Circadian Rhythms in Bacteria and Microbiomes 3030721574, 9783030721572

Citation preview

Carl Hirschie Johnson Michael Joseph Rust   Editors

Circadian Rhythms in Bacteria and Microbiomes

Circadian Rhythms in Bacteria and Microbiomes

Carl Hirschie Johnson • Michael Joseph Rust Editors

Circadian Rhythms in Bacteria and Microbiomes

Editors Carl Hirschie Johnson Department of Biological Sciences Vanderbilt University Nashville, TN, USA

Michael Joseph Rust University of Chicago Chicago, IL, USA

ISBN 978-3-030-72157-2 ISBN 978-3-030-72158-9 https://doi.org/10.1007/978-3-030-72158-9

(eBook)

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG. The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Dedications

Dedication to Dr. Carol Rae Andersson (May 5, 1965–January 24, 2009) She left us too soon!

Dr. Carol Andersson was the first researcher to join Dr. Susan Golden’s lab at Texas A&M University (in 1994) specifically to work on a nascent circadian rhythms project that developed from a collaboration with Golden, Dr. Carl Johnson (Vanderbilt University), and Drs. Takao Kondo and Mashiro Ishiura (initially at the National Institute for Basic Biology in Okazaki). Carol’s Ph.D. work at the Australian National University in Canberra had focused on regulation of hemoglobin genes in plants. She enjoyed jumping into the cyanobacterial world for her postdoctoral research and quickly mastered the genetics needed to explore components of the clock. In her words: “I just really like putting together little bits of DNA!” Her understated sense of humor provided daily joy to Dr. Golden. In addition to conducting foundational experiments and developing the circadian monitoring protocols used in the Golden lab, Carol served as an ambassador for the multi-lab project, spending time in the Kondo/Ishiura group with support of a Human Frontier Science Program grant (1996–1999) designed to facilitate interaction among the groups. She tremendously enjoyed interacting with her Japanese colleagues, who v

vi

Dedications

reflected later upon how gracefully she met the cultural challenge of pursuing her vegetarian diet during the visit. When Dr. Kondo visited Texas A&M in early 1995, armed with many clock-modifying clones identified by Dr. Ishiura, Carol quickly performed a beautiful Southern blot that showed all genes mapped to the same locus: the one we now revere as kaiABC. Because her efforts were so early in the hunt for the cyanobacterial circadian clock, she laid groundwork for projects that many others benefited from later. After her time in the Golden lab she carried her new love of circadian biology back to the plant science realm, to identify components of the Arabidopsis circadian clock in the lab of Dr. Steve Kay. Thereafter, she returned to her native Australia, where she worked in government policy related to food standards. She died sadly young of melanoma, remaining cheerful and finding joy in nature and pets throughout the challenging period of treatment and symptoms. Her memory still brings a smile to the faces of all who knew her. Contributed by Dr. Susan Golden, University of California, San Diego, CA, USA

Dedication to Dr. Yohko Kitayama (1976–2016)

Yohko Kitayama was born in August 1976, and it was April 1995 when she entered the Faculty of Science at Nagoya University. Our relationship began a few years later when she attended one of my lectures. I remember her enthusiasm and concentrated attention during that lecture. The following year, she chose to join my lab for her graduation project. Needless to say, Ms. Kitayama was a very talented person, and I felt very fortunate to have the opportunity to work with her, especially because she steadfastly maintained her own point of view and devoted herself conscientiously to her research. Dr. Kitayama was particularly good at molecular biological analysis, and her first research project led to the discovery of SasA by the two-hybrid method. Later, Dr. Kitayama participated in the most important studies of our group, such as the discovery and analysis of the phosphorylation of KaiC, and was a key member of our laboratory. The first research project in which she was the leader was published as a paper in EMBO J (2003), which analyzed the dynamics and phosphorylation of Kai proteins in cyanobacterial cells. This work, which led to her Ph.D. degree, was a

Dedications

vii

precursor to the quantitative analysis of Kai proteins and a stepping stone for the lab to move from molecular biology to biochemistry. Dr. Kitayama’s measurement of the absolute amount of Kai proteins in cells and her discovery of the nonlinearity in the regulation of KaiC phosphorylation by KaiB were pivotal to our success in 2005 when we reconstituted the circadian rhythm of KaiC phosphorylation with purified Kai proteins in vitro. After receiving her degree, Dr. Kitayama continued her research as a postdoc, Assistant Professor, and Lecturer, working toward understanding the core mechanism of the circadian clock. Among her many accomplishments, I would particularly like to mention two key studies, namely her study showing that the interaction of the six KaiC monomers within the KaiC hexamer regulates the phosphorylation cycle (Nature Communications, 2013) and her discovery that the circadian rhythm of transcription and translation continues even when the phosphorylation cycle of KaiC is suspended (Genes and Development, 2008). All of these were important discoveries that captured the essence of the circadian clock in cyanobacteria. Without her contributions, our understanding of the function of the circadian clock in cyanobacteria would not have progressed as far as it has. She was also indispensable to the management of our laboratory. At our laboratory meetings, it was my habit to look at Dr. Kitayama’s face first whenever I made a delicate proposal. As long as she was smiling, it was always fine. After she was diagnosed to have stomach cancer in May 2016, I could not do those things anymore. I visited her at the hospital many times after her diagnosis. The treatments must have been very painful, but because of the tender care from her parents, she never stopped smiling. Many people prayed for her to survive as long as possible, but three months later she passed away out of our reach. I would like to dedicate this book to her as a tribute to her dedication to the study of cyanobacterial circadian clocks and my gratitude to her. Tenderly contributed by Takao Kondo, Dr. Sc. University Professor, Nagoya University, Nagoya, Japan

To Takao Kondo on the Occasion of His Retirement It is a pleasure to dedicate this book to Dr. Takao Kondo on the event of his retirement. Because Takao is fundamentally a humble person, I do not think that he considered when he started his scientific training that his career might take him to the heights that he has accomplished. I seriously doubt that Takao ever thought that he would ever be awarded a prize of the stature of the Asahi Prize (in 2006), which is one of the most prestigious prizes a Japanese can attain, as it honors those who have outstanding accomplishments in the fields of arts and academics that greatly contributed to the development and progress of Japanese culture and society at large. Or even grander, that he would be granted a personal audience with the Emperor of Japan to honor (and explain!) his accomplishments. But fortunately, there is some

viii

Dedications

justice in this world, and Takao has been recognized for the seminal contributions he has made. In particular, his laboratory’s discovery of the in vitro KaiABC oscillation (Nakajima et al. 2005) has led to a reevaluation of circadian mechanisms in all organisms, and in particular to more cautious interpretations of the transcriptional and translational feedback loop model for circadian clocks (Tomita et al. 2005; Qin et al. 2010; Johnson 2010). Takao’s first publications were on circadian rhythms of ion fluxes in the duckweed Lemna (Kondo 1978; Kondo and Tsuzuki 1978; Kondo 1983). Lemna is a particularly interesting organism being a small angiosperm that exhibits photoperiodic flowering responses, and different strains of Lemna are short-day plants while other Lemna strains are long-day flowerers (Miwa et al. 2006). Therefore, like Colin Pittendrigh, Takao’s interest in circadian rhythms was inspired early by an interest in photoperiodism. And apparently he never lost his love for Lemna and the puzzle of its photoperiodic responses, since he came back to those research questions after his pioneering work with cyanobacteria (Miwa et al. 2006). However, at some point— probably attracted by the potential of its excellent genetics—Takao developed an interest for the potential of using the unicellular alga Chlamydomonas for analyzing circadian mechanisms. It was Takao’s venture into the study of Chlamydomonas that initiated my interaction and, ultimately, my friendship with Takao in the mid-1980s. At that time, I was a postdoc in Dr. J.W. (“Woody”) Hastings’ laboratory and I was attempting to continue the development of the Chlamydomonas circadian system that had begun with the studies of Dr. Victor Bruce (Bruce 1970). I had inherited Victor Bruce’s apparatus for measuring Chlamydomonas phototaxis rhythms, but I was not obtaining reproducibly precise rhythms. Takao has a real talent for designing apparatuses and writing computer programs for data acquisition and analysis, and he had transferred his expertise from the Lemna system to Chlamydomonas phototaxis rhythms. I became aware of the lovely data that Takao was producing from Chlamydomonas upon my first visit to Japan in 1984. I came back from that trip to Japan in a very enthusiastic mood about the quality of Takao’s phototaxis rhythms from Chlamydomonas, so I convinced Woody Hastings to invite Takao to Woody’s lab for three months in 1985 to initiate a collaboration on Chlamydomonas circadian rhythms. Takao’s visit to the USA was followed by my research visit to Takao’s lab for three months in 1986 at the National Institute for Basic Biology (NIBB) in Okazaki, Japan (where Takao was a Research Associate, Fig. 1). The primary project accomplished during that 1986 visit was to study the action spectroscopy of lightinduced phase resetting of the Chlamydomonas clock with Takao, a project that ultimately resulted in three publications about the effects of light and dark on the Chlamydomonas clock (Kondo et al.1991; Johnson et al. 1991; Johnson and Kondo 1992). During that 1986 visit to Okazaki, I learned that Takao is a subtle non-conformist. Some tip-offs to his underlying nature in 1986 were that he liked spicy food (unusual for most Japanese). Also, he was an early adopter of Apple/Macintosh computers in a time when most Japanese scientists used NEC computers (see the Apple computer in Fig. 1). His car was a Subaru when the most popular cars in Japan were Toyota,

Dedications

ix

Fig. 1 Takao Kondo sitting at his desk at the National Institute for Basic Biology, Okazaki, Japan (1986). Note the Apple computer on his desk behind him

Honda, Mazda, Mitsubishi, or Nissan (Subaru was unusual, but it was 4-wheel drive even then!). Finally, Takao was a mountain climber, not only in the Japan Alps but also in the Himalayas. Not your typical scientist! I think that the ability to think “outside the box” in scientific questions is related to some level of non-conformity. The Nobel Laureate Albert Szent-Györgyi was famously quoted as saying, “Research is to see what everybody else has seen, and to think what nobody else has thought” (this quote is thought to be a modification of an earlier statement by Arthur Schopenhauer). In Takao’s case, my observations of his atypical preferences and activities correlate with his “out-of-the-box” discoveries and hypotheses, one example being his hypothesis that the ATPase activity is the core pacemaking mechanism of the cyanobacterial clockwork rather than the KaiC phosphorylation and KaiABC complex cycles that his lab also discovered [(Terauchi et al. 2007) and also see the Chapter “Roles of Phosphorylation of KaiC in the Cyanobacterial Circadian Clock” in this volume]. My collaborative visit to Takao’s lab was followed by his coming to my lab in 1990–1991 with his family for a 10-month sabbatical (Fig. 2). I had assumed that Takao planned to continue our study of rhythms in Chlamydomonas during his sabbatical, so it was a surprise when Takao announced upon his arrival in the USA that he wanted to search for a new model system for studying circadian systems. Apparently, Takao had been conferring with his colleague Masahiro Ishiura, who had convinced Takao that an organism with more molecular genetic tools than Chlamydomonas would be better for an intensive circadian investigation. Therefore, Takao came to the USA to explore the possibility that E. coli or yeast might have a circadian clock. Takao’s interest in testing for rhythms in E. coli dovetailed nicely with a long-term interest of mine in the possibility that bacteria might harbor circadian rhythms. In the late 1970s when I was in graduate school, every other chronobiologist appeared to have concluded that prokaryotes were too “simple” to harbor circadian systems, but I became fascinated with the possibility that the environmental selective pressures experienced by many bacteria were as conducive to the evolution of a circadian clock as they were for eukaryotic cells (see

x

Dedications

Fig. 2 Takao and Carl (left and middle persons) on a balloon ride in Nashville, USA, in 1991, during Takao’s sabbatical visit to Carl’s lab

Introductory Chapter in this volume). I was particularly excited at the time by discoveries with Halobacterium by Walter Stoeckenius and others (Stoeckenius 1985, 1999) that revealed this bacterium used light energy to pump protons across its plasma membrane to generate a chemiosmotic gradient for synthesizing ATP. I reasoned “here was a bacterium that will really care whether it is light or dark and benefit from a timekeeper to anticipate dawn and dusk.” Therefore, as a graduate student in the laboratory of Colin Pittendrigh, I tried to measure daily rhythms of proton pumping in Halobacterium. Abysmal failure. Nevertheless, this graduatestudent passion prepared my mind to think freely with Takao’s mind toward the heretical possibility that bacteria might have clocks. Takao’s sabbatical experiments with E. coli and yeast did not bear direct fruit, but our minds were ready for heterodoxy. Part way through Takao’s sojourn in my lab, I attended the annual meeting of the American Society for Cell Biology (ASCB) in December of 1990. At this meeting, I presented a poster and by a stroke of luck, the neighboring poster was from the laboratory in Taiwan that had reported circadian rhythms of nitrogen fixation in the cyanobacterium Synechococcus RF-1 (Grobbelaar et al. 1986; Huang and Chow 1990). That poster’s presenter was Dr. Tsung-Hsien Chen, who was collaborating with Dr. Tan-Chi Huang. As Tsung-Hsien and I started to discuss circadian rhythms in algae as we stood by our posters, I forgot all about the rest of the meeting in my excitement about the

Dedications

xi

Taiwanese group’s research on cyanobacteria, which was the first persuasive demonstration of circadian rhythms in a prokaryote. Note that like Halobacterium, photoautotrophic cyanobacteria are also an organism for whom sunlight is the energy source, and therefore, the timing of the daily light/dark cycle would intuitively provide a strong selective pressure for evolution of a clock. Upon my return from the meeting, I convinced Takao that cyanobacteria were the new model system to investigate. Takao and I contacted Drs. Huang and Chen to initiate a collaboration, and they graciously mailed Synechococcus RF-1 to my lab. The idea was to clone the promoter for the nitrogenase gene that Dr. Huang had shown to be rhythmically expressed (Huang and Chow 1990), fuse it to a luciferase gene, and introduce it into the organism to create a luminescent organism whose rhythmic data could be automatically collected by the methods pioneered in Woody Hastings’ lab for the endogenously luminescent eukaryotic alga Gonyaulax (Taylor et al. 1982). A genetically malleable prokaryote whose rhythms could be non-invasively measured by a computerized apparatus for many cycles sounded like a winner! A pivotal event happened shortly thereafter; about a month after receiving the sample of Synechococcus RF-1, I happened to be in New York City and decided to “drop in” on Dr. Steve Kay and Andrew Millar, who were working with Dr. Nam-Hai Chua at Rockefeller University. During my visit with Steve and Andrew, I mentioned our plans to make a luminescence reporter strain of Synechococcus RF-1. Remarkably, Steve and Andrew had obtained a sample of Synechococcus RF-1 and were already underway in the process of making a luminescence reporter strain of this cyanobacterium! This was very depressing news, for at that time neither Takao nor I had much experience with molecular genetic techniques, so it seemed hopeless to compete on the identical approach with Steve and Andrew, who were molecular genetic “jocks.” The flaw in both of our plans, however, was that techniques for genetic transformation of Synechococcus RF-1 had not been worked out, but we all had hoped that the methods that had been developed for the transformation of other cyanobacterial species would be successful with Synechococcus RF-1. Though discouraged by the news from Steve and Andrew, Takao and I did not give up. We decided to drop further work with Synechococcus RF-1 and focus instead upon a cyanobacterial species for which molecular genetic techniques had already been developed. The problem was: what to assay as a circadian output in an uncharacterized strain? In Synechococcus RF-1, nitrogen fixation or nitrogenase activity was known to be rhythmic (Grobbelaar et al. 1986; Huang and Chow 1990), but in a new cyanobacterium it was anybody’s guess as to what rhythm to assay. Because my lab was doing a lot of 2D gel electrophoresis assays to discover circadian-regulated protein expression in Chlamydomonas at that time, we chose to look for rhythmic protein expression in a genetically malleable cyanobacterium. Once found, we reasoned that we could clone that gene’s promoter, make a luminescent reporter construct, and transform it into the organism, but we expected to be far behind the “Steve and Andrew team” that was using Synechococcus RF-1. In retrospect, this episode is reminiscent of advice to scientists from Dr. Efraim Racker, who wrote a book in 1976 about mitochondrial electron transport that included the

xii

Dedications

wise statement that “troubles can be good for you” scientifically (as long as you respond to them constructively!) (Racker 1976). In this case, the reason that these competitive troubles were good for us is that they led us to Dr. Susan Golden (then at Texas A&M University). I was calling cyanobacteriologists for advice about the optimal species/approach and everyone encouraged me to call Susan, who was working on the regulation of gene expression in response to changes of light intensity in the genetically tractable cyanobacterium Synechococcus elongatus PCC 7942. When I explained to Susan on the telephone what Takao and I had in mind, she casually mentioned some preliminary data of a postdoc in her lab that suggested the possibility of a daily rhythm in the expression of the key photosynthesis gene, psbAI. Even more exciting, Susan’s technician had already produced a luminescence reporter strain in which the bacterial luciferase gene set (luxAB) was fused to the psbAI promoter and transformed into S. elongatus. This was a windfall, and it established a collaborative team that was well on its way. Susan sent the PpsbAI::luxAB reporter strain of S. elongatus to us just before the end of Takao’s sabbatical in my lab. On his way back to Japan, Takao and his family visited Woody Hastings in Boston for a few days. As mentioned above, Woody had a custom apparatus that had been designed and built by Dr. Walter Taylor for the specific purpose of long-term, continuous, noninvasive measurements of circadian luminescence from Gonyaulax (Taylor et al. 1982). Takao had an opportunity to collect two days of data from the PpsbAI::luxAB reporter strain in Woody’s apparatus before returning to Japan. My remembrance of those data was that only the barest trace of an oscillation could be imagined (Johnson and Xu 2009). But Takao was not discouraged by the data he had collected in Woody’s lab! His talent for designing apparatuses and writing computer programs for data acquisition and analysis was now applied to the cyanobacteria system. After his return to Japan, Takao constructed a clever dual-channel luminometer that automatically closed a lid for a luminescence measurement and reopened the lid for white light irradiation to keep the photosynthetic cyanobacteria happy. Armed with suggestions from Woody about presenting the decanal substrate for bacterial luciferase and his homemade luminometer, Takao was able to measure rhythms that appeared to be entrainable to light/dark cycles. Encouraged by those results, Takao constructed a Japanese version of Woody’s multichannel luminescence monitoring system and was able to measure beautiful rhythms of psbAI promoter activity as assayed by the luxAB luminescence reporter (Kondo et al. 1993). In retrospect, the combination of the psbAI promoter fused to luxAB and expressed in S. elongatus was a happy coincidence, and it remains one of the most robustly rhythmic combinations in cyanobacteria, even after 25 years of extensive research. Along with the earlier studies on Synechococcus RF-1 by Huang and his collaborators (Grobbelaar et al. 1986; Huang and Chow 1990), our first paper on the S. elongatus rhythms (Kondo et al. 1993) established that prokaryotic cyanobacteria exhibit circadian rhythms. At that time, I was content to prove that bacteria were also members of the “circadian club,” and I was ready to refocus my attention on our studies with Chlamydomonas. But Takao had a much larger vision. He recognized that we had established a new model system for circadian studies that could go

Dedications

xiii

Fig. 3 The “Kondotron.” (a) Photograph of a Kondotron (this is the Kondotron in Carl’s lab; the original Kondotron is in Takao’s lab at Nagoya University). The turntable has positions for twelve 100 mm petri dishes. The CCD camera is to the left (and just barely out of view), on top of a feltencased baffle system to exclude incidental light during the imaging of bioluminescent cyanobacterial colonies on the petri dishes. Three circular fluorescent light fixtures are suspended above the turntable to provide light for photosynthesis to the cyanobacterial plates that are not being imaged. The turntable is rotated by a computer-controlled stepper motor that is underneath the Kondotron and therefore out of view. (b) A representative computer screen image of a plate of bioluminescent colonies of S. elongatus. Bioluminescent intensity is encoded by pseudocolor with blue (low intensity) to green (medium intensity) to red (high intensity). (c) Green S. elongatus colonies on an agar plate

further than most other model systems. To take full advantage of a prokaryotic system for analyzing circadian rhythms, a method to facilitate the identification of mutants was necessary. Therefore, with the first publication finished, Takao again drew from his apparatus-designing talent to develop an optimal mutant identification procedure for S. elongatus rhythms. Fortunately, single colonies of the PpsbAI::luxAB reporter strain are luminescent on agar plates so that Takao and Masahiro were able to visualize the rhythms of luminescence emitted by individual colonies using a CCD camera (Kondo and Ishiura 1994). Not content to observe rhythms from just one plate, however, Takao brought his unique talents to bear and designed a Macintosh computer-operated turntable apparatus that could monitor the rhythms of colonies on twelve 100 mm petri dishes simultaneously (Kondo et al. 1994; Johnson and Xu 2009). Up to 12,000 individual colonies could be screened in a single assay. Our laboratory respectfully calls this turntable-screening apparatus the “Kondotron” in honor of its inventor (Fig. 3). The development of the Kondotron was a breakthrough. It was the first highthroughput screening apparatus for circadian rhythms based on luminescence and enabled molecular genetic expertise to be directed toward developing methods for mutagenesis and complementation of S. elongatus that were specifically designed for circadian goals. With the Kondotron, Takao and Masahiro initiated an extensive mutagenesis project. Takao also generously shared the Kondotron technology with me when I was on a sabbatical with Takao in 1994, and we also contributed in a

xiv

Dedications

Fig. 4 A “meeting of the minds” where a group of the collaborators and Japanese students relaxed together at Takao Kondo’s mountain home in 1994. Persons in the photo from left to right are Takao Kondo, Setsuyuki Aoki (student), Jim Golden (husband of Susan Golden), Carl Johnson (kneeling), Susan Golden, Yusuke Watanabe (friend of Shinsuke Kutsuna who is pushing Carl), Shinsuke Kutsuna (student), and Masahiro Ishiura

minor way to identifying more clock mutants (98% of the early screening came from the Japanese team). This approach led to the isolation of many interesting mutants exhibiting short periods, long periods, and arrhythmia; the largest range of variation for circadian period mutants for any organism: from 16 h to 60 h (Kondo et al. 1994). The team of collaborators and our students began to grow (Fig. 4). In the fullness of time, screening for mutants with the Kondotron, complementation of mutants with wild-type DNA, and rescreening for “rescued” clones with the Kondotron again led to the identification of the key cyanobacterial clock genes, kaiA, kaiB, and kaiC (Ishiura et al. 1998). The three kai genes are immediately next to each other on the S. elongatus chromosome in a master clock gene cluster. Takao and Masahiro named the three-gene cluster “kai” for a Japanese word meaning “cycle or rotation number.” The identification of the kai genes was the key that unlocked the cascade of discoveries summarized in the Introduction and other chapters of this volume, and as they say, “the rest is history” (Johnson and Xu 2009). That cascade culminated in another example of Takao’s ability to “think outside the box,” which was the mind-boggling discovery of the first in vitro circadian rhythm composed of the proteins KaiA, KaiB, KaiC, and the energy source ATP (Nakajima et al. 2005). The coincidence of good fortune, “prepared minds,” clever ideas, and hard work transformed the Synechococcus elongatus system into–arguably–the best understood clockwork at the molecular level in any organism, even though it was the

Dedications

xv

newest comer to clock mechanism analyses. An absolutely indispensable element in that transformation was Takao’s unique combination of talents: 1. An excellent scientist by any standards 2. A technical innovator with a talent for designing apparatuses and writing computer programs for data acquisition and analysis 3. A non-conformist who has been willing and able to “think outside the box,” and have the courage to pursue and report unconventional hypotheses and results Takao’s papers are not merely of interest in terms of the cyanobacterial clock system; many of his papers should be read by ALL chronobiologists and others with interest in biological clock mechanisms. For me personally the following papers of Takao are especially important in terms of general concepts whose full implications remain unrealized: the likelihood of biochemical oscillators being the core circadian clockwork rather than transcription/translation feedback loops (Tomita et al. 2005; Nakajima et al. 2005), the core of the KaiABC oscillator being its ATPase activity rather than the rhythm of protein interactions and phosphorylation events (Terauchi et al. 2007), and the first truly molecular explanation for a circadian oscillator running with a period of 24 h and not something faster (Abe et al. 2015). It has been a privilege for me to have been a contributor to the development of a terrific clock-model system with Takao Kondo and our other initiating and ongoing collaborators (Susan Golden and Masahiro Ishiura). The triggering events were Takao’s desire to identify a new model system that was genetically malleable and the chance encounter that I had with Tsung-Hsien Chen at the ASCB poster session in 1990. The odyssey has had its frustrations and heartaches coupled with delightfully unexpected twists and turns (Johnson and Xu 2009). As in a statement attributed (perhaps incorrectly) to Albert Einstein, “If we knew what we were doing, it wouldn’t be called research, would it?” Compared to me at least, Takao Kondo appeared to know what he was doing. Contributed by Carl Hirschie Johnson Vanderbilt University, Nashville, TN, USA

Acknowledgments Note the fuller treatment of the early development and discoveries with the circadian clock in cyanobacteria in Johnson and Xu (2009). I thank our collaborators and mentors and our present/former lab members, but especially I thank the other three members of the “Quadrumvirate” (Takao Kondo, Susan Golden, and Masahiro Ishiura) for an exciting collaboration that was made possible by good science, good fortune, clever ideas, hard work, and friendship. Finally, I am grateful for research support from the National Institute of General Medical Science (NIH), the National Science Foundation, and the Human Frontiers of Science Program that made this project possible.

xvi

Dedications

References Abe J, Hiyama TB, Mukaiyama A et al (2015) Atomic-scale origins of slowness in the cyanobacterial circadian clock. Science 349: 312–316 Bruce VG (1970) The biological clock in Chlamydomonas reinhardtii. J Protozool 17:328–334 Grobbelaar N, Huang T-C, Lin HY, Chow TJ (1986) Dinitrogen-fixing endogenous rhythm in Synechococcus RF-1. FEMS Microbiol Lett 37:173–177 Huang T-C. and Chow T-J (1990) Characterization of the rhythmic nitrogenase activity of Synechococcus sp. RF-1 at the transcription level. Curr Microbiol 20:23–26 Ishiura M, Kutsuna S, Aoki S et al (1998) Expression of a gene cluster kaiABC as a circadian feedback process in cyanobacteria. Science 281:1519–1523 Johnson CH (2010) Circadian clocks and cell division: what’s the pacemaker? Cell Cycle 9:3864–3873 Johnson CH, Kondo T (1992) Light pulses induce “singular” behavior and shorten the period of the circadian phototaxis rhythm in the CW15 strain of Chlamydomonas. J Biol Rhythms 7:313–327 Johnson CH, Xu Y (2009) The decade of discovery: how Synechococcus elongatus became a model circadian system 1990–2000, Chap. 4. In: Ditty JL, Mackey SR, Johnson CH (eds) Bacterial circadian programs. Springer, pp 63–86 Johnson CH, Kondo T, Hastings JW (1991) Action spectrum for resetting the circadian phototaxis rhythm in the CW15 strain of Chlamydomonas. II. Illuminated cells. Plant Physiol 97:1122–1129 Kondo T (1978) Diurnal change in leakage of electrolytes from a long-day duckweed, Lemna gibba G3, under osmotic stress induced by water treatment. Plant Cell Physiol 19:985–995 Kondo T (1983) Phase shift in the potassium uptake rhythm of the duckweed Lemna gibba G3 caused by an azide pulse. Plant Physiol 73:605–608 Kondo T, Tsuzuki T (1978) Rhythm in potassium uptake by a duckweed, Lemna gibba G3. Plant Cell Physiol 19:1465–1473 Kondo T, Ishiura M (1994) Circadian rhythms of cyanobacteria: monitoring the biological clocks of individual colonies by bioluminescence. J Bacteriol 176:1881–1885 Kondo T, Johnson CH, Hastings JW (1991) Action spectrum for resetting the circadian phototaxis rhythm in the CW15 strain of Chlamydomonas. I. Cells in darkness. Plant Physiol 95:197–205 Kondo T, Strayer CA, Kulkarni RD et al (1993) Circadian rhythms in prokaryotes: luciferase as a reporter of circadian gene expression in cyanobacteria. Proc Natl Acad Sci USA 90:5672–5676 Kondo T, Tsinoremas NF, Golden SS et al (1994) Circadian clock mutants of cyanobacteria. Science 266:1233–1236

Dedications

xvii

Miwa K, Serikawa M, Suzuki S et al (2006) Conserved expression profiles of circadian clock-related genes in two Lemna species showing long-day and short-day photoperiodic flowering responses. Plant Cell Physiol 47:601–612 Nakajima M, Imai K, Ito H et al (2005) Reconstitution of circadian oscillation of cyanobacterial KaiC phosphorylation in vitro. Science 308:414–415 Qin X, Byrne M, Xu Y et al (2010) Coupling of a core post-translational pacemaker to a slave transcription/translation feedback loop in a circadian system. PLoS Biol 8:e1000394 Racker E (1976) A new look at mechanisms in bioenergetics. Academic Press, New York, 197p Stoeckenius W (1985) The rhodopsin-like pigments of halobacteria: light-energy and signal transducers in an archaebacterium. Trends Biochem Sci 10:483–486 Stoeckenius W (1999) Bacterial rhodopsins: evolution of a mechanistic model for the ion pumps. Protein Sci 8:447–459 Taylor W, Wilson S, Presswood R et al (1982) Circadian rhythm data collection using the Apple II microcomputer. J Interdiscip Cycle Res 13:71–79 Terauchi K, Kitayama Y, Nishiwaki T et al (2007) ATPase activity of KaiC determines the basic timing for circadian clock of cyanobacteria. Proc Natl Acad Sci USA 104:16377–16381 Tomita J, Nakajima M, Kondo T et al (2005) No transcription-translation feedback in circadian rhythm of KaiC phosphorylation. Science 307:251–254

Preface

Many fundamental biochemical and physiological processes were first discovered in bacteria, and after gaining a firm basic understanding in bacteria, the study of these processes was extended to eukaryotic cells and organisms. The phenomenon of circadian rhythmicity is an exception. The first report was of macroscopic observations in plants in 1729 and thereafter studies extended in the 1800s–1900s throughout the eukaryotic phyla to multicellular creatures as diverse as fungi, arthropods, and vertebrates. Starting in the 1950s, eukaryotic unicellular organisms were examined as well, especially algae and protozoa. These timekeepers were not only implicated in adaptive timing on the daily scale but in most cases are also the timing system that measures day length, thereby underlying seasonality and photoperiodism. Surprisingly, however, bacteria in the form of cyanobacteria (aka “blue-green algae”) did not join the organismal pantheon of Circadiana until the 1980s, and the story of how that happened is described in this volume in the Introduction and in the Dedication to Takao Kondo. And even though cyanobacteria are now well established to have genuine circadian timekeepers, the evidence for circadian clocks in bacteria other than the cyanobacteria remains murky. This book is “timely” (pun intended) because we now stand at a watershed moment in the investigation of circadian phenomena in prokaryotes. The study of circadian properties in cyanobacteria has been tremendously productive, for even though cyanobacteria are the latest of the major model systems to enter the field of chronobiology (see Introduction), we now know more about circadian mechanism and adaptive significance in cyanobacteria than in any other system, which is a testament to the vigor of methods that can be applied to bacteria. Once initiated with the luminescence reporter strain of Synechococcus elongatus, progress on understanding the clockwork mechanism has proceeded at breakneck speed, culminating in the discovery that the KaiA/B/C clock proteins could be used to reconstitute a circadian oscillator in vitro, the only known system where circadian rhythms can be studied outside of a living organism. Moreover, the Kai proteins proved to be readily crystallizable, thereby enabling the full power of structural and biochemical/ xix

xx

Preface

biophysical techniques to be deployed in determining the post-translational oscillator mechanism and its relationship to the transcriptional/translational feedback loop. A further advantage of the cyanobacterial system is that it has allowed direct access to questions about the adaptive significance of biological clocks. Testing even simple hypotheses about fitness can be challenging in higher organisms where many processes may become interdependent over evolutionary time so that the adaptive value of a specific trait divorced from the whole is difficult to ascertain. Cyanobacteria have yielded unambiguous answers to assessments of the relative role of various environmental factors in the natural selection of clocks by introducing the “competition assay” as a rigorous test of fitness. In so doing, the fitness experiments in cyanobacteria subsequently inspired similar fitness tests in plants and animals. As successful as the study of the circadian system in cyanobacteria has been, however, we believe that elucidation of cyanobacterial clocks is merely the overture to a large opera. In other words, this watershed moment marks the blossoming of the study of prokaryotic clocks from cyanobacteria to the much larger arena of all prokaryotes and in some cases the relationships of bacteria with eukaryotic hosts. Homologs of the kaiB and kaiC genes discovered in cyanobacteria that encode the core clock components KaiB and KaiC are found widely among Eubacteria and Archaea. While it is possible that kaiB and kaiC homologs perform different functions in other species of prokaryotes, it seems likely that the analogy of the role of these genes in the cyanobacterial timekeeping clock along with the fact that many bacteria exist in strongly rhythmic environments is indicative of a conservation of function as well as of sequence. Moreover, it is exciting to consider that the fundamental properties of biological timekeeping in other prokaryotes may be much more varied than those found among eukaryotes. There is already a hint of this diversity in that some prokaryotes such as Prochlorococcus and Rhodopseudomonas (both of which harbor kaiB and kaiC but not functional kaiA) may be exhibiting damped-oscillator or even hourglass-timer behavior rather than canonical selfsustained circadian rhythms. Even more titillating are the observations of possible rhythms in the gut bacterium Klebsiella (née Enterobacter) aerogenes, which together with the fascinating observations of rhythms in the gut microbiome opens a new realm of timekeeping interactions between eukaryotic hosts and prokaryotic residents. Because of the diversity of timing functions that appear to exist within the kai gene family, and because the selective pressures on microbial populations are often intense, we may be on the verge of obtaining precise answers to fundamental questions that have lingered since the beginnings of the study of biological rhythms. Namely, why are circadian clocks found in some organisms but not in others? Why do clocks have the properties that they do? How does a clock evolve from a non-oscillating ancestor? In that context, this book is a harbinger of topics that we foresee as the next great wave of discoveries in chronobiology: 1. Mechanism: what is the clockwork mechanism in cyanobacteria and what is the diversity of possible mechanisms that can adaptively keep time (including modeling)? (Chapters in this volume by: Miwa, Golden and LiWang, Rust, Iwasaki, Mori and Uchihashi, Axmann, Kim and Kim, Akiyama, Ito, Nishiwaki-Ohkawa, Byrne)

