to-pole oscillation. A linear field and two antagonists are assumed. A rapidly spreading substrate is responsible for a pattern in space. The locally accumulating inhibitor leads to the locally 'poisoning' of a region shortly after its activation. Two modes are possible: (A, B) The spatial patterning system on its own would lead to a stable pattern (shown in the top part of A). The locally accumulating inhibitor leads to a permanent shift of the maximum into non-poisoned neighbouring regions. Waves of activation are formed that travel back and forth between the two poles. (C, D) The system that generates the spatial pattern would oscillate on its own. The locally acting inhibitor leads to an out-of-phase oscillation: high concentrations emerge at the poles in an alternating fashion since the new maximum will arise at the least poisoned region. Simulation in a linear field; A, C space-time plots of the activator; B, D: profiles of all substances involved.

e.g., at the opposite pole. In the course of time, this maximum will breakdown for the same reason, and so on. The result is an out-of-phase oscillation (Fig. 4C, D).

5.

The MinD/MinE oscillation

The minimal mechanism described above generates an oscillating polar pattern in a reliable way. However, it does not describe the observation of the Min system correctly. None of the known components has the properties of an inhibitor, i.e., a molecule produced with the self-enhancing reaction and down regulating it. In a model as given in Fig. 4, if the long-lasting antagonistic reactions would be non-functional, a stable pattern or a global oscillation is expected. This, however, is not what is observed. A further argument against the involvement of lateral inhibition comes from a particular feature of the pole-to-pole oscillation. A new high MinD

66

Meinhardt

Figure 5. Model for out-of-phase oscillation in E. coli. (A) MinD protein forms an oscillating polar pattern that is visualized by GFP-labelling (Raskin and de Boer, 1999) (number indicate seconds, DIC is a differential phase contrast photograph of the same bacterium). A full cycle occurs in ca. 40 seconds. (B) Simulation: assumed is that MinD (black) binds everywhere to the membrane. On its own, it would not form a pattern. MinE (grey) forms a local high concentration at the membrane but this needs bound MinD. Since MinE removes MinD from the membrane, a MinE maximum destabilizes itself and enforces its own shift. Near the pole, the MinE wave collapses due to a decrease of the MinE precursors and due to the removal of MinD from this part of the membrane. This collapse of MinD at one pole is further enforced by a trigger of a new MinD maximum at the opposite pole, causing an additional drop of MinD precursors. At its flank a new high MinE concentration is generated that removes the newly triggered MinD activation, and so on. Thus, the MinE waves keep the centre free of MinD and thus of MinC, the proper inhibitor of septum formation (for equations, animated simulations and computational details see Meinhardt and de Boer, 2001; with supplementary material).

concentration can appear at the opposite pole before the old MinD activation disappeared completely (see Hale et al., 2001; Fig. 2). An existing MinD activity at one pole has obviously no strong inhibitory influence on the trigger of another one at the opposite pole. This cannot be described by an interaction as given in Fig. 4C, D. In such a mechanism, only after the breakdown of the existing maximum does the precursor concentration in the whole cell rise sufficiently to trigger the new activation at the opposite pole. From the genetic observations it was clear that another essential component, MinE, is required for the MinD oscillation. If MinE is absent no oscillation occurs but also no pattern is formed. Early visualizations of MinE suggested a more central localization (Raskin and de Boer, 1997), raising the question of how MinE becomes more prevalent at the centre and how a centrally located molecule can be the driving force for a highly dynamic process that takes place in a different part of the cell - the flashing up of MinD at alternating poles. Most observations can be described by the following model (Fig. 5; Meinhardt and de Boer, 2001). MinD and MinE are molecules that can assemble at the membrane. This binding has a self-

Models for bacterial pattern formation

67

reinforcing element that is antagonized by the depletion of the unbound MinD and MinE molecules. The later can move freely in the cytoplasm. In the absence of MinE, MinD accumulates evenly along the membrane of the entire cell, in agreement with the experimental observation (Raskin and de Boer, 1999a). In contrast, MinE forms a local maximum. Its association to the membrane depends on MinD. On its own, this would lead to a stable MinE maximum. However, MinE displaces MinD from the membrane. Thus, a MinE maximum causes its own local destabilization. This leads to a shift of the MinE maximum into an adjacent region that is still rich in MinD. The result is a wave of MinE (E-ring), which "peels" MinD off the membrane. This wave comes to rest shortly before it reaches the pole due to both a fading amount of membrane-bound MinD and a shortage of freely diffusible MinE in the remaining portion next to the pole. These MinE waves have been directly visualized (Hale et al., 2001; Fu et al., 2001). At the opposite pole, MinD makes progress in re-assembling at the membrane since sufficient precursor molecules become available. This attracts the assembly of a new MinE-ring. For the modelling it was crucial the observation that the new MinE activity appears at the flank and not at the maximum of the newly formed MinD activation. This occurs if a certain MinD concentration is required for MinE binding but a too high MinD concentration is inhibitory for that. After this new MinE ring is formed, it travels to the pole of this cell half as well, and so on. The simulations in Fig. 5 show that this model generates dynamic patterns that closely resemble the actual pattern of both MinD and MinE. In terms of the general mechanism described above, the primary pattern forming reaction is MinE. Assuming a non-linear first order decay of the molecules, the general theory predicted that the self-enhancing reaction must be non-linear (Gierer and Meinhardt, 1972). MinE satisfies this condition since it undergoes homodimerization (King et al., 1999). Thus, flashing up of local high MinD concentrations do not result from inherent patternforming properties of MinD but because these regions remain untouched for a while by the peeling-off action of the MinE wave. The modelling shows that reliable patterning is possible without the assumption of any prelocalized determinants. This is an important property of the model because a requirement for such factors would immediately raise the question of how they themselves become localized. Starting with homogenous initial conditions, random fluctuations are sufficient to initiate the pattern formation, Experimentally it has been found that an increase of MinE by multiple copies of the MinE gene leads to more rapid oscillations while a higher MinD level slows these down (Raskin and de Boer, 1999; Hale et al., 2001). In the model, more MinE leads to a more rapid removal of MinD and thus to

68

Meinhardt

Figure 6. Out-of phase oscillations. (A) In large cells obtained by suppression of cell division, several patches of high MinD concentrations are formed. Visualization of the maxima at certain time intervals reveals an oscillation out-of-phase (Raskin and de Boer, 1999). (B) Simulation: MinE waves (grey) move towards high MinD regions (black) from both sides. New high MinD regions emerge at the formerly non-activated sites. Note that the system finds very rapidly this mode without any initial patterns (for further details, equations and parameters see Meinhardt and de Boer, 2001; photographs kindly provided by Piet de Boer).

a faster MinE wave. In contrast, if the concentration of MinD is elevated, more time is required for their removal by a given number of MinE molecules, in agreement with the observations (see Meinhardt and de Boer, 2001 for simulation).

6.

Behaviour during growth, division, and in long filaments

Since the bacterium increases in length during growth, the centre finding mechanism must work even if the field size increases two-fold, at least. Moreover, after separation of a mother cell into two daughter cells, the centres of the later must quickly be detected. The proposed model has these properties (Meinhardt and de Boer, 2001). When cell division is blocked, E. coli grows into long filamentous cells. Such filaments show multiple dynamic MinD accumulations flanked by Erings. Their number increases with cell length (Fig. 6A; Raskin and de Boer, 1999a; Hu and Lutkenhaus, 1999; Hale et al., 2001). The model provides a straightforward explanation for this phenomenon. The simulation shows the appearance of multiple high MinD regions that oscillate out of phase (Fig. 6). In large fields several MinE maxima are formed simultaneously. This leads to multiple maxima in the MinD distribution between the MinE peaks. MinE waves move towards the regions of high MinD concentrations for the

Models for bacterial pattern formation

69

reasons described above. Pairs of MinE waves that flank an internal MinD maximum move from both sides towards the centre of these maxima until both activities disappear at near-collision. During this process, new MinD maxima appear in the regions just cleared by the waves. As a result, MinD maxima oscillate in counter-phase over time, in striking agreement with the observations (Hale et al., 2001). These simulations have been performed in one dimension only. However, the circumferential extension is of the same order as the axial extension. Taking the cylindrical geometry into account, the proposed mechanism leads to a single band of MinE activity that sweeps over the bacterium perpendicular to the long axis of the cylinder (see supplementary material in Meinhardt and de Boer, 2001). This pattern is self-organizing. The bands form even after initiation by random fluctuations but otherwise homogeneous initial conditions. The system is obviously able to detect the geometry of the system. This is in agreement with the observation that in round-shaped E. coli cells the MinD and MinE waves can have all sorts of orientation (Corbin et al., 2002).

7.

The signal to initiate the FtsZ ring

For the formation of the tubulin-like FtsZ ring, the Min proteins are not necessary. However, without the Min-System, the FtsZ ring can appear at more or less arbitrary positions. This can lead to mini-cells that lack the DNA. This shows two essential points (i) the FtsZ ring is on its own a pattern-forming reaction. We assumed that it is driven by a self-reinforcing component that is antagonized by depletion of precursor molecules in the cytoplasm (Meinhardt and de Boer, 2001). This is a reasonable assumption given the evidence that FtsZ-rings form by the localized polymerization of the FtsZ protein of which only a limited amount is available (Mukherjee and Lutkenhaus, 1998; Lu et al., 1998). (ii) The Min-System restricts the regions in which the FtsZ ring can be formed. MinC inhibits the formation of the FtsZ ring and its attachment to the membrane requires MinD. Since MinD removal takes place preferentially in the centre of the cell due to the more central back-and-forth movement of MinE, on time average the highest MinC/MinD concentration is at the poles. A process inhibited by MinC has therefore an optimum at the centre.

8.

A more static way of centre finding

70

Meinhardl

Figure 7. A static mode of pattern formation and the problem of stripe formation. (A) normally pattern formation leads to patch-like signals. (B) If the autocatalysis saturates, high activation emerge in regions that have a stripe-like extension. These stripes, however are not straight and may bifurcate. (C-E) With an inhibitory influence from both poles (centres shown in black, see Fig. 1) the septum-forming signal (grey) can be triggered only in the centre and only if, due to growth, the inhibitory poles obtained a certain distance. It forms a stable stripe with the correct orientation.

In the Gram-positive rod, Bacillus subtilis, MinE is absent and no poleto-pole oscillations of MinC/MinD occur. Instead, stable maxima of this division inhibitor are maintained at both poles. Under their repelling influence, the Z-ring assembles at the largest possible distance from both maxima, i.e., at the centre (Edwards and Errington, 1997; Marston and Errington, 1999; Marston et al., 1998). This raises the question of how locally high and temporary stable signals can emerge at the two poles. An activator/inhibitor reaction would be convenient for the generation of two terminal signals. Starting with a small field, initially only a polar pattern is possible. If a certain size is surpassed, a second maximum is generated at the opposite pole. The result is a symmetrical pattern with two terminal hot spots Fig. 1. If these two maxima have an inhibitory influence on septum formation, the later can take place in the centre only if a certain size is surpassed (Fig. 7). In an activator-depleted substrate scheme, however, peaks have more the tendency to shift towards the centre and split.

9.

The problem of generating a narrow belt-like signal

When such mechanisms are used to localize a signal for septum initiation the circumferential extension is too large to be neglected. The signal must have the geometry of a narrow belt, not of a patch. How can signals emerge that have a short extension along the axis and a long circumferential extension? It should be kept in mind that the FtsZ ring maintains this geometry even in the absence of the localizing MinD/E system. Why, if lateral inhibition is involved, does such a stripe-like signal not decay into individual patches?

Models for bacterial pattern formation

71

Figure 8. Stripe formation by local elongation - a possible mode for the FtsZ ring formation. A patch-forming system (black) triggers a stripe formation system (dark grey). The repelling influence of the stripe system on the inducing patch system leads to a shift of the elongating signal in front of the tip of the elongating stripe, causing its further elongation, etc. Thus, the stripe is generated as a trace behind a moving spot-like signal. The initial single signal splits into two, giving rise to two elongation sites at opposite ends of the emerging stripe-like structure. In this simulation, no orienting cues have been assumed. Therefore, the orientation of the stripe is arbitrary. It is assumed that stripe elongation depends on a molecule (light grey background) that becomes depleted during stripe formation. In this way the stripe has a repelling influence on the signal. This suppresses a shift of the signal to the side. Therefore, the elongation is straight. Equations and the parameters are given in the appendix.

One possibility for stripe formation is that the self-enhancement saturates. With an upper boundary, the local signal increase is limited. This restricts the mutual competition and larger regions remain activated. Stripelike patterns are favoured since in this arrangement, activated regions have activated neighbours but close to them there are non-activated regions from which precursor molecules can be obtained (Meinhardt, 1989, 1995). Without initial asymmetries, these stripes will have somewhat random distributions and may bifurcate (Fig. 7B). As shown in Fig. 7C, the inhibitory influence from the poles would lead automatically to an orientation of such a stripe perpendicular to the long axis of the bacterium. After completion of the division, the septum-system could trigger the pole system in order to convert the former septum-signal into a new pole signal.

10. FtsZ ring: a local signal could elongate the FtsZ stripe An alternative mechanism for stripe formation consists of a local signal that initiates stripe formation while the induced stripe has a repelling influence on the signal. Therefore, the signal starts to move, leaving behind the long-extended structure (Fig. 8). The situation can be compared with the long extended vapour trail that forms behind an airplane. At a very different scale, this takes place during elongation of microtubules. Elongation occurs

72

Meinhardt

only at the ends of the existing microtubules, causing an increase of length without an increase of the diameter of the bundle. Indeed, there are arguments that the FtsZ ring is generated by local polymerisation signals that move around the future division site. Addinall and Lutkenhaus (1996) showed that in spherical mutants the Z ring grows as an arc, suggesting bidirectional growth from a single point. In some mutants, FtsZ is no longer a ring but a spiral. In this case, also the constriction of the tube-shaped bacterium has a spiral shape, showing that FtsZ signal has not to be a closed ring to initiate the constriction. In B. subtilis, with the switch from vegetative growth to sporulation, the central FtsZ ring is replaced by two rings that are much closer to the poles. One of them becomes discarded later on. This central-to-polar re-positioning of the ring is connected with transient FtsZ pattern of spiral geometry (Ben-Yehuda and Losick, 2002). Also the reverse occurs. If the cells are shifted from sporulation to vegetative growth before irreversible steps have occurred, the FtsZ activity spirals from the near-pole to the central position. This suggests that the local signal moves around the cell circumference to lay down the ring. This movement of the elongation signal becomes visible if the axial position of local signal changes, causing spiral-shaped intermediates to emerge. The employment of a moving local signal to generate a long extended structure can be found also in other biological systems. The formation of the notochord, the long-extended midline structure of vertebrates, is generated behind a moving local signal, Hensen's node in the chick or the Spemann organizer in amphibians (for models see Meinhardt, 2001). This mode, however, is not the only solution of this problem. Insects use a very different mechanism for midline formation based on a long-ranging inhibition that emanates from a local centre (Meinhardt, 2002), closely resembling the situation shown in Fig. 7C.

11. Comparison with other theoretical approaches For out-of-phase oscillations as observed in the MinD/E patterning of E. coli it was important to have one self-enhancing reaction coupled with two antagonists. In the model of Meinhardt and de Boer (2001) the selfenhancing MinE accretion to the membrane is antagonized by the depletion of long-ranging precursors of MinE and by detachment of MinD from the membrane. As required, both antagonistic reactions have different spread and different time constants (Gierer and Meinhardt, 1972; Meinhardt, 1982). In the model of Kruse (Kruse , 2002) MinE also stimulates membrane detachment of MinD and the result is also MinD/MinE pole-to-pole oscillations. However, MinD and MinE maxima appear out of phase: when

Models for bacterial pattern formation

73

MinD is high at one pole, MinE is high at the other pole, in contrast to experimental observations (Hale et al., 2002; and Shih et al., 2002). In Kruse's model MinE is not assumed to be a pattern-forming reaction but only accomplishes the detachment. Therefore, no MinE waves form. In the model of Howard et al. (2001) it is assumed that all molecules are strictly maintained. Therefore, the reaction consists only of an association to and dissociation from the membrane. In their equations no direct autoregulatory components are involved. The instability has its origin presumably in the mutual inhibition of two processes: MinD/MinE complexes, formed in the cytoplasm, cause a dissociation of MinD from the membrane. Thus, MinE-binding lowers bound MinD. Conversely, MinE hinders the spontaneous association of MinD precursors to the membrane. Thus, MinD and MinE block each other mutually, and an inhibition of an inhibition is in fact equivalent of a self-enhancing reaction. The diffusible precursor molecules play likewise an antagonistic role.

12. Out-of-phase oscillations in other systems The general class of reactions that generate these types of dynamic patterning has been deduced from the pigment patterns on some tropical sea shells (Meinhardt and Klingler, 1987; Meinhardt 2003). A mollusc can enlarge its shell only at the growing edge. Therefore, the two-dimensional shell patterns represent a time record of reactions that took place along the growing edge. A chessboard-like pattern, as occurring on some shells, results from a pigment production that oscillates out-of-phase in adjacent groups of cells in the shell-producing mantle gland. This pattern is analogous to the counter-phase oscillations of MinD in long filaments (Fig. 6). The initiation of new leaves behind the tip of a growing shoot is based on a similar mechanism. It occurs in a narrow ring-shaped zone next to the dome-shaped apical meristem. The actual arrangement of leaves is a time record of the signalling in this leaf-forming zone. For instance, the formation of leaves at alternating opposite positions, the so-called distichous pattern, results from the temporary formation of corresponding signals at opposite positions of the leaf-forming zone, very similar to the pole-to-pole oscillation in E. coli. Since the leaf forming zone has the form of a ring (not a rod as in E. coli), the displacement need not to be 180°. A displacement by 137,5°, the golden angle, is an especially stable pattern in the model, corresponding to a well-known pattern in phyllotaxis (Meinhardt et al., 1998). Many cells are highly sensitive to external signals, a feature that enables chemotactic movement or an oriented outgrowth as in axons. Cells form

74

Meinhardt

protrusions preferentially at the side exposed to the higher signal concentration. These phenomena can be explained by assuming that an internal cell signalling system exists. Hot spots of this internal patterning system emerge at positions favoured by the external signals. The sensitivity of the cell can be maintained only if these signals are extinguished somewhat later, in order that new ones can be generated, possibly at an updated position in relation to the external signal (Meinhardt, 1999). Thus, the destabilization of patterns combined with the generation of new patterns seems to be a widely used strategy in biology for very different purposes. The pole-to-pole oscillation in E. coli is only an especially striking example of this more general mechanism.

13. Conclusions The main application of our pattern formation theory has been in the area of morphogenesis in multicellular organisms. The theory also describes essential steps in intracellular patterning. The required components are well known in biology, reactions that activate or inhibit each other, including autoregulation and the spread of components. No pre-localized determinants are required to account for a reliable patterning within a cell. The resulting patterns are self-organizing and return to normal operation even after severe perturbations. The models presented should be regarded as a toolbox that provides an inroad into the seemingly complex interactions. The future unravelling of how the actual molecular implementation is realized will certainly require a tight interaction between experimental and theoretical approaches.

Acknowledgements I am most grateful to Piet de Boer for introducing me in into the field of bacterial morphogenesis, for many discussions, corrections of the manuscript and the permission to use photographs from his work. The initial work on pattern formation was the result of a fruitful collaboration with Alfred Gierer over many years.

References Addinall, S.G. and Lutkenhaus, J. (1996). FtsZ-spirals and -arcs determine the shape of the invaginating septa in some mutants of Escherichia coli. Mol. Microbiol. 22, 231-237.

Models for bacterial pattern formation

75

Ben-Yehuda, S. and Losick, R. (2002). Asymmetric cell division in B. subtilis involves a spiral-like intermediate of the cytokinetic protein FtsZ. Cell 109, 257-266. Cheng, A.M.S., Thisse, B., Thisse, C. and Wright, C.V.E. (2000). The lefty-related factor xatv acts as a feedback inhibitor of nodal signaling in mesoderm induction and l-r axis development in Xenopus. Development 127, 1049-1061. Corbin, B.D., Yu, X.C. and Margolin, W. (2002). Exploring intracellular space: function of the Min system in round-shaped Escherichia coli. EMBO J. 21, 1998-2008. Edwards, D.H., and Errington, J. (1997). The Bacillus subtilis DivIVA protein targets to the division septum and controls the site specificity of cell division. Mol. Microbiol. 24, 905915. Fu, X.L., Shih, Y.L., Zhang, Y. and Rothfield, L.I. (2001). The MinE ring required for proper placement of the division site is a mobile structure that changes its cellular location during the Escherichia coli division cycle. Proc. Natl. Acad. Sci. U.S.A. 98, 980-985. Gierer, A. and Meinhardt, H. (1972). A theory of biological pattern formation. Kybernetik 12, 30-39. (available at http://www.eb.tuebingen.mpg.de/meinhardt) Hale, C.A., Meinhardt, H. and de Boer, P.A.J. (2001). Dynamic localization cycle of the cell division regulator MinE in Escherichia coli. EMBO J. 20, 1563-1572. Howard, M., Rutenberg, A.D. and de Vet, S. (2001). Dynamic compartmentalization in bacteria: accurate division in E. coli. Phys. Rev. Letters 87, 278102-278104. Hu, Z.L. and Lutkenhaus, J. (1999). Topological regulation of cell division in Escherichia coli involves rapid pole to pole oscillation of the division inhibitor MinC under the control of MinD and MinE. Mol. Microbiol. 341, 82-90. King, G.F., Rowland, S.L., Pan, B., Mackay, J.P., Mullen, G.P. and Rothfield, L.I. (1999). The dimerization and topological specificity functions of MinE reside in a structurally autonomous C-terminal domain. Mol. Microbiol. 314, 1161-1169. Kruse, K. (2002). A dynamic model for determining the middle of Escherichia coli. Biophys. J. 82, 618-627. Lu, C.L., Strieker, J. and Erickson, H.P. (1998). FtsZ from Escherichia coli, Azotobacter vinelandii, and Thermotoga maritima - Quantitation, GTP hydrolysis, and assembly. Cell Motil. Cytoskeleton 40, 71-86. Marston, A.L., Thomaides, H.B., Edwards, D.H., Sharpe, M.E. and Errington, J. (1998). Polar localization of the MinD protein of Bacillus subtilis and its role in selection of the midcell division site. Genes Dev. 122, 3419-3430. Marston, L. and Errington, J. (1999). Selection of the midcell division site in Bacillus subtilis through MinD-dependent polar localization and activation of MinC. Mol. Microbiol. 33, 84-96. Meinhardt, H. (1982). Models of biological pattern formation. Academic Press, London (available at http://www.eb.tuebingen.mpg.de/meinhardt). Meinhardt, H. (1989). Models for positional signalling with application to the dorsoventral patterning of insects and segregation into different cell types. Development (Supplement 1989), 169-180. Meinhardt, H. (1995). Growth and patterning - dynamics of stripe formation. Nature 376, 722-723. Meinhardt, H. (1999). Orientation of chemotactic cells and growth cones: Models and mechanisms. J. Cell Sci. 112, 2867-2874. Meinhardt, H. (2001). Organizer and axes formation as a self-organizing process. Int. J. Dev. Biol. 45, 177-188. Meinhardt, H. (2002). The radial-symmetric hydra and the evolution of the bilateral body plan: an old body became a young brain. Bioessays 24, 185-191.

76

Meinhardt

Meinhardt, H. (2003). The Algorithmic Beauty of Sea Shells. (3rd edition) Springer, Heidelberg, New York. Meinhardt, H. and Gierer, A. (1974). Applications of a theory of biological pattern formation based on lateral inhibition. J. Cell Sci. 15, 321-346. Meinhardt, H. and Gierer, A. (2000). Pattern formation by local self-activation and lateral inhibition. Bioessays 22, 753-760. Meinhardt, H. and Klingler, M. (1987). A model for pattern formation on the shells of molluscs. J. Theor. Biol 126, 63-69. Meinhardt, H., Koch, A.J. and Bernasconi, G. (1998). Models of pattern formation applied to plant development. In: Symmetry in Plants, (D. Barabe and R. V. Jean, Eds), World Scientific Publishing, Singapore, pp. 723-758. Meinhardt, H. and de Boer, P.A.J. (2001). Pattern formation in E. coli: a model for the poleto-pole oscillations of Min proteins and the localization of the division site. Proc. Natl. Acad. Sci. U.S.A. 98, 14202-14207. Meno, C., Gritsman, K.., Ohishi, S., Ohfuji, Y., Heckscher, E., Mochida, K., Shimono, A., Kondoh, H., Talbot, W.S., Robertson, E.J., Schier, A.F. and Hamada, H. (1999). Mouse lefty2 and zebrafish antivin are feedback inhibitors of nodal signaling during vertebrate gastrulation. Mol. Cell 4, 287-298. Mukherjee, A. and Lutkenhaus, J. (1998). Dynamic assembly of FtsZ regulated by GTP hydrolysis. EMBOJ. 17, 462-469. Raskin, D.M. and de Boer, P.A.J. (1997). The MinE ring: an FtsZ-independent cell structure required for selection of the correct division site in E. coli. Cell 91, 685-694. Raskin, D.M. and de Boer, P.A.J. (1999a). MinDE-dependent pole-to-pole oscillation of division inhibitor MinC in Escherichia coli. J. Bacteriol. 181, 6419-6424, Raskin, D.M. and de Boer, P.A.J. (1999b). Rapid pole-to-pole oscillation of a protein required for directing division to the middle of Escherichia coli. Proc. Natl. Acad. Sci. U.S.A. 96, 4971-4976. Shapiro, L. and Losick, R. (2000). Dynamic spatial regulation in the bacterial cell. Cell 100, 89-98. Shih,Y. L., Fu, X., King,G. F., Le, T. and Rothfield, L. (2002). Division site placement in E. coli: mutations that prevent formation of the MinE ring lead to loss of the normal midcell arrest of growth of polar MinD membrane domains. EMBO J. 2002 21,3347-3357 Turing, A. (1952). The chemical basis of morphogenesis. Phil. Trans. B. 237, 37-72.

Models for bacterial pattern formation

77

14. Appendix: Model for stripe formation by local elongation For the simulation Fig. 8 two activator - inhibitor systems have been used. The a, b system has the capacity for stripe formation due to the saturation

It is induced by the c, d-system that has a patch-forming characteristic (no saturation). The induction results from the term

After induction, the stripe system (a) repels the patch system (c) by removing a factor e that is required for c autocatalysis; e is produced everywhere with the same rate

The local depletion of e causes a shift of the spot-like elongation signal (c) in front of the incipient line since there fewer cells of the stripe system contribute to e-removal. Therefore, this is the region that is adjacent to the stripe in which the e-concentration is highest. In this way, the stripe-like structure becomes elongated in a straight way. Fig. 8 has been calculated with the following parameters: random fluctuation; random fluctuation; Since the c system is close to a transition into an oscillating mode, a feature that facilitates local de-activation required to accomplish the shift; (for animated simulations see Meinhardt, 2003).

This page intentionally left blank

Chapter 5 The assembly of proteins at the cell division site

WILLIAM MARGOLIN Department of Microbiology and Molecular Genetics University of Texas Medical School at Houston Houston, Texas 77030 USA

1.

Overview

In order to divide by binary fission, bacterial cells need to meet two broad requirements. First, the division site must be chosen at a location that will allow daughter chromosomes to be equally partitioned without being damaged. Second, a protein machine dedicated to cytokinesis must be assembled at this chosen site at the correct time during the cell cycle. The localization to the cell centre between the partitioning daughter chromosomes is mediated by several factors, including the nucleoid and the Min system (reviewed in chapter 4). This chapter will focus on the assembly of the protein machine and the interactions between its components.

2.

The players

The molecular genetic studies of the 1970s and 1980s identified many of the key players in E. coli cell division and their nucleotide sequences. The conditional mutants found were called fts mutants, which stands for "filamentous temperature sensitive". The phenotypic analysis of mutant filaments allowed a rough classification of these genes in the pathway. The ftsZ gene, for example, was known to act early in the septation pathway because ftsZ (ts) cells grown at the nonpermissive temperature were smooth, with no sign of constriction or septation. The ftsA gene, on the other hand,

Molecules in Time and Space, Edited by Vicente et al. Kluwer Academic/Plenum Publishers, New York 2004

79

80

Margolin

was thought to act later because at the nonpermissive temperature, ftsA (ts) mutant cells formed filaments with regular indentations. While valuable, these thermosensitive mutants also are the source of potential artifacts, as it is not clear what the thermosensitive defects of the mutants are and how much residual activity remains under the nonpermissive conditions. More solid evidence for a large machine and a specific pathway of its assembly has been more recently established by examining the localization of the septal proteins in the cell. New tools for localization, in particular green fluorescent protein, have allowed breakthroughs in our understanding of how the division machine gets assembled. FtsZ was the first of the division proteins to be localized to the cell centre. The initial immunogold localization of FtsZ was a landmark result (Bi and Lutkenhaus, 1991) that suggested that FtsZ was not merely a regulatory protein but had a structural role at the septum. Thanks to the fluorescence localization tools developed in the last few years, we now know that at least 10 essential cell division proteins localize to the septum. Moreover, by asking which of these proteins localize in various conditional cell division mutants, an assembly pathway for these proteins has been proposed for E. coli. This pathway is, from first to last: FtsZ-FtsA/ZipAFtsK-FtsQ-FtsL/YgbQ-FtsW-FtsI-FtsN (see Fig. 1). The same level of detail has not yet been achieved in other model prokaryotes such as Bacillus subtilis or Caulobacter crescentus, but so far the assembly order information for these organisms agrees with that of E. coli. Interestingly, the order of gene expression of ftsZ and ftsA during the cell cycle of Caulobacter reflects the order of addition seen in E. coli (Sackett et al., 1998). The colocalization of all these proteins, and their mutual dependencies for localization, strongly suggest that they form a large protein complex.

3.

FtsZ and the early recruits, FtsA and ZipA

Localization of FtsZ to the division site does not require any known positively acting factors, and it is the self-assembly of FtsZ into the so-called Z ring that initiates the assembly of the division protein machine. The assembly of FtsZ into protofilaments has been studied extensively in vitro and is the subject of recent comprehensive reviews (Scheffers and Driessen, 2001; Addinall and Holland, 2002; Erickson, 2001). Localization of all the other known cell division proteins to the septum absolutely requires the presence of the Z ring. Whether the Z ring serves as a scaffold for the recruitment of the later proteins, or has additional roles in the invagination of the septum, is reviewed in chapter 7.

The assembly of proteins at the cell division site

81

Figure 1. Hypothetical model of a septal subassembly, showing all the known essential cell division proteins with their general membrane topologies as well as AmiC and YgfE. FtsA is shown as a dimer at the end of one protofilament of FtsZ (see section 7). As FtsA may recruit some of the later proteins directly, it is shown contacting them. ZipA and YgfE are shown in a separate location as they are unlikely to recruit the later proteins by direct contact with them. Some of the other proteins, such as FtsL, are likely dimers, but are shown as monomers for clarity. Sizes are not to scale.

3.1

FtsA

In E. coli, the first known proteins to be recruited to the Z ring are FtsA and ZipA (Hale and de Boer, 1999; Liu et al., 1999). FtsA is a membraneassociated cytosolic protein (Pla et al., 1990) and is a member of the ATPase superfamily that includes eukaryotic actin and prokaryotic MreB (Bork et al., 1992). The crystal structures of FtsA, actin and MreB are very similar, although FtsA has an additional domain (1c) that is lacking in the others (van den Ent et al., 2001; van Den Ent and Lowe, 2000). FtsA is required for the localization of the 7 downstream division proteins. In the absence of functional FtsA, Z rings assemble normally (Addinall et al., 1996), but septation does not proceed, presumably because the later division proteins fail to be recruited. It is not known whether this recruitment function of FtsA is its only essential role, or if it has other activities important for cell division. It is also not known whether the ability of FtsA to localize to the Z ring is independent of its ability to recruit downstream proteins. FtsA was originally thought to act at a late stage of septation, because ftsA(ts) mutant filaments contain periodic indentations at the nonpermissive temperature (Begg & Donachie 1985). Constrictions have also been observed in filaments after depletion of FtsA in an ftsA null mutant, indicating that the late septation block is not solely a result of partial activity of the thermosensitive FtsA proteins. Because FtsA arrives early to the Z

82

Margolin

ring and is required to recruit later proteins, these results suggest that FtsA must function both early and late during septation. Despite the central role of FtsA in cell division, surprisingly little is known about its biochemical activities. Purified E. coli FtsA binds ATP and a subset of the protein is phosphorylated (Sanchez et al., 1994). In vivo studies suggest that E. coli FtsA self-interacts, possibly by dimerizing (Yim et al., 2000). Both the extreme C-terminal tail and the 1c domain, the latter which is not present in MreB or actin, have been implicated in the selfinteraction (Yim et al., 2000; Carettoni et al., 2003). Interestingly, purified B. subtilis FtsA also forms dimers and binds ATP; in addition, the B. subtilis protein displays ATPase activity (Feucht et al., 2001). Because of its structural similarity to actin, it is not surprising that FtsA interacts with ATP and can oligomerize. However, to date the physiological role of any of these activities of FtsA is unclear. A mutant allele of E. coli ftsA (ftsA104) that cannot bind ATP or be phosphorylated can complement various ftsA thermosensitive mutants, suggesting that the ATP interactions may be nonessential in E. coli (Sánchez et al., 1994). Mutations that abolish selfinteraction, both in the C-terminal tail and in domain 1c, also prevent full function of the protein as measured by complementation of an ftsA(ts) mutant (Yim et al., 2000; Carettoni et al., 2003). Whether this inhibition of function is because of the loss of self-interaction or another yet undiscovered activity is not yet clear. Its structural similarity with actin and its ability to self-interact suggests that FtsA has the potential to form polymers. However, the levels of FtsA in E. coli have been estimated to be only about 50-200 monomers per cell (Wang and Gayda, 1992), as compared to approximately 15,000 for FtsZ (Lu et al., 1998). Thus, whereas FtsZ levels are sufficient to form a polymer that encircles the E. coli cell several times, FtsA in E. coli is probably sparsely distributed along the Z ring and therefore unlikely to form an extensive polymer. This approximately 100:1 stoichiometry between FtsZ and FtsA is crucial for cell division, as altering the ratio in either direction results in the formation of filamentous cells (Dewar et al., 1992; Dai and Lutkenhaus, 1992). One possible reason for this strict stoichiometry is that FtsA modifies the Z ring in some way, perhaps by controlling the number of contacts between the Z ring and another factor. FtsA may also affect FtsZ polymerization, perhaps by interacting with the ends of FtsZ polymers (see section 7.2). Interestingly, overproduction of either FtsA or FtsZ results in the lack of Z rings (Addinall and Lutkenhaus, 1996). The mechanism of Z ring ablation by overproduction of FtsA or FtsZ is unknown, but it is possible that excess FtsA binds FtsZ monomers and sequesters them from the Z-ring nucleation site. It is still not clear how excess FtsZ prevents Z ring assembly, however.

The assembly of proteins at the cell division site

83

One function of FtsA appears to involve stabilization of FtsZ assembly. Evidence for this comes from studies of the effects of overproduction of the Z-ring inhibitor, MinC. MinC inhibits assembly of FtsZ in vitro (Hu et al., 1999), and MinC overproduction prevents formation of the Z-ring in vivo (Pichoff and Lutkenhaus, 2001). Interestingly, the inhibition of Z rings by excess MinC can be suppressed by overproduction of FtsA (Justice et al., 2000), suggesting that FtsA antagonizes the destabilizing effects of MinC. Altering wild-type FtsA with a point mutation, R286W, also allows cells to become resistant to the effects of excess MinC (Geissler et al., 2003). Cooverproduction of FtsA-GFP and FtsZ results in formation of stable fluorescent helices throughout the cell, suggesting that FtsA-GFP may help to stabilize FtsZ assembly (Ma and Margolin, 1999). Finally, there is good evidence that FtsA shares some functions with ZipA, which stabilizes Z rings (see next section). How might FtsA mediate FtsZ stabilization? One possibility is that FtsA may affect the dynamics of the Z ring. Recent studies using FRAP (fluorescence recovery after photobleaching) have demonstrated that the Z ring exhibits a high rate of turnover, even when it is not visibly constricting (Stricker et al., 2002). By using FRAP to study FtsZ turnover in mutants lacking or overproducing FtsA, it should be possible to address whether FtsA influences Z ring dynamics. Another approach would be to examine the effects of purified FtsA on FtsZ assembly and GTPase activity in vitro. If FtsA can indeed replace the FtsZ-stabilizing effect of ZipA, one might imagine that FtsA could promote bundling of FtsZ protofilaments, similar to what has been shown for ZipA and B. subtilis ZapA (see below).

3.2 ZipA ZipA was originally isolated by its ability to bind strongly to purified FtsZ (Hale and de Boer, 1997). ZipA is essential for cell division and has several proposed roles. One of these, bundling of FtsZ protofilaments, has been demonstrated convincingly with purified ZipA and FtsZ (Hale et al., 2000; Raychaudhuri, 1999). In vivo, doubling the gene dosage of zipA can suppress the thermosensitive ftsZ84 mutant and stabilize the normally labile FtsZ84 rings at the nonpermissive temperature. This result suggests that a modest increase in ZipA levels may increase FtsZ protofilament bundling in vivo, which in turn may stabilize Z rings. However, it is not possible to make a firm conclusion about the physiological relevance of ZipA-mediated FtsZ protofilament bundling. First, the fine structure of the Z ring has not been determined in any cell. Therefore, it is not known whether protofilament bundles themselves are actually present. Second, the actual levels of ZipA protein that could stabilize Z rings in the ftsZ84 mutant were not reported,

84

Margolin

leaving it still not clear how much ZipA is required in vivo to exhibit this effect. Cells lacking functional ZipA can assemble Z rings, although the rings do not constrict, and some Z rings within the resulting filaments are missing (Liu et al., 1999; Hale and de Boer, 1999). Therefore, although ZipA stabilizes Z rings, it is not required for Z ring assembly. Interestingly, inactivation of both FtsA and ZipA prevents formation of new Z rings and destabilizes pre-existing rings (Pichoff and Lutkenhaus, 2002). This suggests that FtsA and ZipA function synergistically in some way to stabilize Z rings. Two other postulated functions of ZipA are recruitment of downstream septation proteins and anchoring the septal ring to the membrane. Localization of GFP fusions to later division proteins including FtsK indicated that ZipA, like FtsA, is required for the recruitment of all downstream division proteins to the ring (Pichoff and Lutkenhaus, 2002; Hale and de Boer, 2002). Theoretically, ZipA and FtsA could function in recruitment by contacting one or more of the later cell division proteins directly, or by enhancing the ability of the Z ring to recruit, or both. ZipA has also been proposed to anchor the Z ring to the membrane, possibly to provide leverage for the constrictive force (Erickson, 2001). Evidence for this idea comes from the three domains of ZipA: a membranebound N-terminus, a central flexible linker domain (Ohashi et al., 2002), and a cytoplasmic C terminus that binds FtsZ, called the FZB domain (Liu et al., 1999; Hale and de Boer, 1997; Hale et al., 2000). If another membrane anchor domain is substituted for that of ZipA, ZipA function is lost, suggesting that the N-terminal domain of ZipA has an essential function in addition to that of a membrane anchor (Hale et al., 2000). However, so far there is no direct evidence that ZipA actually anchors the Z ring to the membrane. There is also no direct evidence that FtsZ can interact with the E. coli cytoplasmic membrane, although purified FtsZ can assemble on a cationic lipid monolayer (Erickson et al., 1996) and on a Langmuir film (Alexandre et al., 2002). Therefore, it remains to be determined exactly how the Z ring maintains a connection to the leading edge of the invaginating membrane during septum growth. Despite these various functions ascribed to it, ZipA is not conserved outside the gamma-proteobacterial group that includes E. coli. While this might suggest that other bacteria use a divergent protein with a similar membrane anchoring function, it is equally possible that this function can be replaced by the more highly conserved proteins. Indeed, recent results indicate that the requirement for ZipA can be bypassed under certain conditions. A mutation in ftsA, changing R286 to a tryptophan, results in nearly normal cell growth and division in the complete absence of ZipA (Geissler et al., 2003). The bypass results either if the mutation is on the

The assembly of proteins at the cell division site

85

chromosome, replacing the wild type ftsA, or expressed on a plasmid as a merodiploid. This gain-of-function mutation in ftsA appears to act by stabilizing FtsZ assembly, because cells normally sensitive to the destabilizing effects of MinC overproduction become resistant to the effects when the ftsA mutation is introduced. Along the same lines, excess mutant FtsA and FtsZ, expressed from a multicopy plasmid, results in the appearance of multiple constrictions along a filamentous cell along with helical FtsZ patterns. This "hyperconstriction" phenotype is dependent on the presence of ZipA. In zipA- cells, co-expression of ftsZ and mutant ftsA results in minicells, similar to the phenotype observed when wild-type ftsA and ftsZ are coexpressed in zipA+ cells. These results are consistent with FtsA and ZipA having additive roles in stabilization of Z rings, which is supported by the failure of Z rings to form when both ZipA and FtsA are absent (Pichoff and Lutkenhaus, 2002). The results are also consistent with the ability of overproduced wild-type FtsA to suppress the effects of overproduced MinC, if one imagines that the R286W mutant FtsA has enhanced ability to stabilize FtsZ assembly. There are several additional important implications of this ability of a ftsA mutant to bypass the requirement for ZipA. One has to do with the proposed recruitment function of ZipA. How can ZipA be required for recruitment of later cell division proteins and yet not be required for cell division in the presence of the mutant FtsA? The possibility that the later cell division proteins are rendered unnecessary is highly unlikely, especially given that ftsAR286W cells are still sensitive to cephalexin and can localize GFP-FtsN to the septum (Geissler et al., 2003). A more likely possibility is that ZipA does not recruit the later proteins directly, but instead modifies the Z ring in some way; this modified ring is then rendered competent for recruitment of later proteins. When the Z ring is modified sufficiently by FtsAR286W, ZipA is no longer needed, and thus can be classified as an accessory factor for FtsZ assembly (Margolin, 2003). Of course, whether the Z ring is modified at all, and what modifications actually occur, is not known. The most likely modification is the bundling of FtsZ protofilaments, mainly because this has been observed in vitro. However, other activities, such as protection of the ends of the protofilaments from rapid disassembly (capping, see final section) or covalent modification analogous to the action of MAPs on microtubules (Drewes et al., 1998) are certainly possible modification scenarios. Two other proteins, ZapA and EzrA, have activities similar to that of ZipA. ZapA, a B. subtilis protein that has weak orthologs in E. coli and other bacteria, is not essential for cell division. However, like ZipA, ZapA can bundle FtsZ protofilaments in vitro and localizes to the Z ring (Gueiros-Filho and Losick, 2002). EzrA, a B. subtilis protein with no obvious homologs in

86

Margolin

E. coli, shares the unusual transmembrane topology of ZipA, but is nonessential and acts to destabilize Z rings in vivo (Levin et al., 1999). Because all three of these proteins can be dispensable under various conditions, ZipA, ZapA, and EzrA can be thought of more as accessory factors for Z ring assembly and dynamics than as essential components of the septation machinery (Levin et al., 2001; Margolin, 2003).

4.

The C terminus of FtsZ recruits FtsA and ZipA to the Z-ring

In order to understand how the Z ring recruits FtsA and ZipA, it is necessary to review briefly what is known about the structure and function of FtsZ. FtsZ has three domains: a highly conserved N terminus of approximately 300 amino acids (Domain I), a linker region variable in both amino acid sequence and length (Domain II), and (iii) a short, conserved Cterminal domain (Domain III). Domain I is necessary and sufficient for GTP binding, hydrolysis, and self-assembly, and all FtsZs found to date have a complete Domain I (Diaz et al., 2001; Scheffers et al., 2002; Mukherjee et al., 2001). Domain II is highly variable in length and composition. In most species, including E. coli, this domain is approximately 60 residues in length. However, in all alpha-proteobacteria examined so far, as well as mitochondria (which presumably evolved from this bacterial group), Domain II is anywhere from 200-250 residues long. Its function is unknown, but it is reasonable to speculate that this domain serves primarily to tether Domains I and III to each other. The significant length increase of Domain II seen in diverse members of the alpha-proteobacteria, but not any other group, suggests that there is a particular need to separate the two outer domains further from each other in these species. Domain III is fairly well-conserved, with a core of about 12 residues that are remarkably similar among widely distributed prokaryote species as well as chloroplast FtsZs. Domain III is dispensable for FtsZ self-assembly, as purified Domain III-deleted FtsZ can assemble readily into large polymers (Wang et al., 1997; Yu and Margolin, 1997). These polymers are highly bundled, suggesting that Domain III may regulate the extent of lateral interactions among protofilaments. Although Domain III is not required for assembly, it is essential for in vivo function of FtsZ. In the absence of wildtype FtsZ, Domain III-truncated FtsZ fails to form Z rings in vivo and instead assembles into concentrated foci between nucleoids that do not function in cell division, resulting in long filamentous cells (Ma and Margolin, 1999). Such deletions also display strong dominant-negative effects in otherwise wild-type cells (Wang et al., 1997; Din et al., 1998). It is

The assembly of proteins at the cell division site

87

likely that this defect in forming an organized ring is the primary reason for the inability of Domain III deletions to substitute for wild-type FtsZ. Domain III has additional functions that may explain why normal FtsZ rings are not formed in its absence. For example, deletion of Domain III of E. coli FtsZ, even just the 12-residue core, prevents the interaction of FtsZ with FtsA or ZipA (Ma and Margolin, 1999). Similarly, deletions of the C termini of FtsZ proteins from B. subtilis, C. crescentus, and Staphylococcus aureus abolish interactions with FtsA (Wang et al., 1997; Din et al., 1998; Yan et al., 2000). Altering some individual residues in the core region within Domain III abolishes detectable interactions with FtsA and/or ZipA, as measured by several different techniques (Ma and Margolin, 1999; Mosyak et al., 2000; Yan et al., 2000). However, the role of each residue within the core is not yet clear, as much of the data are contradictory. For example, Ma et al. found that alanine scanning of the region resulted in greater effects on FtsA binding than on ZipA binding, and in support of this, observed normallooking Z rings made by the mutant proteins in the absence of wild-type FtsZ. In contrast, Haney et al. found that alanine substitutions for many of these residues inhibited both FtsA and ZipA binding, while Yan et al., working with S. aureus proteins in the yeast two hybrid system, found that changing most of the core residues to alanine had no significant effect on FtsA binding. Moreover, Ma et al. found that D373A could complement an ftsZ mutant and could bind FtsA and ZipA, but Haney et al. found that D373G failed to bind or complement. To date, there has been no attempt to resolve these contradictions. The disparate data for Staphylococcus vs. E. coli may be explained if the S. aureus proteins have subtle differences in their structures and hence undergo a different type of interaction. This is reasonable, considering that S. aureus lacks a ZipA homolog, and its FtsZ C terminus may have changed slightly because binding ZipA is not important (see below). The dramatic difference between D373G and D373A may result from a subtle difference in binding affinities. Structural analysis of the ZipAFtsZ C-terminal protein complex has shed some light on the interaction and on the roles of each residue. A crystal structure of ZipA bound to the FtsZ C terminus shows that there are extensive hydrogen bonding and hydrophobic contacts between the C-terminal 17 residues of FtsZ and the FZB domain of ZipA (Mosyak et al., 2000). Interestingly, P375 of FtsZ, which is highly conserved, does not directly contact any residue in ZipA. This is consistent with a normal interaction between FtsZP375 and ZipA in a protein-protein interaction assay in E. coli. However, FtsZP375 is unable to complement an ftsZ mutant, possibly because it binds FtsA more weakly (Ma and Margolin, 1999). Interestingly, Domain III is not only necessary, but by itself is sufficient for binding ZipA (Hale et al., 2000) and FtsA (Mosyak et al., 2000), although

88

Margolin

Domain III of S. aureus FtsZ was unable to interact with S. aureus FtsA (Yan et al., 2000). At least for E. coli, the results suggest that Domain III can be thought of as a truly separate recruitment domain of FtsZ. Despite the general conservation of Domain III across most bacterial species, subtle structural differences in this region of the protein might account for the differences in FtsZ-FtsA interactions and FtsZ-ZipA interactions across species that have been observed. For example, ZipA is not present outside the gamma-proteobacteria, and FtsA, while widespread, is missing in some groups, such as Streptomyces and mycobacteria, that still contain conserved core residues of Domain III. Moreover, FtsZ-FtsA interactions appear to be somewhat species-specific (Ma et al., 1997), underscoring the likelihood of significant structural differences in the FtsZ C terminus resulting from small changes in amino acid composition. Both FtsA and ZipA seem to require the 12 amino acid core of Domain III for binding to FtsZ. How might this occur? Because FtsA and ZipA are recruited independently to the Z ring (Liu et al., 1999; Hale and de Boer, 1999), an attractive model is that some FtsZ subunits within the FtsZ ring bind FtsA, while others bind ZipA. As FtsZ is at least 10-fold more abundant than ZipA and about 100-fold more abundant than FtsA (although this has recently been challenged, (Rueda et al., 2003)), FtsZ would always be in excess, and therefore would function as an ideal scaffold for protein recruitment. Competition between FtsA and ZipA for binding sites on FtsZ would not be an issue unless FtsA or ZipA levels were artificially increased. Indeed, overproduction of ZipA results in filamentation, just like overproduction of FtsZ and FtsA. However, in the case of ZipA, filamentation can be induced by only a several-fold increase in ZipA concentration (Geissler et al., 2003; Hale et al., 2000). As this increased level of ZipA should not be close to the number of FtsZ molecules, it suggests two possible models. The first is that there are proteins in addition to FtsA and ZipA that bind to FtsZ, causing most FtsZs to be occupied under normal conditions. Assuming that these proteins are also important for cell division, then only a small (several fold) increase in any one of the FtsZbinding proteins would result in competition for binding sites, causing filamentation. This model is supported by the fact that filamentation caused by overproduction of ZipA or FtsA can be reversed by increasing levels of FtsZ. The second model proposes that altering the stoichiometry of FtsZ and its binding partners results in instability that is unrelated to competition for binding sites. For example, excess ZipA may cause excessive bundling of FtsZ polymers that may in turn destabilize some other interaction, preventing the Z ring from functioning or even assembling at all. This may be what prevents Z rings from assembling when FtsA or FtsZ is overproduced. These

The assembly of proteins at the cell division site

89

models are testable, and further work needs to be done to investigate the interplay of the various proteins that bind FtsZ.

5.

The later cell division proteins

The later cell division proteins of E. coli have several characteristics in common. First, they all are essential for cell division and localize to the septum. Second, their localization to the septum depends on FtsA and ZipA (and therefore, on FtsZ). Third, they are generally of low abundance, as far as this has been investigated. Finally, they all are integral membrane proteins with periplasmic domains. FtsW and FtsK are predicted to be polytopic, with several transmembrane stretches (Begg et al., 1995; Lara and Ayala, 2002) (Gerard et al., 2002). FtsQ, YgbQ, FtsI, FtsL, and FtsN are all bitopic; each has a short N-terminal cytoplasmic tail, a single transmembrane domain, and a relatively large periplasmic domain. Their dependency was deduced from the localization of GFP fusion proteins or immunostaining in various cell division mutants (Buddelmeijer and Beckwith, 2002). In E. coli, FtsL and YgbQ are co-dependent on each other for localization, possibly interacting via coiled-coil domains (Buddelmeijer et al., 2002), but all other proteins do not require any other known protein for their localization other than those ahead of them in the pathway. In B. subtilis, the situation is a bit different, as a few of the proteins are intrinsically unstable and several exhibit cooperative assembly (Errington et al., 2002; Robson et al., 2002). Other than FtsI, which is also known as penicillin-binding protein 3 and has transpeptidase activity, little is known about the functions of the later proteins in forming the septum other than the requirement for the localization of their partners. More is known about domains involved in septal localization. In FtsK, the N-terminal 17% of the protein is necessary and sufficient for its localization and for its cell division function (Draper et al., 1998) (Yu et al., 1998). The C terminus of FtsK, on the other hand, is involved in resolution of chromosome dimers, and FtsK therefore serves to couple septation with chromosome segregation (Steiner et al., 1999). The localization determinants for the other polytopic protein, FtsW, are not yet known for E. coli, although its position within the order of assembly is clear (Mercer and Weiss, 2002). However, Mycobacterium tuberculosis FtsW has a unique C-terminal extension that contacts a unique acidic patch in the C terminus of FtsZ (Datta et al., 2002). This direct contact with FtsZ suggests that FtsW might not be dependent on ZipA, FtsA, and other late proteins in M. tuberculosis. This is completely consistent with the absence of ZipA and FtsA from the M. tuberculosis genome, and suggests a mechanism for recruitment of the late proteins by FtsZ and FtsA of E. coli (see section 7.2).

90

Margolin

More is known about the localization domains of the bitopic proteins. For example, residues 50-247 in the periplasmic domain of FtsQ are necessary and sufficient for its targeting to the septum (Buddelmeijer et al., 1998; Chen et al., 2002). On the other hand, the last 29 residues (248-276) are required for FtsQ to function but not for its septal targeting. This indicates that the last 29 residues either physically interact with the downstream proteins FtsL and/or YgbQ, or modify the cell wall (or do something else) that now allows downstream recruitment. Therefore, separate domains for localization to the septum and for recruitment have been defined for FtsQ. As FtsQ is estimated to be only 25-50 molecules per cell (Carson et al., 1991), and FtsQ mutant proteins display competition for localization to the Z ring, it suggests that the targeting substrate is limiting (Chen et al., 2002). Because its targeting domain is periplasmic, FtsQ cannot be recruited directly by the cytoplasmic proteins FtsZ or FtsA. As FtsK is the only other known essential conserved protein ahead of FtsQ, it is possible that FtsK directly recruits FtsQ via one of FtsK's two periplasmic domains. FtsN and FtsL are similar to FtsQ in that only their periplasmic domains are required for targeting. Septal targeting of FtsI, in contrast, requires the cytoplasmic, transmembrane, and periplasmic domains. In addition, the transpeptidase function of FtsI appears to be required for FtsI localization by IFM (Wang et al., 1998), although it cannot be ruled out that the mutant protein could not be immunostained because of an antibody recognition problem.

6.

Function of the later proteins

Other than FtsI, the essential functions of the later proteins in septation remain unknown. There are only pieces of the puzzle so far, in part because no good assays have been developed, and because there are probably more proteins yet to be discovered that are part of the machine. What we know so far is mainly based on genetic interactions. For example, FtsW has been proposed to be a homolog of RodA, a protein important for maintenance of cellular rod shape (Ikeda et al., 1989). However, the actual activity of RodA is not known, so the function of FtsW is still obscure. FtsL may function as a structural protein, perhaps to stabilize some of the other late proteins, because its coiled-coil domains have the capacity to form multimers (Ghigo and Beckwith, 2000) and FtsL overproduction suppresses the thermosensitivity of divIB (ftsQ) mutants in B. subtilis (Daniel and Errington, 2000). FtsQ and YgbQ are still enigmas, although FtsQ has weak sequence similarity to Mpl, which is involved in peptidoglycan recycling (Buddelmeijer and Beckwith, 2002). FtsQ may help to stabilize the Z ring, because cells with a lon mutation (which increases the stability of the FtsZ

The assembly of proteins at the cell division site

91

antagonist, SulA) became less viable under certain conditions only in the presence of an ftsQ (ts) allele (Descoteaux and Drapeau, 1987). FtsN, on the other hand, has several distinctive attributes. It is poorly conserved, making it unlikely that it has a universal function. However, it is the last known protein in the assembly pathway, and has the curious ability to suppress several fts mutations, including ftsA(ts), ftsI(ts), and ftsK (ts), although null mutations are not suppressed. Even more puzzling is that an ftsN null mutant exhibits smooth filaments, without any sign of septation, suggesting that inactivating FtsN results in an early division defect similar to inactivation of many of the other division proteins. How could FtsN be recruited last, be required for an early function, and be able to suppress mutations in genes whose products are required for its recruitment? It was recently proposed that based on the weak homology of FtsN to cell wall amidases, FtsN might function to weaken the cell wall in preparation for constriction of the Z ring. Because the fts alleles suppressible by FtsN overproduction are probably partially functional, as they arrest septation at a later stage than do the null/depletion strains, FtsN overproduction may help to stabilize the defective septation machinery in these mutants. This also suggests that the various fts alleles are all defective in some common activity, such as severing bonds in the cell wall, which can be pushed to completion by extra FtsN. This idea is reasonable, especially considering that all the late proteins are absent in wall-less mycoplasmas. Also, all late proteins have periplasmic domains that could physically interact. However, other aspects of FtsN activity are not easily explained. For example, does extra FtsN allow ftsA(ts) to actually recruit the other late proteins, or is septation now independent of them? It is hard to imagine how extra FtsN could allow FtsA(ts) to recruit, as most of FtsN is in the periplasm, and, for that matter, FtsA(ts) proteins normally fail to localize to the Z ring at the nonpermissive temperature. Also, FtsN overproduction in an ftsA(ts) mutant should result in FtsN being distributed throughout the cell without specific localization to the septum, because FtsN localization to the septum depends on functional FtsA. How, then, would FtsN be able to act on the septation machinery? Perhaps FtsN is stable and although it is also everywhere else on the membrane, its overproduction causes its local concentration at the septum to be high enough to mimic its function when it is recruited by FtsI. Other genetic interaction studies, completed prior to our current understanding of the assembly pathway, suggested that excess FtsL, FtsQ and FtsI have an effect that is opposite of FtsN (Jung et al., 1989) (Dai and Lutkenhaus, 1992). For example, a multicopy plasmid containing the ftsL and ftsI genes caused ftsA(ts) ftsQ(ts) and ftsZ(ts) cells to filament at the permissive temperature of 30°C in no NaCl (the cells were normal in 0.5% NaCl, which can often partially suppress the thermosensitivity of various fts

92

Margolin

alleles). Interestingly, a plasmid that contained ftsL but only a fragment of ftsI caused filamentation of ftsQ(ts) or ftsZ(ts) but not of ftsA(ts). Finally, overproduction of FtsQ induced filamentation in ftsZ(ts), ftsA(ts) and ftsI(ts) mutants under permissive conditions. These results suggest that excess FtsL and/or FtsI inhibit septation when ftsQ, ftsA or ftsZ are compromised, but excess FtsI specifically inhibits septation only when ftsA is compromised. This in turn suggests that FtsA and FtsI might interact, and that FtsI-FtsA interactions might prevent interaction of FtsA with another essential cell division factor (excess FtsI would titrate out FtsA, taking it away from the Z ring even at permissive temperature). The hypothesis can be tested by examining the localization of FtsA(ts) at 30°C when FtsI is overproduced. If the localization were abolished, it would suggest that this titration might occur, although there could be other explanations for the effect. Similarly, the effect of FtsQ overproduction might be due to its titration of FtsA or another factor with which it interacts. Also, the negative effects of excess FtsL, FtsQ and FtsI in contrast to the positive effects of FtsN in the same mutants suggest that FtsN antagonizes the activities of FtsL, FtsQ and FtsI. The idea that FtsA and FtsI interact, either directly or indirectly, is also supported by other data. First of all, the N-terminal cytoplasmic tail of FtsI is required for FtsI function (Dai et al., 1996), whereas the N-terminal cytoplasmic tails of the other bitopic Fts proteins can be replaced by alternative domains (Guzman et al., 1997). These results suggest that FtsI is unique among the late septation proteins in that it is capable of making a specific and essential contact with a cytoplasmic protein, such as FtsA. Second, the resistance of an ftsA(ts) mutant to lysis by beta-lactams, and the decreased binding of a radiolabeled derivative of ampicillin to FtsI protein in the ftsA(ts) mutant at the restrictive temperature, suggests that FtsA and FtsI might interact, at least indirectly (Tormo et al., 1986). Interestingly, inhibiting FtsI by several different methods gives different Z-ring localization patterns (Wang et al., 1998). Inhibition by cephalexin or furazlocillin resulted in Z rings being absent at constriction sites (older sites) but present at potential sites; this suggests that Z rings in the process of contracting at the time of drug addition are susceptible to breakdown, while Z rings that have not yet contracted are stable. Depletion of FtsI resulted in Z ring formation at all potential division sites, indicating that gradual removal of FtsI does not destabilize Z rings, although septation is still inhibited. Inhibition by thermoinactivation of ftsI(ts) caused an intermediate phenotype. Although the experiments are indirect, the results suggest that the state of the Z ring, either determined by cell cycle timing or by the components present at the septum, influences its stability. Recently, a nonessential periplasmic protein, AmiC, has been shown to localize to the E. coli septum, and its localization is dependent on FtsN

The assembly of proteins at the cell division site

93

(Bernhardt and de Boer, 2003). AmiC is a member of cell wall amidases that are exported by the Tat pathway. The dependence on FtsN for its localization, as well as its localization to cells only in the process of constriction, indicates that AmiC is the latest known protein recruited to the septal complex. Like FtsN, the function of AmiC is likely to be the weakening of the cell wall in the vicinity of the growing septum. As the invagination of the periplasm and outer membrane need to be coordinated with the constriction of the inner membrane during cell division, we should expect to find more septal proteins in the periplasm. The discovery of AmiC also suggests that perhaps the FtsN-mediated suppression of various fts alleles may be exerted not by FtsN itself but by proteins recruited by FtsN.

7.

A working model for the assembly of the septumsynthesizing machinery

7.1 General models Can a molecular mechanism be invoked to explain the large amount of indirect genetic, phylogenetic, cytological, and biochemical data? Beckwith (Chen et al., 1999; Buddelmeijer and Beckwith, 2002) has proposed two alternative general models that would be consistent with the sequential localization of the cell division proteins in E. coli. In the first model, a large protein complex comprising all the proteins is formed by the sequential addition of each protein at the septum site. This sequential targeting is dictated by specific protein-protein interactions that allow only the next protein to bind (except when there is a codependent assembly, as with YgbQ/FtsL in E. coli or DivIB and DivIC in B. subtilis). This model does not require that the addition of the next protein cause any particular physical alteration at the septum other than the presence of the new protein. In the second model, the targeting of each protein induces a physical change in the membrane structure or protein-membrane interactions. As in a metabolic pathway, each addition of protein changes the substrate (membrane environment or local cell wall structure) in a specific way that allows the next step to proceed. In the case of septation, the next step would be the next protein to be recruited, as well as the additional change to the substrate. Much of the available evidence supports the first model. For example, the direct pairwise interactions between FtsZ and FtsA and FtsZ and ZipA shown primarily by yeast two-hybrid analysis is consistent with the presence of a protein complex, at least among the early proteins. The direct interaction

94

Margolin

between FtsZ and FtsW in M. tuberculosis also supports the idea of a protein complex at the septum as the driving force for sequential protein localization. Support for a complex formed by the later proteins comes from their persistence at the constricting ring; if the system was more like a metabolic pathway (model 2), then there is no obvious reason why the proteins would stay at midcell after they have achieved their enzymatic step. Finally, model 1 is supported by the lack of any obvious changes to the septum during the recruitment of all the septal proteins. For instance, depletion of FtsN results in smooth filaments, indicating that even when all but one of the known septation proteins are present at the septum, the septal ring stays in its "initiating complex". If the second model were true, one might expect further development of the septum, possibly detectable as visible changes, as each protein is recruited. In the wild-type situation, it is likely that only once all proteins are present and a signal is received does the Z ring contracts and the septum grows. Once this happens, the Z ring becomes susceptible to degradation and can cause the septum to abort. The second model cannot be completely discounted, however. The main evidence in support of this model comes from the ability of FtsN overproduction to suppress several thermosensitive mutations in earlier genes. This implies that extra FtsN can bypass the requirement for sequential recruitment of several of the essential cell division proteins. Such a bypass might be analogous to providing a metabolic intermediate in a linear enzymatic pathway, although it should be emphasized that null alleles in the various early genes cannot be suppressed by extra FtsN, only thermosensitive alleles. There is no other strong evidence to support the second model, making it likely that the first model or a combination of the first and second models is correct. To confirm aspects of model 1, evidence of binary protein-protein interactions at each dependency step needs to be established. Requirement of a specific protein for recruitment (such as the requirement of FtsW for recruitment of FtsI) does not mean that FtsW actually recruits FtsI, either by a direct protein-protein interaction or by an interaction through another protein. If FtsW modifies the membrane or the Z ring, allowing the ring to target FtsI, then model 2 would be supported as well. Experimental proof of actual direct recruitment can be achieved in vitro with purified components, in vivo at an ectopic site (somewhere other than the septum) or in vivo in an organism that lacks the other cell division components.

The assembly of proteins at the cell division site

95

7.2 Is FtsA the key to understand the assembly of the septal complex? Several lines of evidence suggest that FtsA is a key regulator of the recruitment of later cell division proteins. The first is its position in the dependency pathway, immediately after FtsZ. ZipA is in the same position, but because ZipA is not well conserved and can be bypassed by altering FtsA, it is more likely that ZipA is an accessory factor for the Z ring that does not play a direct role in recruitment. The requirement of ZipA for late recruitment may be because it stabilizes FtsZ, allowing FtsZ to better recruit either directly or via FtsA. Second, assuming that its abundance is a few hundred molecules per E. coli cell, FtsA is far less abundant than FtsZ and therefore similar in abundance to the later proteins it recruits. This suggests, as originally proposed by Nanninga (Nanninga, 1998), that FtsA may limit the number of periplasmic subassemblies of the later proteins. It is not known whether there are 3, or 30, or more subassemblies, but clearly there are not enough later proteins in the cell to make subassemblies for each FtsZ monomer. The higher abundance of FtsA, DivIB (FtsQ), and possibly other later proteins in B. subtilis (Rowland et al., 1997) (Feucht et al., 2001) suggests that either more subassemblies or larger subassemblies are needed to synthesize the thicker Gram-positive septum. How might FtsA regulate the number of subassemblies, while at the same time interacting with FtsZ and the later proteins? One attractive idea is that FtsA binds preferentially to the ends of FtsZ protofilaments. Such a role would be analogous to actin filament "capping" proteins, which have major effects on polymer stability and dynamics (Cooper and Schafter, 2000). If FtsA were a capping protein for FtsZ, it would be consistent with the apparent ability of FtsA to stabilize FtsZ polymers in vivo and with the FtsA:FtsZ stoichiometry required for normal cell division to occur. The absence of FtsA in some species may also give clues to its function. Why might mycobacteria, cyanobacteria, and actinomycetes contain FtsZ and some other conserved septal proteins, but lack FtsA? One reasonable idea is that FtsA is involved in recycling FtsZ from old rings to new rings via its potential role as a capping protein. By regulating whether FtsZ protofilaments undergo "catastrophe", FtsA would be able to control protofilament degradation, and therefore the availability of FtsZ monomers to assemble new Z rings. If so, then one might rationalize that FtsA might be dispensable in species that do not require such recycling. In Streptomyces, Z rings are not required for vegetative growth (McCormick et al., 1994) but are formed en masse during sporulation, an irreversible process that leads to simultaneous synthesis of spore septa (Schwedock et al., 1997). There would

96

Margolin

be no need to recycle FtsZ for the next cell cycle because the spores are terminally differentiated. Likewise, slow-growing cells of cyanobacteria and mycobacteria would not need to recycle FtsZ, as new FtsZ could easily be synthesized in the long period between division events even if FtsZ polymerization dynamics are slower (White et al., 2000). But FtsA is also important for recruitment of later septation proteins. As these species must need these proteins for septation, how can the loss of FtsA be explained? In the case of M. tuberculosis, because FtsA is absent, FtsZ may have evolved to contact FtsW directly via a small stretch of acidic amino acids on FtsZ that interact with a stretch of basic residues on FtsW (Datta et al., 2002). Both stretches are unique to the M. tuberculosis proteins. However, the FtsZ proteins of Streptomyces coelicolor or Synechocystis do not have any obvious acidic patches at their C termini, so it remains to be seen how their Z rings recruit downstream proteins in the absence of FtsA.

7.3 FtsA as a potential switch for septal synthesis Might FtsA, by potentially binding to the ends of FtsZ protofilaments and contacting later septation proteins, be able to integrate the signal to initiate septum constriction? Evidence in support of this idea comes from the properties of conditional ftsA mutants. Thermosensitive FtsA proteins such as that expressed by ftsA12(ts) do not appear to be recruited to the Z ring (Addinall and Lutkenhaus, 1996), yet filamentous cells of this mutant at the nonpermissive temperature contain indentations (Addinall et al., 1996). This suggests that when FtsA is absent from the Z ring, the ring contracts, causing a visible indentation in the cell, and subsequently aborts. This might happen because the membrane becomes detached from the cell wall. These constrictions should not contain later cell division proteins, because ftsA(ts) filaments fail to recruit later proteins such as FtsK (Yu et al., 1998; Wang and Lutkenhaus, 1998) (Wang et al., 1998). Thus, these constrictions must form independently of their recruitment. While completion of the division septum requires all septal components to be present, these results imply that, unlike what was proposed by Buddelmeijer and Beckwith (Buddelmeijer and Beckwith, 2002), triggering the initiation of cell division does not require that all division proteins be in place. If we accept the idea that Z rings can contract without the later proteins, then why would depletion of later proteins such as FtsK or N, for example, result in smooth filaments (Wang and Lutkenhaus, 1998) (Addinall et al., 1997)? FtsA should be present at Z rings in these filaments (Addinall and Lutkenhaus, 1996). These results suggest a speculative model in which FtsA acts as a negative regulator to "lock" the Z ring, preventing its contraction until the proper signal is received. In the ftsA(ts) mutant at the nonpermissive

The assembly of proteins at the cell division site

97

temperature, the FtsA protein fails to localize to Z rings, preventing the locking mechanism from working and allowing the rings to contract. Such a locking/unlocking function of FtsA should involve a trigger, perhaps transmitted by FtsN or another protein that, when recruited to the septal complex, signals that the machine is now assembled. According to this model, FtsA would be a molecular switch that regulates FtsZ turnover in response to signals transduced from the state of the septal protein machine, resulting in septal peptidoglycan synthesis. This switch might be achieved by ATP binding and hydrolysis or by phosphorylation state, both of which have been shown to be properties of purified FtsA (Sanchez et al., 1994) (Feucht et al., 2001). The attraction of this model is that it can be tested both in vivo, by finding mutants in ftsA that have dysfunctional regulation, and in vitro, by looking for the effects of FtsA modifications on its ability to interact with the division proteins. The localized bulging and bending of cells overproducing FtsA or truncated derivatives of FtsA (Yim et al., 2000) (Wang et al., 1993) (Gayda et al., 1992) already suggest that alteration of FtsA can cause inappropriate synthesis and deposition of "septal" peptidoglycan.

8.

Outlook

The number of proteins involved in septation continues to grow faster than our understanding of their function or their protein-protein contacts. What is really needed is a way of reconstituting the complex independently in order to obtain structural and functional information and to provide a context for the linear order of recruitment observed for E. coli. Such structural data can be obtained by crystallization of membrane proteins, as well as directly measuring intact structures by scanning probe or atomic force microscopy. In combination with the current optical microscopy and genetic tools that the bacterial model systems have, this should prove to be a powerful way to gain insight into the functions of the cell division protein machine. In addition, cell cycle studies can be combined with mass spectrometry to give a more accurate idea of the protein stoichiometries, and whole-cell tomography can be used, once the technology is sufficiently mature, to be able to visualize the protein complexes in cells at different stages in the division process. Based on the tremendous advances in the past ten years, it is expected that the next decade will bring even more fundamental insights into assembly of the cell division machine.

98

Margolin

References Addinall, S. G., Bi, E., and Lutkenhaus, J. (1996). FtsZ ring formation in fts mutants. J. Bacteriol. 178, 3877-3884. Addinall, S. G., Cao, C., and Lutkenhaus, J. (1997). FtsN, a late recruit to the septum in Escherichia coli. Mol. Microbiol. 25, 303-309. Addinall, S. G., and Holland, B. (2002). The tubulin ancestor, FtsZ, draughtsman, designer and driving force for bacterial cytokinesis. J. Mol. Biol. 318, 219-36. Addinall, S. G., and Lutkenhaus, J. (1996). FtsA is localized to the septum in an FtsZdependent manner. J. Bacteriol. 178, 7167-7172. Alexandre, S., Colé, G., Coutard, S., Monnier, C., Norris, V., Margolin, W., Yu, X., and Valleton, J. M. (2002). Interaction of FtsZ protein with a DPPE Langmuir film. Coll. Surf. B23, 391-395. Begg, K. J., Dewar, S. J., and Donachie, W. D. (1995). A new Escherichia coli cell division gene, ftsK. J. Bacteriol. 177, 6211-6222. Bernhardt, T. G., and de Boer, P. A. (2003). The Escherichia coli amidase AmiC is a periplasmic septal ring component exported via the twin-arginine transport pathway. Mol. Microbiol. 48, 1171-1182. Bi, E., and Lutkenhaus, J. (1991). FtsZ ring structure associated with division in Escherichia coli. Nature 354, 161-164. Bork, P., Sander, C., and Valencia, A. (1992). An ATPase domain common to prokaryotic cell cycle proteins, sugar kinases, actin, and hsp70 heat shock proteins. Proc. Natl. Acad. Sci. USA 89, 7290-7294. Buddelmeijer, N., Aarsman, M. E., Kolk, A. H., Vicente, M., and Nanninga, N. (1998). Localization of cell division protein FtsQ by immunofluorescence microscopy in dividing and nondividing cells of Escherichia coli. J. Bacteriol. 180, 6107-6116. Buddelmeijer, N., and Beckwith, J. (2002). Assembly of cell division proteins at the E. coli cell center. Curr. Opin. Microbiol. 5, 553-7. Buddelmeijer, N., Judson, N., Boyd, D., Mekalanos, J. J., and Beckwith, J. (2002). YgbQ, a cell division protein in Escherichia coli and Vibrio cholerae, localizes in codependent fashion with FtsL to the division site. Proc. Natl. Acad. Sci. USA 99, 6316-21. Carettoni, D., Gomez-Puertas, P., Yim, L., Mingorance, J., Massidda, O., Vicente, M., Valencia, A., Domenici, E., and Anderluzzi, D. (2003). Phage-display and correlated mutations identify an essential region of subdomain 1C involved in homodimerization of Escherichia coli FtsA. Proteins 50, 192-206. Carson, M. J., Barondess, J., and Beckwith, J. (1991). The FtsQ protein of Escherichia coli: membrane topology, abundance, and cell division phenotypes due to overproduction and insertion mutations. J. Bacteriol. 173, 2187-2195. Chen, J. C., Minev, M., and Beckwith, J. (2002). Analysis of ftsQ mutant alleles in Escherichia coli: complementation, septal localization, and recruitment of downstream cell division proteins. J. Bacteriol. 184, 695-705. Chen, J. C., Weiss, D. S., Ghigo, J. M., and Beckwith, J. (1999). Septal localization of FtsQ, an essential cell division protein in Escherichia coli. J. Bacteriol. 181, 521-530. Cooper, J. A., and Schafter, D. A. (2000). Control of actin assembly and disassembly at filament ends. Curr. Opin. Cell Biol. 12, 97-103. Dai, K., and Lutkenhaus, J. (1992). The proper ratio of FtsZ to FtsA is required for cell division to occur in Escherichia coli. J. Bacteriol. 174, 6145-6151. Dai, K., Xu, Y., and Lutkenhaus, J. (1996). Topological characterization of the essential Escherichia coli cell division protein FtsN. J. Bacteriol. 178, 1328-1334.

The assembly of proteins at the cell division site

99

Daniel, R. A., and Errington, J. (2000). Intrinsic instability of the essential cell division protein FtsL of Bacillus subtilis and a role for DivIB protein in FtsL turnover. Mol. Microbiol. 36, 278-289. Datta, P., Dasgupta, A., Bhakta, S., and Basu, J. (2002). Interaction between FtsZ and FtsW of Mycobacterium tuberculosis. J. Biol. Chem. 277, 24983-24987. Descoteaux, A., and Drapeau, G. R. (1987). Regulation of cell division in Escherichia coli K12: probable interactions among proteins FtsQ, FtsA and FtsZ. J. Bacteriol. 169, 19381942. Dewar, S. J., Begg, K. J., and Donachie, W. D. (1992). Inhibition of cell division initiation by an imbalance in the ratio of FtsA to FtsZ. J. Bacteriol. 174, 6314-6316. Diaz, J. F., Kralicek, A., Mingorance, J., Palacios, J. M., Vicente, M., and Andreu, J. M. (2001). Activation of cell division protein FtsZ. Control of switch loop T3 conformation by the nucleotide gamma-phosphate. J. Biol. Chem. 276, 17307-17315. Din, N., Quardokus, E. M., Sackett, M. J., and Brun, Y. V. (1998). Dominant C-terminal deletions of FtsZ that affect its ability to localize in Caulobacter and its interaction with FtsA. Mol. Microbiol. 27, 1051-1063. Draper, G. C., McLennan, N., Begg, K., Masters, M., and Donachie, W. D. (1998). Only the N-Terminal Domain of FtsK Functions in Cell Division. J. Bacteriol. 180, 4621-4627. Drewes, G., Ebneth, A., and Mandelkow, E. (1998). MAPS, MARKs and microtubule dynamics. Trends Biochem. Sci. 23, 307-311. Erickson, H. P. (2001). The FtsZ protofilament and attachment of ZipA--structural constraints on the FtsZ power stroke. Curr. Opin. Cell Biol. 13, 55-60. Erickson, H. P., Taylor, D. W., Taylor, K. A., and Bramhill, D. (1996). Bacterial cell division protein FtsZ assembles into protofilament sheets and minirings, structural homologs of tubulin polymers. Proc. Natl. Acad. Sci. USA 93, 519-523. Errington, J., Daniel, R. A., and Scheffers, D. J. (2002). Cytokinesis in bacteria. Microbiol. Mol. Biol. Rev. 67, 52-65. Feucht, A., Lucet, I., Yudkin, M. D., and Errington, J. (2001). Cytological and biochemical characterization of the FtsA cell division protein of Bacillus subtilis. Mol. Microbiol. 40, 115-125. Gayda, R. C., Henk, M. C., and Leong, D. (1992). C-shaped cells caused by expression of an ftsA mutation in Escherichia coli. J. Bacteriol. 174, 5362-5370. Geissler, B., Elraheb, D., and Margolin, W. (2003). A gain of function mutation in ftsA bypasses the requirement for the essential cell division gene zipA in Escherichia coli. Proc. Natl. Acad. Sci. USA 100, 4197-4202. Gerard, P., Vernet, T., and Zapun, A. (2002). Membrane topology of the Streptococcus pneumoniae FtsW division protein. J. Bacteriol. 184, 1925-1931. Ghigo, J. M., and Beckwith, J. (2000). Cell division in Escherichia coli: role of FtsL domains in septal localization, function, and oligomerization. J. Bacteriol. 182, 116-129. Gueiros-Filho, F. J., and Losick, R. (2002). A widely conserved bacterial cell division protein that promotes assembly of the tubulin-like protein FtsZ. Genes Dev. 16, 2544-56. Guzman, L. M., Weiss, D. S., and Beckwith, J. (1997). Domain-swapping analysis of FtsI, FtsL, and FtsQ, bitopic membrane proteins essential for cell division in Escherichia coli. J. Bacteriol. 179, 5094-5103. Hale, C. A., and de Boer, P. A. (1997). Direct binding of FtsZ to ZipA, an essential component of the septal ring structure that mediates cell division in E. coli. Cell 88, 175185. Hale, C. A., and de Boer, P. A. (1999). Recruitment of ZipA to the septal ring of Escherichia coli is dependent on FtsZ and independent of FtsA. J. Bacteriol. 181, 167-176.

100

Margolin

Hale, C. A., and de Boer, P. A. (2002). ZipA is required for recruitment of FtsK, FtsQ, FtsL, and FtsN to the septal ring in Escherichia coli. J. Bacteriol. 184, 2552-6. Hale, C. A., Rhee, A. C., and de Boer, P. A. (2000). ZipA-induced bundling of FtsZ polymers mediated by an interaction between C-terminal domains. J. Bacteriol. 182, 5153-5166. Hu, Z., Mukherjee, A., Pichoff, S., and Lutkenhaus, J. (1999). The MinC component of the division site selection system in Escherichia coli interacts with FtsZ to prevent polymerization. Proc. Natl. Acad. Sci. USA 96, 14819-14824. Ikeda, M., Sato, T., Wachi, M., Jung, H. K., Ishino, F., Kobayashi, Y., and Matsuhashi, M. (1989) Structural similarity among Escherichia coli FtsW and RodA proteins and Bacillus subtilis SpoVE protein, which function in cell division, cell elongation, and spore formation, respectively. J. Bacteriol. 171, 6375-6378. Jung, H. K., Ishino, F., and Matsuhashi, M. (1989). Inhibition of growth of ftsQ, ftsA, and ftsZ mutant cells of Escherichia coli by amplification of a chromosomal region encompassing closely aligned cell division and cell growth genes. J. Bacteriol. 171, 6379-6382. Justice, S. S., Garcia-Lara, J., and Rothfield, L. I. (2000). Cell division inhibitors SulA and MinC/MinD block septum formation at different steps in the assembly of the Escherichia coli division machinery. Mol. Microbiol. 37, 410-423. Lara, B., and Ayala, J. A. (2002). Topological characterization of the essential Escherichia coli cell division protein FtsW. FEMS Microbiol. Lett. 216, 23-32. Levin, P. A., Kurtser, I. G., and Grossman, A. D. (1999). Identification and characterization of a negative regulator of FtsZ ring formation in Bacillus subtilis. Proc. Natl. Acad. Sci. USA 96, 9642-9647. Levin, P. A., Schwartz, R. L., and Grossman, A. D. (2001). Polymer stability plays an important role in the positional regulation of FtsZ. J. Bacteriol. 183, 5449-5452. Liu, Z., Mukherjee, A., and Lutkenhaus, J. (1999). Recruitment of ZipA to the division site by interaction with FtsZ. Mol. Microbiol. 31, 1853-1861. Lu, C., Stricker, J., and Erickson, H. P. (1998). FtsZ from Escherichia coli, Azotobacter vinelandii, and Thermotoga maritima--quantitation, GTP hydrolysis, and assembly. Cell Motil. Cytoskeleton 40, 71-86. Ma, X., and Margolin, W. (1999). Genetic and functional analysis of the conserved Cterminal core domain of Escherichia coli FtsZ. J. Bacteriol. 181, 7531-7544. Ma, X., Sun, Q., Wang, R., Singh, G., Jonietz, E. L., and Margolin, W. (1997). Interactions between heterologous FtsA and FtsZ proteins at the FtsZ ring. J. Bacteriol. 179, 67886797. Margolin, W. (2003). Bacterial division: the fellowship of the ring, Curr. Biol. 13, R16-8. McCormick, J. R., Su, E. P., Driks, A., and Losick, R. (1994). Growth and viability of Streptomyces coelicolor mutant for the cell division gene ftsZ. Mol. Microbiol. 14, 243254. Mercer, K. L., and Weiss, D. S. (2002). The Escherichia coli cell division protein FtsW is required to recruit its cognate transpeptidase, FtsI (PBP3), to the division site. J. Bacteriol. 184, 904-12. Mosyak, L., Zhang, Y., Glasfeld, E., Haney, S., Stahl, M., Seehra, J., and Somers, W. S. (2000). The bacterial cell-division protein ZipA and its interaction with an FtsZ fragment revealed by X-ray crystallography. EMBO J. 19, 3179-3191. Mukherjee, A., Saez, C., and Lutkenhaus, J. (2001). Assembly of an FtsZ mutant deficient in GTPase activity has implications for FtsZ assembly and the role of the Z ring in cell division. J. Bacteriol. 183, 7190-7197. Nanninga, N. (1998). Morphogenesis of Escherichia coli. Microbiol. Mol. Biol. Rev. 62, 110129.

The assembly of proteins at the cell division site

101

Ohashi, T., Hale, C. A., de Boer, P. A., and Erickson, H. P. (2002). Structural evidence that the P/Q domain of Zip A is an unstructured, flexible tether between the membrane and the C-terminal FtsZ-binding domain. J. Bacteriol. 184, 4313-4315. Pichoff, S., and Lutkenhaus, J. (2001). Escherichia coli division inhibitor MinCD blocks septation by preventing Z-ring formation. J. Bacteriol. 183, 6630-6635. Pichoff, S., and Lutkenhaus, J. (2002). Unique and overlapping roles for ZipA and FtsA in septal ring assembly in Escherichia coli. EMBO J. 21,685-693. Pla, J., Dopazo, A., and Vicente, M. (1990). The native form of FtsA, a septal protein of Escherichia coli, is located in the cytoplasmic membrane. J. Bacteriol. 172, 5097-5102. Raychaudhuri, D. (1999). ZipA is a MAP-Tau homolog and is essential for structural integrity of the cytokinetic FtsZ ring during bacterial cell division. EMBO J. 18, 2372-2383. Robson, S. A., Michie, K. A., Mackay, J. P., Harry, E. J., and King, G. F. (2002) The Bacillus subtilis cell division proteins FtsL and DivIC are intrinsically unstable and do not interact with one another in the absence of other septasomal components. Mol. Microbiol. 44, 663-674. Rowland, S. L., Katis, V. L., Partridge, S. R., and Wake, R. G. (1997). DivIB, FtsZ and cell division in Bacillus subtilis. Mol. Microbiol. 23, 295-302. Rueda, S., Vicente, M., and Mingorance, J. (2003). Concentration and assembly of the division ring proteins FtsZ, FtsA, and ZipA during the Escherichia coli cell cycle. J. Bacteriol. 185, 3344-3351. Sackett, M. J., Kelly, A. J., and Brun, Y. V. (1998). Ordered expression of ftsQA and ftsZ during the Caulobacter crescentus cell cycle. Mol. Microbiol. 28, 421-434. Sanchez, M., Valencia, A., Ferrandiz, M.-J., Sander, C., and Vicente, M. (1994). Correlation between the structure and biochemical activities of FtsA, an essential cell division protein of the actin family. EMBO J. 13, 4919-4925. Scheffers, D., and Driessen, A. J. (2001). The polymerization mechanism of the bacterial cell division protein FtsZ. FEBS Lett. 506, 6-10. Scheffers, D. J., de Wit, J. G., den Blaauwen, T., and Driessen, A. J. (2002). GTP hydrolysis of cell division protein FtsZ: evidence that the active site is formed by the association of monomers. Biochemistry 41, 521-529. Schwedock, J., McCormick, J. R., Angert, E. R., Nodwell, J. R., and Losick, R. (1997). Assembly of the cell division protein FtsZ into ladder-like structures in the aerial hyphae of Streptomyces coelicolor. Mol. Microbiol. 25, 847-58. Steiner, W., Liu, G., Donachie, W. D., and Kuempel, P. (1999). The cytoplasmic domain of FtsK protein is required for resolution of chromosome dimers. Mol. Microbiol. 31, 579583. Stricker, J., Maddox, P., Salmon, E. D., and Erickson, H. P. (2002). Rapid assembly dynamics of the Escherichia coli FtsZ-ring demonstrated by fluorescence recovery after photobleaching. Proc. Natl. Acad. Sci. USA 99, 3171-3175. Tormo, A., Ayala, J. A., de Pedro, M. A., Aldea, M., and Vicente, M. (1986). Interaction of FtsA and PBP3 proteins in the Escherichia coli septum. J. Bacteriol. 166, 985-992. van den Ent, F., Amos, L. A., and Lowe, J. (2001). Prokaryotic origin of the actin cytoskeleton. Nature 413, 39-44. van Den Ent, F., and Lowe, J. (2000). Crystal structure of the cell division protein FtsA from Thermotoga maritima. EMBO J. 19, 5300-7. Wang, H., and Gayda, R. C. (1992). Quantitative determination of FtsA at different growth rates in Escherichia coli using monoclonal antibodies. Mol. Microbiol. 6, 2517-2524. Wang, H., Henk, M. C., and Gayda, R. C. (1993). Overexpression of FtsA induces large bulges at the septal regions in Escherichia coli. Curr. Microbiol. 26, 175-181.

102

Margolin

Wang, L., Khattar, M. K., Donachie, W. D., and Lutkenhaus, J. (1998). FtsI and FtsW are localized to the septum in Escherichia coli. J. Bacteriol. 180, 2810-2816. Wang, L., and Lutkenhaus, J. (1998). FtsK is an essential cell division protein that is localized to the septum and induced as part of the SOS response. Mol. Microbiol. 29, 731-740. Wang, X., Huang, J., Mukherjee, A., Cao, C., and Lutkenhaus, J. (1997). Analysis of the interaction of FtsZ with itself, GTP, and FtsA. J. Bacteriol. 179, 5551-5559. White, E. L., Ross, L. J., Reynolds, R. C., Seitz, L. E., Moore, G. D., and Borhani, D. W. (2000). Slow polymerization of Mycobacterium tuberculosis FtsZ. J. Bacteriol. 182, 4028-4034. Yan, K., Pearce, K. H., and Payne, D. J. (2000). A conserved residue at the extreme Cterminus of FtsZ is critical for the FtsA-FtsZ interaction in Staphylococcus aureus. Biochem. Biophys. Res. Commun. 270, 387-392. Yim, L., Vandenbussche, G., Mingorance, J., Rueda, S., Casanova, M., Ruysschaert, J. M., and Vicente, M. (2000). Role of the carboxy terminus of Escherichia coli FtsA in selfinteraction and cell division. J. Bacteriol. 182, 6366-6373. Yu, X.-C., and Margolin, W. (1997). GTP-dependent dynamic assembly of bacterial cell division protein FtsZ into asters and polymer networks in vitro. EMBO J. 16, 5455-5463. Yu, X.-C., Tran, A. H., Sun, Q., and Margolin, W. (1998). Localization of cell division protein FtsK to the Escherichia coli septum and identification of a potential N-terminal targeting domain. J. Bacteriol. 180, 1296-1304.

Chapter 6 Regulation and utilization of cell division for bacterial cell differentiation

JENNIFER WAGNER and YVES V. BRUN* Department of Biology Indiana University Bloomington, Indiana 47405-3700 USA *corresponding author

1.

Introduction

Bacteria employ various strategies to differentiate, many of which are influenced by the compartmentalization that results from cell division. Cell differentiation can be regulated through gene expression cascades, where one step in expression is dependent on a previous step. These cascades are initiated by external and internal stimuli, and may ensure the ordered expression of genes as they are required during the course of the bacterial life cycle. Once a division septum is formed and the cytoplasm of progeny cells becomes distinct, compartment-specific transcription of genes can be established. In addition to transcriptional control, gene products are also influenced post-translationally. The activity of many proteins is regulated by proteolysis, either by proteolytic processing of pre-proteins into their active forms, or through degradation of those proteins that no longer serve a purpose (Gottesman, 1999). The formation of a division septum can again be exploited by the bacterium to carry out differentiation, since proteolysis of a particular protein can be constrained to a specific cell compartment. Perhaps the most important discovery in bacterial differentiation in recent years

Molecules in Time and Space, Edited by Vicente et al. Kluwer Academic/Plenum Publishers, New York 2004

103

104

Wagner and Brun

comes from the examination of the subcellular distribution of macromolecules within cells. It is evident that even in bacteria, which lack extensive structural compartmentalization, proteins and nucleic acids are localized in very specific manners that often correlate with their function. Macromolecular localization can be dynamic and may assist cell differentiation by generating local concentrations of activities where they are required, and/or preventing them where they would be detrimental. Knowledge of subcellular localization allows us to conceptualize how cells, which have a single cytoplasmic space, can still differentiate. Despite the common strategic themes discussed above, different bacterial species undergo development for diverse reasons (Shimkets and Brun, 2000). In some bacteria, development is a process triggered by environmental conditions, such as nutrient limitation. This is the case for Bacillus subtilis and Streptomyces coelicolor. These organisms grow vegetatively until the developmental pathway is triggered, resulting in the generation of a resting cell type called a spore. The spore is more resistant to stresses such as desiccation, heat, and degradative enzymes, and resumes vegetative growth when conditions become favourable. Some organisms, such as the swarming bacteria, instead receive cues from the nature of the surface and probably cell-cell signaling. For example, Proteus mirabilis detects high cell density on a solid substrate and differentiates into a long, hyper-flagellated, multinucleate cell that associates in an orderly way with other differentiated cells in a colony. This raft of associated cells then migrates en masse to fresh substrate, where the cells resume division (Fraser and Hughes, 1999). In this way, the swarming bacteria exploit differentiation as a means of progeny dispersal. Another organism that uses differentiation to disperse is Caulobacter crescentus. One key difference is that in C. crescentus, differentiation is an obligatory event in the cell cycle, and is not elicited by environmental stimuli. The reproductive cell of C. crescentus possesses a polar extension of the cell body called the stalk, is non-motile, and is often attached to surfaces through an adhesive organelle found at the tip of the stalk called the holdfast. After the cell elongates and division occurs, a progeny cell with a single, polar flagellum is released. This swarmer cell serves a dispersal role, since it is obligated to swim for a period of time before losing its flagellum and generating a stalk in its place. Differentiating bacteria can also be found associated with another organism. This may be a pathogenic association, such as the colonization of the urinary tract by P. mirabilis or of a host bacterial cell by Bdellovibrio. In both of these cases, cell division is inhibited while the cell continues to grow and replicate chromosomes. Alternatively, the association may be a symbiotic relationship, such as nodule formation on the roots of legumes by Sinorhizobium. Bacteria belonging to the genus Sinorhizobium enter the host

Cell division and differentiation

105

cell in a series of phases. After host-cell entry, the bacterium differentiates into a polymorphic, non-dividing cell called a bacteroid, which fixes nitrogen. It is not clear whether bacteroids are terminally differentiated or whether they can reprogram to grow vegetatively under the right conditions. The following sections will discuss some of the major model systems in which cell division and differentiation have been studied (see Brun and Shimkets, 2000 for a more comprehensive review). The emphasis of each section will be related back to what is known about the major strategies used to regulate developmental processes. Regardless of the system discussed, the formation of every new cell type requires the manipulation and exploitation of some aspect of cell division. In this way, bacteria have evolved specialized developmental processes centered around the most primordial growth step.

2.

Bacillus subtilis

Bacillus subtilis is a Gram-positive soil bacterium capable of growing by binary fission or, under starvation conditions, entering a resting cell state in which cellular processes are dormant. The resting cell of B. subtilis, termed an endospore, is highly resistant to stresses such as heat and desiccation, and can remain quiescent for years until the appropriate cues trigger the program for vegetative growth. If cells sporulate at a time when nutrients are not critically limiting, they will lose their competitive growth advantage; however, if they sporulate too late, they may perish under hostile growth conditions. Not surprisingly, the switch to sporulation is a highly controlled process, largely regulated by a multi-component signal transduction cascade that integrates signals from the environment and the cell (Perego, 1998). The endospore is formed by asymmetric division of a cell that possesses two copies of the B. subtilis chromosome. The division septum forms proximal to a single cell pole, generating two disproportionate compartments. The larger compartment, termed the mother cell, contains a complete copy of one of the chromosomes and approximately 70% of the second. The remaining 30% of the second chromosome, trapped during septation, passes through the division septum and resides in the smaller compartment, referred to as the prespore (Fig. 1) (Wu and Errington, 1994). At this time, the two compartments carry out cell-specific transcription of genes through the aid of two sigma factors, and in the mother cell and prespore respectively. At the same time, the remainder of the second chromosome is transported into the forespore. The final steps of sporulation involve the engulfment of the prespore by the mother cell, the activation of yet another set of compartment-specific

106

Wagner and Brun

sigma factors, and the maturation of the spore as the mother cell lyses (Errington, 1996). B. subtilis utilizes cell division not only to generate a smaller cell type through an asymmetric division, but also to create a partition between the mother cell and prespore compartments so that the spore may mature.

Figure 1. A sporulating B. subtilis cell. The formation of the asymmetric division septum creates both a genetic asymmetry (with only 30% of the replicated chromosome initially located in the prespore) and a cytoplasmic asymmetry, allowing the compartment-specific gene and protein expression necessary for endospore maturation.

2.1 Regulation of entry into sporulation Sporulation initiation is regulated by two key transcription factors, Spo0A and (Levin and Losick, 1996). Spo0A is a master regulator and the terminal phosphate acceptor in the sporulation initiation phosphorelay (Perego, 1998). In the phosphorylated state, Spo0A acts as both an activator and a repressor of genes involved in the switch from vegetative growth to sporulation, leading to a cascade of events that culminates in the production of a mature endospore (Perego, 1998). The most critical function of Spo0A~P appears to be in activating the expression of a sporulation-specific cell division gene of unknown function, spoIIE (Ben-Yehuda and Losick, 2002). In addition to nutrient availability, Spo0A phosphorylation is dependent on DNA replication. Thus, a developmental checkpoint exists which links the earliest stage of sporulation to the state of chromosome replication (Perego, 1998; Burkholder et al., 2001). is a sporulation sigma factor responsible for the upregulation of another cell division gene, ftsZ, through an alternative promoter (Gonzy-Treboul et al., 1992). FtsZ initiates

Cell division and differentiation

107

the cell division process by polymerizing into a ring-like structure at the future cell division site. The transcription of ftsZ from an alternative promoter is significant, because FtsZ levels are important for polar FtsZ assembly in B. subtilis during sporulation (Ben-Yehuda and Losick, 2002).

2.2 Determination of the division site Like Escherichia coli, which does not divide asymmetrically as part of its life cycle, wild-type B. subtilis has the capacity to divide at three division sites, one near each pole (during sporulation), and one at midcell (during vegetative growth). Division site placement is determined by a myriad of factors including transcriptional activators, division site inhibitors, the essential cell division protein FtsZ, and the proteins that control FtsZ polymerization and the stability of FtsZ polymers. In B. subtilis, division at cell poles during vegetative growth is suppressed by the action of the MinCD system (Levin et al., 1992; Barak et al., 1998; Marston et al., 1998). MinC and MinD act in conjunction with another protein, DivIVA, to inhibit cell division at polar sites (Edwards and Errington, 1997). Null mutations in any of these proteins results in an anucleate, minicell phenotype, where cells divide at polar sites in addition to midcell (Levin et al., 1992; Cha and Stewart, 1997). Cells with mutations in minCD still sporulate with high efficiency, albeit with some morphological abnormalities (Barak et al., 1998). Division site inhibition is probably not the main regulatory mechanism controlling division site placement. In fact, a recent paper by Ben-Yehuda and Losick (2002) strongly suggests that the most important factor in formation of the sporulation septa is the level of FtsZ and SpoIIE. During sporulation, the concentration of FtsZ is greatly increased by the binding of a sporulation-specific sigma factor, to an alternative ftsAZ promoter, P2 (Gonzy-Treboul et al., 1992; Ben-Yehuda and Losick, 2002). Meanwhile, spoIIE transcription is induced by the binding of the sporulation master regulator, Spo0A, to the spoIIE promoter (Khvorova et al., 1998). It is believed that this increase in FtsZ and SpoIIE levels leads to redeployment of Z-rings from medial sites to polar sites. In support of this hypothesis, only 0.6% of B. subtilis cells which have a double mutation in the P2 promoter and spoIIE are able to divide asymmetrically under sporulation conditions (compared with 55% of wild-type cells), while about 25% of vegetatively growing cells are induced to divide asymmetrically when FtsZ and SpoIIE are overexpressed (Ben-Yehuda and Losick, 2002). During the transition to sporulation, FtsZ assembles in a Z-ring medially, then appears as a spiral-like intermediate of varying pitch from the midcell to one pole or across the entire cell, and finally coalesces as a Z-ring near

108

Wagner and Brun

one or both poles (Ben-Yehuda and Losick‚ 2002). The membrane-bound protein SpoIIE‚ colocalizes with FtsZ‚ also forming a spiral intermediate (Ben-Yehuda and Losick‚ 2002). SpoIIE and FtsZ have been shown to interact directly (Lucet et al.‚ 2000)‚ and SpoIIE localization is dependent on FtsZ (Levin and Losick‚ 1996)‚ but it is not yet clear how this protein contributes to the formation of polar Z-rings. SpoIIE may allow FtsZ to "stick" to polar division sites once FtsZ redeployment has occurred. FtsA‚ another cell division protein‚ also colocalizes with FtsZ in the spiral (BenYehuda and Losick‚ 2002). What prevents cell division from initiating in the spiral intermediate? Spirals of FtsZ have been seen transiently in mutants of E. coli that overexpress the protein or possess an FtsZ with a mutant GTPase activity‚ so the ability to form spirals is probably an intrinsic property of the protein (Lutkenhaus‚ 2002). It may be that FtsZ either needs to form a closed ring or exist as a tight spiral for the cell division proteins to be active‚ but this remains an open question. As mentioned before‚ Z-rings often form at both poles of the cell; however‚ it is unusual for more than one sporulation septum to form. Three mother cell-specific proteins‚ SpoIID‚ SpoIIM‚ and SpoIIP are involved in suppressing septum formation at the Z-ring site opposite the developing sporangium (Eichenberger 2001). The mechanism responsible for this suppression is not yet known. It is hypothesized that some limiting cell division protein‚ such as FtsA‚ is unable to support division at both sites simultaneously‚ and that SpoIID‚ SpoIIM‚ and SpoIIP localize to the divisome-free site to inhibit the initiation of a second division event before the limiting protein is free to re-associate at the second site (Eichenberger et al.‚ 2001).

2.3 Chromosome translocation During sporulation‚ asymmetric cell division results in one copy of the chromosome being partially trapped in the mother cell compartment‚ while the origin proximal region is found in the forespore (Wu and Errington‚ 1994). This arrangement leads to a genetic asymmetry between the mother cell and forespore‚ allowing differential gene expression to occur (Frandsen et al.‚ 1999). Thus‚ the genetic and cellular asymmetry required for the differentiation of B. subtilis into an endospore is dependent on a cell division event. Chromosome positioning plays an important role in the initial stages of spore maturation‚ since only certain regions of the translocated chromosome are available for transcription in the prespore (Dworkin and Losick‚ 2001; Zupancic et al.‚ 2001; Pogliano et al.‚ 2002). However‚ once this role is met‚ the chromosome must somehow be completely translocated into the

Cell division and differentiation

109

prespore. This action is mediated by the protein SpoIIIE‚ which is produced preferentially in the mother cell after asymmetric cell division (Wu and Errington‚ 1998; Bath et al.‚ 2000). SpoIIIE is hypothesized to serve as a DNA channel and/or translocase in the division septum (Fig. 1). Artificial induction of SpoIIIE synthesis in the forespore leads to translocation of DNA back into the mother cell‚ indicating that the channel assembles in a polar manner (Sharp and Pogliano‚ 2002a). Recently‚ the same group showed that the MinCD system responsible for inhibiting polar septation in vegetatively growing cells also represses SpoIHE assembly/activation in the forespore (Sharp and Pogliano‚ 2002b). It has not been determined whether or not the MinCD proteins act directly on SpoIIIE or through another factor. However‚ since the activity of MinCD does not prevent assembly of SpoIIIE on the mother cell side‚ forespore-specific activities are probably required.

2.4 Checkpoints and forespore and mother cell specific transcription The genetic asymmetry that exists between the newly formed mother cell and prespore compartments (see Fig. 1) may be the catalyst for the cell-type specific transcription of genes that follows during the remaining stages of sporulation. For example‚ one gene that is important for the inhibition of (the sporulation sigma factor)‚ spoIIAB‚ is found on the untranslocated portion of the chromosome just after septum formation‚ so it is not available for transcription and translation in the forespore compartment. Furthermore‚ the protein SpoIIAB‚ is subject to proteolytic control by the ClpXP protease (Pan et al.‚ 2001). Since SpoIIAB is degraded in the forespore and cannot be re-synthesized on the forespore side‚ there may no longer be enough SpoIIAB to inhibit sigF transcription in the newly formed compartment. This would allow forespore-specific gene expression to occur (Dworkin and Losick‚ 2001). The link between the morphological event of septum formation and gene expression can be viewed as a checkpoint that prevents prespore development in the absence of cell division (Feucht et al.‚ 1999). Even though the formation of the septum is an important event in spore formation‚ it is not the only factor regulating the progression of the developmental program. In fact‚ the cell division checkpoint can be overcome by mutations in the SpoIIE protein which‚ in addition to its role in division site placement‚ possesses a phosphatase activity which dephosphorylates an anti-anti sigma factor‚ SpoIIAA~P (Feucht et al.‚ 2002). SpoIIAA binds the factor SpoIIAB‚ discussed in the previous paragraph‚ thereby leading to activation and continuation of the developmental pathway. In sporulating wild-type cells‚ the phosphatase

110

Wagner and Brun

activity of SpoIIE is coupled to formation of the septum. There are several hypotheses concerning how this regulation is achieved. One idea‚ which is supported by some experimental evidence‚ is that the SpoIIE protein is enriched on the prespore side of the septum (Wu et al.‚ 1998). Alternatively‚ in a manner similar to SpoIIAB‚ the genetic asymmetry created by the chromosome distribution may lead to a net loss of a SpoIIE inhibitor on the prespore side of the septum. Ultimately‚ many factors probably contribute to SpoIIE phosphatase regulation.

3.

Streptomyces coelicolor

Streptomyces coelicolor is different from the other bacteria discussed in this chapter because in addition to differentiation it carries out to produce a new cell type‚ it also forms a multicellular structure when it grows vegetatively on a solid substrate. S. coelicolor grows by filamentation and branching from apical tips‚ coupled with infrequent formation of vegetative septa without cytokinesis. This results in multinucleate‚ long-cell phenotype (Fig. 2). Uniquely‚ this lifestyle makes FtsZ non-essential for colony formation (McCormick et al.‚ 1994). The cells‚ referred to as hyphal filaments (called substrate hyphae during exponential growth)‚ differentiate further by producing multicellular projections called aerial hyphae (Ohnishi et al.‚ 2002). Inside the aerial hyphae are evenly-spaced‚ condensed nucleoids‚ which are compartmentalized by a synchronized cell division process. These compartments eventually become mature endospores (Fig. 2) (Schwedock et al.‚ 1997). Unlike B. subtilis‚ multiple spore progeny are produced because S. coelicolor cells continue to elongate and replicate chromosomes even after committing to sporulation (Chater‚ 2001).

3.1 Early and late sporulation genes Two large classes of sporulation mutants have been characterized in S. coelicolor. Mutants that cannot complete the first stage of sporulation‚ aerial hyphae formation‚ are called bld (for bald) since colonies lack the fuzzy appearance manifested when thousands of aerial hyphae in a colony project out‚ away from the colony base (Fig. 2). Proteolytic degradation contributes to the development of aerial hyphae. ClpP cellular protease mutants do not generate aerial hyphae‚ while overexpression of ClpP leads to aerial hyphae formation earlier in development (de Crecy-Lagard et al.‚ 1999). The second class is called the whi (for white) class‚ because even though these mutants are able to form aerial hyphae‚ they lack the grey pigment characteristic of

Cell division and differentiation

111

spore formation. The whi class of mutants may fall into an early-acting category if they are unable to form division septa (Fig. 2) (Chater‚ 2001).

Figure 2. S. coelicolor differentiation and division. Cells grow vegetatively on solid surfaces as polynucleoidal substrate hyphae with infrequent vegetative septa. When sporulation is triggered‚ growth occurs from apical tips‚ producing aerial hyphae. Tip growth ceases‚ chromosomes (black fill) are evenly partitioned and condensed‚ and cell division occurs in a synchronous process along the entire length of an individual aerial hyphae. This event is followed by maturation and release of mature spores.

3.2 Morphological checkpoints‚ segregation‚ and FtsZ regulation

chromosome

As aerial hyphae form‚ chromosomes are replicated and evenly partitioned along the length of each individual hypha‚ and sporulation septa form (Flärdh et al.‚ 1999). Nucleoid separation and condensation only become detectable by microscopy once septa form (Schwedock et al.‚ 1997). Mutations in an early whi gene‚ whiG‚ lead to continued growth of aerial hyphae that fail to sporulate‚ but segregate chromosomes and produce septa like vegetatively growing cells (Flärdh et al.‚ 1999). It is believed that the expression of whiG constitutes a developmental checkpoint where S. coelicolor becomes committed to sporulation (Flärdh et al.‚ 1999). A developmental program exists which normally represses whiG expression in substrate hyphae‚ since mutants have been isolated which produce spores in substrate hyphae (Ohnishi et al.‚ 2002). Mutations in two other whi genes‚

112

Wagner and Brun

whiA and whiB‚ also lead to a failure to control aerial hyphae growth‚ but the mutants can no longer condense chromosomes‚ form FtsZ-rings‚ or form sporulation septa. The expression of these proteins appears to constitute a developmental checkpoint‚ since aerial growth cessation is apparently critical for sporulation. Finally‚ there is a late whi gene‚ whiH‚ whose transcription peaks during septum formation (Ryding et al.‚ 1998). Strains with a mutation in whiH are able to condense and irregularly partition chromosomes‚ but form only infrequent sporulation septa (Flärdh et al‚ 1999). During sporulation‚ multinucleoidal aerial hyphae are divided into single chromosome-containing compartments through the activity of FtsZ and other cell division proteins. Z-rings form synchronously along the length of individual aerial hyphae‚ resulting in a ladder-like appearance when FtsZ localization is examined by fluorescence microscopy (Schwedock et al.‚ 1997). Considering that the ladders are observed across distances greater than 100 microns‚ it seems highly probable that ftsZ is transcribed from multiple‚ if not all chromosomes. As in B. subtilis‚ FtsZ is strongly upregulated from an alternative promoter during sporulation (Gonzy-Treboul et al.‚ 1992). However‚ unlike the situation in B. subtilis where ftsZ is transcribed from a chromosome present in the undifferentiated cell‚ ftsZ expression from the S. coelicolor sporulation promoter ftsZ2p‚ is spatially restricted to sporulating aerial hyphae (Flärdh et al.‚ 2000). ftsZ2p and whiH mutants closely resemble each other morphologically‚ and it is hypothesized that WhiH activates transcription from ftsZ2p (Flärdh et al.‚ 2000). Obviously‚ spore septum formation is a highly regulated event since morphological checkpoints‚ developmentally regulated gene expression patterns‚ and cell-type specific gene expression all contribute to the fidelity of the process. Once the sporulation septa form‚ spore-specific transcription is carried out. This leads to maturation of the spore coat‚ pigment synthesis and modification‚ and DNA condensation. Several of these activities are dependent on the expression of SigF‚ a late-acting sporulation transcription factor (Keleman et al.‚ 1996). Thus‚ as in B. subtilis‚ development of the spore is tied to compartment-specific gene transcription made possible by septum formation.

4.

Swarm cell differentiation

Swarming motility is a type of surface-growth cell dispersal that occurs in many bacterial genera including Bacillus and Escherichia‚ but has been most extensively studied in Proteus mirabilis. P. mirabilis cells grow

Cell division and differentiation

113

vegetatively until they are triggered to swarm‚ at which point they elongate by inhibiting cell division while continuing chromosome replication. At this time‚ the cells develop numerous peritrichous flagella. These long‚ multinucleoidal cells line up along longitudinal axes‚ then migrate as a unit via harmonized flagellar rotation‚ away from the previously colonized region (Fraser and Hughes‚ 1999). After a period of swarming‚ which is influenced in part by nutrient source and population dynamics‚ the mass of cells stops migrating and recommences vegetative growth in a process referred to as "consolidation" (Fig. 3) (Rauprich et al.‚ 1996). The actual triggers for swarming are still being elucidated‚ but extracellular stimuli such as surface tension appear to play a major role (Fraser and Hughes‚ 1999).

Figure 3. Swarm cell differentiation. A signal to cells growing on a solid substrate triggers expression of flhD‚ which in turn activates transcription of flagellar genes. Cells become hyperflagellated and replicate chromosomes (indicated by filled ovals) without cytokinesis. Masses of mutlinucleoidal cells align‚ migrate as a mass‚ consolidate‚ and resume cell division.

4.1 Regulation of cell division Swarming differentiation is linked to cell division through the flhDC master operon. flhDC governs expression of flagellar proteins‚ and the FlhD

114

Wagner and Brun

protein‚ a transcriptional activator‚ is essential for expression of other flagellar genes. flhD expression represses cell division in several organisms‚ including P. mirabilis (Furness et al.‚ 1997)‚ Serratia liquefaciens (Eberl et al.‚ 1996)‚ and E. coli (Prüß and Matsumura‚ 1996). It is not known how FlhD inhibits cell division in P. mirabilis‚ but in E. coli it acts indirectly through a central metabolic pathway (Prüß and Matsumura‚ 1996). The products of flhDC are subject to various levels of control that prevent division inhibition and swarming motility at inappropriate times. They are transcriptionally regulated through an unknown pathway. In addition‚ the proteins themselves are proteolytically controlled by the Lon protease (Claret and Hughes‚ 2000)‚ a situation that has also been shown to regulate swarmer cell differentiation in another swarming bacterium‚ Vibrio parahaemolyticus (Stewart et al.‚ 1997). In E. coli‚ an mRNA-binding protein‚ CsrA‚ contributes to the stability of flhDC mRNA in vitro (Wei et al.‚ 2001). A homolog of CsrA‚ RsmA‚ is found in many Gram-negative bacteria‚ and has recently been shown to repress swarming motility in P. mirabilis‚ although it remains to be seen whether the protein can bind flhDC mRNA (Liaw et al.‚ 2003). If it does‚ RsmA may be acting to regulate flagellar biosynthesis and cell division inhibition at the post-transcriptional level by either destabilizing or preventing translation of flhDC mRNA.

5.

Caulobacter crescentus

Caulobacter crescentus begins life as a motile‚ polarly flagellated cell called a swarmer. Swarmer cells are morphologically suited for dispersal. They are motile and unable to initiate DNA replication. After an obligatory period of time (about one third of the complete cell cycle) the flagellum is shed‚ and at the same pole a stalk tipped with an adhesive holdfast is generated. The holdfast is likely generated first‚ with stalk elongation resulting in the holdfast being located at the tip of the stalk. Stalked cells are sessile‚ capable of fixation to surfaces via the holdfast‚ and are competent for repeated rounds of growth and DNA replication‚ releasing a new swarmer cell with each division (Fig. 4) (Brun et al.‚ 1994). The predivisional C. crescentus cell provides an excellent example of cellular asymmetry. At one pole of the predivisional cell is the stalk‚ and at the opposite pole‚ is a flagellum. The division septum demarcates the two incipient cells‚ falling at three-fifths of the cell length away from the stalked pole.

Cell division and differentiation

115

5.1 Transcriptional and proteolytic regulation of division Two mechanisms regulating cell division in C. crescentus are the cell cycle-dependent expression of cell division genes‚ and the subsequent degradation of their protein products. Expression of ftsZ begins during swarmer to stalked cell differentiation‚ coincident with the degradation of the CtrA protein‚ which represses ftsZ transcription in swarmer cells (Fig. 4) (Kelly et al.‚ 1998). CtrA is a global response regulator that is inherited only by swarmer cells after cell division (Quon et al.‚ 1996; Domian et al.‚ 1997)‚ and is required for the expression of approximately 150 cell cycle regulated genes (Laub et al.‚ 2000).

Figure 4. C. crescentus life cycle. Chromosomes are indicated by black ovals. Dark shading represents the presence of CtrA. Light shading represents the degradation of CtrA. During progression through the cell cycle‚ the swarmer cell flagellum is ejected and pili are retracted. At the same pole‚ an adhesive holdfast and stalk are synthesized. The cell elongates‚ completes chromosome replication‚ and differentiates into a predivisional cell which possesses both a flagellum and stalk‚ but lacks a division septum. After division‚ a replication-incompetent swarmer cell is released‚ while the mother cell repeats the cycle of chromosome replication‚ growth‚ and division.