Preface

xxi

2. Evolution and adaptiveness of clocks; how do they enhance fitness? (Chapters in this volume by: Jabbur, Rust, Ito) 3. Harnessing clock properties to enhance bioproduction (Chapter in this volume by: Wang) 4. Ecology of clocks in nature; interactions within communities of clocks in nature (Chapters in this volume by: Hörnlein and Bohuis, Hevroni and Philosof) 5. Interactions among clocks, e.g., microbiomes and their hosts, in some cases with health implications (Chapters in this volume by: Heinemann, Tran Graniczkowska and Cassone) On a more personal level, this book marks another turning point. The scientists who initiated the study of circadian rhythms in cyanobacteria are reaching the stage of their careers and lives to “pass the baton” to younger researchers. This book is dedicated to three different researchers of cyanobacterial clocks, but primarily to Takao Kondo on the event of his retirement. And the other pioneers will soon follow (the pioneers of rhythms in Synechococcus RF-1, Drs. Tan-Chi Huang and TsungHsien Chen, have already retired 15 years ago). At this opportune moment, we are looking backward and looking forward. The cycle of investigators and investigations reinitiates, in a fashion reminiscent of a poem by Theodor Seuss Geisel (aka “Dr. Seuss”): “How did it get so late so soon? It’s night before it’s afternoon. December is here before it’s June. My goodness how the time has flewn. How did it get so late so soon?” Nashville, TN, USA Chicago, IL, USA

Carl Hirschie Johnson Michael Joseph Rust

Contents

The Bacterial Perspective on Circadian Clocks . . . . . . . . . . . . . . . . . . . . Carl Hirschie Johnson and Michael Joseph Rust Part I

1

The Circadian Clock System in Cyanobacteria: Pioneer of Bacterial Clocks

Around the Circadian Clock: Review and Preview . . . . . . . . . . . . . . . . . Takao Kondo A Retrospective: On Disproving the Transcription–Translation Feedback Loop Model in Cyanobacteria . . . . . . . . . . . . . . . . . . . . . . . . Hideo Iwasaki Mechanistic Aspects of the Cyanobacterial Circadian Clock . . . . . . . . . . Susan S. Golden and Andy LiWang

21

53 67

Mechanism of the Cyanobacterial Circadian Clock Protein KaiC to Measure 24 Hours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Kumiko Ito-Miwa, Kazuki Terauchi, and Takao Kondo

79

Oscillation and Input Compensation in the Cyanobacterial Kai Proteins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Michael Joseph Rust

93

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 Maria Luísa Jabbur, Chi Zhao, and Carl Hirschie Johnson Reasons for Seeking Information on the Molecular Structure and Dynamics of Circadian Clock Components in Cyanobacteria . . . . . . . . . 137 Shuji Akiyama

xxiii

xxiv

Contents

Single-Molecule Methods Applied to Circadian Proteins with Special Emphasis on Atomic Force Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . 147 Tetsuya Mori and Takayuki Uchihashi Diversity of Timing Systems in Cyanobacteria and Beyond . . . . . . . . . . 179 Nicolas M. Schmelling, Nina Scheurer, Christin Köbler, Annegret Wilde, and Ilka M. Axmann An In Vitro Approach to Elucidating Clock-Modulating Metabolites . . . 203 Pyonghwa Kim and Yong-Ick Kim Damped Oscillation in the Cyanobacterial Clock System . . . . . . . . . . . . 221 Hiroshi Ito, Yoriko Murayama, Naohiro Kawamoto, Motohide Seki, and Hideo Iwasaki Roles of Phosphorylation of KaiC in the Cyanobacterial Circadian Clock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241 Taeko Nishiwaki-Ohkawa Reprogramming Metabolic Networks and Manipulating Circadian Clocks for Biotechnological Applications . . . . . . . . . . . . . . . . . . . . . . . . 259 Bo Wang, Jamey D. Young, and Yao Xu Insights from Mathematical Modeling/Simulations of the In Vitro KaiABC Clock . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297 Mark Byrne Part II

Circadian Phenomena in Microbiomes/Populations and Bacteria Besides Cyanobacteria

Basic Biology of Rhythms and the Microbiome . . . . . . . . . . . . . . . . . . . 317 Melina Heinemann, Karina Ratiner, and Eran Elinav Disease Implications of the Circadian Clocks and Microbiota Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329 Laura Tran, Christopher B. Forsyth, Faraz Bishehsari, Robin M. Voigt, Ali Keshavarzian, and Garth R. Swanson Circadian Organization of the Gut Commensal Bacterium Klebsiella aerogenes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351 Kinga B. Graniczkowska and Vincent M. Cassone Daily Rhythmicity in Coastal Microbial Mats . . . . . . . . . . . . . . . . . . . . . 365 Christine Hörnlein and Henk Bolhuis Daily and Seasonal Rhythms of Marine Phages of Cyanobacteria . . . . . 387 Gur Hevroni and Alon Philosof

The Bacterial Perspective on Circadian Clocks Carl Hirschie Johnson and Michael Joseph Rust

Abstract Prokaryotes were long thought to be incapable of expressing circadian (daily) rhythms. Research on nitrogen-fixing cyanobacteria in the 1980s squashed that dogma and showed that these bacteria could fulfill the criteria for circadian rhythmicity. Development of a luminescence reporter strain of Synechococcus elongatus PCC 7942 established a model system that ultimately led to the best characterized circadian clockwork at a molecular level. The conclusion of this chapter lists references to the seminal discoveries that have come from the study of cyanobacterial circadian clocks.

1 The “No Clocks in Proks” Dogma Circadian (daily) rhythms are fundamental control systems that regulate a wide variety of biological phenomena, such as behavior, development, metabolism, and gene expression (Dunlap et al. 2004). These “biological clocks” help organisms adapt to the dramatic daily transformation of their environment, particularly daily changes in light, temperature, humidity, and so on. Circadian rhythms are defined by three major phenomenological criteria that are well established (see Box 1). The first criterion is that circadian rhythms persist in constant conditions (i.e., constant temperature and either constant light or constant darkness) with a period of approximately 24 h. The second criterion is that these rhythms are temperature compensated, so that they proceed at almost the same rate (¼ same period) within a permissive range of constant ambient temperature. Finally, the third criterion is that these endogenous rhythms of approximately 24 h can be entrained to exactly

C. H. Johnson (*) Department of Biological Sciences, Vanderbilt University, Nashville, TN, USA e-mail: [email protected] M. J. Rust Departments of Molecular Genetics & Cell Biology and of Physics, University of Chicago, Chicago, IL, USA e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_1

1

2

C. H. Johnson and M. J. Rust

24 h by daily cues in the environment, such as light/dark cycles, temperature cycles, social cues, etc. Box 1 Three Criteria Define Circadian Rhythmicity Three major phenomenological criteria, rather than molecular mechanisms, define circadian biological clocks: 1. Circadian clocks display free-running rhythms that oscillate with a selfsustained period that is near, but not exactly, 24 h when environmental conditions such as temperature and either light or darkness are constant. 2. Circadian clocks exhibit “temperature compensation” which ensures that the period of the free-running rhythm remains nearly constant at different constant ambient temperatures; thus, such timing systems remain relatively accurate by not running too fast on warm days or too slowly on cold days. 3. Circadian rhythms entrain to relevant environmental 24-h cycles, e.g., when an organism is exposed to a particular 24-h light-dark cycle, its rhythm will align with a specific phase relationship to that environmental cycle and take on a period of exactly 24 h.

Prior to the 1980s, chronobiologists had decided that there were “no clocks in proks” (aka prokaryotes), so the dogma became that circadian rhythms were exclusively expressed by eukaryotic organisms. A few dissenting reports (Halberg and Conner 1961; Sturtevant 1973) had not been convincing. For example, a suggestion by Franz Halberg that E. coli might have a circadian clock (Halberg and Conner 1961), was based on a statistical analysis of an old study by Rogers and Greenbank (1930) in which E. coli cells were grown in rich nutrient medium in an apparatus whose temperature control was almost certainly poor. Therefore, the possible daily trends extracted by the statistical analyses (Halberg and Conner 1961) were likely to be merely a result of a diurnal cycle of temperature, to which the growth rate of E. coli is exquisitely sensitive. This and other unconvincing studies led chronobiologists to conclude that prokaryotic cells, either unicellular or multicellular, were too

The Bacterial Perspective on Circadian Clocks

3

“simple” to express circadian behavior (Ehret and Wille 1970; Schweiger and Schweiger 1980; Kippert 1986), despite the fact that there were almost no published reports of rigorous tests of the proposition (One exception was that of Taylor [1979] who did perform a rigorous test). Some models of the circadian mechanism in eukaryotic cells relied on this “no clocks in proks” dogma as a necessary factor so that the models implicated eukaryotic-specific features (e.g., organelles, eukaryoticspecific transcription) as intrinsic parts of the circadian clockwork (Ehret and Trucco 1967; Schweiger and Schweiger 1977; Kippert 1986). Kippert argued as late as 1991 that the evidence for circadian rhythms in prokaryotic cyanobacteria was equivocal (Kippert 1991); his concerns were based on the possible influence of an unidentified exogenous factor. However, that possibility was ruled out by experiments in which rhythms were measured simultaneously from cyanobacterial cultures that had been entrained to light/dark cycles that were 180 out of phase, and that continued to freerun out of phase when released into LL (Kondo et al. 1993; Aoki et al. 1995). There is no doubt at this time that at least cyanobacteria among bacteria have a bona fide circadian clockwork mechanism. While the adaptive significance of a daily clock to the photosynthetic cyanobacteria may now seem self-evident in retrospect (see next section), it might still be argued that a 24-h clock is inconsistent with the rapid-growth lifestyle of many non-photosynthetic prokaryotes. Is it possible that other prokaryotes have circadian clocks? This is a challenging question, with momentous implications for understanding the early evolution of circadian rhythmicity. Many bacteria experience substantial environmental changes during the day and night. For example, most free-living bacteria are exposed to daily cycles of light and darkness and/or temperature (Fig. 1). Even the gut microbiota is exposed to daily changes in the intestinal environment, as most animal hosts eat on a daily schedule—usually during the day for diurnal animals and during the night for nocturnal animals—which creates a daily rhythm of feast and fast in the digestive tract (Johnson et al. 2017; Jabbur et al. 2021). Anticipating daily changes might also facilitate protective responses in bacteria that are exposed to the deleterious effects of sunlight (Fig. 1). The ultraviolet (UV) component of sunlight damages DNA, but even visible components of sunlight are absorbed by cytochromes in the electron-transport chain and affect metabolism (Robertson et al. 2013). The effects of sunlight inspired the “Escape From Light” hypothesis for the original evolution of circadian clocks, which proposed that the predominant selective pressure was the daily cycles of light and darkness in which light impaired growth and metabolism, damaged nucleic acids, etc. (Pittendrigh 1965, 1993). Therefore, the fact that many bacteria experience strongly rhythmic environments means that many bacteria might have evolved a circadian system in respond to these selective forces. Compared with clocks in eukaryotes, bacterial clocks may be more similar to the ancestral clocks that first evolved on Earth, and therefore the most informative in terms of understanding the evolution of circadian systems (Johnson et al. 2017; Jabbur et al. 2021). We think that there is an excellent chance that circadian clocks will be found in bacteria other than cyanobacteria. One line of evidence is that the kaiB and kaiC genes of the cyanobacterial kaiABC clock gene cluster (Ishiura et al.

4

C. H. Johnson and M. J. Rust

Fig. 1 Many bacteria are exposed to daily selective pressures. Free-living bacteria can be exposed to daily cycles of light and temperature that affect the viability (for example, exposure to UV light) and/or provide energy (for example, through photosynthesis). Even the gut microbiota is often exposed to daily cycles of nutrients, owing to the rhythmic eating habits of the host and temperature, as most animal hosts have daily rhythms of body temperature that are metabolically controlled in endotherms and behaviorally controlled in ectotherms. These same bacteria may be exposed to daily environmental cycles of light and temperature following excretion from the gut (Modified from Johnson et al. 2017)

1998) have homologs among many other eubacterial and Archaeal species, including playing important roles in signal transduction among Archaea (Makarova et al. 2017). It is also quite reasonable that circadian systems may have evolved in noncyanobacterial prokaryotic species independently of the kai system; therefore, the absence of kaiBC in any given bacterial species should not be taken to be evidence of the absence of a circadian clock. The key to discover circadian clocks in noncyanobacterial prokaryotes will be to find the proper conditions and to discover an appropriate parameter to measure. For example, E. coli cells may not express a circadian rhythm under optimal growth conditions, such as those encountered in the rich intestinal environment or in the laboratory. But E. coli cells can live under very different conditions in nature, for example, in soil or water after excretion from a host, barely surviving until they are again ingested (Fig. 1). In these sub-optimal conditions, a daily clock could again become adaptive. A final evolutionary aspect to consider is whether the ancestral relationship of cyanobacteria to the progenitor of

The Bacterial Perspective on Circadian Clocks

5

the chloroplast of higher plants (Sánchez-Baracaldo et al. 2017) means that the clock mechanism of a cyanobacterium-like prokaryote was passed by endosymbiosis into its host? While this remains an open question, at this time there is no evidence for such a postulate since no kai gene homologs have been found in the nuclear or chloroplast genomes of higher plants.

2 Data Dethroned the Dogma: Circadian Rhythms in Cyanobacteria In the mid-1980s, several papers were published that began to unseat the “no clocks in proks” dogma (Stal and Krumbein 1985; Mitsui et al. 1986; Grobbelaar et al. 1986). Those researchers reported that diazotrophic cyanobacteria (unicellular Synechococcus sp. Miami BG 43511 and 43522; unicellular Synechococcus sp. RF-1, and filamentous but non-heterocystous Oscillatoria) display daily rhythms of nitrogen fixation in light/dark cycles and in constant light. One of these studies showed that the cyanobacteria also express a daily rhythm of photosynthesis which was 180 out of phase from the nitrogen-fixation rhythm; photosynthesis peaked at midday, while nitrogen fixation occurred at night (Mitsui et al. 1986). These reversed-phase relationships are particularly interesting from the perspective of adaptive significance. The nitrogen-fixing enzyme nitrogenase is inactive in the presence of even tiny amounts of oxygen, which creates a major design problem for photosynthetic diazotrophs like nitrogen-fixing cyanobacteria whose photosynthesis releases oxygen (Gallon 1992). Nitrogen-fixing organisms have contrived differing solutions to this problem, notably the spatial separation of photosynthesis and nitrogen fixation in filamentous cyanobacteria that develop specialized nonphotosynthesizing cells (heterocysts) for fixing nitrogen. Spatial segregation may not be a practical solution in tiny unicellular bacteria, however. As one of several tactics to accomplish mutually incompatible tasks, some unicellular cyanobacterial strains separate photosynthesis and nitrogen fixation temporally (Gallon 1992; Mitsui et al. 1986; Grobbelaar et al. 1987; Schneegurt et al. 1994). Rhythms of both photosynthesis and nitrogen fixation continued in constant light and were entrained by prior light/dark cycles (Stal and Krumbein 1985; Mitsui et al. 1986; Grobbelaar et al. 1986; Huang and Grobbelaar 1995), so two of the crucial properties of circadian rhythmicity were satisfied. The third crucial circadian characteristic—temperature compensation—was initially demonstrated in several strains within the genus Synechococcus. In the marine Synechococcus WH7803, Sweeney and Borgese (1989) found temperature-compensated rhythms of cell division. In the freshwater Synechococcus RF-1 isolated from rice fields, the Taiwanese group reported temperature-compensated rhythms of nitrogen fixation and amino acid uptake (Huang et al. 1990; Chen et al. 1991). Using bacterial luciferase reporters of gene expression (Liu et al. 1995a; Andersson et al. 2000), temperature compensation and the other salient properties of circadian

6

C. H. Johnson and M. J. Rust

rhythms were demonstrated in Synechococcus elongatus PCC 7942 (Kondo et al. 1993), Synechocystis PCC 6803 (Aoki et al. 1995), Anabaena PCC 7120 (Kushige et al. 2013), and Thermosynechococcus elongatus (Onai et al. 2004). The temperature compensation exhibited by the thermophilic Thermosynechococcus is particularly spectacular: a Q10 of 1.08 over a range of more than 30  C, from 30  C to 60  C (Onai et al. 2004)!

3 Establishing the Synechococcus elongatus PCC 7942 Model System The aforementioned studies on the species Synechococcus RF-1, Synechococcus WH7803, Synechococcus 43511/43522, and Oscillatoria demonstrated that the canonical properties of circadian clocks (Box 1) were demonstrable in prokaryotic cyanobacteria. Strangely, the paradigm-shifting significance of those studies was not widely appreciated among the Chronobiology community. It was at that time that Takao Kondo, Carl Johnson, Susan Golden, and Masahiro Ishiura entered the study of cyanobacterial clocks. Other chronobiologists started to take seriously the idea that cyanobacteria were ready to join the menagerie of circadian organisms, probably because Carl Johnson and Takao Kondo were already “card-carrying” chronobiologists who were well known in the field (which is a sad reflection on the predominance of personalities over evidence in persuasion). Kondo/Johnson/Golden/Ishiura recognized that rapid future progress would depend upon using an organism with excellent genetic techniques and automatable data collection. The previously used strains (Synechococcus RF-1 et al.) did not have well-established genetic techniques, and moreover, the rhythms that were measured (nitrogen fixation, amino acid uptake, cell division, et al.) required laborious manual measurements. Consequently, while the earlier studies demonstrated the existence of circadian clocks in prokaryotes, they did not identify an optimal organism for an extensive molecular/genetic analysis of the clock mechanism. To reap the technical benefits that prokaryotes potentially offer for an in-depth analysis of clock mechanisms and evolution, Kondo/Johnson/Golden/Ishiura identified a cyanobacterium that was amenable to molecular/genetic analyses, and that could be engineered to express circadian rhythms of a parameter that can be assayed continuously for many cycles by an automated system. Susan Golden provided an optimal cyanobacterium and genetic techniques, and Kondo/Johnson provided the clock background and technical expertise for rhythm collection (Johnson and Xu 2009; Dedication to Takao Kondo in this volume). The “optimum cyanobacterium” that Susan Golden contributed was the genetically tractable Synechococcus elongatus PCC 7942 with the Vibrio harveyi luciferase gene cassette (luxAB) expressed under control by the promoter for the photosystem II gene, psbAI (Kondo et al. 1993; Andersson et al. 2000). The luminescence rhythm expressed by this reporter strain in liquid cultures or from single colonies on agar medium was

The Bacterial Perspective on Circadian Clocks

7

assayable by an automated monitoring system such that the luminescence rises during the day and falls during the night (Johnson and Xu 2009; Dedication to Takao Kondo in this volume). The luminescent glow rhythm was an accurate reporter of gene expression, confirming the hypothesis that this glow rhythm reflects circadian control over the promoter of the psbAI gene (Liu et al. 1995a). While it could be argued that the first paper on circadian rhythms in Synechococcus elongatus 7942 (Kondo et al. 1993) did not add very much conceptually to the demonstration of circadian clocks in prokaryotes that had been accomplished by the aforementioned studies in Synechococcus RF-1, Synechococcus WH7803, Synechococcus 43511/43522, and Oscillatoria (Hall and Rosbash 1993), several key characteristics of the Synechococcus elongatus system were the harbinger of a terrific new model system: 1. The luxAB luciferase reporter enabled the automated recording of rhythms and relatively high throughput screening of mutant strains. 2. S. elongatus 7942 has only one copy of the kaiABC cluster, which simplified circadian genotype/phenotype relationships (many other cyanobacterial species have several copies of kaiB and/or kaiC, which would have complicated the genetic analyses of the clockwork if they had been the first system to be tested). 3. Genetic tools (developed by Susan Golden’s lab and other laboratories): S. elongatus undergoes homologous recombination, which allows precise genetic targeting and complementation. S. elongatus is naturally competent and therefore is easily transformed. Both transformation and conjugation techniques had been developed for S. elongatus when the circadian analyses started. S. elongatus was optimal for a saturational mutagenesis approach because it has a small genome (smaller than E. coli !) for which the complete genome sequence became available early in the circadian analyses. 4. S. elongatus is easy to grow in liquid cultures or on agar plates and exhibits “simple” prokaryotic genetic organization. The coincidence of good fortune, clever ideas, “prepared minds,” and hard work has transformed the Synechococcus elongatus system into the best-understood clockwork at the molecular level in any organism. Jacques Monod, in his Nobel Prize acceptance speech, said, “The ambition of molecular biology is to interpret the essential properties of organisms in terms of molecular structures” (Monod 1966). That is an ambitious goal. But, in the case of the circadian clockwork in cyanobacteria, we are VERY close to attaining Monod’s goal. With complete 3-D structures of the key clock proteins individually and in complexes, an in vitro oscillator, and molecular explanations for why the oscillator has a period close to ~24 h (and not, for example, 8 h), we are on the threshold of a truly molecular explanation for the ticking of a circadian clock. The conclusion of this chapter is a listing of some references to the key discoveries that have come from the study of the (predominantly) S. elongatus circadian clock system with a special emphasis on the seminal studies that initiated that

8

C. H. Johnson and M. J. Rust

approach/topic (this is an incomplete and subjective listing and some key references may have been inadvertently neglected; if so, please forgive us). Topics of Bacterial Clocks and Selected Key References Demonstration of Circadian Clocks in Cyanobacteria. Cyanobacteria were the first and still the only prokaryotes definitively known to express bona fide circadian rhythms. This volume: Introduction; Dedication to Takao Kondo Grobbelaar et al. (1986) Kondo et al. (1993) Johnson et al. (1996) Johnson and Xu (2009) Possible Circadian Clocks in Other Bacteria. Other prokaryotes besides cyanobacteria have been suggested to harbor circadian clocks (or “proto” clocks), but this conclusion is not yet widely accepted. This volume: Introduction; chapters by Jabbur et al.; Axmann; Cassone Min et al. (2005) Ma et al. (2016) Paulose et al. (2016) Johnson et al. (2017) Sartor et al. (2019) Eelderink-Chen et al. (2021) Communities: The interface between biological clocks in communities, such as within the microbiome and in the natural environment. The temporal relationships among different species living in communities is an under-studied but fascinating topic. Can clock information be passed to cells/organisms of other species? Beyond entrainment, how do circadian clocks in microbes interact with natural environments? This volume: chapters by Jabbur et al.; Elinav; Keshavarzian; Hörnlein et al.; Hevroni and Philosof Thaiss et al. (2014) Voigt et al. (2014) Leone et al. (2015) Sartor et al. (2019) Hellweger et al. (2020) Clocks in Cyanobacteria; Establishing the Synechococcus elongatus System. As discussed in this Chapter, the first circadian clocks discovered in prokaryotes were in cyanobacteria, but not in S. elongatus. Nevertheless, once the luciferase reporter system was established in S. elongatus, its advantageous characteristics enabled rapid progress in understanding the clock system of cyanobacteria. This volume: Introduction; Dedication to Takao Kondo pp. XXX Kondo and Ishiura (1994) Kondo et al. (1993)

The Bacterial Perspective on Circadian Clocks

9

Kondo et al. (1994) Johnson and Xu (2009) Clocks in Cyanobacteria; Identification of Core Clock Genes. The initial EMS-generated period mutants in S. elongatus ranged from as short as 16 h to as long as 60 h, which is the largest range of circadian period phenotypes yet described from any organism. These mutations were localized within the kaiABC cluster by genetic complementation of mutants with wild-type DNA. Homologs to kaiB and kaiC have been found widely among Eubacteria and Archaea. This volume: chapters by LiWang and Golden; Axmann Kondo et al. (1994) Ishiura et al. (1998) Dvornyk et al. (2003) Schmelling et al. (2017) Clocks in Cyanobacteria; Structural Biology of Core Clock Proteins. The core cyanobacterial clock proteins KaiA, KaiB, and KaiC were the first circadian clock proteins for which full-length 3-D structures were derived, and now 3-D structures of the clock protein complexes are also known. The structure of KaiB undergoes an almost unprecedented fold-switching to accomplish its binding to KaiC. This volume: chapters by LiWang and Golden; Akiyama; Rust Mori et al. (2002) Pattanayek et al. (2004) Garces et al. (2004) Ye et al. (2004) Vakonakis and LiWang (2004) Hitomi et al. (2005) Johnson et al. (2008) Chang et al. (2015) Tseng et al. (2017) Snijder et al. (2017) Clocks in Cyanobacteria; In Vitro Oscillator. The in vitro kaiABC oscillator is certainly the most unexpected and important finding that has come from the study of cyanobacterial circadian rhythms. This volume: chapters by LiWang and Golden; Miwa et al.; Akiyama; Mori and Uchihashi; Iwasaki; Rust Nakajima et al. (2005) Tomita et al. (2005) Murayama et al. (2017) Clocks in Cyanobacteria; Core PTO Mechanism. The in vitro kaiABC oscillator is thought to correspond to a Post-Translational Oscillator (PTO) that functions in vivo as a mass-action biochemical oscillator. The mechanism of this PTO involves the KaiA-stimulated autophosphorylation of KaiC, assembly/disassembly of clock

10

C. H. Johnson and M. J. Rust

protein complexes, mechanisms of interhexamer synchronization (monomer exchange, etc.), ATPase activity, KaiC dephosphorylation via ATP synthase activity, and undoubtedly as yet undiscovered delights. This volume: chapters by LiWang and Golden; Miwa et al.; Nishiwaki; Akiyama; Mori and Uchihashi; Iwasaki; Rust Xu et al. (1999) Iwasaki et al. (1999) Nishiwaki et al. (2000) Xu et al. (2000) Iwasaki et al. (2002) Xu et al. (2004) Kageyama et al. (2006) Terauchi et al. (2007) Nishiwaki et al. (2007) Rust et al. (2007) Mori et al. (2007) Ito et al. (2007) Qin et al. (2010b) Nishiwaki and Kondo (2012) Egli et al. (2012) Kitayama et al. (2013) Abe et al. (2015) Mori et al. (2018) Clocks in Cyanobacteria; Relationships Between the PTO and the TTFL. In vivo, the Post-Translational Oscillator (PTO) is embedded within and controls a Transcriptional/Translational Feedback Loop that reinforces the robustness of the emergent clock system and controls output pathways. This volume: chapters by LiWang and Golden; Miwa et al.; Akiyama Ishiura et al. (1998) Xu et al. (2003) Tomita et al. (2005) Nakajima et al. (2005) Kitayama et al. (2008) Qin et al. (2010a) Teng et al. (2013) Clocks in Cyanobacteria; Identification of Input Pathways. Unlike clock photobiology in eukaryotes, there appears to not be a specific photoreceptor and signal transduction pathway that entrains the cyanobacterial clock in vivo. Rather, the current data suggest that changes in intracellular redox and/or energy metabolism regulated by photosynthesis are the entraining cues. This volume: chapters by Kim; LiWang and Golden; Miwa et al.; Rust Schmitz et al. (2000)

The Bacterial Perspective on Circadian Clocks

11

Rust et al. (2011) Kim et al. (2012) Pattanayak et al. (2015) Clocks in Cyanobacteria; Identification of Output Pathways. Global gene expression is linked to the KaiABC oscillator via coupling kinase pathways involving SasA, LabA, CikA, and the transcriptional factor RpaA. This volume: chapters by LiWang and Golden; Miwa et al. Iwasaki et al. (2000) Takai et al. (2006) Taniguchi et al. (2010) Markson et al. (2013) Clocks in Cyanobacteria; Global Gene Regulation and Chromosomal Topology. The global control of promoters by the circadian system is accompanied by a circadian rhythm in supercoiling/compaction of the entire chromosome that could modulate/potentiate the global circadian control of gene expression in cyanobacteria. Because of these torsional effects, even heterologous promoters are rhythmically active in S. elongatus. This volume: chapters by LiWang and Golden; Miwa et al. Kondo et al. (1993) Liu et al. (1995a) Liu et al. (1995b) Kucho et al. (2005) Smith and Williams (2006) Woelfle et al. (2007) Ito et al. (2009) Vijayan et al. (2009) Xu et al. (2013b) Clocks in Cyanobacteria; Clocks and Cell Division. While the timing of cell division within the day is timed by the circadian system in cyanobacteria, even when the average doubling time is much faster than 24 h (as fast as 5–6 h), the circadian clock ticks along stably at the endogenous free-running period. This has been shown in both populations of cells and in individual cells, and it means that the circadian clock is coupled to the timing of cell division in a one-way control circuit whereby the circadian clockwork is mostly unperturbed by DNA replication or cell division. This is consistent with the hypothesis that the PTO is the core circadian pacemaker as a stochastic mass-action oscillator that is not dependent upon transcription (transcription could be modulated by the presence versus absence of DNA replication). Mori et al. (1996) Kondo et al. (1997) Mori and Johnson (2001) Mihalcescu et al. (2004)

12

C. H. Johnson and M. J. Rust

Dong et al. (2010) Johnson (2010) Clocks in Cyanobacteria; Adaptive Fitness and Evolution. S. elongatus was the first circadian system in any organism to provide pioneering rigorous tests of adaptive fitness by competition assays. This volume: chapters by Jabbur et al. Ouyang et al. (1998) Woelfle et al. (2004) Xu et al. (2013a) Lambert et al. (2016) Diamond et al. (2017) Puszynska and O’Shea (2017) Clocks in Cyanobacteria; Modeling. The combination of quantitative data and a small number of core components in the S. elongatus oscillator has created a remarkably productive conversation between experimental science and mathematical modeling. This literature has become quite extensive, so rather than attempting to be exhaustive, we list a small subset of papers that address many of the biochemical concepts that have been used in creating models of the KaiABC system. This volume: chapters by Byrne; Ito et al.; Rust van Zon et al. (2007) Mori et al. (2007) Zwicker et al. (2010) Qin et al. (2010a) Hatakeyama and Kaneko (2012) Paijmans et al. (2017) Clocks in Cyanobacteria; Biotech Applications. Cyanobacteria are being engineered to become low-cost bioreactors for producing biofuels, pharmaceuticals, and other economically valuable products. The fact that the circadian clock globally and profoundly regulates gene expression means that manipulating clock properties can be exploited to enhance bioproduction. This volume: chapters by Wang et al. Xu et al. (2013b) Acknowledgments We thank Mr. Ian Dew for his artwork in Fig. 1. Research in Dr. Johnson’s laboratory is supported by the USA NIH/NIGMS (GM 067152 and GM 107434), and in Dr. Rust’s laboratory by NIH/NIGMS (GM 107369) and an HHMI-Simons Faculty Scholar award.

The Bacterial Perspective on Circadian Clocks

13

References Abe J, Hiyama TB, Mukaiyama A et al (2015) Atomic-scale origins of slowness in the cyanobacterial circadian clock. Science 349:312–316 Andersson CR, Tsinoremas NF, Shelton J et al (2000) Application of bioluminescence to the study of circadian rhythms in cyanobacteria. Methods Enzymol 305:527–542 Aoki S, Kondo T, Ishiura M (1995) Circadian expression of the dnaK gene in the cyanobacterium Synechocystis sp. strain PCC 6803. J Bacteriol 177:5606–5611 Chang YG, Cohen SE, Phong C et al (2015) A protein fold switch joins the circadian oscillator to clock output in cyanobacteria. Science 349:324–328 Chen T-H, Chen T-L, Hung L-M et al (1991) Circadian rhythm in amino acid uptake by Synechococcus RF-1. Plant Physiol 97:55–59 Diamond S, Rubin BE, Shultzaberger RK et al (2017) Redox crisis underlies conditional light–dark lethality in cyanobacterial mutants that lack the circadian regulator, RpaA. Proc Natl Acad Sci USA 114:E580–E589 Dong G, Yang Q, Wang Q et al (2010) Elevated ATPase activity of KaiC applies a circadian checkpoint on cell division in Synechococcus elongatus. Cell 140:529–539 Dunlap JC, Loros JJ, PJ DC (eds) (2004) Chronobiology: biological timekeeping. Sinauer, Sunderland, MA. 406 p Dvornyk V, Vinogradova O, Nevo E (2003) Origin and evolution of circadian clock genes in prokaryotes. Proc Natl Acad Sci USA 100:2495–2500 Eelderink-Chen Z, Bosman J, Sartor F et al (2021) A circadian clock in a nonphotosynthetic prokaryote. Sci Adv 7:eabe2086 Egli M, Mori T, Pattanayek R et al (2012) Dephosphorylation of the core clock protein KaiC in the cyanobacterial KaiABC circadian oscillator proceeds via an ATP synthase mechanism. Biochemistry 51:1547–1558 Ehret CF, Trucco E (1967) Molecular models for the circadian clock. I. The chronon concept. J Theor Biol 15:240–262 Ehret CF, Wille JJ (1970) The photobiology of circadian rhythms in protozoa and other eukaryotic microorganisms. In: Halldal P (ed) Photobiology of microorganisms, Chap 13. Wiley, New York, pp 369–416 Gallon JR (1992) Reconciling the incompatible: N2 fixation and O2. New Phytol 122:571–609 Garces RG, Wu N, Gillon W et al (2004) Anabaena circadian clock proteins KaiA and KaiB reveal a potential common binding site to their partner KaiC. EMBO J 23:1688–1698 Grobbelaar N, Huang T-C, Lin HY et al (1986) Dinitrogen-fixing endogenous rhythm Synechococcus RF-1. FEMS Microbiol Lett 37:173–177 Grobbelaar N, Lin H-Y, Huang TC (1987) Induction of a nitrogenase activity rhythm in Synechococcus and the protection of its nitrogenase against photosynthetic oxygen. Curr Microbiol 15:29–33 Halberg F, Conner RL (1961) Circadian organization and microbiology: variance spectra and a periodogram on behavior of Escherichia coli growing in fluid culture. Proc Minnesota Acad Sci 29:227–239 Hall JC, Rosbash M (1993) Oscillating molecules and how they move circadian clocks across evolutionary boundaries. Proc Natl Acad Sci USA 90:5382–5383 Hatakeyama TS, Kaneko K (2012) Generic temperature compensation of biological clocks by autonomous regulation of catalyst concentration. Proc Natl Acad Sci USA 109:8109–8114 Hellweger F, Jabbur ML, Johnson CH et al (2020) Circadian clock helps cyanobacteria manage energy in coastal and high latitude ocean. ISME J 14:560–568 Hitomi K, Oyama T, Han S et al (2005) Tetrameric architecture of the circadian clock protein KaiB: a novel interface for intermolecular interactions and its impact on the circadian rhythm. J Biol Chem 280:18643–18650 Huang T-C, Grobbelaar N (1995) The circadian clock in the prokaryote Synechococcus RF-1. Microbiology 141:535–540

14

C. H. Johnson and M. J. Rust

Huang T-C, Tu J, Chow T-J et al (1990) Circadian rhythm of the prokaryote Synechococcus sp. RF1. Plant Physiol 92:531–533 Ishiura M, Kutsuna S, Aoki S et al (1998) Expression of a gene cluster kaiABC as a circadian feedback process in cyanobacteria. Science 281:1519–1523 Ito H, Kageyama H, Mutsuda M et al (2007) Autonomous synchronization of the circadian KaiC phosphorylation rhythm. Nat Struct Mol Biol 14:1084–1088 Ito H, Mutsuda M, Murayama Y et al (2009) Cyanobacterial daily life with Kai-based circadian and diurnal genome-wide transcriptional control in Synechococcus elongatus. Proc Natl Acad Sci USA 106:14168–14173 Iwasaki H, Taniguchi Y, Ishiura M et al (1999) Physical interactions among circadian clock proteins KaiA, KaiB and KaiC in cyanobacteria. EMBO J 18:1137–1145 Iwasaki H, Williams SB, Kitayama Y et al (2000) A kaiC-interacting sensory histidine kinase, SasA, necessary to sustain robust circadian oscillation in cyanobacteria. Cell 101:223–233 Iwasaki H, Nishiwaki T, Kitayama Y et al (2002) KaiA-stimulated KaiC phosphorylation in circadian timing loops in cyanobacteria. Proc Natl Acad Sci USA 99:15788–15793 Jabbur ML, Zhao C, Johnson CH (2021) Insights into the evolution of circadian clocks gleaned from bacteria. In: Johnson CH, Rust M (eds) Circadian rhythms in bacteria and microbiomes. Springer, Cham Johnson CH (2010) Circadian clocks and cell division: what’s the pacemaker? Cell Cycle 9:3864– 3873 Johnson CH, Xu Y (2009) The decade of discovery: how Synechococcus elongatus became a model circadian system 1990–2000. In: Ditty JL, Mackey SR, Johnson CH (eds) Bacterial circadian programs, Chap 4. Springer, Berlin, pp 63–86 Johnson CH, Golden SS, Ishiura M et al (1996) Circadian clocks in prokaryotes. Mol Microbiol 21:5–11 Johnson CH, Egli M, Stewart PL (2008) Structural insights into a circadian oscillator. Science 322:697–701 Johnson CH, Zhao C, Xu Y et al (2017) Timing the day: what makes bacterial clocks tick? Nat Rev Microbiol 15:232–242 Kageyama H, Nishiwaki T, Nakajima M et al (2006) Cyanobacterial circadian pacemaker: Kai protein complex dynamics in the KaiC phosphorylation cycle in vitro. Mol Cell 23:161–171 Kim YI, Vinyard DJ, Ananyev GM et al (2012) Oxidized quinones signal onset of darkness directly to the cyanobacterial circadian oscillator. Proc Natl Acad Sci USA 109:17765–17769 Kippert F (1986) Endocytobiotic coordination, intracellular calcium signalling, and the origin of endogenous rhythms. In: Lee & Frederick (eds), Endocytobiology III. Ann NY Acad Sci 503:476–495 Kippert F (1991) Essential clock proteins/circadian rhythms in prokaryotes; what is the evidence? Bot Acta 104:2–4 Kitayama Y, Nishiwaki T, Terauchi K et al (2008) Dual KaiC-based oscillations constitute the circadian system of cyanobacteria. Genes Dev 22:1513–1521 Kitayama Y, Nishiwaki-Ohkawa T, Sugisawa Y et al (2013) KaiC intersubunit communication facilitates robustness of circadian rhythms in cyanobacteria. Nat Commun 4:2897 Kondo T, Ishiura M (1994) Circadian rhythms of cyanobacteria: monitoring the biological clocks of individual colonies by bioluminescence. J Bacteriol 176:1881–1885 Kondo T, Strayer CA, Kulkarni RD et al (1993) Circadian rhythms in prokaryotes: luciferase as a reporter of circadian gene expression in cyanobacteria. Proc Natl Acad Sci USA 90:5672–5676 Kondo T, Golden SS, Johnson CH et al (1994) Circadian clock mutants of cyanobacteria. Science 266:1233–1236 Kondo T, Mori T, Lebedeva NV et al (1997) Circadian rhythms in rapidly dividing cyanobacteria. Science 275:224–227 Kucho K, Okamoto K, Tsuchiya Y et al (2005) Global analysis of circadian expression in the cyanobacterium Synechocystis sp. strain PCC 6803. J Bacteriol 187:2190–2199

The Bacterial Perspective on Circadian Clocks

15

Kushige H, Kugenuma H, Matsuoka M et al (2013) Genome-wide and heterocyst-specific circadian gene expression in the filamentous cyanobacterium Anabaena sp. strain PCC 7120. J Bacteriol 195:1276–1284 Lambert G, Chew J, Rust MJ (2016) Costs of clock-environment misalignment in individual cyanobacterial cells. Biophys J 111:883–891 Leone V, Gibbons SM, Martinez K et al (2015) Effects of diurnal variation of gut microbes and high-fat feeding on host circadian clock function and metabolism. Cell Host Microbe 17:681– 689 Liu Y, Golden SS, Kondo T et al (1995a) Bacterial luciferase as a reporter of circadian gene expression in cyanobacteria. J Bacteriol 177:2080–2086 Liu Y, Tsinoremas NF, Johnson CH et al (1995b) Circadian orchestration of gene expression in cyanobacteria. Genes Devel 9:1469–1478 Ma P, Mori T, Zhao C et al (2016) Evolution of KaiC-dependent timekeepers: a proto-circadian timing mechanism confers adaptive fitness in the purple bacterium Rhodopseudomonas palustris. PLoS Genet 12:e1005922 Makarova KS, Galperin MY, Koonin EV (2017) Proposed role for KaiC-like ATPases as major signal transduction hubs in Archaea. mBio 8:e01959–e01917 Markson JS, Piechura JR, Puszynska AM et al (2013) Circadian control of global gene expression by the cyanobacterial master regulator RpaA. Cell 155:1396–1408 Mihalcescu I, Hsing W, Leibler S (2004) Resilient circadian oscillator revealed in individual cyanobacteria. Nature 430:81–85 Min H, Guo H, Xiong J (2005) Rhythmic gene expression in a purple photosynthetic bacterium, Rhodobacter sphaeroides. FEBS Lett 579:808–812 Mitsui A, Kumazawa S, Takahashi A et al (1986) Strategy by which nitrogen-fixing unicellular cyanobacteria grow photoautotrophically. Nature 323:720–722 Monod J (1966) From enzymatic adaptation to allosteric transitions. Science 154:475–483 Mori T, Johnson CH (2001) Independence of circadian timing from cell division in cyanobacteria. J Bacteriol 183:2439–2444 Mori T, Binder B, Johnson CH (1996) Circadian gating of cell division in cyanobacteria growing with average doubling times of less than 24 hours. Proc Natl Acad Sci USA 93:10183–10188 Mori T, Saveliev SV, Xu Y et al (2002) Circadian clock protein KaiC forms ATP-dependent hexameric rings and binds DNA. Proc Natl Acad Sci USA 99:17203–17208 Mori T, Williams DR, Byrne M et al (2007) Elucidating the ticking of an in vitro circadian clockwork. PLoS Biol 5:e93 Mori T, Sugiyama S, Byrne M et al (2018) Revealing circadian mechanisms of integration and resilience by visualizing clock proteins working in real time. Nat Commun 9:3245 Murayama Y, Kori H, Oshima C et al (2017) Low temperature nullifies the circadian clock in cyanobacteria through Hopf bifurcation. Proc Natl Acad Sci USA 114:5641–5646 Nakajima M, Imai K, Ito H et al (2005) Reconstitution of circadian oscillation of cyanobacterial KaiC phosphorylation in vitro. Science 308:414–415 Nishiwaki T, Kondo T (2012) Circadian autodephosphorylation of cyanobacterial clock protein KaiC occurs via formation of ATP as intermediate. J Biol Chem 287:18030–18035 Nishiwaki T, Iwasaki H, Ishiura M et al (2000) Nucleotide binding and autophosphorylation of the clock protein KaiC as a circadian timing process of cyanobacteria. Proc Natl Acad Sci USA 97:495–499 Nishiwaki T, Satomi Y, Kitayama Y et al (2007) A sequential program of dual phosphorylation of KaiC as a basis for circadian rhythm in cyanobacteria. EMBO J 26:4029–4037 Onai K, Morishita M, Itoh S et al (2004) Circadian rhythms in the thermophilic cyanobacterium Thermosynechococcus elongatus: compensation of period length over a wide temperature range. J Bacteriol 186:4972–4977 Ouyang Y, Andersson CR, Kondo T et al (1998) Resonating circadian clocks enhance fitness in cyanobacteria. Proc Natl Acad Sci USA 95:8660–8664

16

C. H. Johnson and M. J. Rust

Paijmans J, Lubensky DK, ten Wolde PR (2017) A thermodynamically consistent model of the post-translational Kai circadian clock. PLoS Comput Biol 13:e1005415 Pattanayak GK, Lambert G, Bernat K et al (2015) Controlling the cyanobacterial clock by synthetically rewiring metabolism. Cell Rep 13:2362–2367 Pattanayek R, Wang J, Mori T et al (2004) Visualizing a circadian clock protein: crystal structure of KaiC and functional insights. Mol Cell 15:375–388 Paulose JK, Wright JM, Patel AG et al (2016) Human gut bacteria are sensitive to melatonin and express endogenous circadian rhythmicity. PLoS One 11:e0146643 Pittendrigh CS (1965) Biological clocks: the functions, ancient and modern, of circadian oscillations. In: Science and the sixties, Proceedings of the Cloudcraft symposium. Air Force Office of Scientific Research, pp 96–111 Pittendrigh CS (1993) Temporal organization: reflections of a Darwinian clock-watcher. Annu Rev Physiol 55:17–54 Puszynska AM, O’Shea EK (2017) Switching of metabolic programs in response to light availability is an essential function of the cyanobacterial circadian output pathway. eLife 6:e23210 Qin X, Byrne M, Xu Y et al (2010a) Coupling of a core post-translational pacemaker to a slave transcription/translation feedback loop in a circadian system. PLoS Biol 8:e1000394 Qin X, Byrne M, Mori T et al (2010b) Intermolecular associations determine the dynamics of the circadian KaiABC oscillator. Proc Natl Acad Sci USA 107:14805–14810 Robertson JB, Davis CR, Johnson CH (2013) Visible light alters yeast metabolic rhythms by inhibiting respiration. Proc Natl Acad Sci USA 110:21130–21135 Rogers LA, Greenbank GR (1930) The intermittent growth of bacterial cultures. J Bacteriol 19:181–190 Rust MJ, Markson JS, Lane WS et al (2007) Ordered phosphorylation governs oscillation of a threeprotein circadian clock. Science 318:809–812 Rust MJ, Golden SS, O’Shea EK (2011) Light-driven changes in energy metabolism directly entrain the cyanobacterial circadian oscillator. Science 331:220–223 Sánchez-Baracaldo P, Raven JA, Pisani D et al (2017) Early photosynthetic eukaryotes inhabited low-salinity habitats. Proc Natl Acad Sci USA 114:E7737–E7745 Sartor F, Eelderink-Chen Z, Aronson B et al (2019) Are there circadian clocks in non-photosynthetic bacteria? Biology (Basel) 8:41 Schmelling NM, Lehmann R, Chaudhury P et al (2017) Minimal tool set for a prokaryotic circadian clock. BMC Evol Biol 17:1–20 Schmitz O, Katayama M, Williams SB et al (2000) CikA, a bacteriophytochrome that resets the cyanobacterial circadian clock. Science 289:765–768 Schneegurt MA, Sherman DM, Nayar S et al (1994) Oscillating behavior of carbohydrate granule formation and dinitrogen fixation in the cyanobacterium Cyanothece sp. strain ATCC 51142. J Bacteriol 176:1586–1597 Schweiger H-G, Schweiger M (1977) Circadian rhythms in unicellular organisms: an endeavor to explain the molecular mechanism. Int Rev Cytol 51:315–342 Schweiger H-G, Schweiger M (1980) Molecular mechanisms of cellular circadian clocks. Eur J Cell Biol 21:335–336 Smith RM, Williams SB (2006) Circadian rhythms in gene transcription imparted by chromosome compaction in the cyanobacterium Synechococcus elongatus. Proc Natl Acad Sci USA 103:8564–8569 Snijder J, Schuller JM, Wiegard A et al (2017) Structures of the cyanobacterial circadian oscillator frozen in a fully assembled state. Science 355:1181–1184 Stal LJ, Krumbein WE (1985) Nitrogenase activity in the non-heterocystous cyanobacterium Oscillatoria sp. grown under alternating light-dark cycles. Arch Microbiol 143:67–71 Sturtevant RP (1973) Circadian variability in Klebsiella demonstrated by cosinor analysis. Int J Chronobiol 1:141–146 Sweeney BM, Borgese MB (1989) A circadian rhythm in cell division in a prokaryote, the cyanobacterium Synechococcus WH7803. J Phycol 25:183–186

The Bacterial Perspective on Circadian Clocks

17

Takai N, Nakajima M, Oyama T et al (2006) A KaiC-associating SasA-RpaA two-component regulatory system as a major circadian timing mediator in cyanobacteria. Proc Natl Acad Sci USA 103:12109–12114 Taniguchi Y, Takai N, Katayama M et al (2010) Three major output pathways from the KaiABCbased oscillator cooperate to generate robust circadian kaiBC expression in cyanobacteria. Proc Natl Acad Sci USA 107:3263–3268 Taylor WR (1979) Studies on the bioluminescent glow rhythm of Gonyaulax polyedra. Ph.D. dissertation, University of Michigan, Chap 5, pp 78–110 Teng SW, Mukherji S, Moffitt JR et al (2013) Robust circadian oscillations in growing cyanobacteria require transcriptional feedback. Science 340:737–740 Terauchi K, Kitayama Y, Nishiwaki T et al (2007) ATPase activity of KaiC determines the basic timing for circadian clock of cyanobacteria. Proc Natl Acad Sci USA 104:16377–16381 Thaiss CA, Zeevi D, Levy M et al (2014) Transkingdom control of microbiota diurnal oscillations promotes metabolic homeostasis. Cell 159:514–529 Tomita J, Nakajima M, Kondo T et al (2005) No transcription-translation feedback in circadian rhythm of KaiC phosphorylation. Science 307:251–254 Tseng R, Goularte NF, Chavan A et al (2017) Structural basis of the day-night transition in a bacterial circadian clock. Science 355:1174–1180 Vakonakis I, LiWang AC (2004) Structure of the C-terminal domain of the clock protein KaiA in complex with a KaiC-derived peptide: implications for KaiC regulation. Proc Natl Acad Sci USA 101:10925–10930 van Zon JS, Lubensky DK, Altena PR et al (2007) An allosteric model of circadian KaiC phosphorylation. Proc Natl Acad Sci USA 104:7420–7425 Vijayan V, Zuzow R, O’Shea EK (2009) Oscillations in supercoiling drive circadian gene expression in cyanobacteria. Proc Natl Acad Sci USA 106:22564–22568 Voigt RM, Forsyth CB, Green SJ et al (2014) Circadian disorganization alters intestinal microbiota. PLoS One 9:e97500 Woelfle M, Ouyang Y, Phanvijhitsiri K et al (2004) The adaptive value of circadian clocks: an experimental assessment in cyanobacteria. Curr Biol 14:1481–1486 Woelfle MA, Xu Y, Qin X et al (2007) Circadian rhythms of superhelical status of DNA in cyanobacteria. Proc Natl Acad Sci USA 104:18819–18824 Xu Y, Piston D, Johnson CH (1999) A bioluminescence resonance energy transfer (BRET) system: application to interacting circadian clock proteins. Proc Natl Acad Sci USA 96:151–156 Xu Y, Mori T, Johnson CH (2000) Circadian clock-protein expression in cyanobacteria: rhythms and phase-setting. EMBO J 19:3349–3357 Xu Y, Mori T, Johnson CH (2003) Cyanobacterial circadian clockwork: roles of KaiA, KaiB, and the kaiBC promoter in regulating KaiC. EMBO J 22:2117–2126 Xu Y, Mori T, Pattanayek R et al (2004) Identification of key phosphorylation sites in the circadian clock protein KaiC by crystallographic and mutagenetic analyses. Proc Natl Acad Sci USA 101:13933–13938 Xu Y, Ma P, Shah P et al (2013a) Non-optimal codon usage is a mechanism to achieve circadian clock conditionality. Nature 495:116–120 Xu Y, Weyman PD, Umetani M et al (2013b) Circadian Yin-Yang regulation and its manipulation to globally reprogram gene expression. Curr Biol 23:2365–2374 Ye S, Vakonakis I, Ioerger TR et al (2004) Crystal structure of circadian clock protein KaiA from Synechococcus elongatus. J Biol Chem 279:20511–20518 Zwicker D, Lubensky DK, ten Wolde PR (2010) Robust circadian clocks from coupled proteinmodification and transcription–translation cycles. Proc Natl Acad Sci USA 107:22540–22545

Part I

The Circadian Clock System in Cyanobacteria: Pioneer of Bacterial Clocks

Around the Circadian Clock: Review and Preview Takao Kondo

Abstract Dr. Johnson kindly arranged for me to take this opportunity to review my 50 years of research on circadian clocks. I was first attracted by Dr. Bunning’s textbook and Dr. Oota’s lecture on entrainment model of Dr. Pittendrigh’s Drosophila circadian clock. It has been a lot of fun. It took me a long way, but I finally found what I was looking for. My research life can be divided into three periods: the era of duckweed, the era of cyanobacteria, and the era of KaiC. The era of duckweed from 1971 was to experience the unique properties of the circadian clock while watching the cloning of the per gene with a sideways glance. When the per gene was cloned in 1984, I realized the need to find a new system and thanks to the cooperation of Dr. Johnson, Dr. Golden, and Dr. Ishiura, we were able to develop an experimental system for cyanobacteria using bioluminescence around 1991. In 1998, we cloned its clock gene group, kaiABC. When we examined the expression of kaiC, we found a clear negative feedback regulation on the expression of KaiC, and we thought that the transcription–translation feedback model of clock genes that preceded that of eukaryotes would also apply to cyanobacteria. However, the factors constituting the feedback loops were not found as in eukaryotes. In 2005, we found the phosphorylation rhythm of KaiC under dark conditions where transcription and translation do not occur, and they were able to sustain the phosphorylation rhythm of KaiC by simply mixing three purified Kai proteins with ATP in vitro. Initially, we expected that the phosphorylation rhythm of KaiC in vitro would be incomplete and that it would acquire stable properties in cells. However, the in vitro rhythms were found to be better than the circadian rhythms of cyanobacterial cells in terms of cycle characteristics (circadian period and temperature compensation), phase response curves and their tuning, and oscillation persistence, indicating that they are almost perfect as clock mechanisms. The interaction of the three Kai proteins, the phosphorylation cycle, and other details of the cycle were investigated, but the major turning point was the measurement of ATPase activity. KaiC encodes two ATPases. We measured ATPase activity carefully, and found that their activity is very weak,

T. Kondo (*) Nagoya University, Nagoya, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_2

21

22

T. Kondo

degrading only 10–15 ATPs per day but the activity was temperature-compensated and stable. Furthermore, the activity was proportional to the inverse of the cycle (frequency). It is surprising that the activity of definite enzyme reaction determines the rate of the circadian clock. Furthermore, this relationship holds true even when no rhythm is generated by KaiC alone incubation. This paradoxical fact refutes the implicit assumption that the circadian cycle is determined by the sum of chemical reactions, and suggests a mechanism that is quite different from the conventional model of the circadian clock. It appears to be rather similar to mechanical clocks that use harmonic oscillations (physical movement) as their pacemaker. An unpublished discussion of our model of such a KaiC clock is presented as a preview at the last of this review.

As an undergraduate and graduate student at Nagoya University, I was attracted to physics and mathematics, but biology was the most appealing to me. In the biology department, I studied mainly molecular biology and developmental biology. I was also interested in ecology and systematics. In the end, out of the various options, I chose to study circadian clocks in the summer of my senior year. I would like to write about the process. Fifty years have already passed, but I am still in the process of extending that choice. I was confronted with this choice in 1969, but a year or two earlier, university conflicts had spread widely across the country, and the effects were still felt in many parts of the university. The movement began as a concrete demand by students against the old university structure, but as it spread worldwide, there was also a sincere debate on the nature of universities and their research. The movement then shifted away from its original goals and eventually came to a halt with the intervention of the authorities. At the same time, questions were being raised about the progress of science and technology, coinciding with the many distortions caused by high economic growth at that time in Japan. This background also had a great impact on my choice. After much consideration, I wondered if there was any research field that I could do that was truly interesting and not too crowded with other scientists to compete with. I wanted to study a phenomena found in many organisms, without being concerned whether it was “useful,” e.g., medically relevant. Although I would not recommend it to others, in fact, for myself I may have been deliberately looking for “useless” topics. On the other hand, I was also wondering if there were any biological functions that are not essential mechanisms for the survival of life, but are related to the subtleties of survival that are subject to Natural Selection. However, I could not find any specific field of study, so I was just going back and forth, and my interest tended to focus on mountain climbing, which I had started at that time. As a person who dislikes exercise, it was a very new experience for me to be moving my body in the mountains. It was during this time that I listened to a lecture on “Plant Developmental Physiology” by Prof. Oota (pronounced Oh-oh-tah). It seemed that plants use autonomous clocks, observed as circadian rhythms with a period of about 24 h, to measure the length of the day in order to select the season in which they should produce flowers. Prof. Oota explained that when a certain stimulus is given at

Around the Circadian Clock: Review and Preview

23

various times of the ongoing circadian rhythm, the phase of the oscillation moves forward or backward depending on the time of the stimulus, but the pattern (phase response curve) is the same regardless of the species, and repeats with the period of the circadian clock. When this stimulus is given repeatedly at various periods, the phase of the circadian oscillation is consistent with what can be logically predicted from the phase response curve. Therefore, from the entrainment that is expected to be established when a light program with varying day/night length is given, the day length judgment of the plant can be inferred, and flowering occurs accordingly. From this explanation, neither the nature of the circadian clock nor the mechanism of the day length measurement was clear, but the phase of the circadian oscillation and the flowering curve predicted from the light schedule are in perfect agreement. Such research is called physiological analysis. It may seem outdated, but to me, it seemed “pretty cool.” This was the beginning of my research, and as a result, this method has continued to the present. It may have been a roundabout way to go, but I think I was fortunate to have been taught this style. This analysis of the circadian clock is based on the physiological analysis by the three pioneers who established this field in the 1960s, who were fascinated by this phenomenon and had bold ideas (Erwin Bunning, Colin Pittendrigh, and Jurgen Aschoff).

1 With Duckweed I started graduate school in 1971, the year Konopka and Benzer discovered the per mutant of Drosophila. I remember well the surprise I felt when I found the paper while flipping through PNAS in the library to prepare for a seminar. For a moment, I thought I had made a mistake by studying plants rather than Drosophila, but as I was already fascinated by plant physiology, I did not care and concentrated on the analysis of flowering of duckweed (Lemna). Initially, I spent about 2 years counting the number of flower primordia while dissecting them under a microscope in order to quantify the formation of flowers cultivated in various conditions. It took a lot of patience, but once I got used to it, I found it quite enjoyable. Looking through the microscope for about 10 h a day, I had various fantasies and felt as if I could understand a little bit how the plants felt. However, 2 years into the master’s program, I decided that I wanted to focus on studying the circadian clock, which is the basis of photoperiodism, and after much experimentation, I discovered that Lemna has a potassium ion absorption rhythm. This is measured by placing the Lemna in a small chamber and perfusing it with a culture solution that is diluted enough to be affected by the oyster grass absorption. A fraction collector used in chromatography was employed to collect the effluent every hour to determine its potassium ion concentration. Collecting 8 channels of samples with one fraction collector was easy, as we found out from the operation of the fraction collector in the lab. Opening the back of a machine was something I had learned to do as a child. A physiologist would never try to open the back of a living

24

T. Kondo

organism, but a machine is a different story. The radio engineering I had done as a child also came in handy. The flow medium culture (FMC) I created in this way became my first system. I set myself the basic task of creating a measurement system that could function in an automated way and collect data without stopping. This is because the study of circadian rhythms requires continuous measurements over a long period of time. With this system, I followed the circadian clock of Lemna for about 10 years. In 1978, the year I published my work on circadian rhythms in Lemna, Prof. Bunning visited Prof. Oota at Nagoya University. When I explained my work on Lemna to him, he became so interested that he often introduced my work on Lemna in his review articles. This was a great encouragement to me, since work on plants was often not given much attention, even in the field of biological clocks. Later, in 1994, when I found the rhythm of cyanobacteria, I was able to visit Tubingen, Germany (where Prof. Bunning’s lab was located) and visited Prof. Bunning’s graveside under the guidance of his student Prof. Wolfgang Engelmann.

2 At the National Institute for Basic Biology Although I had completed the last 3 years of my graduate school course, I was in no hurry to finish my degree, and on the contrary, I was busy planning to climb the Himalayan Mountains. One day, Prof. Oota asked me to help him set up a laboratory at the National Institute for Basic Biology (NIBB), which had been established in Okazaki, Japan. He told me that if everything went well, I would be able to become an assistant professor of the NIBB (“Assistant Professor” was a tenured position, but not fully independent). The establishment of the National Institute for Basic Biology (NIBB) was an aspiration of the scientists who were interested in “basic biology” in Japan, but I thought it was none of my business. The new institute was established at Myodaiji Temple in Okazaki, where I had spent my junior and senior high school years, so I suppose I was “invited by the geographical advantage” of having gone to high school in Okazaki. The first few years that I spent in this temporary laboratory of Prof. Oota with people from different fields was an experience that expanded my worldview several times faster than would otherwise have happened. The building of the institute was not yet completed, so it was a special and precious experience to be able to talk daily with Dr. Masutaro Kuwabara, Dr. Nobuo Kamiya, and Dr. Haruo Kanatani, as well as with my contemporaries in other departments of the Institute for Basic Biology and at the attached Institute of Physiology and Institute for Molecular Science. Prof. Oota’s laboratory, named “Timekeeping,” was probably the first laboratory in Japan to study biological timing, and I was hired as an assistant professor in October 1978. I remember how elated I was when I arrived at the base camp in Karakoram (a mountain range spanning the borders of China, India, and Pakistan) a few months earlier and started a 40-day climb with my good friends after arduous preparations, and likewise I also still remember how I felt when we started the experiments in the new laboratory in Okazaki.

Around the Circadian Clock: Review and Preview

25

In Okazaki, I continued my research on Lemna using the FMC I had developed in Nagoya. My goal was to take advantage of the ability to measure rhythms with high precision, and to analyze the phase response to environmental factors that change the period and phase, and to pulse administration of chemicals. FMCs, which perfuse the culture medium, allow pulse administration of chemical substances, which is difficult in other organisms, and I hoped that systematic investigation of this would reveal the most basic processes of the circadian clock. After more than 6 years of these experiments, the result was to reconfirm that Bunning and Pittendrigh had been right. The rhythms that can be measured (overt rhythms) are amplified versions of the fundamental oscillations of the circadian clock (covert rhythms or pacemakers) through a number of processes, and the amplitude of the rhythms to be measured is of less significance. In most cases when the rhythm disappears because the overt rhythm ceases to function, the pacemaker cannot be analyzed. On the other hand, if the period changes, it can be expected that the pacemaker has been affected, but since the measurement of the period requires a long duration of continuous administration of the drug, careful confirmation is needed whether this effect is due to the expected primary target of action of the drug or to a secondary effect. On the other hand, the analysis of phase displacement by a brief pulse of a drug may aid the interpretation of the drug’s site of action, although the number of experiments needed is much larger. In any case, precise confirmation of the phase shift and chemical effects is difficult. Although I found phase responses due to various chemicals in Lemna, it was difficult to identify the mechanistic target sites corresponding to the phase responses. On the contrary, it was very impressive that most of the phase responses followed the same curve even though I tried a variety of chemicals that were supposed to have different target points. Interestingly, amino acid analogues showed exceptionally pronounced phase responses in Lemna, and I spent several years analyzing it. My impressions of the circadian clock pacemaker from these experiences, which are summarized in this report, are that they reaffirmed the importance of the fundamental properties of the circadian clock across species as summarized by Bunning, Pittendrigh, and Aschoff in 1960 at the Cold Spring Harbor Symposium on Biological Clocks, and that they suggested the possibility of a highly conserved endogenous clock hidden in the cell. It seemed invaluable to me to realize the importance of the basic properties of circadian clocks across species, and to recognize the commonality of stable endogenous clocks hidden in cells. I thought that the combination of the two different types of oscillations that they were discussing might be the essence of the circadian clock. However, I could not adapt this goal to my experimental design of that time, and I was forced to change direction. As a side note, I conducted all of my research on Lemna alone during my 6 years in Okazaki. It was Prof. Oota’s style to not be a coauthor of the papers of his younger lab members, although he was always ready to help or advise if the junior member requested. So, I always worked alone from the planning of the experiments to the implementation and data analysis, which I thought was natural considering that there were only about 10 researchers in Japan working on plant circadian clocks. In fact, during this period, I wrote 11 papers on Lemna, all in single-authored volumes,

26

T. Kondo

which seemed quite strange to the outside world and made me seem like a difficult and strange person. In my situation, I had no choice but to work alone, and no one who knows me has ever called me a “weirdo.” However, doing research alone and writing single-authored papers, which is similar to climbing a mountain alone, may have been a good training for later research.

3 Plant Physiology When I was asked the question of “what do you do?” by a landlord whom I did not know well, I was embarrassed at first, but when I answered that I was a plant physiologist, he apparently understood. The fact that my research was in physiology was not clear to me even after my degree, but I realized the reality of it when I was in my mid-30s. But by the time I was in my mid-30s and realized that I was a physiologist, I looked around, and I saw that I was already in the minority of biologists. This was a little sad, but I remember thinking that if this was a choice I had made unconsciously, I would cherish it. I guess I was also already aware that many of my choices were often in the minority. Looking back, I was very happy that many people in the USA recognized me as a physiologist. Of course, physiologists are in the minority there as well. At that time, I was often advised that it would be bad for my career if I did not do molecular biology and biochemistry, which everyone else was doing. I visited a biochemist and a molecular geneticist whom I secretly respected, and they both agreed that I should not take the view that I should do what everyone else was doing it because it was not logical. Since they unexpectedly shared the same opinion, I was fortunate to be able to continue with my own method. However, it was obvious that such a scientist could not continue his valuable research without coming up with a new method of observation, as he could not compete with only a single “ruler.” As a result, I decided to develop a device that would automatically and repeatedly measure the activity of life, automatically control it to perform detailed continuous measurements, analyze the response to environmental changes, and visualize the activity of life in high resolution. To this end, my laboratory became increasingly equipped not with centrifuges and electrophoresis chambers, but with electronic components, oscilloscopes, and computers. For the electronic components, I planned to use the sensors and actuators that were already used for various measurements, and to use the microcomputers that became available at that time. Fortunately, the Apple II+ computer had been released in Japan, allowing me to develop a system without any basic knowledge. Using this, I was able to create my own method.

Around the Circadian Clock: Review and Preview

27

4 Cloning of the per Gene Nevertheless, the cloning of the per gene of Drosophila in 1984 had a great impact on me. I felt that it was time to “pay my dues,” and I began to seriously consider a change of direction, as my research on Lemna had just stalled. At any rate, the first substance that was definitely involved in the circadian clock mechanism had been identified, and I thought that if I looked at the sequence, I would see a picture of a clock. It was not that simple, but the impact was huge. When I attended the Gordon Conference in 1985, the people involved in the cloning process for per looked like intimidating people who would take my breath away in a heartbeat. By 1990, the hypothesis that a looped central dogma determines the 24-h period was proposed, and by 1998, the loop was connected to form the currently understood paradigm of the circadian clock. The hypothesis that the central dogma is what we have been chasing was irresistible, but I felt that it was somewhat disappointing. Ironically, it was not until 10 years later that I experienced for myself that the clock gene sequence contained a picture of another type of clock mechanism rather than the genetic feedback loop of the Central Dogma, but that is a story for later (see below).

5 Phototaxis of Chlamydomonas reinhardtii I tried to continue my research using physiological methods as much as possible, but this seemed to be outdated and I needed to focus on identified genes and proteins. So, at the suggestion of Dr. Hideaki Nakashima, a senior researcher of Prof. Oota’s lab at the NIBB, I decided to study the phototaxis rhythm of Chlamydomonas. Dr. Bruce and Dr. Pittendrigh had reported the phototaxis rhythm of Chlamydomonas, and mutants of its circadian period had also been reported. I decided to develop a device to measure the phototaxis rhythm. Fortunately, the Apple II+ was now available on the market, so I decided to use it and make everything myself from individual parts. This was my first attempt to develop an automatic measurement device using a computer, but the Apple II+ was open to the public, from the circuit diagram to the control software, and the system for connecting it to the outside world was ready. Therefore, by concentrating on the instructions for the integrated circuit chips for the input and output peripheral circuits and the development of the machine language software using an assembler, I was able to freely create a measurement device. Because I did all this development on my own, it was an invaluable experience to learn how to design and manufacture each step of the electronics technology. Of course, I had not mastered all aspects of the technology, but being able to culture the organism I wanted to measure, measure its activity, and analyze the data all by myself became an asset for my subsequent research because it gave me the confidence to design instruments for circadian research in the future. The start of my research on Chlamydomonas brought me another great fortune. This was the development of collaborative research. I made a prototype of the

28

T. Kondo

phototaxis apparatus, entrusted it to Dr. Nakashima, and went to Hokkaido with my family for 10 days to attend the Sapporo Symposium on Biological Clocks and the Japanese Society of Plant Physiologists in Sapporo in 1984. During this time, Dr. Carl Johnson visited Dr. Nakashima in Okazaki, whom he had met when Dr. Nakashima was on sabbatical in the USA. Carl, who was also working on Chlamydomonas rhythms at the time, was impressed by the clarity of the rhythms that my apparatus recorded. He therefore asked his postdoctoral supervisor, Prof. J. W. Hastings (“Woody,” who was a professor at Harvard University), to invite me to come to his laboratory for a collaborative visit. I had not been very enthusiastic about studying abroad as I disliked English, but I had changed my mind after climbing overseas twice and therefore I decided to accept Woody’s and Carl’s invitation. Looking back, it was a rare stroke of good fortune to be invited by Carl and Woody because that visit led to the development of my subsequent research. The fact that I was able to make an acquaintance with so many researchers through these two men was invaluable in my life. Although the visit lasted only about 5 months in 1985, it was a very valuable experience for me to have direct contact with researchers from around the world for the first time. I still remember some of the places at Harvard University in Cambridge Massachusetts (USA) where I spent a summer, and I can still recall the exuberant feeling I had at that time. I also had discussions with Dr. Till Roenneberg in the yard in front of the Harvard Biolabs building, with Dr. Walter Taylor, the developer of the “Taylortron” (the first automated instrument for monitoring circadian rhythms of bioluminescence from Gonyaulax), about the fun of electronic circuits, and with Carl about our fascination with circadian clocks. In addition, the backyard of Professor Woody Hastings’ home was a special place for me. Woody held parties for many circadian clock researchers—both famous and unknown—whenever they visited, and I, who lived right next door to Woody, was always hunted down to join them. I was not sure how much I could understand with my poor English, and I always complained that it was bad for my digestion, but getting to know the personalities of my respected seniors, whom I only knew from their papers, became a basis for my subsequent research.