116

Wagner and Brun

After swarmer cell differentiation‚ transcription of ftsZ continues while the chromosome is replicated. FtsZ is at maximal levels in pre-divisional cells that have a visible constriction‚ and greatly decreases in abundance as division progresses‚ coincident with the reappearance of CtrA in predivisional cells (Kelly et al.‚ 1998). The transcription of ftsZ nicely correlates with FtsZ protein levels. FtsZ is absent in swarmer cells (where it is not expressed)‚ begins appearing during differentiation‚ peaks in predivisional cells‚ and then begins decreasing in half-life as the cell constricts (Quardokus et al.‚ 1996; Kelly et al.‚ 1998). Following the expression of ftsZ‚ the expression of other cell division genes occurs at a distinct time in the cell cycle with the order ftsI‚ ftsW‚ ftsQ‚ and ftsA (Sackett et al.‚ 1998; Laub et al.‚ 2000). The late expression of ftsQ and ftsA is consistent with their requirement for the maturation of the initial FtsZ-dependent constriction (Ohta et al.‚ 1997). CtrA‚ the global cell cycle response regulator that represses ftsZ expression‚ also serves as a transcriptional activator of ftsQ and ftsA‚ which are co-transcribed from the promoter Pqa (Sackett et al.‚ 1998; Wortinger et al.‚ 2000). The burst of Pqa transcription occurs following the completion of DNA replication‚ when CtrA synthesis begins in predivisional cells. Thus‚ by repressing ftsZ and activating ftsQA‚ the cycling of CtrA levels ensures that the expression of ftsZ precedes that of ftsQ and ftsA‚ linking the progression of cell division to the process of differentiation. Recent experiments indicate that FtsQ and FtsA are also regulated by cell cycle proteolysis (Martin and Brun‚ in preparation). Finally‚ genome-wide location analysis of CtrA binding sites has shown that CtrA directly regulates ftsW expression (Laub et al.‚ 2002).

5.2

Coordination of DNA replication and cell division

Swarmer cells are unable to initiate chromosome replication and cell division. Different mechanisms combine to ensure that these two processes are coordinated during the normal cell cycle and in response to a block in DNA replication. Swarmer cell DNA replication is inhibited by CtrA‚ which binds to five sites in the origin of replication (Quon et al.‚ 1998; Marczynski and Shapiro‚ 2002). Although the mechanism is not fully understood‚ CtrA binding at the origin probably functions‚ at least in part‚ to prevent binding of the DnaA protein. This would prevent unwinding of the DNA and loading of the replisome protein complex. In addition to blocking DNA replication‚ CtrA inhibits transcription of ftsZ‚ thus coordinating the initiation of DNA replication and cell division (Kelly et al.‚ 1998). Degradation of CtrA by the ClpXP protease (Jenal and Fuchs‚ 1998) during swarmer cell differentiation (Domian et al.‚ 1997) relieves the repression of replication‚ and allows FtsZ expression. However‚ FtsZ levels alone do not govern whether cell division

117

Cell division and differentiation

will be initiated in C. crescentus‚ since artificially expressing FtsZ in swarmer cells does not lead to Z-ring formation in those cells (Quardokus et al.‚ 2001). In fact‚ cell division in C. crescentus is regulated by at least two DNA replication checkpoints. Cells inhibited for DNA replication do not form midcell Z-rings and generate long‚ smooth‚ filamentous cells (Quardokus and Brun‚ 2002). Inhibition of DNA replication also prevents resynthesis of CtrA after swarmer cell differentiation‚ preventing transcription of the late cell division genes ftsQ and ftsA (Wortinger et al.‚ 2000). Since CtrA is required for the expression of many developmental genes‚ this DNA replication checkpoint is of major importance in coordinating development and cell division with DNA replication.

5.3 Regulation of cell partitioning proteins

division

by

chromosome

In addition to DNA replication‚ proper chromosome partitioning is required for cell division. The ParA and ParB homologues of plasmid and phage partitioning proteins are required for the initiation of cell division‚ while the ParC and ParE topoisomerase IV (Topo IV) subunits are required for late stages of cell division. ParA and ParB are chromosomally encoded and essential in C. crescentus. ParB binds to parS sites located close to the origin of replication‚ and in swarmer cells‚ ParB colocalizes with the origin of replication at the flagellated pole (Mohl and Gober‚ 1997; Mohl et al.‚ 2001). Soon after the initiation of DNA replication‚ one copy of the origin of replication moves to the opposite pole of the cell (Jensen and Shapiro‚ 1999; Jensen et al.‚ 2001). ParB exhibits a bipolar localization pattern (Mohl and Gober‚ 1997; Mohl et al.‚ 2001)‚ suggesting that it is bound to parS sites at the replicated origin of replication. ParA also exhibits a bipolar localization pattern in predivisional cells‚ and it probably forms a complex with ParB at the origin (Mohl and Gober‚ 1997). Overexpression of ParA or ParB inhibits cell division and leads to a chromosome partitioning defect (Mohl and Gober 1997). Depletion of ParB prevents Z-ring formation and thus blocks the initiation of cell division (Mohl et al.‚ 2001). It is thought that a fraction of ParA is free in the cell until a second partitioning complex is formed. Free ParA would act to inhibit cell division by preventing the formation of stable Z-rings‚ either by interfering with Z-ring formation itself or by regulating the transcription of a gene whose product modulates Z-ring formation or stability. Formation of a second ParB-parS partitioning complex‚ once replication has been initiated‚ would recruit ParA and allow cell division initiation. In support of this model is the fact that ParB localization at the second pole occurs 20 minutes prior to Z-ring formation (Mohl et al.‚ 2001).

118

Wagner and Brun

ParB-stimulated nucleotide exchange of ParA may provide a switch that regulates cell division. ParA can exist in an ATP or ADP bound form (Easter and Gober‚ 2002). Upon hydrolysis of ATP‚ ParA is bound to ADP and exhibits single-stranded DNA binding activity that may regulate the expression of a cell division gene. ParB promotes nucleotide exchange‚ thus returning ParA to the ATP bound form. An increase of ParA relative to ParB results in more ADP bound ParA and the inhibition of cell division (Easter and Gober‚ 2002). Interestingly‚ some parB alleles block cell division in the presence of wild-type parB but do not cause chromosome partitioning defects. This observation raises the interesting possibility that the main function of ParAB is to regulate cell division initiation using chromosome localization as a signal (Figge et al.‚ 2003). Whether ParA and ParB are involved in chromosome partitioning per se or are involved in communicating the state of chromosome replication and/or partitioning to the cell division machinery remains to be determined. However‚ it is clear that the completion of chromosome segregation is required for cell division. Conditional mutants in parC and parE‚ encoding the two subunits of Topo IV‚ are deficient in cell division at the nonpermissive temperature. Under these conditions‚ the par mutants form multiply constricted‚ filamentous cells that are unable to complete late stages of cell division (Ward and Newton‚ 1997). DNA segregation defects can only be observed by combining the parC and parE mutations with ftsA mutations that block cell division at an earlier stage. The parC and parE C. crescentus mutants do not produce anucleate cells‚ in contrast to similar mutants in E. coli‚ because they have no residual cell division at the nonpermissive temperature. The difference between the segregation and division phenotypes of Topo IV mutants in C. crescentus and E. coli is thought to be due to the tight regulation of cell cycle progression in C. crescentus. Thus‚ at least chromosome decatenation is required for the latest stages of cell division. The division inhibition that results from inactivation of Topo IV may be caused by the presence of DNA throughout the sites of constriction. Alternatively‚ proper chromosome segregation could provide a checkpoint for the late stages of cell division.

5.4 Regulation of cell Division by flagellum assembly In addition to being regulated by the DNA replication cycle‚ cell division is influenced by flagellum assembly. Flagellum assembly is a complex process resulting from the expression of more than 50 genes whose cell cycle expression coincides with the time of action of their gene products (Brun et al.‚ 1994). Class II genes‚ which encode the flagellum secretion machinery‚ the innermost flagellar ring‚ and the flagellar transcription factors

Cell division and differentiation

119

and FlbD are transcribed first‚ followed by Class III genes‚ which encode outer ring‚ rod and hook proteins‚ and finally by Class IV genes‚ encoding flagellins. Expression of genes in early classes is required for the transcription of the genes in the subsequent classes. Mutations in Class II genes cause a delay in cell division‚ resulting in the formation of filamentous cells. The cell division defect of Class II mutants is due to a delay or instability in FtsZ ring formation (Muir and Gober‚ 2001). The effect of Class II mutations on late flagellar gene transcription and on cell division is due to an alteration in the activity of the FlbD response regulator. Gain-offunction mutations in FlbD‚ called bfa‚ restore transcription of Class III and IV genes in Class II mutant backgrounds. These bfa alleles of FlbD also restore normal FtsZ ring formation and cell division in Class II mutants (Muir and Gober‚ 2002). Thus‚ the assembly of Class II flagellar components is thought to exert a stimulatory effect on cell division via the activation of FlbD. FlbD is likely to activate the transcription of a gene required for optimal FtsZ ring assembly. This checkpoint coupling early flagellar assembly to cell division may be important because of the timing and regulation of Class II flagellar genes. Class II gene transcription is activated by CtrA early in the cell cycle and prior to cell division initiation. Later in the cell cycle‚ when the predivisional cell is compartmentalized by the cell division barrier‚ CtrA is only present in the swarmer compartment where it represses DNA replication and Class II gene transcription. Therefore‚ positive regulation of cell division by the assembly of Class II proteins ensures that sufficient expression of these proteins has occurred before cell division causes the transcriptional shut down of their genes. This is in contrast to the situation in swarming bacteria‚ such as P. mirabilis‚ S. liquefaciens‚ and E. coli‚ where expression of an early flagellar gene‚ flhD‚ leads to an inhibition of cell division (Furness et al.‚ 1997; Eberl et al.‚ 1996; and Matsumura‚ 1996).

5.5 Cell division regulates polar development C. crescentus maintains cellular asymmetry by utilizing checkpoints that ensure the coordination of development and the cell cycle. Cell division and polar development are regulated by a DNA replication checkpoint. Cells inhibited for DNA replication are blocked for cell division and form long filamentous cells (Degnen and Newton‚ 1972). Blocking DNA replication inhibits development prior to flagellum synthesis and prevents any further polar development (Degnen and Newton‚ 1972). The replication checkpoint is probably mediated by the control of CtrA synthesis in predivisional cells‚ as explained previously (Wortinger et al.‚ 2000). In addition to DNA

120

Wagner and Brun

replication‚ cell division is required for polar development. Cells that can replicate DNA but that are blocked in cell division are affected in their progression through development (Terrana and Newton‚ 1976; Huguenel and Newton‚ 1982; Ohta and Newton‚ 1996; Newton and Ohta‚ 2003). Normally‚ a flagellum is synthesized at the new pole of the cell. When cell division is inhibited‚ a flagellum is synthesized; however‚ its rotation cannot be activated (Fig. 5). Under these conditions‚ DNA replication continues‚ but the shedding of the flagellum and the synthesis of pili‚ stalk and holdfast is blocked. If cells are blocked at a late stage of cell division‚ long filaments with regularly spaced constrictions are formed. These cells are able to activate the development of the flagellar pole to yield filamentous cells that have a stalk and holdfast at both poles (Fig. 5). Thus‚ polar development is not simply coupled to progression through the DNA replication cycle or to increase in cell mass; it is dependent on early stages of cell division as well.

Figure 5. DNA replication and cell division checkpoints and their influence on polar development in C. crescentus. The diagram shows the fate of a stalked cell (S) in different conditions. The black dot follows the fate of the new pole of the beginning stalked cell.

It has recently been shown that RpoN and the response regulator TacA are required for the cell division checkpoint (Quardokus and Brun‚ in preparation). The molecular mechanism of the cell division

Cell division and differentiation

121

checkpoint is unknown‚ but one possible target for the checkpoint is the proteolytic processing of PodJ. PodJ is localized at the flagellar pole of predivisional cells following its expression in mid-cell cycle (Viollier et al.‚ 2002; Hinz et al.‚ 2003). PodJ is required for localization of the histidine kinase PleC‚ a critical regulator of polar development (Newton and Ohta‚ 2003)‚ and is thought to regulate PleC activity. The full length form of PodJ‚ is processed to the short form‚ during the late stages of cell division. It is possible that the processing of couples polar development to cell division by regulating the activity of PleC.

6.

Other models

6.1 Bdellovibrio Bdellovibrio is a small‚ parasitic bacterium that invades the periplasm of Gram-negative bacteria‚ where it differentiates into a growth-phase cell which is a nonmotile‚ polynucleoidal‚ aseptate filament that feasts on the cytoplasm of the dead host cell (Fig. 6) (Ruby and Rittenberg‚ 1983). When the reserves from the host cell become limiting‚ a developmental process is triggered that results in the formation of multiple septa‚ followed by flagellation‚ and release of replication-incompetent attack-phase progeny (Ruby and Rittenberg‚ 1983). This is similar to the arrangement in C. crescentus‚ where the motile‚ dispersal cell is incapable of initiating chromosome replication (Brun et al.‚ 1994). After penetration into the host‚ Bdellovibrio can be synchronized to proceed through their lifecycle. This method can be utilized to examine protein and DNA synthesis at different stages of intracellular growth (Thomashow and Rittenberg‚ 1978)‚ and has been used in conjunction with 2D gel electrophoresis to reveal over 30 proteins that are developmentally regulated (McCann et al.‚ 1998). 6.1.1 Differentiation and DNA replication The penetration of Bdellovibrio into the host-cell perils is followed by flagellum ejection and a host-factor dependent conversion of the host cell into a stable‚ spherical structure called a bdelloplast (Thomashow and Cotter‚ 1992). The bdelloplast provides Bdellovibrio with an appropriate environment to begin growth and DNA replication‚ without division. Based on DNA synthesis rates‚ it is likely that replication proceeds from multiple chromosomes (Gray and Ruby‚ 1989).

122

Wagner and Brun

In Bdellovibrio‚ there also appear to be developmental checkpoints controlling differentiation and cell division processes. Growth-phase cells can be prematurely released from bdelloplasts with a lysozyme treatment‚ allowing the commitment to different stages of host dependent growth to be examined (Ruby and Rittenberg‚ 1983). Cells in the transition between bdelloplast formation and DNA replication (30-50 minutes) are unable to initiate DNA synthesis or fragment into individual cells‚ but form flagella after release (Ruby and Rittenberg‚ 1983). Cells which are released between 75 and 120 minutes‚ when DNA replication has already initiated‚ are able to finish the initiated round of replication‚ divide‚ and become flagellated‚ although with some delay (Ruby and Rittenberg‚ 1983). Since the delay corresponds to the time it would take for chromosome replication to complete‚ it is hypothesized that DNA synthesis and cell division are coupled processes (Ruby and Rittenberg‚ 1983). Additionally‚ host cell extracts can be obtained that allow attack-phase progeny to initiate DNA synthesis without differentiation into growth-phase cells (Ruby and Rittenberg‚ 1983). These results indicate that sometime after invasion‚ a host factor-dependent commitment to bdelloplast formation and growth occurs‚ and that this commitment is separate from the host factor-dependent commitment to DNA synthesis. Both stages probably constitute developmental checkpoints (Fig. 6).

Figure 6. Bdellovibrio infection of a Gram-negative bacterium. Chromosomes are indicated by black ovals. Bdellobvibrio enters the periplasm of the host cell and becomes a non-motile growth-phase cell which undergoes repeated rounds of DNA replication without cell division. When host cell reserves become limiting‚ chromosome replication initiation is inhibited and cell division is triggered. This is followed by lysis of the bdelloplast and release of motile attack phase cells.

Cell division and differentiation

123

6.2 Sinorhizobium Bacteria belonging to the genus Sinorhizobium have two distinct lifestyles. In the environment‚ Sinorhizobium exists as a free-living soil bacterium that grows vegetatively as a rod. It can also form a symbiotic relationship with leguminous plants such as clover and alfalfa. In the rhizosphere around root systems‚ compounds released by plants induce the bacterium to produce a factor that in turn stimulates cell division in the plant root (Oke and Long‚ 1999a). This factor is called Nod factor because it ultimately leads to the formation of a nodule on the plant root. The bacteria infect the nodule‚ enter the plant cells through an endocytic-type process‚ and initiate a developmental program that results in a non-dividing cell type referred to as a bacteroid (Oke and Long‚ 1999b). Bacteroids are capable of fixing atmospheric nitrogen‚ which they pass on to the plant in the form of ammonia. It is not known how the bacterium benefits the plant (Oke and Long‚ 1999b). It is also not known whether differentiated bacteroids are capable of resuming vegetative growth when they are released back into the soil when the plant dies (Oke and Long‚ 1999b). 6.2.1 Cell division and differentiation

Sinorhizobium bacteroids are pleomorphic‚ forming Y-shaped cells with branches and buds (Latch and Margolin‚ 1997; Oke and Long‚ 1999b). These unusual forms may be the result of the inhibition in cell division that occurs in the symbiotic relationship with the plant‚ since pleomorphism can be induced during vegetative growth under conditions that inhibit cell division in S. meliloti (previously called Rhizobium meliloti) (Latch and Margolin‚ 1997). In plant associations‚ Sinorhizobium which have entered host cells generally only continue dividing if the plant cells themselves are dividing (Oke and Long‚ 1999b). Although the signals which inhibit division in bacteroids are not known‚ the connection between plant division and bacterial division suggests that the process is regulated by the host (One and Long‚ 1999b). S. meliloti has two FtsZ proteins that differ only in their Cterminal domains. FtsZ1‚ which contains an extended C-terminal domain‚ is essential‚ whereas FtsZ2‚ which lacks the domain‚ is non-essential during vegetative growth (Margolin and Long‚ 1994; Latch and Margolin‚ 1997). It will be of interest to determine if the FtsZ2 plays a specific role in bacteroid formation‚ especially with respect to cell division and pleomorphism.

124

Wagner and Brun

6.3 Hyphomonas Hyphomonas neptunium (previously Hyphomicrobium neptunium) is a type of prosthecate bacterium‚ which like C. crescentus‚ reproduces in an asymmetric manner. The daughter cell‚ or swarmer‚ possesses a single polar flagellum. During development‚ the flagellum is shed and the cell increases in size until it is between two and four times larger in volume. At this time‚ H. neptunium cells begin synthesizing a prosthecum‚ a structure that is continuous with the cell body. Unlike the stalk of C. crescentus‚ the H. neptunium prosthecum can accommodate ribosomes and DNA (Poindexter 1964). At the tip of the prosthecum‚ distal to the cell body‚ a swelling bud forms. Macromolecules‚ including DNA‚ are shuttled from the mother cell to the maturing bud via pseudovesicles‚ which appear to be continuous with the cytoplasmic membrane of the mother cell (Zerfas et al.‚ 1997). Bud formation is followed by septum formation‚ synthesis of a polar flagellum‚ and upon division‚ subsequent release of the new swarmer cell (Fig. 7) (Weiner and Blackman‚ 1973). The mother cell can generate several more swarmer cells by budding from the tip of the previously synthesized prosthecum. Although little is known about the processes that regulate development and cell division in H. neptunium‚ the sequencing of its genome in 2003 will likely spur investigations into the organism's reproductive lifestyle. At least one checkpoint exists that links development to DNA replication; swarmer cells that are inhibited for DNA replication by the antibiotic nalidixic acid are able to increase in volume and elongate a prosthecum‚ but are unable to form buds (Fig. 7) (Weiner and Blackman‚ 1973). Thus‚ as in many other bacteria that undergo developmental processes‚ DNA replication is tied to morphogenic events. In H. neptunium and other prosthecate‚ budding bacteria‚ cellular compartmentalization between the mother and daughter cells is largely brought about by the spatial separation of the two cells. Even though they are connected by the prosthecum before division‚ it is unlikely that molecules travel freely between the two cell types. In the future it will be of interest to determine the mechanisms governing the morphogenic asymmetry between the mother cell and swarmer‚ including how replicated DNA is translocated down the prosthecum in a directional manner‚ the role of FtsZ and other cell division proteins in regulating cell division‚ and what factors regulate the various stages of the H. neptunium life cycle.

Cell division and differentiation

125

Figure 7. H. neptunium life cycle. A. During growth‚ the swarmer cell ejects its flagellum‚ begins enlarging‚ and initiates DNA replication. A prosthecum elongates‚ and at the tip a bud swells. The replicated chromosome is translocated to the bud‚ and upon cell division a new swarmer is released. B. When DNA replication is inhibited in swarmer cells‚ cells enlarge and elongate prosthecae‚ but are unable to form buds.

7.

Conclusion

Cell division and differentiation are interwoven processes in biology. Nowhere is this more apparent than in differentiating bacteria‚ where division creates cellular compartmentalization and allows differential gene expression‚ and therefore different fates‚ in progeny cells. This is certainly the situation for B. subtilis and S. coelicolor‚ where formation of the division septum is a prerequisite for maturation of the progeny spores. Many bacteria also coordinate developmental processes and cell division in order to control progeny dispersal. In Bdellovibrio‚ P. mirabilis‚ and S. coelicolor‚ chromosomes are replicated and cells continue to grow without cell division. This produces "dispersal-ready" filaments of cells that will produce a crop of progeny upon division. C. crescentus and H. neptunium utilize a somewhat

126

Wagner and Brun

similar strategy‚ in that they use cell cycle checkpoints to ensure that cell division does not occur until the dispersal cell is ready for release. These checkpoints likely evolved to ensure the faithful execution of the developmental programs in C. crescentus and H. neptunium. Finally‚ there is the case of S. meliloti‚ where cell division in the bacteroid appears to be inhibited by signals from the plant symbiont. In this instance‚ regulation of bacterial division may have been commandeered by the legume in order to promote differentiation of the bacteroid. Until it is clear whether bacteroids are terminally differentiated‚ this will remain an open question. Bacteria have evolved various mechanisms to exploit cell division during differentiation. These include developmental checkpoints‚ compartmentspecific proteolysis‚ compartment-specific gene expression‚ genetic asymmetry‚ and subcellular localization of cell-cycle regulated proteins. It is not surprising that cell division‚ one of the earliest reproductive steps‚ is tightly linked to more evolved developmental processes. The next big challenge will be to determine how cells integrate each of these mechanisms for regulating differentiation with the changes in cell morphology and physiology that occur during division and development.

Acknowledgements We would like to thank the members of the Brun Laboratory for their helpful comments on the text. Work in our laboratory is supported by National Institutes of Health Grant (GM51986) and a National Science Foundation CAREER Award (MCB-9733958) to Y.V.B‚ and by a National Institutes of Health Predoctoral Fellowship (GM07757) to J.W.

References Barak‚ I.‚ Prepiak‚ P. and Schmeisser‚ F. (1998) MinCD proteins control the septation process during sporulation of Bacillus subtilis. J. Bacteriol. 180‚ 5327-5333. Bath‚ J.‚ Wu‚ L.‚ Errington‚ J., Wang‚ J. (2000) Role of Bacillus subtilis SpoIIIE in DNA transport across the mother cell-prespore division septum. Science 290‚ 995-997. Ben-Yehuda‚ S.‚ and Losick‚ R. (2002) Asymmetric cell division in B. subtilis involves a spiral-like intermediate of the cytokinetic protein FtsZ. Cell 109‚ 257-266. Brun‚ Y.‚ Marczynski‚ G. and Shapiro‚ L. (1994) The expression of asymmetry during Caulobacter cell differentiation. Ann. Rev. Biochem. 63‚ 419-450. Brun‚ Y. and Shimkets‚ L. (eds). (2000) Prokaryotic Development. Washington‚ D.C: ASM Press.

Cell division and differentiation

127

Burkholder‚ W.‚ Kurtser‚ I.‚ and Grossman‚ A. (2001) Replication initiation proteins regulate a developmental checkpoint in Bacillus subtilis. Cell 104‚ 269-279. Cha‚ J. and Stewart‚ G. (1997) The divIVA minicell locus of Bacillus subtilis. J. Bacteriol. 179‚ 1671-1683. Chater‚ K. (2001) Regulation of sporulation in Streptomyces coelicolor A3(2): a checkpoint multiplex? Curr. Opin. Microbiol. 4‚ 667-673. Claret‚ L.‚ and Hughes‚ C. (2000) Rapid turnover of FlhD and FlhC‚ the flagellar regulon transcriptional activator proteins‚ during Proteus swarming. J. Bacteriol. 182‚ 833-836. de Crecy-Lagard‚ V.‚ Servant-Moisson‚ P.‚ Viala‚ J.‚ Grandvalet‚ C.‚ and Mazodier‚ P. (1999) Alteration of the synthesis of the Clp ATP-dependent protease affects morphological and physiological differentiation in Streptomyces. Mol. Microbiol. 32‚ 505-517. Degnen‚ S. and Newton‚ A. (1972) Dependence of cell division on the completion of chromosome replication in Caulobacter crescentus. J. Bacteriol. 110‚ 852-856. Domian‚ I.‚ Quon‚ K. and Shapiro‚ L. (1997) Cell type-specific phosphorylation and proteolysis of a transcriptional regulator controls the G1-to-S transition in a bacterial cell cycle. Cell 90‚ 415-424 Dworkin‚ J. and Losick‚ R. (2001) Differential gene expression governed by chromosomal spatial asymmetry. Cell 107‚ 339-346. Easter‚ J.‚ Jr. and Gober‚ J. (2002) ParB-stimulated nucleotide exchange regulates a switch in functionally distinct ParA activities. Mol. Cell 10‚ 427-434. Eberl‚ L.‚ Christiansen‚ G.‚ Molin‚ S. and Givskov‚ M. (1996) Differentiation of Serratia liquefaciens into swarm cells is controlled by the expression of the flhD master operon. J. Bacteriol. 178‚ 554-559. Edwards‚ D.H. and Errington‚ J. (1997) The Bacillus subtilis DivIVA protein targets to the division septum and controls the site specificity of cell division. Mol. Microbiol. 24‚ 905915. Eichenberger‚ P.‚ Fawcett‚ P. and Losick‚ R. (2001) A three-protein inhibitor of polar septation during sporulation in Bacillus subtilis. Mol. Microbiol. 42‚ 1147-1162. Errington‚ J. 1996. Determination of cell fate in Bacillus subtilis. Trends in Genet. 12‚ 31-34. Feucht‚ A.‚ Abbotts‚ L. and Errington‚ J. (2002) The cell differentiation protein SpoIIE contains a regulatory site that controls its phosphatase activity in response to asymmetric septation. Mol. Microbiol. 45‚ 1119-1130. Feucht‚ A.‚ Daniel‚ R. and Errington‚ J. (1999) Characterization of a morphological checkpoint coupling cell-specific transcription to septation in Bacillus subtilis. Mol. Microbiol. 33‚ 1015-1026. Figge‚ R. M.‚ Easter Jr.‚ J. and Gober‚ J. (2003) Productive interaction between the chromosome partitioning proteins‚ ParA and ParB‚ is required for the progression of the cell cycle in Caulobacter crescentus. Mol. Microbiol. 47‚ 1225-1237. Flärdh‚ K.‚ Findlay‚ K. and Chater‚ K. (1999) Association of early sporulation genes with suggested developmental decision points in Streptomyces coelicolor A3(2). Microbiology 145‚ 2229-2243. Flärdh‚ K.‚ Leibovitz‚ E.‚ Buttner‚ M.‚ and Chater‚ K. (2000) Generation of a non-sporulating strain of Streptomyces coelicolor A3(2) by the manipulation of a developmentally regulated ftsZ promoter. Mol. Microbiol. 38‚ 737-749. Frandsen‚ N.‚ Barak‚ I.‚ Karmazyn-Campelli‚ C. and Stragier‚ P. (1999) Transient gene asymmetry during sporulation and establishment of cell specificity in Bacillus subtilis. Genes and Dev. 13‚ 394-399. Fraser‚ G.M.‚ and Hughes‚ C. (1999) Swarming Motility. Curr. Opin. Microbiol. 2‚ 630-635.

128

Wagner and Brun

Furness‚ R.‚ Fraser‚ G.‚ Gay‚ N. and Hughes‚ C. (1997) Negative feedback from a Proteus class II flagellum export defect to the flhDC master operon controlling cell division and flagellum assembly. J. Bacteriol. 179‚ 5585-5588. Gonzy-Treboul‚ G.‚ Karmazyn-Campelli‚ C. and Stragier‚ P. (1992) Developmental regulation of transcription of the Bacillus subtilis ftsAZ operon. J. Mol. Biol. 224‚ 967-979. Gottesman‚ S. 1999. Regulation by proteolysis: developmental switches. Curr. Opin. Microbiol. 2‚ 142-147. Gray‚ K.‚ and Ruby‚ E. (1989) Unbalanced growth as a normal feature of development of Bdellovibrio bacteriovorus. Arch. Micro. 152‚ 420-424. Hinz‚ A.‚ Larson‚ D.‚ Smith‚ C. and Brun‚ Y. (2003) The Caulobacter crescentus polar organelle development protein PodJ is differentially localized and is required for polar targeting of the PleC development regulator. Mol Microbiol. 47‚ 929-941. Huguenel‚ E. D. and Newton‚ A. (1982) Localization of surface structures during procaryotic differentiation: Role of cell division in Caulobacter crescentus. Differentiation 21‚ 71-78. Jenal‚ U. and Fuchs‚ T. (1998) An essential protease involved in bacterial cell cycle control. EMBO J. 17‚ 5658-5669. Jensen‚ R.‚ and Shapiro‚ L. (1999) The Caulobacter crescentus smc gene is required for cell cycle progression and chromosome segregation. Proc. Natl. Acad. Sci. U.S.A. 96‚ 1066110666. Jensen‚ R.‚ Wang‚ S. and Shapiro‚ L. (2001) A moving DNA replication factory in Caulobacter crescentus. EMBO J. 20‚ 4952-4963. Keleman‚ G.‚ Brown‚ G.‚ Kormanec‚ J.‚ Potuckova‚ L.‚ Chater‚ K. and Buttner‚ M. (1996) The positions of the sigma-factor genes‚ whiG and sigF‚ in the hierarchy controlling the development of spore chains in the aerial hyphae of Streptomyces coelicolor A3(2). Mol. Microbiol. 21‚ 593-603. Kelly‚ A.‚ Sackett‚ M.‚ Din‚ N.‚ Quardokus‚ E. and Brun‚ Y. (1998) Cell cycle-dependent transcriptional and proteolytic regulation of FtsZ in Caulobacter. Genes Dev. 12‚ 880893. Khvorova‚ A.‚ Zhang‚ L.‚ Higgins‚ M. and Piggot‚ P. (1998) The spoIIE locus is involved in the Spo0A-dependent switch in the location of FtsZ rings in Bacillus subtilis. J. Bacteriol. 180‚ 1256-1260. Latch‚ J.‚ and Margolin‚ W. (1997) Generation of buds‚ swellings‚ and branches instead of filaments after blocking the cell cycle of Rhizobium meliloti. J. Bacteriol. 179‚ 23732381. Laub‚ M.‚ McAdams‚ J.‚ Feldblyum‚ T.‚ Fraser‚ C. and Shapiro‚ L. (2000) Global analysis of the genetic network controlling a bacterial cell cycle. Science. 290‚ 2144-2148. Laub‚ M.‚ Chen‚ S.‚ Shapiro‚ L. and McAdams‚ H. (2002) Genes directly controlled by CtrA‚ a master regulator of the Caulobacter cell cycle. Proc. Natl. Acad. Sci. U.S.A. 99‚ 46324637. Levin‚ P.‚ Margolis‚ P.S.‚ Setlow‚ P.‚ Losick‚ R. and Sun‚ D. (1992) Identification of Bacillus subtilis genes for septum placement and shape determination. J. Bacteriol. 174‚ 67176728. Levin‚ P.A.‚ and Losick‚ R. (1996) Transcription factor Spo0A switches the localization of the cell division protein FtsZ from a medial to a bipolar pattern in Bacillus subtilis. Genes and Dev. 10‚ 478-488. Liaw‚ S.J.‚ Lai‚ H.C.‚ Ho‚ S. W.‚ Luh‚ K.T. and Wang‚ W.B. (2003) Role of RsmA in the regulation of swarming motility and virulence factor expression in Proteus mirabilis. J. Med. Microbiol. 52‚ 19-28.

Cell division and differentiation

129

Lucet‚ I.‚ Feucht‚ S.‚ Yudkin‚ M. and Errington‚ J. (2000) Direct interaction between the cell division protein FtsZ and the cell differentiation protein SpoIIE. EMBO J. 19‚ 1467-1475. Lutkenhaus‚ J. (2002) Unexpected twist to the Z ring. Dev. Cell. 2‚ 519-521. Marczynski‚ G. and Shapiro‚ L. (2002) Control of chromosome replication in Caulobacter crescentus. An. Rev. Microbiol. 56‚ 625-656. Margolin‚ W.‚ and Long‚ S. (1994) Rhizobium meliloti contains a novel second homolog of the cell division gene ftsZ. J. Bacteriol. 176‚ 2033-2043. Marston‚ A.‚ Thomaides‚ G.‚ Edwards‚ D.‚ Sharpe‚ M. and Errington‚ J. (1998) Polar localization of the MinD protein of Bacillus subtilis and its role in selection of the midcell division site. Genes and Dev. 12‚ 3419-3430. McCann‚ M.‚ Solimeo‚ H.‚ Cusick Jr.‚ F.‚ Panunti‚ B. and McCullen‚ C. (1998) Developmentally regulated protein synthesis during intraperiplasmic growth of Bdellovibrio bacteriovorus 109J. Can. J. Microbiol. 44 50-55. McCormick‚ J.‚ Su‚ E.‚ Driks‚ A. and Losick‚ R. (1994) Growth and viability of Streptomyces coelicolor mutant for the cell division gene ftsZ. Mol. Microbiol. 14‚ 243-254. Mohl‚ D.‚ Easter‚ Jr.‚ J. and Gober‚ J. (2001) The chromosome partitioning protein‚ ParB‚ is required for cytokinesis in Caulobacter crescentus. Mol. Microbiol. 42‚ 741-755. Mohl‚ D. and Gober‚ J. (1997) Cell cycle-dependent polar localization of chromosome partitioning proteins in Caulobacter crescentus. Cell 88‚ 675-684. Muir‚ R.‚ and Gober‚ J. (2001) Regulation of late flagellar gene transcription and cell division by flagellum assembly in Caulobacter crescentus. Mol. Microbiol. 41‚ 117-130. Muir‚ R.‚ and Gober‚ J. (2002) Mutations in FlbD that relieve the dependency on flagellum assembly alter the temporal and spatial pattern of developmental transcription in Caulobacter crescentus. Mol. Microbiol. 43‚ 597-615. Newton‚ A. and Ohta‚ N. (2003) Role of Multiple Sensor Kinases in Cell Cycle Progression and Differentiation in Caulobacter crescentus. Histidine Kinases in Signal Transduction. M. Inouye and R. Dutta. San Diego‚ CA‚ Elsevier Science‚ 377-396. Ohnishi‚ Y.‚ Seo‚ J. and Horinouchi‚ S. (2002) Deprogrammed sporulation in Streptomyces. FEMS Micro. Lett. 216‚ 1-7. Ohta‚ N. and Newton‚ A. (1996) Signal transduction in the cell cycle regulation of Caulobacter differentiation. Trends Microbiol. 4‚ 326-332. Ohta‚ N.‚ Ninfa‚ A.‚ Allaire‚ A.‚ Kulick‚ L. and Newton‚ A. (1997) Identification‚ characterization‚ and chromosomal organization of cell division cycle genes in Caulobacter crescentus. J. Bacteriol. 179‚ 2169-2180. Oke‚ V.‚ and Long‚ S. (1999a) Bacterial genes induced within the nodule during the Rhizobium-legume symbiosis. Mol. Microbiol. 32‚ 837-849. Oke‚ V.‚ and Long‚ S. (1999b) Bacteroid formation in the Rhizobium-legume symbiosis. Curr. Opin. Microbiol. 2‚ 641-646. Pan‚ Q.‚ Garsin‚ D. and Losick‚ R. (2001) Self-reinforcing activation of a cell-specific transcription factor by proteolysis of an anti-sigma factor in B. subtilis. Mol. Cell. 8‚ 873883. Perego‚ M. (1998) Kinase-phosphatase competition regulates Bacillus subtilis development. Trends in Microbiol. 6‚ 366-370. Pogliano‚ J.‚ Sharp‚ M. and Pogliano‚ K. (2002) Partitioning of chromosomal DNA during establishment of cellular asymmetry in Bacillus subtilis. J. Bacteriol. 184‚ 1743-1749. Poindexter‚ J. and Cohen-Bazire‚ G. (1964) The fine structure of stalked bacteria belonging to the family Caulobacteraceae. J. Cell Biol. 23‚ 587-607. Prüß‚ B. and Matsumura‚ P. (1996) A regulator of the flagellar regulon of Escherichia coli‚ flhD‚ also affects cell division. J. Bacteriol. 178‚ 668-674.

130

Wagner and Brun

Quon‚ K. C.‚ Marczynski‚ G. T. and Shapiro‚ L. (1996) Cell cycle control by an essential bacterial two-component signal transduction protein. Cell. 84‚ 83-93. Quon‚ K.‚ Yang‚ B.‚ Domian‚ I.‚ Shapiro‚ L. and Marczynski‚ G. (1998) Negative control of bacterial DNA replication by a cell cycle regulatory protein that binds at the chromosome origin. Proc. Natl. Acad. Sci. U.S.A. 95‚ 120-125. Quardokus‚ E.‚ Din‚ N. and Brun‚ Y. (1996) Cell cycle regulation and cell type-specific localization of the FtsZ division initiation protein in Caulobacter. Proc. Natl. Acad. Sci. U.S.A. 93‚ 6314-6319. Quardokus‚ E.‚ Din‚ N. and Brun‚ Y. (2001) Cell cycle and positional constraints on FtsZ localization and the initiation of cell division in Caulobacter crescentus. Mol. Microbiol. 39‚ 949-959. Quardokus‚ E.‚ and Brun‚ Y. (2002) DNA replication initiation is required for mid-cell positioning of FtsZ rings in Caulobacter crescentus. Mol. Microbiol. 45‚ 605-616. Rauprich‚ O.‚ Matsushita‚ M.‚ Weijer‚ C.‚ Siegert‚ F.‚ Esipov‚ S. and Shapiro‚ J. (1996) Periodic phenomena in Proteus mirabilis swarm colony development. J. Bacteriol. 178‚ 6525-6538. Ruby‚ E.‚ and Rittenberg‚ S. (1983) Differentiation after premature release of intraperiplasmically growing Bdellovibrio bacteriovorous. J. Bacteriol. 154‚ 32-40. Ryding‚ N.J.‚ Kelemen‚ G.‚ Whatling‚ C.‚ Flärdh‚ K.‚ Buttner‚ M. and Chater‚ K. (1998) A developmentally regulated gene encoding a repressor-like protein is essential for sporulation in Streptomyces coelicolor A3(2). Mol. Microbiol. 29‚ 343-357. Sackett‚ M.‚ Kelly‚ A. and Brun‚ Y. (1998) Ordered expression of ftsQA and ftsZ during the Caulobacter crescentus cell cycle. Mol. Microbiol. 28‚ 421-434. Schwedock‚ J.‚ McCormick‚ J.‚ Angert‚ E.‚ Nodwell‚ J. and Losick‚ R. (1997) Assembly of the cell division protein FtsZ into ladder-like structures in the aerial hyphae of Streptomyces coelicolor. Mol. Microbiol. 25‚ 847-858. Sharp‚ M.‚ and Pogliano‚ K.. (2002a) Role of cell-specific SpoIIIE assembly in polarity of DNA transfer. Science. 295‚ 137-139. Sharp‚ M.‚ and Pogliano‚ K. (2002b) MinCD-dependent regulation of the polarity of SpoIIIE assembly and DNA transfer. EMBO J. 21‚ 6267-6274. Shimkets‚ L. and Brun‚ Y. (2000) Prokaryotic development: Strategies to enhance survival. In: Prokaryotic Development. Brun‚ Y.V.‚ and Shimkets‚ L.J. (eds). Washington‚ D.C: ASM Press‚ pp. 1-7. Stewart‚ B.‚ Enos-Berlage‚ J. and McCarter‚ L. (1997) The lonS gene regulates swarmer cell differentiation of Vibrio parahaemolyticus. J. Bacteriol. 179‚ 107-114. Terrana‚ B. and Newton‚ A. (1976) Requirement of Cell Division Step for Stalk Formation in Caulobacter crescentus. J. Bacteriol. 128‚ 456-462. Thomashow‚ M.‚ and Rittenberg‚ S. (1978) Penicillin-induced formation of osmotically stable spheroplasts in nongrowing Bdellovibrio bacteriovorus. J. Bacteriol. 133‚ 1484-1491. Thomashow‚ M.‚ and Cotter‚ T. (1992) Bdellovibrio host dependence: the search for signal molecules and genes that regulate the intraperiplasmic growth cycle. J. Bacteriol. 174‚ 5767-5771. Viollier‚ P.‚ Sternheim‚ N. and Shapiro‚ L. (2002) Identification of a localization factor for the polar positioning of bacterial structural and regulatory proteins. Proc. Natl. Acad. Sci. U.S.A. 99‚ 13831-13836. Ward‚ D. and Newton‚ A. (1997) Requirement of topoisomerase IV parC and parE genes for cell cycle progression and developmental regulation in Caulobacter crescentus. Mol. Microbiol. 26‚ 897-910.

Cell division and differentiation

131

Wei‚ B.‚ Brun-Zinkernagel‚ A.‚ Simecka‚ J.‚ Pruss‚ B.‚ Babitzke‚ P. and Romeo‚ T. (2001) Positive regulation of motility and flhDC expression by the RNA-binding protein CsrA of Escherichia coli. Mol. Microbiol. 40‚ 245-256. Weiner‚ R. and Blackman‚ M. (1973) Inhibition of deoxyribonucleic acid synthesis and bud formation by nalidixic acid in Hyphomicrobium neptunium. J. Bacteriol. 116‚ 1398-1404. Wortinger‚ M.‚ Sackett‚ M. and Brun‚ Y. (2000) CtrA mediates a DNA replication checkpoint that prevents cell division in Caulobacter crescentus. EMBO J. 19‚ 4503-4512. Wu‚ L.‚ and Errington‚ J. (1994) Bacillus subtilis SpoIIIE protein required for DNA segregation during asymmetric cell division. Science. 264‚ 572-575. Wu‚ L.‚ Feucht‚ A. and Errington‚ J. (1998a) Prespore-specific gene expression in Bacillus subtilis is driven by sequestration of SpoIIE phosphatase to the prespore side of the asymmetric septum. Genes Dev. 12‚ 1371-1380. Wu‚ L. and Errington‚ J. (1998b) Use of asymmetric cell division and spoIIIE mutants to probe chromosome orientation and organization in Bacillus subtilis. Mol. Microbiol. 27‚ 777-786. Zerfas‚ P.‚ Kessel‚ M.‚ Quintero‚ E. and Weiner‚ R. (1997) Fine structure evidence for cell membrane partitioning of the nucleoid and cytoplasm during bud formation in Hyphomonas species. J. Bacteriol. 179‚ 148-156. Zupancic‚ M.‚ Tran‚ H. and Hofmeister‚ A. (2001) Chromosomal organization governs the timing of cell type-specific gene expression required for spore formation in Bacillus subtilis. Mol. Microbiol. 39‚ 1471-1481.

This page intentionally left blank

Chapter 7 FtsZ folding‚ assembly

self-association‚

activation

and

Comparison with tubulin and implications for bacterial cell division *

JOSÉ M. ANDREU ‚ MARÍA A. OLIVA and SONIA HUECAS Centro de Investigaciones Biológicas (CSIC) Ramiro de Maeztu 9 28040 Madrid Spain *corresponding author

1.

Introduction. FtsZ and tubulin

FtsZ is a central protein of the prokaryotic cell division machinery‚ which recruits other cell division proteins to the septal site; FtsZ is the product of the ftsZ (filament-forming temperature sensitive Z) gene (for reviews see Rothfield et al.‚ 1999; Addinall and Holland‚ 2002; Buddelmeijer and Beckwith‚ 2002; Lutkenhaus‚ 2002; Errington et al.‚ 2003). Before cell division‚ FtsZ concentrates into a Z-ring beneath the plasma membrane at the division site (Bi and Lutkenhaus‚ 1991; Ma et al.‚ 1996). Tubulin (for a review see Nogales‚ 2000) is the main constituent of microtubules‚ which‚ in addition to their cytoskeletal roles‚ form the mitotic spindle during eukaryotic cell division (Borisy and Taylor‚ 1967; Weisenberg‚ 1972). It has been recognized that FtsZ has very little sequence homology with tubulin‚ but that both proteins share several nucleotide binding motifs and GTPase activity (Raychaudhuri and Park‚ 1992; de Boer et al.‚ 1992; Mukherjee et al.‚ 1993)‚ the ability to polymerize into filaments (Mukherjee and Lutkenhaus‚ 1994; Erickson et al.‚ 1996) and homologous secondary structures (de Pereda et al.‚ 1996). When the X-ray crystal structure of FtsZ from the archaeon Methanococcus jannaschii (Löwe and Amos‚ 1998) and

Molecules in Time and Space‚ Edited by Vicente et al. Kluwer Academic/Plenum Publishers‚ New York 2004

133

134

Andreu‚ Oliva and Huecas

the electron-crystallographic structure of from brain (Nogales et al.‚ 1998a) were independently determined it was found that both proteins share a common fold in two domains and form a distinct family of GTPases with highly conserved nucleotide binding sites (Nogales et al.‚ 1998b). The nucleotide sites participate in the contacts employed to form similar protofilaments (Löwe and Amos‚ 1999)‚ which associate to perform different functions. Thus the microtubules of the mitotic spindle are long cylindrical polymers‚ roughly perpendicular to the cell division plane‚ which push and pull chromosomes for segregation‚ whereas the Z-ring marks the cell division plane for prokaryotic cytokinesis. The prokaryotic origin of the eukaryotic cytoskeleton has been further evidenced by the structural homology of the bacterial cell shape protein MreB (Jones et al.‚ 2001; van den Ent et al.‚ 2001) with eukaryotic actin (Straub‚ 1942; Kabsch et al.‚ 1990). Other prokaryotic members of actin protein family of ATP-binding proteins are the plasmid segregation protein ParM (van den Ent et al.‚ 2002) and FtsA‚ which is believed to anchor FtsZ to the membrane and recruit downstream cell division proteins (van den Ent and Lowe‚ 2000). Whereas the assembly‚ structure and functions of actin microfilaments and of microtubules are relatively well understood‚ those of the physiological polymers of their prokaryotic homologous proteins are mostly unknown‚ including the structure and functional mechanism of the Z-ring. This chapter addresses key biochemical issues which we have found more relevant in the pathway going from the FtsZ polypeptide chains to the FtsZ polymer‚ including its folding‚ association‚ activation‚ the energetics and dynamics of FtsZ assembly and its polymer structure and regulation; without intending a comprehensive review of the literature‚ a better insight into several of these problems is gained by appropriate comparisons with the tubulin system.

2.

FtsZ polypeptide folding

Folding of actin‚ tubulin and many newly synthesized proteins in the cytosol is assisted by molecular chaperones‚ including the cytoplasmic chaperonin CCT (Hartl and Hayer-Hartl‚ 2002). FtsZ was not identified as an in vivo substrate of the bacterial chaperonin GroEL‚ although the less abundant cell division protein MinD was (Houry et al.‚ 1999). FtsZ did not bind to GroEL or CCT in vitro (Dobrzynski et al.‚ 2000). HscA and DnaK have been implicated in the FtsZ ring formation through a chaperone-like activity (Uehara et al.‚ 2001). The ease of overproducing soluble FtsZ and MreB in bacteria suggested that these proteins might be able to fold spontaneously. It has been found that the unfolding by guanidinium hydrochoride of FtsZ from M. jannaschii and from E. coli is reversible; their

FtsZ folding‚ self-association‚ activation and assembly

135

isolated polypeptide chains are capable of spontaneous refolding followed by GTP-dependent self-assembly‚ whereas tubulin monomers misfold in solution. Both FtsZ from E. coli and tubulin are very sensitive to relatively low denaturant concentrations‚ and release their nucleotide in a first stage of a multi-step unfolding‚ suggesting a nucleotide-dependent structural stabilization of their native state (Andreu et al.‚ 2002). This first stage leads to an unfolding intermediate in the denaturation of FtsZ from E. coli (Santra and Panda‚ 2003). The nature of this unfolding intermediate is not known‚ and it might be worth to investigate whether it corresponds to the possible separation of the two structural domains of FtsZ‚ which might fold independently (see a ribbon diagram of FtsZ in Figure 1).

Figure 1. Ribbon diagram of the structure of FtsZ from M. jannaschii (Protein Data Bank code 1FTSZ). The position of loops T3 and T7‚ the GDP nucleotide and the side chain of tryptophan 319 are indicated.

The structure based alignment of the FtsZ and tubulin sequences (Nogales et al.‚ 1998b) shows two types of differences. One of them consists in the specialized C-terminal domain that tubulin uses to bind microtubule associated proteins and motor proteins at the microtubule surface (Nogales‚ 2000). This domain has no counterpart in FtsZ; instead‚ the C-terminal peptide of FtsZ binds to FtsA (Yan et al.‚ 2000) and to ZipA (Mosyak et al.‚ 2000). The others are the extensive insertions which locate to tubulin loops between the secondary structure elements of the FtsZ/tubulin common core (Nogales et al.‚ 1998b). Fig. 2 shows these regions marked onto the structure

136

Andreu‚ Oliva and Huecas

of The tubulin loop insertions are excluded from the nucleotide binding site and the top and bottom contact interfaces between tubulin monomers‚ and they map to more peripheral positions at the sides and back of tubulin monomers. Analysis of the loop insertions of tubulin has shown that they include the contacts made in tubulin-CCT model complexes‚ with the exception of loop T7 (Llorca et al.‚ 2001) and the lateral contacts between protofilaments in model microtubules (Andreu et al.‚ 2002). We have therefore proposed that the loop insertions of tubulin and its CCTassisted folding co-evolved with the lateral association interfaces responsible for two-dimensional polymerization into microtubules (Andreu et al.‚ 2002). Thus FtsZ folding appears simpler than that of tubulin‚ and this is related to its simpler assembly‚ which will be discussed later. An interesting analogy is that the subunit interactions in actin filaments involve sequence insertions which are not present in its bacterial homologue MreB‚ and these zones can make different contacts in F-actin polymorphs (Galkin et al.‚ 2002).

Figure 2. Ribbon diagrams of beta-tubulin (from the tubulin dimer Protein Data Bank code 1JFF) in which the residues that are not present in FtsZ have been coloured dark grey. These are the insertions in the FtsZ/tubulin common core and the C-terminal twohelix domain of tubulin. A‚ view from the plus end of a protofilament with GDP in the center and a taxol molecule bound. B‚ tangential view from the side of a protofilament‚ with GTP and magnesium bound to the alpha-subunit shown at the bottom. C‚ back view corresponding to the inside of a microtubule and to Figure 1 (modified from Andreu et al.‚ 2002).

FtsZ folding‚ self-association‚ activation and assembly

3.

137

FtsZ self-association

Bacterial FtsZ from E. coli is a monomer at low and protein concentrations and high ionic strength. FtsZ monomers undergo a oligomerization that is best described as an indefinite linear selfassociation with a slight decrease of the affinity of protomer addition with increasing oligomer size. This quasi-isodesmic self-association is weakly modulated by GDP or GTP binding to FtsZ. However, under closely related solution conditions FtsZ undergoes and GTP-induced assembly into long polymers. It has been proposed that the E. coli FtsZ oligomers and polymers share a similar interaction between monomers (Rivas et al., 2000) (see Fig. 3). FtsZ from the hypertermophile M. jannaschii undergoes an oligomerization insensitive to and nucleotide, and polymerizes upon temperature increase with GTP and (Huecas and Andreu, 2003). In order to avoid frequent confusion, we recall that protein self-association refers to the formation of oligomers and quaternary structures, whereas protein assembly usually means the formation of large polymers and biological superstructures (Jaenicke and Lilie, 2000). Similarly, tubulin are known to undergo an end-wise self-association leading to double ring formation (Frigon and Timasheff, 1975), and under closely related conditions with and GTP, tubulin undergoes cooperative assembly into microtubules (Lee and Timasheff, 1975). Tubulin double rings correspond to curved protofilament segments of disassembled microtubules (Melki et al., 1989; Díaz et al., 1994). In the case of FtsZ, the circular closure of linear oligomers into rings has not been observed, except with cationic phospholipid or DEAE dextran additives (Erickson et al., 1996; Lu et al., 2000).

Figure 3. A scheme illustrating the concepts of linear self-association and cooperative polymerization of FtsZ monomers (modified from Rivas et al.‚ 2000).

138

Andreu, Oliva and Huecas

The FtsZ self-association observed in dilute buffer is moderately enhanced by macromolecular crowding (Rivas et al., 2001) but it is significantly weakened under ionic conditions resembling the physiological osmolytes and pH of the E. coli cytosol, suggesting that the non-assembled cellular FtsZ may have an average degree of association close to dimers, instead of larger oligomers (González et al., 2003).

4.

The functional assembly cycle

Purified FtsZ assembles with GTP and forming dynamic polymers, which hydrolyze GTP and disassemble upon nucleotide consumption (Mukherjee and Lutkenhaus 1998, 1999; Lu et al., 1998; Rivas et al., 2000), in addition induces polymer bundling and retards GTP hydrolysis (Yu and Margolin, 1997). Thus a FtsZ monomer experiences an assembly cycle which, similarly to the tubulin assembly cycle, includes polymerization, nucleotide hydrolysis, depolymerization and reloading with nucleotide triphosphate (see Fig. 4). Like in other GTPases the GTP bound protein is functionally active and the GDP liganded form is normally inactive. Unlike small GTPases and G-proteins, in FtsZ and tubulin the target protein and the GTPase activating protein are more identical molecules, which provide both the appropriate interfaces for polymerization and GTPase activating residues (Fig. 4, GAR). The interplay of polar polymerization, nucleotide hydrolysis and exchange confers characteristic dynamic properties to the polymer. The nucleotide binds at the axial association interface with the next monomer in the polymer. In the case of tubulin, the hydrolyzable nucleotide becomes buried by assembly and is only exchanged by the unassembled protein. Tubulin hydrolyses GTP shortly after assembly, generating microtubules that contain GDP-tubulin and are stabilized against disassembly by small terminal caps of GTP-bound subunits. As a result of these properties, microtubules are molecular machines that experience treadmilling (head to tail polymerization) (Maly and Borisy, 2002) and dynamic instability (phases of endwise growth and shrinkage) (Howard and Hyman, 2003). Whether the nucleotide binding site is really buried or the nucleotide is exchangeable in FtsZ polymers (Desai and Mitchison, 1998), what is the intrinsic nucleotide hydrolysis rate and the nucleotide content of FtsZ polymers are key isssues for their dynamics which are still subject to some controversy (Mingorance et al., 2001; Scheffers and Driessen, 2002). Microtubule dynamics is a well-known target of assembly-inducing or inhibitory antitumour drugs, such as paclitaxel and vinblastine (Jordan and Wilson, 1998). FtsZ can be regarded as a potential target for small molecules

FtsZ folding, self-association, activation and assembly

139

that by modulating its assembly may lead to new antibiotics selectively inhibiting bacterial cell division.

Figure 4. Scheme of the assembly cycle of FtsZ or tubulin. Soluble protomers bound to GTP (labelled the filled circle represents GTP) associate into GTP-bound polymer following which, with the help of a co-catalytic GTPase activating residue (GAR) from the next incoming protomers the nucleotide is hydrolyzed, forming a GDP-liganded polymer Following disassembly, the inactive GDP-liganded protomers exchange-in new GTP, regenerating the active species There has been a report (Mingorance et al., 2001) proposing that reloading with GTP also takes place within the FtsZ polymer.

5.

GTPase activity

The GTPase activity of E. coli FtsZ is known to be dependent on protein concentration, compatible with the notion that it is activated upon polymerization (Wang and Lutkenhaus, 1993; Romberg et al., 2001), and also by oligomerization (Sossong et al., 1999). The GTPase activity of FtsZ from M. jannaschii parallels its polymerization above a critical protein

140

Andreu , Oliva and Huecas

concentration, indicating that GTPase activation is induced by the contacts between the FtsZ molecules in the polymer (Oliva et al. 2003). Structural comparisons between FtsZ and tubulin dimers have suggested that the GTPase-activating residues of FtsZ and tubulin are provided by the bottom association interface of the next subunit, specifically by loop T7, which contains the acidic residue Asp 238 in FtsZ from M. jannaschii (Nogales et al., 1998b). This residue corresponds to Asp 212 in E. coli FtsZ, whose mutation to Ala inhibits the GTPase activity (Dai et al., 1994). The equivalent residues in tubulin come in proximity with the of the next tubulin subunit. They are Glu 254 of which is substituted by Lys 254 in does not activate the hydrolysis of the GTP permanently bound to in the same heterodimer. Several other mutations in loop T7 of FtsZ, also termed synergy loop (Erickson, 1998), inhibit the GTPase activity (Lu et al., 2001; Scheffers et al., 2001, 2002) providing further evidence that the FtsZ active site is formed by the association of monomers. Mutation of the single tryptophan residue Trp 319 of M. jannaschii FtsZ (corresponding to Ile 294 in E coli FtsZ), which is located at the bottom association interface of FtsZ, less than 10 Å away from loop T7 (see Fig. 1), into Tyr abolishes the GTPase activity and results in non-hydrolyzing FtsZ polymers (Oliva et al., 2003).

6.