6 Sabbatical: Toward a New Experimental System for Circadian Clocks My research expanded greatly when I started experimenting with Chlamydomonas, but I was not optimistic about my initial goal of finding new clock genes in this alga. It was clear that a more efficient platform for possible molecular genetics was essential, but the requirement was that circadian rhythms could be measured efficiently, and this was difficult in Chlamydomonas. E. coli and yeast seemed to be more efficient organisms for isolating and characterizing mutants, but there were no reports of functioning circadian rhythms in those cells. Around that time, I heard a presentation by Dr. Steve Kay at the Gordon Conference in 1989, in which he

Around the Circadian Clock: Review and Preview

29

introduced the luciferase gene into tobacco and showed that transgenic living cells can be engineered to emit bioluminescence, paving the way for an automated measurement of circadian rhythms in organisms that—unlike Woody’s Gonyaulax—do not endogenously express bioluminescence. When I returned to Okazaki and talked with my colleague, Dr. Masahiro Ishiura, he replied that it would be possible and that he would be willing to help, so we decided to try it together. Considering the cloning of clock genes, we decided that yeast or E. coli, which have well-developed genetic systems, would be advantageous. At that time, it was possible to obtain an overseas research fellowship (a kind of sabbatical), so I decided to use that time to develop a new experimental system. I visited Dr. Carl Johnson’s laboratory in Nashville, USA (where he was now an assistant professor at Vanderbilt University) on this sabbatical. I spent most of my 10 months there trying to measure the circadian rhythms of yeast and E. coli, not Chlamydomonas (which was the organism I was collaborating on with Carl), but doing something completely unpredictable. I am grateful to Carl for cooperating with me in such an unselfish way, even though I was on sabbatical in Carl’s lab and therefore he was probably expecting me to focus on Chlamydomonas. However, even when I was about to run out of time on sabbatical, I could not get a definite response from E. coli or yeast. So, after consulting with Carl, who had just returned from a scientific meeting in which he had seen a poster about the circadian rhythm of nitrogen-fixing activity of cyanobacteria that had already been reported by Chen et al. in Taiwan, we decided to look into cyanobacteria. This nitrogen-fixation rhythm had been confirmed to be temperature-compensated and was a genuine circadian rhythm, but it had been ignored by the larger chronobiology community due to the “eukaryotic only” dogma that almost all chronobiologists believed at that time. Also, other laboratories had not attempted to repeat the experiments of Chen et al. because of the laboriousness of measuring nitrogen fixation and the false preconception that molecular genetics is not easy in cyanobacteria.

7 Meeting Dr. Susan Golden Carl looked at the list of cyanobacterial researchers and began to call them one by one. I could not help but be impressed by the way he talked to them as if they were old friends even though he didn’t know any of them at that time. After some calls, he called Dr. Susan Golden, and everything started to happen. A few days later, a small package arrived from Texas. When I opened it, I found 50 ml of green liquid in a Falcon tube, labeled AMC149. I was told that Susan’s lab had incorporated the gene for bacterial luciferase with the promoter of the main gene encoding a key protein of photosynthesis (PsbAI), and to my surprise, I was free to use it. However, we only had a few days left before I was due to leave Nashville, so we only put some AMC149 cells in the scintillation counter and checked for very weak luminescence. By the way, I had always wanted to try traveling to a new place with my family and luggage in a horse-drawn carriage, if only as an imitation. So, instead of flying

30

T. Kondo

home, I loaded my luggage and three family members into a rather unreliable Renault automobile and put the AMC149 Falcon tube that Susan had given me in the rear window, thinking it would be better in the light. Everyone saw me off from Nashville and my family and I set off on Interstate 40, heading northeast toward Boston. After about 1300 miles and 10 days of visiting national parks and a few old friends on the east coast, I arrived at Woody’s house at Harvard University. Woody and his wife Hanna were as warm and welcoming as ever, and the backyard yard was just as I remembered it. Since I had about 10 days to spare before returning to Japan, my plan was to see if I could repeat the results that I had obtained with Lemna on the Gonyaulax luminescence rhythms on the Taylortron in Woody’s lab while my family toured Boston, and also to visit the famous Walden Pond. The Gonyaulax experiment did not go well, but I put the AMC149 cyanobacteria into an extra channel, and Woody, being a bioluminescence expert, advised me not to put the decanal substrate directly into the culture medium, but to moisten a piece of filter paper and stick it on the top of the vial and feed it through the air. This did the trick, but the resulting chart was not very impressive; it was almost a straight line. However, when I looked at the chart as a runway from the perspective of a pilot on the verge of landing, it seemed to have a faint rhythm. I told myself that this was promising even if I detected that I was biased in my interpretation, packed AMC 149 in the corner of my suitcase, and headed home for Japan.

8 Whispers of Bioluminescence When I returned to Okazaki in August, I thought a lot about it, but decided to give up on E. coli and yeast and bet on AMC149. I had learned from my experience with E. coli and yeast that even though the rhythms were unstable, there were some rhythms that seemed to be present in E. coli and yeast, but the processing and final measurements (mainly high temperature resistance and UV resistance, which were all measured manually) contained errors from the experimental manipulations. We had learned from our experiences with Lemna and Chlamydomonas that an automatable measuring system was better. Therefore, I preferred the faint rhythm of the fully automated Taylortron data and consequently I started by building a measurement device. Since I had no experience in weak light measurement, I consulted Dr. Shigeru Itoh, a photosynthesis researcher at the National Institute for Basic Biology (NIBB). He kindly explained about photomultiplier tubes and confirmed the optimal emission wavelength, and told me that I could easily ask Hamamatsu TV (later known as Hamamatsu Photonics) for the same wavelength. When I called the sales staff at Hamamatsu and told them what I wanted, they suggested that I should try the 931B photomultiplier tube (PMT) first. The 931B was the cheapest photomultiplier tube at around 9000 yen (about 100 USD). When I ordered this PMT and opened the box, I was relieved to find that it looked the same as the GT-type vacuum tube I had used as a child.

Around the Circadian Clock: Review and Preview

31

The next problem was to find the culture conditions under which the cyanobacteria would show their rhythm. My experience with Lemna taught me that the amplitude of the circadian rhythm is not necessarily highest at the optimal growth rate. When I was in graduate school, I remembered that when I brought to Prof. Oota a large amount of blurred data on flower bud induction in Lemna, he said, “You haven’t taken good care of the Lemna yet,” and asked me to start over. In this area, it is important to treat even bacteria in such a way that we can imagine their feelings, and it is important for physiologists to have the imagination to understand the whispers of the organisms. Konrad Lorenz’s book, King Solomon’s Ring may be the essence of physiology. I once had a disagreement with a biochemist about which should be cleaned more carefully, the glassware used for culturing organisms or the glassware used for storing reagents. The biochemist naturally said the latter, but I insisted on the former. For the physiologist, the experimental organism is not only an object to be measured but also a measuring instrument. If this is the case, isn’t the culture vessel more important for physiology? In fact, I remembered how careful Prof. Oota was with culturing for his experiments on flower bud induction in Lemna, and I wondered what conditions were important for observing circadian rhythms. As a result, I learned that for many organisms, the amplitude of the rhythm was weaker under the conditions that led to high metabolic activity, and that conditions with low activity were more suitable for observing the rhythm (e.g., most circadian rhythms in plants are more robust under dim light than under bright light). Thus, grasping the whisper of life with a new tactic is an obvious pleasure of applying a physiologist’s approach, and this is where new research begins. In fact, after 3 months of trial and error, the first data were obtained on Christmas Eve, 1991. There was still room for improvement, but the rhythm was clear for all to see. I was so excited that I showed it to Dr. Ken-ichi Homma, who happened to be in Okazaki. A respected physiologist, he immediately understood the significance of those data and invited me to present a seminar in Sapporo.

9 Japan–USA Joint Research The subsequent work to clone the three kai genes proceeded very smoothly with some good fortune: the Synechococcus elongatus PCC7942 that Susan had developed (strain AMC149) had all the best properties, as if it had been designed for our purpose beforehand. The luminescence and stability of the Vibrio luciferase employed as a reporter was optimal for the analysis of circadian rhythms. The colonies on the agar medium remained as discrete, easily imaged colonies and could be measured over a week. The DNA added to the medium enters the cells in normal culture and causes highly efficient homologous recombination. The circadian oscillation is extremely stable and highly accurate. Even the fact that the growth rate of the cells is much slower than that of E. coli is extremely favorable for the study of the circadian clock, which requires long-term measurements. On the other hand, the fact that genetic techniques were not as well developed as in E. coli was not a

32

T. Kondo

Fig. 1 Beginning of joint research at Jibuzaka, Japan in 1994

problem, as gene transfer in the naturally competent cyanobacteria is actually easier than in E. coli and the frequency of homologous recombination is higher. It was as if the organism had been optimally tuned for the study of the circadian clock. I reasoned that if we could test the rhythm of 1,000 colonies in a single petri dish— which had been necessary to clone a nutrient-requiring gene in E. coli—we would be able to obtain a clock gene. This was the spirit in which I set out to develop an experimental system to measure circadian rhythms by bioluminescence using microorganisms, a method I will describe later. How many times have I thanked myself for my good fortune? Another more important aspect of my good fortune was that I had the full cooperation of three reliable and top-class researchers (i.e., Ishiura, Susan, and Carl; see Fig. 1). The team of four, including myself, had clear and complementary responsibilities, and as this was a completely new system, there was work for each of us to do. The team was able to maintain close cooperation for 6 years starting around 1991. In fact, my professional position was the most precarious of the four, so this cooperation was a matter of life and death for me, and I was probably the most dedicated to sustaining the team. In Japan, assistant professors were tenured but not allowed full independence like assistant professors in the USA, and in the context of biology’s rush into molecular biology, I was truly an endangered species, because I was sticking to physiology. Susan pioneered the development of the molecular biology platform for Synechococcus elongatus and provided me with various essential technologies for my research. Carl was a cell biologist from the basic research field of circadian clocks in the USA and had a wide network of contacts. He also had extensive experience with many organisms and had a unique biological sense. Dr. Ishiura was a cell biologist who had mastered molecular biology early in his career and was a strong collaborator of mine in Okazaki. I was a plant physiologist who was fascinated by the circadian clock, and I made my own measuring devices to observe the rhythms and think about this and that.

Around the Circadian Clock: Review and Preview

33

The fact that this team of four collaborated so closely for 6–8 years was often wondered at by American researchers who value independence. But when it came to developing a completely new experimental system, this was a mutually beneficial arrangement. For me, my collaboration with the three of them was not only successful, but also very enjoyable, and I am still grateful for this good fortune. The joy of such collaborative research may have been something I learned from my mountaineering experiences in the past. I started mountaineering when I was in my fourth year of college, and once I had experienced mountaineering, I wanted to go to various places more freely. This meant walking freely on rocky ridges and snowcovered ridges without trails in various seasons, but it required a certain level of skill, and since I couldn't invite my old friends to join me, I started walking alone. Thinking about it now, it was quite a dangerous situation, but fortunately, an experienced friend invited me and we enjoyed climbing together for almost 10 years. Climbing with a friend opened up a world of possibilities, and I learned the joy of climbing difficult routes together. Somewhere along the way, these experiences helped me to foster the togetherness of our project on the circadian clock of cyanobacteria. And it was much more fulfilling than when I was working on projects that led to single-authored papers.

10

Design and Fabrication of Bioluminescence Measurement System

My task was to make the measurement device, but since I did not have very much money for research, I decided to make the apparatus by myself as much as possible. Even if it took a lot of time and effort at first, I knew from my previous experience that it was more efficient because it allowed me to make adjustments by my own will, so I bought a photomultiplier tube from Hamamatsu Photonics and assembled the measurement device. It could only measure four vials at one time, but I was still able to repeat the measurement automatically. This worked well, so the next step was to build a smaller version of the Taylortron in Woody’s lab. This also worked quite well, and we were able to obtain high-quality data. In particular, we were able to increase the sensitivity by several orders of magnitude by introducing the photoncounting method, which was a great achievement. I would like to express my gratitude to Hamamatsu Photonics for providing me with a modularized circuit that made such advanced technology much easier to use. In the same way, I am indebted to Tateishi Electric Co., Ltd (now OMRON Corporation) for modularizing and providing various electrical control systems. These types of technologies were probably a basis for Japan’s economic success. The second step in developing the instrument was to measure the luminescence rhythm from colonies on Petri dishes. As mentioned above, the most important step in molecular genetics is to make mutants based on identifiable traits in colonies grown on agar plates. If we could do this, we could easily find the desired mutant

34

T. Kondo

trait among tens of thousands of colonies, and we could expect to catch up with the earlier research on Drosophila and Neurospora (bread mold). So, I borrowed a Hamamatsu Photonics ARGUS image processor from a friend at the Institute and stayed up all night for 4 days to measure the luminescence images on petri dishes. The data obtained were excellent. Since the expected luminescence was very weak but long exposure times were possible (for circadian rhythms, measurements were over long time spans), a system with low thermal noise was suitable, and therefore I chose a cooled CCD camera as the detector. I wanted to keep the equipment as simple as possible and to minimize mechanical operation, so I decided to use an aluminum disk of 75 cm diameter as a turntable on which to place 12 petri dishes. Cyanobacteria are cultivated under illumination for photosynthesis, but they needed to be placed into complete darkness for the bioluminescence measurement. I did not want to use complicated shutters or sample carriers, so I used the simple rotating turntable. The measurement position itself had to be completely shaded. For this purpose, 5 mm black felt was used. The positioning of the disk depended on the accuracy of the motor, and we chose a stepper motor (made by Oriential Motor). The precision of the turntable’s positioning from cycle to cycle using the stepper motor was excellent. The most important part, the metalwork of the disk, was beyond my control, so I asked the Machine Shop at the Institute for Molecular Science next door to machine the turntable and frame. I was very grateful for the support I received. The design was a great success and has already lasted 25 years with little maintenance and is breaking records.

11

Development of LCM and LCA: With Inside Macintosh

I chose a cooled CCD camera manufactured by Princeton Instruments of the USA. The main reason was that this company was the only one that provided the driver software for use with Macintosh computers, which I wanted to use for a variety of reasons. For example, there were many existing image analysis software packages available, and their performance was excellent, but for the measurement of circadian rhythms, the system must be controlled by a fully automated system, and therefore I had to create our own image handling software. In order to do this, I also had to create our own software to import images from the camera, and such a driver is essential, but this is a software component that only the camera’s manufacturer can write. On the other hand, the Macintosh computer was excellent for developing software to handle images, since Apple published a software manual called Inside Macintosh, which could be used with your own software, and you could build procedures using the basic control structures provided by so-called general-purpose compilers. These general-purpose compilers required detailed processing, but allowed the programmer to freely control every detail. At that time, Pascal or C were the mainstream compilers, but since only Pascal was offered for the Macintosh, I used Pascal.

Around the Circadian Clock: Review and Preview

35

I decided on the software structure I would have to develop and ordered a cooled CCD camera from Princeton Instruments. The price was more than my salary for an entire year, but I managed to pay for it in installments. One day, a large parcel arrived. I attached my favorite Nikon lens to the camera and tried to capture the images. In order to do this, I needed to match the functions described in Volume 5 of Inside Macintosh, with the drivers provided by Princeton Instruments, and display them on the screen using Macintosh software. After 2 weeks of researching, writing programs, and testing, I was at a loss. It didn’t work. Then one Sunday, my 6-yearold daughter visited the lab to play with me. I stood her in front of the camera, changed some settings almost randomly (I can’t remember what I changed!), and was surprised after taking the next exposure to see her smiling back at me from the computer screen. Finally, I could use the image as my own. Since then, the routine has been used in my laboratory without interruption for about 25 years. I cannot write about the details of the program here, but I would like to describe what I did. Without this program, the analysis of the circadian clock of cyanobacteria would not have progressed, and I would like to explain it in the hope that young people will create their own programs in the future and open up a new field of physiology. There are two ways to use technology, one is to pursue the maximum potential and freedom, and the other is to pursue the maximum efficiency. You can choose the programming environment you want to use depending on your needs, but the standard C is probably the way to go. In my case, I did not have C at my disposal, so I used Pascal, which is almost as good. It is necessary to organize the specifications of the developed system and examine them carefully. The next step is to decide what equipment is needed, and then carefully study the interface and driver software provided by the manufacturer. The first step in the system configuration is to design the data structure to be handled. In the case of LCM (my data acquisition program, called Luminescent Colony Monitor, or LCM), it was difficult to recognize the colonies from the luminescence images, but I could find the standard routines and transfer them to my own data. The important thing here is not to be misled by the higher-order tools provided by the manufacturers or other programmers, but to understand the details of your own program; otherwise you will not be able to achieve your goal. Trying to make things easy by using existing software, often leads to a dead end. Programming on the Macintosh with Pascal was a lot of fun, and I still recall spending a summer programming on a Mac Plus computer at my second home in the mountains (Jibuzaka). Pascal can handle a wide variety of data, and can be combined to create structured data. It’s a bit of a pain to get used to, but once you do, it’s a very reliable system and you can develop your own programs with confidence. However, it is also true that when you are trying to develop a practical program, you need to be prepared to avoid tangents and move forward. It took me about 3 months to develop the basic software as a starting point, and another 3 months to be able to screen the cells efficiently. I knew that colony screening was essential and that without it, the collaborative research would be meaningless, so I worked hard to accomplish this. Once I could get this working, we could quickly make up for the time that I spent developing the software, because it did all the work that requires a lot of effort, such

36

T. Kondo

as the circadian clock research. It took me about a year to create all the software. When I printed out the source code on paper as a hard copy, it was about 20 cm thick. I called the software that I created LCM (Luminescent Colony Monitor) and LCA (Luminescent Colony Analysis). LCM can repeatedly measure the bioluminescence of 12 petri dishes in a 45-min cycle and record the time course of the luminescence of up to 1190 colonies per image. The whole process is fully automated, and once started there is nothing to do until the end. In other words, up to 14,280 circadian rhythms can be recorded simultaneously, and the period and phase of each can be calculated automatically with high accuracy by LCA, making it possible to screen for mutants.

12

Mutant Screening and Complementation by Library

The subsequent screening was like a treasure hunt: mutagenesis by the mutagen ethyl methanesulfonate (EMS), sowing in petri dishes, measurement by LCM, and analysis by LCA. The system that I designed performed better than I had expected, and one person was able to screen about 70,000 colonies in 1 month. The first circadian period mutant was 22 h and the next one was 28 h. After about 3 months of screening, I was able to isolate about 30 different mutants. In the CCD system, the resolution of the rhythm measurement was very high, so it seemed that many mutants could be found, including those with abnormal waveforms. Since the 24-h circadian rhythm has many steps, we dreamed that we could catch all of them.

13

Discovery of the KaiABC Clock Gene Cluster

Next, we started searching for the causative genes of the obtained mutants. Since a high rate of homologous recombination was expected from simple transformation, we introduced a random library of wild-type genomes created by Setsuyuki Aoki and Shinsuke Kutsuna (graduate students at that time) under the guidance of Dr. Ishiura into a cyanobacterial homologous recombination vector. This was then incorporated into four period mutants, and the CCD system was used to search for mutants that had reverted to the wild-type phenotype. According to Dr. Ishiura, it should be possible to find mutants by screening 20,000 to 30,000 mutants, so we got to work. We first tried the 48-h period mutants, but could not find any “rescued” colonies in the first and second screenings. On the third screening, we found one clone that showed almost the same rhythm as the wild type. At that time, the rhythm of about two colonies per second was displayed sequentially and watched by LCA, but suddenly the rhythm of a “wild-type” phenotype was displayed for 0.5 s. I remember well the shock I felt as if I had been nudged for a moment. The next day, I checked the chart several times and picked up the colony with trembling hands. In this clone, we predicted that the rescuing clock gene with the tag should be included in the DNA

Around the Circadian Clock: Review and Preview

37

that had been transferred. For the other three mutants, we also found DNA that could revert the mutant phenotypes to the wild-type phenotype. Since each of the four mutants had a very different periodicity, we thought that they were not alleles of a single gene. However, Susan’s lab suggested that we do a Southern analysis just to be sure, so we asked her to look into it, and to our surprise, the DNA that reverted the four mutants to the wild type all hybridized with each other. That result meant that the four mutants were alleles of a single gene, so we examined the other mutants. To do this, we simply put the same DNA fragment that had rescued the first four mutants into the other mutants. To our astonishment, the same DNA fragment rescued 30 or more different period mutants! The results meant that all the mutants were mutants of one gene or one gene cluster. Eventually, the DNA sequence was determined and the genetic analysis of the kai gene cluster consisting of three adjacent genes, kaiA, kaiB, and kaiC, began. We named the cluster “Kai” because this word means “cycle” or “rotation” in Japanese.

14

Return to Nagoya University

Since I was not a Principal Investigator (PI) at the NIBB in Okazaki, there were various restrictions on how I could proceed with my research. So, I tried to apply here and there for other positions, but was not successful. However, after I got the idea of cloning the clock gene, I received several offers. Among those offers, I decided to move to Nagoya University because I could move with Dr. Ishiura. Coincidentally, the post was in the same laboratory where I had been a graduate student, so I was returning to my old home. However, since several generations of professors in other fields had already taken the post since the days of Prof. Oota, I had to set up a completely new laboratory. It was also a hectic year for my family and me as we had to move out of our house. As a PI at Nagoya University, I now had to supervise graduate and postgraduate students, which made me much busier, but I was grateful for the much greater potential for space and human resources.

15

What “Not to Do”

Sometime after I returned to Nagoya, I had a visit from a senior colleague at Prof. Oota's laboratory. Although we did not work together at the same time, he was a senior who had made great progress in the analysis of enzymes, and I was very grateful that he visited me to give advice to junior colleagues. He gave me a lot of advice as I was still anxious after becoming a professor, but the main point was that I should think carefully about “what NOT to do” rather than “what TO do” because the time for research is not infinite. He was right, and I appreciated that he pointed it out clearly. In my own way, I decided to (1) focus on clarifying the most basic mechanism of the circadian clock, (2) place the highest importance on physiological

38

T. Kondo

methods that suit my personal style, and (3) actively develop technologies and methods as I like to do so, but only for the purpose of clarifying the basic clock mechanism (i.e., Purpose#1) and not for the application of those technologies. I have been following this policy ever since. There have been many people who were interested in the methods I had developed and came to me for consultation. I was very happy to know of their interest, and tried to help them as much as I could, but I limited my help to what I could spare of my time. Some people were particularly interested in programs and automatic measurements, and were eager to learn more. My advice was that I would be happy to provide them with my software, etc., but they would have to figure it out on their own for their own purposes, which would take a lot of time, so they had to be prepared for that kind of time investment. I told them that learning programming would pay for itself in the future if they did it properly, but that it would be a complete waste of effort if they stopped halfway.

16

Cyanobacterial Transcription and Translation Model: Central Dogma in a Loop?

By the end of 1995, the three kai genes had been identified, but the paper was not published until 1998, which ultimately included an analysis of their expression according to a transcription–translation model. Native kaiC expression was found to be completely suppressed by overexpression of kaiC introduced downstream of the inducible promoter, suggesting that the cyanobacterial circadian clock was similar to the eukaryotic Transcription–Translation Feedback Loop (TTFL) model. This conclusion was readily accepted by many researchers and contributed to the straightforward conclusion that the core of the clock is the same in prokaryotes. On the other hand, although it was clearly pointed out in the 1998 paper that KaiC is an ATPase, its significance was unclear; and nothing was said to me by my colleagues in this field or by the readers about its potential significance. I just remember that a few close friends told me that it was troubling. In any case, the original plan of splitting the clock genes among the four collaborators was no longer feasible because there was essential just one clock cluster, so we agreed to suspend the collaboration and let each group of us work freely. However, as friends, we agreed to keep each other informed of the status of our research and to avoid unnecessary competition. Since then, there have been some actual conflicts, which may have caused some stress to the members of each laboratory, but I am very grateful that we still maintain a good relationship as friends and colleagues, and I am thankful for that. In the following days, we pursued the usual path of molecular biology: studying the interaction among the three Kai proteins, the scenario of the KaiC phosphorylation cycle, and the search for proteins that interact with KaiC. The research proceeded smoothly and yielded results expected from previous eukaryotic studies,

Around the Circadian Clock: Review and Preview

39

but some differences were found: the three Kai proteins showed clear interactions with each other, while Hideo Iwasaki’s two-hybrid search in yeast showed that there were few clear interactions of the Kai proteins with other cyanobacterial proteins. An exception was SasA, known as a histidine kinase, whose deficiency almost eliminated the rhythm, but Susan’s lab also focused on this gene and found a weak rhythm under different conditions. This result was reported jointly by our two laboratories, and the fact that the period did not change even when the transcription rate of the Kai proteins was severely reduced by the SasA knockout raised questions about the TTFL model. In addition, the fact that many related genes had been found in Drosophila and mammalian circadian clock studies raised the question of whether a 24-h TTFL was possible in cyanobacteria. Our lab and Carl’s lab also found an unexpected fact about the regulation of the expression of the kaiBC operon. In other words, if the expression level is carefully regulated, a perfect rhythm can be maintained even when kaiBC is expressed under the control of a heterologous promoter from E. coli. In hindsight, these results suggested a possible next development, but at the time, they were just puzzles.

17

Obligate Photoautotrophy: Key to the Circadian Paradox

The 2003 paper written by Yohko Kitayama, who was a graduate student during this period, was not a study bound by the ideology of molecular biology, but was a biochemical paper that dealt with the intracellular movement of Kai proteins and KaiC phosphorylation quantitatively, and its contents became the basis of our subsequent analyses. Those analyses would not have been possible without her, and they became the basis for a major research shift in 2005. However, it was Jun Tomita’s analysis that directly triggered the change. He was interested in the mechanism of synchronization of circadian rhythms. Since the cyanobacterial circadian clock can be synchronized by a 12-h dark period, he was investigating the fluctuations in the amount and phosphorylation of KaiC protein during this dark period. The Synechococcus elongatus we were using is absolutely photoautotrophic and gene expression quickly drops to zero during the dark period. Since Tomita already knew that the phosphorylation of KaiC showed a rhythm, he investigated it in the dark, but as he analyzed it, he extended the dark period to 72 h. He found that the amount of KaiC protein was constant over the 72 h of darkness, neither synthesized nor degraded, and the KaiC molecule maintained its phosphorylation rhythms in the cells in darkness even though the metabolic activity was completely lost. I was really surprised when Dr. Tomita showed me the data together with Hideo Iwasaki. It was obvious that the single western blot he had brought was a complete rejection of the TTFL model because transcription and translation were not occurring in the darkness. They confirmed the temperature-compensated nature of the period and the absence of transcription–translation activity, and submitted the paper

40

T. Kondo

after experimentally confirming that the rhythm was due to the KaiC protein that remained undegraded by using mutants and inhibitors. The paper was published by Science in early 2005.

18

Reconstitution Experiment

After submitting the Tomita/Iwasaki paper, I asked the members of the lab to get together and discuss the future of our research. Since almost everyone in the lab was involved in the research plan based on the TTFL model of the kaiC gene, the discovery of the phosphorylation rhythm in the absence of transcription–translation activity meant that the assumed model had become questionable. Everyone was very serious. There were many opinions and suggestions, but we could not reach a consensus. In the midst of all this, I remembered that at the end of my application for the Grant-in-Aid for Scientific Research, I had written “reconstruction of the circadian clock” as my dream for the future. If we stuck to the TTFL model, we would have to prepare all the transcription–translation processes, prepare a system where oscillations are generated by feedback, and supply energy continuously. It was obvious that this was almost impossible. It took me a while to reconsider these preconceptions in my mind, but when I thought about it, the only possible explanation for the rhythm that Tomita had discovered was that transcription and translation had completely stopped, and that the KaiC protein in the cell from the beginning of the dark period was being phosphorylated cyclically. If this were the case, the rhythm could be sustained by simply mixing the three Kai proteins and ATP in an appropriate buffer. Fortunately, we had a good idea of the correct protein concentrations from Kitayama’s measurements, and since the importance of Kai proteins in biochemistry had been predicted, we had already developed the methods to purify them. Why don’t we all try to reconstitute an oscillation with the three Kai proteins and ATP? Many members were interested in this suggestion, and enough members agreed to divide the work for all-night sampling, so Masato Nakajima, Taeko Nishiwaki, and Hiroshi Itoh took the lead in continuing the arduous (every 2 h) sampling. As we tried various conditions, the first data came out in about 2 months. We rushed to get the necessary data and sent the manuscript to Science. When Nakajima brought me the first data with a smile, I was strangely convinced that it was true. We had “opened the lid” of the cell and found three Kai proteins quietly ticking away the 24 h. My intuition that this was a big deal came true. I was 57 years old and had only 6 years left until the usual retirement age for Nagoya University. In addition, by this time, I had no choice but to get involved in the management of the university. In fact, the Dean of the Graduate School was to be appointed the following year and I was asked to serve as Dean. So, first of all, I took our reconstitution data to Prof. Omine—the Dean of the Graduate School at that time—explained its significance to him, and asked him seriously if I could decline the offer to be Dean. Prof. Omine, who had been listening in silence, looked at me

Around the Circadian Clock: Review and Preview

41

seriously and said that unfortunately it was too late for me to avoid the Deanship and encouraged me to work hard on my research as much as I was able. After submitting the manuscript to Science, we gathered the lab members again and I told them that we would concentrate on in vitro experiments from now on and asked them to choose a theme to pursue. The basic analysis of the system had been going well, including the binding of Kai proteins in vitro (Kageyama), the analysis of the KaiC phosphorylation cycle in vitro (Nishiwaki), and the ratio and rhythmic properties of the three Kai proteins (Nakajima and Itoh). The basic analysis of this system progressed smoothly. As for me, I was not able to have sufficient discussions with my lab members due to my new position as Dean of the Graduate School, but there were several results that awakened my interest.

19

Perfect Circadian Oscillation

Initially, we assumed that in vitro circadian oscillations would be less perfect than those exhibited by living cyanobacteria; we imagined that the “primitive circadian oscillations” seen in vitro would be optimized inside the cell, gaining stability and phase properties to become a circadian clock that would perform extraordinary physiological functions. However, this was a simple assumption, and in fact it seems closer to the truth that the three Kai protein clocks are nearly perfect circadian oscillations on their own, and that fluctuations of intracellular metabolism is more of a noise to the oscillations. I would like to explain each parameter below. First of all, if ATP is supplied from the outside and its concentration is kept above 0.1 mM, the in vitro oscillation will continue for more than 2 weeks without decay. We took measurements at 30 min intervals in our laboratory, and found no signs of rhythm decay over 2 weeks. Analysis of the position of the peak from curve fitting shows that the stability of the period (fluctuation of the peak time) is about 10 min at most, which is as accurate as—or more accurate than—the circadian rhythms of many eukaryotes. This accuracy means that the fluctuation of the period is only a few percent, which should be kept in mind as a condition for a pacemaker model. On the other hand, in the temperature range where the rhythm is found, the period is almost completely compensated. Next, we examined the many KaiC period mutants that Keiko Imai had collected, and found that the period was temperaturecompensated in almost all of them, with periods ranging from 14 to 60 h. The fact that this compensation is achieved not in a living cell, but in a system of three Kai proteins and a small fraction of ATP, makes it very difficult to explain how the mechanism for determining the period balances the temperature coefficient of the time required for multiple processes. Rather, it is more natural to think that the period is determined by a mechanism that is not affected by temperature. It may seem far-fetched to think of a physical pendulum as a mechanism that enables the period to be unaffected by temperature, but I would like to explain that this analogy is not unrelated.

42

T. Kondo

Galileo’s discovery of the isochronous nature of the pendulum means that the period of the pendulum is determined only by its length, and that the period remains the same even if the amplitude changes. In other words, it is the amplitude compensation of the period. Since the amplitudes of various non-circadian oscillations exhibited by living organisms are strongly affected by temperature, circadian oscillations, whose period does not change with temperature, can be regarded as having amplitude compensation properties. If this is the case, the fact that the mutation of KaiC maintains the amplitude compensation property even when the period is changed can be easily explained if the mutation “shortens the length of the pendulum” so that the period is shortened, but the principle of the pendulum is maintained. The initial phase of the KaiC phosphorylation oscillation is usually determined by the initial phosphorylation state of KaiC and the onset of the temperature shift from ice storage to the measuring temperature (e.g., 30
C). We have analyzed the phase fluctuations of these oscillations by stimulation, and have confirmed that the phase changes with temperature shifts, although light has no effect in vitro because none of the Kai proteins show light absorption. A team composed of Takuya Yoshida, Yoriko Murayama, and Hiroshi Ito studied this phase shift in detail to determine the phase shift caused by temperature shift. The phase response curve obtained was typical of circadian oscillations, so they repeatedly applied temperature pulses to analyze the subsequent synchronization of the rhythm. Surprisingly, the rhythms of the three proteins in vitro were identical to those reported for Drosophila eclosion (Zimmerman et al.), and the three Kai proteins thus almost completely achieved the characteristics that I believed could only be achieved by “living life” in an intact cell. In addition, the in vivo analysis of KaiC had showed us yet another unexpected feat. It was to Itoh’s credit that he observed that the phosphorylation rhythm continued for 2 weeks without decay. It is very strange that the rhythm does not decay at all in a completely constant situation without any external information. So he tried to test this using the in vitro system. He mixed in vitro reactions that were at different phases. The results were clear. When the different phases of the KaiABC system were mixed, the phases quickly unified to the preferred phase (early stage of dephosphorylation). This mechanism is still largely unexplained. However, this function is expected to be important for cells such as cyanobacteria, where new, as-yet-unsynchronized KaiC is actively synthesized (the most effective noise in the cell is probably newly synthesized KaiC) and then needs to be synchronized to the current circadian phase of the cell. Inherent function is also important to remember when considering a KaiC pacemaker.