The nucleotide activation/disassembly switch

Once the nucleotide is hydrolyzed, which are the structural changes that induce FtsZ polymer disassembly? FtsZ has been reported to assemble into straight filaments with DEAE dextran and GTP, but into curved filaments with the same additive and GDP (Lu et al. 2000), resembling the so-called straight (active, microtubule forming) and curved conformation (inactive, ring forming) of tubulin. An ultimate answer will require determining the atomic structures of the GTP-bound FtsZ polymer and of its GDP-bound disassembly product. A related approach consists in asking which is the nucleotide switch that makes monomers of GTP-bound FtsZ active for assembly and GDP-bound FtsZ inactive. The crystal structure of FtsZ from M. jannaschii mainly contains GDP (Löwe and Amos, 1998; confirmed by Diaz et al., 2001), and a different GTP-bound crystal is lacking for comparison. On the other hand, the available structures of tubulin are that of the active form locked into a two-dimensional zinc-induced polymer (Nogales et al., 1998a; Löwe et al., 2001), and a lower resolution structure of GDP-tubulin in complex with stathmin (Gigant et al., 2000); therefore the same question can be asked for tubulin. A molecular dynamics study,

FtsZ folding, self-association, activation and assembly

141

starting from the crystal structure of M. jannaschii FtsZ, has simulated the effect of bound nucleotide on the conformation of FtsZ (Díaz et al., 2001). Among other changes, it has been predicted that the of GTP induces a conformational change in loop T3 (the Gly88 to Gly99 segment) that is pulled into a more compact conformation than with GDP (possibly by hydrogen-bonding the N-H of Gly 99) (see Fig. 5). The existence of a nucleotide-induced structural change in loop T3 has been confirmed by mutating Thr 92 into Trp (T92W-W319Y FtsZ), and observing the large difference in tryptophan fluorescence induced by GTP versus GDP. Loop T3 is in a position structurally equivalent to switch II of ras-p21, at the top contact interface between FtsZ monomers. The structural change in the GDP-bound conformation may make loop T3 push against helix H8 of the next monomer, thus potentially explaining filament curvature upon GTP hydrolysis, and a related mechanism may operate for tubulin.

Figure 5. Conformation of loop T3 GTP-bound FtsZ (alpha-carbon tracing and nucleotide in light grey) compared with the conformation in GDP-bound FtsZ (in black), from a molecular dynamics simulation (Díaz et al., 2001).

7.

Energetics and dynamics of FtsZ assembly

Protein polymerization is typically cooperative, which requires the formation of multiple-stranded polymers (Oosawa and Asakura, 1975). The GTP-induced assembly of bacterial FtsZ was reported cooperative (Mukherjee and Lutkenhaus, 1998; Lu et al., 1998; White et al., 2000), or on the contrary to consist of the non-cooperative isodesmic association of monomers into single stranded filaments observed by transmission scanning electron microscopy (Romberg et al., 2001); more recently, the same group has reported apparently cooperative assembly of FtsZ (Caplan and Erickson, 2003). However, single stranded filaments have only one type of bond between monomers and cannot conceivably assemble in a cooperative

142

Andreu, Oliva and Huecas

fashion, which requires a minimum of two types of bond between the monomers in the polymer (Oosawa and Asakura, 1975). In fact, a majority of the polymers formed by bacterial FtsZ from E. coli have a doublestranded structure in negative stain electron microscopy (Oliva et al., 2003) and assembly is cooperative (González et al., 2003). FtsZ from M. jannaschii has been shown to cooperatively assemble following a condensation polymerization mechanism (see Fig. 6); the use of buffer results in suppression of the hydrolysis of the GTP or GMPCPP nucleotide, allowing rigorous equilibrium measurements of the thermodynamics of assembly elongation (the addition of one monomer to the polymer), which proceeds with a negative apparent heat capacity change and positive enthalpy and entropy changes, compatible with the hydrophobic character of the axial model association interfaces between FtsZ monomers. Polymer elongation entails the participation of one medium affinity ion, probably coordinated with the bound nucleotide. Rather than completely depolymerizing by GTP hydrolysis, GDP-bound FtsZ from M. jannaschii increases its critical concentration and forms helically coiled polymers different from the straight polymers formed before GTP hydrolysis. These GDP-FtsZ polymers may be the FtsZ equivalent to the GDP-tubulin rings (Huecas and Andreu, 2003).

Figure 6. Cooperative assembly of FtsZ from M. jannaschii, with magnesium and GMPCPP (a slowly hydrolyzable GTP analogue) in potassium-less phosphate buffer at pH 6.5 at 55 °C. The FtsZ polymers were sedimented. The void and filled circles are the FtsZ concentrations in the supernatant and pellet respectively. No polymer is formed below a critical concentration ( under these conditions) above which all FtsZ excess goes into the polymer (Huecas and Andreu, 2003).

FtsZ folding, self-association, activation and assembly

143

The Z rings of growing E. coli cells are very dynamic. A fusion of FtsZ with green fluorescent protein (FtsZ-GFP) localizes to the new Z ring (Ma et al., 1996) in a few minutes after the previous division and stays until quickly constricting to divide again (Sun and Margolin, 1998). In the temperature sensitive mutant ftsZ84 (Gly105Ser) the Z rings rapidly disassemble upon shifting to 42 °C (Addinall et al., 1997). The exchange of FtsZ subunits in the Z-ring has been examined by fluorescence recovery after photobleaching (FRAP) of FtsZ-GFP, with the finding that the half time for remodelling the Z-GFP-ring is of only tens of seconds (Strieker et al., 2002), which is comparable to the in vitro rate of GTP hydrolysis by FtsZ in physiological buffer (Lu et al., 1998). The mutant FtsZ84 has one tenth of the wild type GTPase and its Z-GFP-ring has a similarly reduced turnover, suggesting that the assembly dynamics is determined by the rate of GTP hydrolysis (Stricker et al., 2002). On the other hand, macromolecular crowding resembling the volume exclusion conditions of the bacterial cytosol, significantly retards the GTP hydrolysis and polymer disassembly (González et al., 2003), raising the question of how the Z-ring dynamics is made so fast in vivo. The kinetic mechanisms of FtsZ in vitro polymer assembly/disassembly remain to be studied. It has not been shown whether the FtsZ polymers exchange their subunits by end-wise treadmilling or dynamic instability, or by fragmentation and reannealing, or whether they rapidly cycle, transiently assembling to fully disassemble upon nucleotide hydrolysis, and reassemble upon nucleotide exchange. Distinguishing between these possibilities may require solution kinetics or single polymer biophysical approaches.

8.

FtsZ polymer structure and the Z-ring

Insight into the possible structure and mechanism of the Z-ring can be obtained by examining the structures of in vitro assembled FtsZ polymers. This is complicated by the fact that assembly of purified FtsZ is typically polymorphic. A number of studies have shown the formation of different filamentous polymers from FtsZ with GTP and divalent cations (Yu and Margolin, 1997; Mukherjee and Lutkenhaus, 1998; Rivas et al., 2000; Lu et al., 2000; White et al., 2000; Mingorance et al., 2001; Romberg et al., 2001; Andreu et al., 2002). FtsZ polymers have the characteristic tubulin ~40Å axial spacing between consecutive monomers (Erickson et al., 1996). In the presence of high and GTP concentrations at pH 6, a small proportion of M. jannaschii FtsZ with a six histidines tag formed crystalline sheets made of symmetric protofilament pairs associated in an antiparallel arrangement. Manual fitting of the structure of GDP-bound FtsZ into a 3D electron microscopy map of the sheet polymers lead to the tubulin-like model of FtsZ

144

Andreu, Oliva and Huecas

protofilament (Löwe and Amos, 1999). The GTPase inactivated mutant FtsZ-W319Y quantitatively assembles under biochemically-regulated conditions into similar sheets, whose crystallization has been shown to be favoured by the six-histidine tag. On the other hand wild type M.jannaschii FtsZ without a histidine tag assembles into associated and isolated filaments, and most isolated filaments are made of two parallel protofilaments with a 43 Å longitudinal spacing between monomers, as shown by image processing of electron micrographs This structure is also observed in FtsZ from E. coli (see Fig. 7; Oliva et al. 2003).

Figure 7. Structure of FtsZ polymers. Panel A shows a gallery of individual filaments of FtsZ from E. coli in negatively stained electron micrographs; the bar indicates 10 nm. Panel B is the average image obtained after processing 268 areas extracted from individual filaments, showing two protofilaments with a 43 Å axial spacing between monomers Panel C is a scheme of the proposed symmetric association of two FtsZ protofilaments, and panel D depicts the mode of association of tubulin protofilaments to form a microtubule (from Oliva et al., 2003).

FtsZ folding, self-association, activation and assembly

145

We have proposed that the basic FtsZ polymer structure is a pair of symmetric tubulin-like protofilaments. The lateral interactions between two symmetric protofilaments in archaeal FtsZ are clearly different from those between microtubule protofilaments (Fig. 7; Oliva et al. 2003). In this assembly model the two FtsZ protofilaments cannot use their common lateral bonding interface to further associate, whereas the same lateral interaction between microtubule protofilaments are known to propagate until reaching cylindrical closure. The common fold of FtsZ and tubulin is thus employed to construct similar protofilaments that associate to perform the different functions of cytokinesis and chromosome segregation. It may be speculated that these functions have evolved from a common ancestor, by means of two different modes of protofilament association, a simple FtsZ protofilament pair or the more robust assembly of tubulin into microtubules. It is possible that the Z-ring may be made of a few double stranded filaments, which have resisted observation by electron microscopy. From current estimates of the number of FtsZ molecules per bacterial cell, it may be calculated that if only a 30% of FtsZ was incorporated into the Z-ring, as for GFP-FtsZ (Stricker et al., 2002), there would be about three and a half continuous turns of double stranded FtsZ filament in a spiral Z-ring in E. coli, and only one double stranded filament in a B. subtilis dividing cell (discussed by Oliva et al., 2003). Alternatively, the Z-ring may be a more consistent scaffold, such as a ribbon made of ten or more FtsZ filaments, whose in vitro formation is favoured by macromolecular crowding (Gonzalez et al., 2003) and have a width closer to that of the zone apparently labelled in immunoelectron microscopy (Bi and Lutkenhaus, 1991); however, this would require more FtsZ molecules to form the Z-ring. It has been hypothesized that upon GTP hydrolysis, a change from straight to curved FtsZ protofilament, flexibly attached to the membrane through ZipA, should accommodate the decrease in diameter of the bacterium during constriction and may even provide the force powering constriction (Erickson, 2001). However, a bacterial septation circle is much larger than an FtsZ mini-ring; the curvature angle between FtsZ monomers in mini-rings is 22°, whereas simple calculations show that for a septal constriction with a diameter of one tenth of micrometre, the corresponding angle between FtsZ monomers would be 2.3°, which is one order of magnitude smaller. However, a spiral curvature of the FtsZ polymers observed upon nucleotide hydrolysis (Huecas and Andreu, 2003) would result in an effective shortening which might help in septal constriction or accommodate the decreasing size of the FtsZ ring. There is no actual evidence that FtsZ polymers generate force upon curvature or depolymerization. It is not known to which extent the FtsZ polymers in the Z

146

Andreu, Oliva and Huecas

ring are force-generating machines or just signals for other proteins to be recruited to the septal site.

9.

FtsZ interactions and the regulation of the Z-ring

Several proteins are known to regulate the Z-ring formation and stability, including some of the downstream cell division proteins that are recruited by FtsZ. This is in contrast with the many proteins that interact with tubulin, including the microtubule stabilizing and destabilizing factors that regulate microtubule dynamics (Heald and Nogales, 2002), and the motor proteins, which have not been found in prokaryotes. The mitotic spindle and the Zring are both self-made machines, but so far, no structurally homologous interactions have been evidenced for FtsZ polymers and microtubules, except in protofilament formation. FtsZ interacts with assembly regulatory proteins in different fashions, resulting in FtsZ assembly induction or inhibition, which are summarized below (see Fig. 8) The C-terminal peptide of FtsZ contains a conserved motif (consensus sequence: LDXPXFOR/K, where X is any residue and O a hydrophobic residue) that is bound by the membrane-associating essential division protein FtsA; this motif is absent in archaea, where FtsA homologues are not present (Yan et al., 2000). It might be possible that the C-terminus of FtsZ binds to the groove between subdomains 1A and 1C of FtsA (van den Ent and Löwe, 2000). The integral membrane protein ZipA, which is believed to tether FtsZ to the cytoplasmic membrane (Hale and de Boer, 1997), contains in its Cterminal domain a shallow hydrophobic cavity that binds a 17 residue FtsZ C-terminal fragment including the conserved motif (Mosyak et al., 2000). The FtsZ binding domain of ZipA induces FtsZ polymer bundling (Hale et al., 2000). Either ZipA or FtsA are capable of supporting the formation and stabilization of Z-rings, and they are both required for recruitment of further downstream cell division proteins (Pichoff and Lutkenhaus, 2002). A gainof-function mutant of FtsA can bypass the requirement for ZipA in E. coli cell division (Geissler at al., 2003). The C-terminal end of FtsZ from M. tuberculosis, in which no FtsA or ZipA have been identified, interacts in vitro with the C-terminal tail of FtsW; both FtsZ and FtsW from M. tuberculosis carry complementarity charged C-terminal extensions not present in the E. coli proteins (Datta et al., 2002). ZapA is a conserved bacterial cell division protein, discovered in B. subtilis, which induces the assembly of FtsZ polymer bundles and promotes Z-ring assembly (GueirosFilho and Losick, 2002). SulA, an inhibitor of FtsZ assembly, GTPase and Z-ring formation (Mukherjee et al., 1998), binds to the C-terminal domain of FtsZ at its T-7

FtsZ folding, self-association, activation and assembly

147

loop side, thus capping one of the FtsZ protofilament ends; the crystal structure of the FtsZ:SulA complex from P. aeruginosa shows a SulA dimer binding two FtsZ molecules in exactly opposite directions (Cordell et al., 2003). On the other hand, MinC was reported to inhibit FtsZ polymerization without apparently inhibiting its GTPase (Hu et al., 1999). MinC is the component of the oscillating MinCD-MinE system that inhibits assembly of FtsZ except at the midcell (Lutkenhaus, 2002). MinC is a dimer hold together by the C-terminal domains, attached through flexible linkers to the N-terminal domains (Cordell et al., 2001). The N-terminal domain binds to FtsZ whereas the C-terminal domain binds to MinD (Hu and Lutkenhaus, 2000), and MinD binds to the membrane. MinCD prevented the addition of FtsA to the FtsZ ring (Justice et al., 2000). The part of FtsZ that binds to MinC and the mechanism of FtsZ assembly inhibition by MinC remain to be determined. Finally, ErzA is a negative regulator of FtsZ ring formation that localizes throughout the plasma membrane of B. subtilis, and is thought to destabilize FtsZ polymers (Levin et al., 1999).

Figure 8. A scheme of the interactions of FtsZ filaments with assembly inducers FtsA, ZapA and ZipA (on the left) and inhibitors ErzA, MinC and SulA (on the right side).

148

Andreu, Oliva and Huecas

Acknowledgements Work in the authors laboratory was supported by grants from MCyT BIO200203665 and CAM Programa de Grupos Estratégicos.

References Addinall, S.G., Cao, C., and Lutkenhaus, J. (1997). Temperature shift experiments with an ftsZ84(Ts) strain reveal rapid dynamics of FtsZ localization and indicate that the Z ring is required throughout septation and cannot reoccupy division sites once constriction has initiated. J. Bacteriol. 179, 4277-4284. Addinall, S.G., and Holland, B. (2002). The tubulin ancestor, FtsZ, draughtsman, designer and driving force for bacterial cytokinesis. J. Mol. Biol. 318, 219-236. Andreu, J.M., Oliva, M.A., and Monasterio, O. (2002). Reversible unfolding of FtsZ cell division proteins from archaea and bacteria. Comparison with eukaryotic tubulin folding and assembly. J. Biol. Chem. 277, 43262-43270. Bi, E.F., and Lutkenhaus, J. (1991). FtsZ ring structure associated with division in Escherichia coli. Nature 354, 161-164. Borisy, G.G. and Taylor E.W. (1967). The mechanism of action of colchicine. Binding of colchicine-H3 to cellular protein. Colchicine binding to sea urchin eggs and the mitotic apparatus. J. Cell Biol. 34, 525-548. Buddelmeijer, N., and Beckwith, J. (2002). Assembly of cell division proteins at the E. coli cell center. Curr. Opin. Microbiol. 5, 553-557. Caplan, M.R., and Erickson, H.P. (2003). Apparent cooperative assembly of the bacterial cell division protein FtsZ demonstrated by isothermal titration calorimetry. J. Biol. Chem. 278, 13784-13788. Cordell, S.C., Anderson, R.E., and Löwe, J. (2001). Crystal structure of the bacterial cell division inhibitor MinC. EMBO J. 20, 2454-2461. Cordell S.C., Robinson, E.J.H. and Löwe, J. (2003). Crystal structure of the SOS cell division inhibitor SulA and in complex with FtsZ. Proc. Natl. Acad. Sci. USA 100, 7889-7804. Dai, K., Mukherjee, A., Xu, Y., and Lutkenhaus, J. (1994), Mutations in ftsZ that confer resistance to SulA affect the interaction of FtsZ with GTP. J. Bacteriol. 176, 130-136. Datta, P., Dasgupta, A., Bhakta, S., and Basu, J. (2002). Interaction between FtsZ and FtsW of Mycobacterium tuberculosis. J. Biol. Chem. 277, 24983-24987. de Boer, P., Crossley, R., and Rothfield, L. (1992). The essential bacterial cell-division protein FtsZ is a GTPase. Nature 359, 254-256. de Pereda, J.M., Leynadier, D., Evangelio, J.A., Chacon, P., and Andreu, J.M. (1996). Tubulin secondary structure analysis, limited proteolysis sites, and homology to FtsZ. Biochemistry 35, 14203-14215. Desai, A., and Mitchison, T.J. (1998). Tubulin and FtsZ structures, functional and therapeutic implications. Bioessays 20, 523-527. Diaz, J.F., Pantos, E., Bordas, J., and Andreu, J.M. (1994). Solution structure of GDP-tubulin double rings to 3 nm resolution and comparison with microtubules. J. Mol. Biol. 238, 214-225. Diaz, J.F., Kralicek, A., Mingorance, J., Palacios, J.M., Vicente, M., and Andreu, J.M. (2001). Activation of cell division protein FtsZ. Control of switch loop T3 conformation by the nucleotide gamma-phosphate. J. Biol. Chem. 276, 17307-17315.

FtsZ folding, self-association, activation and assembly

149

Dobrzynski, J.K., Sternlicht, M.L., Peng, I., Farr, G.W., and Sternlicht, H. (2000). Evidence that beta-tubulin induces a conformation change in the cytosolic chaperonin which stabilizes binding, implications for the mechanism of action. Biochemistry 39, 39884003. Erickson, H.P., Taylor, D.W., Taylor, K.A., and Bramhill, D. (1996). Bacterial cell division protein FtsZ assembles into protofilament sheets and minirings, structural homologs of tubulin polymers. Proc. Natl. Acad. Sci. U S A. 93, 519-523. Erickson, H.P. (1998). Atomic structures of tubulin and FtsZ. Trends Cell Biol. 8, 133-137. Erickson, H.P. (2001). The FtsZ protofilament and attachment of ZipA--structural constraints on the FtsZ power stroke. Curr. Opin. Cell Biol. 13, 55-60. Errington, J., Daniel, R.A., and Scheffers, D.J. (2003). Cytokinesis in bacteria. Microbiol. Mol. Biol. Rev. 67, 52-65. Frigon, R.P., and Timasheff, S.N. (1975). Magnesium-induced self-association of calf brain tubulin. Biochemistry 14, 4559-4573. Galkin, V.E., VanLoock, M.S., Orlova, A., and Egelman, E.H. (2002). A new internal mode in F-actin helps explain the remarkable evolutionary conservation of actin's sequence and structure. Curr. Biol. 12, 570-575. Geissler, B., Elraheb D., and Margolin W. (2003). A gain-of-function mutation in ftsA bypasses the requirement for the essential cell division gene zipA in Escherichia coli. Proc. Natl. Acad. Sci. USA 100, 4197-4202. Gigant, B., Curmi, P.A., Martin-Barbey, C., Charbaut, E., Lachkar, S., Lebeau, L., Siavoshian, S., Sobel, A., and Knossow, M. (2000). The 4 Å X-ray structure of a tubulin:stathmin-like domain complex. Cell 102, 809-816. González, J.M., Jiménez, M., Vélez, M., Mingorance, J., Andreu, J.M., Vicente, M. and Rivas, G. (2003) Essential cell division protein FtsZ assembles into one-monomer thick ribbons under conditions resembling the crowded intracellular environment. J. Biol. Chem. 278, 37664-37671. Gueiros-Filho, F.J. and Losick, R. (2002). A widely conserved bacterial cell division protein that promotes assembly of the tubulin-like protein FtsZ. Genes Dev. 16, 2544-2556. Hale, C.A., and de Boer, P.A.J. (1997). Direct binding of FtsZ to ZipA, an essential component of the septal ring structure that mediates cell division in E. coli. Cell. 88, 175185. Hale, C.A., Rhee, A.C. and de Boer, P.A.J. (2000). ZipA-induced bundling of FtsZ polymers mediated by an interaction between C-terminal domains. J. Bacteriol. 182, 5153-5166. Hartl, F.U., and Hayer-Hartl, M. (2002). Molecular chaperones in the cytosol, from nascent chain to folded protein. Science 295, 1852-1858. Heald, R. and Nogales A. (2002). Microtubule dynamics. J. Cell Sci. 115, 3-4. Houry, W.A., Frishman, D., Eckerskorn, C., Lottspeich, F., and Hartl, F.U. (1999). Identification of in vivo substrates of the chaperonin GroEL. Nature 402, 147-154. Howard, J. and Hyman, A.A. (2003). Dynamics and mechanics of the microtubule plus end. Nature 422, 753-758. Hu, Z., Mukherjee, A., Pichoff, S., and Lutkenhaus, J. (1999). The MinC component of the division site selection system in Escherichia coli interacts with FtsZ to prevent polymerization. Proc Natl Acad Sci USA 96, 14819-14824. Hu, Z., and Lutkenhaus, J. (2000). Analysis of MinC reveals two independent domains involved in interaction with MinD and FtsZ. J Bacteriol. 182, 3965-3971. Huecas, S. And Andreu, J.M., 2003, Energetics of the cooperative assembly of cell division protein FtsZ and the nucleotide hydrolysis switch. J. Biol. Chem., 278, 46146-46154.

150

Andreu , Oliva and Huecas

Jaenicke, R., and Lilie, H. (2000). Folding and association of oligomeric and multimeric proteins. Adv. Protein Chem. 53, 329-401. Jones, L.J., Carballido-Lopez, R., and Errington, J. (2001). Control of cell shape in bacteria, helical, actin-like filaments in Bacillus subtilis. Cell 104, 913-922. Jordan, M.A. and Wilson, L. (1998). Microtubules and actin filaments, dynamic targets for cancer chemotherapy. Curr. Opin. Cell Biol. 10:123-130 Justice, S.S., Garcia-Lara, J., and Rothfield, L.I. (2000). Cell division inhibitors SulA and MinC/MinD block septum formation at different steps in the assembly of the Escherichia coli division machinery. Mol. Microbiol. 37, 410-423. Kabsch, W., Mannherz, H.G., Suck, D., Pai, E.F., and Holmes, K.C. (1990). Atomic structure of the actin:DNase I complex. Nature 347, 37-44. Lee, J.C., and Timasheff, S.N. (1975). The reconstitution of microtubules from purified calf brain tubulin. Biochemistry 14, 5183-5187. Levin, P.A., Kurtser, I.G., and Grossman, A.D. (1999). Identification and characterization of a negative regulator of FtsZ ring formation in Bacillus subtilis. Proc. Natl. Acad. Sci. USA 96, 9642-9647. Llorca, O., Martin-Benito, J., Gomez-Puertas, P., Ritco-Vonsovici, M., Willison, K.R., Carrascosa, J.L., and Valpuesta, J.M. (2001). Analysis of the interaction between the eukaryotic chaperonin CCT and its substrates actin and tubulin. J. Struct. Biol. 135, 20518. Löwe, J. and Amos, L.A. (1998). Crystal structure of the bacterial cell-division protein FtsZ. Nature 391, 203-206. Löwe, J., and Amos, L.A. (1999). Tubulin-like protofilaments in FtsZ sheets. EMBOJ. 18, 2364-2371. Löwe, J., Li, H., Downing, K.H., and Nogales, E. (2001). Refined structure of alpha betatubulin at 3.5 A resolution. J. Mol. Biol. 313, 1045-1057. Lu, C., Stricker, J., and Erickson, H.P. (1998). FtsZ from Escherichia coli, Azotobacter vinelandii, and Thermotoga maritima-quantitation, GTP hydrolysis, and assembly. Cell Motil. Cytoskeleton 40, 71-86. Lu, C., Reedy, M., and Erickson, H.P. (2000). Straight and curved conformations of FtsZ are regulated by GTP hydrolysis. J. Bacteriol. 182, 164-170. Lu, C., Stricker, J., and Erickson, H.P. (2001). Site-specific mutations of FtsZ--effects on GTPase and in vitro assembly. BMC Microbiol. 1, 7. Lutkenhaus, J. (2002). Dynamic proteins in bacteria. Curr. Opin. Microbiol. 5, 548-552. Ma, X., Ehrhardt, D.W., and Margolin, W. (1996). Colocalization of cell division proteins FtsZ and FtsA to cytoskeletal structures in living Escherichia coli cells by using green fluorescent protein. Proc. Natl. Acad. Sci. U S A 93, 12998-13003. Maly, I.V. and Borisy, G.G. (2002). Self-organization of treadmilling microtubules into a polar array. Trends Cell Biol. 12, 462-465. Melki, R., Carlier, M.F., Pantaloni, D., and Timasheff, S.N. (1989). Cold depolymerization of microtubules to double rings, geometric stabilization of assemblies. Biochemistry 28, 9143-9152. Mingorance, J., Rueda, S., Gómez-Puertas, P., Valencia, A., and Vicente, M. (2001). Escherichia coli FtsZ polymers contain mostly GTP and have a high nucleotide turnover. Mol. Microbiol. 41, 83-91. Mosyak, L., Zhang, Y., Glosfeld, E., Haneys, S., Stahl, M., Seehra, J., and Somers, W.S. (2000). The bacterial cell-division protein ZipA and its interaction with an FtsZ fragment revealed by X-ray crystallography. EMBO J. 13, 3179-3191.

FtsZ folding, self-association, activation and assembly

151

Mukherjee, A., Dai, K., and Lutkenhaus, J. (1993). Escherichia coli cell division protein FtsZ is a guanine nucleotide binding protein. Proc. Natl. Acad. Sci. U S A 90, 1053-1057. Mukherjee, A., and Lutkenhaus, J. (1994). Guanine nucleotide-dependent assembly of FtsZ into filaments. J. Bacteriol. 176, 2754-2758. Mukherjee, A., Cao, C. and and Lutkenhaus, J. (1998). Inhibition of FtsZ polymerization by SulA, an inhibitor of septation in Escherichia coli. Proc. Natl. Acad. Sci. USA 95, 28852890. Mukherjee, A., and Lutkenhaus, J. (1998). Dynamic assembly of FtsZ regulated by GTP hydrolysis. EMBO J. 17, 462-469. Mukherjee, A., and Lutkenhaus, J. (1999). Analysis of FtsZ assembly by light scattering and determination of the role of divalent metal cations. J. Bacteriol. 181, 823-832. Nogales, E., Wolf, S.G., and Downing, K.H. (1998a) Structure of the alpha beta tubulin dimer by electron crystallography. Nature 391, 199-203. Nogales, E., Downing, K.H., Amos, L.A., and Lowe, J. (1998b). Tubulin and FtsZ form a distinct family of GTPases. Nat. Struct. Biol. 5, 451-458. Nogales, E. (2000). Structural insights into microtubule function. Annu. Rev. Biochem. 69, 277-302. Oliva, M.A., Huecas, S., Palacios, J.M., Martín-Benito, J., Valpuesta, J.M. and Andreu, J.M. (2003). Assembly of archaeal cell division protein FtsZ and a GTPase-inactive mutant into double stranded filament. J. Biol. Chem. 278, 33562-33570. Oosawa, F., and Asakura, S. (1975). Thermodynamics of the polymerization of protein, Academic Press, London. Pichoff, S. and Lutkenhaus J. (2002). Unique and overlapping roles for ZipA and FtsA in septal ring assembly in Escherichia coli. EMBO J 21, 685-693. RayChaudhuri, D., and Park, J.T. (1992). Escherichia coli cell-division gene ftsZ encodes a novel GTP-binding protein. Nature 359, 251-254. Rivas, G., Lopez, A., Mingorance, J., Ferrandiz, M.J., Zorrilla, S., Minton, A.P., Vicente, M., and Andreu, J.M. (2000). Magnesium-induced linear self-association of the FtsZ bacterial cell division protein monomer. The primary steps for FtsZ assembly. J. Biol. Chem. 275, 11740-11749. Rivas, G., Fernandez, J.A., and Minton, A.P. (2001). Direct observation of the enhancement of noncooperative protein self-assembly by macromolecular crowding, indefinite linear self-association of bacterial cell division protein FtsZ. Proc. Natl. Acad. Sci. USA 98, 3150-3155. Romberg, L., Simon, M., and Erickson, H.P. (2001). Polymerization of Ftsz, a bacterial homolog of tubulin: is assembly cooperative? J. Biol. Chem. 276, 11743-11753. Rothfield, L., Justice, S., and Garcia-Lara, J. (1999). Bacterial cell division. Annu. Rev. Genet. 33, 423-448. Santra, M.K., and Panda, D. (2003). Detection of an Intermediate during Unfolding of Bacterial Cell Division Protein FtsZ, loss of functional properties precedes the global unfolding of Ftsz. J. Biol. Chem. 278, 21336-21343. Scheffers, D.J., de Wit, J.G., den Blaauwen, T., and Driessen, A.J. (2001). Substitution of a conserved aspartate allows cation-induced polymerization of FtsZ. FEBS Lett. 494, 3437. Scheffers, D.J., and Driessen, A.J. (2002). Immediate GTP hydrolysis upon FtsZ polymerization. Mol. Microbiol. 43, 1517-1521. Scheffers, D.J., de Wit, J.G., den Blaauwen, T., and Driessen, A.J. (2002). GTP hydrolysis of cell division protein FtsZ, evidence that the active site is formed by the association of monomers. Biochemistry 41, 521-529.

152

Andreu , Oliva and Huecas

Sossong, T.M. Jr, Brigham-Burke, M.R., Hensley, P., and Pearce, K.H. Jr. (1999). Selfactivation of guanosine triphosphatase activity by oligomerization of the bacterial cell division protein FtsZ. Biochemistry 38, 14843-14850. Straub, F.B. (1942). Actin. Stud. Inst. Med. Chem. Univ. Szeged 2, 3-15. Stricker, J., Maddox, P., Salmon, E.D., and Erickson, H.P. (2002). Rapid assembly dynamics of the Escherichia coli FtsZ-ring demonstrated by fluorescence recovery after photobleaching. Proc. Natl. Acad. Sci. U S A 99, 3171-3175. Sun, Q., and Margolin, W. (1998). FtsZ dynamics during the division cycle of live Escherichia coli cells. J. Bacteriol. 180, 2050-2056. Uehara, T., Matsuzawa, H., and Nishimura, A. (2001). HscA is involved in the dynamics of FtsZ-ring formation in Escherichia coli K12. Genes Cells 6, 803-814. van den Ent, F., and Lowe, J. (2000). Crystal structure of the cell division protein FtsA from Thermotoga maritima. EMBO J. 19, 5300-5307. van den Ent, F., Amos, L.A., and Lowe, J. (2001). Prokaryotic origin of the actin cytoskeleton. Nature 413, 39-44. van den Ent, F., Moller-Jensen, J., Amos, L.A., Gerdes, K., and Lowe, J., 2002 F-actin-like filaments formed by plasmid segregation protein ParM. EMBO J. 21, 6935-43. Wang, X., and Lutkenhaus, J., 1993, The FtsZ protein of Bacillus subtilis is localized at the division site and has GTPase activity that is dependent upon FtsZ concentration. Mol. Microbiol. 9, 435-442. Weisenberg, R.C. (1972). Microtubule formation in vitro in solutions containing low calcium concentrations. Science 177, 1104-1105. White, E.L., Ross, L.J., Reynolds, R.C., Seitz, L.E., Moore, G.D., and Borhani, D.W. (2000). Slow polymerization of Mycobacterium tuberculosis FtsZ. J. Bacteriol. 182, 4028-4034. Yan, K., Pearce, K.H. and Payne, D.J. (2000). A conserved residue at the extreme C-terminus of FtsZ is critical for the FtsA-FtsZ interaction in Staphylococcus aureus. Biochem. Biophys. Res. Commun. 270, 387-392. Yu, X.C., and Margolin, W. (1997). GTP-dependent dynamic assembly of bacterial cell division protein FtsZ into asters and polymer networks in vitro. EMBO J. 16, 5455-5463.

NOTE ADDED IN PROOF Significant advances in several of the subjects reviewed have appeared since this chapter was submitted. These include, regarding the dynamics of FtsZ (reviewed by Romberg and Levin, 2003), the demonstration that at steady state polymers of FtsZ from E. coli are predominantly GTP bound, and the measurement of the intrinsic GTP hydrolysis rate, which is ratelimiting for nucleotide turnover (Romberg and Mitchison, 2004). Two different screening strategies have been employed to discover small molecules that respectively inhibit FtsZ assembly (Wang et al. 2003) and the interaction of FtsZ with ZipA (Kenny et al., 2003). The MinCDE proteins are organized into extended membrane-associated spiral structures in E. coli,

FtsZ folding, self-association, activation and assembly

153

and their pole-to-pole oscillation reflects changes in their distribution within the coiled structure (Shih et al., 2003).

ADDITIONAL REFERENCES Kenny, C.H., Ding, W., Kelleher, K., Bernard, S., Dushin, E.G., Sutherland, A.G., Mosyak, L., Kriz, R., and Ellestad, G., 2003, Development of a fluorescence polarization assay to screen for inhibitors of the FtsZ/ZipA interaction. Anal. Biochem. 323, 224-233. Romberg, L., and Mitchison, T.J., 2004, Rate-limiting guanosine-5'-triphosphate hydrolysis during nucleotide turnover by FtsZ, a prokaryotic tubulin homologue involved in bacterial cell division. Biochemistry 43, 282-288. Romberg, L., and Levin, P.A., 2003, Assembly dynamics of bacterial cell division protein FtsZ, poised at the edge of stability. Annu. Rev. Microbiol. 57, 125-154. Shih, Y.L., Le, T., and Rothfield, L., 2003, Division site selection in Escherichia coli involves dynamic redistribution of Min proteins within coiled structures that extend between the two cell poles. Proc. Natl. Acad. Sci. USA 100, 7865-7870. Wang, J., Galgoci, A., Kodali, S., Herath, K.B., Jayasuriya, H., Dorso, K., Vicente, F., González, A., Cully, D., Bramhill, D., and Singh, S., 2003, Discovery of a small molecule that inhibits cell division by blocking FtsZ, a novel therapeutic target of antibiotics. J. Biol. Chem. 278, 44424-44428.

This page intentionally left blank

Chapter 8 Sequence and structural alignments of eukaryotic and prokaryotic cytoskeletal proteins

EDUARDO LÓPEZ-VIÑAS and PAULINO GÓMEZ-PUERTAS* Centro de Astrobiología (CSIC-INTA) Instituto Nacional de Técnica Aeroespacial Ctra de Ajalvir, km 4 28850-Torrejón de Ardoz, Madrid. Spain *corresponding author

1.

Introduction

Cell division is an essential bacterial process driven by a complex machinery that involves at least ten proteins (FtsZ, FtsA, ZipA, FtsK, FtsQ, FtsL, FtsB, FtsW, FtsI and FtsN, as described in E. coli and B. subtilis; Chen and Beckwith, 2001; Errington and Daniel, 2001; Errington et al., 2003). The roles of most of these proteins, and the complexes that they form are still unknown. In eukaryotes, the process of cell division needs of the presence of a cytoskeleton, whose lack has been a defining characteristic of the prokaryotes until recently. In 1992, a series of published works started to establish evolutionary relationships between bacterial proteins FtsA and FtsZ and their eukaryotic counterparts actin and tubulin. FtsA and MreB were suggested to have a similar tertiary structure to the ATPase domains of hexokinase, Hsc70 and actin, proposing a common evolutionary origin (Bork et al., 1992). Comparison of sequences and structures of FtsZ and tubulin suggested in the same manner a common ancestor for both proteins (de Boer et al., 1992; de Pereda et al., 1996; RayChaudhuri and Park, 1992). Studies on the behavior of the Z ring in bacterial cell division showed that the structure is highly dynamic, continuously remodeled in relation with its

Molecules in Time and Space, Edited by Vicente et al. Kluwer Academic/Plenum Publishers, New York 2004

155

156

López-Viñas and Gómez-Puertas

GTPase activity (Lu et al., 2000; Mukherjee and Lutkenhaus, 1998; Stricker et al., 2002), suggesting a functional relationship with the dynamic instability of the polymerization processes of tubulin (Mitchison and Kirschner, 1984). The crystallization and resolution of the three-dimensional structures of tubulin (Nogales et al., 1998b), FtsZ (Lowe and Amos, 1998), actin (Kabsch et al., 1990), MreB (van den Ent et al., 2001b) and FtsA (van den Ent and Lowe, 2000) demonstrated, through the alignment of structures and the comparison of the aligned amino acidic sequences, a common origin for tubulin and FtsZ (Erickson, 1997; Erickson, 1998; van den Ent et al., 200la) and actin, MreB and FtsA (van den Ent et al., 200la). Multiple alignment of proteins, and mainly those based on the comparison of experimentally obtained-three dimensional atomic structures (structural alignments), are a very valuable source of information related to the evolutionary strategies followed by the different members of a family of proteins to conserve or modify their functions and structures (Zuckerkandl and Pauling, 1965). During evolution, the three dimensional structures and, usually, the amino acidic sequences of proteins that diverged from a common ancestor, called homologous proteins, conserve enough similarity to allow detailed comparison among them. The analysis of structural alignments allows the detection of at least three types of regions or multiple alignment positions according to conservation, and related with their common and different functions. The most evident is the presence of completely or almost completely conserved amino acids. Since natural selection works at the level of molecular function, amino acid residues that are involved in the common function of a given protein family are more likely to remain unchanged during evolution, and might be identified by their characteristic conservation in multiple sequence alignments and the conservation of their relative positions in the three dimensional structures of different members of the family. More interesting is the detection of positions with patterns of variation related to the differences in function of the corresponding groups or subfamilies of proteins: residues or regions that are specifically conserved only in subfamilies with a certain functional specialization, that implies functional changes that were selected during evolution. This type of residues, named tree-determinant residues as they are identified as conserved for specific subfamilies, could be related with specific active sites, substrate binding sites or protein-protein interaction surfaces (del Sol Mesa et al., 2003; Lopez-Romero et al., 2004). The third type of residues carrying evolutionary information as can be deduced from multiple alignment of sequences are those positions that correspond to compensatory mutations that stabilize the mutations in one protein with changes in the other (correlated mutations) (Göbel et al., 1994; Olmea and Valencia, 1997). Correlated mutations can be used to predict three-

Sequence and structural alignments of cytoskeletal proteins

157

dimensional contacts in proteins (Fariselli et al., 200la, b) and they are very effective in the detection of protein-protein interaction contacts, being able to select the correct structural arrangement of two proteins based on the accumulation of signals in the proximity of interacting surfaces, or to detect interacting protein pairs based on the differential accumulation of correlated mutations between the interacting partners (Pazos et al., 1997a; Valencia and Pazos, 2002). The comparison, using structural alignments of protein sequences, of the homologous proteins tubulin/FtsZ and actin/MreB/FtsA, will allow us to detect patches of common residues and structural subdomains, related to the conserved function of the homologous proteins. Conserved residues specific of the prokaryotic subfamilies of sequences, not present in the eukaryotic ones, will be related with differences in function and macromolecular interactions. Finally, the analysis of correlated mutations can be used to build models of protein complexes, in the absence of crystallized pairs of interacting proteins.

2.

Protein sequence data sources and alignment of structures

Representative sequences of the FtsA, MreB, FtsZ, actin, and tubulin families of proteins were obtained from Pfam (Bateman et al., 2002; http://www.sanger.ac.uk/Software/Pfam) and HSSP (Sander and Schneider, 1993; ftp://ftp.embl-heidelberg.de/pub/databases/protein_extras/hssp). The three dimensional structures of (entry 1TUB-B), yeast actin (entry 1YAG-A), FtsZ from Methanococcus jannaschii (entry 1FSZ), FtsA from Thermotoga maritima (entry 1E4F-T) and MreB from Thermotoga maritima (entry 1JCE) were obtained from the Protein Data Bank. The secondary structure elements of 1YAG-A, 1FSZ, 1JCE and 1E4F-T were extracted from the DSSP database (Kabsch and Sander, 1983; ftp://ftp.cmbi.kun.nl/pub/molbio/data/dssp). The secondary structure of 1TUB-B was obtained from the original PDB file as reported by the authors (Nogales et al., 1998b). The structure-based sequence alignments of actin/MreB/FtsA and tubulin/FtsZ were performed using DALI (Holm and Sander, 1993; http://www2.ebi.ac.uk/dali). The alignments were visualized and colored according to residue conservation (average BLOSUM62 score) using Belvu 2.8 (http:// www.sanger.ac.uk/;esr/Belvu.html). For clarity, the numbering of residues in the text will be referred to E. coli sequences in the case of FtsA, MreB and FtsZ proteins (SwissProt names: ftsa_ecoli, mreb_ecoli, ftsz_ecoli) and to the yeast actin (act_yeast) and pig (tbb_pig), respectively.

158

3.

López-Viñas and Gómez-Puertas

Structural alignment of actin, MreB and FtsA

Filamentous actin is a major component of the eukaryotic cytoskeleton, involved in processes of cytokinesis, shape conformation, cell movement and organelle distribution. The three dimensional structure of actin from different organisms has been solved or modeled both in presence of absence of ligands or in a polymerized state (Chik et al., 1996; Kabsch et al., 1990; McLaughlin et al., 1993; Vorobiev et al., 2003). The actin family of proteins contains two structural domains (1 and 2) divided in two subdomains (A and B). Subdomains 1A (yeast actin residues 1-34, 70-137 and 339-375) and 2A (residues 138-181 and 270-338) are the larger ones and have a common fold, comprising a strand of five surrounded by (RnaseH fold). The subdomains 1B and 2B correspond to residue numbers 35-69 and 182-269, respectively. The structure and ATP binding properties of subdomains 1A and 2A are common to the so-called actin superfamily of proteins, that includes Hsp70 (Flaherty et al., 1990), some sugar kinases (Bork et al., 1992) and the bacterial proteins FtsA and MreB (reviewed in Carballido-López and Errington, 2003; van den Ent et al., 2001a). Figure 1 shows the structure-based multiple alignment of three sequences of actin, three sequences of MreB and three sequences of FtsA. Despite the absence of a high degree of similarity in the primary sequence, the atomic structures of FtsA, MreB and actin show a very similar threedimensional architecture (Figure 2) mainly in the subdomains 1A, 2A and 2B and in the nucleotide- binding site (Kabsch et al., 1990; van den Ent et al., 2001b; van den Ent and Lowe, 2000). The alignment of the secondary structure elements (Figure 1) shows as expected a similar distribution, being the main difference the presence, in FtsA, of a new subdomain, 1C, comprising residues 88-164 in the E. coli FtsA sequence, as well as the lack of subdomain 1B, present both in MreB and actin structures (Figure 2). The analysis of the structures and sequences of the FtsA, MreB and actin subfamilies allows the identification of specific regions present only in one or two groups of sequences. During evolution, the different subfamilies have varied and adapted to specific functions and interactions. The variations in structure and sequence could be a source of information on the specific characteristics of each subfamily. One of this specializations, in the transition from prokaryotic to eukaryotic organisms, was the co-evolution with a family of proteins specialized in assisting the folding of other proteins, as the chaperonins GroEL -bacteria- thermosome -archaea- or CCT -eukarya- (Gómez-Puertas et al., 2004; Llorca et al., 2001; Valpuesta et al., 2002), or the eukaryotic prefoldin (Martín-Benito et al., 2002). Prokaryotic GroEL binds the unfolded polypeptides mainly through non-specific hydrophobic contacts, while, in contrast, the eukaryotic chaperonin CCT

Sequence and structural alignments of cytoskeletal proteins

159

shows a high specificity for the cytoskeletal proteins actin and tubulin (Sternlicht et al., 1993).

Figure 1. Structure-based alignment of three representative sequences of actin (act_yeast, actin from Saccharomyces cerevisiae; actb_rabit, cytoplasmic b-actin from Oryctolagus cuniculus; act1_orysa, actin 1 from Oryza sativa), three sequences of MreB (mreb_ecoli, MreB protein from Escherichia coli; mreb_bacsu, from Bacillus subtilis; mreb_thmar, from Thermotoga maritima) and three sequences of FtsA (ftsa_ecoli, from E.coli; ftsa_haein, from Haemophilus influenzae; ftsa_thmar, T. maritima). Residues are colored by average BLOSUM62 score using Belvu v.2.8 (Erik Sonnhammer, http://www.sanger.ac.uk/;esr/Belvu.html). The secondary structures of actin from yeast (1YAG_A), FtsA protein from T. maritima (1E4F) and MreB from T.maritima (1JCE) are also depicted in the alignment (E, H, ). The CCT-binding motifs of the actin sequence are marked as X.

When the specific binding sites of actin to CCT are compared with the multiple sequence alignment of actin, FtsA and MreB (Figure 1) and with the structure alignment of the three-dimensional structures of the three proteins, it is surprising to see that the four binding sites (marked as X in

160

López-Viñas and Gómez-Puertas

Figure 1 and as bold trace in Figure 2) are localized mainly in segments of actin absent in FtsA or in both, FtsA and MreB.

Figure 2. Structural comparison of Protein data bank entries 1YAG-A (yeast actin), 1E4F (FtsA from T. maritima) and 1JCE (MreB from T. maritima). Positions of subdomains 1A, 1B, 2A and 2B of actin and subdomain 1C of FtsA are indicated. The regions in actin structure corresponding to the CCT-binding sites are depicted in bold trace. A molecule of ATP is included in all structures to indicate the conserved position of the nucleotide-binding site.

Sequence and structural alignments of cytoskeletal proteins

161

The first CCT binding site of actin is placed in a loop comprising residues 37-51 in the yeast actin sequence. This loop is smaller in the MreB structure, lacking most of the corresponding residues. This site and the second one, residues 62-66, are completely absent in FtsA, as this protein lacks the entire subdomain 1B, where both of them are located. The situation is similar in the case of the third and fourth fragments, actin residues 195206 and 229-250 respectively, located in the subdomain 2B, which are, in part, absent in the prokaryotic structures (Llorca et al., 2001). The location of these same CCT binding sites of monomeric actin in the atomic model of polymerized filamentous actin (Holmes et al., 1990) indicates that a large part of the interactions generating the filament occur through residues that form part of the CCT-binding sites involving, among others, residues 40–45, 63–64, 195–197, 202–204, and 243–245 (Holmes et al., 1990; Llorca et al., 2001). This fact suggests that these regions, specific to the actin subfamily, have appeared during evolution to play roles different that those common to MreB and FtsA, and that they can be easily detected through sequence and structure comparison of the members of the family.

4.

A model for FtsA polymerization

One of the most surprising characteristics of the structure of FtsA when compared to its homologues of the actin family of proteins is the absence of the subdomain 1B and the presence of the extra subdomain 1C (van den Ent and Lowe, 2000). This type of subfamily-specific regions can be related with specific function or protein-protein interaction sites, as indicated above in the case of the CCT binding sites of actin. Indeed, changes in the relative position of subdomain 1C were suggested to be related with ATP binding in the active center of the protein (van den Ent and Lowe, 2000). FtsA has been postulated to be able to form homodimers and polymers based on biochemical studies made using both E. coli and B. subtilis model organisms (Feucht et al., 2001; Sanchez et al., 1994; Yim et al., 2000). In agreement with all these information, and using an integrative approach of bioinformatic and biochemical techniques, recently a role for the subdomain 1C in FtsA homodimerization has been proposed (Carettoni et al., 2003). The bioinformatic approach was based on the use of correlated mutations. This term is used to define concerted changes in protein sequence, likely to compensate substitutions that occur independently, indicating a tendency of positions in proteins to mutate coordinately. This tendency is measured by analyzing the correlation between changes in pairs of positions in multiple sequence alignments. As the compensation is believed to be favorably disposed between residues in physical contact, correlated mutation

162

López-Viñas and Gómez-Puertas

calculations can be used for contact prediction, either for inter-domain docking, or between two different proteins. Using the algorithm plotcorr for correlation coefficients calculation in multiple alignments of proteins (Olmea et al., 1999; Pazos et al., 1997a; Pazos et al., 1997b), a particularly notable correlation was found between a group of amino acids present in the loop connecting S6 and S7 strands of the FtsA-specific subdomain 1C (residues 138-151 in the sequence of E. coli FtsA) and another group of amino acids surrounding the ATP binding site (Table I and Figure 3).

Based on this information, a model was proposed for the homodimerization of FtsA. In this model (Figure 3), the correlated pairs of amino acids from both monomers (white and black spheres) are closely positioned, being the cleft located between subdomains 1A and 1C of one monomer in contact to the apical S12 and S13 strands in the 2B subdomain of the second monomer. This disposition is conceivable to allow not only homodimerization but also homopolymerization in elongated fibers. The model was compatible with the disposition of predicted residues implicated in protein-protein interaction obtained using a second independent bioinformatic approach to predict three-dimensional contacts (Carettoni et al., 2003; Fariselli et al., 2001a, b). More interestingly, the residues corresponding to the cleft in the first monomer of the multiple sequence alignment of orthologous FtsA proteins (amino acids 126-133 in E. coli sequence) are conserved and very similar to a peptide sequence (KLWVIPQ; peptide T8A-11) obtained through a phage-panning procedure that showed binding capabilities to native FtsA protein. Deletion mutants in this region of FtsA showed alterations in the functionality of the protein proposed to be

Sequence and structural alignments of cytoskeletal proteins

related dimer, second related

163

to the homodimerization of FtsA. In the proposed model for the FtsA the position of subdomain 1C of one monomer occupies, in the monomer, the space corresponding to the domain 1B of the spatially F-actin structure.

Figure 3. A model for FtsA homodimerization. The first molecule of FtsA (black trace) locates its subdomain 1C close to the ATP binding site of the second molecule (white trace). The molecule of ATP is represented as sticks. The head-tail homodimerization model allows the formation of elongated oligomers. Correlated residues in both structures (Table I) are depicted as white spheres in the subdomain 1C of the first monomer and black spheres in the ATP surrounding area in the second monomer.

164

López-Viñas and Gómez-Puertas

The apical S12 and S13 strands proposed to be implicated in the dimerization process correspond the E. coli FtsA residues 272-289, and located in the same structural position as the actin residues 195-250, located in the subdomain 2B and implicated in CCT binding and polymerization processes (Holmes et al., 1990; Llorca et al., 2001). Functionally, due to the proposed implication of the ATP binding in the flexibility of the subdomain 1C, or at least the possibility that the binding of the nucleotide could fix the subdomain 1C in a specific position (van den Ent and Lowe, 2000), the model suggest a interesting mechanism of nucleotide-regulated FtsA homodimerization or polymerization.

5.

Structural alignment of tubulin and FtsZ

Microtubules are composed of longitudinal filaments of heterodimers of and The subunits are aprox. 50% identical in sequence and bind the nucleotide GTP. In GTP can be hydrolyzed, leading to filament destabilization (for a review, see Nogales, 2000). The division Z ring in bacteria is composed mainly of FtsZ polymers formed in presence of GTP, resembling tubulin protofilaments (Bramhill and Thompson, 1994; Lowe and Amos, 1999; Lu et al., 2000). The Z-ring is also found in chloroplasts (Vitha et al., 2001; Osteryoung, 2001), but it is absent in mitochondria, where a ring-like structure is formed by dynamin (Erickson, 2000). Early after formation of FtsZ-ring, FtsA assembles to the structure. In 1998, both the structures of M. jannaschii FtsZ (Lowe and Amos, 1998) and heterodimer from bovine brain (Nogales et al., 1998b) were solved, providing conclusive evidence that FtsZ and tubulin are homologues (Nogales et al., 1998a). Despite low sequence similarity (10%–18% on the amino acid level), mainly in their C-terminal domain (Figure 4), the threedimensional structures of tubulin and FtsZ are closely similar (Figure 5). Both structures have a Rossmann fold in their N-terminal domain, with a parallel of six strands surrounded by In both tubulin and FtsZ, a loop (T7) from one subunit in is located into the active site of the next subunit and activates GTPase activity. The structure based alignment of FtsZ and tubulin (Figure 4) shows clear similarities between the two groups of sequences. However, in contrast, some differences in both sequence and structure can be observed. The main ones are the presence of two large helices in the C-terminus of tubulin (used to bind microtubule associated proteins) that are not present in FtsZ and the existence in tubulin of larger loops among secondary structure elements. In the same manner as actin sites when compared to FtsA and MreB structures, the specific sites of binding of tubulin to CCT are located mainly in these

Sequence and structural alignments of cytoskeletal proteins

165

loops (Llorca et al., 2001), suggesting again the co-evolution of the two families of proteins. The loop insertions in tubulin are related with specific functions, as CCT-binding sites (Figure 4, marked as X) and lateral interactions between protofilaments in microtubules (Andreu et al., 2002).

Figure 4. Structure-based alignment of four representative sequences of and (tbb_pig, from porcine brain -Sus scrofa-; tbb1_human, tubulin chain, brain specific -Homo sapiens-; tba_pig, from porcine brain -Sus scrofa-, tba1_yeast, tubulin chain -Saccharomyces cerevisiae-) and FtsZ (ftsz_ecoli, cell division protein FtsZ from Escherichia coli; ftsz_ pseae, from Pseudomonas aeruginosa; ftsz_bacsu, from Bacillus subtilis; ftsz_metja, from Methanococcus janaschii). All the symbols are as in Fig. 1. The secondary structures of alpha tubulin from porcine brain (1TUB_A) and FtsZ protein from M. janaschii (1FSZ) are also depicted.

Due to the extreme similarity between the structures of tubulin and FtsZ, a structural alignment can be easily performed between FtsZ (Protein Data Bank entry 1FSZ) and the tubulin dimer (entries 1TUB-A and 1TUB-B, corresponding to and respectively). In (Mingorance et al., 2001), each monomer of the tubulin dimer was substituted by the coordinates of a molecule of FtsZ according to the optimal alignment using the Dali algorithm (Holm and Sander, 1993). The resulting model (Figure 5), almost equivalent to the proposed by (Lowe and Amos, 1999), allowed to make a comparison between the interacting surfaces of tubulin and FtsZ

166

López-Viñas and Gómez-Puertas

dimers, showing that the nucleotide GTP was immobilized in the tubulin interface while the same structure leave an opening in the FtsZ interface that might allow the exchange of the nucleotide in FtsZ polymers. The model was compatible with the high nucleotide turnover of GTP in FtsZ polymers observed in vitro.

Figure 5. A model for FtsZ polymerization. The structure of the FtsZ dimerization model (left) based on the structure of the dimer (right, Protein Data Bank entry 1TUB) was obtained thought computer-assisted substitution of each one of the tubulin monomers in the crystallized structure by the coordinates of FtsZ. A molecule of GTP is depicted in the interface of both dimers to indicate the position of the nucleotide-binding site.

Sequence and structural alignments of cytoskeletal proteins

6.

167

Comparison of subdomain 1C of FtsA and the Cterminal domain of FtsZ

During the comparison of the sequences and structures of FtsA to actin and FtsZ to tubulin, two domains of the prokaryotic proteins were shown to be less similar to their eukaryotic counterparts, a characteristic usually indicative of specific functional differences. These regions are the subdomain 1C of FtsA (van den Ent and Lowe, 2000) and the C-terminal domain of FtsZ (Erickson, 1998; Lowe and Amos, 1998). The function of the domain 1C of FtsA is not known, although an essential role in protein homodimerization has been proposed based on bioinformatic and phage display approaches (Carettoni et al., 2003). The C-terminal domain of FtsZ seems to be a site for interaction between this protein and both FtsA (Din et al., 1998; Ma and Margolin, 1999; Yan et al., 2000) and ZipA (Mosyak et al., 2000). Despite the lack of homology between both proteins, a comparison of the structures of domain 1C of FtsA (residues 88-164 in the E. coli sequence) and the C-terminal domain of FtsZ (E.coli FtsZ residues 221-316, with a gap between residues 270-297) indicates that a significant structural alignment can be obtained, with a root-mean-square-displacement (rmsd) between backbone alpha carbons of the two domains of 3.0 angstroms (Figure 6, upper panel). Although the rmsd value is in the limit of structural similarity, the structure-based sequence alignment between sequences of both domains from a set of various bacterial organisms showed a similarity in sequence not far from the observed between the C-terminal domain of FtsZ and the correspondent domain of tubulin (Figure 6, lower panel). As the C-terminal domain of FtsZ appears to be a site specialized in the interaction with other division proteins, as FtsA and ZipA, and the FtsA subdomain 1C has been suggested to have a role in FtsA dimerization, some questions can be formulated. The first is related with the origin of the subdomain 1C, as FtsA is the only protein in the actin superfamily of proteins that presents this structure. It could be speculated that a segment of FtsZ was inserted into an FtsA ancestor (probably more similar to MreB and actin that the present structure of the protein) and evolved to conform the subdomain, loosing a and a corresponding to the FtsZ residues 270-297. This is related with the second question. The C-terminal domain of FtsZ (as well as the corresponding one of tubulin) collaborates in the interaction with the next subunit of the FtsZ or tubulin polymers. The similarity in structure and sequence of subdomain 1C of FtsA and the Cterminal domain of FtsZ suggests that this structure might be implied in the interaction of FtsA with FtsZ. This model, in the same manner as the model proposed for the homodimerization of FtsA, suggests a interesting

168

López-Viñas and Gómez-Puertas

mechanism of nucleotide-regulated FtsA binding to FtsZ, due to the dependence of ATP binding in the flexibility of the subdomain 1C.

Figure 6. Comparison of sequences and structures of FtsA subdomain 1C and FtsZ C-terminal domain. Upper panel. The structures of the subdomain 1C of the FtsA protein from T. maritima -white trace- and the C-terminal domain of the FtsZ protein from M. janaschii black trace- were aligned using the algorithm Dali (Holm and Sander, 1993). Lower panel. Structure-based alignment of four representative sequences of FtsA subdomain 1C and Cterminal domain of FtsZ. Colours, symbols and sequences are as in Figures 1 and 3.

Sequence and structural alignments of cytoskeletal proteins

169

This functional model, as well as the above described ones for the homodimerization of FtsA and FtsZ based on bioinformatic approaches, open new possibilities of research to elucidate the complex machinery that governs the assembly of the division ring proteins during the bacterial ceil cycle.

Acknowledgements We want to thank Jesüs Mingorance, Miguel Vicente and Alfonso Valencia for their continuous support and useful discussions. This work has been financed by a research grant from the "Fundación Ramón Areces". E. L-V. is recipient of a pre-doctoral fellowship from the "Fundación Ramón Areces".

References Andreu, J. M, Oliva, M. A. and Monasterio, O. (2002) Reversible unfolding of FtsZ cell division proteins from archaea and bacteria. Comparison with eukaryotic tubulin folding and assembly, J. Biol. Chem. 277, 43262-43270. Bateman, A., Birney, E., Cerruti, L., Durbin, R., Etwiller, L., Eddy, S. R., Griffiths-Jones, S., Howe, K. L., Marshall, M. and Sonnhammer, E. L. (2002) The Pfam protein families database, Nucleic Acids Res. 30, 276-280. Bork, P., Sander, C. and Valencia, A. (1992) An ATPase domain common to prokaryotic cell cycle proteins, sugar kinases, actin, and hsp70 heat shock proteins. Proc. Natl. Acad. Sci. USA 89, 7290-7294. Bramhill, D. and Thompson, C. M. (1994) GTP-dependent polymerization of Escherichia coli FtsZ protein to form tubules. Proc. Natl. Acad. Sci. USA 91, 5813-5817. Carballido-López, R. and Errington, J. (2003) A dynamic bacterial cytoskeleton. Trends Cell Biol. 13. 577-583. Carettoni, D., Gómez-Puertas, P., Yim, L., Mingorance, J., Massidda, O., Vicente, M., Valencia, A., Domenici, E. and Anderluzzi, D. (2003) Phage-display and correlated mutations identify an essential region of subdomain 1C involved in homodimerization of Escherichia coli FtsA. Proteins 50, 192-206. Chen, J. C. and Beckwith, J. (2001) FtsQ, FtsL and FtsI require FtsK, but not FtsN, for colocalization with FtsZ during Escherichia coli cell division, Mol. Microbiol. 42, 395-413. Chik, J. K., Lindberg, U. and Schutt, C. E. (1996) The structure of an open state of beta-actin at 2.65 Å resolution. J. Mol. Biol. 263, 607-623. de Boer, P., Crossley, R. and Rothfield, L. (1992) The essential bacterial cell-division protein FtsZ is a GTPase, Nature 359, 254-256. de Pereda, J. M., Leynadier, D., Evangelio, J. A., Chacón, P. and Andreu, J. M. (1996) Tubulin secondary structure analysis, limited proteolysis sites, and homology to FtsZ. Biochemistry 35, 14203-14215. del Sol Mesa, A., Pazos, F. and Valencia, A. (2003) Automatic methods for predicting functionally important residues. J. Mol. Biol. 326, 1289-1302.

170

López-Viñas and Gómez-Puertas

Din, N., Quardokus, E. M., Sackett, M. J. and Brun, Y. V. (1998) Dominant C-terminal deletions of FtsZ that affect its ability to localize in Caulobacter and its interaction with FtsA. Mol. Microbiol. 27, 1051-1063. Erickson, H. P. (1997) FtsZ, a tubulin homologue in prokaryote cell division, Trends in Cell Biology 7, 362-367. Erickson, H. P. (1998) Atomic structures of tubulin and FtsZ. Trends Cell Biol. 8, 133-137. Erickson, H. P. (2000) Dynamin and FtsZ. Missing links in mitochondrial and bacterial division. J. Cell Biol. 148, 1103-1105. Errington, J. and Daniel, R. A. (2001) Cell division during growth and sporulation, in, Bacillus subtilis and its relatives, from genes to cells. (L. Sonenshein, R. Losick, and J. A. Hoch, eds.), American Society for Microbiology, Washington, D.C, pp. 97–109. Errington, J., Daniel, R. A. and Scheffers, D. J. (2003) Cytokinesis in bacteria. Microbiol. Mol. Biol. Rev. 67, 52-65. Fariselli, P., Olmea, O., Valencia, A. and Casadio, R. (2001a) Prediction of contact maps with neural networks and correlated mutations. Protein Eng. 14, 835-843. Fariselli, P., Olmea, O., Valencia, A. and Casadio, R. (2001b) Progress in predicting interresidue contacts of proteins with neural networks and correlated mutations. Proteins. Suppl 5, 157-162. Feucht, A., Lucet, I., Yudkin, M. D. and Errington, J. (2001) Cytological and biochemical characterization of the FtsA cell division protein of Bacillus subtilis. Mol. Microbiol. 40, 115-125. Flaherty, K. M., DeLuca-Flaherty, C. and McKay, D. B. (1990) Three-dimensional structure of the ATPase fragment of a 70K heat-shock cognate protein. Nature 346, 623-628. Göbel, U., Sander, C., Schneider, R. and Valencia, A. (1994) Correlated mutations and residue contacts in proteins. Proteins 18, 309-317. Gómez-Puertas, P., Martín-Benito, J., Carrascosa, J. L., Willison, K. R. and Valpuesta, J. M. (2004) The substrate recognition mechanisms in chaperonins. J. Mol. Recognit. 17, 1-10. Holm, L. and Sander, C. (1993) Protein structure comparison by alignment of distance matrices. J. Mol. Biol. 233, 123-138. Holmes, K. C., Popp, D., Gebhard, W. and Kabsch, W. (1990) Atomic model of the actin filament. Nature 347, 44-49. Kabsch, W., Mannherz, H. G., Suck, D., Pai, E. F. and Holmes, K. C. (1990) Atomic structure of the actin, DNase I complex. Nature 347, 37-44. Kabsch, W. and Sander, C. (1983) Dictionary of protein secondary structure, pattern recognition of hydrogen-bonded and geometrical features. Biopolymers 22, 2577-2637. Llorca, O., Martín-Benito, J., Gómez-Puertas, P., Ritco-Vonsovici, M., Willison, K. R., Carrascosa, J. L. and Valpuesta, J. M. (2001) Analysis of the interaction between the eukaryotic chaperonin CCT and its substrates actin and tubulin. J. Struct. Biol. 135, 205218. López-Romero, P., Gómez, M. J., Gómez-Puertas, P. and Valencia, A., 2004, Prediction of functional sites in proteins by evolutionary methods. in Principles and practice. Methods in proteome and protein analysis (R. M. Kamp, J. Calvete, and T. Choli-Papadopoulou, eds.), Springer-Verlag, Berlin Heidelberg, pp. 319-340. Löwe, J., and Amos, L. A. (1998) Crystal structure of the bacterial cell-division protein FtsZ. Nature 391, 203-206. Löwe, J., and Amos, L. A. (1999) Tubulin-like protofilaments in Ca2+-induced FtsZ sheets, EMBO J. 18, 2364-2371. Lu, C., Reedy, M. and Erickson, H. P. (2000) Straight and curved conformations of FtsZ are regulated by GTP hydrolysis. J. Bacteriol. 182, 164-170.

Sequence and structural alignments of cytoskeletal proteins

171

Ma, X., and Margolin, W. (1999) Genetic and functional analyses of the conserved C-terminal core domain of Escherichia coli FtsZ. J. Bacteriol. 181, 7531-7544. Martín-Benito, J., Boskovic, J., Gómez-Puertas, P., Carrascosa, J. L., Simons, C. T., Lewis, S. A., Bartolini, F., Cowan, N. J. and Valpuesta, J. M. (2002) Structure of eukaryotic prefoldin and of its complexes with unfolded actin and the cytosolic chaperonin CCT. EMBO J 21, 6377-6386. McLaughlin, P. J., Gooch, J. T., Mannherz, H. G. and Weeds, A. G. (1993) Structure of gelsolin segment 1-actin complex and the mechanism of filament severing. Nature 364, 685-692. Mingorance, J., Rueda, S., Gómez-Puertas, P., Valencia, A. and Vicente, M. (2001) Escherichia coli FtsZ polymers contain mostly GTP and have a high nucleotide turnover. Mol. Microbiol. 41, 83-91. Mitchison, T. and Kirschner, M. (1984) Dynamic instability of microtubule growth. Nature 312, 237-242. Mosyak, L., Zhang, Y., Glasfeld, E., Haney, S., Stahl, M., Seehra, J. and Somers, W. S. (2000) The bacterial cell-division protein ZipA and its interaction with an FtsZ fragment revealed by X-ray crystallography. EMBO J. 19, 3179-3191. Mukherjee, A. and Lutkenhaus, J. (1998) Dynamic assembly of FtsZ regulated by GTP hydrolysis, EMBO J. 17, 462-469. Nogales, E. (2000) Structural insights into microtubule function. Annu. Rev. Biochem. 69, 277-302. Nogales, E., Downing, K. H., Amos, L. A. and Löwe, J. (1998a) Tubulin and FtsZ form a distinct family of GTPases. Nat. Struct. Biol. 5, 451-458. Nogales, E., Wolf, S. G. and Downing, K. H. (1998b) Structure of the alpha beta tubulin dimer by electron crystallography. Nature 391, 199-203. Olmea, O., Rost, B. and Valencia, A. (1999) Effective use of sequence correlation and conservation in fold recognition. J. Mol. Biol. 293, 1221-1239. Olmea, O. and Valencia, A. (1997) Improving contact predictions by the combination of correlated mutations and other sources of sequence information. Folding & Design 2, S25-S32. Osteryoung, K. W. (2001) Organelle fission in eukaryotes. Curr. Opin. Microbiol. 4, 639-646. Pazos, F., Helmer-Citterich, M., Ausiello, G. and Valencia, A. (1997a) Correlated mutations contain information about protein-protein interaction. J. Mol. Biol. 271, 511-523. Pazos, F., Olmea, O. and Valencia, A. (1997b) A graphical interface for correlated mutations and other protein structure prediction methods. Comput. Appl. Biosci. 13, 319-321. RayChaudhuri, D. and Park, J. T. (1992) Escherichia coli cell-division gene ftsZ encodes a novel GTP-binding protein. Nature 359, 251-254. Sánchez, M., Valencia, A., Ferrándiz, M. J., Sander, C. and Vicente, M. (1994) Correlation between the structure and biochemical activities of FtsA, an essential cell division protein of the actin family. EMBO J. 13, 4919-4925. Sander, C. and Schneider, R. (1993) The HSSP data base of protein structure-sequence alignments. Nucl. Acids Res. 21, 3105-3109. Sternlicht, H., Farr, G. W., Sternlicht, M. L., Driscoll, J. K., Willison, K. and Yaffe, M. B. (1993) The t-complex polypeptide 1 complex is a chaperonin for tubulin and actin in vivo. Proc. Natl. Acad. Sci. USA 90, 9422-9426. Stricker, J., Maddox, P., Salmon, E. D. and Erickson, H. P. (2002) Rapid assembly dynamics of the Escherichia coli FtsZ-ring demonstrated by fluorescence recovery after photobleaching. Proc. Natl. Acad. Sci. USA 99, 3171-3175.

172

López-Viñas and Gómez-Puertas

Valencia, A. and Pazos, F. (2002) Computational methods for the prediction of protein interactions. Curr. Opin. Struct. Biol. 12, 368-373. Valpuesta, J. M., Martín-Benito, J., Gómez-Puertas, P., Carrascosa, J. L. and Willison, K. R. (2002) Structure and function of a protein folding machine, the eukaryotic cytosolic chaperonin CCT. FEBS Lett. 529, 11-16. van den Ent, F., Amos, L. and Löwe, J. (2001a) Bacterial ancestry of actin and tubulin. Curr. Opin. Microbiol. 4, 634-638. van den Ent, F., Amos, L. A., and Löwe, J. (2001b) Prokaryotic origin of the actin cytoskeleton. Nature 413, 39-44. van den Ent, F. and Löwe, J. (2000) Crystal structure of the cell division protein FtsA from Thermotoga maritima. EMBO J. 19, 5300-5307. Vitha, S., McAndrew, R. S. and Osteryoung, K. W. (2001) FtsZ ring formation at the chloroplast division site in plants. J. Cell Biol. 153, 111-120. Vorobiev, S., Strokopytov, B., Drubin, D. G., Frieden, C., Ono, S., Condeelis, J., Rubenstein, P. A. and Almo, S. C. (2003) The structure of nonvertebrate actin, implications for the ATP hydrolytic mechanism. Proc. Natl. Acad. Sci. USA 100, 5760-5765. Yan, K., Pearce, K. H. and Payne, D. J. (2000) A conserved residue at the extreme C-terminus of FtsZ is critical for the FtsA-FtsZ interaction in Staphylococcus aureus. Biochem. Biophys. Res. Commun. 270, 387-392. Yim, L., Vandenbussche, G., Mingorance, J., Rueda, S., Casanova, M., Ruysschaert, J. M., and Vicente, M. (2000) Role of the carboxy terminus of Escherichia coli FtsA in selfinteraction and cell division. J. Bacteriol. 182, 6366-6373. Zuckerkandl, E. and Pauling, L. (1965) Evolutionary divergence and convergence in proteins. In Evolving Genes and Proteins (V. Bryson, and H. Vogel, eds.), Academic Press, New York, pp. 97–166.

Chapter 9 Bacterial morphogenes JESÚS MINGORANCE1*, ANABEL RICO1 and PAULINO GÓMEZPUERTAS2

1 Centro Nacional de Biotecnología (CSIC) Campus de Cantoblanco-UAM 28049-Madrid 2 Centro de Astrobiología (CSIC-INTA) InstitutoNacional de Técnica Aeroespacial Ctra de Ajalvir, km 4 288S0-Torrejón de Ardoz, Madrid Spain *corresponding author

1.

Introduction

The shape of the individual cells is the most prominent feature of bacteria upon microscopical examination, and it is not surprising that the earliest bacterial classification schemes relied almost exclusively on this trait (Cohn, 1875), and it is still a relevant character in descriptive and determinative bacteriology. In bacteria, like in other unicellular organisms, cell morphology governs several properties with significant influence over the cell life. Among these are the hydrodynamic properties of the cells, the way that cells interact with other cells and with the external medium, the cell surface to volume ratio and the way that this ratio changes with cell growth. In its turn, the surface to volume ratio affects the kinetics of nutrient uptake (Koch, 1995). In addition, the shape of the cell affects, and is affected by the processes of cell growth and division. Thus, in the bacteria the shape of the cells is not an anecdotical character, but a trait selected during the

Molecules in Time and Space, Edited by Vicente et al. Kluwer Academic/Plenum Publishers, New York 2004

173

174

Mingorance, Rico and Gómez-Puertas

evolutionary history of the organism, and integrated with its physiology and ecology (Zinder and Dworkin, 2001). Cell shape in bacteria is conferred by the murein sacculus, a stressbearing fabric that responds to the physical forces acting from the cytoplasm and from the external medium (Koch, 1995; Vollmer and Höltje, 2001). For this reason it has been argued that the analysis of the bacterial cell shape might be a subject better suited to physics than to genetics (Nanninga, 1998; Harold, 2002). This is true to some extent, as the structure of the murein sacculus is not in itself encoded by the genome, and its properties cannot be deduced from gene sequences. Understanding bacterial morphogenesis involves understanding the physical properties of the sacculus, how it grows, and how it responds to turgor pressure, questions addressed by the Surface Stress Theory (Koch, 1995). But the sacculus is a dynamic entity that is constantly recycled and results from the collective action of many gene products working in the context of the cell physiology. Indeed, cell shape is a stable trait in the bacteria, remains remarkably constant whithin lineages, or changes in reproducible ways, and may as well change by mutation, i.e. it is a genetically determined trait. Bacteria are often described in textbooks in terms of simple, basic shapes like rods, cocci, filaments or spiral cells. This is probably an oversimplification, as many other cell morphologies have been described (Zinder and Dworkin, 2001), but the simplest shapes are the most widespread among the different bacterial groups. Recently, several independent lines of research have proposed that the ancestor of extant bacteria was a rod-shaped cell and that all the other morphologies evolved from this (Koch, 2003; see also Chapters 1 and 13). On the other hand, it has been postulated that the bacterial default shape, i. e. the shape of the cell when morphogenesis is suppressed, is that of a coccus (Jones et al., 2001). This seems quite reasonable, as the spherical shape is the simplest, and indeed, it is the shape that wall-less cells have when they are grown in isotonic culture media. Yet it is important to bear in mind that cocci do have cell wall and have also active cell morphogenesis. Genetic analysis of bacterial morphogenesis has relied upon the study of morphological mutants, and most research has been done with the rodshaped Escherichia coli and Bacillus subtilis. This work will be restricted therefore to the genes described in these two organisms. These genes have been usually called "morphogenes", although the use of this term has been criticized because molecular and biochemical studies have shown that most of them are not regulatory genes controlling a "morphogenetic" program (Harold, 1995), so their primary functions are not to produce a certain cell shape, but to catalize some step in the complex peptidoglycan biosynthesis network, or to regulate some of those reactions. Keeping this in mind, we

Bacterial morphogenes

175

Figure 1. Schematic view showing some cell morphologies that can be derived from a basic rod-shaped cell like E. coli by mutation. The cell processes affected in each case are indicated below. Mutations in cell elongation or division are usually conditional mutations.

use the term morphogene in a broad sense simply to refer to genes that are closely related to the determination or the maintenance of the cell morphology. Several mutations are known in E. coli that result in stable or reproducible changes in the shape of the cells from rods to filaments, coiled filaments, cocci or irregular L-forms (Fig. 1; Donachie, 1993). Rod shape in E. coli results from the balance between two processes: cell elongation and cell division. Whenever cell division is delayed or suppressed without blocking cell growth the cells elongate continuously resulting in the formation of long multinucleated filaments. On the contrary, inhibition of elongation without interfering with division produces small coccoid cells. It follows then that all the genes involved in the processes of cell elongation and division enter into the category of morphogenes. In most cases the biochemical or cellular activities of the products of these morphogenes are not known; and even in those proteins in which some biochemical activitiy is known, the relation between the activity and the cellular architecture is not obvious. This lack of insight is due to the fact that these proteins work in large multiprotein complexes crossing the cytoplasmic membrane, or in the periplasmic space, bound to the peptidoglycan network, and are not amenable to standard biochemical analysis. This has changed during the last decade by the application of GFP fusions and immunolabelling combined with fluorescence microscopy. The application of these techniques has shown that bacterial cells contain several intracellular protein structures related with cell shape (Jones et al., 2001), cell division (Addinall et al., 1996) and the selection of division sites (Shih et al., 2003). In addition X-ray crystallography has shown that some of the proteins that form these

176

Mingorance, Rico and Gómez-Puertas

structures are related to eukaryotic cytoskeletal proteins (Löwe and Amos, 1998; van den Ent and Löwe, 2000; van den Ent et al., 2001). In addition, in the last few years the sequencing of several bacterial genomes has provided a wealth of data that has made possible the application of comparative genomics approaches to microbiology. One of these new approaches is the analysis of protein phylogenetic profiles, that is, the determination of the presence or absence of a gene in a group of bacterial genomes. This data are used to construct vectors, called phylogenetic profiles, which contain information about the distribution of the gene in the different taxonomic groups. The finding that two different genes have identical or similar phylogenetic profiles suggests that they might be functionally linked, meaning that they are involved in the same biochemical pathway, protein-protein interaction network, or cellular process. This analysis can be done on a large scale (Pazos and Valencia, 2001; Pellegrini et al., 1999), but it can be also applied to small groups of proteins, and it can be greatly enriched if additional biological information is included in the analysis. In this work the comparative approach is applied to the problem of cell shape determination in the bacteria.

2.

Data sources and criteria

The data were taken from the COGs (http://www.ncbi.nlm.nih.gov/COG/: Tatusov et al., 2001), CMR (http://www.tigr.org/; Uchiyama, 2003) or MGDB (http://mbgd.genome.ad.jp/; Peterson et al., 2001) public databases. Generally blast scores smaller than were considered acceptable in homology comparisons, although additional criteria were used to assign paralogous genes (duplicated genes that have taken on different functions, and evolve separately, but still have high sequence homology). In the case of the ftsL gene, a short and poorly conserved gene, additional genetic information was taken into account to identify as ftsL two sets of sequences (one belonging to Gram- bacteria and another belonging to the Gram+) that do not have significant homology between them, but have similar predicted secondary structures, have been found by experiments to play equivalent roles in E. coli and B. subtilis, and are found in equivalent positions within the dcw cluster (Daniel et al., 1998). For the paralogous pairs ftsW-rodA and ftsI-pbpA the genomic context was taken also into account, so the genes found within the dcw cluster were considered to be either ftsW and ftsI, while the genes outside the cluster were called rodA and pbpA respectively. When only one of the two genes was present the name was chosen taking into account the genomic context in the same manner.

Bacterial morphogenes

177

For simplicity, the genes that have been named differently in E. coli and in B. subtilis are denoted here using the E. coli nomenclature. There are several genes that affect cell shape but affect also basic cellular processes and have well documented pleiotropic effects, like dnaK (Bukau and Walker, 1989), or the genes involved in chromosome replication and segregation (Donachie, 1993). These will not be considered in this work.

3.

Genes for cell wall synthesis

The enzymes responsible for the synthesis of the peptidoglycan precursor subunits are coded by the mur genes: murA, murB, murC, murD, murE, murG, murF, mraY and ddlA/B (van Heijenoort, 1998; El Zoeiby et al., 2003). With the exception of murA, all the others can be found in the dcw gene cluster at least in some organisms (Ayala et al., 1994; Tamames et al., 2001). The products of these genes constitute a single and unique biosynthetic pathway, therefore when the pathway is present all the genes must be present, and when it is not, all of them are useless and can be lost. In agreement with this, these genes are found in most bacterial genomes sequenced. An important exception are the mycoplasma that do not have cell wall and accordingly do not have any of these genes. Another exception is the obligate wall-less endosymbiont Buchnera sp. that lacks the ddl gene. The presence of all the other genes in these intracellular bacteria, and the fact that the gene is present in the genome of B. aphidicola suggests that they might still need part of the pathway at some point in their life cycle, or alternatively, that the ancestor of the extant lineages lost the pathway after an initial evolutionary period of extensive gene loss, and these still conserve most of the genes due to the high stability of their genomes (Silva et al., 2003). The Chlamydia do not have cell-wall, but they contain the entire pathway, and indeed they are sensitive to penicillin, this has been dubbed the "Chlamydial anomaly", and suggests that at some point in their life-cycle the Chlamydiae need some kind of peptidoglycan structure (Chopra et al., 1998; Ghuysen and Goffin, 1999). All the genes of this pathway are essential in B. subtilis (Kobayashi et al., 2003) and are thought to be essential also in E. coli (Boyle and Donachie, 1998; van Heijenoort, 1996). Inhibition of any of the enzymes, either chemically or genetically, blocks the synthesis of peptidoglycan precursors, the cells swell and eventually lyse (van Heijenoort, 1996). Once the precursor is formed, the peptidoglycan is synthesized, recycled and modified by a series of enzymes collectively known as PBPs (penicillinbinding proteins). Most of these proteins are acyltransferases that belong to the SxxK acyltransferase superfamily. This superfamily includes a great

178

Mingorance, Rico and Gómez-Puertas

variety of proteins that share three short sequence motifs, SxxK, SxN and KTG, that occur at equivalent relative positions within the protein sequence, and form part of the catalytic centre (Goffin and Ghuysen, 2002). Some of the PBP's have clearly defined functions in cell growth, division, and morphogenesis, and will be treated in detail below in other sections. For others, nevertheless, the physiological roles are not known, and are the subject of current research (Young, 2001). It is worth to note here that the number of PBP's encoded in the different bacterial genomes ranges from total absence in the genomes of the mycoplasma, to more than 30 putative genes in the genome of Streptomyces coelicolor.

4.

Cell division genes

This group includes several genes specifically involved in cell division. These are often called fts genes, for "filamentous temperature sensitive" because they were first identified in E. coli or B. subtilis as conditional mutants that are unable to divide at the restrictive temperature and then grow as multinucleated filaments. Most of them are essential genes (Begg et al., 1980; Boyle et al., 1997; Donachie et al., 1979; Guzman et al., 1992; Lutkenhaus et al., 1980; Kobayashi et al., 2003; Suzuki et al., 1978), and several are found in the dcw gene cluster (Ayala et al., 1994). During division their protein products localize at midcell (Errington et al., 2003; Buddelmeijer and Beckwith, 2002; Margolin, 2000; see also Chapter 5) where they are thought to be part of a hypothetical multiprotein complex variously called septator (Pla et al., 1990), septosome (Witte et al., 1998) or divisome (Nanninga, 1991). Under non-permissive conditions the cell division mutants usually form long straight filaments, some of them with incipient constrictions. Other shapes have been described in double mutants of division and elongation genes (Begg and Donachie, 1985), and coiled cells and filaments have been found in C-terminal deletion mutants of FtsA in E. coli (Gayda et al., 1992; Yim et al., 2000), suggesting that some of these proteins might play a more direct role in morphogenesis. The strongest evidence is for FtsZ, that in addition to its central role in cell division, is involved also in the synthesis and localization of inert peptidoglycan (de Pedro et al., 1997), essential in the determination of cell shape (Chapter 3; Young, 2003).

Bacterial morphogenes

179

180

Mingorance, Rico and Gómez-Puertas

According to their phylogenetic profiles these genes can be divided in two groups (Table I). A group of highly conserved genes including ftsA, ftsI, ftsK, ftsQ, ftsW and ftsZ. Among this group interactions have been described or proposed for the products of the gene pairs ftsA-ftsZ, ftsI-ftsW, ftsA-ftsI and ftsA-ftsQ (Descoteaux and Drapeau, 1987; Matsuhashi, 1994; Wang et al., 1997; Yan et al., 2000). The phylogenetic profiles of these six genes are very similar, suggesting that all of them might be part of a protein complex that might be the core of the macromolecular cell division machinery. A second group of less conserved genes is formed by ftsB, ftsL, ftsN and zipA. The FtsB protein has been proposed to interact with FtsL (Buddelmeijer et al., 2002), and accordingly their profiles are very similar within the proteobacteria. The Gram+ ftsL gene has no sequence homology with the proteobacterial one, and a Gram+ homologue of ftsB has not been identified. Instead, a gene called divIC has been described (Levin and Losick, 1994) that closely follows the phylogenetic pattern of ftsL among the Gram+ bacteria. FtsL and DivIC are unstable proteins that localize to the division site in B. subtilis during septation and have been proposed to interact (Sievers and Errington, 2000), though it has been proposed too that the interaction is not direct but mediated by some other septal proteins (Robson et al., 2002). Finally, ZipA interacts with FtsZ (Haney et al., 2001; Moy et al., 2000; Mosyak et al., 2000) and FtsN seems to be functionally related to FtsA (Dai et al., 1993), but the similarity of their phylogenetic profiles suggests that there might be a direct ZipA-FtsN interaction, or that they might have a common role in septation. There are two genes of unknown function in the dcw cluster, mraZ and mraW The mraW gene is the most conserved gene found in this study, as it is found in all the sequenced bacterial genomes. It codes for an S-adenosylmethionine-dependent methyltransferase, but its substrate is unknown (Carrion et al., 1999; Miller et al., 2003). The mraZ gene, is also a gene of unknown function, is not as highly conserved but has a wide phylogenetic distribution, with a phylogenetic profile that does not match that of any other protein. In addition to these groups there are two genes originally described in B. subtilis whose protein products modulate the assembly of the FtsZ ring. The ezrA gene, that codes for a negative regulator of FtsZ ring formation (Levin et al., 1999) seems to be restricted to the low GC Gram+ group. The zapA gene codes for a positive regulator that promotes ring formation (GueirosFilho and Losick, 2002). It co localizes with the FtsZ ring, works stimulating the formation of high-order FtsZ assemblies, and is not necessary for the recruitment of additional cell division proteins to the septum. It has a wide distribution but does not correlate with any particular cell shape.

Bacterial morpbogenes

5.

181

Genes for the selection of the division site

The minCDE gene products are responsible for the determination of the correct division site in E. coli (Rothfield et al., 2001; Shih et al., 2003; see also Chapter 4). The mutations in any of these genes produce different phenotypes, depending on the mutated gene, but are characterized by filamentous growth and production of small cells devoid of DNA (minicells). The morphological anomalies found in these mutants are very likely due to alterations in Z-ring localization. In most cases the three genes are present or the three are absent (Table II), and often they are found in adjacent positions forming an operon. In the low GC Gram+ group the minCD genes appear next to the mreBCD operon. In this group the minE gene is absent, and the divIVA gene has been proposed to participate in the selection of the division site. The DivIVA protein is targeted to the division site and controls the topological specificity of MinCD activity (Edwards and Errington, 1997; see Chapter 4). In addition DivIVA has a function in chromosome segregation during sporulation in B. subtilis (Thomaides et al., 2001). This bifunctionality might explain the fact that divIVA is found in all the members of this group, while the minCD genes are found only in the rod-shaped members, and not in the cocci (Table II). The divIVA gene is found also in the actinobacteria, which do not have any of the min genes. In the Streptomyces coelicolor mycelium DivIVA localizes to hyphal tips and seems to be involved in apical growth and morphogenesis (Flärdh, 2003). The presence of the min genes in the sequenced genomes is strongly biased, being present in most bacilli, and absent in the majority of the bacteria that have other cell morphologies (Table II). Hence, the genes are absent in the actinobacteria, the chlamydiae, the Gram+ cocci, the mycoplasma, and the spirochaetae. They are absent also in some like the stalked C. crescentus, the coccus Magnetococcus sp., the rickettsiae and the coccobacilli Haemophilus influenzae and Pasteurella multocida (intriguingly, this still has a minC gene). The genes are present in the Gram- cocci Neisseria sp., where they have been found to be essential (Ramirez-Arcos et al., 2002). The broad taxonomic distribution of the min genes suggests that they were present in the ancestor of extant bacteria, which was probably a rodshaped organism (Koch, 2003; Chapter 1) that might have used the Min system for the selection of the division site. The genes would have been lost later in some species in which the cell shape or the mode of division had changed. Therefore the strong correlation between the morphology of the cells and the presence of the Min system is probably not due to the morphogenetic attributes of the system but, quite the opposite, it seems more

182

Mingorance, Rico and Gómez-Puertas

likely that it is the cell shape the factor that determines or sets restrictions to the mode of division site selection.

6.

Cell elongation genes

6.1

rodA-pbpA

The products of the rodA (mrdB) and pbpA (mrdA) genes have been proposed to form a complex that would participate in peptidoglycan synthesis during elongation and would be analogous to a complex formed by the FtsW and FtsI proteins and involved in septation (Matsuhashi, 1994). In E. coli they form a single operon that contains these two genes, and the dacA gene (Stoker et al., 1983). The conditional E. coli mutants lose the capacity to elongate, and therefore grew as osmotically stable spherical cells. Similarly, E. coli cells treated with mecillinam, which specifically inhibits the product of the pbpA gene, PBP2, become spherical. It might be expected therefore that cocci lack these genes (Table II), and indeed they are absent in some of them, like Synechocystis sp., Deinococcus radiodurans, Neisseria sp., and Buchnera aphidicola. They are absent also in Streptococcus pyogenes, but are present in the other streptococci and Gram+ cocci. It has been found that these genes are not essential in Streptococcus thermophilus, and upon mutation of any of the two the ovoid streptococcal cells became spherical suggesting that rodA and pbpA participate in cell elongation in the streptococci (Thibessard et al., 2002). Most striking is the absence of rodA-pbpA among the rhizobiales, as these have rod-shaped cells that must have peptidoglycan elongation activity. It has been described that in Sinorhizobium meliloti inhibition of cell division with cephalexin treatment, or by overproducing FtsZ, does not induce filamentation, but cell swelling and branching, suggesting that in these bacteria the elongation system does not work as in other rod-shaped bacteria, and that the septal peptidoglycan synthesis system is involved in cell elongation (Latch and Margolin, 1997), possibly by localized cell growth, either from the cell center like the cocci, or from the poles like Streptomyces coelicolor or Corynebacterium glutamicum (Daniel and Errington, 2003; see below).

Bacterial morphogenes

183

184

Mingorance, Rico and Gómez-Puertas

6.2

mreBCD

These genes where originally isolated in screenings for mecillinam resistance in E. coli (Doi et al., 1988; Wachi et al., 1989). The resistant mutants grew as cocci, and had increased expression of the ftsI gene, so it was proposed that the MreB protein might be a regulator of the processes of cell division and elongation (Wachi and Matsuhashi, 1989). The three genes usually appear together forming an operon, and in some groups are followed by the min genes (Levin et al., 1992; Varley and Stewart, 1992). MreB belongs, with FtsA, to the actin/hsp70/sugar kinase superfamily, and X-ray crystallography has shown that it has a fold very similar to actin, and it has been found that it polymerizes in an ATP-dependent form (van den Ent et al., 2001). In B. subtilis MreB is an essential protein that forms a helical structure that encircles the cell (Jones et al., 2001). In addition to mreB, B. subtilis has a second, homologous gene called mbl whose protein product also forms a helical structure (Jones et al., 2001; Carballido-López and Errington, 2003). The subcellular distribution of these proteins, together with their homology to eukaryotic actin suggest that the MreB-like proteins have a cytoskeletal role. An analysis of several sequenced genomes showed a strong bias in the distribution of the mreB gene, that was found to be present in bacilli and absent in cocci, so it was proposed that the bacterial default cell shape would be the spherical coccus shape, and that the mreB gene product would be a rod-shape determinant (Jones et al., 2001). Later the sequencing of additional genomes has shown that the mreBCD genes are lacking not only in the cocci, but also in the actinobacteria, characterized by branching and filamentous growth, and in the rhizobiales, a group of rodshaped proteobacteria. The study of the distribution of newly made peptidoglycan in several Gram+ bacteria has shown that there are two modes of growth among the bacilli: dispersed insertion through the lateral wall, and polar growth. It has been shown also that in the actinobacteria, either rodshaped or filamentous, cell elongation occurs by apical growth (Daniel and Errington, 2003). This led to the proposal that there are two different morphogenetic pathways in bacilli, an mreB-dependent pathway in organisms that grow by dispersed insertion of new material in the cell wall, like E. coli or B. subtilis, and an mreB-independent pathway in organisms that have apical growth, like the actinobacteria (Daniel and Errington, 2003). In the rhizobiales the entire cell elongation system (rodA-pbpA, mreBCD) is absent, suggesting that in this group the cells grow through a different mechanism, probably involving localized synthesis of new cell wall mediated by the division apparatus (see above). The functions and structures of the MreC and MreD proteins are not known, but their phylogenetic profiles are very similar to that of mreB,

Bacterial morphogenes

185

suggesting that they might be closely related to the MreB "cytoskeleton". Intriguingly, the Gram+ cocci, that do not have mreB, still have mreC and mreD.

7.

Other genes

7.1

dacA

The dacA gene codes for the periplasmic carboxypeptidase PBP5. In the proteobacteria it is found downstream of rodA and pbpA, possibly forming an operon. Its is closely related to dacC (PBP6), dacD (PBP6b) and pbpG (PBP7) and like these it is a non-essential gene that can be deleted without apparent phenotype, but its overexpression produces spherical cells (Stoker et al., 1983), Its deletion, in E. coli strains lacking multiple PBPs produces several morphological alterations, including changes in the cell diameter and branching (Nelson and Young, 2000). These alterations can be rescued with by a plasmid containing the wild type dacA gene, but not by the dacC gene. This has been exploited to produce dacA/dacC hybrids, and the dacA morphological effects were pinned down to two amino acid residues (Asp218 and Lys219) in the PBP5 sequence (Ghosh and Young, 2003). In addition it has been shown that the DD-carboxypeptidase activity of DacA is essential for rescuing the aberrant morphology, therefore it has been proposed that residues 218-219 participate in selecting some specific substrate, or in the response to some regulator, but the possible differences in substrate specificity between PBP5 and the other DD-carboxipeptidases are not known. The phylogenetic profile of the dacA gene does not correlate with the cell morphology (Table II), but intriguingly, it is very similar to those of ftsK and ftsQ (Table I), suggesting that dacA might interact with these proteins, or that it might be involved in septation. This is further supported by the fact that deletion of dacA suppresses the block on septation imposed by some ftsK mutations (Begg et al., 1995), while overexpression suppresses some ftsI mutations (Begg et al., 1990). Finally, in a recent model of bacterial morphogenesis it has been suggested that PBP5 affects the morphology of the cell through an FtsZ-dependent process: the synthesis and localization of inert peptidoglycan (Young, 2003), perhaps by regulating the levels of a pentapeptide substrate necessary for this type of peptidoglycan.

186

Mingorance, Rico and Gómez-Puertas

7.2 bolA The bolA gene was initially identified as a gene that produced osmotically stale spherical cells when it was overexpressed in the presence of a functional FtsZ protein in E. coli (Aldea et al., 1988). Overexpression produced also an increase in the levels of the DD-carboxypeptidases PBP5 and PBP6 (coded by the dacA and dacC genes; Aldea et al., 1989) suggesting a regulatory role for the BolA protein. It was found that the gene is not essential for E. coli as it could be deleted without any apparent effect (Aldea et al., 1989), although more recently it has been found that bolA deletion mutants growing in minimal medium fail to reduce their length upon entry into stationary phase or when subject to glucose depletion (Santos et al., 2002). These authors also found that the round morphology induced by BolA overproduction is strictly dependent on PBP5 and PBP6. The analysis of the BolA sequence suggested that it might have a helixturn-helix motif in the amino terminal part, and so it was proposed that BolA might be a DNA-binding protein that could function as a transcriptional regulator (Aldea et al., 1989). A multiple alignment of the BolA predicted aminoacid sequences from several bacteria has shown that this motif is not well conserved, and is unlikely to correspond to a real DNA-binding motif (J. Mingorance and M. Vicente, unpublished observations). The recent report of the solution structure of a mouse (Mus musculus) bolA homologue shows no sign of any known DNA-binding domain (Protein Data Bank entry 1IW5; Kasai et al., unpublished). Therefore, it seems unlikely that BolA regulates the expression of the dacA, dacC and ampC genes by directly regulating their transcription, though it may well be an indirect mechanism. Among the bacteria the bolA gene is found exclusively in the proteobacteria and its profile does not correlate with the cell morphology (Table II). Interestingly, in the eukaryotic genomes sequenced to date there are several bolA homologues. This, toghether with the fact that bolA is a proteobacterial gene suggests that they might be derived from the ancestral mitochondrial genome. The function of bolA homologues in eukaryotes is unkown, although one of them has been related to the control of cell division in yeast (Kim et al., 2002).

8.

Morphogene content and cell shape

There are some genes that seem to play more prominent roles in the determination of cell shape. Among these are the mreB gene (Jones et al., 2001; Daniel and Errington, 2003), the ftsZ gene (Young, 2003) and the dcw gene cluster (Tamames et al., 2001; see also Chapter 13). To evaluate their

Bacterial morphogenes

187

contributions to cell shape determination and maintenance a comparative approach was taken. Using the data shown in Tables I and II, a matrix of presence/absence of each gene in each genome was constructed that could be analysed using standard phylogeny reconstruction methods. No significant pattern was found when all the genes from both tables were included in the analysis, or when only the cell division genes (Table I) were included (the dcw cluster genes are analysed in Chapter 13). On the other hand, an interesting pattern arose from the analysis of the other morphogenes (Table II, Fig. 2). Most rod-shaped bacteria branch together forming a group of organisms (upper right branch) that posses a Min system, the mreBC and the rodA-pbpA genes (with two subgroups, with and without an mreD gene).

Figure 2. Urooted tree obtained from the pattern of presence/absence in several genomes of the morphogenes listed in Table II. The tree was done using the Wagner parsimony method with the MIX program of the WebPHYLIP package (Lim and Zhang, 1999). The boxes enclose bacterial species that share the same set of genes.

188

Mingorance, Rico and Gómez-Puertas

Next (in clockwise direction) there is a group that includes the rhizobiales (rod-shaped) and several cocci, all of them lacking the elongation genes (rodA-pbpA) and the mreBCD genes. All the species remaining (left half of the tree) lack the Min system, and most of them have morphologies different from the rod-shape. These are further grouped into three branches: one that includes the Chlamydia and the Mycoplasma, and in addition to the Min system lack the mreCD genes, one that only lacks the Min genes and includes the Spirochaetae and several proteobacteria, and one that includes the Gram+ cocci and several actinobacteria, that lack also the mreB gene. It has been proposed that there are two distinct ways to make rod-shaped cells (Daniel and Errington, 2003), but from these data another group is suggested that includes the rod-shaped rhizobiales. There would be then three different ways to make rods: i) in most bacilli the cells grow by insertion of peptidoglycan through the lateral wall, this growth is mediated by the RodAPBP2 proteins and probably directed by the MreB cytoskeleton. In these organisms the Min system selects the division site (probably assisted by other mechanisms like nucleoid occlusion), which is also the site of insertion of inert peptidoglycan. ii) in the second group, rod-shaped or filamentous cells grow by apical insertion of new material in an MreB-independent manner. In these bacteria it is the lateral cell wall the one that is formed by inert peptidoglycan, while the poles remain active for growth (Daniel and Errington, 2003). These bacteria have RodA-PBP2 homologues that might be involved in cell wall elongation. They do not have min genes, and the mechanism for selection of the division site is not known. iii) finally, the third group includes the Rhizobiales, in which cell growth is dependent on the division machinery (Latch and Margolin, 1997), in agreement with this, these organisms do not have cell wall elongation genes (rodA-pbpA), and they lack also the mreBCD genes. On the other hand they have a Min system that is likely involved in the selection of the division site, a process specially important if the division site is at the same time the site of cell growth in these bacteria.

9.

Conclusion

Cell shape is not the product of a particular gene or protein, but the result of the collective actions of many of them. These are involved in several processes, including peptidoglycan precursor synthesis, peptidoglycan synthesis and recycling, cell elongation, cell septation and division site selection. The analysis of the "morphogene" content of several bacterial genomes suggests that there are three different ways to make rod-shaped cells. The fact that most bacilli appear in a single branch of the tree supports

Bacterial morphogenes

189

previous works that suggest that the ancestor of all extant bacteria was a rodshaped organism (Siefert and Fox, 1998; Tamames et al., 2001; Koch, 2003; Chapters 1 and 13). This ancestor presumably had a dcw gene cluster (Nikolaichik and Donachie, 2000; Tamames et al., 2001; Chapter 13), a RodA-PBP2 system for cell elongation, an MreB cytoskeleton, and a Min system for selection of the division site.

Acknowledgements We wish to thank Miguel Vicente for his continued support. This work was financed by grants BIO2000-0451-P4-02 from the Ministerio de Ciencia y Tecnología, and QLRT-2000-00079 from the European Comission DG XII, both to Miguel Vicente, and by a research grant from the "Fundación Ramón Areces" to P.G-P.

References Addinall, S. G., Bi, E. and Lutkenhaus, J. (1996). FtsZ ring formation in fts mutants. J. Bacteriol. 178, 3877-3884. Aldea, M., Hernandez-Chico, C., de la Campa, A. G., Kushner, S. R. and Vicente, M. (1988). Identification, cloning, and expression of bolA, an ftsZ-dependent morphogene of Escherichia coli. J. Bacteriol. 170, 5169-5176. Aldea, M., Garrido, T., Hernandez-Chico, C., Vicente, M. and Kushner, S. R. (1989).

Induction of a growth-phase-dependent promoter triggers transcription of bolA, an

Escherichia coli morphogene. EMBO J. 8, 3923-3931. Ayala, J. A., Garrido, T., De Pedro, M. A. and Vicente, M. (1994). Molecular biology of bacterial septation. New comprehensive biochemistry. Bacterial cell wall. R. Hakenbeck and J. M. Ghuysen. Amsterdam, Elsevier. 27: 73-101. Begg, K. J., Hatfull, G. F. and Donachie, W. D. (1980). Identification of new genes in a cell envelope-cell division gene cluster of Escherichia coli: cell division gene ftsQ. J. Bacteriol. 144, 435-437. Begg, K. J. and Donachie, W. D. (1985). Cell shape and division in Escherichia coli: experiments with shape and division mutants. J. Bacteriol. 163, 615-622. Begg, K. J., Takasuga, A., Edwards, D. H., Dewar, S. J., Spratt, B. G., Adachi, H., Ohta, T., Matsuzawa, H. and Donachie, W. D. (1990). The balance between different peptidoglycan precursors determines whether Escherichia coli cells will elongate or divide. J. Bacteriol. 172, 6697-6703. Begg, K. J., Dewar, S. J. and Donachie, W. D. (1995). A new Escherichia coli cell division gene, ftsk. J. Bacteriol. 177, 6211-6222. Boyle, D. S., Khattar, M. M., Addinall, S. G., Lutkenhaus, J. and Donachie, W. D. (1997). ftsW is an essential cell-division gene in Escherichia coli. Mol. Microbiol. 24, 1263-1273. Boyle, D. S. and Donachie, W. D. (1998). mraY is an essential gene for cell growth in Escherichia coli. J. Bacteriol. 180, 6429-6432.

190

Mingorance, Rico and Gómez-Puertas

Buddelmeijer, N. and Beckwith, J. (2002). Assembly of cell division proteins at the E. coli cell center. Curr. Opin. Microbiol. 5, 553-557. Buddelmeijer, N., Judson, N., Boyd, D., Mekalanos, J. J. and Beckwith, J. (2002). YgbQ, a cell division protein in Escherichia coli and Vibrio cholerae, localizes in codependent fashion with FtsL to the division site. Proc. Natl. Acad. Sci. USA 99, 6316-6321. Bukau, B. and Walker, G. C. (1989). Cellular defects caused by deletion of the Escherichia coli dnaK gene indicate roles for heat shock protein in normal metabolism. J. Bacteriol. 171, 2337-2346. Carballido-Lopez, R. and Errington, J. (2003). The bacterial cytoskeleton: in vivo dynamics of the actin-like protein Mbl of Bacillus subtilis, Dev. Cell 4, 19-28. Carrión, M., Gómez, M. J., Merchante-Schubert, R., Dongarra, S. and Ayala, J. A. (1999). mraW, an essential gene at the dcw cluster of Escherichia coli codes for a cytoplasmic protein with methyltransferase activity. Biochimie 81, 879-888. Chopra, I., Storey, C., Falla, T. J. and Pearce, J. H. (1998). Antibiotics, peptidoglycan synthesis and genomics: the chlamydial anomaly revisited. Microbiology 144 ( Pt 10), 2673-2678. Cohn, F. J. (1875). Studies on bacteria. Milestones in microbiology. T. D. Brock, ASM Press: 210. http://www.asmusa.org/mbrsrc/archive/pdfs/milestone/l875p210/1875p210.pdf Daniel, R. A., Harry, E. J., Katis, V. L., Wake, R. G. and Errington, J. (1998). Characterization of the essential cell division gene ftsL(yIID) of Bacillus subtilis and its role in the assembly of the division apparatus. Mol. Microbiol. 29, 593-604. Daniel, R. A. and Errington, J. (2003). Control of cell morphogenesis in bacteria: two distinct ways to make a rod-shaped cell. Cell 113, 767-776. de Pedro, M. A., Quintela, J. C., Holtje, J. V. and Schwarz, H. (1997). Murein segregation in Escherichia coli. J. Bacteriol. 179, 2823-2834. Descoteaux, A. and Drapeau, G. R. (1987). Regulation of cell division in Escherichia coli K12: probable interactions among proteins FtsQ, FtsA, and FtsZ. J. Bacteriol. 169, 19381942. Doi, M., Wachi, M., Ishino, F., Tomioka, S., Ito, M., Sakagami, Y., Suzuki, A. and Matsuhashi, M. (1988). Determinations of the DNA sequence of the mreB gene and of the gene products of the mre region that function in formation of the rod shape of Escherichia coli cells. J. Bacteriol. 170, 4619-4624. Donachie, W. D., Begg, K. J., Lutkenhaus, J. F., Salmond, G. P., Martinez-Salas, E. and Vincente, M. (1979). Role of the ftsA gene product in control of Escherichia coli cell division. J. Bacteriol. 140, 388-394. Donachie, W. D. (1993). The cell cycle of Escherichia coli. Annu. Rev. Microbiol. 47, 199230. Edwards, D. H. and Errington, J. (1997). The Bacillus subtilis DivIVA protein targets to the division septum and controls the site specificity of cell division. Mol. Microbiol. 24, 905915. El Zoeiby, A., Sanschagrin, F. and Levesque, R. C. (2003). Structure and function of the Mur enzymes: development of novel inhibitors. Mol. Microbiol. 47, 1-12. Errington, J., Daniel, R. A. and Scheffers, D. J. (2003). Cytokinesis in bacteria. Microbiol. Mol. Biol. Rev. 67, 52-65. Flärdh, K. (2003). Essential role of DivIVA in polar growth and morphogenesis in Streptomyces coelicolor A3(2). Mol. Microbiol. 49, 1523-1536. Gayda, R. C., Henk, M. C. and Leong, D. (1992). C-shaped cells caused by expression of an ftsA mutation in Escherichia coli. J. Bacteriol. 174, 5362-5370.

Bacterial morphogenes

191

Ghosh, A. S. and Young, K. D. (2003). Sequences near the active site in chimeric penicillin binding proteins 5 and 6 affect uniform morphology of Escherichia coli. J. Bacteriol. 185, 2178-2186. Ghuysen, J. M. and Goffin, C. (1999). Lack of cell wall peptidoglycan versus penicillin sensitivity: new insights into the chlamydial anomaly. Antimicrob. Agents Chemother. 43, 2339-2344. Goffin, C. and Ghuysen, J. M. (2002). Biochemistry and comparative genomics of SxxK superfamily acyltransferases offer a clue to the mycobacterial paradox: presence of penicillin-susceptible target proteins versus lack of efficiency of penicillin as therapeutic agent. Microbiol. Mol. Biol. Rev. 66, 702-738. Gueiros-Filho, F. J. and Losick, R. (2002). A widely conserved bacterial cell division protein that promotes assembly of the tubulin-like protein FtsZ. Genes Dev. 16, 2544-2556. Guzman, L. M., Barondess, J. J. and Beckwith, J. (1992). FtsL, an essential cytoplasmic membrane protein involved in cell division in Escherichia coli. J. Bacteriol. 174, 77167728. Haney, S. A., Glasfeld, E., Hale, C., Keeney, D., He, Z. and de Boer, P. (2001). Genetic analysis of the Escherichia coli FtsZ.ZipA interaction in the yeast two-hybrid system. Characterization of FtsZ residues essential for the interactions with ZipA and with FtsA. J. Biol. Chem. 276, 11980-11987. Harold, F. M. (1995). From morphogenes to morphogenesis. Microbiology 141, 2765-2778. Harold, F. M. (2002). Force and compliance: rethinking morphogenesis in walled cells. Fungal Genet. Biol. 37, 271-282. Jones, L. J., Carballido-Lopez, R. and Errington, J. (2001). Control of cell shape in bacteria: helical, actin-like filaments in Bacillus subtilis. Cell 104, 913-922. Kim, M. J., Kim, H. S., Lee, J. K., Lee, C. B. and Park, S. D. (2002). Regulation of septation and cytokinesis during resumption of cell division requires uvi31+, a UV-inducible gene of fission yeast. Mol. Cells 14, 425-430. Kobayashi, K., Ehrlich, S. D., Albertini, A., Amati, G., Andersen, K. K., Arnaud, M., et al. (2003). Essential Bacillus subtilis genes. Proc. Natl. Acad. Sci. USA 100, 4678-4683. Koch, A. L. (1995). Bacterial Growth and Form. New York, Chapman & Hall. Koch, A. L. (2003). Were Gram-positive rods the first bacteria? Trends Microbiol. 11, 166170. Latch, J. N. and Margolin, W. (1997). Generation of buds, swellings, and branches instead of filaments after blocking the cell cycle of Rhizobium meliloti. J. Bacteriol. 179, 23732381. Levin, P. A., Margolis, P. S., Setlow, P., Losick, R. and Sun, D. (1992). Identification of Bacillus subtilis genes for septum placement and shape determination. J. Bacteriol. 174, 6717-6728. Levin, P. A. and Losick, R. (1994). Characterization of a cell division gene from Bacillus subtilis that is required for vegetative and sporulation septum formation. J. Bacteriol. 176, 1451-1459. Levin, P. A., Kurtser, I. G. and Grossman, A. D. (1999). Identification and characterization of a negative regulator of FtsZ ring formation in Bacillus subtilis. Proc. Natl. Acad. Sci. USA 96, 9642-9647. Lim, A. and Zhang, L. (1999). WebPHYLIP: A Web Interface to PHYLIP. Bioinformatics. 15, 1068-1069. Löwe, J. (1998). Crystal structure determination of FtsZ from Methanococcus jannaschii. J Struct Biol 124, 235-243.

192

Mingorance, Rico and Gómez-Puertas

Lutkenhaus, J. F., Wolf-Watz, H. and Donachie, W. D. (1980). Organization of genes in the ftsA-envA region of the Escherichia coli genetic map and identification of a new fts locus (ftsZ). J. Bacteriol. 142, 615-620. Margolin, W. (2000). Themes and variations in prokaryotic cell division. FEMS Microbiol. Rev. 24, 531-548. Matsuhashi, M. (1994). Utilization of lipid-linked precursors and the formation of peptidoglycan in the process of cell growth and division: membrane enzymes involved in the final steps of peptidoglycan synthesis and mechanism of their regulation. New comprehensive biochemistry. Bacterial cell wall. R. Hakenbeck and J. M. Ghuysen. Amsterdam, Elsevier. 27: 55-71. Miller, D. J., Ouellette, N., Evdokimova, E., Savchenko, A., Edwards, A. and Anderson, W. F. (2003). Crystal complexes of a predicted S-adenosylmethionine-dependent methyltransferase reveal a typical AdoMet binding domain and a substrate recognition domain. Protein Sci. 12, 1432-1442. Mosyak, L., Zhang, Y., Glasfeld, E., Haney, S., Stahl, M., Seehra, J. and Somers, W. S. (2000). The bacterial cell-division protein ZipA and its interaction with an FtsZ fragment revealed by X-ray crystallography. EMBO J. 19, 3179-3191. Moy, F. J., Glasfeld, E., Mosyak, L. and Powers, R. (2000). Solution structure of ZipA, a crucial component of Escherichia coli cell division. Biochemistry 39, 9146-9156. Nanninga, N. (1991). Cell division and peptidoglycan assembly in Escherichia coli. Mol. Microbiol. 5, 791-795. Nanninga, N. (1998). Morphogenesis of Escherichia coli. Microbiol. Mol. Biol. Rev. 62, 110129. Nelson, D. E. and Young, K. D. (2000). Penicillin binding protein 5 affects cell diameter, contour, and morphology of Escherichia coli. J. Bacteriol. 182, 1714-1721. Nikolaichik, Y. A. and Donachie, W. D. (2000). Conservation of gene order amongst cell wall and cell division genes in Eubacteria, and ribosomal genes in Eubacteria and Eukaryotic organelles. Genetica 108, 1-7. Pazos, F. and Valencia, A. (2001). Similarity of phylogenetic trees as an indicator of proteinprotein interaction. Protein Eng. 14, 609-614. Pellegrini, M., Marcotte, E. M., Thompson, M. J., Eisenberg, D. and Yeates, T. O. (1999). Assigning protein functions by comparative genome analysis: protein phylogenetic profiles. Proc. Natl. Acad. Sci. USA 96, 4285-4288. Peterson, J. D., Umayam, L. A., Dickinson, T., Hickey, E. K. and White, O. (2001). The Comprehensive Microbial Resource. Nucleic Acids Res. 29,123-125. Pla, J., Dopazo, A. and Vicente, M. (1990). The native form of FtsA, a septal protein of Escherichia coli, is located in the cytoplasmic membrane. J. Bacteriol. 172, 5097-5102. Ramirez-Arcos, S., Szeto, J., Dillon, J. A. and Margolin, W. (2002). Conservation of dynamic localization among MinD and MinE orthologues: oscillation of Neisseria gonorrhoeae proteins in Escherichia coli. Mol. Microbiol. 46, 493-504. Robson, S. A., Michie, K. A., Mackay, J. P., Harry, E. and King, G. F. (2002). The Bacillus subtilis cell division proteins FtsL and DivIC are intrinsically unstable and do not interact with one another in the absence of other septasomal components. Mol. Microbiol. 44, 663-674. Rothfield, L. I., Shih, Y. L. and King, G. (2001). Polar explorers: membrane proteins that determine division site placement. Cell 106, 13-16. Santos, J. M., Lobo, M., Matos, A. P., De Pedro, M. A. and Arraiano, C. M. (2002). The gene bolA regulates dacA (PBP5), dacC (PBP6) and ampC (AmpC), promoting normal morphology in Escherichia coli. Mol. Microbiol. 45, 1729-1740.