20

The CI-ATPase Activity of KaiC Determines the Period

The most significant discovery using the in vitro reconstitution system was the measurement of ATPase activity in KaiC by Kazuki Terauchi. ATPase is, of course, the most fundamental enzyme in life, catalyzing the synthesis and degradation of ATP, especially in the degradation process, where they have physical functions that

Around the Circadian Clock: Review and Preview

43

other enzymes cannot perform and achieve the characteristic functions of life. In particular, the degradation process has physical functions that cannot be performed by other enzymes. On the other hand, given the implicit assumption of molecular biologists that the most important phenomena of life are controlled by gene expression, ATPases are thought to be merely energy conversion devices that supply necessary energy but are not expected to control life-like phenomena. ATPases are not expected to measure time, are they? That’s what I thought, too. However, when I thought about it, I realized that if the TTFL model was no longer possible, and given that the diversity of the periods exhibited by the KaiC mutations was much greater than that of KaiA and KaiB, this all suggested that KaiC was the main determinant of the period. If this is the case, we must seriously investigate its ATPase activity. Another reason I thought this was worth investigating was the controversy surrounding how ATP is synthesized in the research field of oxidative phosphorylation that I had learned about as a student. The chemical theory that ATP is synthesized by the production of high-energy intermediates through a combination of chemical reactions was the mainstream hypothesis until the alternative chemiosmosis theory was proposed by Peter Mitchell in 1966 in response to the lack of progress on the search for such intermediates. Mitchell’s hypothesis clarified that the hydrogen ion concentration gradient in the space separated by the biological membrane causes ATPase to reverse its rotation and synthesize ATP. For me, the physical structure of biological membranes allows concentration gradients and is the basis of functions of life that cannot be accomplished by chemical reactions. It is hard to imagine how ATPase activity is related to the circadian clock, but I thought there must be something to these correlations and I wanted to examine this hypothesis further. The first report was that the ATPase activity was very low, but it seemed to be temperature-compensated. Immediately, I asked her to examine the ATPase activity of period mutants. The results were on my desk the night I left for the SRBR conference in Florida in 2006. The graph plotted the ATPase activity of three mutant KaiCs, including the wild type, against the period. I looked for a piece of graph paper and plotted the ATPase activities against the frequencies and was surprised to see that they fell on a straight line. Furthermore, the line appeared to pass through the origin. I knew immediately that this was something unusual, so I rushed to Florida, leaving a note asking for a follow-up test. I was mildly excited when I received an e-mail the day after I arrived there. Previously a member of my lab had asked me, “What kind of data are you expecting?” I replied identifying which activity of the enzyme correlates with the period and is not affected by temperature should be our target to understand the circadian clock. The ATPase activity of KaiC found by Terauchi was exactly what I had expected. I thought I had finally captured the essence of the circadian clock. In Florida, I did not feel like going to the meeting. By the end of the conference, I had a vague idea that the intramolecular tension created by the ATPase of CI could be the pacemaker of the harmonic oscillation and that the phosphorylation cycle of CII was synchronized and coupled to it. (To see the members of my laboratory in 2003, see Fig. 2.)

44

T. Kondo

Fig. 2 Lab members under cherry tree in bloom in Nagoya (2003). I greatly thank all of them. The members who are mentioned in the text are labeled with their names

21

Review and Preview

Up to this point, this chapter has been a review of my research from graduate school to the present. I don’t show the data for each research topic, but I have already published those data as research papers, so please refer to them if you are interested. What follows below is an overview of the ongoing research. Many of the contents have been introduced at conferences, but the data have not yet been published as papers, so I hope you will read the following as a “preview.” Also, please read the chapter by Kumiko Ito-Miwa, Kazuki Terauchi, and me in this volume. We are planning to publish these data and interpretations as a paper as soon as possible. This part of the chapter requires a change of paradigm for us, and may not be familiar to the readers. I hope you will enjoy reading the remainder of this chapter as an introduction to a new view of the circadian oscillator.

22

Inside KaiC

It took us 8 years to realize the significance of this, even though we reported in our 1998 paper that the KaiC sequence clearly describes an ATPase with two typical P-loops. In the meantime, we, and probably many other researchers, were not thinking deeply about this fact. We needed to be as specific as possible. The following are the three points that were known about the ATPase activity of KaiC up to that point.

Around the Circadian Clock: Review and Preview

45

1. The circadian period of cyanobacteria is greatly changed by mutation of KaiC, but its ATPase activity is directly proportional to the circadian frequency (the reciprocal of the period) and is also temperature-compensated. This activity is also observed in KaiC with only CI, and is due to the ATPase of CI. 2. This relationship is perfectly valid even in the absence of KaiA and KaiB and no rhythm at all; when KaiA and KaiB are added and a rhythm is generated, a small additional ATPase activity (2–3 ATP/day) is added, although it is thought to be necessary for phosphorylation. This part is temperature-dependent. 3. The ATPase activity of CI is 10 ATP/day in the wild type, which is extremely low. With this small amount of activity, KaiC is able to maintain its period accurately (within 10 min/24 h). Terauchi’s data that the speed of the circadian oscillation is directly proportional to the activity of ATPase seemed extraordinary to me. The fact that the most important parameter of the circadian clock, the temperature-compensated period, is determined by a single enzyme activity made me think that this activity is the most fundamental relationship in the circadian clock. In physiology, when a measurable activity is directly proportional to a parameter of a phenomenon, we assume a close causal relationship between the two. More importantly, whether KaiC is phosphorylated or not, whether it binds KaiA or KaiB or not, KaiC “remembers” its temperature-compensated circadian period and exhibits it as its rate of ATPase activity. When I obtained this ATPase activity in the lab, the phosphorylation of KaiC and its binding to KaiB and KaiC temporarily disappeared from my scope. The period is not determined by the phosphorylation rhythm or the discrete binding of KaiC to KaiB and KaiA. However, few people seemed to be surprised by these data. I have never experienced a circadian clock parameter of period being directly proportional to a simple enzyme activity. And since the activity is the ATPase activity of KaiCI, and not of some unspecified enzyme in the cell, it is already clear that period of the clock is encoded in the structure/function of KaiC, and therefore KaiC “remembers” its period. Furthermore, the main function of KaiC is first and foremost its ATPase activity. I have often heard the objection, “You call it a period, but there is no rhythm.” This criticism sounds plausible, but it implicitly assumes that the period is determined as a result of rhythm. How then do we explain the relationship between ATPase and frequency, where no rhythm is occurring? I didn’t think this was a coincidental meaningless relationship, but the reason I thought this was that the frequency of an oscillator made up of vacuum tubes and transistors is determined by coils and capacitors even when the power is turned off and the oscillator is not operating at all, and these passive components store a stable period as a constant. This insight was based on my experiences with radio work in the past. The core of these electric circuits are vacuum tubes, transistors, and operation amplifiers, which are called active components because they consume power constantly and are used to amplify signals. In contrast, components such as resistors, capacitors, and coils are components that characterize the features of a circuit, but their electrical characteristics are fixed to each component and remain constant regardless of their operation.

46

T. Kondo

These passive components, when well combined (condenser and coil or resistor), do not consume energy and can vibrate in harmony. By successfully connecting them to the control input and output circuits of a vacuum tube or transistor, stable electrical oscillations can be generated. This circuit is equivalent to a mechanical clock, or even simpler, a stationary pendulum that “remembers” its period. KaiC is an enzyme that hydrolyzes ATP, but is it possible to precisely control the 24-h period with only 10 chemical reactions per day? In addition, how can this activity be temperature-compensated? It seems almost impossible to measure time by a single chemical reaction. It is also difficult to explain how temperaturecompensated activity can be measured with a precision of 10 min in 24 h. Although this may seem contradictory, it is necessary to consider the generation of physical properties by chemical reactions. Although this may not be possible in an enzymatic reaction as a catalyst, Mitchell’s idea may be possible if we take advantage of the fact that the ATPase is characterized by the generation of mechanical torque. The precise temperature-compensated ATPase activity of 10 catalytic events per day in KaiC cannot be explained by chemical reaction control; the CI-ATPase activity would have to be controlled by negative feedback rather than by its elementary activity. If the negative feedback is strong enough, the high activity can be suppressed to almost zero. The activity controlled by the negative feedback will be stable even if the elementary activity (activity without feedback) is subjected to changes due to temperature and other factors. It is important to note that if the feedback circuit is affected by temperature or other factors, the stability will be lost. Many feedback systems of intracellular metabolism are performed by biochemical reactions, so it is possible to suppress the activity by that kind of process, but for the suppression to be temperature-compensated, the feedback pathway cannot contain biochemical reactions. Is it possible to create a feedback pathway that does not contain biochemical reactions? We need to remember that KaiC is an ATPase, which can generate mechanical torque, and if that torque acts back on the molecule that generated the torque (i.e., KaiC), it will no longer only include biochemical reactions, but will be a physical/mechanical process. In this way, the torque generated by the CI-ATPase would be an activity that is not affected by temperature. Thereby biochemical reactions could be converted into physical/mechanical processes using direct feedback of the ATPase. Intramolecular mechanical feedback can create strains in KaiC or KaiC hexamers and establish local tensions. And the tension dynamics would most commonly be expected to follow Hooke’s law. Of course, it is not entirely clear where this tension is generated, but we might expect it to be generated in KaiC or in a specific region of the hexamer. Then, if such tension is generated, the dynamic state will cause a small amount of energy leakage proportional to the strength of the tension, and the activity of KaiC will be allowed by that value, which will be the temperature-compensated rate of ATPase activity that we discovered. Since these KaiCs obey Hooke’s law, if the equilibrium is shifted by any means, the restoring force back to the balanced position will generate harmonic oscillations. In normal experimental operation this would be possible by stepping up to a

Around the Circadian Clock: Review and Preview

47

rhythmically allowable temperature. In this way, KaiC can maintain its physical spring using the energy of ATPase decomposition that occurs 10 times a day. But can it sustain the oscillation for 24 h? It may seem almost impossible, but note that harmonic oscillation does not consume energy. If there is damping due to friction, for example, the oscillation can continue as long as the energy can be replenished at every cycle.

23

Harmonic and Relaxation Oscillation

It is basically difficult to imagine a circadian oscillation that has both a perioddetermining mechanism and an amplitude-maintaining mechanism. It is important to recognize that the characteristics of the circadian period, namely, temperature compensation and accuracy, are quite different from those of relaxation oscillations, which tend to explain stable amplitudes. It must be understood that, as entities, they are assumed to be physical phenomena and chemical reactions, respectively, and as components of life phenomena, they correspond to passive processes with no (or little) energy consumption and active processes linked to energy consumption, with completely opposite properties. These two oscillations have been variously discussed as opposing models (Bunning 1967). A comparison of these models is given in Table 1 (Table 1). There is not enough time to discuss them in detail here, but what is clear from the analysis of KaiC is that the combination of these two oscillations constitutes a circadian clock that combines the features of the two opposing models. If the circadian clock is to adapt to the physical phenomenon of the earth’s rotation, then physics-based harmonic oscillations are expected to have properties as a mechanism to determine the circadian period. On the other hand, it is natural to assume that the persistence of the oscillation is maintained by chemical reactions. Bunning often discussed the contrast between harmonic and relaxation oscillations, Table 1 Harmonic oscillation and relaxation oscillation

Example

Harmonic oscillation Physics Hooke’s law F ¼ -kx Mass connected to Spring

Period Temperature effect Amplitude effect Amplitude Energy for cycle Waveform Persistence

Fixed by physical constant Q10 ca. 1 No Variable by energy Zero Linear (sine wave) No

Mechanism

Relaxation oscillation Cycle of chemical reactions Tension and relaxation Pipet washer/deer repeller “Shishi-odoshi” Sum of steps (variable) Q10 ¼ 2–3 (if chemical process) Yes Fixed by threshold level Constant consumption Discontinuous(triangular/square) Yes

48

T. Kondo

and Pittendrigh’s A oscillator/B oscillator model was the basic model for the analysis of circadian rhythms. A starting point for understanding the circadian clock is to understand that the mechanism that determines the period, the pacemaker, may be a very difficult process to design within a cell. On the other hand, many biological oscillations other than the circadian oscillation are self-excited. It should be noted that the selfsustained oscillations of life are much more diverse than the circadian oscillations, and their analysis is usually rather different from that of the circadian oscillations. It seems to me that it would have been futile to seek commonality among circadian rhythms and ultradian rhythms with short periods or infradian rhythms with long periods, although there may be exceptions. If the harmonic oscillation is the pacemaker, then the oscillation is a continuous process. That is, there would be no discontinuous steps in the cycle, and no energetically stable state. This means that the peaks and troughs of the oscillation would be able to transition smoothly to the next state. On the other hand, in a relaxation oscillation with discontinuous state transitions, a stable state is expected to exist in the waiting time between steps. This difference may perhaps have significant implications for structural biology, which attempts to analyze crystal structures.

24

About Mechanical Clocks

Time in our lab has passed without us being able to identify a specific mechanism for the coupling between CI and CII. It is clear that the period is determined by CI and the phosphorylation cycle is driven by CII, and that the two are coupled, but it was difficult to see how exactly they are coupled. Forty years have already passed since I began my research on circadian clocks, but I realized that I had never been interested in mechanical clocks because I thought they were completely different from circadian clocks. I was aware of the similarities between biological mechanisms and engineering design, but I never imagined that it would apply to my research. When I was a child, I remembered taking apart a mechanical clock and being scolded by my parents for not being able to put it back together, but I had never been interested in how it worked. I had looked into the literature and found that there was a device called an escapement built into mechanical watches and clocks, but I could not figure out how it worked. One night, I decided to see how it worked, since seeing is believing, but there were no more mechanical clocks in my house. So I took a look at Amazon.com. When I searched for mechanical timekeepers, I found almost all were watches. The prices were astonishing, and it seemed difficult to see the details inside. Not giving up, I scrolled further and found a tabletop pendulum clock. This one was made in Germany, but it cost about US$400, which seemed manageable, and it didn’t have a cover so I could see the escapement. I was a little hesitant when I saw the look on my wife’s face when I told her that I was thinking about paying US$400 for an old-fashioned mechanical clock, but I nevertheless ordered it, and 2 days later the package arrived.

Around the Circadian Clock: Review and Preview

49

I scrutinized the movements of the gears and cogs and other parts within the clock, but I couldn’t tell whether the pendulum or the spring came first. As a result, I was able to understand the mechanism of the escapement, in which the harmonic and relaxation oscillations are skillfully combined to move the harmonic oscillation (pendulum) as freely as possible and engage the escapement wheel, which is rotated by the energy of the spring with a period shorter than that of the pendulum, to drive the hands of the timekeeper with the precise movement of the harmonic oscillation. At the same time, the escapement uses this control mechanism to provide the pendulum with an extremely small amount of energy, called the impulse, to sustain the harmonic oscillation. The design of this mechanical clock is said to have been developed by Huygens using Galileo’s isochronism, but when I looked at the design that combines two oscillations to measure accurate time autonomously, I could not believe that it was a coincidence that KaiC was composed of two ATPases. Later, I discovered that the escapement of a mechanical watch/clock has various visualizations on YouTube, so I recommend that you search for “escapement” on YouTube and watch the videos. The escapement works by the harmonic oscillation temporarily stopping the relaxation oscillation and the energy of the relaxation oscillation accelerating the harmonic oscillation, but the two actions are performed in a very short period of time, so please watch a version on YouTube that can be run in slow motion. Also note that the harmonic vibrations need to be as free as possible to allow the oscillation to tick away with accurate timing.

25

Dual ATPases Coupling Model for KaiC Circadian Oscillator

As explained earlier, the KaiCI-ATPase could function as a harmonic pacemaking oscillator. However, with this pacemaker alone, the oscillation will stop sooner or later, and even if it “remembers” an intrinsic 24-h period it cannot transmit it to the cyanobacterial cell. The KaiCII-ATPase (i.e., the ATPase of the CII domain of KaiC), in cooperation with KaiA and KaiB, is responsible for the phosphorylation and dephosphorylation of the two neighboring amino acids of CII. The CII-ATPase of KaiC cooperates with KaiA and KaiB to autonomously repeat the phosphorylation and dephosphorylation of the two neighboring amino acids of CII. The periodicity of this cycle is naturally affected by temperature and various environmental factors, but if its progression is gated by the pacemaker of CI, the periodicity can take on the harmonic nature of the CI oscillation. If that is the case, CI can repeatedly receive a very small amount of mechanical energy from CII by its phosphorylation status (called “impulse”) at the optimal phase of the CI cycle so that the CI oscillation is stimulated to continue cycling. In this way, the CI can sustain a harmonic oscillation without damping by a very small impulse. The CII cycle thereby becomes more stable and various metabolic activities can be controlled in a 24-h cycle. In other words, two coupled Kai oscillators (CI and CII) now function

50

T. Kondo

as an integrated circadian oscillator with two different characteristics of stable period and robust amplitude. Note that this configuration is the same as the design of mechanical clocks coupled with an escapement mechanism. How is such a mechanism possible in KaiC? It will take a long time to figure out the specifics, since the movement is probably very small at the atomic level.

26

ATPase Measurement

The ATPase activity of KaiC is the degradation of 10–15 ATP per day. If this activity were intracellular, it would be impossible to measure. However, we were able to measure it by taking advantage of the in vitro experimental system. First of all, since the biological clock mechanism works perfectly within a reaction solution of only the three proteins and ATP, we can introduce as much of the reaction solution in vitro as possible directly into ADP-measuring HPLC column without any treatment, and expand it to make sure that there is no interference from unknown substances, and quantify the ADP produced. In this way, we were able to perform stable measurements with maximum sensitivity. I am deeply grateful to the four researchers (Kazuki Terauchi, Yoriko Murayama, Kumiko Miwa, and Tomoaki Muranaka) who took on the challenge of this seemingly impossible measurement and patiently contributed to the observation of the movement of the circadian clock. In particular, Miwa greatly improved the method to measure the fluctuating waveform of ATPase activity reproducibly, and also to separate the two ATPase activities of KaiC. Table 2 is an energy balance sheet of the timekeeping mechanism of KaiC as examined by Miwa and Muranaka. Their measurements also allow us to infer the waveform of each activity, albeit under limited conditions. Table 2 Balance sheet for ATP consumption during the KaiC phosphorylation rhythm Usage CI Sustain tension at structurally defined level (equal to leak of energy from tension in CI) Impulse to CI by CII to sustain CI oscillation (equal to energy loss by friction) Harmonic oscillation of CI tension CII phosphorylation of CII at S and T Drive CII phosphorylation cycle Total

ATP consumption (ATP/days/KaiC) 12 0.1 mM) causes the CI-ATPase to hydrolyze ATP–probably at rates that are several times higher than observed later on—and generate a mechanical torque. The torque will act mechanically on the nearby structures within KaiC and produce a reversible strain according to Hooke’s law. This strain will act on the CI-ATPase structure to deform it, causing the KaiC or its hexamer to deform elastically as stated by Hooke’s law.

28

Fundamental Frequency Problem

When you examine a mechanical clock, you will notice that the basic period is about 0.1–100 s. In any case, these rapid oscillations, which are much faster than 24 h, are slowed down by the gears that ultimately move the hands. This is due in part to the mechanical constraints of the pacemaker, such as the pendulum and balance, but also because the gears accurately reduce the frequency. In the case of a quartz clock, the harmonic oscillation of 30 MHz is slowed down by an electronic circuit called a counter. Such a short fundamental period is effective in increasing the resolution of time measurement in the case of digital clocks, but in the case of analog clocks, it seems to be limited by mechanical constraints (although it is still very interesting to note that a range of 1000 times is possible). What is the basic cycle of the circadian clock? At the current time, it is difficult to answer this question. The fact that the dynamic range of the period of harmonic oscillations is very large, and moreover that it is difficult to imagine gears and counters in protein solutions, suggests the possibility of harmonic oscillations with a period of 24 h, but the question of how long a period can be fulfilled by a harmonic oscillation is not easy to answer. Of course, the harmonic oscillation of a pendulum would be impossible, but if the torsion of the protein structure can achieve a relatively large vibrational mass and a weak spring constant, it may be possible to realize a circadian harmonic oscillation according to Hooke’s law. Recently, Kumiko Miwa used genetics and the CCD camera screening methodology to track the period displacement and found that the period changes from 0.6 days to 1 week with a change at the 402nd amino acid. The fact that the period of the resulting mutant is determined by the size of the side chain of the 402nd amino acid and that temperature compensation can be achieved even for a spectrum of periods that span about an order of magnitude (from ~16 h to ~160 h) would provide an opportunity to analyze the physical mechanism of period determination. On the other hand, the problem with the fundamental cycle is also the empirical rule that the phase response curve pointed out by the three pioneers (Bunning,

52

T. Kondo

Pittendrigh, and Aschoff) is basically a curve, even though the magnitude of the phase displacement may vary. In fact, the Atlas of Phase Response Curves compiled by Carl back in 1990 summarizes the phase response behavior of the rhythms of various organisms; the phase response curves are remarkably similar and can be used to explain the synchronization of repeated stimuli. This means that the phase response curve is repeated in a circadian period, suggesting that the basic cycle of circadian oscillations is also circadian and not composed of sub-frequencies.

29

Acknowledgment

I am indebted to the editors of this book (Carl and Dr. Michael Rust) for their kindness and patience in allowing me to write this recollection and commentary. The reminiscences are the history of my research over the past 45 years or so. I have a mixed feeling that I have done my best and that I have inconvenienced a lot of people by taking detours. However, I am very fortunate to have been able to devote 50 years of my life to the study of circadian clocks, and I would like to thank many people for their support. I am very fortunate to have been able to spend 50 years just studying the circadian clock. This was a very difficult essay for me to write, and I am afraid it may be annoying for you. However, as a researcher, I would like to share the secrets of the circadian clock revealed by cyanobacteria and their KaiC, so I have written this as well as I can. I hope you have enjoyed reading it.

30

Editors’ Note

For complementary views on the first part of this chapter (historical background), please also see: 1. “To Takao Kondo on the occasion of his retirement,” Dedication by Carl Johnson in this volume. 2. “A retrospective: on disproving the transcription-translation feedback loop model in cyanobacteria,” by Hideo Iwasaki in this volume. 3. Johnson CH, Xu Y (2009) The Decade of Discovery: How Synechococcus elongatus became a model circadian system 1990–2000, Chapter 4. In: Ditty JL, Mackey SR, Johnson CH (eds) Bacterial circadian programs. Springer, pp 63–86 For complementary views on the second part of this chapter (“Review and Preview”), please also see: “Mechanism of the cyanobacterial circadian clock protein KaiC to measure 24 hours,” by Kumiko Ito-Miwa, Kazuki Terauchi, and Takao Kondo in this volume.

A Retrospective: On Disproving the Transcription–Translation Feedback Loop Model in Cyanobacteria Hideo Iwasaki

1 Transcription–Translation Feedback Loop Model As most readers know, the “transcription–translation feedback loop” (TTFL, hereafter) model in the circadian system is, in short, as follows. Clock proteins encoded by the clock genes essential to biological clocks negatively regulate (repress) their own expression (transcription/translation), thereby generating an autonomous oscillation in gene expression as the basis of the endogenous 24-h rhythms. This regulatory loop is composed of a number of clock genes and related factors. Since the mid-1990s, clock genes have been analyzed in Drosophila, Neurospora, Arabidopsis, cyanobacteria, and mammals, and many groups worked hard under synergistically competitive conditions to elucidate the functions of the proteins encoded by these clock genes. If the transcription–translation feedback loop model is correct, we can simply predict the following. 1. The expression levels of clock gene mRNAs and clock proteins are state variables of the oscillator, and they oscillate with a time lag. 2. Complete abolishment of the rhythm in clock gene disrupted strains/lines. 3. Nullification of the rhythm by inhibiting transcription or translation. 4. Period length changes due to mutations in transcription, translation, and stability of the clock gene mRNAs and clock proteins. 5. Overexpression of clock genes attenuates or stops the rhythm. 6. Transient overexpression or transient reduction of the clock gene causes a shift in the phase of the rhythm.

H. Iwasaki (*) Department of Electrical Engineering and Bioscience, Waseda University, Tokyo, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_3

53

54

H. Iwasaki

Of these, 1, 2, 4, and 5 were first reported by studies on Drosophila pioneered by Hardin et al. (1990), but clearer results were reported for 3 and 6 in Neurospora (Aronson et al. 1994; Crosthwaite et al. 1995). It should be noted that the scheme of a genetic oscillation circuit via TTFLs that was established in the 1990s had a spillover effect to basic biology in a broader sense than just as an achievement of chronobiology. It was a good example of systems biology and synthetic biology as they emerged (more precisely, rearranged) as distinct fields around 2000. Although there are many definitions of systems biology, I will refer to systems biology as an approach to understanding biological phenomena based on mathematical and molecular biological analyses on dynamics contributed by multiple factors. Systems biology tries to find the essence of control in the relationships among multiple factors rather than in the characteristics of individual factors. An oscillation itself is not a substance, but it is an event or process. In the 1990s, the study of the molecular biology of biological clocks was a favorable application of concepts of systems biology. Although the sequences of clock genes are conserved between insects and mammals, the clock genes discovered in plants, fungi, and bacteria were not. This observation suggested that biological clocks were not inherited and conserved from a single ancestor but have been acquired independently through convergent evolution. The key idea from the studies of biological clocks in eukaryotic organisms, however, is that even though the clock proteins do not share any sequence in common, they form a transcription–translation feedback loop that generates circadian oscillations. In other words, different “actors” play the same “roles.” This is a very systems-biological perspective; namely, not to address the detailed function of each gene, but rather to study interrelationships among the clock components with the goal of understanding the essential dynamic biological phenomena. Understanding the regulatory networks among molecules is also a basis of synthetic biology, which seeks to modify or create new genetic networks. A defining example of such an approach was the creation of a synthetic genetic oscillation by Elowitz and Leibler (2000) named the repressilator. Obviously, this work builds on the preceding studies of the TTFL model. The repressilator is composed of three transcriptional repressors that negatively regulate each other. The authors initially proposed that the oscillatory network composed of the triadic oscillatory circuit was an artificial one that does not exist in natural biological clocks (unlike the simple inhibitor–activator system). However, it is an interesting irony that this triadic network has since been found in the molecular networks of plant and mammalian clocks (Ukai-Tadenuma et al. 2011; Pokhilko et al. 2012). In any case, however, it should be clearly recognized that the TTFL model has had a significant impact on other areas of biology as well, such as synthetic biology. Regarding the affinity of the TTFL study with systems biology and synthetic biology, it might be illuminating to point out another good correspondence, namely, that between mathematical analysis and chronobiology. The study of circadian rhythms has been approached mathematically since its early days. An early mathematical model that included a TTFL was proposed by Goodwin (1963). It was a novel idea devised at the dawn of the central dogma of molecular biology and in

A Retrospective: On Disproving the Transcription–Translation Feedback. . .

55

response to the genetic feedback model by Jacob and Monod (1961). The Goodwin model did not attract the attention of biologists at first. Rather, it is a work that was rediscovered and revived after the proposal of the TTFL loop model a quarter of a century later.

2 Beyond the TTFL Model In the 1990s, data supporting the TTFL model accumulated for most of the model organisms in chronobiology. The model became widely recognized as a “principle of the genesis of circadian rhythms,” regardless of species (Dunlap 1998). It was not only a guiding principle for many studies, but also an authoritarian principle at that time. For example, researchers involved in molecular studies of circadian rhythms were expected to consider their results in the context of discussing the link to a TTFL. Clock-related molecules were expected to be involved in the TTFL process in some way, and were otherwise not likely to be published in so-called prestigious journals. This trend became decisive, with the discovery of the transcription factor CLOCK in mice, which was published in 1997 (Antoch et al. 1997; King et al. 1997). There was of course a lot of data that honestly matched the transcription– translation feedback model, but there were also many papers that were forced to provide a rationale for a TTFL even when the data might have suggested other possibilities. The TTFL model became a dogma that had to be followed. From spring 1995, I was myself involved in the cloning of cyanobacterial clock genes in a small research group led by Takao Kondo and Masahiro Ishiura initially at the National Institute for Basic Biology (NIBB) and later at Nagoya University in Japan. Although I was the first person to witness the amino acid sequence of kai genes (the sequence was finally confirmed in July 1995), the most important hurdle that ultimately led to the cloning of the kai genes was accomplished by Takao Kondo, who developed an automated bioluminescence measurement device for mutant screening and complementation analyses, Masahiro Ishiura who devised clever cloning tactics, and Shinsuke Kutsuna and the late Carol Andersson for early genetic mapping. Carol was a postdoc in Susan Golden’s lab, and she came to NIBB in Okazaki, Japan for some months for collaboration. She was a vegetarian and for us it was a difficult task to take Carol to restaurants that served vegetarian food in such a small city as Okazaki: vegetarian food was not popular in Japan at that time. Anyway, the results of the collaboration were reported 3 years later in 1998 (Ishiura et al. 1998). This paper contained quite a lot of information, and it argued in favor of the TTFL model because KaiC negatively regulates its own transcription under continuous light conditions, while KaiA positively regulates the transcription of the kaiC gene. However, in reality there were some results that did not fit the TTFL model, such as the fact that the kaiBC gene expression level was not enhanced after disruption of the “negative element” kaiC gene. After being involved in the cloning of the kai gene cluster, I started to conduct biochemical analysis of the Kai proteins (Figs. 1 and 2, Iwasaki et al. 1999, 2000, 2002; Nishiwaki et al. 2000).

56

H. Iwasaki

Fig. 1 The first note describing KaiC’s autophosphorylation in vitro (my lab notebook on October 29th, 1998). After my identification of SasA as KaiC-binding protein using the yeast two-hybrid analysis (March 26th, 1998), I attempted to test if KaiC affected phosphoryl transfer from SasA to E. coli response regulator protein, OmpR. GST-tagged fusion proteins were used for SasA and KaiC without removing GST-tag in this initial assay. OmpR protein was a gift from Dr. Hirofumi Aiba (Nagoya University). Though I failed to see KaiC’s effect on SasA at that time, KaiC was unexpectedly found to be autophosphorylated. It was before we found KaiC phosphorylation rhythm in vivo: Taeko was aware that band shift appeared cycling on western blot on December 29th, 1998, but it took much longer before she finally showed that it was derived from phosphorylation using the phosphatase assay on November 28th, 1999

While molecular studies in animals and fungi were focused on DNA-binding factors, the relationship between the KaiC protein and transcription/translation processes was not clear, which bewildered us. Regardless to say, however, the transcription– translation feedback model was a MODEL, and not a rational necessity. It states that oscillations can occur through gene networks, but even to this day, the TTFL model

A Retrospective: On Disproving the Transcription–Translation Feedback. . .

57

Fig. 2 The first data showing stimulation of KaiC phosphorylation by KaiA in vitro (my lab notebook on October 14th, 1999). GST-fused KaiA, KaiB, and KaiC and the N-terminal segment of SasA were mixed to test if KaiC phosphorylation was altered. At this initial assay, KaiA was found to activate KaiC’s autophosphorylation, while I failed to detect the effect of KaiB. The effect of KaiB to antagonize the KaiA’s function was found later by Stanly Williams, a postdoc in Susan’s lab

does not explain why circadian rhythms are 24 h in duration or why they are temperature compensated. Thus, I felt the necessity to seek for alternative possibilities and began to carefully read scientific papers that described results and ideas that were incompatible with the TTFL model. In the 1960s, Sweeney and colleagues had reported that the photosynthetic rhythm of the giant unicellular green alga, Acetabularia, persisted for many days after removal of the rhizoid that contained the nucleus (Sweeney and Haxo 1961). Moreover, Sweeney et al. (1967) showed that the rhythm of photosynthesis persists even after the addition of transcription and translation inhibitors. This approach was retrospectively similar to our later paper in cyanobacteria (Tomita et al. 2005) in its experimental design. Although these papers were cited before the 1980s, after the rapid development of clock gene studies in Drosophila, these studies were relatively ignored, partly because this species was not genetically tractable for molecular studies. Moreover, most researchers did not consider it to be representative of “normal” cells, because Acetabularia cells can survive for several days without a nucleus. The controversy over whether transcription and translation in nonnuclear organelles is important for the development of circadian rhythms has also not been

58

H. Iwasaki

settled. Initially, Sweeney et al. (1967) showed that at least partial inhibition of protein synthesis in organelle had little effect on the period of the photosynthesis rhythm in Acetabularia. On the other hand, Schweiger and colleagues argued on the basis of using other inhibitors that the 80S ribosome-mediated translation process is essential for maintaining the rhythm (Mergenhagen and Schweiger 1975). Moreover, many studies in Acetabularia, especially performed by Schweiger’s lab, had some difficulties in terms of the number of experiments and reproducibility. Therefore, many chronobiologists were skeptical of the results reported from Acetabularia. Unfortunately, due to Sweeney’s sudden and unexpected death and the lack of a successor to Schweiger, these findings were not confirmed for a long time. In the 1990s, many molecular chronobiologists were totally silent about those reports. Since I was fascinated by the old papers on Acetabularia, to me their stance showed an unavoidable bias. There had also been a small number of reports that null-mutant strains of some classical clock genes such as per in Drosophila also retain their rhythms, even if only partially. For example, Helfrich and Engelmann (1987) reported that the per0 flies may show circadian periodicity, but this paper did not attract much attention. Akira Matsumoto also confirmed a clear locomotion rhythm of this mutant under certain conditions and noted it in Japanese in 2002 (Matsumoto 2002). In fact, Jeff Hall told him, “That’s absolutely due to your wrong choice of strains,” including some abusive words (Akira Matsumoto, personal communication). In addition, Yang and Sehgal (2001) had discovered that constitutive (non-rhythmic) induction of per and tim rescued rhythms in per;tim double knockout background in Drosophila: this paper reported that the translational rhythm of the clock proteins was also restored at the same time, and they argued that the TTFL was complemented by some compensatory mechanism. A similar situation occurred in Neurospora, in which the TTFL model had been strongly advanced by the labs led by Jay Dunlap and Jennifer Loros. Martha Merrow was originally involved in the quantitative work in the Dunlap/Loros labs, but she later teamed up with Till Roenneberg to report in 1999 that circadian properties remain even in functionally deficient strains of the frequency clock gene in Neurospora; they used imaginative and thoughtful physiological experiments to demonstrate these circadian properties (Merrow et al. 1999). This was a somewhat complicated paper, but it was of high quality and I was very impressed and encouraged by it. However, Till and Martha said to me that it was greeted with bashing and open disregard by most supporters of the TTFL model. Moreover, Pat Lakin-Thomas reported that Neurospora strains with mutations within two genes in the lipid metabolic system showed a long-period rhythm in null-mutant backgrounds without frequency (Lakin-Thomas 1998; Lakin-Thomas and Brody 2000). However, these studies of Pat were also ignored or disputed by some TTFL-supporters to be considered as non-circadian metabolic rhythms because of the lack of a stable 24-h oscillation. Each of these results was inconsistent with the “brilliant” story of the dogmatic TTFL theory, but left enough “wiggle-room” for TTFL-supporters to rationalize the results to be apparently interpretable in the context of TTFL-based oscillations. Thus, the TTFL model itself was inviolable.