Bacterial morphogenes

193

Shih, Y. L., Le, T. and Rothfield, L. (2003). Division site selection in Escherichia coli involves dynamic redistribution of Min proteins within coiled structures that extend between the two cell poles. Proc. Natl. Acad. Sci. USA 100, 7865-7870. Siefert, J. L. and Fox, G.E. (1998). Phylogenetic mapping of bacterial morphology. Microbiology 144, 2803-2808. Sievers, J. and Errington, J. (2000). The Bacillus subtilis cell division protein FtsL localizes to sites of septation and interacts with DivIC. Mol. Microbiol. 36, 846-855. Silva, F. J., Latorre, A. and Moya, A. (2003). Why are the genomes of endosymbiotic bacteria so stable? Trends Genet. 19, 176-180. Stoker, N. G., Broome-Smith, J. K., Edelman, A. and Spratt, B. G. (1983). Organization and subcloning of the dacA-rodA-pbpA cluster of cell shape genes in Escherichia coli. J. Bacteriol. 155, 847-853. Suzuki, H., Nishimura, Y. and Hirota, Y. (1978). On the process of cellular division in Escherichia coli: a series of mutants of E. coli altered in the penicillin-binding proteins. Proc. Natl. Acad. Sci. USA 75, 664-668. Tamames, J., Gonzalez-Moreno, M., Mingorance, J., Valencia, A. and Vicente, M. (2001). Bringing gene order into bacterial shape. Trends Genet. 17, 124-126. Tatusov, R. L., Natale, D. A., Garkavtsev, I. V., Tatusova, T. A., Shankavaram, U. T., Rao, B. S., Kiryutin, B., Galperin, M. Y., Fedorova, N, D. and Koonin, E. V. (2001). The COG database: new developments in phylogenetic classification of proteins from complete genomes. Nucleic Acids Res. 29, 22-28. Thibessard, A., Fernandez, A., Gintz, B., Leblond-Bourget, N. and Decaris, B. (2002). Effects of rodA and pbp2b disruption on cell morphology and oxidative stress response of Streptococcus thermophilus CNRZ368. J. Bacteriol. 184, 2821-2826. Thomaides, H. B., Freeman, M., El Karoui, M. and Errington, J. (2001). Division site selection protein DivIVA of Bacillus subtilis has a second distinct function in chromosome segregation during sporulation. Genes Dev. 15, 1662-1673. Uchiyama, I. (2003). MBGD: microbial genome database for comparative analysis. Nucleic Acids Res. 31, 58-62. van den Ent, F. and Löwe, J. (2000). Crystal structure of the cell division protein FtsA from Thermotoga maritima. EMBO J. 19, 5300-5307. van den Ent, F., Amos, L. A. and Löwe, J. (2001). Prokaryotic origin of the actin cytoskeleton. Nature 413, 39-44. van Heijenoort, J. (1996). Murein synthesis. Escherichia coli and Salmonella. Cellular and molecular biology. F. C. Neidhardt, R. Curtis III, J. L. Ingraham et al. Washington, ASM Press. van Heijenoort, J. (1998). Assembly of the monomer unit of bacterial peptidoglycan. Cell Mol Life Sci 54, 300-304. Varley, A. W. and Stewart, G. C. (1992). The divIVB region of the Bacillus subtilis chromosome encodes homologs of Escherichia coli septum placement (minCD) and cell shape (mreBCD) determinants. J. Bacteriol. 174, 6729-6742. Vicente, M. and Löwe, J. (2003). Ring, helix, sphere and cylinder: basic geometry of prokaryotic division. EMBO Reports 4, 655-660. Vollmer, W. and Holtje, J. V. (2001). Morphogenesis of Escherichia coli. Curr Opin. Microbiol. 4, 625-633. Wachi, M., Doi, M., Okada, Y. and Matsuhashi, M. (1989). New mre genes mreC and mreD, responsible for formation of the rod shape of Escherichia coli cells. J. Bacteriol. 171, 6511-6516.

194

Mingorance, Rico and Gómez-Puertas

Wachi, M. and Matsuhashi, M. (1989). Negative control of cell division by mreB, a gene that functions in determining the rod shape of Escherichia coli cells. J. Bacteriol. 171, 31233127. Wang, X., Huang, J., Mukherjee, A., Cao, C. and Lutkenhaus, J. (1997). Analysis of the interaction of FtsZ with itself, GTP, and FtsA. J. Bacteriol. 179, 5551-5559. Witte, A., Brand, E., Mayrhofer, P., Narendja, F. and Lubitz, W. (1998). Mutations in cell division proteins FtsZ and FtsA inhibit phiX174 protein-E-mediated lysis of Escherichia coli. Arch. Microbiol. 170, 259-268. Van, K., Pearce, K. H. and Payne, D. J. (2000). A conserved residue at the extreme Cterminus of FtsZ is critical for the FtsA-FtsZ interaction in Slaphylococcus aureus. Biochem. Biophys. Res. Commun. 270, 387-392. Yim, L., Vandenbussche, G., Mingorance, J., Rueda, S., Casanova, M., Ruysschaert, J. M. and Vicente, M. (2000). Role of the carboxy terminus of Escherichia coli FtsA in selfinteraction and cell division. J. Bacteriol. 182, 6366-6373. Young, K. D. (2001). Approaching the physiological functions of penicillin-binding proteins in Escherichia coli. Biochimie 83, 99-102. Young, K.. D. (2003). Bacterial shape. Mol. Microbiol. 49, 571-580. Zinder, S. H. and Dworkin, M. (2001). Morphological and physiological diversity. The Prokaryotes: An evolving electronic resource for the microbiological community. M. Dworkin et. al. New York, Springer-Verlag.

Chapter 10 Genome structures, operating systems and the image of the machine ANTOINE DANCHIN* and STANISLAS NORIA HKU-Pasteur Research Centre Dexter HC Man Building, 8, Sassoon Road Pokfulam, Hong Kong Present address: Genetics of Bacterial Genomes Institut Pasteur, 23 rue du Docteur Roux, 75724 Paris Cedex 15 France * Corresponding author

1.

Introduction

Despite scattered reflections about the relationships tying genes and genomes, our view of genomes remains static. Worse, many scientists still speak about "junk" DNA for a large part of the genome, demonstrating that they consider it as a collection of genes rather than an integrated structure. When some speak about the genome "fluidity" or "plasticity" this is just to account for variation in the distribution of genes along the chromosome, not for the time-dependent expression of genes as a function of their distribution in the genome (Brunder and Karch, 2000; Dobrint and Hacker, 2001). Our view is commonly to split the genome text into the multiple collections of its individual genes. Some even advocate "draft sequencing" as a sufficient approach for studying genomes (Bhattacharyya, 2002), without understanding that there might exist rules of genome organisation that would only be visible in completed sequences. We ought to think of life, however, more as the set of relationships between objects: genes, gene products and metabolites, than of a simple collection of objects (Danchin, 2003). On august 7th 2003, at least 747 genome-sequencing programmes were

Molecules in Time and Space, Edited by Vicente et al. Kluwer Academic/Plenum Publishers, New York 2004

195

196

Danchin and Noria

underway or completed, including 140 complete and 355 ongoing prokaryotic genomes. This allows scientists, despite a significant bias in the species chosen for sequencing (often pathogens, because of our anthropocentric view of the world), to explore the sequences and compare them with one another. At first sight, especially when the organisms are widely divergent, the structures of their genomes look very different. This is certainly not unexpected in view of the very long time of divergence between organisms. As a case in point it is admitted (but this rests on a very shaky set of interpretations of molecular clocks and rRNA divergence) that the two model Bacteria, Escherichia coli and Bacillus subtilis would have diverged from a common ancestor some 1-1.5 billion years ago. All genes should have been randomly reshuffled if there had not been some selective pressure keeping some of them together. And, as a matter of fact, at least 100 segments of operons are conserved between these organisms, sometimes as full operons (in the case of molecular machines such as the ribosome or ATP synthase, for example). Does this mean that there is some kind of rule that maintains genes together? Is there a constraint in the organization of the chromosome? At the time when the B. subtilis and E. coli genome sequences were known we argued that this might be related to the structure of the cell (Danchin and Henaut, 1997). What can we say today, as we have so many genome sequences available?

2.

Cells as Turing machines

2.1 Data, programme and machine The discovery of the processes setting up regulation of gene expression, followed by that of the genetic code, spread the representation of life as the result of the expression of a "programme" (Yockey, 1992). This happened at a time when, in what was to become Computer Sciences or using the Gallic neologism "Informatics", the concept of programming (or writing algorithms) was already widely understood. The first computers had already been shown to operate as predicted by Turing, von Neumann and the many theoreticians and scientists who had uncovered the link between the arithmetics of whole numbers and logic. In a famous metaphor - that was to be implemented as a concrete object - Turing proposed that all computations involving integers as well as all operations of logic could be performed by a simple machine reading and modifying a tape carrying a linear sequence of symbols, the Universal Turing Machine. He could show that this required only the physical separation between the string of symbols and the machine.

Genome structures, operating systems and the image of the machine

197

More precisely he showed that the tape was carrying the data that allowed the machine to proceed. In terms of their anthropocentric meaning the data could be split into two types, a programme, that embedded the meaning of the logical sequence recognized by the machine, and the data, that were needed for the programme to operate (in a way they provided the context). However this distinction, which implies a purpose (in the implicit signification of programme), is not an intrinsic property of the machine: the algorithm that is read can be considered as purely declarative. This has important consequences in terms of the origin of the "genetic programme" since it would not formally require a pre-existing purpose. We shall not discuss this further here. The metaphor of genetic programme was initially only perceived as a convenient way to describe how cells live and develop. It was however understood early on that some strange organisms (were they living, or not alive?), the viruses, behaved as individual pieces of programmes, using the cell as the machine needed to make them multiply and subsequently propagate (often by destroying the machine): no virus can survive without a host living cell. Interestingly, the phenomenon of lysogeny demonstrated that viruses could be embedded into the very genetic programme of the cell. Later on, when computer programming developed, pieces of programmes were found to behave formally as do biological viruses, and were named "viruses" accordingly. This was a further indication that there was perhaps a deeper meaning of the "programme" metaphor in what is life. Furthermore, when it became possible to manipulate DNA in vitro, the analogy appeared even deeper: working on a string of symbols, on a paper or in a computer (in silico) is enough to allow scientists to construct in vitro, and then in vivo experiments that correspond to the very concrete action of reprogramming cells. Many bacteria, today, produce human proteins. Finally, the identification of widely spread horizontal gene transfer (Medigue et al., 1991), and subsequently nuclear cloning (Wilmut et al., 1997), perfected the analogy to a point where it can be considered as highly revealing, if not (of course) explaining life in totality.

2.2 The first von Neumann's conjecture: operating systems and "housekeeping genes" At this point it became interesting to explore more in depth the analogy: is it possible to really separate between "data + programme" and "machine" in living organisms? And, if so, can one find more detailed organization of the data + programme in the cell? Can we see how it evolved from the time when life appeared? The first question asks to demonstrate that there is a

198

Danchin and Noria

physical separation between the carrier of the programme and the machine. The DNA molecule is the obvious candidate, as argued frequently. But one must also find particular properties of the programme itself that should be present. It must be possible to represent it as a string of symbols: this is indeed the case of the DNA when represented as a text written in a fourletter alphabet, and genome sequencing programmes are the ultimate demonstration that this is quite fit to reality. The programme must specify "instructions" that make the machine manipulate it. The basic concepts of what became Molecular Biology are remarkably appropriate for that. "Gene expression" is organized in an algorithmic way, and the coding process introduced at the level of translation allows for recursive processes to be implemented (Hofstadter, 1979). And recursivity, in itself, displays remarkable properties in terms of information, properties that are quite fit to account for the "creative" features of life (Bennett, 1988; Danchin, 2003; Yockey, 1992). Now, von Neumann looked for a kind of hierarchy in the various processes that needed to be implemented for a Turing Machine to function as a concrete physical machine. And he introduced the concept of what is now named the Operating System (OS), a particular piece of the programme that is indispensable for the machine to operate (Silberschatz et al., 2001). This is defined by a series of functions that can be summarized as follows. A "virtual machine" hides from the "users" all the engineering details of the computer as a physical entity (we shall go to this point again below). Associated to it there must be a "resource manager" providing efficient and effective sharing among "users" of the machine (each one using and creating data and running programmes). It should be noticed at this point that the concept of user is not directly associated to the human interaction with the machine (Man-Machine interaction). In fact users can also be machines (printers, screens, memory storage devices, all kinds of peripherals; the corresponding algorithms are usually named "drivers") and some are programmes. Finally several classes of programmes managed by the OS can be summarized as system programs (compilers, editors, loaders, etc.), application support programs (database management systems, networking systems, etc.) and finally, the programmes that correspond to the goals of the machine, application programs. In different OSs the data can be organized in a variety of ways, and in particular, in "object oriented" OSs resource management can be associated to data serving as output or input of application programmes. As one can immediately see, there is no reason why there should exist only one type of OS. And, as a matter of fact, many exist and are in

Genome structures, operating systems and the image of the machine

199

competition, since no machine would work without an OS. But an OS is a programme, hence can be separated from a machine, and a given machine can run several OSs. In short, an OS plays the role of a "housekeeping" programme. Do we find similar properties in living organisms? At first sight, living cells display overall features similar to one another, and the (almost) universal rule of the genetic code, as well as of the DNA replication machineries would argue for universality. However, the process of cell division is remarkably different between the eukaryotes and the prokaryotes, for example. The compartmentalization is also very different in these organisms, with the former having a well-formed nucleus. In the class of prokaryotes, the components of the division machinery are different between Bacteria and Archaea, too. And even Bacteria are not homogeneous (see the interesting - and heated - debate about the origin of prokaryotes; Gupta, 1998; Mayr, 1998; Woese, 1998). Despite similarities, there are large differences in the housekeeping genes controlling replication, transcription and translation in cells, and these are so large that the question remains: do they all have a common origin, or did they evolve independently? Even in "monoderm" Bacteria, there are at least two classes in terms of replication: one class uses only one housekeeping DNA polymerase, for the management of both DNA strands, while in another class two polymerases are needed (Rocha, 2002). All this points to the idea of a common functional class, that of the analogue of an OS (Bozinovski et al., 2001), which would differ in different types of cells, thereby making classes within which horizontal gene transfer is possible, while it would be impossible or quite difficult otherwise. This may have been selected through evolution as a means to screen out a number of viruses, for example. We think that this may account for the many discrepancies observed when scientists compare phylogenetic trees constructed from ribosomal RNA, or from diverse families of proteins (creating virulent controversies, that with the present view would simply look irrelevant). In summary, a common formal feature is needed for all life to develop, and in particular, to allow for DNA replication and gene expression. This requires the presence of what has precisely been named "housekeeping" functions. If we explore the analogy of the Turing Machine revisited by von Neumann it is easy to place in parallel many of the functions needed by Operating Systems. And, exactly as is the case in the domain of real computers, there can be (and there is) more than one OS. Single cells would need an OS similar to that of personal computer OSs, with some timesharing properties. In some instances the OS could degenerate to that of a simple batched OS or more often to multiprogrammed batched OSs. For more complex organisms, one would obviously need parallel and/or

200

Danchin and Noria

distributed systems. And in general, because the analogue of the OS must manage many nanomachines, it would probably be highly modular and hierarchically structured, of the object-oriented type (i.e. managing resources inside data files). As a consequence, while it is important for each organism to identify its housekeeping genes, there is no compelling reason that would state that these genes should have exact counterparts in all organisms. The only good reason for universality would be historical: if it is difficult to create this or that function, it is likely that once it has appeared somewhere it will spread everywhere. This implies divergent evolution (but horizontal transfer as well). In contrast, for functions that are more straightforward one could witness diversity and/or convergent evolution.

2.3 The second von Neumann's conjecture: where is the image of the machine? All this may look fine. However, a Turing machine does not make Turing machines. John von Neumann argued convincingly that, if this were to happen there should be somewhere an image of the machine, and this image should be reproduced in parallel to replication of the programme (von Neumann, 1958). In addition to DNA replication, many structures are duplicated when cells multiply. In some cases, such as that of the centriole, this is particularly remarkable (and still not really understood; Rice and Agard, 2002). All structures of the cells must at some point duplicate, and one might think that its overall organisation is tightly linked to a duplication process. What would be the link, however? Interestingly there seems to exist a link between the control of centrosome duplication and maintenance of genomic stability that is particularly visible in cancer (Kramer et al., 2002; Wong and Stearns, 2003). Chromosomes of course duplicate, and in eukaryotic cells this means not only DNA structures, but also the whole chromatin, with its network of histones and regulatory proteins and RNAs. In bacteria, the daughter cells are not equivalent, as clearly seen in Caulobacter crescentus (Judd et al., 2003), and indirectly seen by whole genome analysis (Rocha et al., 2003). It is therefore interesting to explore the conjecture that the very structure for which we know how duplication proceeds, DNA, might also somehow carry the overall map of the cell. In short, there would be a map of the cell in the chromosome. Or stated in another way, the gene order in the chromosome should not be random. In most Bacteria and some Archaea, the existence of a higher number of genes in the leading strand, relative to the lagging strand has long been known after it was observed that rDNA and ribosomal proteins are

Genome structures, operating systems and the image of the machine

201

systematically coded in the leading strand of the E. coli chromosome (Ellwood and Nomura, 1982), then in several genomes (McLean et al., 1998). The preferential positioning of translation-related highly expressed genes (rDNA and ribosomal proteins) in the leading strand has been widely admitted to result from selection for lower transcription abortion and higher replication rates while avoiding the head-on collisions that happen when genes are located in the lagging strand. Within this view selection would only be effective for highly expressed genes and acting on the positioning of genes in the leading strand, thus creating a gene strand bias proportional to the expression level in replicating bacteria. Because the bias would match the frequency of replication, fast-growing bacteria should display a more important strand bias. It was however demonstrated recently, that, at least in B. subtilis, where essential genes have been identified experimentally (Kobayashi et al., 2003), the level of expression was not the cause of the bias, but, rather, this bias was due to the essential character of the genes (Rocha and Danchin, 2003). This suggested that replication/transcription collisions are selected against because this creates abortive transcripts, not replication pausing. It also suggests that essentiality is a major determinant of the chromosome structure. That this is a general feature in many bacteria has also been observed (Rocha and Danchin, unpublished). Other biases exist in the distribution of nucleotides, words or genes in bacterial genomes. We have mentioned distribution of genes in the leading, vs. lagging strand. Another bias between the strands has been known for some time, and used to identify the origins of replication in chromosomes where it was unknown (Lobry, 1996). And this bias is often so large that it could identify a universal principle (A+C enrichment in the lagging strand, vs. G+T in the leading strand) that can be seen when genes are shifted from one strand to the complementary one (Rocha and Danchin, 2001). Remarkably, this bias is the same under conditions of DNA shuffling and recombination that are highly variable, including situations where many Insertions Sequences are present or when they are absent (Rocha et al., 1999). As a consequence this indicates, at least in Bacteria, that there exist selective constraints that organize the genome irrespective of the way the DNA nucleotide sequence is managed. What could be these constraints?

2.4 Drosophiloculus, homunculus? Curiously, hints for a possible answer come from the study of eukaryotic organisms, in particular multicellular organisms. In the early eighties extraordinary mutations were discovered in the drosophila fly: mutants arose that had legs in the place of antennae. These mutants carried a mutation at

202

Danchin and Noria

the Antennapedia locus (Bachiller et al., 1994; Balavoine and Telford, 1995; Duboule and Dolle, 1989; Gaunt, 1991; Gehring, 1987; Kaufman et al. 1990). Many such homeotic genes were later on discovered, including in plants (for reviews see (Akam, 1998; Brock and van Lohuizen, 2001; Cobourne, 2000; Jagla et al., 2001; Kammermeier and Reichert, 2001; Lohmann and Weigel, 2002; Reichert, 2002; Stock, 2001)). Interestingly, homeotic genes are clustered on the genome in the same sequence as they are expressed on the anteroposterior axis of the animal, and this is true both for vertebrates and invertebrates (Bachiller et al., 1994; Balavoine and Telford, 1995; Duboule and Dolle, 1989; Gaunt, 1991). In short, they are collinear (as a string of symbols in a computer programme). A knock-out of a homeotic gene often results in the transformation of a segment towards a more anterior type of segment. In general it can be concluded that insects have one such set of homeotic genes, while mammals have four (Bachiller et al., 1994). Whether this is true in the case of fish is a matter of discussion (Holland, 1999). One interesting feature is the comparison between insect and Crustacea: the old observation by Geoffroy Saint Hilaire that the body plan of crustacea was reversed under the thorax (the abdomen becomes the back and vice versa) has been substantiated by the observation that modification of homeotic genes between insects and Crustacea affected their body plan (Averof, 1997; Averof and Akam, 1995). An interesting summary of all these discussions can be found at Merks (1997). There is, as yet, no convincing explanation to account for the selective forces that maintained this order in these control genes. A general conclusion of an enormous amount of work is that there is, in multicellular organisms, a relationship between the order of at least one family of genes controlling gene expression, and the plan of the organism itself. In short there is an "animalculus" in animals, similar to the "homunculus" that preformists thought to see at the origin of the development of Man (Danchin, 2003). This structure however, is not a minute animal, but a physical organisation of the programme that makes the animal. We do not have, at this point, any idea of the selective constraints that account for this relationship, or any idea of the reason why it can indeed control the development of the embryo from the egg to the adult animal. We wish however to point out here that this fits very well with the conjecture of von Neumann about computers that would make computers.

2.5 Celluloculus? Once admitted that the order of the genes making a multicellular organism is not random, it becomes tempting to ask the question in the case

Genome structures, operating systems and the image of the machine

203

of the cell itself. Biochemical studies of cells, usually based on "cell-free" experimental approaches, have the tendency to consider the cell as a tiny test-tube in which biological objects move more or less randomly. Considering the extremely tight level of packaging inside cells suggests that this is certainly not plausible, and not even an interesting simplification (Ellis, 2001). A cell, even a bacterial cell, must be a highly organized medium (Danchin, 2003; Danchin et al., 2000). This must be taken into account in any description of phenomena involving gene expression in the construction or duplication of a cell. A rapid observation of the position of orthologous genes in various genomes may give the impression that they can be located anywhere, and certainly not always at the same position in different genomes. However, it is clear that genes have a strong tendency to form clusters. This clustering is certainly not random: genes coding for related functions, such as those coding for the subunits of an enzyme for example, are generally located in a same transcription unit. Indeed this was the very basis of the concept of the operon. Another type of clustering also associated to a correlation between the architecture of the genome and that of the organism was found when lysogenic phages were described. There, genes are clustered into operons that correspond to specialized functions: replication and construction of the head and tail of the virus. We have seen that genes do not move very frequently, or, more exactly that most gene moves are incompatible with the long-term survival of the species. What we see as a move represents only what remains, under the constraint of long term survival, preserving usually a privileged orientation (and maybe privileged distances) with respect to the origin of replication. As a consequence, interestingly, one observes in bacteria a strong constraint in the distribution of the genes with respect to the origin and terminus of replication: genome rearrangements tend to be symmetrical around the origin of replication and thus do not change the replicating strand coding for the gene, even though they disrupt synteny (Tillier and Collins, 2000). Furthermore, although the laterally transferred genes (that contribute much for the apparent genome diversity) appear to be present almost anywhere, they are not randomly located. When comparing strains (for example comparing E. coli K12 and pathogenic E. coli), one remarks that the foreign genes differ, but that they are placed at the same position in the chromosome, often near tRNA genes (Blum-Oehler et al., 2000; Hou, 1999). This may just be because tRNAs allow recombination hot spots, but this is also compatible with the notion of a strong architectural constraint in the gene distribution, if tRNAs play a role in this organisation. This makes

204

Danchin and Noria

interesting to analyze the distribution of codons in protein coding genes, since this would be directly related to the translation process. Careful studies of the codon usage bias in genomes indeed point to the existence of strong constraints in the gene order along the chromosome. The genes within an operon have usually a similar codon usage bias. Interestingly, comparison between organisms where genes are clustered into an operon in one and dispersed in the other reveals that the codon usage bias is preserved (Rocha et al., 2000). The usage of a particular codon is coupled to a particular transfer RNA. It is therefore likely that repeating the same codon in a gene tends to maintain the cognate tRNA local concentration at a high level (in the vicinity of the ribosome that translates the messenger RNA). A large bias in the codon usage might therefore imply a local enrichment in some tRNAs, while in the same local environment the corresponding ribosome is depleted in other tRNAs. Another cause could also account for this observation: a gene transcribing a tRNA creates a source for that tRNA that would maintain a certain codon preference in its vicinity. In either case, the fact that genes distant in the chromosome share the same bias strongly suggests that the corresponding mRNAs are sitting next to each other and are translated from the same or very closely packed ribosomes. This indicates that there is a relation between the position of some ribosomes in the cell, and the position of some genes in the chromosome. The most remarkable observation that we could make in favour of this hypothesis was that the codon usage bias in orthologous genes in organisms as different as E. coli and B. subtilis shows a similar deviation of the bias with respect to its average value (different in both bacteria) for genes with orthologous functions, indicative of the existence of a common selection pressure in these quite different organisms (Danchin et al., 2000). We could not visualize any other cause except for an architectural selection pressure that would cause such a strong correlation. This should prompt further exploration of gene and gene product organization in bacteria. Of course, one needed to find reasons to account for the selective pressures that would tend to cluster genes together. We have already referred to nanomachines, such as the ribosome, RNA polymerase and ATP synthase: the corresponding genes are indeed clustered into well-organized operons. But one would like to see more. Gasses, which diffuse freely, play an important role in life, and not only at its origin. and naturally, but also and NO, CO or Maintaining and together, in the presence of metal ions is like having a lit match in a gas station. Hence the idea that there must be compartmentalization of at least some metabolic complexes, in particular those dealing with the sulphur atom, quite prone to

Genome structures, operating systems and the image of the machine

205

oxido-reduction (from -2 to +6). We looked for the distribution of sulphur related genes: they are without doubt clustered together into islands (Rocha et al., 2000). Much more must be performed in this direction, but this is a hint that some phenotypic characters, such as chemical reactivity, might exert sufficient pressure to get genes clustered together. Bacteria have a variety of shapes. For a long time this was ascribed to some unknown process building up the cell wall, the murein layer(s) in particular. Within this frame of thought the cell's interior was still understood as a simple bag of chemicals, with no specific organisation. Work by many groups however recently challenged this view, when the distribution of MreB and MreB-like proteins was analyzed using GFP fusions (Carballido-López and Errington, 2003; Egelman, 2003). It seems clear that for Bacteria of the Firmicute family, those that have a bacillus form, in contrast to a coccus form possess a kind of cytoskeleton that determines the shape (Carballido-López and Errington, 2003). There are however certainly a variety of means to construct rod-shaped bacteria (Daniel and Errington, 2003), but it is most likely that this corresponds to a highly organized substratum (Errington, 2003; Errington et al., 2003).

3.

Conclusions

Many studies are now demonstrating that many gene products are compartmentalized in Bacteria, whether this is related to the organization of the genome is still open to question, but the many new genome sequences recently uncovered, rather than challenging this conjecture, make it all the more intriguing. Furthermore clustering appears not to be restricted to Bacteria but to extend to Eukarya as well (Blumenthal and Gleason, 2003; Lee and Sonnhammer, 2003). Future studies will tell us whether it was worth exploring.

Acknowledgements We would like to thank all the colleagues who have throughout the years contributed to shape the content of the conjecture presented here. Their contribution is indicated as that of Stanislas Noria, who is the collective Team who creates a seminar in conceptual biology maintained at the Institut Pasteur in Paris, and, for some time, at the Department of Mathematics of the University of Hong Kong.

206

Danchin and Noria

References Akam, M. (1998). Hox genes, homeosis and the evolution of segment identity: no need for hopeless monsters. Int. J. Dev. Biol. 42, 445-451. Averof, M. (1997). Arthropod evolution: same Hox genes, different body plans. Curr. Biol. 7, R634-636. Averof, M., and Akam, M. (1995). Hox genes and the diversification of insect and crustacean body plans. Nature 376, 420-423. Bachiller, D., Macias, A., Duboule, D., and Morata, G. (1994). Conservation of a functional hierarchy between mammalian and insect Hox/HOM genes. EMBO J. 13, 1930-1941. Balavoine, G., and Telford, M, J. (1995). Identification of planarian homeobox sequences indicates the antiquity of most Hox/homeotic gene subclasses. Proc. Natl. Acad. Sci. USA 92, 7227-7231. Bennett, C. H. (1988). Logical Depth and Physical Complexity. In Herken, R. (ed.), The Universal Turing Machine: a Half-Century Survey. Oxford University Press, Oxford. Bhattacharyya, A., Stilwagen, S., Reznik, G., Feil, H., Feil, W. S., Anderson, I., Bernal, A., D'Souza, M., Ivanova, N., Kapatral, V., Larsen, N., Los, T., Lykidis, A., Selkov, E., Jr., Walunas, T. L., Purcell, A., Edwards, R. A., Hawkins, T., Haselkorn, R., Overbeek, R., Kyrpides, N. C., and Predki, P. F. (2002). Draft sequencing and comparative genomics of Xylella fastidiosa strains reveal novel biological insights. Genome Res. 12, 1556-1563. Blum-Oehler, G., Dobrindt, U., Janke, B., Nagy, G., Piechaczek, K., and Hacker, J. (2000). Pathogenicity islands of uropathogenic E. coli and evolution of virulence. Adv. Exp. Med. Biol. 485, 25-32. Blumenthal, T., and Gleason, K. S. (2003). Caenorhabditis elegans operons: form and function. Nat. Rev. Genet. 4, 112-120. Bozinovski, S., Jovancevski, G., and Bozinovska, N. (2001). DNA as a Real Time, Database Operating System. Presented at the International Conference on Information Systems, Analysis and Synthesis, La Habana, Cuba. Brock, H. W., and van Lohuizen, M. (2001). The Polycomb group--no longer an exclusive club? Curr. Opin. Genet. Dev. 11, 175-181. Brunder, W., and Karch, H. (2000). Genome plasticity in Enterobacteriaceae. Int. J. Med. Microbiol. 290, 153-165. Carballido-Lopez, R., and Errington, J. (2003). The bacterial cytoskeleton: in vivo dynamics of the actin-like protein Mbl of Bacillus subtilis. Dev. Cell 4, 19-28. Cobourne, M. T. (2000). Construction for the modern head: current concepts in craniofacial development. J. Orthod. 27, 307-314. Danchin, A. (2003). The Delphic Boat. Harvard University Press, Cambridge (Mass). Danchin, A., Guerdoux-Jamet, P., Moszer, I., and Nitschke, P. (2000). Mapping the bacterial cell architecture into the chromosome. Philos. Trans. R. Soc. Lond. B Biol. Sci. 355, 179190. Danchin, A., and Henaut, A. (1997). The map of the cell is in the chromosome. Curr. Opin. Genet. Dev. 7, 852-854. Daniel, R. A., and Errington, J. (2003). Control of cell morphogenesis in bacteria: two distinct ways to make a rod-shaped cell. Cell 113, 767-776. Dobrindt, U., and Hacker, J. (2001). Whole genome plasticity in pathogenic bacteria. Curr. Opin. Microbiol. 4, 550-557. Duboule, D., and Dolle, P. (1989). The structural and functional organization of the murine HOX gene family resembles that of Drosophila homeotic genes. EMBO J. 8, 1497-1505. Egelman, E. H. (2003). Cell walls, cell shape, and bacterial actin homologs. Dev. Cell 5, 4-5.

Genome structures, operating systems and the image of the machine

207

Ellis, R. J. (2001). Macromolecular crowding: obvious but underappreciated. Trends Biochem. Sci. 26, 597-604. Ellwood, M., and Nomura, M. (1982). Chromosomal locations of the genes for rRNA in Escherichia coli K-12. J. Bacteriol. 149, 458-468. Errington, J. (2003). Dynamic proteins and a cytoskeleton in bacteria. Nat. Cell Biol. 5, 178178. Errington, J., Daniel, R. A., and Scheffers, D. J. (2003). Cytokinesis in bacteria. Microbiol. Mol. Biol. Rev. 67, 52-65. Gaunt, S. J. (1991). Expression patterns of mouse Hox genes: clues to an understanding of developmental and evolutionary strategies. Bioessays 13, 505-513. Gehring, W. J. (1987). Homeo boxes in the study of development. Science 236, 1245-1252. Gupta, R. S. (1998). Protein phylogenies and signature sequences: A reappraisal of evolutionary relationships among archaebacteria, eubacteria, and eukaryotes. Microbiol. Mol. Biol. Rev. 62, 1435-1491. Hofstadter, D. (1979). Gödel, Escher, Bach: an Eternal Golden Braid. Basic Books, New York. Holland, P. W. (1999). Gene duplication: past, present and future. Semin. Cell Dev. Biol. 10, 541-547. Hou, Y. M. (1999). Transfer RNAs and pathogenicity islands. Trends Biochem. Sci. 24, 295298. Jagla, K., Bellard, M., and Frasch, M. (2001). A cluster of Drosophila homeobox genes involved in mesoderm differentiation programs. Bioessays 23, 125-133. Judd, E. M., Ryan, K. R., Moerner, W. E., Shapiro, L., and McAdams, H. H. (2003). Fluorescence bleaching reveals asymmetric compartment formation prior to cell division in Caulobacter. Proc. Natl. Acad. Sci. USA 100, 8235-8240. Kammermeier, L., and Reichert, H. (2001). Common developmental genetic mechanisms for patterning invertebrate and vertebrate brains. Brain Res. Bull. 55, 675-682. Kaufman, T. C., Seeger, M. A., and Olsen, G. (1990). Molecular and genetic organization of the antennapedia gene complex of Drosophila melanogaster. Adv. Genet. 27, 309-362. Kobayashi, K., Ehrlich, S. D., Albertini, A., Amati, G., Andersen, K. K., Amaud, M., et al. (2003). Essential Bacillus subtilis genes. Proc. Natl. Acad. Sci. USA 100, 4678-4683. Kramer, A., Neben, K., and Ho, A. D. (2002). Centrosome replication, genomic instability and cancer. Leukemia 16, 767-775. Lee, J. M., and Sonnhammer, E. L. (2003). Genomic gene clustering analysis of pathways in eukaryotes. Genome Res. 13, 875-882. Lobry, J. R. (1996). Asymmetric substitution patterns in the two DNA strands of bacteria. Mol. Biol. Evol. 13, 660-665. Lohmann, J. U., and Weigel, D. (2002). Building beauty: the genetic control of floral patterning. Dev. Cell 2,135-142. Mayr, E. (1998). Two empires or three? Proc. Natl. Acad. Sci. USA 95, 9720-9723. McLean, M. J., Wolfe, K. H., and Devine, K. M. (1998). Base composition skews, replication orientation, and gene orientation in 12 prokaryote genomes. J. Mol. Evol. 47, 691-696. Medigue, C., Rouxel, T., Vigier, P., Henaut, A., and Danchin, A. (1991). Evidence for horizontal gene transfer in Escherichia coli speciation. J. Mol. Biol. 222, 851-856. Merks, R. (1997). The molecular Bauplan.. http://carol.wins.uva.nl/~roel/bauplan/. Reichert, H. 2002. Conserved genetic mechanisms for embryonic brain patterning. Int. J. Dev. Biol. 46, 81-87. Rice, L. M., and Agard, D. A. (2002). Centriole duplication: centrin in on answers? Curr. Biol. 12, R618-619.

208

Danchin and Noria

Rocha, E. (2002). Is there a role for replication fork asymmetry in the distribution of genes in bacterial genomes? Trends Microbiol. 10, 393-395. Rocha, E. P., and Danchin, A. (2003). Essentiality, not expressiveness, drives gene-strand bias in bacteria. Nat. Genet. 34, 377-378. Rocha, E. P., and Danchin, A. (2001). Ongoing evolution of strand composition in bacterial genomes. Mol. Biol. Evol. 18, 1789-1799. Rocha, E. P., Danchin, A., and Viari, A. (1999). Analysis of long repeats in bacterial genomes reveals alternative evolutionary mechanisms in Bacillus subtilis and other competent prokaryotes. Mol. Biol. Evol. 16,1219-1230. Rocha, E. P., Fralick, J., Vediyappan, G., Danchin, A., and Norris, V. (2003). A strandspecific model for chromosome segregation in bacteria. Mol. Microbiol. 49, 895-903. Rocha, E. P., Guerdoux-Jamet, P., Moszer, I., Viari, A., and Danchin, A. (2000). Implication of gene distribution in the bacterial chromosome for the bacterial cell factory. J. Biotechnol. 78, 209-219. Rocha, E. P., Sekowska, A., and Danchin, A. (2000). Sulphur islands in the Escherichia coli genome: markers of the cell's architecture? FEBS Lett. 476, 8-11. Silberschatz, A., Galvin, P. B., and Gagne, G. (2001). Operating System Concepts, 6th Edition ed. John Wiley & Sons, New York. Stock, D. W. (2001). The genetic basis of modularity in the development and evolution of the vertebrate dentition. Philos. Trans. R. Soc. Lond. B Biol. Sci. 356, 1633-1653. Tillier, E. R., and Collins, R. A. (2000). The contributions of replication orientation, gene direction, and signal sequences to base-composition asymmetries in bacterial genomes. J. Mol. Evol. 50, 249-257. von Neumann, J. 1958 (reed 1979 ). The Computer and the Brain. Yale University Press, New Haven. Wilmut, I., Schnieke, A. E., McWhir, J., Kind, A. J., and Campbell, K. H. (1997). Viable offspring derived from fetal and adult mammalian cells. Nature 385, 810-813. Woese, C. R. (1998). Default taxonomy: Ernst Mayr's view of the microbial world. Proc. Natl. Acad. Sci. USA 95, 11043-11046. Wong, C., and Steams, T. (2003). Centrosome number is controlled by a centrosome-intrinsic block to reduplication. Nat. Cell Biol. 5, 539-544. Yockey, H. P. (1992). Information theory and molecular biology. Cambridge University Press, Cambridge (UK).

Chapter 11 Gene order in Prokaryotes: conservation and implications

MANUEL J. GÓMEZ, ILDEFONSO CASES and ALFONSO VALENCIA* Protein Design Group, Centre Nacional de Biotecnología, CSIC. Campus Universidad Autónoma de Madrid, Cantoblanco. Madrid 28049 Spain *corresponding author

1.

Summary

Genes in Prokaryotes are often organised in operons, groups of contiguous genes that function as single transcription units, or clusters, groups of contiguous genes subject to complex regulation that code for several transcripts. Several models suggest that the grouping of genes in operons or clusters provides physiological and genetic advantages that positively select their formation and maintenance. However, gene order along the chromosome is an evolutionary trait that is lost relatively quickly, since frequent chromosomal reorganisations and acquisition of foreign DNA shuffle the genetic material. As result, operons are generally conserved only among closely related species and widely conserved operons are scarce, although gene neighbourhood may be a more conserved property. Interestingly, the conservation of operons, gene clusters or neighbourhoods can be used as indicator of functional relations between gene products.

Molecules in Time and Space, Edited by Vicente et al. Kluwer Academic/Plenum Publishers, New York 2004

209

210

Gómez, Cases and Valencia

2.

Operons, clusters and regulons

2.1 Historical background and definitions The operon model was proposed, more than 40 years ago, by Jacob and Monod to explain the lac phenotype of Escherichia coli (Jacob et al., 1960; Jacob and Monod, 1961a; Jacob and Monod, 1961b; Jacob, 1966; Beckwith, 1996; Snyder and Champness, 1997). The original concept, referring to negatively regulated, inducible sets of cotranscribed genes, applied mainly to catabolic operons. The operon concept, however, had to be widened when positively regulated systems were discovered, in which the regulatory protein is a transcription activator, rather than a repressor (Snyder and Champness, 1997; Baumberg, 1999). Nowadays it is accepted that operons can be regulated by a variety of mechanisms. Transcription may be initiated at one or more promoters, and some of these promoters can be controlled by more than one regulatory protein. This is, in fact, the case of the E. coli lac operon, initially studied by Jacob and Monod, which is coregulated by CRP and LacI. In addition to a variety of mechanisms controlling transcription initiation, transcription attenuation mechanisms regulate transcription elongation and termination. For instance, anti-termination mechanisms abolish the function of transcription terminators, thus enabling the inclusion of additional sequences in the transcripts (Landick et al., 1996; Snyder and Champness, 1997; Baumberg, 1999; Henkin and Yanofsky, 2002). Therefore, the most complex operon architectures may include, in addition to the upstream promoter elements and downstream terminator sites, other promoters and terminators located within the group of consecutive genes. The activity of promoters and terminators may be regulated in such a way that not all genes in the group are cotranscribed in every situation, and a variety of alternative transcripts is possible. This makes the operon even harder to define precisely and, in some cases, the term gene cluster is often preferred over the term operon. The former has also a more general meaning, to denote a group of genes that is located in close proximity and that show some special property, as for example, to be relatively conserved among distinct species, or homogeneous in terms of the functional class of its genes, without making any assumption about their regulation. A notable example of gene cluster, as a group of genes that are to some extent coregulated, but also display a complex transcription pattern, is the dcw cluster (Vicente et al., 1998). In E. coli, the dcw cluster includes sixteen genes involved in cell division and peptidoglycan synthesis, all transcribed

Gene order in Prokaryotes: conservation and implications

211

in the same direction. The absence of transcription terminators within the cluster may allow the extension of the transcripts from the first promoter (Pz, located upstream of mraZ) to the last of the genes (envA) and, in fact, complementation tests suggest that transcripts extend at least to ftsW, in the central part of the cluster. The promoter Pz may be regulated by the LexA repressor. Therefore, the dcw cluster can be regarded as a bona fide operon. However, promoters for each of the genes in the cluster have been detected or predicted within the previous structural gene, and the gene ftsZ (located immediately upstream of envA) is transcribed from at least six promoters, located within the open reading frames of the previous three genes (ddlB, ftsQ and ftsA). In addition, the activity of several of the promoters of the cluster is known to be regulated by other mechanisms. Therefore, although the coregulation and coexpression of a number of the genes in the cluster may be required under certain circumstances, the presence of multiple control elements and the possibility to produce a variety of transcripts reflect the complex requirements of cell division control (see Chapters 12 and 13). Another extreme example in which it is impossible to define an operon is the set of transcription units formed by the E. coli cluster plsX-fabH- fabDfabG-acpP-fabF, which encodes acyl carrier protein and several fatty acid biosynthetic enzymes. All the genes are transcribed with the same orientation but the cluster contains multiple promoters. Each gene is cotranscribed with at least one other gene, and monocistronic transcripts are also produced (Zhang and Cronan, 1996).

2.2 The bacterial transcriptional regulatory network In contrast to the idea of operons, regulons comprise sets of genes located in separate positions in the genome that are regulated in the same way by the same transcription factor (Baumberg, 1999). This kind of proteins is frequent in bacterial genomes, and represents between 1% and 10% of their gene content. The amount of transcription factors in a given genome is influenced by two properties: the size of the genome, with larger genomes harbouring more transcription factors; and the species life-style, being almost absent in intracellular symbionts and abundant in free-living organisms (Cases et al., 2003). This fact that has been interpreted as the requirement of the latter to respond to and integrate a wider range of environmental signals. The variety of regulatory proteins seems to have arisen by two main mechanisms, protein domain recombination and gene duplication (Madam Babu and Teichmann, 2003). It has been estimated that there are around 350 transcription factors in E. coli (Pérez-Rueda and Collado-Vides, 2000). When 271 of them were analysed, only 11 different DNA binding domains were found. However, 90% of them exhibited a multi-domain structure.

212

Gómez, Cases and Valencia

While up to 46 different non-DNA binding domains were present in these proteins, they are assembled in 74 distinct domain architectures, reflecting extensive domain shuffling. At the same time, since proteins with the same sequential arrangement of domains are likely to be direct duplicates of each other (Apic et al., 2001, Bashton and Cothia, 2002) these data suggest that almost three-quarters of the E. coli transcription factors have arisen by gene duplication. While common regulation may be accomplished by direct binding of transcription factors to the promoter regions, it is also common the presence of more complex regulatory cascades, in which genes encoding transcription factors are in turn regulated by other transcription factors. In addition, promoter regions can be bound by more than one regulatory protein, configuring and intricate network of regulatory relations, in which regulons frequently overlap. This network has been only defined in detail, although not completely, in a few model organisms, among which E. coli is the one for which more information is available. This information has been compiled in RegulonDB (Salgado et al., 2001) and its analysis has revealed several interesting features: Regulon size, or the number of genes a given transcription factor regulates, is highly variable, and in E. coli ranges between 1 and almost 200, with an average of about 14. The regulon sizes follow a power-law distribution, with 85% of regulators controlling the expression of 15 or fewer genes. Similar numbers were obtained in an early analysis on a previous version of RegulonDB (Thieffry et al., 1998). Only 7 regulators (CRP, FNR, IHF, FIS, ArcA, NarL and Lrp) modulate the expression of more than 50 genes, which is consistent with the aforementioned power-law distribution, and together they account for the regulation of 51% of the E. coli genome. The strong influence exerted by these proteins has granted them the qualification of "Global Regulators" (Martinez-Antonio and Collado-Vides, 2003). An additional level of complexity comes form the fact that genes can be regulated by more than one transcription factor. In other words, regulons frequently overlap. Again the number of transcription factors per gene varies between 1 and 8, with an average of approximately 2. In this case the distribution is exponential, with almost 50% of the genes regulated only by one transcription factor. Interestingly, this distribution is also found in the yeast S. cerevisiae, although in the case of eukaryotes the average is significantly higher (Guelzim et al., 2002). A high degree of overlapping occurs among the regulons of the global regulators mentioned above, with many cases in which two or more of these act in concert to tune the expression of a particular gene (Martínez-Antonio and Collado-Vides,

Gene order in Prokaryotes: conservation and implications

213

2003), as it is the case of the stationary phase response (Shen-Orr et al., 2002). When the topology of the network configured by all this overlapping regulons is observed in detail, a number of repetitive structures can be found, called network motifs. These structures are small sub-networks composed of a few genes in which the connections are established in a particular manner. Interestingly, among all the possible network motifs, some appear in the transcription network more frequently than what would be expected, and in particular, the so-called "feed-forward loops", in which a transcription factor X, regulates both another transcription factor Y and a gene Z, with Y also regulating Z, are extremely abundant (Shen-Orr et al., 2002). The overrepresentation of feed-forward loops is not exclusive of the E. coli network, and can also be observed in S. cerevisiae (Milo et al., 2002), suggesting that the formation of these sub-structures has been positively selected both in prokaryotes and eukaryotes. Although it could be thought that enrichment in abundant network motifs is the product of duplications of complete modules, this does not seem to be the case. On the contrary, they may have appeared by convergent evolution. When Conant and Wagner (Conant and Wagner, 2003) looked in E. coli for pairs of motifs in which all the components were homologous among them, they could not find any, suggesting that each of them was assembled independently along evolution. It is believed that the substrate of this selection lays on the kinetics properties of these motifs. Cascades can act as filters for noise or transient fluctuations of signals, while feed-forward loops can increase the response of the system when the signal is turned off (Shen-Orr et al., 2002).

2.3 Distribution of gene functional classes in operons and clusters Operons were first considered groups of adjacent genes, transcribed into a single polycistronic mRNA, under the control of a contiguous operator, coding for products that participate in a common metabolic pathway or function (Demerec and Hartman, 1959). This principle applied to the lac operon, which includes the genes lacZ, lacY and lacA, encoding galactoside permease and thiogalactoside transacetylase, respectively. It also applied to many other examples of catabolic and biosynthetic operons, negatively or positively regulated. However, it was later recognised that genes within operons often participate in different pathways. As a consequence, the concept of operon that remains today and that is taught in textbooks fits better with the structural definition of a polycistronic unit and its associated regulatory elements.

214

Gómez, Cases and Valencia

The original assumption that genes in operons tend to be involved in similar functions, however, is still valid although it has been difficult to test in a systematic fashion. First, decades had to pass since operons were first described, until the complete sequence of a number of microbial genomes became available. Second, the absence, or scarcity, of experimental data regarding the organisation of transcription units, especially for organisms other than E. coli or B. subtilis, still precluded the study of operons. Researchers thus have resorted to studying the patterns of gene clustering in chromosomes, in terms of their functional classification. Therefore, the question to address was: do genes with related function cluster in prokaryotic genomes? This approach has the advantage that it does not require the definition of transcription units, while it focuses in the idea of gene cluster, a group of genes that are physically close in the chromosome, which is a relevant concept in the context of operon evolution models, as discussed later. The issue of deciding whether two neighbour genes have related functions is not trivial. For some genomes, such as the E. coli genome, a functional classification was established at the time of its annotation (Riley 1997). That scheme considers five functional superclasses (Energy, Information, Communication, Regulation and Transport) divided into 20 additional specific classes, and can be used to address whether two genes have related functions or not. In other cases, functionally related genes can be identified by the similarity of keywords in their SwissProt annotations, or because they are part of the same metabolic pathway as defined in the KEGG database (Kanehisha et al., 2002). Regardless of the approach used to define functional classes and functional similarity, the common facts are that functional descriptions tend to be of a general character and are subjective, by definition. The use of a more universal and comprehensive functional scheme, such as that proposed by the Gene Ontology consortium (Blake and Harris, 2003) should facilitate functional comparisons. One of the first studies that used complete genomic sequences to address whether functionally related genes are clustered in microbial chromosomes was focused on Haemophilus influenzae and E. coli. The functional classification of genes was derived from the class scheme proposed by M. Riley (Riley, 1997) and the analysis of relative gene locations was performed at the level of consecutive pairs of genes along the chromosome. Both in E. coli and in H. influenzae, functionally related genes tend to be neighbours more often than unrelated genes do. Although this trend is particularly strong for genes of the "Metabolism" and "Transport" classes, it is true for all functional classes (Tamames et al., 1997). This conclusion was later confirmed and expanded by studies about the distribution of gene functional classes in E. coli operons (Salgado et al.,

Gene order in Prokaryotes: conservation and implications

215

2000; de Daruvar et al., 2002). Operons for the study carried out by de Daruvar and collaborators were obtained from RegulonDB, and the functional classification scheme used was that proposed by M. Riley. A strong tendency for operons in E. coli to include genes of a single functional class was observed, and a number of functional classes that contained a significant number of operons with four or five genes of the same functional class, such as the classes "energy metabolism", "amino acid biosynthesis and metabolism" and "biosynthesis of cofactors", were identified. Interestingly, when functional superclasses were considered (this is, less detailed functional descriptions were used), at least 85% of the genes in operons of the Energy, Information, Communication and Transport superclasses were classified as belonging to the same superclass. This suggests that genes in operons that could have apparently unrelated functions, require coregulation because they are functionally related at a higher level of complexity. For example the operon containing the genes rpsU, dnaG and rpoD, encodes ribosomal protein S21, DNA primase and which are essential for the general processes of translation, replication and transcription, respectively (Burton et al., 1983). The fact that they are part of the same operon reflects the need to coordinate these three processes at a very basic level, which would correspond to the "Information" superclass, as defined by M. Riley (Riley, 1997). The universality of the tendency of operons to include genes with related function is supported by a study in which operons were predicted for 50 microbial genomes. The predicted operons were found to contain a high proportion of functionally related genes (Moreno-Hagelsieb and Collado Vides, 2002).

2.4 Prediction of operons Operons are believed to be common in bacteria. However, most experimental evidence about polycistronic transcription units comes from just a few species. For this reason, most methods implemented for the prediction of operons are developed and tested using E. coli data, and their validity for other organisms is, some times, difficult to assess. Several operon features have been used to implement operon prediction methods. The basic approach involves identifying runs, or groups of collinear genes that are contiguous in the chromosome. Obviously, not all runs correspond to real operons. However, a method to predict the gross number of operons in a genome and their size distribution in terms of number of genes, based on the fact that operon genes are transcribed in the same direction, was implemented by J. L. Cherry. Although it cannot be applied to genomes for which the direction of transcription is highly biased

216

Gómez, Cases and Valencia

toward the direction of replication, it predicts a number of operons in E. coli that is close to that obtained with other methods, and it has been used to analyse various microbial genomes (Cherry, 2003). Yada and collaborators used Hidden Markov Models to model and predict E. coli operons using structural elements such as open reading frames, promoters and terminators. The method showed an accuracy of about 60%. The system, however, cannot cope with the presence of promoters and terminators internal to operon structures. Moreover, organism-specific models would be required for operon prediction in distinct species, given that regulatory elements are usually poorly conserved (Yada et al., 1999). Salgado and collaborators used experimentally defined E. coli transcription units from RegulonDB to study the distribution of intergenic distances between genes in operons and the distribution of functional classes of genes in operons. The intergenic distance between genes in operons is short and within a clearly defined range, and genes in operons tend to have related functions, as discussed in the previous section. This information was used to develop a method to predict operons. The method identifies pairs of adjacent genes in the same operon with 88% accuracy and correctly identifies 75% of the known E. coli transcription units (Salgado et al., 2000). Later, the method was successfully tested with a set of known B. subtilis operons and, since the tendency to keep short intergenic distances in operons was identified as a common feature of all bacterial genomes analysed, it was suggested that it probably works in most prokaryotes (Moreno-Hagelsieb and Collado -Vides, 2002). Craven and collaborators used machine-learning methods to combine sequence information with the functional annotation of genes and expression data derived from micro-array experiments, from which information regarding coregulation was extracted. The evaluation of the approach, using E. coli data, indicated that 78% of the known operons could be identified, with a false positive rate of 10%. However, this method is only applicable to well studied organisms, such as E. coli, for which information from sequencing and functional genomics projects is available (Craven et al., 2000; Bockhorst et al., 2003a; Bockhorst et al., 2003b). Finally, other methods make use of evolutionary information, by focusing on the identification of conserved runs, i.e., collinear gene clusters in which gene order and orientation are conserved in two or more genomes (Overbeek et al., 1999). As it is discussed later, gene order is evolutionarily unstable and its maintenance is restricted to the conservation of operon structures, mainly between closely related species. Since conserved runs are not necessarily operons, a statistical method was later implemented to assign to each conserved run the probability that the run is an operon (Ermolaeva et

Gene order in Prokaryotes: conservation and implications

217

al., 2001). The specificity of the method was estimated in 30-50%, by comparing the predictions for E. coli against the collection of known operons in RegulonDB.

2.5 Operon and Regulon Databases RegulonDB is the most complete collection of information on the organisation of E. coli transcription units (Huerta et al., 1998; Salgado et al., 2001). Its current version (v4.0, October 2003) includes data about 4408 genes, 750 known and 2326 predicted transcription units, known and predicted promoters and other regulatory sequences, regulatory proteins and protein complexes, and the influence of growth conditions on gene expression. EcoCyc is another E. coli specific database that provides information about metabolic pathways and genome organisation (Karp et al., 2002). The current release (v7.5) includes information about 4393 genes and 810 transcription units that, according to this database definition, include operons and individually transcribed genes. EcoReg is a database focused on modelling E. coli transcriptional regulation and is being developed as a complement to EcoCyc. To our knowledge, there is no public database about experimentally or predicted B. subtilis operons with a systematic scope similar to that of RegulonDB. However, a collection of about 100 operons of B. subtilis compiled from the literature is available (Itoh et al., 1999). In addition, a database with information about B. subtilis transcription units as determined in Northern blot experiments, BSORF, is accessible through the WWW, and describes 467 transcription units, 266 monocistronic and 201 polycistronic, that include 565 open reading frames (Okuda et al., 2002).

3.

Gene order evolution

3.1 Orthology and paralogy To compare the organisation of genomes it is first required to define which genes are equivalent in them. Equivalent genes, in this context, are those that are related by orthology. Two genes are defined as orthologous if they belong to genomes of distinct species and they have independently evolved from a single gene in the last common ancestor of the two species. In contrast, paralogous genes are those that have been generated, from a

218

Gómez, Cases and Valencia

common ancestral gene, by a process of duplication. Additional subtypes exist: two genes of a given genome are classified as in-paralogs or outparalogs, if the duplication event happened, respectively, after or before the speciation event for that genome (Sonnhammer and Koonin, 2002). Orthology and paralogy are two cases of homology and both are defined only with respect to the phylogeny of the genes. However, orthologous genes are assumed to have the same or similar functions, unlike paralogous ones. Duplicated genes, unless they differentiate by mutation to acquire new functions, are redundant and prone to inactivation, if there is no special selective pressure to keep both copies active. Therefore, the detection of orthology relations is the basis of homology-based function prediction methods. Since it is seldom possible to trace back the evolutionary history of genes, orthology relations are predicted on the basis of sequence similarity detection, at the protein level. Given two genomes and the complete collection of predicted gene products from both of them, algorithms such as BLAST, FASTA or PSI-BLAST are used to perform all-against-all comparisons. The sequences that are reciprocally most similar, and that fulfil certain additional requirements, are often referred to as Best Bidirectional Hits (BBHs) and are considered orthologous. When using PSI-BLAST, for example, a pairwise identity significance (E value) lower than 0.001 and an overlap of at least 60% in the alignments, have been used as additional requisites for two sequences that are the most reciprocally similar, to be considered BBH's (Huynen and Bork, 1998). The identification of orthologous proteins may be complicated by the presence of paralogs, multidomain proteins and several evolutionary processes, such as sequence divergence, non-orthologous gene displacement, gene loss and horizontal gene transfer (Huynen and Bork, 1998). A number of bioinformatics projects have dealt with the identification and maintenance of collections of orthologous proteins. These can be used as a starting point for genome organisation comparisons, but also as a tool in protein function prediction and a source of sequence alignments of orthologous proteins. Among them, the National Center for Biotechnology Information (NCBI) maintains the COG database (Clusters of Orthologous Groups of Proteins) (Tatusov et al., 1997; Tatusov et al., 2001), which currently includes 74.059 proteins from 43 organisms, grouped into 3307 COGs. The criteria for the inclusion of a new protein into a given preexisting COG, consists of the presumable relation of homology to two members of the COG that belong to distinct phylogenetic lineages, rather than considering just single linkage. Each COG is assumed to have evolved form an individual ancestral gene through a series of speciation and duplication events, and may include both orthologous and paralogous

Gene order in Prokaryotes: conservation and implications

219

proteins. The INPARANOID database provides automatically generated collections of orthologous gene pairs from two genomes. In addition, each orthologous group includes other similar sequences that are classified as inparalogs with a certain confidence value. The current version (v. 2.5, April 2003) includes more than 165.000 sequences from E. coli and six eukaryotic genomes (Remm et al., 2001). The Kyoto Encyclopaedia of Genes and Genomes (KEGG) is a database of genomes, genes, metabolic pathways and chemical reactions and compounds. The last version (March 2003) also includes a database of 84 orthology tables (Kanehisha et al., 2002). The Microbial Genome Database (MBGD) is a database for the comparative analysis of completely sequenced microbial genomes, and also allows the generation of orthologous gene cluster tables. To this end, similarities between all genes are precomputed and stored, in addition to functional annotations (Uchiyama, 2003).

3.2 Conservation of gene order, operon organisation, and gene neighbourhood Genome evolution has received considerable attention for a few decades. However, genome-wide comparisons at the sequence level and the identification, or prediction, of orthologous genes in a systematic fashion could be performed only when complete genome sequences became available. Several strategies have been used to detect gene order conservation between prokaryotic genomes: first, by studying the fraction of conserved pairs of neighbour genes (as, for example, in Dandekar et al., 1998; Tamames, 2001); second, by analysing the conservation of genes in runs (as, for example, in Overbeek et al., 1999); and third, by measuring the degree of conservation of known operons from a given organism, usually E. coli or B. subtilis, in other organisms (as, for example, in de Daruvar et al., 2002). Shortly after the first microbial genomic sequences were released, studies on the correspondence of orthologous gene locations between genomes indicated that orthologous genes appear almost randomly placed and that gene order is lost easily (Mushegian and Koonin, 1996; Watanabe et al., 1997). From the comparison of the first nine completely sequenced microbial genomes, a hierarchy of rates at which genomes evolve was established. Protein identity is the most conserved trait, followed by the composition of genomes in terms of genes (gene content), and then, by gene order along the chromosome. Finally, the sequence of intergenic regions was found to be the least conserved trait, suggesting that gene regulatory features evolve at the highest rate (Huynen and Bork, 1998). There is an almost linear relationship between the divergence of sequence and gene order degradation,

220

Gómez, Cases and Valencia

in such a way that gene order is considerably disrupted when the protein sequence identity shared by orthologous genes in two genomes is below 50% (Huynen and Bork, 1998; Suyama and Bork, 2001). A study about the conservation of gene order in prokaryotes, conducted with 35 genomes, confirmed that gene order is lost easily during evolution, but well preserved at close phylogenetic distances (Tamames, 2001). Gene order is unstable not only at the genomic level, but also at the level of operon conservation (Watanabe et al., 1997). The percentage of operons that are identical between E. coli and H. influenzae, two organisms that have diverged relatively recently, is 56%. This percentage goes down to 13% when comparing E. coli and Helicobacter pylori. Thus, the destruction of operon structures is almost selectively neutral in long-term evolution (Itoh et al., 1999). The outcome of this lack of stability in operon structures is that there are very few widely conserved operons, and many operons that are conserved in a few genomes (Ermolaeva et al., 2002). Classical examples of operon structures that are partially conserved, even in divergent species, are the dcw cluster and the so-called S10 region, which includes operons S10, spc and containing genes involved in translation (Watanabe et al., 1997; Tamames, 2001). Although gene order along the chromosome is lost rapidly, and operon structures are unstable in long-term evolution, another conservation mode, known as gene neighbourhood conservation, has been described (Lathe et al., 2000). Gene neighbourhood conservation refers to the fact that, in some cases, chromosomal rearrangements, although they alter gene order, preserve the neighbourhood of the genes in terms of the functional class and regulatory features of the neighbouring genes. The set of genes that is conserved at this level of organisation is called an uber-operon (Lathe et al., 2000). As an extension of the uber-operon concept, that of connected gene neighbourhoods has been developed (Rogozin et al. 2002). Connected gene neighbourhoods are identified first, by locating pairs of contiguous genes that are conserved among several genomes, then, by assembling the gene pairs to generate arrays of conserved gene pairs and by finally merging the arrays to construct clusters of connected genes. If connected gene neighbourhoods are represented as graphs, nodes represent genes, and edges represent proximity in the chromosome of the connected genes. Two connected nodes in the graph would represent a gene pair that is conserved in a certain number of genomes. A connected gene neighbourhood is a summary of the possible architectures that a group of genes may have in different genomes. Most connected gene neighbourhoods happen to consist of genes that share a common function, sometimes described at low level of detail. The largest neighbourhood identified includes 79 genes and consists of overlapping, rearranged ribosomal protein operons (Rogozin et al. 2002).

Gene order in Prokaryotes: conservation and implications

221

The correlation between gene order degradation and sequence divergence indicates that gene order conservation can be used to measure evolutionary distances between species, and a web server (SHOT) for the construction of whole genome, gene order based phylogenies has been implemented (Korbel et al., 2002). Gene order based trees have also been constructed with restricted collections of genes, to establish whether their organisation reflects the evolutionary history of the species involved, or it is rather the result of some specific selective pressure. In the case of the already mentioned dcw cluster, gene order seems to correlate with the type of bacterial morphology. In bacilli, the dcw cluster is apparently more conserved and compact than in bacteria of other shapes, bearing a type of architecture that seems to be required for bacillar morphology. The molecular bases are not completely clear, though. Most likely the organisation of the dcw cluster in bacilli fulfils the requirement for coregulation of some of the genes, and also the need of clustering coadapted alleles and the possibility of promoting certain protein interactions (Tamames et al., 2001; Chapter 13).

3.3 Models about the evolutionary origin and maintenance of the operon organisation Three reasons can explain the observed conservation of operon structures in bacteria: first, recent divergence; second, recent horizontal gene transfer; and third, strong regulatory or structural constraints that select against the reorganisation of the conserved operons or clusters (Tamames, 2001). Several models have been proposed to explain how such constraints may direct the evolutionary origin and maintenance of operons (reviewed in Lawrence and Roth, 1996; Roth et al., 1996; Lawrence, 1997; Lawrence, 1999). This issue is highly controversial and all models have limitations and inconsistencies. It seems therefore reasonable to assume that various operons may have originated under distinct selective forces, and that natural selection most likely functions in different ways for the generation and maintenance of operons. The coregulation model is based on the fact that the expression of genes in operons is coordinated, and that this is required, or beneficial, for some phenotypes. While this type of selective force is very likely involved in the maintenance of operons, it seems unclear how the advantage of common regulation was first acquired, since coordinated expression would be achieved only once all genes and regulatory elements in the operon were properly placed. It would be highly unlikely that a number of rearrangements happened, each of them contributing to sequentially generate an operon, in the absence of positive selection that would be possible only once the operon structure had been organised.

222

Gómez, Cases and Valencia

The Natal model obviates this inconsistency by proposing that some gene clusters originated not by the grouping of pre-existing genes, but in situ, by gene duplication and differentiation by mutation. Then, coregulation, or other properties related with the organisation of genes in operons, would be advantageous enough to select against the separation of duplicated genes as consequence of chromosomal rearrangements. This may be the case of operons involved in metabolic pathways, in which several genes code for enzymes that accept related substrates, such as the histidine biosynthetic operon his. Although this model may explain the origin of some operons, or part of them, most bacterial operons are in fact composed of genes that show no obvious homology. Other models propose that the mere proximity of some genes in the chromosome may be advantageous, and this is the selective force that drives the generation of operons in a number of intermediate clustering steps. In this case, the future component genes would become increasingly closer, until they would fuse in a single transcription unit. These type of models can be divided into three categories: (1) those that consider the benefits of transcribing and translating the clustered genes in a restricted space within the cell volume; (2) those that point to the amplification of clustered genes as a primitive mechanism to achieve coregulation; and (3) those that take into account the effect of gene proximity on the frequency of genetic cotransfer and recombination. The molarity model, within the first category, suggests that gene clustering results, upon transcription and translation, in high local concentration of the products of the clustered genes. This may facilitate the participation of the gene products in series of pathway reactions, or accelerate the formation of protein complexes by allowing early protein interactions. The resulting increase in fitness would be a selective force for both operon generation and maintenance. In addition, protein interactions may be required to protect unstable proteins or to guide protein folding. In this regard, it has been suggested that some proteins fold in an interdependent fashion, referred to as cotranslational folding (Thanaraj and Argos, 1996). As a variation, the thermophilic model postulates that gene clusters and operons originated in ancestral thermophilic cells. Protein complex assembly would be necessary to stabilise thermolabile individual protein components or to quickly channel thermolabile compounds produced along a metabolic pathway. A possible example of this situation may be the trp operon, whose predicted structure in the common ancestor for all bacteria (trpEGDCFBA) is precisely the one found in Thermotoga maritima (Glansdorff, 1999). Within the second category, the amplification model assumes that if a block of functionally related genes is frequently amplified in response to

Gene order in Prokaryotes: conservation and implications

223

selection to increase gene dosage, clustering may be beneficial and positively selectable, because it would provide a simple mechanism for coregulation. Finally, the Fisher coadapted model, in the third category, is based on the concept of coevolution. Genes whose products are involved in the same metabolic pathway, especially if they are part of a protein complex, are supposed to evolve in a coordinated fashion. The physical proximity of these genes may be positively selected because it reduces the frequency of recombination events that separate the coadapted loci. On the other hand, the multigene repair model proposes that, in populations under weak selective pressure, mutations may accumulate in several non-essential genes that are involved in a pathway or that code for components of a complex. The pathway, or the complex, may be restored by horizontal gene transfer and changing selective pressure, but only if all components are transferred together. Another view of the last situation is proposed in the selfish operon model, which suggests that gene clustering is beneficial for the genes in an operon, but not necessarily for the organisms that host them. Genes in clusters can be transferred together both vertically and horizontally, while unclustered genes can be transferred together only in a vertical fashion. If the combination of component genes confers a positively selectable phenotype, the gene cluster or operon will propagate successfully. New genes may be sequentially recruited to the cluster, if the new combination is positively selectable. A well-known system that reflects this type of dynamics is that of pathogenicity islands (Hacker et al., 1997). The validity of the selfish operon model is supported by the finding that the contribution of horizontal gene transfer to prokaryotic evolution is far more relevant that what had been estimated in the past (Gogarten et al., 2002). About 18% of the current E. coli genome has been acquired by horizontal gene transfer since this species diverged from Salmonella 100 million years ago (Lawrence and Ochman, 1998). The main inconsistency of the selfish operon model is that it does not explain how clusters of essential genes originated, since positive selection is possible only if lack of function is not deleterious. And yet, there are multiple examples of large clusters or operons of essential genes, such as the clusters encoding ribosomal proteins or the (Lawrence and Roth, 1996). Moreover, the operons associated with essential functions in the Energy superclass show the highest degrees of functional homogeneity in terms of gene composition (de Daruvar et al., 2002). Thus, the operons of essential genes may have originated by mechanisms different from that proposed by the selfish operon model (Lawrence and Roth, 1996). The first possible explanation is that some of these clusters were assembled before the divergence of all known life, when natural selection may have operated in different ways. The second

224

Gómez, Cases and Valencia

explanation applies to operons that code for proteins that must interact, such as ribosomal protein operons; in this case, the genes have probably coevolved and the Fisher model predicts that the clustering of coadapted alleles is positively selected.

4.

Gene order based predictions in the context of other experimental and theoretical methods for the prediction of interactions

4.1 Protein interactions The study of protein interactions has become a hot topic during the last years because several experimental and computational methods have been implemented that allow their detection at large scale. The discovery of large number of protein interactions is paving the path to the study of cells as systems, by facilitating the prediction of new metabolic and regulatory pathways. The characterisation of protein interactions is also a novel strategy for predicting protein function, because proteins that interact physically or functionally are assumed to have related functions (following the so called guilty by association principle). Since the different experimental and computational methods to predict or identify protein interactions yield predictions at a different level, the expressions "protein interaction", "functional interaction", "functional association" or "functional coupling" are often used to imply either physical interaction or indirect functional association between proteins (for example, involvement in the same cellular process). In some cases, the different types of interaction have been defined in the form of a hierarchical classification. For example, Huynen and collaborators considered seven classes of protein interaction: 1: direct physical interaction; 2: indirect physical interaction (for example, between proteins that are part of the same complex); 3: involvement in the same metabolic pathway; 4: involvement in the same non-metabolic pathway (for example, a regulatory pathway); 5: involvement in the same process; 6: interaction between proteins of which at least one is hypothetical; 7: absence of interaction between proteins with known function (Huynen et al., 2000b). Among the experimental methods to identify protein interactions at large scale, the application of the yeast two hybrid method (Y2H) and affinity purification of protein complexes (TAP) coupled with high-throughput mass spectrometry to identify their components (HMS-PCI), are providing information about physical interactions in several proteomes (Uetz et al.,

Gene order in Prokaryotes: conservation and implications

225

2000; Ito et al., 2000; Gavin et al., 2002; Ho et al., 2002). On the other hand, genetic approaches and mRNA coexpression provide information about functional interactions (Ge et al., 2001; Grigoriev 2001; Jansen et al., 2002; Van Noort et al., 2003).

4.2 Computational prediction of protein interactions Several computational methods for the prediction of protein interactions that are based on the analysis of sequences have been developed (Table 1). Three of them, known collectively as genomic-context methods, are based in the comparison of complete genomic sequences to identify local or global patterns of gene co-occurrence. The other two are based in the analysis of multiple sequence alignments to detect patterns of co-evolution in pairs of proteins, which are expected if two proteins interact physically or functionally and orthologs of them exist in several organisms (Huynen et al., 2000a; Valencia and Pazos, 2002).

One of the genomic-context methods for the prediction of protein interactions is based on the detection of gene order or gene neighbourhood conservation. The first studies about gene order conservation in complete microbial genomes revealed that genes in operons code for proteins that may interact physically (Mushegian and Koonin, 1996). Later, the examination of collections of conserved gene pairs and operons indicated that, for many of them, there was experimental evidence of physical interactions between gene products (Huynen and Bork, 1998; Dandekar et al., 1998). Accordingly, two of the models that propose mechanisms for operon generation and maintenance, the molarity model and the Fisher coadapted models, require that proteins coded by operons interact physically. Also, it is important to note that, (I) regardless of their degree of conservation, operons tend to be composed of genes that belong to the same functional class (see section 2.3); (II) a previously mentioned study that analysed the functional classification of neighbour pairs of genes in E. coli and H. influenzae, also described that pairs of functionally related genes tend to be more conserved than pairs of

226

Gómez, Cases and Valencia

unrelated genes (Tamames et al., 1997). Moreover, the comparison of the organisation of E. coli operons with the distribution of related genes in the B. subtilis genome indicated that there are clear examples of very conserved structures of homologous genes in both genomes. This tendency is particularly striking in the case of genes belonging to functional classes associated with energy processes, suggesting stronger evolutionary constraints perhaps linked to the maintenance of metabolic pathways (de Daruvar et al., 2002); and (III), even if gene order it is not conserved, it is possible sometimes to identify conserved neighbourhoods, which include genes of related function (section 3.2). Therefore, conserved operons and conserved gene neighbourhoods tend to encode proteins that either interact physically or have related functions. These evolutionary tendencies are the basis of several computational strategies that use gene order conservation to predict protein interactions (Dandekar et al., 1998; Overbeek et al., 1999; Yanai et al., 2002; Snel et al., 2002). Other genomic-context method to predict protein interactions is based on the detection of gene fusions (Enright et al., 1999; Marcotte et al., 1999; Yanai et al., 2001). In this case, two proteins, or protein domains, coded in different genes, are assumed to interact physically or, at least functionally, if in some species they are coded by a single gene, presumably originated by a gene fusion event. It has been shown that fusion events are particularly common in metabolic enzymes, probably because it is advantageous to facilitate the channelling of substrates (Tsoka and Ouzounis, 2000). Two additional studies support that gene fusion events may be positively selected along evolution. In the first, it was shown that gene fusion events have been more prevalent along evolution than gene fission events, probably because gene fusion allows for the physical coupling of functions that were already biologically coupled. Also, it was observed that there is a direct correlation between genome size and the number of genes that have been originated by fusion events (Snel et al., 2000). In the second study, the evolutionary history of a number genes originated by gene fusion was analysed. The findings indicated that gene fusions have often disseminated by horizontal gene transfer and that the evolution of gene fusions often involves an intermediate stage in which the future fusion components exist as contiguous and coregulated genes within operons (Yanai et al., 2002). The third type of genomic-context method is phylogenetic profiling. Two genes have the same phylogenetic profile if they are both present or absent in a number of genomes (Tatusov et al., 1997; Huynen and Bork, 1998; Pellegrini et al., 1999). The similarity of the phylogenetic profiles of two genes may indicate that the corresponding gene products need to be simultaneously present in order to perform a given function. However, a direct physical interaction between the proteins is not necessarily implied.

Gene order in Prokaryotes: conservation and implications

227

The main limitations of this approach are that it can only be applied to complete genomes (as only then it is possible to rule out the absence of a given gene), and that it cannot be used with the essential proteins that are common to most organisms. Finally, among the methods based on the detection of co-evolution patterns in pairs of proteins, the MirrorTree method is based on the observation that interacting protein pairs co-evolve. In such cases, the corresponding phylogenetic trees of the interacting proteins show a greater degree of similarity (symmetry) than the trees of non-interacting proteins (Pazos and Valencia, 2001). On the other hand, the In Silico 2 Hybrid method (I2H) is based on the search of pairs of positions in two proteins that show a correlated mutational behaviour (Pazos et al., 1997; Pazos and Valencia, 2002). Each of the three genomic-context methods has a bias for the detection of a given type of interaction, and it has been found that there is a correlation between the spatial proximity of genes in genomes and the type of interaction that it is possible to predict. Physical interactions can be predicted when genes occur fused or as conserved neighbours. On the other hand, co-occurrence in operons or in genomes allows less detailed predictions, as for example, the involvement in the same metabolic pathway (Huynen et al., 2000b). Several studies have focused in assessing the coverage and specificity of genomic-context methods in comparison to homology based approaches (detection of orthology) and experimental methods, in terms of their ability to predict protein function and physical interactions, respectively. The coverage of genomic-context methods has been steadily growing, as the sequences of more complete genomes have become available, and it was predicted that by the year 2003 the number of proteins in databases for which function could be predicted by genomic-context methods would be as high as the number of proteins for which function could be predicted by homology (Huynen et al., 2003). In any case, it is considered that while homology based predictions may indicate the molecular function of a given protein, genomic-context based predictions point to higher order functions, since they provide information about the pathway or process in which a given protein is involved. In this sense, both types of prediction are complementary (Huynen et al., 2000b). Also, a comparison of the performances of genomic-context and experimental methods to predict protein physical interactions was carried out on the yeast proteome, using manually curated catalogues of protein complexes as a trusted reference. The better combination of accuracy and coverage was obtained by TAP, followed by genomic-context methods, that showed higher accuracy and similar coverage than mRNA coexpression, and higher coverage and similar

228

Gómez, Cases and Valencia

accuracy than Y2H (Von Mering et al., 2002). It can be therefore concluded that genomic-context methods are a valuable resource for both protein function and protein physical interaction prediction.

Figure 1. Protein interactions predicted for E. coli Isocitrate lyase (ICL) at the ECID database (Juan et al., 2003). (A) The number of predicted protein interactions with ICL are shown for each of five methods: PP, phylogenetic profiles; IH, In Silico 2 Hybrid; MT, MirrorTree; GC, gene context (gene order or neighbourhood conservation); GF, Gene fusion. (B) Score distribution for GC predictions. When the allowed minimum score was set to zero, 3785 protein pairs were reported; 99.9% of them had scores below 0.4 and probably they correspond to false positives (not shown). Then, the minimum score was set to one, and only five predicted protein pairs were reported. (C) The highest score has a value of 4 and corresponds to a predicted interaction between Malate synthase A (MSA) and ICL. ICL and MSA are in fact both involved in the glyoxylate and dicarboxylate metabolism and appear in the same metabolic pathway at the KEGG database.

Gene order in Prokaryotes: conservation and implications

229

230

Gómez, Cases and Valencia

4.3 Predicted protein interaction databases Several public databases store information about predicted protein interactions. Among them, COG provides gene neighbourhood and phylogenetic profile information, based on the same database's information about orthologous groups of proteins, although potential interactions are not explicitly calculated (Tatusov et al., 2001). ALLFUSE is a database of predicted functional associations of proteins in complete genomes, based on the detection of gene fusions (Enright et al., 2001). PREDICTOME, on the other hand, provides interaction predictions from genomic-context methods and also from experimental approaches such as Y2H or mRNA coexpression (Mellor et al., 2002). STRING combines the three genomic-context methods and provides a unified score to the predicted interactions (Von Mering et al., 2003). ECID is a database devoted to the definition of protein interactions in E. colt (Juan et al., 2003). For that end, interaction predictions based on five computational methods (gene order conservation, gene fusion, phylogenetic profiles, MirrorTree and I2H) are separately calculated (an example is shown in Figure 1). The type of information offered by these databases is complementary to that stored in databases as DIP (Xenarios et al., 2002), BIND (Bader et al., 2003), MIPS (Mewes et al., 2002) and MINT (Zanzoni et al., 2002), which are devoted to experimentally identified protein interactions and protein complexes. Finally, INTACT is a project sponsored by the European Bioinformatics Institute that aims to define a standard for the representation and annotation of protein-protein interactions, and to develop a public database of experimentally identified and predicted interactions.

4.4 Protein interaction networks By combining predicted or identified pairwise protein interactions, interaction networks for whole proteomes may be constructed. The overlap between the networks derived from the different individual methods is relatively small, in such a way that their combination is largely additive, and a giant connected component that includes a large proportion of an organism's complete genetic complement is usually generated (Yanai and DeLisi, 2002). The analysis of the properties of such interaction networks promises to bring new insights in the field of system biology. For example, a protein interaction network constructed only from the co-occurrence of genes in operons showed scale-free properties, similarly to what has been described for other complex biological networks (Jeong et al., 2000; Jeong et al., 2001; Wolf et al., 2002). Also, by identifying highly connected regions it was possible to identify clusters of genes of homogeneous functional

Gene order in Prokaryotes: conservation and implications

231

composition that represent specific metabolic pathways and processes (Snel et al., 2002). These functional modules are probably similar to the connected gene neighbourhoods defined by Rogozin and collaborators (Rogozin et al., 2002).

References Apic, G., Gough, J. and Teichmann S. (2001) Domain combination in archeal, eubacterial, and eukaryotic proteomes. J.Mol. Biol. 310, 311-325. Bader, G. D., Betel, D. and Hogue, C.W. (2003) BIND: the Biomolecular Interaction Network Database. Nucleic Acids Res. 31, 248-250.

232

Gómez, Cases and Valencia

Bashton, M. and Cothia, C. (2002) The geometry of domain combinations in proteins. J. Mol. Biol. 315, 927-939. Baumberg, S. (ed) (1999) Prokaryotic gene expression. Oxford University Press, New York. Beckwith, J. (1996) The Operon: an Historical Account. In Escherichia coli and Salmonella Typhimurium. Cellular and Molecular Biology, 2nd ed. (Neidhardt F.C., Curtiss III R., Ingraham J.L., Lin E.C.C., Low K.B., Magasanik B,. Reznikoff W., Schaechter M., Umbarger H.E. and Riley M., Eds.), ASM Press, Washington D.C. Chap 78: pp 12271231. Blake, J. A. and Harris, M. (2003) The Gene Ontology Project: Structured vocabularies for molecular biology and their application to genome and expression analysis. In Current Protocols in Bioinformatic (A.D. Baxevanis, D.B. Davison, R. Page, G. Stormo and L. Stein, Eds.). Wiley and Sons, Inc., New York. Bockhorst, J., Craven, M., Page, D., Shavlik, J. and Glasner J. (2003a) A Bayesian network approach to operon prediction. Bioinformatics 19, 1227-1235. Bockhorst, J., Qiu, Y., Glasner, J., Liu, M., Blattner, F. and Craven, M. (2003b) Predicting bacterial transcription units using sequence and expression data. Bioinformatics 19(Suppl 1), I34-I43. Burton, Z. F., Gross, C. A., Watanabe, K. K. and Burgess, R.R. (1983) The operon that encodes the sigma subunit of RNA polymerase also encodes ribosomal protein S21 and DNA primase in E. coli K12. Cell 32, 335-349 . Cases, I., de Lorenzo, V. and Ouzounis, C.A. (2003) Transcription regulation and environmental adaptation in Bacteria. Trends Microbiol. 11, 248-253. Cherry, J. L. (2003) Genome size and operon content. J. Theor. Biol. 221, 401-410. Conant, G. C. and Wagner, A. (2003) Convergent evolution of gene circuits. Nat. Genet. 34, 264-266. Craven, M., Page, D., Shavlik, J., Bockhorst, J. and Glasner, J. (2000) A probabilistic learning approach to whole-genome operon prediction. Proc. Int. Conf. Intell. Syst. Mol. Biol. 8, 116-127. Dandekar, T., Snel, B., Huynen, M. and Bork P. (1998) Conservation of gene order: a fingerprint of proteins that physically interact. Trends Biochem. Sci. 23, 324-328 . de Daruvar, A., Collado-Vides, J. and Valencia, A. (2002) Analysis of the cellular functions of Escherichia coli operons and their conservation in Bacillus subtilis. J. Mol. Evol. 55, 211-221. Demerec, M. and Hartman, P. (1959) Annu. Rev. Microbiol. 13: 377-406. Enright, A. J., Iliopoulos, I., Kyrpides, N. C. and Ouzounis, C. A. (1999) Protein interaction maps for complete genomes based on gene fusion events. Nature. 402, 86-90. Enright, A. J. and Ouzounis, C.A. (2001) Functional associations of proteins in entire genomes via exhaustive detection of gene fusion. Genome Biology. 2, 341-347. Ermolaeva, M. D, White, O. and Salzberg, S. L. (2001) Prediction of operons in microbial genomes. Nucleic Acids Res. 29, 1216-1221. Gavin, A. C., Bösche, M., Krause, R., Grandi, P., Marzioch. M., et al. (2002) Functional organisation of the yeast proteome by systematic analysis of protein complexes. Nature. 415, 141-147. Ge, H., Liu, Z., Church, G. M. and Vidal, M. (2001) Correlation between transcriptome and interactome mapping data from Saccharomyces cerevisiae. Nat Genet. 29, 482-486. Glansdorff, N. (1999) On the origin of operons and their possible role in evolution toward thermophily. J. Mol. Evol. 49, 432-438. Gogarten, J. P., Doolittle, W.F. and Lawrence, J. G. (2002) Prokaryotic evolution in light of gene transfer. Mol. Biol. Evol. 19, 2226-2238.

Gene order in Prokaryotes: conservation and implications

233

Grigoriev, A. (2001) A relationship between gene expression and protein interactions on the proteome scale: analysis of the bacteriophage T7 and the yeast Saccharomyces cerevisiae. Nucleic Acids Res. 29, 3513-3519. Guelzim, N., Bottani, S., Bourgine, S., Képès, F. (2002) Topological and causal structure of the yeast transcriptional regulatory network. Nat. Genet. 31, 60-63. Hacker, J., Blum-Oehler, G., Muhldorfer, I. and Tschape, H. (1997) Pathogenicity islands of virulent bacteria: structure, function and impact on microbial evolution. Mol. Microbiol. 23, 1089-1097. Henkin, T. M. and Yanofsky, C. (2002) Regulation by transcription attenuation in bacteria: how RNA provides instructions for transcription termination/antitermination decisions. Bioessays 24, 700-707. Ho, Y., Gruhler, A., Heilbut, A., Bader, G.D., Moore, L. et al. (2002) Systematic identification of protein complexes in Saccharomyces cerevisiae by mass spectrometry. Nature. 415, 123-124. Huerta, A. M., Salgado, H., Thieffry, D. and Collado-Vides, J. (1998) RegulonDB: a database on transcriptional regulation in Escherichia coli. Nucleic Acids Res. 26, 55-59. Huynen, M., Snel, B., Lathe, W. and Bork, P. (2000a) Exploitation of gene context. Curr. Opin. Struct. Biol. 10, 366-370. Huynen, M., Snel, B., Lathe, W. and Bork, P. (2000b) Predicting protein function by genomic context: quantitative evaluation and qualitative inferences. Genome Res. 10, 1204-1210. Huynen, M. A. and Bork, P. (1998) Measuring genome evolution. Proc. Natl. Acad. Sci. USA. 95, 5849-5856. Huynen, M. A., Snel, B., von Mering, C. and Bork, P. (2003) Function prediction and protein networks. Curr. Opin. Cell. Biol. 15, 191-198. Itoh, T., Takemoto, K., Mori, H. and Gojobori, T. (1999) Evolutionary instability of operon structures disclosed by sequence comparisons of complete microbial genomes. Mol. Biol. Evol. 16, 332-346. Ito, T., Tashiro, K., Muta, S., Ozawa, R., Chiba, T., el al. (2000) Toward a protein-protein interaction map of the budding yeast: A comprehensive system to examine two-hybrid interactions in all possible combinations between the yeast proteins. Proc. Natl. Acad. Sci. USA. 97, 1143-1147. Jacob, F. (1966) Genetics of the Bacterial Cell. Science. 152, 1470-1478. Jacob, F., Perrin, D., Sánchez, C. and Monod, J. (1960) L'Operon: groupe des gènes à expression cordonée par un operateur. C. R. Seances Acad. Sci. 250, 1727-1729. Jacob, F. and Monod, J. (1961a) Genetic regulatory mechanisms in the synthesis of proteins. J. Mol. Biol. 3, 318-356. Jacob, F. and Monod, J. (1961b) On the regulation of gene activity. Cold Spring Harb. Symp. Quant. Biol. 26, 193-211. Jansen, R., Greenbaum, D. and Gerstein, M. (2002) Relating whole-genome expression data with protein-protein interactions. Genome Res. 12, 37-46. Jeong, H., Mason, S.P., Barabasi, A. L. and Oltvai, Z. N. (2001) Lethality and centrality in protein networks. Nature 411, 41-42. Jeong, H., Tombor, B., Albert, R., Oltvai, Z. N. and Barabasi, A. L. (2000) The large-scale organization of metabolic networks. Nature 407, 651-654. Juan, D., Devos, D., Pazos, F., Ouzounis, C., Blaschke, C. and Valencia, A. (2003) Reconstruction of the E. coli interactome. Small is still beautiful. Submitted for publication.

234

Gómez, Cases and Valencia

Kanehisha, M., Goto, S., Kavashima, S. and Nakaya, A. (2002) The KEGG databases at GenomeNet. Nucleic Acid Res. 30, 42-46. Karp, P. D., Riley, M., Saier, M., Paulsen, I. T., Collado-Vides, J., Paley, S. M., PellegriniToole, A., Bonavides, C., Gama-Castro, S. (2002) The EcoCyc Database. Nucleic Acids Res. 30, 56-58. Korbel, J. O., Snel, B., Huynen, M. A. and Bork, P. (2002) SHOT: a web server for the construction of genome phylogenies. Trends Genet. 18, 158-162. Landick, R., Turnbough, Jr. C. L. and Yanofsky, C, (1996) Transcription Attenuation. In Escherichia coli and Salmonella typhimurium. Cellular and Molecular Biology, 2nd ed. (Neidhardt, F. C., Curtiss III, R., Ingraham, J. L., Lin, E. C. C., Low, K. B., Magasanik B, Reznikoff, W., Schaechter, M., Umbarger, H. E. and Riley, M., Eds). ASM Press, Washington D.C. Chap 81: pp 1263-1286. Lathe, W. C. 3rd, Snel, B. and Bork, P. (2000) Gene context conservation of a higher order than operons. Trends Biochem. Sci. 25, 474-479. Lawrence, J. G. (1999) Selfish operons: the evolutionary impact of gene clustering in prokaryotes and eukaryotes. Curr. Opin. Genet. Dev. 9, 642-648. Lawrence, J. G. (1997) Selfish operons and speciation by gene transfer. Trends Microbiol. 5, 355-359. Lawrence, J. G. and Ochman H. (1998) Molecular archaeology of the Escherichia coli genome. Proc. Natl. Acad. Sci. U S A. 95, 9413-9417. Lawrence, J. G. and Roth, J. R. (1996) Selfish operons: horizontal transfer may drive the evolution of gene clusters. Genetics 143, 1843-1860 . Maas, W. K. (1964) Studies on the Mechanism of Repression of Arginine Biosynthesis in Escherichia coli II. Dominance of Repressibility in Diploids. J. Mol. Biol. 8, 365-370. Madan Babu, M. and Teichmann, S. (2003) Evolution of transcription factors and the gene regulatory network in Escherichia coli. Nucleic Acids Res. 31, 1234-1244. Marcotte, E. M., Pellegrini, M., Ho-Leung, N., Rice, D. W., Yeates, T. O. and Eisenberg, D. (1999) Detecting protein function and protein-protein interactions from genome sequences. Science. 285, 751-753. Martínez-Antonio, A. and Collado-Vides J. (2003) Identifying global regulators in transcriptional regulatory networks in bacteria. Curr. Opin. Microbiol. 6, 482-489. Mellor, J. C., Yanai, I., Clodfelter, K. H., Mintseris, J. and DeLisi, C. (2002) Predictome: a database of putative functional links between proteins. Nucleic Acids Res. 30, 306-309. Mewes, H. W., Frishman, D., Guldener, U., Mannhaupt, G., Mayer, K. et al. (2002) MIPS: a database for genomes and protein sequences. Nucleic Acids Res. 30, 31-34. Milo, R., Shen-Orr, S., Itzkovitz, S., Kashtan, N., Chklovskii, D. and Alon U. (2002) Networks motifs: simple building blocks of complex networks. Science. 298, 824-827. Moreno-Hagelsieb, G. and Collado-Vides J. (2002) A powerful non-homology method for the prediction of operons in prokaryotes. Bioinformatics. 18 (Suppl 1), S329-336. Mushegian, A. R. and Koonin, E. V. (1996) Gene order is not conserved in bacterial evolution. Trends. Genet. 12, 289-290. Okuda, S., Kawashima, S. and Kanehisa, M. (2002) Database of operons in Bacillus subtilis. Genome Informatics 13, 496-497. Overbeek, R., Fonstein, M., D'Souza, M., Pusch, G. D., and Maltsev N. (1999) The use of gene clusters to infer functional coupling. Proc. Natl. Acad. Sci. USA. 96, 2896-2901. Pazos, F., Helmer-Citterich, M., Ausiello, G. and Valencia, A. (1997) Correlated mutations contain information about protein-protein interaction. J. Mol. Biol. 271, 511-523. Pazos, F. and Valencia, A. (2001) Similarity of phylogenetic trees as indicator of proteinprotein interaction. Protein Eng. 14, 609-614.

Gene order in Prokaryotes: conservation and implications

235

Pazos, F. and Valencia A. (2002) In silico two-hybrid system for the selection of physically interacting protein pairs. Proteins. 47, 219-227. Pellegrini, M., Marcotte, E. M., Thompson, M. J., Eisenberg, D. and Yates, T. O. (1999) Assigning protein functions by comparative genome analysis: protein phylogenetic profiles. Proc. Natl. Acad Sci. USA 96, 4285-4288. Pérez-Rueda, E. and Collado-Vides, J. (2000) The repertoire of DNA-binding transcriptional regulators in Escherichia coli K-12. Nucleic Acids Res. 28, 1838-1847. Remm, M., Storm, C. E. V. and Sonnhammer, E. L. L. (2001) Automatic Clustering of Orthologs and In-paralogs from Pairwise Species Comparisons. J. Mol. Biol., 314, 10411052. Riley, M. (1997) Genes and proteins of Escherichia coli K-12 (GenProtEC). Nucleic Acids Res. 25, 51-52. Rogozin, I. B., Makarova, K. S., Murvai, J., Czabarka, E., Wolf, Y. I., Tatusov, R. L., Szekely, L. A. and Koonin, E. V. (2002) Connected gene neighborhoods in prokaryotic genomes. Nucleic Acids Res. 30, 2212-2223. Roth, J. R., Benson, N., Galitski, T., Haack, K., Lawrence, J. G. and Miesel, L. (1996) Rearrangements of the Bacterial Chromosome: Formation and Applications. In Escherichia coli and Salmonella typhimurium. Cellular and Molecular Biology, 2nd ed. (Neidhardt, F. C., Curtiss III, R., Ingraham, J. L., Lin, E. C. C., Low, K. B., Magasanik, B., Reznikoff, W., Schaechter, M., Umbarger, H. E. and Riley, M., Eds). ASM Press, Chap 120: pp 2256-2276. Salgado, H., Moreno-Hagelsieb, G., Smith, T. F. and Collado-Vides, J. (2000) Operons in Escherichia coli: Genomic analyses and predictions. Proc. Natl. Acad. Sci. USA. 97, 6652-6657. Salgado, H., Santos-Zavaleta, A., Gama-Castro, S., Millán-Zárate, D., Díaz-Peredo, E., Sanchez- Solano, F., Pérez-Rueda, E., Bonavides-Martínez, C., and Collado-Vides, J. (2001) RegulonDB (version 3.2): Transcriptional Regulation and Operon Organization in Escherichia coli K-12 Nucleic Acids. Res. 29, 72-74 . Shen-Orr, S. S., Milo, R., Mangan, S. and Alon, U. (2002) Network motifs in the transcriptional regulation network of Escherichia coli. Nat Genet. 31, 64-68. Snel, B., Bork, P. and Huynen, M. A. (2000) Genome evolution. Gene fusion versus gene fission. Trends Genet. 16, 9-11. Snel, B., Bork, P. and Huynen, M. A. (2002) The identification of functional modules from the genomic association of genes. Proc. Natl. Acad. Sci. USA 99, 5890-5895. Snyder, L. and Champness W. (1997) Regulation of Gene Expression. In Molecular Genetics of Bacteria. ASM Press, Washington D.C. Sonnhammer, E. L. L. and Koonin, E. V. (2002) Orthology, paralogy and proposed classification for paralog subtypes. Trends Genet. 18, 619-620. Suyama, M. and Bork, P. (2001) Evolution of prokaryotic gene order: genome rearrangements in closely related species. Trends Genet. 17, 10-13. Tamames, J., Casari, G., Ouzounis, C. and Valencia, A. (1997) Conserved clusters of functionally related genes in two bacterial genomes. J. Mol. Evol. 44, 66-73. Tamames, J., González-Moreno, M., Mingorance, J., Valencia, A. and Vicente, M. (2001) Bringing gene order into bacterial shape. Trends Genet. 17,124-126. Tamames, J. (2001) Evolution of gene order conservation in prokaryotes. Genome Biol. 2, R0020. Tatusov, R. L., Koonin, E. V. and Lipman, D. J. (1997) A genomic perspective on protein families. Science. 278, 631-637.

236

Gómez, Cases and Valencia

Tatusov, R. L., Natale, D. A., Garkavtsev, I. V., Tatusova, T. A., Shankavaram, U. T., Rao, B. S., Kiryutin, B., Galperin, M. Y., Fedorova, N. D. and Koonin, E. V. (2001) The COG database: new developments in phylogenetic classification of proteins from complete genomes. Nucleic Acid Res. 29, 22-28. Thanaraj, T. A. and Argos, P. (1996) Ribosome-mediated translational pause and protein domain organization. Protein Sci. 5, 1594-1612. Thieffry, D., Huerta, A. M., Pérez-Rueda, E. and Collado-Vides, J. (1998) From specific gene regulation to genomic networks: a global analysis of transcriptional regulation in E. coli. BioEssays 20, 433-440. Tsoka, S, and Ouzounis, C. A. (2000) Prediction of protein interactions: metabolic enzymes are frequently involved in gene fusion. Nature Genet. 26, 141-142. Uchiyama, I. (2002) MBGD: microbial genome database for comparative analysis. Nucleic Acid Res. 31, 58-62. Uetz, P., Giot, L., Cagney, G., Mansfield, T. A., Judson, R. S. et al. (2000) A comprehensive analysis of protein-protein interactions in Saccharomyces cerevisiae. Nature. 403, 601603. Valencia, A., Pazos, F. (2002) Computational methods for the prediction of protein interactions. Curr. Opin. Struct. Biol. 10, 368-373. van Noort, V., Snel, B. and Huynen, M.A. (2003) Predicting gene function by conserved coexpression. Trends Genet, 19, 238-242. Vicente, M., Gómez, M. J. and Ayala, A. (1998) Regulation of transcription of cell division genes in E. coli dcw cluster. Cell Mol. Life Sci. 54, 221-271. von Mering, C., Huynen, M., Jaeggi, D., Schmidt, S., Bork, P. and Snel, B. (2003) STRING: a database of predicted functional associations between proteins. Nucleic Acids Res. 31, 258-61. von Mering, C., Krause, R., Snel, B., Cornell, M., Oliver, S. G., Fields, S. and Bork, P. (2002) Comparative assessment of large-scale data sets of protein-protein interactions. Nature. 417, 399-403. Watanabe, H., Mori, H., Itoh, T., Gojobori, T. (1997) Genome plasticity as a paradigm of eubacteria evolution. J. Mol. Evol. 44 Suppl 1, S57-64. Wolf, Y. I., Karev, G. and Koonin, E.V. (2002) Scale-free networks in biology: new insights into the fundamentals of evolution? Bioessays 24, 105-109. Xenarios, I., Salwinski, L., Duan, X. J., Higney, P., Kim, S. and Eisenberg, D. (2002) DIP: The Database of Interacting Proteins. A research tool for studying cellular networks of protein interactions. Nucleic Acids Res. 30, 303-5. Yada, T., Nakao, M., Totoki, Y. and Nakai, K. (1999) Modeling and predicting transcriptional units of Escherichia coli genes using hidden Markov models. Bioinformatics 15, 987-993. Yanai, I. and DeLisi, C. (2002) The society of genes: networks of functional links between genes from comparative genomics. Genome Biol. 3, R0064. Yanai, I., Derti, A. and DeLisi, C. (2001) Genes linked by fusion events are generally of the same functional category: a systematic analysis of 30 microbial genomes. Proc. Natl. Acad. Sci. USA. 98, 7940-7945. Yanai, I., Mellor, J. C. and DeLisi, C. (2002) Identifying functional links between genes using conserved chromosomal proximity. Trends Genet.18, 176-179. Yanai, I., Wolf, Y. I. and Koonin, E. V. (2002) Evolution of gene fusions: horizontal transfer versus independent events. Genome Biol. 3, R0024. Zanzoni, A., Montecchi-Palazzi, L., Quondam, M., Ausiello, G., Helmer-Citterich, M. el al. (2002) MINT: a Molecular INTeraction database. FEBS Letters. 513, 135-140.

Gene order in Prokaryotes: conservation and implications

237

Zhang, Y. and Cronan, J. E. Jr. (1996) Polar allele duplication for transcriptional analysis of consecutive essential genes: application to a cluster of Escherichia coli fatty acid biosynthetic genes. J. Bacteriol. 178, 3614-3620.

This page intentionally left blank

Chapter 12 How similar cell division genes are located and behave in different bacteria

MIGUEL VICENTE*, JAVIER ÁLVAREZ and ROCÍO MARTÍNEZARTEAGA Centro National de Biotecnología (CSIC) Campus de Cantoblanco-UAM 28049-Madrid Spain. *corresponding

1.

author

Introduction

Present day bacteria are the survivors of countless individual cells in which mutations were selected along the time to produce a variety of genomes, shapes and metabolic pathways that now are adapted to specific environments. During the last decade the sequencing of the full genome of many bacterial species has allowed to know some interesting facts on the conservation of genes and groups of genes in the genomes of several bacteria. It is nevertheless obvious that genome sequences do not tell the full story of the cell. In some cases, such as the genes that are grouped in the dcw cluster, that comprise a substantial part of the genes that encode proteins involved in cell division and peptidoglycan synthesis, both the sequence of the genes and their relative order in the cluster are conserved in many bacteria, particularly in rod-shaped ones (Tamames et al., 2001). It is obvious that if the genomes are similar and the bacteria are not, there must be something that differs. One difference is the way in which the genome is expressed. We may have to wait for the completion of extensive studies on the expression of full bacterial genomes to gain a panoramic view

Molecules in Time and Space, Edited by Vicente et al. Kluwer Academic/Plenum Publishers, New York 2004

239

240

Vicente, Alvarez and Martínez-Arteaga

of the different regulatory webs that connect all the functions of a bacterial cell to its division process. Our present knowledge, although limited, tells us anyway that different bacteria possess regulatory mechanisms that finely tune the expression of the dcw cluster genes to some of their particular behaviours. In this chapter we will summarise the features of three cases, Escherichia coli, Bacillus subtilis and Neisseria gonorrhoeae, in which different regulatory patterns operate on very similar dcw clusters to adjust their expression in ways that we believe cater for the particular needs of each organism.

2.

Expression of genes in the Escherichia coli dcw cluster

The dcw cluster of E. coli (figure 1) may be a close direct descendant of the primitive structure of the cluster, the one that could be found in a hypothetical bacterial ancestor (Nikolaichik and Donachie, 2000). It maps at minute 2 of the genetic map and includes sixteen genes in a 17.9 kb fragment (Vicente et al., 1998). It is formed by four gene families: i- the mur genes (murE, F, D, G, C, mraY and ddlB), that code for the enzymes that synthesize the peptidoglycan precursors in the cytoplasm; ii- three genes that code for proteins that connect the cytoplasmic steps with the periplasm, i.e. ftsW that due to its membrane topology is assumed to translocate precursors from the cytoplasm to the perisplasm (Lara et al., 2002), ftsI that processes the precursors to produce septal peptidoglycan (Spratt et al., 1977) and ftsL that spans the region from the cytoplasmic membrane to the peptidoglycan (Ghigo et al., 2000); iii- the ftsQ, A and Z genes that encode septal proteins that, together with those from the preceding group, assemble sequentially into a division ring marking the cell centre before division, and in addition recruit into it the ZipA, FtsK, FtsB and FtsN division proteins that are encoded by genes that do not belong to the cluster; and iv- the mraZ and mraW genes whose role in septation is not fully determined (Lutkenhaus and Mukehrjee, 1996; Dewar and Dorazi, 2000).

Figure 1. The Escherichia coli dcw cluster. In dark grey the septation genes, in light grey the genes that code for enzymes involved in peptidoglycan precursor synthesis. The stem-loop structure indicates a transcriptional terminator. The white arrowhead marks the location of the ftsQ1p gearbox promoter.

How similar cell division genes are located and behave in different bacteria

241

Although the mur genes in the E. coli cluster do not follow exactly the same order as the biochemical steps catalysed by their products, their order can be theoretically derived from them by allowing for a few inversions and insertions. The murB gene is conspicuously absent from the E. coli dcw cluster, it maps at 89.88 min of the genetic map instead. A group of three genes, ftsZ, A, and Q, and three more separate ones, ftsW, I and L, are arranged relative to each other in nearly the inverse order in which their products seem to be incorporated into the division ring. In E. coli there are no transcriptional terminators in the dcw cluster and its genes are transcribed from several promoters that are located even in the structural regions of the preceding genes (Vicente et al., 1998). Transcription proceeds in the same direction in which the replication fork progresses, corresponding to genes that should be actively expressed. As many of the genes overlap or are separated by short gaps, this transcriptional arrangement yields a complex and not yet fully characterised set of transcripts. Transcription from mraZ1p (=Pmra), the promoter located most upstream in the cluster, is necessary to yield sufficient levels of all the proteins encoded down to the ftsW gene that already has a proximal promoter located immediately upstream (Hara et al., 1997). While up to 66% of the messengers that reach ftsZ, the gene located near the end of the cluster, are originated from distant promoters (de la Fuente et al., 2001; Flärdh et al., 1998), the contribution of all the promoters, the distal and the numerous proximal ones, may be required for an expression of ftsZ that satisfies the physiological needs of the cell. Transcription from the most proximal promoters to ftsZ, responsible for the remainder 34% of the transcripts that reach ftsZ, may be particularly intricate. Under conditions of undisturbed growth it proceeds from at least six promoters (Flärdh et al., 1997). The importance of this complex regulation was first shown using a strain in which the expression of ftsZ is dissociated from the upstream promoters in the chromosome, and placed under the control of an artificially inducible tac promoter (Palacios et al., 1996). In this case the levels of the FtsZ protein required by the engineered strain to sustain division conserving the same mean cell length as the wild type are 140% those needed by the wild type strain. Similarly, the substitution of mraZ1p by Plac causes a 30% reduction in the levels of the FtsZ protein in the absence of IPTG (Mengin-Lecreulx et al., 1998). While only one of the promoters in the dcw E. coli cluster, ftsQ1p, conforms to the definition of a gearbox (promoters that yield constant amounts of transcript per cell at any growth rate, and are preferentially recognised by RpoS) (Ballesteros et al., 1998), the ftsZ proximal promoters, although transcribed by the RpoD-containing holoenzyme, also show a similar yield (Dewar et al., 1989; Flärdh et al., 1997; Smith et al., 1993).

242

Vicente, Alvarez and Martínez-Arteaga

This may reflect the fact that the division proteins, and specifically FtsQ, A and Z are required only once in the life of a cell for the production of one single division ring. During exponential growth the expression from the quorum-sensing SdiA protein induces the ftsQ2p promoter, resulting in increased levels of ftsZ transcript. A 50 to 80% decrease in sdiA expression occurs under conditions in which conditioned medium (medium in which E. coli cells have previously been grown to saturation and removed) is used to grow an indicator strain containing an sdiA-lacZ reporter fusion. The down-regulation of is associated with a decrease in expression of the SdiA target promoter ftsQ2p (García-Lara et al., 1996). Transcription from ftsQ1p and 2p decreases when the levels of the ppGpp alarmone increase, but it is not clear how the change in the ppGpp levels modifies the expression of ftsZ (Powell and Court 1998; Navarro et al., 1998; Joseleau-Petit et al., 1999). The expression from ftsA1p is induced in vitro by RcsB, a protein that forms part of a two-component system involved in the activation of core lipopolysaccharide synthesis (Carballés et al., 1999). On the other hand the expression of ftsZ is reduced both in the presence of misfolded proteins (Lesley et al., 2002) and during osmotic shock (Weber et al., 2002). It is not likely that this summarised description is complete, and other regulatory pathways that modulate the expression of the dcw cluster are likely to be discovered in the future. How and why the cell uses such a complex signalling system for the expression of the genes in the dcw cluster is simply unknown. Considering that E. coli has been sometimes defined as the organism from which biologists know more, this should be a sobering statement.

3.

The ftsI and spoVE, duplicated genes in the dcw cluster of Bacillus subtilis

While E. coli cells are Gram-negative rods that inhabit the lower intestinal tract of vertebrates, a relatively sheltered and constant environment, B. subtilis cells, also rod shaped but Gram positive, live freely in open environments such as soils. B. subtilis cells have the ability to sporulate when the conditions become hostile. Sporulation is a form of cell differentiation that is initiated by an asymmetrical division of one cell to produce a pre-spore and a mother cell. Differentiation in B subtilis is a form of primitive bacterial apoptosis in which some material of the mother cell, that is destined to die, is used by the pre-spore that survives. Early during sporulation the spore becomes wrapped in peptidoglycan layers that are later integrated into a fairly resistant cover (Stragier and Losick 1996).

How similar cell division genes are located and behave in different bacteria

243

Despite their widely divergent phylogeny, the dcw cluster of B. subtilis (figure 2) is remarkably similar to the E. coli one (Henriques et al., 1992). There is nevertheless a conspicuous difference between the two: the B. subtilis cluster contains one internal transcription terminator that separates two genes, ftsI and spoVD, that encode two similar (33% identical over 641 residues) forms of the Penicillin-Binding Protein that participates in the synthesis of septal peptidoglycan. While ftsI encodes the protein that functions during vegetative growth, the spoVD copy is the gene that is expressed during sporulation and provides the activity to synthesize the peptidoglycan layers of the prespore.

Figure 2. The Bacillus subtilis dcw cluster. In black the spoVD, a sporulation specific gene presumably a duplication of the ftsI gene that codes for a penicillin-binding protein. The black arrow indicates a promoter that transcribes spoVD. Other symbols as in figure 1.

During vegetative growth and early stationary phase, spoVE, a gene that encodes a protein with an aminoacid sequence similar to the E. coli FtsW, is co transcribed with murD and murG in the form of very long polycistronic messages. After entering into stationary phase, the long polycistronic molecule is shut–off and switched to a monocistronic mode (Henriques et al., 1992).

4.

Additional coding sequences and inverted repeats that may alter transcription are present in the dcw cluster of coccal Neisseria

Given the absence of transcriptional terminators the whole dcw cluster may function in E. coli as a single transcriptional unit. We believe that this kind of transcriptionally compact dcw cluster is a trait that is associated to rod-shaped cells as E. coli and B. subtilis (Tamames et al., 2001). As the surface of a cell plays physiological roles similar to what lungs, stomach and kidneys have in our body those cells having the shape of a rod enjoy a clear advantage during growth, because their surface to volume ratio is constant along their entire cell cycle. On the contrary the amount of surface per unit

244

Vicente, Alvarez and Martínez-Arteaga

of volume diminishes progressively during the growth of a sphere. The dcw cluster of Neisseria gonorrhoeae (Francis et al., 2000), an extracellular pathogen that causes severe infections in the genital tract of humans and that has a coccal instead of a rod shape, is nevertheless almost similar to the E. coli one. Neisseria is then a notable exception to the rule that correlates dcw gene clustering and cell shape. The dcw cluster of N. gonorrhoeae (figure 3) contains no clear-cut transcriptional terminator, but it differs from the dcw cluster of E. coli as it contains some additional coding sequences and inverted repeats (Snyder et al., 2001; Snyder et al., 2003). At least one of these extra elements (Snyder et al, 2003.) functions as an attenuator of transcription while another adjacent element contains an additional promoter. It is not clear if these elements are able to disrupt transcription to a sufficient extent to separate the cluster into two different transcriptional units. They have been found in several species of Neisseria and their incorporation into the cluster may be contributing to the divergence of Neisseria from the E. coli transcription pattern.

Figure 3. The Neisseria gonorrhoeae FA 1090 dcw cluster. The insertion of a CREE element is shown as an open arrowhead at the point of insertion. The white arrowhead is a gearbox promoter functional in E. coli but not in Neisseria. The grey stem and loop structure represents an uptake element that may attenuate transcription. It is followed by a promoter, represented as a black arrow, preceding the dcaC gene. Other symbols as in figure 1.

5.

Expression of cell division genes during Caulobacter development

An unexplained observation is that in E. coli the combined consequence of the signals that contribute to the transcription of the dcw cluster and the subsequent processing of the transcripts, results in an accumulation of ftsZ transcripts that shows a periodic pattern (Garrido et al., 1993; Zhou and Helmstetter, 1994). E. coli follows a relatively simple cell cycle, in bacteria

How similar cell division genes are located and behave in different bacteria

245

with more complex cell division cycles, as exemplified by the obligatory dimorphic organism Caulobacter crescentus the cellular concentration of the essential cell division protein FtsZ is subject to both transcriptional and proteolytic controls (Kelly et al., 1998). The dcw cluster in Caulobacter is split in three separate locations (figure 4). After division to form a stalked (sessile) and a swarmer (motile) cell, a two-component cell cycle-dependent response regulator, CtrA, represses both DNA replication and the transcription of ftsZ in the swarmer cell. In addition, the FtsZ protein is degraded in the swarmer cell preventing further cell division. However, once the swarmer cell differentiates into a stalked cell, CtrA is degraded allowing DNA to replicate and ftsZ to be expressed in preparation for the next cell division.

Figure 4. The Caulobacter crescentus dcw cluster. Large insertions or rearrangements are represented as arrowheads pointing to the sites in which the dcw cluster has been disrupted. Other symbols as in figure 1.

6.

Dispersal of the cluster in Helicobacter

Clustering of the genes that encode cell division and peptidoglycan synthesising proteins may result from a selective advantage conferred by the clustering to the rod-shaped bacteria. Channelling of peptidoglycan precursors to the septation machinery would ensure that septation proceeds in bacilli simultaneously with elongation. Although in cocci growth and septation have not been described yet in sufficient detail it seems that in many cases a single peptidoglycan synthesising reaction, namely septation, suffices both to duplicate the volume of the cell and to separate the two daughters. The loss of the cylindrical shape would relief the constrain that imposes a strict conservation of the dcw cluster, and the precise ordering of the genes could be progressively lost. Coccal Neisseria may represent one of the first stages in the loss of the cylindrical shape in which the physical

246

Vicente, Alvarez and Martínez-Arteaga

integrity of the dcw cluster is still maintained, although it transcriptional continuity may be affected. If such is the case, the future fate of the N. gonorrhoeae dcw cluster may well be the physical dispersal of the genes to distant chromosomal positions. Progressively more dispersed dcw clusters, that could be examples of present day stages of gene dispersal, may be represented by the dcw cluster of Helicobacter pylori. H. pylori is a coilshaped pathogen in which the dcw cluster appears distributed at separate positions, some of them, such as the ftsA and Z, conserving a remnant of the clustering (figure 5).

Figure 5. The Helicobacer pylori dcw cluster. Although the continuity of the cluster has been severely disrupted some traces of clustering are still recognised. Symbols as in figure 1.

7.

How our knowledge on the clustering and behaviour of bacterial cell division genes may advance

As the sequence of more microbial genomes becomes available, more examples of the correlation between the cell shape and the compactness of the dcw cluster may be found that would strengthen its validity. In addition more exceptions are likely to be found as well. If an equivalent progress in the description of gene expression regulation occurs, it may then be possible to extract additional conclusions on how the clustering may have been favoured in the first place, and how some of the exceptions may reflect adaptations to specific environments, e.g. to parasitic lifestyles. Regrettably the study of gene expression, even with the recently available functional genomic technologies that allow the analysis of the full set of transcripts from a given cell, requires more effort than present-day sequencing technologies. As a consequence, given these limitations, many of the possible exceptions that may be found are likely to remain temporarily unexplained due to the difficulty to obtain sufficient evidence.