A Retrospective: On Disproving the Transcription–Translation Feedback. . .

59

What about in cyanobacteria? The first paper on the cloning of kai genes (Ishiura et al. 1998) also advocated and supported the TTFL with considerable accuracy; this fact weighed heavily on us for the next decade. We felt a strong sense of urgency to find a transcriptional regulator that should close the TTFL for kai gene expression. A great deal of time was spent testing KaiC’s DNA-binding ability and screening for transcriptional regulators, but the results were not always encouraging. In this context, Tetsuya Mori and colleagues in Carl Johnson’s lab reported a DNA-binding activity of KaiC and proposed the possibility that KaiC directly drives genome-wide expression rhythms by affecting the chromosomal topology, rather than via specific transcription factors (Mori et al. 2002). Because of its DNA-binding activity and RecA-like formation of KaiC hexameric structures observed by electron microscopy, the authors suggested KaiC might be a factor that alters the structure of DNA, such as a DNA helicase. At about the same time, Yao Xu and Tetsuya in Carl’s lab and Yoichi Nakahira in Takao’s lab independently found that replacing the kaiBC promoter with the IPTG-inducible trc promoter derived from an E. coli promoter allowed the bioluminescent rhythm to continue that monitors the kaiBC promoter activity (Xu et al. 2003; Nakahira et al. 2004). A strict TTFL interpretation would have predicted that there was a specific clock-regulated transcriptional factor that interacted with a specific motif in the kaiBC promoter; the generic trc promoter derived from a different organism certainly did not fit the TTFL prediction. Based on these results, there was a growing realization that kaiBC promoter-specific transcription factors might not be important. These were major policy shifts, and little by little cyano-clock researchers navigated their own course away from that of researchers who studied the eukaryotic TTFL. However, we did not yet necessarily deny that the core of the oscillation was based on kaiBC gene expression. In retrospect, it was not these analyses around DNA binding/metabolism that most closely approached our anti-TTFL paper in 2005 (Tomita et al. 2005) but some results reported in a paper by Yao and Tetsuya in Carl’s lab in 2000 (Xu et al. 2000). This paper was the first to report oscillations in the abundances of KaiB and KaiC proteins in the light. More importantly, they found that the phase of the bioluminescence rhythm hardly changed when the length of the dark period, which started at twilight (“CT12”), was gradually extended to about 48 h. This means that the clock continued to tick even during the dark period. Moreover, they showed that the phase did not change when the protein synthesis inhibitor chloramphenicol was given. Nevertheless, the authors showed a faint rhythmicity in the KaiC signal on a western blot in the continuous darkness, and concluded that KaiC abundance continued to oscillate in the dark, even when chloramphenicol was given. Retrospectively, the rhythmic signal of KaiC observed at that time was not total KaiC abundance level, but a band shift associated with a rhythm of KaiC phosphorylation. The phosphorylation rhythm of KaiC (Iwasaki et al. 2002), however, was not known at that time. In any case, the flow of this paper, which was so close to the truth but whose results were interpreted in the context of the TTFL scheme, is itself a testimony to the strength of the TTFL dogma at that time. One of the authors, Tetsuya Mori, was a rebellious debater (in a good sense), and he, like me, had a strong sense of discomfort with the TTFL model. Nevertheless, the TTFL model was too big a barrier to

60

H. Iwasaki

overcome in the paper. For Carl Johnson, there was another personal background in favor of a possible rhythm of KaiC abundance in the dark, because he was the very person to find the first protein abundance rhythm using immunoblot in any organisms. In 1984, the same year of the cloning of period in Drosophila, Carl reported circadian changes in abundance of the luciferase protein in the dinoflagellate alga Gonyaulax (Johnson et al. 1984). According to him, in 1984 very few people believed that there would be daily rhythms of protein abundance because they thought it would be too “wasteful” to synthesize and degrade proteins on a 24 h timescale. So, Carl had a personal bias toward finding other examples of circadian changes in the abundance of key proteins. Since he did not know much about posttranslational modification, he recalls, “I never considered the possibility that Yao’s data might indicate a rhythm of protein modification,” (Carl Johnson, personal communication). By the way, I had also been investigating the accumulation profiles of KaiB and KaiC since I obtained antibodies to KaiB and KaiC in 1997, but at first we could not confirm any accumulation rhythms at all, and I once started to write a paper to report no oscillation in protein abundances of the Kai proteins. Then, I received a preprint of Yao and Tetsuya’s paper, and I remember feeling relieved that I hadn’t made a premature conclusion to publish a wrong paper. In 2003, Jun Tomita, a graduate student in Takao’s lab, was analyzing the expression pattern of kai genes under various light–dark cycles or during darkpulses for driving phase shifts. He found that transcription of the kai genes was immediately suppressed and the mRNA was gone during the dark period. From the above-mentioned analysis by Xu et al. (2000), we knew that the circadian oscillation seems to continue during the dark period. Therefore, we investigated what happens to the protein during longer exposures to darkness: while all clock gene mRNAs disappeared immediately after the dark onset, the level of the clock proteins remained constant and the oscillation in the phosphorylation level of the KaiC protein clearly persisted without any decay (first discovered on July 10th, 2003; see Fig. 3). However, we could not rule out the possibility that another loop may be operating at the level of transcription and translation in the dark period. Therefore, Jun and I added excess amounts of transcriptional and translational inhibitors, and we confirmed the phosphorylation rhythm continued. This was the first time that the TTFL model was completely disproved. We also found that the phosphorylation rhythm of KaiC without transcription/translation processes was clearly temperaturecompensated. Furthermore, the cycle was similarly altered in mutants that altered the cycle of gene expression rhythms. I summarized these results and first submitted them to the journal Nature in February 2004, but it was rejected without going to reviewers, and the editor stated that “the journal was not interested in the cyanobacterial clock.” ...What?? Up to this point, Takao Kondo had been relatively passive as far as these results were concerned, because there were some differences of opinion between me—who had been thinking about possibilities other than the TTFL for a long time—and Takao, who had pioneered the cyanobacterial TTFL model since the initial kai paper. There was also a difference between us in what we had been focusing on as methods for measuring rhythms. Takao was the researcher who had originally developed the

A Retrospective: On Disproving the Transcription–Translation Feedback. . .

61

B

kaiC 0 4 7.5 (D 0) 12.5 16 20 24 28 32 36 40 44 48 (h) kaiC

A

Fig. 3 The first data showing that KaiC phosphorylation rhythm persists in the dark in vivo (Jun’s lab notebook on July 10 2003). (a) Photo of the original X-ray film (western blotting). (b) Cropped image of the KaiC profile. After entraining the clock with a single 12-h dark, cells were grown in a continuous culture system for 7.5 h in the light, and then incubated in the continuous dark. The upper and lower bands represent phosphorylated and unphosphorylated form of KaiC

automated bioluminescence monitoring systems. Naturally, Takao was proud that his group’s most important weapon was the use of his advanced systems for monitoring rhythms of gene expression. In fact, they were a magical weapon that enabled many novel results. But the bioluminescent reporter is a device that monitors transcription in the first place, so it cannot be applied to investigate rhythms that do not involve transcription. Also, in the case of Synechococcus, the bioluminescent luxAB reporter cannot be used in the dark, not only because it requires transcription, but also because the enzyme’s substrates ATP and FMNH2 are at reduced levels in the dark; consequently, even if luciferase is present in the cells in the dark, there is minimal bioluminescence emission in the absence of the substrates. Therefore, rhythm analyses by this method relied heavily on monitoring the promoter activity in the light in the early phase of cyano-clock studies. The results of Tomita’s and my experiments may have prompted Takao to make a major policy change. I guess it took him some time to accept this change, though he immediately understood its implications when we brought our data to him. On the other hand, for us young scientists, it was a huge relief to be released from the demanding TTFL dogma, and in addition, also from the necessity of always using the bioluminescence reporter. At that time, I had almost finished writing a revised manuscript for the journal Science with additional data that reinforced our conclusions, and then Takao said that the paper should show a new possibility, and came up with an idea to include a strange preliminary result that had been performed about half year earlier by Masato Nakajima, a postdoc in Takao’s lab. It was a biochemical experiment showing that

62

H. Iwasaki

the phosphorylation and dephosphorylation rates of purified KaiC in vitro were temperature-compensated. When I told Masato about Takao’s suggestion, he had forgotten that he had conducted that experiment. In response to these suggestions, Masato reanalyzed the data and we included it in the paper. It was accepted on November 5th in 2004 (Tomita et al. 2005). But 4 days before the paper was accepted, a magical moment happened in the lab.

3 Establishment of the In Vitro Reconstitution System The question left to us was, “What is the alternative to a TTFL for the mechanism of the oscillation?” Cells producing the KaiC phosphorylation rhythm in the presence of an excess amount of rifampicin in the light cannot propagate even after washing the inhibitor out, implying that the cells had died or irreversibly passed beyond a critical point at which propagation is possible (Tomita et al. 2005; Takano et al. 2015). It meant that the circadian oscillation was maintained even in dead (or ghost) cells. It was clear that the biochemistry of the Kai proteins was at the core. However, other cellular functions, such as subcellular localization of the clock proteins or interaction of the Kai proteins with DNA or other proteins, were not yet ruled out. We therefore had to explore a variety of directions. For example, I planned to follow a process of standard biochemistry to observe enzymatic activity (here, the KaiC phosphorylation rhythm) using partially destroyed cells, crude extract, or subcellular fractionations derived from cyanobacterial cells. On the other hand, Takao suggested to mix KaiA, KaiB, and KaiC to see if we could create a rhythm, and asked Masato to take charge. This idea had actually been a joke in the past. We were so desperate to search for the putative transcription factors and so we asked each other, “Can’t you just get the rhythm only from Kai proteins?” Even in 2004, this was still a joke: starting from recombinant proteins to get rhythm sounded too optimistic to me (and even Masato) at the initial phase. But we had reached the point where we thought we might be able to figure out some interesting protein dynamics, if not a rhythm. Masato has a background in protein biochemistry and dramatically improved the quality of the purified Kai proteins. He began the reconstitution experiments by fixing the amount of KaiA and changing the concentrations of KaiB and KaiC in various ways. Although he sometimes observed a transient increase in KaiC phosphorylation, he did not see any periodic fluctuations. Then, Masato asked Taeko Nishiwaki and Keiko Imai to test more experimental conditions. In the early evening of November 1st in 2004, Masato smiled at me and said, “maybe we’re seeing some kind of rhythm.” It was on an SDS-PAGE gel that Taeko and Keiko Imai had sampled with increased KaiA concentration, which was being destained after CBB staining. Taeko had already returned home, but when we looked carefully at the gel with Masato and a few others, we could see that the phosphorylation bands were indeed changing periodically. According to Taeko, she had also seen the results, but thought, “This is a big deal . . . I’ll think about it tomorrow,” and went home. She came back the next day and found a crowd of people around the gel.

A Retrospective: On Disproving the Transcription–Translation Feedback. . .

63

Fig. 4 The very first data of the KaiC phosphorylation cycle reconstituted in vitro (Taeko’s lab notebook on November 1st, 2004). Taeko incubated recombinant KaiA (0.04μg/ml), KaiB (0.04μg/ ml), and KaiC (0.2μg/ml) in the presence of 1 mM ATP at 30  C. Taeko and Keiko sampled from October 27th for 2 days at the indicated times, and subjected to SDS-PAGE and Coomassie Brilliant Blue (CBB) staining. The upper and lower bands represent phosphorylated and unphosphorylated form of KaiC. The period length of these initial data was ~28 h. Taeko and Keiko tried five different conditions on the same date, and they found a rhythm in one of them

Anyway, we all witnessed the establishment of the world’s first in vitro reconstitution of the circadian clock: at that time, the period of the KaiC phosphorylation rhythm in this preliminary result was ~28 h (Fig. 4). It was a very strange shock. I do not follow a particular religion, but it was a sublime moment, like something coming down from heaven. “How can this thing really happen?” It was a glimpse into the abyss of nature. Richard Feynman once said, “What I cannot create, I do not understand,” which could be paraphrased “If you cannot make it, you do not understand it.” In vitro reconstitution of the KaiC phosphorylation rhythm (Nakajima et al. 2005) has opened up a whole new horizon of biological clock research, but not necessarily because we understood it. Many researchers, including Takao and many colleagues contributing to this book, are making progress in their quest to understand it. We are still in the midst of our research, but we now know that there is a beautiful alternative to the TTFL model. In eukaryotes, the TTFL model is still basically supported, and its status as a conservative mainstream mechanism has not wavered. In cyanobacteria, in addition to the posttranslational oscillator (that is the in vivo counterpart of the in vitro reconstituted oscillator) the TTFL mechanism is still considered to play an important role in the light as well. On the other hand, it has now been reported that some processes other than TTFL play an important role in clock systems even in mammals and fungi. At the very least, hence, it is no longer taboo to refer to possibilities other than the TTFL model.

64

H. Iwasaki

Acknowledgments There are many colleagues and collaborators to whom I owe a debt of gratitude, but I would like to thank initially Takao Kondo, Masahiro Ishiura, Carl Johnson, and Susan Golden for pioneering the field and guiding me in the early days. Until around 2003, there was active interaction among these groups, especially among students, and we had a habit of showing each other manuscripts before they were submitted. When I think about it now, it was an honor for me to have my first manuscript very carefully corrected by Susan even though she was not an author. Without the friendly collaboration with Takao Kondo, Shinsuke Kutsuna, Taeko Nishiwaki, Jun Tomita, Masato Nakajima, Yohko Kitayama, Keiko Imai, Hakuto Kageyama, Hiroshi Ito, Kazuki Terauchi, and other excellent teammates, the results I have described here would not have been possible. Also, Tetsuya Mori, Akira Matsumoto, Till Roenneberg, Martha Merrow, Pat Lakin-Thomas, and Hiroki Ueda, with whom I discussed possibilities outside of the TTFL, were my few comrades at the time. I would like to thank them for their support. I thank Taeko and Jun for kindly providing copies of their valuable notebook for this chapter. Finally, I thank Carl for reading the manuscript carefully with valuable comments, and Taeko for providing detailed information on the process of in vitro reconstitution studies.

References Antoch MP, Song EJ, Chang AM, Vitaterna MH, Zhao Y, Wilsbacher LD, Sangoram AM, King DP, Pinto LH, Takahashi JS (1997) Functional identification of the mouse circadian Clock gene by transgenic BAC rescue. Cell 89:655–657 Aronson BD, Johnson KA, Loros JJ, Dunlap JC (1994) Negative feedback defining a circadian clock: autoregulation of the clock gene frequency. Science 263:1578–1584 Crosthwaite SK, Loros JJ, Dunlap JC (1995) Light-induced resetting of a circadian clock is mediated by a rapid increase in frequency transcript. Cell 81:1003–1012 Dunlap JC (1998) Common threads in eukaryotic circadian systems. Curr Opin Genet Dev 8:400–406 Elowitz MB, Leibler S (2000) A synthetic oscillatory network of transcriptional regulators. Nature 403:335–338 Goodwin BC (1963) Temporal organization in cells; a dynamic theory of cellular control processes. Academic, London Hardin PE, Hall JC, Rosbash M (1990) Feedback of the Drosophila period gene product on circadian cycling of its messenger RNA levels. Nature 343:536–540 Helfrich C, Engelmann W (1987) Evidences for circadian rhythmicity in the per0 mutant of Drosophila melanogaster. Z Naturforsch C J Biosci 42:1335–1338 Ishiura M, Kutsuna S, Aoki S, Iwasaki H, Andersson CR, Tanabe A, Golden SS, Johnson CH, Kondo T (1998) Expression of a gene cluster kaiABC as a circadian feedback process in cyanobacteria. Science 281:1519–1523 Iwasaki H, Taniguchi Y, Ishiura M, Kondo T (1999) Physical interactions among circadian clock proteins, KaiA, KaiB and KaiC, in cyanobacteria. EMBO J 18:1137–1145 Iwasaki H, Williams SB, Kitayama Y, Ishiura M, Golden SS, Kondo T (2000) A KaiC-interacting sensory histidine kinase, SasA, necessary to sustain robust circadian oscillation in cyanobacteria. Cell 101:223–233 Iwasaki H, Nishiwaki T, Kitayama Y, Nakajima M, Kondo T (2002) KaiA-stimulated KaiC phosphorylation in circadian timing loops in cyanobacteria. Proc Natl Acad Sci U S A 99:15788–15793 Jacob F, Monod J (1961) On the regulation of gene activity. Cold Spring Harb Symp Quant Biol 26:193–211 Johnson CH, Roeber JF, Hastings JW (1984) Circadian changes in enzyme concentration account for rhythm of enzyme activity in Gonyaulax. Science 223:1428–1430

A Retrospective: On Disproving the Transcription–Translation Feedback. . .

65

King DP, Zhao Y, Sangoram AM, Wilsbacher LD, Tanaka M, Antoch MP, Steeves TD, Vitaterna MH, Kornhauser JM, Lowrey PL, Turek FW, Takahashi JS (1997) Positional cloning of the mouse circadian clock gene. Cell 89:641–653 Lakin-Thomas P (1998) Choline depletion, frq mutations, and temperature compensation of the circadian rhythm in Neurospora crassa. J Biol Rhythm 13:268–277 Lakin-Thomas PL, Brody S (2000) Circadian rhythms in Neurospora crassa: lipid deficiencies restore robust rhythmicity to null frequency and white-collar mutants. Proc Natl Acad Sci U S A 97:256–261 Matsumoto A (2002) Temperature-dependent circadian locomotion rhythm in the per0 mutant of Drosophila melanogaster. J Jpn Soc Chronobiol 8:11–15. (in Japanese) Mergenhagen D, Schweiger HG (1975) The effect of different inhibitors of transcription and translation on the expression and control of circadian rhythm in individual cells of Acetabularia. Exp Cell Res 94:321–326 Merrow M, Brunner M, Roenneberg T (1999) Assignment of circadian function for the Neurospora clock gene frequency. Nature 399:584–586. https://doi.org/10.1038/21190 Mori T, Saveliev SV, Xu Y, Stafford WF, Cox MM, Inman RB, Johnson CH (2002) Circadian clock protein KaiC forms ATP-dependent hexameric rings and binds DNA. Proc Natl Acad Sci U S A 99:17203–17208 Nakahira Y, Katayama M, Miyashita H, Kutsuna S, Iwasaki H, Oyama T, Kondo T (2004) Global gene repression by KaiC as a master process of prokaryotic circadian system. Proc Natl Acad Sci U S A 101:881–885 Nakajima M, Imai K, Ito H, Nishiwaki T, Murayama Y, Iwasaki H, Oyama T, Kondo T (2005) Reconstitution of circadian oscillation of cyanobacterial KaiC phosphorylation in vitro. Science 308(5720):414–415 Nishiwaki T, Iwasaki H, Ishiura M, Kondo T (2000) Nucleotide binding and autophosphorylation of the clock protein KaiC as a circadian timing process of cyanobacteria. Proc Natl Acad Sci U S A 97:495–499 Pokhilko A, Fernández AP, Edwards KD, Sounthern MM, Halliday KJ, Millar AJ (2012) The clock gene circuit in Arabidopsis includes a repressilator with additional feedback loops. Mol Syst Biol 8:574 Sweeney BM, Haxo FT (1961) Persistence of a photosynthetic rhythm in enucleated Acetabularia. Science 134:1361 Sweeney BM, Tuffli CF, Rubin RH (1967) The circadian rhythm in photosynthesis in Acetabularia in the presence of actinomycin D, puromycin, and chloramphenicol. J Gen Physiol 50:647–659 Takano S, Tomita J, Sonoike K, Iwasaki H (2015) The initiation of nocturnal dormancy in Synechococcus as an active process. BMC Biol 13:36 Tomita J, Nakajima M, Kondo T, Iwasaki H (2005) No transcription-translation feedback in circadian rhythm of KaiC phosphorylation. Science 307(5707):251–254 Ukai-Tadenuma M, Yamada RG, Xu H, Ripperger JA, Liu AC, Ueda HR (2011) Delay in feedback repression by Cryptochrome 1 is required for circadian clock function. Cell 144:268–281 Xu Y, Mori T, Johnson CH (2000) Circadian clock-protein expression in cyanobacteria: rhythms and phase setting. EMBO J 19:3349–3357 Xu Y, Mori T, Johnson CH (2003) Cyanobacterial circadian clockwork: roles of KaiA, KaiB and the kaiBC promoter in regulating KaiC. EMBO J 22:2117–2126 Yang Z, Sehgal A (2001) Role of molecular oscillations in generating behavioral rhythms in Drosophila. Neuron 29:453–467

Mechanistic Aspects of the Cyanobacterial Circadian Clock Susan S. Golden and Andy LiWang

Abstract Circadian clocks are intracellular systems that provide an internal representation of time to regulate metabolism in preparation for day and night. A variety of mechanisms arose throughout the evolutionary tree as an adaptation to predictable daily swings in ambient light and temperature. In this chapter we will focus exclusively on the circadian clock shared by the cyanobacteria Synechococcus elongatus and Thermosynechococcus elongatus, with a noncomprehensive scope that emphasizes mechanism. As will hopefully become apparent here, the cyanobacterial system is valuable to the broader circadian clocks field because it yields conceptual insights into biological timekeeping at a level of detail higher and more comprehensive than those achieved so far in other systems. Major takeaways of this chapter include the following: (1) coordination of moving clock components is achieved through long-range allostery mediated by changes in structure and dynamics such that the clock runs clockwise and not counterclockwise, and (2) the fold-switching behavior of a clock protein underpins circadian rhythms of gene expression.

In 1993, it was demonstrated that the cyanobacterium Synechococcus elongatus possesses a bona fide circadian clock (Kondo et al. 1993); 5 years later these same labs identified the kaiA and kaiBC gene cluster as encoding central clock components (Ishiura et al. 1998), which initiated efforts to understand the functions of their

S. S. Golden Division of Biological Sciences, University of California, San Diego, La Jolla, CA, USA Center for Circadian Biology, University of California, San Diego, La Jolla, CA, USA A. LiWang (*) Center for Circadian Biology, University of California, San Diego, La Jolla, CA, USA Department of Chemistry and Biochemistry, University of California Merced, Merced, CA, USA Center for Cellular and Biomolecular Machines, University of California, Merced, CA, USA Health Sciences Research Institute, University of California, Merced, CA, USA e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_4

67

68

S. S. Golden and A. LiWang

protein products, KaiA, KaiB, and KaiC. A key advance in this regard was finding KaiC autophosphorylates and autodephosphorylates (Nishiwaki et al. 2000). KaiC is a member of the RecA superfamily (Leipe et al. 2000) and has two domains that share 42% sequence similarity, named CI and CII for the N- and C-terminal domains, respectively. Early on, it was found that CI, but not CII, enhances KaiA– KaiB interactions (Iwasaki et al. 1999), although the significance of this observation would not be appreciated until much later. Around this time, a sensor histidine kinase, SasA, whose loss dramatically reduces the amplitude of rhythmic gene expression in S. elongatus, was discovered to bind KaiC through a domain that shares 60% sequence similarity with KaiB (Iwasaki et al. 2000). Only many years later would it become clear how interactions between SasA and KaiB form part of the clock mechanism (Tseng et al. 2014; Chang et al. 2015; Heisler et al. 2020). By 2004, individual structures of KaiA, KaiB, and KaiC were solved using X-ray crystallography and NMR spectroscopy. These structures guided biochemistry experiments over the next several years as parallel efforts were underway to solve crystal structures of their complexes. KaiA was found to have two domains, an N-terminal pseudo-receiver domain and an all α-helical C-terminal domain that stimulates KaiC autophosphorylation (Taniguchi et al. 2001; Iwasaki et al. 2002; Williams et al. 2002; Vakonakis and LiWang 2004; Ye et al. 2004; Pattanayek and Egli 2015). KaiA crystallizes as a domain-swapped homodimer (Ye et al. 2004) in which the N-terminal domain of one subunit packs against the C-terminal domain of the other subunit. The two C-terminal domains of KaiA share a hydrophobic interface, which maintains the protein as a stable homodimer. However, the interface between the N- and C-terminal domains is mostly polar, and the domains when separated are well behaved by themselves at high concentrations. Experiments with the isolated domains suggested early on that conformations are likely accessible to KaiA in solution other than the one with the N- and C-terminal domains packed together. More than 10 years later, it became apparent that there are indeed such large-scale motions: the N- and C-terminal domains compete for binding an α-helix, α5, of the linker that connects the two domains, regulating the dynamic equilibrium between two conformations of KaiA that differ in their ability to stimulate KaiC phosphorylation (Tseng et al. 2017). KaiB crystallizes in a unique fold, either as a homodimer (Garces et al. 2004) or homotetramer (Iwase et al. 2005; Hitomi et al. 2005). These homotypic interactions were first detected by in vivo bioluminescence resonance energy transfer between luciferase-KaiB and GFP-KaiB fusion proteins expressed in Escherichia coli (Xu et al. 1999). Due to the sequence similarity between KaiB and the N-terminal domain of SasA, N-SasA, it was expected that they would adopt similar structures and compete for binding KaiC. However, N-SasA was found to adopt a thioredoxinlike fold and is monomeric even at high concentrations (Vakonakis et al. 2004), which was puzzling for over 10 years. KaiC crystallizes as a homohexamer with the N-terminal domains forming the so-called CI ring and the C-terminal domains forming the CII ring (Pattanayek et al. 2004). Leading up to these structures were reports that KaiC phosphorylates with a circadian rhythm in vivo (Iwasaki et al. 2002), KaiA stimulates KaiC autophosphorylation at seronyl and threonyl residues

Mechanistic Aspects of the Cyanobacterial Circadian Clock

69

(Iwasaki et al. 2002), specifically at S431 and T432 (Nishiwaki et al. 2004; Xu et al. 2004), and KaiB blocks KaiA activity (Xu et al. 2003; Kitayama et al. 2003; Williams et al. 2002). It was also shown that KaiA, KaiB, KaiC, and SasA rhythmically interact to form large transient complexes (Kageyama et al. 2003). Up to this point in time, it was assumed that the hearts of all circadian clocks, including that of Synechococcus elongatus, were transcription–translation feedback loops (TTFLs). Then, the lab of Takao Kondo published two transformative papers in the journal Science. The first showed that circadian rhythms of KaiC phosphorylation persist in vivo even when transcription and translation are halted (Tomita et al. 2005), demonstrating that a TTFL like the oscillators of eukaryotic clocks is not necessary to drive cyanobacterial circadian rhythms. Even more extraordinarily, the second showed that a simple mixture of recombinant forms of KaiA, KaiB, and KaiC generates autonomous KaiC phosphorylation rhythms in vitro that fulfill circadian criteria (Nakajima et al. 2005). Thus, in cyanobacteria the mechanism of the core circadian oscillator is entirely posttranslational. These papers set the stage for dissection of the cyanobacterial clock outside of the complex milieu of the cell, and many labs took to this task in earnest. For example, it was shown that the KaiABC in vitro oscillator tracks midday (Leypunskiy et al. 2017) and is reset by pulses of temperature (Yoshida et al. 2009), ADP (Rust et al. 2011), and oxidized quinones (Kim et al. 2012, 2020) like in cyanobacteria, demonstrating that the ability to entrain to the environment is a property inherent to the proteins themselves. A major leap in understanding the Kai phosphorylation cycle was that residues S431 and T432 in the CII ring of KaiC autophosphorylate and autodephosphorylate in a specific order (Nishiwaki et al. 2007; Rust et al. 2007) – S;T!S;pT!pS; pT!pS;T!S;T!. . . – although the mechanistic basis of this “clockwise” order was unclear (Fig. 1). It was later found that phosphorylation of T432 loosens the CII ring, whereas phosphorylation of S431 tightens it (Chang et al. 2011). Thus, the CII ring loosens and tightens every day as follows: loose (S;T)!looser (S;pT)!tight (pS;pT)!tighter (pS;T)!loose (S;T)!. . . and that these ring dynamics regulate accessibility of the KaiA-binding site on KaiC (Tseng et al. 2014). When the CII ring is loose, it undergoes “breathing” motions. The integrity of the hexameric state of KaiC when the CII ring is loose is maintained by a CI ring that is tightly bound to ATP (Hayashi et al. 2004). CI is a slow ATPase (Murakami et al. 2008), and thus it makes occasional and transient excursions to a looser ADP-bound post-ATP hydrolysis state. KaiA stimulates the phosphorylation phase by binding to the extreme C-terminal peptide on the CII side of KaiC, which includes a 10-residue segment called the A-loop (Kim et al. 2008). The A-loop is in a dynamic equilibrium between buried and solvent-exposed conformations, the latter of which is captured by KaiA (Tseng et al. 2014); this KaiA-bound A loop-out state catalyzes exchange of ADP for ATP in CII (Nishiwaki-Ohkawa et al. 2014). When the CII ring is loose and breathing, A-loops more easily transition to the solvent-exposed state (Tseng et al. 2014). In contrast, these loops are more buried in a tight CII ring and thus less accessible to KaiA. Indeed, mutations in the A-loop region of KaiC that stabilize the solvent-exposed state of the A-loop promote KaiC autophosphorylation even in the absence of KaiA (Kim et al. 2008). Corroboration is provided by crystallography,

70

S. S. Golden and A. LiWang

Fig. 1 Model of the cyanobacterial clock. For simplicity, only three and four subunit domains of CII and CI are drawn. Small gray disks in CII represent phosphoryl groups at S431 and/or T432. This model explains why the clock can only run in the clockwise direction. The CI ring has pre- and post-ATP hydrolysis states, the latter of which stacks with the CII ring whenever it is tightened by S431 phosphorylation. The two rings are allosterically coupled, allowing them to coordinate daytime processes on CII and nighttime interactions on CI. The fold-switching behavior of KaiB temporally separates SasA-mediated activation and CikA-mediated deactivation of the master transcription factor RpaA, thereby underpinning circadian rhythms of RpaA-mediated gene expression

high-speed atomic force microscopy experiments on KaiA-KaiC complexes, NMR, and other studies (Ma and Ranganathan 2012; Mori et al. 2018; Pattanayek and Egli 2015; Vakonakis and LiWang 2004). The phosphorylation states of S431 and T432 have opposing effects on KaiC dynamics, structure, and function, creating an ultrasensitive switch in KaiC that enhances the robustness of the timing mechanism (Lin et al. 2014). Recent molecular dynamics simulations of KaiC and Bayesian modeling provided detailed insights into the mechanism of nucleotide exchange

Mechanistic Aspects of the Cyanobacterial Circadian Clock

71

(Hong et al. 2018) and KaiA-mediated KaiC phosphorylation under different phosphorylation and nucleotide-bound states (Hong et al. 2020). KaiB binds to KaiC whenever residue S431 in the CII domain is phosphorylated (Nishiwaki et al. 2007). As such, reports that KaiB binds to CII seemed reasonable (Pattanayek et al. 2008, 2013; Villarreal et al. 2013; Snijder et al. 2014). Thus, a report that KaiB binds to the CI domain with a 1:1 stoichiometry and sequesters a dimeric KaiA there, breaking KaiA’s twofold symmetry (Chang et al. 2012), surprised many in spite of yeast 2-hybrid data 13 years earlier that implicated CI in KaiB binding (Iwasaki et al. 1999). The new data raised the question of how information on the phosphorylation status of S431 in CII is transmitted tens of angstroms to the CI ring of KaiC. A key observation was that the rings stack together whenever S431 is phosphorylated (Chang et al. 2012). Another was that KaiB does not bind to isolated CI rings that are associated solely with ATP, but does bind when some ADP (post-ATP hydrolysis) is present in the CI ATPase to loosen the ring (Chang et al. 2012). Thus, it was proposed that ring stacking occurs when a CII ring tightened by phosphorylation at S431 stacks and thereby stabilizes a somewhat open CI ring in the post-ATP hydrolysis state with KaiB-binding sites exposed (Chang et al. 2012). In this case, the hexameric integrity of KaiC when the CI ring is unstable is maintained by the tight CII ring. Corroboration is provided by work showing that ATP hydrolysis by the CI ring is necessary for KaiB–KaiC binding (Mutoh et al. 2013; Phong et al. 2013). The slow ATPase allows CI to make infrequent transitions between pre- and post-ATP hydrolysis states, and, thus, sample whether the CII ring is receptive for stacking. This process is likely why altering ATPase activities of the CI ring can affect clock period (Abe et al. 2015; Ito-Miwa et al. 2020). A conceptual movie of KaiC emerges in which the two rings communicate, oscillating out of phase to one another, such that KaiC maintains its integrity as a homohexamer and the CII and CI rings alternate as sites of protein–protein interactions during the day and night, respectively. An orthogonal demonstration of the coupling between the CI and CII rings was by using EPR to observe the effect of spiking a pre-equilibrated sample of KaiB–KaiC complexes (where KaiC was a phosphomimetic of pS;pT-KaiC) with a substoichiometric amount of KaiA, which caused some KaiB to transiently dissociate from KaiC (Chow et al. 2020). This KaiB-unbinding phase was most likely due to sudden KaiA–A loop interactions that induced CII ring loosening and thus transient CI–CII ring unstacking before KaiA could be sequestered in KaiABC complexes. SasA also binds to CI (Chang et al. 2011) and competes with KaiB for a stretch of residues called the B-loop (Tseng et al. 2014). The basis of this competition lies in one of the more remarkable features of the cyanobacterial circadian oscillator: a fold switch in KaiB (Chang et al. 2015). Although many proteins are known to undergo large conformational rearrangements, their underlying secondary structures and overall folds are largely preserved. A handful of proteins, however, are known to exist in equilibrium between two distinctly different folds, and they are classified as metamorphic (Murzin 2008). The unique ground-state fold into which KaiB crystallizes was assumed to represent the active and only state of KaiB. Attempts to crystallize complexes of KaiB with KaiC were unsuccessful, however, because the complexes

72

S. S. Golden and A. LiWang

would dissociate during crystallization. Intriguingly, the kinetics of binding KaiB to KaiC in vitro is very slow, even when S431 is replaced with an aspartyl or glutamyl residue to mimic a constitutively phosphorylated state of KaiC that is receptive for KaiB binding. A comparison of NMR spectra of isotopically enriched KaiB alone and bound to CI revealed that the form of KaiB that binds KaiC undergoes a global change in structure (Chang et al. 2012), adopting a thioredoxin-like fold in the complex (Chang et al. 2015). Thus, KaiB is a metamorphic protein with an inactive oligomeric state and a distinctly different and active monomeric fold-switched state. The reason this fold-switched state went undetected for many years is because it is much less stable than the “ground-state” fold found in isolated samples of KaiB. This thermodynamic difference is probably also why KaiB–KaiC complexes were not sufficiently stable for crystallization. Mutations that lock KaiB in its active foldswitched state produced instantaneous KaiB–KaiC binding (Chang et al. 2015) and very stable complexes, including complexes with KaiA and the pseudo-receiver domain of a sensor histidine kinase, CikA (Tseng et al. 2017). Crystal structures of KaiB–KaiC and KaiA–KaiB–CI complexes (Tseng et al. 2017) showed that one subunit of homodimeric KaiA increases the interface between KaiB and CI by interacting with both KaiB and CI, and thereby explaining KaiA–KaiB–CI binding cooperativity and the nature of asymmetric KaiA–KaiB binding observed earlier (Tseng et al. 2014). Cryo-EM structures of KaiA–KaiB–KaiC complexes were consistent with the crystal structures (Snijder et al. 2017). These structures also corroborate an earlier finding that KaiA and KaiB do not interact with one another when mixed unless CI is also present (Iwasaki et al. 1999). An unexpected observation was that when KaiB sequesters KaiA on CI, KaiA adopts an autoinhibited conformation that is dramatically different than when it binds to the A-loops of KaiC (Tseng et al. 2017). The KaiB–KaiC crystal structure also revealed conformational changes in the post-ATP hydrolysis state of the CI ring that exposes the KaiBbinding site, corroborating earlier studies (Chang et al. 2012; Abe et al. 2015). KaiC ring dynamics and stacking provide an explanation for why the clockwise direction of phosphorylation is allowed and the counterclockwise direction is prevented even in the presence of stochastic thermal fluctuations (Chang et al. 2011). Autophosphorylation in the counterclockwise direction, S;T!pS;T!pS; pT, is prohibited as follows: The first reaction, S;T!pS;T, maximally tightens the CII ring and A-loops become highly inaccessible to KaiA (Tseng et al. 2014). Additionally, the CII and CI rings stack, allowing KaiB to sequester the autoinhibited state of KaiA on the CI side, which initiates autodephosphorylation and return to the S;T state (Chang et al. 2012). In contrast, in the first clockwise step, S;T!S;pT, which occurs during subjective morning, the CII ring increases its breathing motions and thereby promotes further A-loop accessibility to KaiA, providing positive feedback for this step. In the second clockwise step, S;pT!pS;pT, the CII ring tightens but modestly as the tightening effect of pS431 is offset somewhat by the loosening effect of pT432 (Chang et al. 2011). Thus, although the A-loops become moderately less accessible, KaiA is still able to stimulate near complete autophosphorylation of KaiC, whereupon the CI ring uses its ATPase activity to stack onto the tight CII ring. With the B-loops now exposed,

Mechanistic Aspects of the Cyanobacterial Circadian Clock

73

when KaiB stochastically transitions to its fold-switched state it binds to CI and sequesters KaiA there in its autoinhibited conformation (Tseng et al. 2017, 2014). It is now subjective night and KaiC commences to autodephosphorylate until subjective dawn. As with counterclockwise autophosphorylation, counterclockwise autodephosphorylation is also prohibited, but as follows: In the first counterclockwise reaction, pS;pT!S;pT, the rings would unstack, prematurely dissolving the KaiABC nighttime complex and liberating KaiA to stimulate autophosphorylation, which would reestablish the doubly phosphorylated state. In the clockwise direction, pS;pT!pS;T, the KaiC rings remain tightly stacked together and thus integrity of the KaiABC complex is maintained throughout the rest of the night. In the final clockwise step of the cycle, pS;T!S;T, the CII ring returns to its loose state, the rings unstack, and KaiB releases KaiA so that KaiA can stimulate a new cycle of KaiC autophosphorylation. Thus, CII ring dynamics dictate that the overall direction of the reactions is clockwise rather than counterclockwise (Chang et al. 2011, 2012). CII ring dynamics also provides an explanation for maintaining ensemble-wide phase coherence. During the phosphorylation phase, there is a distribution of KaiC phosphostates where the more slowly phosphorylating KaiC particles, the laggards, have more accessible A-loops (Tseng et al. 2014) and thus receive more stimulation from (substoichiometric) KaiA than the more rapidly phosphorylating KaiC particles with tighter CII rings. In this way, the laggards speed up, the front runners slow down, and ensemble-wide phase coherence is maintained, which was first predicted by a mathematical model (van Zon et al. 2007). Signals are transduced from the core oscillator via interactions with histidine kinases, SasA and CikA, which physically engage with the Kai complex. Roles have been recognized for several years in which SasA activates the master transcription factor, RpaA, through phosphoryl transfer and later CikA dephosphorylates RpaA to terminate promoter activation (Gutu and O’Shea 2013). This delay between RpaA activation and deactivation generates circadian rhythms of gene expression, but the origin of this delay was unclear until recently. Autophosphorylation of SasA occurs only when SasA is activated by KaiC (Smith and Williams 2006). This activation is maximal whenever S431 is phosphorylated (after subjective dusk) and the B-loop is exposed on CI. SasA binds to the B-loop like KaiB does, but does so much faster because the SasA pool is already fully in the active thioredoxin-like fold (Chang et al. 2012; Tseng et al. 2014). Thus, whenever the B-loop is presented, SasA is available to engage before it is displaced from KaiC by the cooperative binding of six KaiB proteins to the CI ring (Tseng et al. 2014, 2017; Chang et al. 2012). The cooperativity of KaiB–KaiC binding was first demonstrated by native mass spectrometry (Snijder et al. 2014) and then corroborated structurally (Tseng et al. 2017). In contrast to SasA, CikA binds to KaiB, which stimulates CikA phosphatase activity, but only after fold-switched KaiB is bound to KaiC timed later in the circadian cycle (Tseng et al. 2017; Chavan et al. 2020). These phase differences have been directly observed in vitro (Chavan et al. 2020). Thus, fold-switching by KaiB creates the delay between SasA activation and CikA deactivation of RpaAmediated gene expression. It is also interesting to note that CikA and KaiA have overlapping binding sites on fold-switched KaiB (Tseng et al. 2017), similar to how

74

S. S. Golden and A. LiWang

SasA and KaiB have overlapping binding sites on KaiC. This intimate coupling between oscillator and input/output components blurs their distinction. Indeed, the cycle generated by KaiA, KaiB, and KaiC alone in vitro tolerates only narrow concentration ranges and ratios, but the presence of SasA and CikA increases this tolerance significantly; SasA and CikA can separately rescue KaiABC-only reactions that fail from limiting KaiB or KaiA, respectively (Heisler et al. 2020). Moreover, inclusion of these histidine kinases tunes the period of the oscillation (Kaur et al. 2019; Chang et al. 2015; Heisler et al. 2020). Together, these results expand the concept of the oscillator to include these components that have heretofore been considered auxiliary. As this chapter comes to a close, we should consider the future role of the cyanobacterial circadian clock in the context of benefitting the broader clocks field. An empirically based model of the mechanism of temperature compensation, a criterion of all bona fide circadian clocks, has not yet been fully developed, but the cyanobacterial system, because of its structural, biochemical, and biophysical tractability, offers a tremendous opportunity to fill this long-standing knowledge gap. The development of a real-time fluorescence-based reporting system that increases the throughput and precision, and reduces the labor, of oscillator monitoring (Heisler et al. 2019) opens a door to build greater complexity into the in vitro system, which should be of special interest to systems and synthetic biologists. For example, now it is possible to reconstitute circadian DNA binding by RpaA in vitro (Chavan et al. 2020), setting a new standard for how the cyanobacterial circadian clock should be studied in vitro and making it possible to ask questions about how changing environments, such as temperature, metabolites, and protein levels, are reflected in the core oscillator and propagated to regulation of transcription, providing deeper mechanistic understanding into clock biology. Acknowledgments This work was supported by US National Institutes of Health grants R35GM118290 (to S.S.G.) and R01GM107521 (to A.L.), the Department of the Army Research Office grant W911NF-17-1-0434 (to A.L.), and NSF-CREST: Center for Cellular and Biomolecular Machines at the University of California, Merced (NSF-HRD-1547848).

Bibliography Abe J, Hiyama TB, Mukaiyama A et al (2015) Circadian rhythms. Atomic-scale origins of slowness in the cyanobacterial circadian clock. Science 349(6245):312–316 Chang Y-G, Kuo N-W, Tseng R, LiWang A (2011) Flexibility of the C-terminal, or CII, ring of KaiC governs the rhythm of the circadian clock of cyanobacteria. Proc Natl Acad Sci U S A 108 (35):14431–14436 Chang Y-G, Tseng R, Kuo N-W, LiWang A (2012) Rhythmic ring-ring stacking drives the circadian oscillator clockwise. Proc Natl Acad Sci U S A 109(42):16847–16851 Chang Y-G, Cohen SE, Phong C et al (2015) Circadian rhythms. A protein fold switch joins the circadian oscillator to clock output in cyanobacteria. Science 349(6245):324–328 Chavan AG, Ernst DC, Fang M et al (2020) Reconstitution of an intact clock that generates circadian DNA binding in vitro. BioRxiv

Mechanistic Aspects of the Cyanobacterial Circadian Clock

75

Chow GK, Chavan AG, Heisler JC, Chang Y-G, LiWang A, Britt RD (2020) Monitoring proteinprotein interactions in the cyanobacterial circadian clock in real time via electron paramagnetic resonance spectroscopy. Biochemistry 59(26):2387–2400 Garces RG, Wu N, Gillon W, Pai EF (2004) Anabaena circadian clock proteins KaiA and KaiB reveal a potential common binding site to their partner KaiC. EMBO J 23(8):1688–1698 Gutu A, O’Shea EK (2013) Two antagonistic clock-regulated histidine kinases time the activation of circadian gene expression. Mol Cell 50(2):288–294 Hayashi F, Itoh N, Uzumaki T et al (2004) Roles of two ATPase-motif-containing domains in cyanobacterial circadian clock protein KaiC. J Biol Chem 279(50):52331–52337 Heisler J, Chavan A, Chang Y-G, LiWang A (2019) Real-time in vitro fluorescence anisotropy of the cyanobacterial circadian clock. Method Protoc 2(2):42 Heisler J, Swan JA, Palacios JG et al (2020) Structural mimicry confers robustness in the cyanobacterial circadian clock. BioRxiv Hitomi K, Oyama T, Han S, Arvai AS, Getzoff ED (2005) Tetrameric architecture of the circadian clock protein KaiB. A novel interface for intermolecular interactions and its impact on the circadian rhythm. J Biol Chem 280(19):19127–19135 Hong L, Vani BP, Thiede EH, Rust MJ, Dinner AR (2018) Molecular dynamics simulations of nucleotide release from the circadian clock protein KaiC reveal atomic-resolution functional insights. Proc Natl Acad Sci U S A 115(49):E11475–E11484 Hong L, Lavrentovich DO, Chavan A et al (2020) Bayesian modeling reveals metabolite-dependent ultrasensitivity in the cyanobacterial circadian clock. Mol Syst Biol 16(6):e9355 Ishiura M, Kutsuna S, Aoki S et al (1998) Expression of a gene cluster kaiABC as a circadian feedback process in cyanobacteria. Science 281(5382):1519–1523 Ito-Miwa K, Furuike Y, Akiyama S, Kondo T (2020) Tuning the circadian period of cyanobacteria up to 6.6 days by the single amino acid substitutions in KaiC. Proc Natl Acad Sci U S A 117 (34):20926–20931 Iwasaki H, Taniguchi Y, Ishiura M, Kondo T (1999) Physical interactions among circadian clock proteins KaiA, KaiB and KaiC in cyanobacteria. EMBO J 18(5):1137–1145 Iwasaki H, Williams SB, Kitayama Y, Ishiura M, Golden SS, Kondo T (2000) A kaiC-interacting sensory histidine kinase, SasA, necessary to sustain robust circadian oscillation in cyanobacteria. Cell 101(2):223–233 Iwasaki H, Nishiwaki T, Kitayama Y, Nakajima M, Kondo T (2002) KaiA-stimulated KaiC phosphorylation in circadian timing loops in cyanobacteria. Proc Natl Acad Sci U S A 99 (24):15788–15793 Iwase R, Imada K, Hayashi F et al (2005) Functionally important substructures of circadian clock protein KaiB in a unique tetramer complex. J Biol Chem 280(52):43141–43149 Kageyama H, Kondo T, Iwasaki H (2003) Circadian formation of clock protein complexes by KaiA, KaiB, KaiC, and SasA in cyanobacteria. J Biol Chem 278(4):2388–2395 Kaur M, Ng A, Kim P, Diekman C, Kim Y-I (2019) Cika modulates the effect of kaia on the period of the circadian oscillation in kaic phosphorylation. J Biol Rhythm 34(2):218–223 Kim Y-I, Dong G, Carruthers CW, Golden SS, LiWang A (2008) The day/night switch in KaiC, a central oscillator component of the circadian clock of cyanobacteria. Proc Natl Acad Sci U S A 105(35):12825–12830 Kim Y-I, Vinyard DJ, Ananyev GM, Dismukes GC, Golden SS (2012) Oxidized quinones signal onset of darkness directly to the cyanobacterial circadian oscillator. Proc Natl Acad Sci U S A 109(44):17765–17769 Kim P, Porr B, Mori T et al (2020) Cika, an input pathway component, senses the oxidized quinone signal to generate phase delays in the cyanobacterial circadian clock. J Biol Rhythm 35 (3):227–234 Kitayama Y, Iwasaki H, Nishiwaki T, Kondo T (2003) KaiB functions as an attenuator of KaiC phosphorylation in the cyanobacterial circadian clock system. EMBO J 22(9):2127–2134

76

S. S. Golden and A. LiWang

Kondo T, Strayer CA, Kulkarni RD et al (1993) Circadian rhythms in prokaryotes: luciferase as a reporter of circadian gene expression in cyanobacteria. Proc Natl Acad Sci U S A 90 (12):5672–5676 Leipe DD, Aravind L, Grishin NV, Koonin EV (2000) The bacterial replicative helicase DnaB evolved from a RecA duplication. Genome Res 10(1):5–16 Leypunskiy E, Lin J, Yoo H, Lee U, Dinner AR, Rust MJ (2017) The cyanobacterial circadian clock follows midday in vivo and in vitro. eLife 6:e23539 Lin J, Chew J, Chockanathan U, Rust MJ (2014) Mixtures of opposing phosphorylations within hexamers precisely time feedback in the cyanobacterial circadian clock. Proc Natl Acad Sci U S A 111(37):E3937–E3945 Ma L, Ranganathan R (2012) Quantifying the rhythm of KaiB-C interaction for in vitro cyanobacterial circadian clock. PLoS One 7(8):e42581 Mori T, Sugiyama S, Byrne M, Johnson CH, Uchihashi T, Ando T (2018) Revealing circadian mechanisms of integration and resilience by visualizing clock proteins working in real time. Nat Commun 9(1):3245 Murakami R, Miyake A, Iwase R, Hayashi F, Uzumaki T, Ishiura M (2008) ATPase activity and its temperature compensation of the cyanobacterial clock protein KaiC. Genes Cells 13(4):387–395 Murzin AG (2008) Biochemistry. Metamorphic proteins. Science 320(5884):1725–1726 Mutoh R, Nishimura A, Yasui S, Onai K, Ishiura M (2013) The ATP-mediated regulation of KaiBKaiC interaction in the cyanobacterial circadian clock. PLoS One 8(11):e80200 Nakajima M, Imai K, Ito H et al (2005) Reconstitution of circadian oscillation of cyanobacterial KaiC phosphorylation in vitro. Science 308(5720):414–415 Nishiwaki T, Iwasaki H, Ishiura M, Kondo T (2000) Nucleotide binding and autophosphorylation of the clock protein KaiC as a circadian timing process of cyanobacteria. Proc Natl Acad Sci U S A 97(1):495–499 Nishiwaki T, Satomi Y, Nakajima M et al (2004) Role of KaiC phosphorylation in the circadian clock system of Synechococcus elongatus PCC 7942. Proc Natl Acad Sci U S A 101 (38):13927–13932 Nishiwaki T, Satomi Y, Kitayama Y et al (2007) A sequential program of dual phosphorylation of KaiC as a basis for circadian rhythm in cyanobacteria. EMBO J 26(17):4029–4037 Nishiwaki-Ohkawa T, Kitayama Y, Ochiai E, Kondo T (2014) Exchange of ADP with ATP in the CII ATPase domain promotes autophosphorylation of cyanobacterial clock protein KaiC. Proc Natl Acad Sci U S A 111(12):4455–4460 Pattanayek R, Egli M (2015) Protein-protein interactions in the cyanobacterial circadian clock: structure of KaiA dimer in complex with C-terminal KaiC peptides at 2.8 Å resolution. Biochemistry 54(30):4575–4578 Pattanayek R, Wang J, Mori T, Xu Y, Johnson CH, Egli M (2004) Visualizing a circadian clock protein: crystal structure of KaiC and functional insights. Mol Cell 15(3):375–388 Pattanayek R, Williams DR, Pattanayek S et al (2008) Structural model of the circadian clock KaiBKaiC complex and mechanism for modulation of KaiC phosphorylation. EMBO J 27 (12):1767–1778 Pattanayek R, Yadagiri KK, Ohi MD, Egli M (2013) Nature of KaiB-KaiC binding in the cyanobacterial circadian oscillator. Cell Cycle 12(5):810–817 Phong C, Markson JS, Wilhoite CM, Rust MJ (2013) Robust and tunable circadian rhythms from differentially sensitive catalytic domains. Proc Natl Acad Sci U S A 110(3):1124–1129 Rust MJ, Markson JS, Lane WS, Fisher DS, O’Shea EK (2007) Ordered phosphorylation governs oscillation of a three-protein circadian clock. Science 318(5851):809–812 Rust MJ, Golden SS, O’Shea EK (2011) Light-driven changes in energy metabolism directly entrain the cyanobacterial circadian oscillator. Science 331(6014):220–223 Smith RM, Williams SB (2006) Circadian rhythms in gene transcription imparted by chromosome compaction in the cyanobacterium Synechococcus elongatus. Proc Natl Acad Sci U S A 103 (22):8564–8569

Mechanistic Aspects of the Cyanobacterial Circadian Clock

77

Snijder J, Burnley RJ, Wiegard A et al (2014) Insight into cyanobacterial circadian timing from structural details of the KaiB-KaiC interaction. Proc Natl Acad Sci U S A 111(4):1379–1384 Snijder J, Schuller JM, Wiegard A et al (2017) Structures of the cyanobacterial circadian oscillator frozen in a fully assembled state. Science 355(6330):1181–1184 Taniguchi Y, Yamaguchi A, Hijikata A et al (2001) Two KaiA-binding domains of cyanobacterial circadian clock protein KaiC. FEBS Lett 496(2–3):86–90 Tomita J, Nakajima M, Kondo T, Iwasaki H (2005) No transcription-translation feedback in circadian rhythm of KaiC phosphorylation. Science 307(5707):251–254 Tseng R, Chang Y-G, Bravo I et al (2014) Cooperative KaiA-KaiB-KaiC interactions affect KaiB/ SasA competition in the circadian clock of cyanobacteria. J Mol Biol 426(2):389–402 Tseng R, Goularte NF, Chavan A et al (2017) Structural basis of the day-night transition in a bacterial circadian clock. Science 355(6330):1174–1180 Vakonakis I, LiWang AC (2004) Structure of the C-terminal domain of the clock protein KaiA in complex with a KaiC-derived peptide: implications for KaiC regulation. Proc Natl Acad Sci U S A 101(30):10925–10930 Vakonakis I, Klewer DA, Williams SB, Golden SS, LiWang AC (2004) Structure of the N-terminal domain of the circadian clock-associated histidine kinase SasA. J Mol Biol 342(1):9–17 van Zon JS, Lubensky DK, Altena PRH, ten Wolde PR (2007) An allosteric model of circadian KaiC phosphorylation. Proc Natl Acad Sci U S A 104(18):7420–7425 Villarreal SA, Pattanayek R, Williams DR et al (2013) CryoEM and molecular dynamics of the circadian KaiB-KaiC complex indicates that KaiB monomers interact with KaiC and block ATP binding clefts. J Mol Biol 425(18):3311–3324 Williams SB, Vakonakis I, Golden SS, LiWang AC (2002) Structure and function from the circadian clock protein KaiA of Synechococcus elongatus: a potential clock input mechanism. Proc Natl Acad Sci U S A 99(24):15357–15362 Xu Y, Piston DW, Johnson CH (1999) A bioluminescence resonance energy transfer (BRET) system: application to interacting circadian clock proteins. Proc Natl Acad Sci U S A 96 (1):151–156 Xu Y, Mori T, Johnson CH (2003) Cyanobacterial circadian clockwork: roles of KaiA, KaiB and the kaiBC promoter in regulating KaiC. EMBO J 22(9):2117–2126 Xu Y, Mori T, Pattanayek R, Pattanayek S, Egli M, Johnson CH (2004) Identification of key phosphorylation sites in the circadian clock protein KaiC by crystallographic and mutagenetic analyses. Proc Natl Acad Sci U S A 101(38):13933–13938 Ye S, Vakonakis I, Ioerger TR, LiWang AC, Sacchettini JC (2004) Crystal structure of circadian clock protein KaiA from Synechococcus elongatus. J Biol Chem 279(19):20511–20518 Yoshida T, Murayama Y, Ito H, Kageyama H, Kondo T (2009) Nonparametric entrainment of the in vitro circadian phosphorylation rhythm of cyanobacterial KaiC by temperature cycle. Proc Natl Acad Sci U S A 106(5):1648–1653

Mechanism of the Cyanobacterial Circadian Clock Protein KaiC to Measure 24 Hours Kumiko Ito-Miwa, Kazuki Terauchi, and Takao Kondo

Abstract By only mixing the cyanobacterial (Synechococcus elongatus PCC 7942) circadian clock proteins KaiC, KaiA, and KaiB with ATP in a test tube, the phosphorylation status of KaiC shows a circadian rhythm. The major component KaiC is a typical adenosine triphosphatase (ATPase). Although ATPase activity of KaiC is extremely low, the ATPase activity of KaiC by itself shows temperature compensation and correlates with the rate of circadian rhythm (oscillation frequency), indicating that it defines the temperature compensation and period of the circadian rhythm as a pacemaker. The Kai clock system based on this activity shows the existence of a completely novel circadian clock mechanism that does not require transcription/translation or molecular interactions, etc., which has been assumed so far for prokaryotic and eukaryotic organisms. In this chapter, we describe a model for the cyanobacterial circadian clock system that we have derived from our analyses of KaiC ATPase activity.

1 Introduction Having a circadian clock, many organisms are able to adapt to the day/night environment on Earth. A circadian clock is advantageous for life on Earth and is widely developed as a basic system for life in the evolutionary process by fulfilling the three characteristics: an approximate 24-h period, temperature compensation of the period, and entrainment to external time cues derived from day/night alterations (Dunlap et al. 2004). Therefore, to understand the nature of circadian clocks, we must explain not only the stable oscillation scenario, but also the mechanistic basis

K. Ito-Miwa (*) · T. Kondo (*) Division of Biological Science, Graduate School of Science and Institute for Advanced Studies, Nagoya University, Nagoya, Japan e-mail: [email protected]; [email protected] K. Terauchi (*) College of Life Sciences, Ritsumeikan University, Kusatsu, Shiga, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_5

79

80

K. Ito-Miwa et al.

supporting these three characteristics. Cyanobacteria acquired the capacity for photosynthesis with water photolysis and oxygen evolution during the early course of evolution and further evolved into plant chloroplasts. Cyanobacteria and plants also exhibit distinct circadian rhythms and are important models for elucidation of circadian clocks. Here, we explain the study of the cyanobacterial clock system performed in our laboratory, particularly the mechanisms of period determination and temperature compensation, and propose a new model of a circadian oscillator composed of only three proteins.

2 Clock Systems of Cyanobacteria The cyanobacterial circadian clock genes kaiA, kaiB, kaiC (kai means cycle in Japanese) were found by molecular genetic analysis using bioluminescence around 1996 (Ishiura et al. 1998). Because kaiC expression is self-regulated, it was initially thought that transcriptional and translational feedback models applied to the clock system of cyanobacteria as in eukaryotes. However, in 2005 it was revealed that the phosphorylation status of KaiC generates a stable circadian rhythm (KaiC phosphorylation rhythm), by only mixing three purified Kai-proteins with ATP in vitro (Fig. 1a) (Nakajima et al. 2005). This KaiC phosphorylation rhythm exhibits temperature compensation over a wide range of temperatures and can be entrained to temperature cycles, fulfilling three prominent characteristics of circadian clocks (Nakajima et al. 2005; Yoshida et al. 2009; Murayama et al. 2011). We believe that this simple clock mechanism composed of only three proteins, which exhibits properties comparable to the circadian clock in higher organisms, is the most suitable system for elucidating the enigma of the circadian clock, and we have advanced our analysis of its mechanisms.

3 Characteristics of the Circadian Clock in Terms of Temperature Compensation of Period The rate of most biochemical reactions in living organisms varies in a temperaturedependent manner, with the rate being 2–3 times higher when the temperature is increased by 10
C (the temperature coefficient Q10 ¼ 2–3, Q10 represents how many times the reaction rate increases when the reaction temperature increases by 10
C) (Nobel 1991). A circadian clock maintains a constant period by a special mechanism called temperature compensation, even if there is a difference in external temperature, e.g., the seasonal changes. Temperature compensation is also preserved even in KaiC period mutants that exhibit a wide range of period (Ito-Miwa et al. 2020). The fact that temperature compensation is universally preserved even when the period is changed by mutation suggests that the temperature compensation of period is a more fundamental property than the mechanism for the period determination.

Mechanism of the Cyanobacterial Circadian Clock Protein KaiC to Measure. . .

81

Fig. 1 Characteristics of circadian oscillations in cyanobacteria. (a) Reconstruction of the circadian clock in vitro. Recombinant KaiC, KaiA, and KaiB proteins, in the presence of ATP, were mixed in a test tube and incubated at 30
C. Aliquots of reaction mixtures were collected every 4 h, and the phosphorylation status of KaiC was analyzed by SDS-PAGE. The upper and lower bands correspond to phosphorylated (P-KaiC) and non-phosphorylated KaiC (NP-KaiC), respectively. The phosphorylation status of KaiC oscillated with a period of approximately 24 h. (b) Harmonic oscillation and relaxation oscillation

Erwin Bünning, a pioneer in the study of the circadian clock, examined the characteristics of circadian rhythms, and suggested that a circadian rhythm has two characteristics, that of a harmonic oscillation and also as a relaxation oscillation (Fig. 1b) (Bünning 1967). The harmonic oscillation, represented by the oscillation of an object connected to a spring, is a physical oscillation that obeys Hooke’s law, and does not consume energy to sustain the oscillations. The amplitude is easily changed by the energy given to the oscillatory system, but the period is determined by a constant of Hooke’s law and is independent of amplitude (“isochronism” of harmonic oscillation). On the other hand, relaxation oscillations, if assumed to occur in the cell, are explained by a loop of programmed biochemical reactions (sequential switching occurs due to the accumulation of the reaction and proceeds to the next step). Thus, the biochemical reactions are rate-limiting for the rate of the relaxation oscillator, and the period can be easily altered by temperature, but the amplitude is determined by the threshold for switching and is invariant. The temperature compensation of the period suggests that the basic mechanism of circadian rhythmicity has the property of harmonic oscillators, while the persistence of the rhythm is easy to explain as a relaxation oscillation. Indeed, the harmonic nature of the KaiC phosphorylation rhythm has been suggested by the decrease in amplitude of the

82

K. Ito-Miwa et al.

rhythm associated with the decrease in the reaction temperature, and by the disappearance of oscillations caused by even lower temperature (Murayama et al. 2017). The isochronism of the harmonic oscillation (as in the “isochronism of the pendulum”) discovered by Galileo was incorporated into the pacemakers of mechanical clocks by Huygens and Hooke, and its accuracy was dramatically improved. The period of the circadian clock is also accurate and it is known that some organisms maintain their rhythms with fluctuations of period ranging from only a few minutes to about 10 min per day (Dunlap et al. 2004; Njus et al. 1981). Our analysis also confirmed that the period of the KaiC phosphorylation rhythm shows a similar accuracy. The accuracy of circadian clocks is important to accommodate the extremely precise period of the Earth’s rotation, and the harmonic nature of circadian clocks is advantageous to achieve this accuracy.

4 ATPase Activity and Intramolecular Feedback of KaiC How can the circadian rhythmicity created by Kai proteins exhibit a periodicity that is unaffected by temperature? The answer seems to be inherent in the activity of KaiC to hydrolyze ATP (ATPase activity) (Terauchi et al. 2007; Abe et al. 2015). KaiC is a protein consisting of 519 amino acids, and whose amino acid sequence is similar to that of ATPases such as RecA (Leipe et al. 2000). However, the ATPase activity of KaiC is nearly zero; a single molecule of KaiC hydrolyzes only 10–15 ATPs per day (less than 10,000 times the activity of many enzymes) (Terauchi et al. 2007). KaiC has two ATPase domains, termed CI on the N-terminal side and CII on the C-terminal side (Iwasaki et al. 1999). CI and CII exhibit significant similarities to each other, and are linked in tandem via a hinge sequence. The ATPase activity of KaiC was reported considerably later than the autophosphorylation activity of KaiC (Nishiwaki et al. 2000; Kitayama et al. 2003; Xu et al. 2003; Terauchi et al. 2007). KaiC exhibits a very low but stable ATPase activity (15 ATP per KaiC monomer per day) (Terauchi et al. 2007). The exact contributions of the CI versus CII domains to the total ATPase activity is unknown, but it is thought that the contribution of CI is higher than that of CII, as the CI domain of KaiC without the CII domain retains approximately 70% of the activity of full-length KaiC (Terauchi et al. 2007; Abe et al. 2015). On the other hand, the phosphorylation status of the CII domain also affects the total ATPase activity. The KaiC-AA mutant protein that mimics a dephosphorylation status, in which S431 and T432 are replaced by alanine, has a higher ATPase activity than does native (wild-type) KaiC, whereas KaiC-DE that mimics the phosphorylated status, in which S431 and T432 are replaced by aspartic acid and glutamic acid, has a reduced ATPase activity (Terauchi et al. 2007). Although the ATPase activity of KaiC is extremely low, it is stable and consistent over a wide physiological range of temperatures and exhibits temperature compensation (Terauchi et al. 2007). This finding showed that the temperature compensation of this circadian clock is based on one chemical reaction, and was a major step toward elucidating the molecular mechanisms of temperature compensation of the

Mechanism of the Cyanobacterial Circadian Clock Protein KaiC to Measure. . .

83

circadian clock, which was previously enshrouded in mystery. As shown by Arrhenius’ equation, the rate constant of a typical enzymatic reaction increases exponentially with increasing temperature. However, the ATPase activity of KaiC is obviously not a typical enzymatic reaction and must be understood from perspectives that are distinctly different from previous biochemistry. Interestingly, temperature compensation of the ATPase activity is deficient in the KaiC homologues of the Rhodopseudomonas palustris, which is believed to have a proto-type circadian clock, and in KaiC3, the KaiC homolog of the cyanobacterium Synechocystis sp. PCC 6803 (Ma et al. 2016; Wiegard et al. 2020). In addition, wild-type KaiC and five KaiC period mutants show the important proportionality of the rate of the ATPase activity to frequency (reciprocal of the period) for both the in vivo bioluminescence rhythm or for the in vitro KaiC phosphorylation rhythm (Terauchi et al. 2007; Abe et al. 2015; Ito-Miwa et al. 2020). The ATPase activities of short-period mutant proteins are higher than those of wild-type KaiC, whereas those of long-period mutant proteins are lower. It has been pointed out that the ATPase activity of CI is particularly important in period determination (Abe et al. 2015). It has also been reported that increased hydrostatic pressure accelerates ATPase activity, resulting in a shorter period length of KaiC phosphorylation rhythm under high pressure (Kitahara et al. 2019). These two results about KaiC ATPase activity, i.e., temperature compensation and proportionality to frequency, imply that the ATPase activity of KaiC defines the most critical features of the circadian clock, namely, temperature compensation and period, suggesting that this reaction is the most fundamental pacemaker of the circadian clock. Interestingly, a relationship between ATPase activity of KaiC and cell division has also been pointed out (Dong et al. 2010). Based on these findings, we propose that the ATPase activity of KaiC functions as the pacemaker responsible for period determination and temperature compensation, and controls the cyanobacterial circadian system. These two properties that are exhibited by the ATPase activity of KaiC are realized by KaiC acting alone. This means that KaiC comprises an essential mechanism as the pacemaker of the circadian clock. On the other hand, note that KaiC by itself does not generate any circadian oscillations, including the phosphorylation rhythm. The circadian period and its temperature compensation are properties that are not only expressed when circadian oscillations are generated but are potentially encoded within the structure of the nonoscillating KaiC protein. This seemingly incomprehensible fact evokes the image of a stopped pendulum whose intrinsic properties have “memorized” its period even when it is not swinging, and strongly suggests that KaiC possesses the harmonic oscillator nature of a pendulum. How is such a thing possible? We assume intramolecular feedback within the KaiC molecule as one of the possibilities. In the normal ATPase, the energy generated by ATP hydrolysis is rapidly transferred to other molecules and used for physical or chemical work that follows the subsequent ATP hydrolysis. In the case of KaiC, the energy generated by ATP hydrolysis would accumulate within the same KaiC molecule, and generate tension in KaiC, such as in the winding of a spring. In the case of KaiC, this intramolecular tension acts to suppress its own ATPase

84

K. Ito-Miwa et al.

activity. We suggest that the energy accumulated in the KaiC molecule is released (leaked) over time. If the energy leakage is proportional to the tension, we propose that the energy production and energy leakage from ATP hydrolysis may be balanced at the same point, thereby maintaining a stable tension. In contrast to the many feedback systems seen in life, which are established by loops of chemical reactions by multiple interacting proteins, this is an “intramolecular feedback” that occurs within a single protein. Its essence can be assumed to be autoregulation by the mechanical properties of KaiC, such as distortion occurring in the molecule. In the case of the harmonic oscillation of a physical spring, the period becomes shorter as the spring becomes stronger and the same phenomenon may happen in KaiC. Unlike chemical reactions that are highly dependent on temperature, mechanical properties derived from molecular structures are expected to be less influenced by temperature, and equilibrium states due to intramolecular feedback are also expected to be less influenced by temperature. In other words, the temperature compensation of the period of KaiC phosphorylation rhythm could be achieved by a mechanical feedback system within a KaiC molecule.