How similar cell division genes are located and behave in different bacteria

247

References Ballesteros, M., Kusano, S., Ishihama, A., and Vicente, M. (1998) The ftsQ1p gearbox promoter of Escherichia coli is a major sigma S-dependent promoter in the ddlB-ftsA region. Mol. Microbiol. 30, 419-430. Carballés, F., Bertrand, C., Bouche, J. P., and Cam, K. (1999) Regulation of Escherichia coli cell division genes ftsA and ftsZ by the two-component system rcsC-rcsB. Mol. Microbiol. 34, 442-450. de la Fuente, A., Palacios, P., and Vicente, M. (2001) Transcription of the Escherichia coli dcw cluster, evidence for distal upstream transcripts being involved in the expression of the downstream ftsZ gene. Biochimie 83, 109-115. Dewar, S. J., Kagan-Zur, V., Begg, K. J., and Donachie, W. D. (1989) Transcriptional regulation of cell division genes in Escherichia coli. Mol. Microbiol. 3, 1371-1377. Dewar, S. J., and Dorazi, R. (2000) Control of division gene expression in Escherichia coli. FEMS Microbiol. Lett. 187, 1-7. Flärdh, K., Garrido, T., and Vicente, M. (1997) Contribution of individual promoters in the ddlB-ftsZ region to the transcription of the essential cell-division gene ftsZ in Escherichia coli. Mol. Microbiol. 24, 927-936. Flärdh, K., Palacios, P., and Vicente, M. (1998) Cell division genes ftsQAZ in Escherichia coli require distant cis-acting signals upstream ddlB for full expression. Mol. Microbiol. 30, 305-315. Francis, F., Ramírez-Arcos, S., Salimnia, H., Victor, C., and Dillon, J.R. (2000) Organization and transcription of the division cell wall (dcw) cluster in Neisseria gonorrhoeae. Gene 251, 141-151. García-Lara, J., Shang, L. H., and Rothfield, L. I. (1996) An extracellular factor regulates expression of sdiA, a transcriptional activator of cell division genes in Escherichia coli. J. Bacteriol. 178, 2742-2748. Garrido, T., Sánchez, M., Palacios, P., Aldea, M., and Vicente, M. (1993) Transcription of ftsZ oscillates during the cell cycle of Escherichia coli. EMBO J. 12, 3957-3965. Ghigo, J. M., and Beckwith, J. (2000) Cell division in Escherichia coli, role of FtsL domains in septal localization, function, and oligomerization. J. Bacteriol. 182, 116-129. Hara, H., Yasuda, S., Horiuchi, K., and Park, J. T. (1997) A promoter for the first nine genes of the Escherichia coli mra cluster of cell division and cell envelope biosynthesis genes, including ftsI and ftsW. J. Bacteriol. 179, 5802-5811. Henriques, A. O., de Lancastre, H., and Piggot, P. J. (1992) A Bacillus subtilis morphogene cluster that includes spoVE is homologous to the mra region of Escherichia coli. Biochimie 74, 735-748. Joseleau-Petit, D., Vinella, D., and D'Ari, R. (1999) Metabolic alarms and cell division in Escherichia coli. J. Bacteriol. 181, 9-14. Kelly, A. J., Sackett, M. J., Din, N,. Quardokus, E., and Brun, Y. V. (1998) Cell cycledependent transcriptional and proteolytic regulation of FtsZ in Caulobacter. Genes Dev. 12, 880-93. Lara, B. and J. A. Ayala (2002). Topological characterization of the essential Escherichia coli cell division protein FtsW. FEMS Microbiol. Lett. 216, 23-32. Lesley, S. A., Graziano, J., Cho, C. Y., Knuth, M. W., and Klock, H. E. (2002) Gene expression response to misfolded protein as a screen for soluble recombinant protein. Protein Eng. 15, 153-160.

248

Vicente, Alvarez and Martínez-Arteaga

Lutkenhaus, J., and Mukherjee, A. (1996) Cell division. In, Neidhardt et al (eds). Escherichia coli and Salmonella, cellular and molecular biology. 2nd ed. ASM Press, Washington, D.C. pp. 1615-1626. Mengin-Lecreulx, D., Ayala, J., Bouhss, A., van Heijenoort, J., Parquet, C., and Hara, H. (1998) Contribution of the Pmra promoter to expression of genes in the Escherichia coli mra cluster of cell envelope biosynthesis and cell division genes. J. Bacteriol. 180, 44064412. Navarro, F., Robin, A., D'Ari, R. and Joseleau-Petit, D. (1998) Analysis of the effect of ppGpp on the ftsQAZ operon in Escherichia coli. Mol. Microbiol. 29, 815-823. Nikolaichik, Y. A., and Donachie, W. D. (2000) Conservation of gene order amongst cell wall and cell division genes in eubacteria, and ribosomal genes in eubacteria and eukaryotic organelles. Genetica 108, 1-7. Palacios, P., Vicente, M., and Sánchez, M. (1996) Dependency of Escherichia coli celldivision size, and independency of nucleoid segregation on the mode and level of ftsZ expression. Mol. Microbiol. 20, 1093-1098. Powell, B. S., and Court, D. L. (1998) Control of ftsZ expression, cell division, and glutamine metabolism in Luria-Bertani medium by the alarmone ppGpp in Escherichia coli. J. Bacteriol., 180, 1053-1062. Smith, R. W., Masters, M., and Donachie, W. D. (1993) Cell division and transcription of ftsZ. J. Bacteriol. 175, 2788-2791. Spratt, B. G. (1977) Temperature-sensitive cell division mutants of Escherichia coli with thermolabile penicillin-binding proteins. J. Bacteriol. 131, 293-305. Stragier, P., Losick, R. (1996) Molecular genetics of sporulation in Bacillus subtilis. Annu. Rev. Genet. 31, 297-341. Snyder, L. A., Saunders, N. J., and Shafer, W. M. (2001) A putatively phase variable gene (dca) required for natural competence in Neisseria gonorrhoeae but not Neisseria meningitidis is located within the division cell wall (dcw) gene cluster. J. Bacteriol. 183, 1233-1241. Snyder, L. A., Shafer, W. M., and Saunders, N. J. (2003) Divergence and transcriptional analysis of the division cell wall (dcw) gene cluster in Neisseria spp. Mol. Microbiol. 47, 431-442. Tamames, J., González-Moreno, M., Mingorance, J., Valencia, A., and Vicente, M. (2001) Bringing gene order into bacterial shape. Trends. Genet. 17, 124-126. Vicente, M., Gómez, M. J., and Ayala, J. A. (1998) Regulation of transcription of cell division genes in the Escherichia coli dcw cluster. Cell. Mol. Life Sci. 54, 317-324. Weber, A., and Jung, K. (2002) Profiling early osmostress-dependent gene expression in Escherichia coli using DNA macroarrays. J. Bacteriol. 184, 5502-5507. Zhou, P., and Helmstetter, C. E. (1994) Relationship between ftsZ gene expression and chromosome replication in Escherichia coli. J. Bacteriol. 176, 6100-6106.

Chapter 13 The bacterial dcw gene cluster: an island in the genome? JESÚS MINGORANCE1* and JAVIER TAMAMES2

1 Centro Nacional de Biotecnología (CSIC) Campus de Cantoblanco 28049 Madrid 2 ALMA Bioinformática. Ronda de Poniente, 4. Tres Cantos. 28760 Madrid Spain. *corresponding author

1.

Introduction

The bacterial dcw cluster is a group of genes (16 in Escherichia coli, Fig. 1) involved in the synthesis of peptidoglycan precursors and cell division (Ayala et al., 1994). A prominent feature of this cluster is that it is conserved in many bacterial genomes over a broad taxonomic range (Vicente et al., 1998; Nikolaichik and Donachie, 2000; Tamames et al., 2001). A finding that is further underscored by the fact that in the prokaryotes genome organization is very dynamic in evolutionary terms, meaning that the gene content as well as the gene order within operons and gene clusters diverge quickly with increasing evolutionary distance (Tamames et al., 1997; Itoh et al., 1999; Tamames, 2001; Xie et al., 2003; Chapter 11). An analysis of gene order conservation in the genomes of several bacteria and archaea found eighteen highly conserved groups including nearly 120 genes. This means that less than 3% of the E. coli genes have a genome organization highly conserved across the prokaryotes (Tamames, 2001). Nine of the groups included ribosomal protein genes and other genes related with the processes

Molecules in Time and Space, Edited by Vicente et al. Kluwer Academic/Plenum Publishers, New York 2004

249

250

Mingorance and Tamames

Figure 1. Schematic representation of the dcw cluster in several bacterial genomes. Adjacent boxes represent contiguous genes. White boxes correspond to genes that are not known to be related with division or cell wall synthesis. Groups of genes separated by spaces represent groups that are separated in the genome. Isolated genes are not represented but most are present in all the genomes (Chapter 8).

of transcription and translation. Among these, there was a large cluster of ribosomal protein genes that is the most conserved, and is found in bacterial, archaeal, and organelle genomes (Wächtershäuser, 1998; Hansmann and Martin, 2000; Tamames, 2001). Other conserved groups included the dcw cluster, the atp operon (coding for the subunits of ATP synthase), and the his operon (coding for the enzymes involved in the biosynthesis of histidine). In most cases it is obvious that the clustered genes are functionally related (Tamames et al., 1997), what is not evident at all is which are the selective forces that hold these genes together in a precise order over long evolutionary distances. Comparison of the dcw cluster from several bacterial genomes using phylogenetic analysis tools has shown that there is a strong relationship between the organization of this gene cluster and the morphology of the cells (Tamames et al., 2001). This correlation suggests that the organization of these genes in the chromosome is related to the processes of cell growth and division. We postulate that for the rod-shaped bacteria the mechanism

The dcw gene cluster

251

linking gene order in the chromosome with cell shape is the cotranslational assembly of the products of the mur and fts genes to form the multiprotein complexes responsible for cell wall growth and cell division.

2.

The dcw cluster: structure and function

Part of the dcw cluster was initially identified in Escherichia coli as a region containing genes related to the synthesis of murein precursors, and was called mra region (Miyakawa et al., 1972) or mra operon (Donachie, 1993). Later several cell division genes were found to map adjacent to this region, and therefore the name dcw cluster, standing for division and cell wall, was proposed to include the whole set of linked genes (Ayala et al., 1994). In the context of comparative genomics the term cluster seems more appropriate than operon, because in most organisms the transcriptional structure of this region is not known. And in fact in some organisms, like Neisseria gonorrhoeae, the cluster might not be an operon but be composed of several independent transcriptional units (Francis et al., 2000; but see also Snyder et al., 2003). In E. coli the cluster contains 16 tightly packed genes (Fig. 1), some overlapping, and all of them transcribed in the same direction (Vicente et al., 1998; Dewar and Dorazi, 2000). There is one initial promoter upstream of mraZ (mraZ1p) several internal promoters, and a single transcription terminator downstream of the gene after ftsZ, There are four known promoters in the 5' end of the cluster and six in the 3' end, in the region upstream of ftsZ (Vicente et al., 1998). The first promoter, mraZ1p, is essential for transcription of at least the first nine genes of the cluster (Mengin-Lecreulx et al., 1998). Six internal promoters have been characterized in the region containing the genes ddlB-ftsQ-ftsA-ftsZ, contributing to about one third of the total expression of the ftsZ gene (Flärdh et al., 1998). The remaining two thirds are derived from promoters located upstream of ddlB (de la Fuente et al., 2001). The absence of internal transcription terminators suggests that in E. coli there might exist a large transcript arising from mraZp1 and spanning the whole cluster, but experimental evidence for the existence of this transcript is still lacking. In any case this hypothetical transcript could be differentially processed to yield shorter transcripts due to the presence of RNase targets at specific regions of the sequence, for example the ftsZ transcript (Cam et al., 1996). A high-resolution transcript analysis of the whole cluster using DNA microarrays has shown two peaks with higher steady-state transcript levels just downstream of the promoters mraZ1p and ftsQ2p1p, and a central region with lower levels. In addition it was found that the transcripts from this

252

Mingorance and Tamames

central region are more stable than those from the lateral ones, which might have a faster turnover (Selinger et al., 2003). The transcriptional pattern may be different in other organisms, for example in Bacillus sp. there are two ftsI homologs (pbpB and spoVD) separated by a transcriptional terminator (Daniel et al., 1996) and responsible respectively for the synthesis of peptidoglycan during vegetative and sporulation septum formation (Daniel et al., 1994). The function of the first gene of the cluster, is not known. The next gene, codes for an S-adenosylmethioninedependent methyltransferase of unknown substrate and unknown function (Carrión et al., 1999). The mur genes, together with mraY and ddlB code for the enzymes responsible for the synthesis of the murein precursors (van Heijenoort, 1996). These genes are considered to be unique and essential (Boyle and Donachie, 1998; Kobayashi et al., 2003). The gene ddlB has an additional copy in the E. coli chromosome, ddlA, and therefore although it is not essential in itself in E. coli, the activity of the enzyme coded by these two genes, D-alanine:D-alanine ligase, is essential for cell wall synthesis (Zawadzke et al., 1991; van Heijenoort, 1996). The cluster contains six fts genes (ftsL, ftsI, ftsW, ftsQ, ftsA and ftsZ), all of them are essential for cell division, and all of them code for proteins that have been shown to localize to the division site during septation (Chapter 5). Downstream of ftsZ there are some genes that are specific of some bacterial groups, like the envA gene found in the or a cluster of up to six additional genes, namely ylmD, ylmE, ylmF, ylmG, ylmH and divIVA, found in the Gram-positive bacteria (Massidda et al., 1998).

3.

Conservation of the genes in the dcw cluster

Some generalizations can be made from the comparison of the dcw cluster in several genomes (Nikolaichik and Donachie, 2000; Tamames et al., 2001). First, although most of the genes of the cluster are conserved in the majority of the genomes studied (see Chapter 9), they are not invariably found in the cluster, and most of them can be found in other locations in some instances. The frequency with which the genes appear in the cluster is also variable, and there are genes, like murB, that are found more often in other locations, while others, like ftsQ, ftsA and ftsZ only rarely are found isolated.

The dcw gene cluster

253

Figure 2. Schematic representation of the dcw cluster from two deep branching bacterial lineages.

Second, the relative order of the genes is highly conserved, and a hypothetical archetypal, or ancestral (Nikolaichik and Donachie, 2000), cluster can be defined that contains all the genes (Fig. 1). Most of the clusters analyzed can be derived from this archetypal cluster by deletion, and sometimes insertion of extra genes, without changing the relative order of the remaining genes (by relative order it is meant the ordering of neighbour genes in relation to the direction of transcription). In some organisms, like Caulobacter crescentus, or Neisseria gonorrhoeae the cluster has been split by insertions. In others, like Thermotoga maritima, or Rickettsia prowazekii it is dispersed into two or three separated groups, and even more in Helicobacter pylori and Campylobacter jejuni, but the same relative gene order is almost invariably maintained within each group (Figs. 1, 2 and 3). As a result, a pattern of conservation can be outlined in the hypothetical cluster: the 5' region is variable, as it contains the mraZ and ftsL genes that are missing in many genomes; then there is a conserved block comprising the genes ftsI to murG; this is followed by another variable region that may contain the genes murC, murB and ddl, though only rarely the three genes are found together; and finally there is another conserved block comprising the division genes ftsQAZ. This pattern was clearly seen when conservation of the cluster was measured as occurrence of gene pairs (Fig. 4 in Tamames et al., 2001).

4.

On the origin of the dcw cluster

254

Mingorance and Tamames

Is the dcw cluster an ancient and conserved trait, or is it of recent origin? It seems very unlikely that a complex gene structure like this, with such precise gene ordering, might be a recent acquisition arisen independently by convergent evolution in phylogenetically distant bacterial groups. Other possibilities are that the cluster might be an ancient trait that was already present in the last common ancestor (LCA) of extant bacteria, or alternatively, that it might have originated after the LCA, and might have been later spread through the bacterial domain by lateral gene transfer. The phylogenetic analysis of the sequences of individual genes shows trees that are congruent with the standard 16S rRNA tree. There are a few exceptions but these correspond to different genes in different organisms showing that in general the genes have evolved in a manner consistent with that of the organisms to which they belong, and arguing against extensive lateral gene transfer after the LCA (Faguy and Doolittle, 1998; Fig. 8). The cluster is found in Thermotoga maritima, and in a very reduced version in Aquifex

Figure 3. Hypothetical pathway of assembly of the dcw cluster indicating the four gene groups implicated. The order of the different events shown is arbitrary. The pathway starts with a primordial mur operon containing the genes responsible for the synthesis of the peptidoglycan precursor. An operon containing the rodA/ftsW and the pbpA/ftsI ancestors was duplicated, and then the pathway could diverge into septal and lateral peptidoglycan synthesis pathways. Incorporation of the fts genes might have facilitated the localization of the Mur complex to the septal zone during division.

The dcw gene cluster

255

aeolicus (Fig. 2), both belonging to the deepest branches of the bacterial phylogenetic tree (Bocchetta et al., 2000). In T. maritima the cluster is split in two groups plus several isolated genes. One of the groups contains some mur genes and the other contains the ftsA-ftsZ pair. In A. aeolicus the cluster is reduced to the murB and ddlB genes clustered with the ftsA-ftsZ pair. The fact that both organisms have clusters but these do not overlap might be explained in several ways, but the most parsimonious explanation is that the genomic organization of the cluster in these two bacteria is derived secondarily by reduction from a large cluster that contained all the genes, and that this has happened also in other groups. A corollary of this is that the LCA had a large dcw cluster that has been maintained in some lineages and has been reduced in others (Fig. 1). How did this cluster arise? Although this is rather speculative, examination of gene function and gene order suggests that it might have arisen from the blending of four gene subsets (Fig. 3): i) the mur genes (including mraY and ddlA/B), responsible for the synthesis of peptidoglycan precursors; these genes code for the enzymes of a single biosynthetic pathway, and might have been maintained together forming an operon before the LCA; ii) the ftsI and ftsW genes that participate in the specialized peptidoglycan synthesis at the septum, and have counterparts in the pbpA and rodA genes, involved in the synthesis of lateral wall peptidoglycan; these two pairs of genes must have arisen also by duplication from a single ancestor pair; iii) the ftsL and ftsQAZ genes, that code for proteins that localize at the division ring; iv) the mraZ and mraW genes, that are of unknown function, and might belong to some of the other groups. The order in which the different rearrangements shown in Fig. 3 occurred is unknown.

5.

Phylogenetic distribution of the dcw cluster and bacterial shape

A phylogenetic analysis of the clusters from several completed genomes, based not on gene sequences, but on gene content and gene order, produced a tree different from the standard 16S rRNA-based tree (Tamames et al., 2001). A complete description of the procedure for coding gene order information is given in Tamames et al. (2001). Briefly, a set of vectors is created, one vector for every genome. The vectors contain as many positions as possible pairs of genes (only dcw cluster genes), and each position codes for a given pair of genes, being set to one if the pair exists, and to zero if it does not. A pair of genes might be defined as a pair of adjacent genes (distance 1), or a pair of genes separated by n intervening genes (distance n+1). The resulting set of vectors contains information about the gene pairs

256

Mingorance and Tamames

Figure 4. Schematic representation of the dcw cluster of bacteria with different shapes belonging to two different lineages. T. thermophilus (rod) and D. radiodurans (coccus) belong to the Thermus/Deinococcus group, while B. subtilis (rod) and S. aureus (coccus) belong to the high GC Gram-positives, but regarding the composition of the dcw cluster the two rod-shaped species are clearly more similar, and the two cocci have in common a reduced cluster.

that are found in the genomes analyzed. If a given position is zero in all vectors, meaning that the corresponding gene pair is never found, it is considered uninformative, and it is removed. These sets of vectors allow the building of phylogenetic trees by several methods. Two different approaches were used: a parsimony analysis (Felsenstein, 1996) (using the "mix" program in the Phylip package), and a hierarchical clustering algorithm, SOTA (Herrero et al., 2001). Both methods produce very similar results, with minimal disagreements. The tree showed two main branches, one that included organisms with a compact cluster, and the other formed by organisms in which the genes are more dispersed through the genome and the cluster is reduced, or does not exist at all. The distribution of bacterial species in the tree was found to correlate with their cell morphologies. Bacilli and filamentous bacteria fell in the group that had a large dcw gene cluster, while cocci, helicobacteria and spirochetes formed the branch in which the dcw cluster genes are more dispersed or lost altogether. In some instances the trees separated organisms

The dcw gene cluster

257

Figure 5. Dendrogram of the dcw cluster based on gene content and gene order. The tree shown in this figure was obtained with distance 2 as explained in the text. Names in black correspond to rod-shaped or filamentous bacteria, while those in grey color correspond to bacteria with other shapes.

that belong to the same phylogenetic group but have different shapes, like the low GC Gram-positive bacilli and cocci, the rod-shaped Thermus thermophilus and the coccus Deinococcus radiodurans, or the helicobacteria (Campylobacter and Helicobacter, and the rest of the proteobacteria (Fig. 4; Tamames et al., 2001). To test the robustness of the conclusions drawn in the original work with twenty-six sequenced genomes, the same study has been repeated including twenty-four additional genomes. The results of this new study fully confirmed the previous results, the rod-shaped and filamentous bacteria tend to branch together, and separate from cocci, spirochetes and helicobacteria, although there are a few exceptions. The pattern was less clear (there were more exceptions) when the analysis was done with adjacent gene pairs, probably because even in the organisms in which the cluster is reduced or dispersed several gene pairs are conserved (see for example H. pylori, Fig. 1). When the analysis was done with a distance 2 (Fig. 5) or a distance 3 (Fig. 6) between genes (meaning that there are respectively one and two

258

Mingorance and Tamames

Figure 6. Distance 3 dendrogram of the dcw cluster. The tree was obtained as described in the legend to Fig. 5 but using distance 3 (two intervening genes between each pair of genes). Names in black correspond to rod-shaped or filamentous bacteria, while those in grey color correspond to bacteria with other shapes.

intervening genes between the pair being considered) the pattern was more pronounced. The fact that increasing the distance between genes improves the pattern of separation of rod-shaped cells from other cell types supports the conclusion that the pattern found is due to the clustering of the genes, and not to the individual genes. The branching pattern observed might be due to differences in gene content between genomes, and independent of the positions of the genes in their genomes. To test this possibility an analysis was done taking into account only the presence or absence of the genes from the archetypal cluster, independently of their positions in the genome. In this case the correlation with cell morphology was not found (Fig. 7). The tree obtained was poorly informative, there were some groups consistent with the standard bacterial phylogeny, but several others were mixed and no trends could be discerned, probably because the tree is based on few and highly conserved (uninformative) characters. This shows that the correlation between the organization of the dcw gene cluster and the cell morphology is not an

The dcw gene cluster

259

Figure 7. Dendrogram based on the dcw gene content in the different bacterial genomes. For each genome the presence or absence of the dcw genes (those from the hypothetical ancestor cluster shown in Fig. 1) were recorded independently of their chromosomal positions, and the resulting vectors where used to construct a tree using the methods described in the text.

artifact due to the different gene contents of the genomes, but is due to the fact that these genes are clustered and that the degree of clustering is different in different genomes. Sequence-based phylogenies were done for the individual genes and were found to be fairly similar to the standard 16S rRNA phylogeny (Fig. 8). Most groups were correctly identified, although the relations between them were sometimes inaccurate. Only the helicobacteria, C. jejuni and H. pylori, were found to branch consistently apart from the other proteobacteria, suggesting that in this group at least some of the genes of the cluster might have been received by horizontal gene transfer. The differences between the individual sequence-based gene trees and the tree of the cluster indicate that while individual genes have evolved at the pace of the chromosomes in which they are contained, there have been genomic rearrangements within the dcw cluster that have affected differently to different bacterial lineages. Assuming that the ancestor of extant bacteria had a dcw cluster similar to the

260

Mingorance and Tamames

Figure 8. FtsZ sequence tree. Tree based on the FtsZ predicted protein sequence using a neighbour-joining method. The H. pylori and C. jejuni branch in an anomalous position, the rest of the tree agrees fairly well with the standard 16S rRNA phylogeny.

archetypal cluster represented in Fig. 1, it can be concluded that the cluster has been maintained in most rod-shaped bacteria and tends to be lost from bacteria with other morphologies. This agrees with the inference that coccoid, helical, and other cell shapes have arisen several times independently from a rod-shaped ancestor (Siefert and Fox, 1998; Chapter 1).

6.

How is the dcw cluster maintained?

If the cluster originated in the ancestor of extant bacteria then what are the forces that have maintained the cluster for billions of years? And why in some lineages these forces are relaxed or do not operate anymore? We have shown that there is a relation between the genomic structure of the dcw cluster and the shape of the cells, but it is very unlikely that the cluster in

The dcw gene cluster

261

itself has any morphogenetic properties. Transferring the cluster from a rodshaped bacterium into a coccus would not transform it into a bacillus because most cocci lack several other genes, like mreB, also involved in the determination of the cell morphology (Chapter 9). Furthermore, breaking the continuity of the cluster in E. coli has several effects, and may cause filamentation, but does not convert E. coli cells into cocci (Palacios et al., 1996; Mengin-Lecreulx et al., 1998). There are currently several models to explain the clustering of genes and the persistence of clusters for long evolutionary periods (reviewed in Lawrence and Roth, 1996). These models are based on different mechanisms, and therefore do not exclude each other, but might operate simultaneously. The natal model postulates that clusters originate by gene duplication and subsequent sequence divergence. In this case the order of genes in the cluster simply reflects the order in which duplications occurred. This model might apply to the origins of part of the dcw cluster (Fig. 3), as four of the genes (murC, murD, murE and murF) are paralogues that must have arisen by gene duplication in the ancestor of all eubacteria, but it cannot explain the presence in the cluster of other, unrelated, genes, the long persistence of the cluster, or its relation with cell morphology. In the selfish operon model clustering benefits the cluster itself, and conservation of gene order reflects the spreading of the cluster by lateral gene transfer. This model applies successfully to weakly selected functions, but it has difficulties with housekeeping genes. To apply the selfish operon model to the dcw cluster it should be assumed that 1) a full cluster is acquired via LGT, and 2) the old (endogenous) division genes, initially dispersed through the genome, are immediately lost, since in most cases they will interfere with the new (exogenous) genes (for example, exogenous ftsZ genes are often toxic for E. coli). A simple mechanism to effect the rapid and simultaneous loss of the dispersed genes is difficult to envisage, and the model is useless if it assumes that they were already clustered. Moreover, the selfish operon model might explain why the cluster has such a broad phylogenetic distribution, but not why it is related to cell shape. The co-regulation model proposes that clustering facilitates the coordinate expression and regulation of the clustered genes (Lawrence and Roth, 1996). Contrary to this, there are many instances of known genes that are co-regulated, and are in different chromosomal locations, or are found together in some organisms and not in others (Tamames et al., 1997; Itoh et al., 1999). Therefore, although gene grouping contributes to the coordinate regulation of gene expression at the DNA level, it is clear also that these mechanisms are very flexible in evolutionary terms, and do not require conservation of gene order (Xie et al., 2003). An alternative to this model

262

Mingorance and Tamames

proposes that clustering might be related to the co-regulation of gene expression at the RNA level (Siefert et al., 1997). This might apply to the cluster of ribosomal protein genes and other clusters that are known to be regulated at the level of mRNA translation in at least some organisms, but so far this level of regulation has not been described in the dcw cluster, and it is not clear at all why such a strict conservation of gene order would be required. Finally, the Fisher model postulates that clustering is the result of selection acting on a co-adapted set of genes to maintain them together because physical proximity decreases the probability of recombination occurring between them. This model is well suited to sexual organisms, where recombination is frequent. The discovery that prokaryotes may undergo extensive horizontal gene transfer (Nelson et al., 1999) made the application of this model to bacteria more plausible, although the frequency of gene exchange in most species is not known. It has been postulated that the conservation of gene order in the dcw cluster might be explained by this model as the result of the lateral exchange of a set of linked genes that code for a co-adapted set of proteins (Nikolaichik and Donachie; 2000). This argument has two parts: first, that the linked genes are co-adapted; and second that the linkage is maintained by selection against the mixture of genes ("alleles") from different sources. When applied to the dcw gene cluster, the first part of the argument is likely to be correct, as co-adapted proteins are those that interact either physically or functionally, which is the case in this cluster. On the other hand, it is unlikely that the cluster might have been maintained by horizontal gene transfer. It is true that exogenous cell division genes are usually toxic, at least for E. coli, and lateral gene transfer is certainly more frequent in prokaryotes than previously thought, but it is doubtful that the frequency of lateral gene transfer has been so high as to spread the cluster throughout the entire bacterial domain. Moreover, the division and cell wall genes are ancient genes likely to belong to the group of orthologous genes that are not likely to participate in lateral gene transfer (Daubin et al., 2003). The sequence-based phylogenetic trees obtained for the individual genes are congruent with the standard 16S rRNA tree topology (Fig. 8), indicating that the cluster has not been subjected to extensive lateral gene transfer after the LCA. Therefore, none of these models is able to explain satisfactorily the existence of the dcw gene cluster, and although some of them might be partially correct, a new mechanism should be invoked that takes into account the properties of the cluster.

The dcw gene cluster

263

7.

A model on the relation of bacterial cell shape and the dcw gene cluster

7.1

Gene order reflects protein-protein interactions

It has been found that conservation of gene order in several gene pairs is a predictor of protein-protein interactions (Dandekar et al., 1998; Overbeek et al., 1999; Huynen et al., 2000; Chapter 11). As gene order is maintained during transcription and translation, it has been suggested that this correlation results from an interdependence of the folding of the proteins involved (cotranslational folding; Dandekar et al., 1998). This interdependence should be weak in the case of the dcw cluster, as some of the genes have been successfully expressed in isolation from plasmids. Instead, we postulate that the maintenance of gene order is related to the cotranslational assembly and localization of protein complexes, and that the order might be important to promote the correct interactions among the different components of the complexes. Cotranslational assembly and localization might be especially important for complexes that are in low copy numbers in the cells. It has been already pointed that some of the genes of the cluster have been expressed from plasmids and have been found to be functional and able to rescue conditional mutations in the chromosome, therefore clustering is hot essential for the correct folding of the proteins produced, and is not essential for cell growth and division. Yet it has to be remembered that in most cases the cloned genes were in multicopy plasmids, and therefore the number of protein molecules produced was probably much higher than in the wild type cells. The high local concentrations of the polypeptide chains that may be reached in polysomes might favour assembly even if only a few molecules of some of the components exist in the cell (for example, in E. coli FtsQ is present at ca. 50 molecules/cell; Carson et al., 1991). In addition, as some of the products of the genes in the cluster are membrane proteins, the coupled transcription-translation-insertion of these proteins might direct the localization of the protein complexes to specific domains in the cell membrane (Norris et al., 1996). A cotranslation mechanism would not necessarily mean that the order of the genes should be essential, only that it should have a significant effect on fitness. It might work on housekeeping genes, and it might be operative over large evolutionary distances, as far as there is some selective pressure to maintain the function of the protein complexes involved. The products of the fts genes of the cluster are good candidates for such mechanism, as they are known to localize in a ring at the septation site during division, and it is thought that there they form a protein complex

264

Mingorance and Tamames

responsible for septation (Nanninga, 1991; Rothfield et al., 1999). Moreover, several interactions among the products of ftsI, ftsQ, ftsA and ftsZ genes have been described (Dai and Lutkenhaus, 1992; Descoteaux and Drapeau, 1987; Weiss et al., 1999). Among the mur genes, on the other hand, some code for soluble and others for membrane proteins, all of them involved in the same biochemical pathway. Most of these proteins have been purified and characterized, and they are active in vitro in isolation (Duncan et al., 1990; Michaud et al., 1990; Pratviel-Sosa et al., 1991; Liger et al., 1995). However analysis of the pool levels of peptidoglycan precursors in E. coli has shown that the cellular concentrations of the metabolites in the pathway are below the Km values for the corresponding enzymes (van Heijenoort, 1996). The pools of lipids I and II in particular are very low (700 and 1000-2000 molecules/cell respectively), and seem to be a limiting factor for the synthesis of peptidoglycan. It has been suggested that as these enzymes are not saturated by substrate, the efficiency and the control of the pathway could be greatly improved if they were part of a multiprotein complex and could transfer sequentially the products of every enzyme to the next enzyme of the pathway (van Heijenoort, 1996). This improvement might provide a selective force favouring the presence of a cluster of mur genes if the grouping of genes facilitates complex assembly.

7.2

Connecting the mur and fts complexes: the two-competing sites model for peptidoglycan synthesis

The dcw cluster contains both mur and fts genes interspersed, therefore the cotranslational assembly model requires that additional interactions between the mur complex(es) and the fts complex(es) should be operative and physiologically advantageous to explain the persistence of the cluster and its broad taxonomic range. We propose that the genomic structure of the dcw cluster, as well as its phylogenetic distribution result from the simultaneous operation of cotranslational assembly of Mur and Fts proteins, and the existence of two-competing sites for peptidoglycan synthesis in rodshaped cells (Satta et al., 1994). In these, peptidoglycan synthesis involves different precursors and multienzyme complexes during elongation and during septation (Canepari et al., 1997; Höltje, 1998; de Pedro et al., 1997; Alaedini and Day, 1999). Synthesis of new peptidoglycan during elongation occurs in patches, bands and areas of diffuse insertion in the sidewall of the cell (de Pedro et al., 2003), while the insertion of new peptidoglycan during septation occurs in a narrow zone at the division site. The two-competing sites model for peptidoglycan synthesis states that these two processes are in fact two competing reactions (or pathways), and that the balance between

The dcw gene cluster

265

Figure 9. Schematic representation of the cotranslational assembly and localization of multiprotein complexes involved in the elongation of the lateral cell wall, and in division. Coupled transcription-translation together with the insertion of membrane proteins, a process called "transertion" (Norris et al., 1996), facilitates the localized assembly of protein complexes at the division site. After division the Mur moiety of this complex might migrate through the membrane to interact with the elongation complex.

them regulates cell shape in rod-shaped bacteria, while in the cocci the lateral-wall elongation reaction is severely reduced or absent (Satta et al., 1994). We propose that in rod-shaped bacteria, prior to cell division, cotranslational folding and assembly of the products of the mur and fts genes results in the assembly and localization of a Mur-Fts complex able to synthesize murein precursors and channel them towards septal peptidoglycan synthesis. After cell division this complex disassembles, and the Mur moiety might diffuse through the membrane, and then interact with the lateral wall elongation machinery, again to channel the murein precursors towards the synthesis of lateral peptidoglycan (Fig. 9). The localization of a Mur-Fts complex that synthesizes the murein precursors at the division site provides a mechanism to channel the flux of precursors (that provide matter and energy) towards the synthesis of septal peptidoglycan during division. This is an intense, and highly localized

266

Mingorance and Tamames

process that would be substrate-limited if the precursors were synthesized and used throughout the whole cell surface. This mechanism might be even more important in filamentous cells that grow apically to avoid diffusion of the precursors through the long cell body. On the contrary, the channeling of murein precursors might not be as important in cocci, in which there is one narrow zone of murein synthesis either for elongation and for division (Morlot et al., 2003; Pinho and Errington, 2003). The localization of the complex responsible for the synthesis of the peptidoglycan precursors would not be necessary in the cocci, and therefore the selective pressure to maintain the clustering would be relieved, and as a consequence the dcw genes have often been dispersed. In other organisms, like the spirochetes, or the helicobacteria the lack of a compact cluster might also be related to their particular modes of cell growth and division. The localization of a Mur-Fts complex might need assistance because it involves the concentration of few molecules in a narrow area to form protein complexes that are predicted to be weak and transient, as the Z-ring is not a static protein assembly, but a highly dynamic structure (Stricker et al., 2002; Rueda et al., 2003). Therefore, gene clustering and cotranslational assembly and localization of protein complexes might help rod-shaped cells to grow and divide more efficiently. However these are not by themselves morphogenetic mechanisms. Other genes have been associated to the morphogenesis of E. coli (Chapter 9). The mreB gene seems to have a key role, as it is found in prokaryotes (both bacteria and archaea) with rod or similar shapes, but not in the genomes of several cocci (Jones et al., 2001). As more genomes are sequenced exceptions arise, for example Methylococcus capsulatus and Magnetococcus sp., are cocci but both contain complete dcw clusters, and mreB homologs. These exceptions do not invalidate the model, that assumes that clustering may be present in all kind of cells, but it is a favourable trait only for the rod-shaped, and is dispensable in others. Some predictions can be made from our model that might be experimentally tested. For example, it would be expected that the dcw mRNA transcript (or part of it) and the products of the mur genes should colocalize with the Z-ring during septation in rod-shaped bacteria. Moreover, several weak interactions between the proteins coded by the genes of the cluster should be expected, and these should occur between protein pairs in a manner consistent with the order of the genes in the cluster. This last prediction is partially supported by the intriguing observation that in the Chlamydia the genes murC and ddlA are fused (Chopra et al., 1998), as it has been argued that the fusion of two genes in one genome indicates that the protein products of these two genes may interact and form complexes, even in other organisms in which the genes are not fused (Enright et al.,

The dcw gene cluster

267

1999). Although the Chlamydia lack both peptidoglycan, and the ftsQAZ genes, they have the other genes of the cluster, and are sensitive to penicillin, suggesting that the pathway is active in these organisms (Ghuysen and Goffin, 1999; McCoy et al., 2003).

8.

Conclusion

It has been shown that extant bacterial lineages descend from rod-shaped, peptidoglycan containing bacteria (Siefert and Fox, 1998). It seems likely that those bacteria already contained a dcw cluster that had evolved in parallel with the peptidoglycan synthesizing machinery to allow an efficient co-ordination between growth and division by alternatively channelling the precursor synthesis towards the division site or to the lateral wall. This coordination might still favour gene clustering in the bacilli, but might become useless when the shape of the cells or their life cycles change during evolutionary history making the cluster susceptible to dispersal, reduction or reorganization.

Acknowledgements We are very grateful to Thomas Harsch and Anke Henne (Göttingen Genomics Laboratory, Georg-August-University, Göttingen, Germany) for kindly providing us unpublished Thermus thermophilus data. This work has been financed through grant BIO2000-0451-P4-02 from the Ministerio de Ciencia y Tecnologia to the laboratory of Miguel Vicente.

References Alaedini, A. and Day, R. A. (1999) Identification of two penicillin-binding multienzyme complexes in Haemophilus influenzae. Biochem. Biophys. Res. Commun. 264, 191-195. Ayala, J. A., Garrido, T., de Pedro M.A. and Vicente, M. (1994) Molecular biology of bacterial septation. In: Ghuysen JM, Habenbeck R, editors. Bacterial cell wall. Amsterdam. Elsevier Science B. V. Bocchetta, M., Gribaldo, S., Sanangelantoni, A. and Cammarano, P. (2000) Phylogenetic depth of the bacterial genera Aquifex and Thermotoga inferred from analysis of ribosomal protein, elongation factor, and RNA polymerase subunit sequences. J. Mol. Evol. 50, 366380. Boyle, D. S. and Donachie, W. D. (1998) mraY is an essential gene for cell growth in Escherichia coli. J. Bacteriol. 180,6429-6432.

268

Mingorance and Tamames

Cam, K., Rome, G., Krisch, H. M. and Bouche, J.P. (1996) RNase E processing of essential cell division genes mRNA in Escherichia coli. Nucleic Acids Res. 24, 3065-3070. Canepari, P., Signoretto, C., Boaretti, M. and Lleò, M. M. (1997) Cell elongation and septation are two mutually exclusive processes in Escherichia coli. Arch. Microbiol. 168, 152-159. Carrión, M., Gómez, M. J., Merchante-Schubert, R., Dongarra, S. and Ayala, J. A. (1999) mraW, an essential gene at the dcw cluster of Escherichia coli codes for a cytoplasmic protein with methyltransferase activity. Biochimie 81,879-888. Carson, M. J., Barondess, J. and Beckwith, J. (1991) The FtsQ protein of Escherichia coli: membrane topology, abundance and cell division phenotypes due to overproduction and insertion mutations. J. Bacteriol. 178, 1328-1334. Chopra, I., Storey, C., Falla, T. J. and Pearce, J. H. (1998) Antibiotics, peptidoglycan synthesis and genomics: the chlamydial anomaly revisited. Microbiology 144, 2673-2678. Dai, K. and Lutkenhaus, J. (1992) The proper ratio of FtsZ to FtsA is required for cell division to occur in Escherichia coli. J. Bacteriol. 174, 6145-6151. Dandekar, T., Snel, B., Huynen, M. and Bork, P. (1998) Conservation of gene order: a fingerprint of proteins that physically interact. Trends Biochem. Sci. 23, 324-328. Daniel, R. A., Drake, S., Buchanan, C.E., Scholle, R. and Errington, J. (1994) The Bacillus subtilis spoVD gene encodes a mother-cell-specific penicillin-binding protein required for spore morphogenesis. J. Mol. Biol. 235, 209-220. Daniel, R. A., Williams, A. M. and Errington J. (1996) A complex four-gene operon containing essential cell division gene pbpB in Bacillus subtilis. J. Bacteriol. 178, 23432350. Daubin, V. Moran, N. A. and Ochman, H. (2003) Phylogenetics and the cohesion of bacterial genomes. Science 301, 829-832. de la Fuente, A., Palacios, P. and Vicente, M. (2001) Transcription of the Escherichia coli dcw cluster: Evidence for distal upstream transcripts being involved in the expression of the downstream ftsZ gene. Biochimie 83, 109-115. de Pedro, M. A., Quintela, J. C., Höltje, J. V. and Schwarz, H. (1997) Murein segregation in Escherichia coli. J. Bacteriol. 179, 2823-2834. de Pedro, M. A., Schwarz, H. and Koch, A. L. (2003) Patchiness of murein insertion into the sidewall of Escherichia coli. Microbiology 149, 1753-1761. Descoteaux, A. and Drapeau, G. R. (1987) Regulation of cell division in Escherichia coli K12: probable interactions among proteins FtsQ, FtsA, and FtsZ. J. Bacteriol. 169, 19381942. Dewar, S. J. and Dorazi, R. (2000) Control of division gene expression in Escherichia coli. FEMS Microbiol. Lett. 187, 1-7. Donachie, W. D. (1993) The cell cycle of Escherichia coli. Annu. Rev. Microbiol. 47, 199230. Duncan, K., van Heijenoort, J. and Walsh, C.T. (1990) Purification and characterization of the D-alanyl-D-alanine-adding enzyme from Escherichia coli. Biochemistry 29, 2379-2386. Enright, A. J., Iliopoulos, I., Kyrpides, N.C. and Ouzounis, C.A. (1999) Protein interaction maps for complete genomes based on gene fusion events. Nature 402, 86-90. Faguy, D. M. and Doolittle, W. F. (1998) Cytoskeletal proteins: The evolution of cell division. Curr. Biol. 8, R338-R341. Felsenstein, J. (1996) Infering phylogenies from protein sequences by parsimony, distance and likelihood methods. Methods Enzymol. 266, 418-427.

The dcw gene cluster

269

Flärdh, K., Palacios, P. and Vicente, M. (1998) Cell division genes ftsQAZ in Escherichia coli require distant cis-acting signals upstream of ddlB for full expression. Mol. Microbiol. 30, 305-316. Francis, F., Ramírez-Arcos, S., Salimnia, H., Victor, C. and Dillon, J-A. R. (2000) Organization and transcription of the division and cell wall (dcw) cluster in Neisseria gonorrhoeae. Gene 251, 141-151. Ghuysen, J. M. and Goffin, C. (1999) Lack of cell wall peptidoglycan versus penicillin sensitivity: new insights into the chlamydial anomaly. Antimicrob. Agents Chemother. 43, 2339-2344. Hansmann, S. and Martin, W. (2000) Phylogeny of 33 ribosomal and six other proteins encoded in an ancient gene cluster that is conserved across prokaryotic genomes: influence of excluding poorly alignable sites from analysis. Int. J. Syst. Evol. Microbiol. 50, 1655-1663. Herrero, J., Valencia, A. and Dopazo, J. (2001) A hierarchical unsupervised growing neural network for clustering gene expression patterns. Bioinformatics 17, 126-136. Höltje, J. V. (1998) Growth of the stress-bearing and shape-maintaining murein sacculus of Escherichia coli. Microbiol. Mol. Biol. Rev. 62, 181-203. Huynen, M. A., Snel, B., Lathe, W. and Bork P. Predicting protein function by genomic context: quantitative evaluation and qualitative inferences. Genome Res. (2000) 10, 12041210. Itoh, T., Takemoto, K., Mori, H., and Gojobori, T. (1999) Evolutionary instability of operon structures disclosed by sequence comparisons of complete microbial genomes. Mol. Biol. Evol. 16, 332-346. Koch, A. (2000) The bacterium's way for safe enlargement and division. Appl. Environ. Microbiol. 66, 3657-3663. Lawrence, J. G. and Roth, J. R. (1996) Selfish operons: horizontal transfer may drive the evolution of gene clusters. Genetics 143, 1843-1860. Liger, D., Masson, A., Blanot, D., van Heijenoort, J. and Parquet, C. (1995) Over-production, purification and properties of the uridine-diphosphate-N-acetylmuramate:L-alanine ligase from Escherichia coli. Eur. J. Biochem. 230, 80-87. Massidda, O., Anderluzzi, D., Friedli, L. and Feger, G. (1998) Unconventional organization of the division and cell wall gene cluster of Streptococcus pneumoniae. Microbiology 144, 3069-3078. Mengin-Lecreulx, D., Ayala, J., Bouhss, A., van Heijenoort, J., Parquet, C. and Hara, H. (1998) Contribution of the Pmra promoter to expression of genes in the Escherichia coli mra cluster of cell envelope biosynthesis and cell division genes. J. Bacteriol. 180, 44064412. Michaud, C., Mengin-Lecreulx, D., van Heijenoort, J. and Blanot, D. (1990) Over-production, purification and properties of the uridine-diphosphate-N-acetylmuramoyl-L-alanyl-Dglutamate: meso-2,6-diaminopimelate ligase from Escherichia coli. Eur. J. Biochem. 194, 853-861. Miyakawa, T., Matsuzawa, H., Matsuhashi, M. and Sugino, Y. (1972) Cell wall peptidoglycan mutants of Escherichia coli K.-12: existence of two clusters of genes, mra and mrb, for cell wall peptidoglycan biosynthesis. J. Bacteriol. 112, 950-958. Morlot, C., Zapun, A., Dideberg, O. and Vemet, T. (2003) Growth and division of Streptococcus pneumoniae: localization of the high molecular weight penicillin-binding proteins during the cell cycle. Mol. Microbiol. 50, 845-855. Nanninga, N. (1991) Cell division and peptidoglycan assembly in Escherichia coli. Mol. Microbiol. 5, 791-795.

270

Mingorance and Tamames

Nelson, K. E., Clayton, R. A., Gill, R. S., Gwinn, M. L., Dodson, R. J., Haft, D. H., et al., (1999) Evidence for lateral gene transfer between Archaea and bacteria from genome sequence of Thermotoga maritima. Nature 399, 323-329. Nikolaichik, Y. A. and Donachie, W. D. (2000) Conservation of gene order amongst cell wall and cell division genes in Eubacteria, and ribosomal genes in Eubacteria and Eukaryotic organelles. Genetica 108, 1-7. Norris, V., Turnock, G. and Sigee, D. 1996 The Escherichia coli enzoskeleton. Mol. Microbiol. 19, 197-204. Overbeek, R., Fonstein, M., D'Souza, M., Pusch, G. D. and Maltsev, N. (1999) The use of gene clusters to infer functional coupling. Proc. Natl. Acad. Sci. USA 96, 2896-2901. Palacios, P., Vicente, M. and Sánchez, M. (1996) Dependency of Escherichia coli celldivision size, and independency of nucleoid segregation on the mode and level of ftsZ expression. Mol. Microbiol. 20, 1093-1098. Pinho, M. G. and Errington, J. (2003) Dispersed mode of Staphylococcus aureus cell wall synthesis in the absence of the division machinery. Mol. Microbiol. 50, 871-881. Pratviel-Sosa, F., Mengin-Lecreulx, D. and van Heijenoort, J. (1991) Over-production, purification and properties of the uridine diphosphate N-acetylmuramoyl-L-alanine:Dglutamate ligase from Escherichia coli. Eur. J. Biochem. 202, 1169-1176. Rothfield, L., Justice, S. and García-Lara, J. (1999) Bacterial cell division. Anna Rev. Genet. 33, 423-448. Rueda, S., Vicente, M. and Mingorance, J. (2003) Concentration and assembly of the division ring proteins FtsZ, FtsA and ZipA during the Escherichia coli cell cycle. J. Bacteriol. 185:3344-3351. Satta, G., Fontana, R. and Canepari, P. (1994) The two-competing site (TCS) model for cell shape regulation in bacteria: the envelope as an integration point for the regulatory circuits of essential physiological events. Adv. Microb. Physiol. 36, 180-245. Selinger, D. W., Saxena, R. M., Cheung, K. J., Church, G. M. and Rosenow, C. (2003) Global RNA Half-life analysis in Escherichia coli reveals positional patterns of transcript degradation. Genome Res. 13, 216-223. Siefert, J. L. and Fox, G.E. (1998) Phylogenetic mapping of bacterial morphology. Microbiology 144, 2803-2808. Siefert, J. L., Martin, K. A., Abdi, F., Widger, W. R. and Fox, G. E. (1997) Conserved gene clusters in bacterial genomes provide further support for the primacy of RNA. J. Mol. Evol. 45, 467-472. Snyder, L. A., Shafer, W. M. and Saunders, N. J. (2003) Divergence and transcriptional analysis of the division cell wall (dcw) gene cluster in Neisseria spp. Mol. Microbiol. 47, 431-441. Strieker, J., Maddox, P., Salmon, E. D. and Erickson, H. P. (2002) Rapid assembly dynamics of the Escherichia coli FtsZ-ring demonstrated by fluorescence recovery after photobleaching. Proc. Natl. Acad. Sci. USA 99: 3171-3175. Tamames, J. (2001) Evolution of gene order conservation in prokaryotes. Genome Biol. 2:0020.1-0020.11. Tamames, J., González, M., Mingorance, J., Valencia, A. and Vicente, M. (2001) Bringing gene order into bacterial shape. Trends in Genet. 17, 124-126. Tamames, J., Casari, G., Ouzounis, C. and Valencia, A. (1997) Conserved clusters of functionally related genes in two bacterial genomes. J. Mol. Evol. 44, 66-73. van Heijenoort, J. (1996) Murein synthesis. In: Neidhardt FC, Curtis III R, Ingraham JL, Lin ECC, Brooks Low K, Magasanik B, Reznikoff WS, Riley M, Schachter M, Umbarger

The dcw gene cluster

271

HE, editors. Escherichia coli and Salmonella. Cellular and molecular biology. Washington. ASM Press. Vicente, M., Gómez, M. J. and Ayala, J. A. (1998) Regulation of transcription of cell division genes in the Escherichia coli dcw cluster. Cell Mol. Life Sci. 54, 317-324. Vollmer, W., Höltje, J. V. (2001) Morphogenesis of Escherichia coli. Curr. Opin. Microbiol. 4, 625-633. Wätchtershäuser, G. (1998) Towards a reconstruction of ancestral genomes by gene cluster alignment. System Appl. Microbiol. 21, 473-477. Weiss, D. S., Chen, J. C., Ghigo, J.M., Boyd, D. and Beckwith, J. (1999) Localization of FtsI (PBP3) to the septal ring requires its membrane anchor, the Z ring, FtsA, FtsQ, and FtsL. J. Bacteriol. 181, 508-520. Xie, G., Keyhani, N. O., Bonner, C. and Jensen, R. A. (2003) Ancient origin of the tryptophan operon and the dynamics of evolutionary change. Microbiol. Mol. Biol. Rev. 67, 303-342. Zawadzke, L. E., Bugg, T.D. and Walsh, C. T. (1991) Existence of two D-alanine:D-alanine ligases in Escherichia coli: cloning and sequencing of the ddlA gene and purification and characterization of the DdlA and DdlB enzymes. Biochemistry 30, 1673-1682.

This page intentionally left blank

Index actin, 155, 158–61 Agrobacterium, 8 Aquifex aeolicus, 255 Bacillus subtilis, 8, 70, 104, 105–10, 196, 201, 226, 242–43 division site determination, 107 sporulation, 105 bacterial cytoskeleton, 7 Bdellovibrio, 121–22 bolA, 186 branching cells, 47 Buchnera aphidicola, 182 Buchnera sp., 177 Campylobacter jejuni, 253 Caulobacter crescentus, 104, 114–21, 181, 200, 245, 253 chromosome partitioning, 117 DNA replication, 116 flagellum assembly, 118 life cycle, 115 polar development, 119 cell differentiation, 103, 112 Cell division genes, 178, 239 cell division site, 79, 107 cell elongation genes, 182 COG Database, 176, 218 Comprehensive Microbial Resource, 176 connected gene neighbourhoods, 220 correlated mutations, 156 Corynebacterium, 8, 182 CtrA, 115 cytoskeletal proteins, 155 Chlamydia, 177, 266 dacA, 182, 185, 186 dacC, 185, 186 dacD, 185

dcw cluster, 6, 210, 220, 239, 242, 249, 261 bacterial shape, 255–59 conservation, 252–53 origin, 253–55 structure and function, 251–52 D-Cys, 37 ddlA, 252 ddlB, 252 DD-transpeptidation, 29 Deinococcus radiodurans, 182, 257 divIC, 180 DivIVA, 107, 181, 252 domains cytoplasmic membrane, 49 outer membrane, 48 EcoCyc, 217 EcoReg, 217 envA, 252 Escherichia coli, 7, 27, 59, 196, 203, 210, 212, 214, 220, 225, 240–42, 251 EzrA, 85. See first cell, 2, 15, 16, 21 FlhD, 113 fts genes, 178, 252 fts mutants, 79 FtsA, 80, 95–97, 95–97, 116, 155, 95–97, 167, 180, 240, 252, 255, 263 ftsB, 180, 240 FtsI, 89, 180, 185, 240, 242, 253, 255, 263 FtsK, 89, 180, 185, 240 FtsL, 89, 176, 180, 240, 253, 255 FtsN, 89, 91, 94, 180, 240 FtsQ, 89, 116, 180, 185, 240, 252, 263

273

274 ftsQAZ, 253, 255, 266 FtsW, 89, 180, 211, 240, 255 FtsZ, 80, 107, 115, 116, 123, 133–47, 155, 133–47, 180,211, 240, 251, 252, 255, 263 assembly cycle, 138 C terminus, 86 energetics of assembly, 141 folding, 134–36 GTPase, 139 interactions, 146 nucleotide switch, 140 polymer structure, 143 self-association, 137 FtsZ-ring, 43, 60, 69, 71 gene cluster, 203, 210, 211, 214 gene content, 219 gene order, 204, 209, 217, 219, 263 global regulators, 212 glycan strands cross-linking, 28 length, 30 orientation, 31 Haemophilus influenzae, 181, 214, 220, 225 Helicobacter pylori, 31, 220, 246, 253 helicobacteria, 257, 266 homeotic genes, 202 Hyphomonas, 124 indels, 5 inert murein. See inert peptidoglycan inert peptidoglycan, 7, 41, 185 inert poles, 7 INPARANOID, 219 KEGG, 219 LD-transpeptidation, 29 Lpp, 32. See Magnetococcus sp., 181, 266 map of the cell, 200

mapping cell morphology, 6 mbl, 184 Methanococcus jannaschii, 133 Methylococcus capsulatus, 266 Microbial Genome Database, 176, 219 MinC, 69, 83, 85, 107, 147, 181 MinCDE, 7, 152, 181 MinD, 7, 59, 64, 65, 67, 68, 72, 107, 134, 147 MinE, 7, 65, 66, 68, 72, 181 morphogenes, 173 mraW, 180, 240, 252, 255 mraZ, 180, 240, 252, 253, 255 MreB, 7–9, 155, 7–9, 184, 205, 266 mreBCD, 184 mur genes, 177, 240, 252, 253, 255, 261, 264, 266 murein polar cap, 40 segregation, 38–40 murein layers, 28 Mur-Fts complex, 265 muropeptides, 29 nanomachines, 200, 204 Neisseria gonorrhoeae, 243–44, 251,253 Neisseria sp., 181, 182 network motifs, 213 nucleoid occlusion model, 36 OmpA, 33 operating systems, 195, 198 operon, 210, 213, 217 prediction, 215–17 orthology, 217 PAL, 32 paralogy, 217 Pasteurella multocida, 181 pattern formation, 59 oscillating polar pattern, 64 periodic patterns, 63

275 stable patterning, 60 PBP2, 44, 182 PBP3, 42 PBP5, 185, 186 PBP6, 186 pbpA,182, 255 pbpG, 185 PBPs, 177, 185 phylogenetic profiles, 176 PodJ, 121 pole-to-pole oscillation, 7, 59, 64, 153 ppGpp, 242 prebiotic synthesis, 14–17, 20 precellular membranes, 18–20 protein interactions, 93, 224, 225, 263 databases, 230 gene fusions, 226 gene order, 225 phylogenetic profiling, 226 prediction, 225 Proteus mirabilis, 104, 112–14 RcsB, 242 regulon, 211, 217 RegulonDB, 212, 215, 216, 217 Rhizobium., 8 Rickettsia prowazekii, 253 rickettsiae, 181 RNA world, 17–18 RodA, 44, 182, 255 sacculus D-amino acid labelling, 37 growth, 33 interactions with membranes, 32 precursor-insertion, 35 radioactive precursors, 36 septal gaps, 41 structure, 28 SdiA, 242 septosome, 35, 39, 40, 42, 178

Sinorhizobium, 104, 123, 182 spirochetes, 266 Spo0A, 106 SpoIIAA, 109 SpoIIAB, 109 SpoIID, 108 SpoIIE, 107 SpoIIM, 108 SpoIIP, 108 spoVE, 242, 243 strand bias, 201 Streptococcus, 8 Streptococcus pyogenes, 182 Streptococcus thermophilus, 182 Streptomyces coelicolor, 104, 110-12, 181, 182 structural alignment, 156, 158, 164 sulphur related genes, 205 Synechocystis sp., 182 Thermotoga maritima, 253, 254 Thermus thermophilus, 257 tree-determinant residues, 156 tubulin, 133, 155, 164–66 Turing machine, 196, 198, 200 two competing sites model, 264–67 uber-operon, 220 von Neumann's conjecture, 197, 200 wall synthesis genes, 177 yabB. See mraZ yabC. See mraW YgbQ, 89. See FtsB ylmDEFGH, 252 ZapA, 85, 180 ZipA, 80, 86–89, 145, 180, 240. See Z-ring, 117, 266. See FtsZ-ring 109 106