5 Stable Circadian Oscillations Due to Interactions Between Two ATPase Domains of KaiC Although the energy of the harmonic oscillation of springs and pendulums is a conservative quantity, it is gradually lost by friction, etc., and the oscillation eventually damps. To generate a self-sustained oscillation that maintains a stable amplitude, it is necessary to supply an external force. The same is true for KaiC, which does not generate sustained oscillations by itself. When KaiA and KaiB are added to KaiC, a sustained phosphorylation rhythm is generated. At the same time, when KaiA and KaiB are added to the reaction, the ATPase activity also shows a circadian rhythm, which peaks as early as 4 h before the peak of the phosphorylation rhythm (Terauchi et al. 2007). The main action of KaiA is to stimulate the KaiC ATPase activity and phosphorylation of CII in KaiC (Iwasaki et al. 2002; Terauchi et al. 2007; Kim et al. 2008; Nishiwaki-Ohkawa et al. 2014). KaiB directly suppresses the ATPase activity of KaiC, but has no direct effect on KaiC phosphorylation. On the other hand, KaiB indirectly suppresses KaiC phosphorylation by inhibiting the stimulatory function of KaiA that promotes KaiC phosphorylation (Kitayama et al. 2003; Terauchi et al. 2007). The binding of KaiB to KaiC is dependent on the nucleotide-binding state of KaiC, and the hydrolysis of ATP bound to the protomer interface of KaiC changes the conformation of the KaiC hexamer, allowing KaiB to bind to the CI domain (Phong et al. 2013; Mutoh et al. 2013; Oyama et al. 2016, 2018; Tseng et al. 2017; Mukaiyama et al. 2018). Thus, the ATPase activity of KaiC is essential for the generation of stable rhythms, and the actions of KaiA and KaiB allow changes in the status of CI and CII that provide an external force to sustain harmonic oscillations of KaiC.

Mechanism of the Cyanobacterial Circadian Clock Protein KaiC to Measure. . .

A

Anchor linked to pendulum

85

B

Escape wheel linked to mainspring

WT Impulse from CII to CI

CI

CI locks CII

CII KaiC

CI ATP → ADP + Pi + Energy

Pacemaker

Tension in KaiC Impulse

Timing

Leak

E scapement

Mutant defective in temperature compensation

S/T

KaiA pS/T P-cycle S/pT KaiB

pS/pT

Driving force

CII runs freely

CII

Fig. 2 The clock model for the KaiC protein. (a) CI-CII coupling model. CI-ATPase and CII-phosphorylation cycle functions as a harmonic pacemaker and a driver of oscillation, respectively. The energy from ATP hydrolysis of CI generates a tension in the KaiC structure by intramolecular feedback. The tension in CI regulates the timing of the phosphorylation cycle in CII, and CII-driver supplies energy (“impulse”) back to CI to sustain the oscillation. This mechanism assumes the escapement of a mechanical pendulum clock (Fig. 3), and the CI and CII can be regarded as a pendulum and a mainspring, respectively. (b) Two domains of wild-type KaiC (CI and CII) interact in a time-dependent manner (upper). On the other hand, in KaiC mutants in which interaction between CI and CII is aberrant, CII is no longer optimally regulated by CI. Therefore, it is predicted that the phosphorylation cycle of CII in such a mutant runs freely, and exhibits the temperature-dependent nature that CII inherently possesses (mutant defective in temperature compensation) (lower)

For a stable rhythm of the Kai clock system, it is crucial that KaiC is composed of duplicate ATPase domains (CI and CII) and that these two domains interact with each other with the assistance of KaiA and KaiB (Fig. 2a). Among the two ATPase domains of KaiC, CI is responsible for more than half of the total ATPase activity and functions as a pacemaker with the properties of a harmonic oscillator (Terauchi et al. 2007; Abe et al. 2015). The ATPase activity of CII drives the phosphorylation cycle (KaiC autophosphorylation and autodephosphorylation) (Nishiwaki et al. 2004). Because this phosphorylation cycle of CII consists of loops composed of multiple chemical reactions (Kitayama et al. 2003; Xu et al. 2003; Nishiwaki et al. 2007), we hypothesize that CII has the properties of a relaxation oscillator, one of which is that the period is dependent on temperature. It is possible that the CI-pacemaker controls the timing of the activity of CII (autophosphorylation and autodephosphorylation), and therefore the period of the phosphorylation rhythm in CII becomes the same as that of the harmonic oscillation in CI, thereby achieving the temperature compensation. On the other hand, the harmonic oscillation of CI may

86

K. Ito-Miwa et al.

A

Escapement Anchor

B

Lock

Impulse

Escapement wheel

Rotate

Mainspring Impulse

Lock

Pendulum

Rotate Fig. 3 An escapement of a mechanical pendulum clock. (a) An escapement. It is composed of an “anchor” interlinked with the pendulum and a “gear wheel (escapement wheel)” driven by the mainspring. (b) Movement of escapement. The movement of the escapement wheel, which is powered by the spring, is alternately stopped (locked) or released by the two claws of the anchor, which are linked with the harmonic oscillation of the pendulum, and thereby the escapement wheel rotates at fixed intervals stepwise. On the other hand, the rotation of the escapement wheel provides the pendulum with energy (“impulse”) to sustain its oscillations

acquire the energy to sustain its harmonic oscillation from the “movement” of CII produced by the phosphorylation cycle in CII. Since CII has the same period as the harmonic oscillation of CI, CI could resonate with CII by even a very small amount of energy transferred from CII to CI, and consequently sustain a stable oscillation. Therefore, our model proposes two oscillators within KaiC with opposing properties of harmonic oscillations (CI) and relaxation oscillations (CII) that interact within one KaiC hexameric molecule to produce stable self-sustained oscillations. There are many examples of stable harmonic oscillations that persist by coupling oscillators with opposed properties that can be found in nature and in engineering (see Sect. 6). More specifically, the operations of KaiC are performed by functional specialization of the two ATPases, and the coupling of these complementary specializations is accomplished by the mechanical coupling of the two domains. We propose that the coupling is due to tiny movements of the internal structure of the protein. It is difficult to directly demonstrate these tiny movements within the KaiC structure, and hopefully future structural analyses will support our hypothesis. On the one hand, we may be able to experimentally approach the mechanical coupling of CI and CII from the viewpoint of temperature compensation (Fig. 2b). That is, in KaiC mutants in which CII is no longer optimally regulated by CI, the intrinsic temperaturedependent nature of CII may become partially or completely unmasked (because it is a relaxed oscillation), leading to loss of temperature compensation of the period and therefore a shorter period of the circadian rhythm at higher temperatures. We would therefore predict that the period of circadian rhythms in such a mutant protein

Mechanism of the Cyanobacterial Circadian Clock Protein KaiC to Measure. . .

87

will be much shorter than that in the wild type. We expect that when the intrinsic period of CII is shorter than that of CI, CI can effectively overrule the oscillation of CII by a simple mechanism such as stopping the CII reaction at a specific phase. On the other hand, if the intrinsic period of CII is longer than that of CI, it might not be easy for CI to accelerate the reaction of CII.

6 Design of Mechanical Clocks and the Design of Circadian Clocks As described in Sect. 3, when a deviation from an equilibrium state occurs in an object and a restoring force proportional to the deviation occurs, Hooke’s law is established and a harmonic oscillation is generated. The period of this oscillation does not change even if the amplitude changes, and the oscillatory energy is conserved. This oscillation is not a chemical reaction based on the statistical behavior of a large number of molecules, but a physical movement of macro objects as defined by Hook’s rule, and which is practically unaffected by temperature. Therefore, a harmonic oscillatory mechanism is advantageous as the basis for circadian oscillations that are characterized by the temperature compensation of period, as pointed out by Bünning (1967). Many ATPases that are involved in intracellular movement and structural maintenance can generate mechanical torque. Therefore, it is conceivable that the machinery generating harmonic oscillations based on physical mechanisms is hidden inside the higher-order structure of the KaiC ATPase. On the other hand, another important feature of the circadian clock is the persistence of oscillations. When looking for a mechanism to explain this persistence in biological phenomena, the first candidate is a relaxation oscillator, and indeed, this model can explain oscillations with various periods. A transcription–translation feedback model, which is considered to be the mechanism of the circadian clock in eukaryotic organisms, is also a model of a relaxation oscillation. At first, we thought that this model would also work for cyanobacteria (Ishiura et al. 1998), but it was difficult to explain the temperature compensation of the period. Therefore, the mechanical pendulum clock invented in the seventeenth century is a very good alternative that exemplifies a harmonic oscillation that stably persists. In a pendulum clock, the isochronous pendulum and the rotation of the gears powered by the mainspring are coupled in a device called an escapement (Fig. 3). The escapement is composed of a claw (anchor), which is connected to a pendulum, and a special form of a gear wheel (an escapement wheel), that is driven by the mainspring. The rotation speed of the escapement wheel is controlled by the pendulum to achieve a high degree of accuracy in the timing. By the shape of the gear, the mainspring transiently gives the energy (“impulse” in Fig. 3b) to the pendulum, and the pendulum thereby acquires the ability to persistently oscillate. The escapement is designed not to be constantly engaged like ordinary gears but to have a phase-limited

88

K. Ito-Miwa et al.

and directional action, achieving the oscillations with the combined advantages of two oscillators, namely, the reciprocal movement of the anchor and the rotation of the escapement wheel. These oscillations are known from many devices and physical phenomena (i.e. musical wind instruments, stringed musical instruments, and electrical oscillators) that require a stable period in addition to clocks. All of them combine harmonic oscillators that define the period of the oscillation (respectively corresponding to elasticity and volume of air columns in the tube, length of strings/cords that define the frequency, or time constant circuits by the condenser and the coil), with oscillators that amplify and stabilize them (respectively corresponding to airflow by breath or the motion of bows that sustain amplitude, or amplification by vacuum tubes and transistors). Assuming this design within the KaiC molecule, we can assign the ATPase activity of CI as the pendulum (pacemaker) and the phosphorylation cycle of CII as the mainspring (driving force) (Fig. 2a). The energy from ATP hydrolysis of CI accumulates within KaiC molecules and sustains the intramolecular tension that functions as a spring of the balance wheel. And then, CI acquires the potential to generate harmonic oscillations and functions as a pacemaker. On the other hand, the phosphorylation cycle of CII occurs through the action of KaiA and KaiB, and functions as an oscillator that essentially oscillates in a temperature-dependent manner. This cycling of the CII phosphorylation would function as a mainspring (driving force) and give CI an energy (impulse) to sustain the harmonic oscillation of CI. These two processes are interlocked by the escapement in the KaiC molecule, resulting in a phosphorylation cycle with a stable amplitude and a temperaturecompensated period. It is possible that the interaction between CI and CII, which results in temperature compensation of the period, is not a static and rigid coupling that constantly engages like ordinary gears, but is a time- (phase-) limited coupling. And such a loose coupling might need directionality for the action. This is the design principle of the cyanobacterial clock mechanism that we envision.

7 Conclusions Cyanobacteria have been considered to inhabit Earth for over 3 billion years. To stably keep track of time in harsh environments such as the ancient Earth, cyanobacteria may have evolved a harmonic oscillatory clock that can maintain a stable period against disturbances in the external environment. Recently, we have discovered that, depending on the size of a specific amino acid in KaiC (Tyr402), the period of the KaiC phosphorylation rhythm changes over a wide dynamic range from 0.6 days to 1 week (Ito-Miwa et al. 2020). These period mutants basically maintain temperature compensation of period, indicating that the mechanism of temperature compensation is more difficult to perturb than the period, which suggests that it is a more rigidly defended property and therefore a more fundamental property than that of the period determination. In addition, KaiC has two ATPase

Mechanism of the Cyanobacterial Circadian Clock Protein KaiC to Measure. . .

89

domains (CI and CII) in one protein, each of which promotes distinct functions as period-determining ATPase activity (CI) versus phosphorylation rhythmicity that allows the “spring” of CI to sustain oscillations (CII). In the future, by uncovering the biochemical activities and atomic-level structures of CI and CII in KaiC, we will clarify the mechanisms of these three prominent characteristics of KaiC, i.e., (1) the intramolecular feedback that establishes the “spring,” (2) the phosphorylation cycle that supplies CI the driving force and regulates the intracellular environment, and (3) the intramolecular coupling between CI and CII that functions as an “escapement.” Erwin Bünning, a founding pioneer of the circadian clock, suggested that circadian clocks have properties of both harmonic oscillators and relaxation oscillators (Bünning 1967), and Colin Pittendrigh, known as a founding “father” of the circadian clock, hypothesized multiple oscillators controlling circadian behavior where the coupling of an A oscillator with the property of harmonic oscillation and a B oscillator with the property of relaxation oscillation can explain many properties of circadian rhythms (Pittendrigh 1960). Although there is no homolog of KaiC in eukaryotes, ATP hydrolysis is the most fundamental enzymatic reaction in life, so it would be premature to deny the existence of an ATPase with KaiC-like functions (Edgar et al. 2012). It is obvious that life evolved the mechanism of “spring” and “escapement” in this protein, long before the invention by humans of a mechanical pendulum clock. Acknowledgment This review summarizes our studies on the mechanisms of circadian rhythm generation based on the functions of KaiC ATPase that we have performed since around 2010 at Nagoya University. The results summarizing this study are now in preparation for submission as a paper, together with the results of a number of experiments supporting a coupling model of CI and CII. We thank Dr. Keiko Okano-Imai (Kansai Medical University), Dr. Yoriko Murayama (Waseda University), Dr. Naoki Takai (Tokyo Metropolitan Institute of Medical Science), and Dr. Tomoaki Muranaka (Kagoshima University), who contributed greatly to our analyses. This study was supported by JST CREST Grant number JPMJCR07O2 to T.K. and Grants-in-Aid for Scientific Research (24000016 and 17H01427 to T.K). Editor’s Note: For a complementary view on this chapter, please also see Takao Kondo’s chapter, “Around the Circadian Clock: Review 1 and Preview”.

References Abe J, Hiyama TB, Mukaiyama A et al (2015) Atomic-scale origins of slowness in the cyanobacterial circadian clock. Science 349:312–316 Bünning E (1967) The physiological clock. SpringerVerlag, Berlin Dong G, Yang Q, Wang Q et al (2010) Elevated ATPase activity of KaiC applies a circadian checkpoint on cell division in Synechococcus elongatus. Cell 140(4):529–539 Dunlap JC, Loros JJ, DeCoursey PJ (eds) (2004) Chronobiology: biological timekeeping. Sinauer Associates Inc, Sunderland, MA Edgar RS, Green EW, Zhao Y et al (2012) Peroxiredoxins are conserved markers of circadian rhythms. Nature 485:459–464

90

K. Ito-Miwa et al.

Ishiura M, Kutsuna S, Aoki S et al (1998) Expression of a gene cluster kaiABC as a circadian feedback process in cyanobacteria. Science 281:1519–1523 Ito-Miwa K, Furuike Y, Akiyama S et al (2020) Tuning the circadian period of cyanobacteria up to 6.6 days by the single amino acid substitutions in KaiC. Proc Natl Acad Sci U S A 117 (34):20926–20931 Iwasaki H, Taniguchi Y, Ishiura M, Kondo T (1999) Physical interactions among circadian clock proteins KaiA, KaiB and KaiC in cyanobacteria. EMBO J 18(5):1137–1145 Iwasaki H, Nishiwaki T, Kitayama Y et al (2002) KaiA-stimulated KaiC phosphorylation in circadian timing loops in cyanobacteria. Proc Natl Acad Sci U S A 99(24):15788–15793 Kim YI, Dong G, Carruthers CW et al (2008) The day/night switch in KaiC, a central oscillator component of the circadian clock of cyanobacteria. Proc Natl Acad Sci U S A 105 (35):12825–12830 Kitahara R, Oyama K, Kawamura T et al (2019) Pressure accelerates the circadian clock of cyanobacteria. Sci Rep 9:12395 Kitayama Y, Iwasaki H, Nishiwaki T et al (2003) KaiB functions as an attenuator of KaiC phosphorylation in the cyanobacterial circadian clock system. EMBO J 22(9):2127–2134 Leipe DD, Aravind L, Grishin NV et al (2000) The bacterial replicative helicase DnaB evolved from a RecA duplication. Genome Res 10:5–16 Ma P, Mori T, Zhao C et al (2016) Evolution of KaiC-dependent timekeepers: A proto-circadian timing mechanism confers adaptive fitness in the purple bacterium Rhodopseudomonas palustris. PLoS Genet 12(3):e1005922 Mukaiyama A, Furuike Y, Abe J et al (2018) Conformational rearrangements of the C1 ring in KaiC measure the timing of assembly with KaiB. Sci Rep 8:8803 Murayama Y, Mukaiyama A, Imai K et al (2011) Tracking and visualizing the circadian ticking of the cyanobacterial clock protein KaiC in solution. EMBO J 30(1):68–78 Murayama Y, Kori H, Oshima C et al (2017) Low temperature nullifies the circadian clock in cyanobacteria through Hopf bifurcation. Proc Natl Acad Sci U S A 114(22):5641–5646 Mutoh R, Nishimura A, Yasui S et al (2013) The ATP-mediated regulation of KaiB-KaiC interaction in the cyanobacterial circadian clock. PLoS One 8(11):e80200 Nakajima M, Imai K, Ito H et al (2005) Reconstitution of circadian oscillation of cyanobacterial KaiC phosphorylation in vitro. Science 308:414–415 Nishiwaki T, Iwasaki H, Ishiura M et al (2000) Nucleotide binding and auto- phosphorylation of the clock protein KaiC as a circadian timing process of cyanobacteria. Proc Natl Acad Sci U S A 97 (1):495–499 Nishiwaki T, Satomi Y, Nakajima M et al (2004) Role of KaiC phosphorylation in the circadian clock system of Synechococcus elongatus PCC 7942. Proc Natl Acad Sci U S A 101 (38):13927–13932 Nishiwaki T, Satomi Y, Kitayama Y et al (2007) A sequential program of dual phosphorylation of KaiC as a basis for circadian rhythm in cyanobacteria. EMBO J 26(17):4029–4037 Nishiwaki-Ohkawa T, Kitayama Y, Ochiai E et al (2014) Exchange of ADP with ATP in the CII ATPase domain promotes autophosphorylation of cyanobacterial clock protein KaiC. Proc Natl Acad Sci U S A 111(12):4455–4460 Njus D, Gooch VD, Hasting JW (1981) Precision of the Gonyaulax circadian clock. Cell Biophys 3 (3):223–231 Nobel PS (1991) Characteristics of crossing membranes. In: Nobel PS (ed) Physicochemical and environmental plant physiology. Academic, Cambridge, pp 139–156 Oyama K, Azai C, Nakamura K et al (2016) Conversion between two conformational states of KaiC is induced by ATP hydrolysis as a trigger for cyanobacterial circadian oscillation. Sci Rep 6:32443 Oyama K, Azai C, Matsuyama J et al (2018) Phosphorylation at Thr432 induces structural destabilization of the CII ring in the circadian oscillator KaiC. FEBS Lett 592(1):36–45 Phong C, Markson JS, Wilhoite CM et al (2013) Robust and tunable circadian rhythms from differentially sensitive catalytic domains. Proc Natl Acad Sci U S A 110(3):1124–1129

Mechanism of the Cyanobacterial Circadian Clock Protein KaiC to Measure. . .

91

Pittendrigh CS (1960) Circadian rhythms and the circadian organization of living systems. Cold Spring Harb Symp 125:159–184 Terauchi K, Kitayama Y, Nishiwaki T et al (2007) ATPase activity of KaiC determines the basic timing for circadian clock of cyanobacteria. Proc Natl Acad Sci U S A 104(41):16377–16381 Tseng R, Goularte NF, Chavan A et al (2017) Structural basis of the day-night transition in a bacterial circadian clock. Science 355:1174–1180 Wiegard A, Köbler C, Oyama K et al (2020) Synechocystis KaiC3 displays temperature- and KaiBdependent ATPase activity and is important for growth in darkness. J Bacteriol 202(4):e00478– e00419 Xu Y, Mori T, Johnson CH (2003) Cyanobacterial circadian clockwork: roles of KaiA, KaiB and the kaiBC promoter in regulating KaiC. EMBO J 22(9):2117–2126 Yoshida T, Murayama Y, Ito H et al (2009) Nonparametric entrainment of the in vitro circadian phosphorylation rhythm of cyanobacterial KaiC by temperature cycle. Proc Natl Acad Sci U S A 106(5):1648–1653

Oscillation and Input Compensation in the Cyanobacterial Kai Proteins Michael Joseph Rust

Abstract The purified Kai protein oscillator remarkably recapitulates many of the qualitative features of circadian clocks in living organisms. Here I review the phenomenology of this system with a focus on how the Kai oscillator is special compared to generic biochemical oscillators. A key feature is that many external perturbations couple much more strongly to the amplitude than to the period of oscillations. This property is the basis of the ability of the oscillator to entrain to cycling environments while maintaining free-running rhythms that are always near 24 h. In contrast, internal perturbations that alter the structure of KaiC can cause large changes in the period. Explaining the molecular mechanism that underlies the sensitivity of period and amplitude to perturbations is a major remaining challenge that lies at the heart of clock physiology. I propose a toy mathematical model based on speculative biochemical assumptions that may shed some light on how the period and amplitude of the Kai protein oscillator arise.

1 The Cyanobacterial Oscillator as a Biochemical Model for Chronobiology The discovery of the reconstituted KaiABC oscillator by Takao Kondo’s group was a watershed moment in the effort to understand the mechanistic basis of circadian rhythms (Nakajima et al. 2005). The existence of a well-defined biochemical system that generates stable circadian oscillations has made it possible to unleash the tools of molecular genetics, biochemistry, and biophysics on the Kai proteins to uncover the molecular basis of the 24-h timescale of the rhythm. This is an exciting project being pursued by multiple research groups that has already met with remarkable successes, many of which are described elsewhere in this book.

M. J. Rust (*) Department of Molecular Genetics and Cell Biology, University of Chicago, Chicago, IL, USA Department of Physics, University of Chicago, Chicago, IL, USA e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_6

93

94

M. J. Rust

The potential of the Kai system to shed light on the mechanisms of circadian biology extends beyond the ability of purified proteins to generate a slow timescale. Further investigation has revealed that the KaiABC system has the ability to react to physiologically relevant inputs in a way that allows efficient entrainment while keeping the period of oscillation nearly invariant. This property is a hallmark of circadian physiology, and a famous example is temperature compensation. The period of oscillation of a circadian clock is typically only weakly dependent on temperature over the ecologically relevant range for an organism. Teleologically, we say that a clock should not run markedly faster or slower in the winter versus the summer since the 24 h length of the day is unchanged. Nevertheless, and perhaps paradoxically, temperature changes usually act as an efficient entraining cue for circadian clocks, causing large phase shifts. Thus, what must be explained is not immunity to temperature, but the coupling of temperature to the oscillator in such a way that preserves period (Francois et al. 2012). In this chapter, we will refer to this general phenomenon, when it occurs in response to external changes beyond temperature, as “input compensation.” First, we will review basic biochemical properties of the Kai proteins. Then, we will review experimental evidence that the KaiABC oscillator shows this input compensation property, with a focus on the varying the ATP/ADP ratio and varying temperature. We will show that one path to achieving input compensation can be visualized with the help of the phase plane, a tool from dynamical systems theory. Finally, we construct a toy mathematical model of the Kai oscillator where, under certain biochemical assumptions, there is a formal analogy to a harmonic oscillator with a positive feedback loop that pumps energy into the system. The structure of this toy model naturally puts period dependence into the internal catalytic rates of KaiC and suggests biochemical experiments that could deepen our understanding of the system.

2 Phenomenology of the Cyanobacterial Oscillator Mixtures of purified KaiA, KaiB, and KaiC proteins can generate circadian rhythms in a test tube, most obviously manifest as stable oscillations in the phosphorylation of KaiC (Nakajima et al. 2005). Numerous studies on this system have revealed several qualitative features of the oscillations that we view as essential features related to the ability of the Kai proteins to entrain to the environment and properly regulate circadian physiology. A major goal of the field is to explain the mechanistic origin of these phenomena: • Perturbing the system by changing the [KaiA]/[KaiC] ratio, the ATP/ADP ratio, temperature, etc. typically results in oscillations where the period is changed only slightly (~15% over the operational range) while the amplitude and baseline of the phosphorylation rhythm change markedly (Lin et al. 2014; Murayama et al. 2017; Nakajima et al. 2010; Phong et al. 2013; Yoshida et al. 2009). • In contrast, mutation of the KaiC protein sequence can produce enormous changes in period, from 18 h to >6 days. Often, these period mutation in KaiC

Oscillation and Input Compensation in the Cyanobacterial Kai Proteins

95

retain normal amplitude and temperature compensation properties (Ishiura et al. 1998; Ito-Miwa et al. 2020). • KaiC phosphorylation rhythms are more-or-less sinusoidal in waveform and do not show obvious relaxation oscillator-like character. This appears to hold even when the period is dramatically altered, so that long period mutants elongate the entire cycle, not just one phase. These biochemical observations are similar to the genetic structure of the circadian rhythm in vivo. Mutations have been identified outside the kai locus that alter the rhythm, but even mutation of key components like cikA or labA tends to alter period by at most ~15% while having a severe impact on amplitude (Schmitz et al. 2000; Taniguchi et al. 2007). Point mutations within the KaiC coding sequence can cause a large period change to the in vivo rhythm, often precisely analogous to the period of the purified system with these mutant proteins (Abe et al. 2015). To put the problem of explaining these phenomena in context, we will briefly summarize some of the known biochemical properties of the Kai proteins, a topic covered in detail by other chapters in this volume. KaiC is a double-domain member of the RecA superfamily of P-loop ATPases, and the two homologous ATPase domains are termed CI and CII. KaiC forms hexameric rings, giving the two domains a “double doughnut” appearance. As in related enzymes, the active sites are formed at subunit interfaces with residues from two adjacent subunits contributing to the reaction. The steady-state rate of ATP hydrolysis catalyzed by KaiC is on the order of hours (~15 ATP/subunit/day), strongly indicating that the slow timescale of the circadian rhythm is encoded in the KaiC structure and is not an emergent property of a network of interactions between proteins (Terauchi et al. 2007). The CII domain has the apparently unusual property that, in addition to the ATP hydrolysis reaction, it can catalyze phosphotransfer from the terminal phosphate of bound ATP to two conserved residues on an adjacent subunit, Ser431 and Thr432, in an ordered manner, i.e., autokinase activity. Remarkably, dephosphorylation proceeds through the reverse molecular pathway where bound ADP is required to serve as a phosphoacceptor for an autophosphatase reaction, regenerating ATP as a product (Egli et al. 2012; Nishiwaki and Kondo 2012). Thus, the observed rhythms of KaiC phosphorylation are rhythms in the balance of autokinase and autophosphatase activities occurring within a hexameric particle. Because of this mechanism, the bound nucleotide in CII plays a key role in determining whether KaiC will act as an autokinase or autophosphatase. During the day, KaiA stimulates the switch from autophosphatase mode to autokinase mode, in part by acting as a nucleotide exchange factor allowing ATP to be reloaded into the active site (Hong et al. 2018; Nishiwaki-Ohkawa et al. 2014). KaiA binds to sites on C-terminal “tentacles” that extend from the KaiC hexamer, restructuring the “Aloops” that form part of subunit interface (Kim et al. 2008). As Ser431 phosphorylation increases throughout the day, the A-loops retract into the protein, pushing KaiC toward a nighttime conformation, and it becomes increasingly difficult for KaiA to act. Critically, this nighttime state of KaiC involves a stacking interaction between the CII and CI domains (Chang et al. 2012). This

96

M. J. Rust

interaction modulates the catalytic cycle in CI and allows CI to accumulate bound ADP (Phong et al. 2013). This then allows the “B-loops” on the N-terminal face of the hexamer to become exposed, allowing KaiB to bind (Snijder et al. 2017; Tseng et al. 2017). The binding of KaiB to KaiC involves a dramatic refolding of the KaiB protein so that it adopts an alternative secondary structure. While the free form of KaiB appears to be biochemically inert, fold-switched KaiB can trap and inhibit KaiA (Chang et al. 2015). Thus, the nighttime state of KaiC is self-reinforcing: by inhibiting KaiA, it suppresses the ability of KaiA to convert KaiC back to the daytime state (Rust et al. 2007). Because the nighttime state is active as a phosphatase, Ser431 phosphorylation is removed throughout the night. This ultimately leads to disassembly of the inhibitory KaiB•KaiC complexes that are trapping KaiA, allowing KaiC to return to the daytime state and begin the cycle again.

3 Metabolic Input and Input Compensation An important question about the KaiABC protein oscillator is how it is coupled to the light–dark cycle in the context of a living cell to achieve entrainment and thus appropriate timing of clock-controlled physiology. The Kai proteins themselves are not light-sensitive, and investigations so far have not revealed a dedicated lightsensing clock input pathway in S. elongatus. An emerging consensus is that, instead, metabolic changes induced by photosynthesis are sufficient to provide resetting cues to the clock. Our test of this hypothesis made use of an engineered strain that can import glucose as an alternative energy source. We found that supplementing the culture medium with glucose could block the normal clock resetting effect of a dark pulse, and that, in the absence of light–dark cues, cycling external glucose can efficiently entrain the clock (Pattanayak et al. 2015). Thus, the circadian oscillator and the metabolic status of the cell are closely intertwined. How do metabolic fluctuations caused by the light–dark cycle entrain the circadian clock? There are probably multiple mechanisms. KaiA is sensitive to the redox state of the quinone pool, an indicator of the status of the electron transport chain (Wood et al. 2010). Darkness also inhibits the expression of the Kai proteins, along with a global suppression of gene expression (Ito et al. 2009). But the most fundamental influence of metabolism of the Kai system may be the ability of the ATP/ADP ratio to alter the phosphorylation of KaiC. As mentioned above, the bound nucleotide in CII is critical for setting the balance between kinase and phosphatase activity, and the relative concentrations of ATP and ADP in solution can therefore shift this balance. Mimicking changes in the ATP/ADP ratio observed in the cell during a pulse of darkness can produce an in vitro phase response curve with a similar shape to that observed in the cell (Rust et al. 2011). The KaiABC protein oscillator will entrain to repeated cycling of nucleotide concentrations with different “photoperiods” in a way that is qualitatively similar to the behavior of cells in different photoperiods—the

Oscillation and Input Compensation in the Cyanobacterial Kai Proteins

97

phase of KaiC phosphorylation closely follows the midpoint of the day (Leypunskiy et al. 2017).

4 Phase Plane Picture of Input Compensation and Entrainment What is the effect of various ATP/ADP ratios on the phosphorylation rhythm? As already mentioned, the period of oscillation is nearly unchanged by varying ATP/ADP but the amplitude and midpoint of the phosphorylation rhythm changes—when ATP/ADP is lower, fewer KaiC molecules are phosphorylated on average while still producing a ~ 24 h rhythm (Phong et al. 2013) (Fig. 1a). One way

A

B 60 %P-KaiC

%P-KaiC

60 40

40 20

20 0

80

100% 70% 50%

80

0 0

24

48

72

24

36

48

C

72

84

96

D

100% ATP 50% ATP

30

60 time (hrs)

time (hrs)

%pS431-KaiC

25 20 15 10

lights on

5 0 0

5

10

15

20

% pT432-KaiC

25

30

lights o

Fig. 1 Input compensation in the Kai oscillator. (a) KaiABC reactions in buffers with various % ATP achieved by mixing ATP and ADP (10 mM total). Changing the ATP/ADP ratio changes the amplitude and midpoint of the P-KaiC rhythm with little effect on period. (b) Phase shifts can be induced with pulses of lowered ATP/ADP. Orange curve is a KaiABC reaction in 100% ATP, cyan curve is shifted to 50% ATP for 5 h during the gray bar. (c) KaiC phosphorylation orbits for reactions at 100% ATP and 50% ATP in the plane of pS431 vs pT432. (d) Schematic showing a simple geometric model for entrainment of a uniform velocity oscillator. Blue arrows show rapid transition from the black orbit (when lights are off) to the blue orbit. Efficient entrainment requires the two orbits be significantly separated relative to their radius

98

M. J. Rust

of understanding the relationship between these properties and entrainment is via a simple picture from dynamical systems theory. We imagine that the oscillations correspond to a stable limit cycle on a two-dimensional manifold, for concreteness we illustrate this with experimental data in the plane of Ser431 phosphorylation and Thr432 phosphorylation (Fig. 1b). When the external ATP/ADP conditions are altered, this limit cycle is deformed—the size of the orbit changes and the midpoint moves. Nevertheless, it still takes ~24 h to transit around the cycle. For simplicity, we assume that there is a unique orbit for a given metabolic condition, independent of history—this seems to be true in the purified protein system, but may not be true in the living cell. We now approximate the effect of a light–dark or dark–light transition as causing the state of the system to move from its initial point to the new limit cycle defined by the change in metabolites. We can get an empirical view of the limit cycle by measuring two variables, such as KaiC phosphorylation on either Ser431 or Thr432, and plotting the points visited during the oscillation. The result shows a projection of the oscillator onto a plane (Fig. 1c). If the velocity around the cycles is uniform, entrainment is only possible if the effect of the environment is to shift the light limit cycle and dark limit cycle relative to each other, which is indeed what is observed with ATP/ADP shifts in the in vitro oscillator. Thus, this is a “nonparametric” formalism for studying entrainment where discrete transitions in the environment, e.g., between light and dark, cause abrupt jumps in the clock state by moving and deforming the limit cycle attractor. The detailed conditions for how entrained phase varies with the period of the environment and the day length follow from the geometry of the two limit cycles. In general, a separation of the centers of the light and dark cycles comparable to their radius is sufficient to create the midday tracking effect observed in vivo (Leypunskiy et al. 2017). This phase plane picture is a way to visualize the relationship between entrainment and the coupling of external stimuli to amplitude and midpoint (while leaving period relatively unchanged). Qualitatively similar phenomena can be seen for other influences that vary with the light–dark cycle, most notably temperature. Changing the ambient temperature also alters the amplitude and midpoint of the phosphorylation rhythm while having a mild impact on period (Yoshida et al. 2009). This is most dramatically seen in a recent study of the in vitro oscillator as temperature is lowered: amplitude goes smoothly to zero as temperature is lowered to the critical point where oscillations fail, while period is almost unchanged (Murayama et al. 2017). As with metabolite cues, the in vitro oscillator can entrain efficiently to temperature cycles, showing again the importance of zeitgeber cues coupling biochemically to amplitude and midpoint but not to the period of the oscillator.

Oscillation and Input Compensation in the Cyanobacterial Kai Proteins

99

5 A Toy Model with Integral Feedback Can Decouple Period and Amplitude Can we construct a toy model of the KaiABC oscillator that generates harmonic-like oscillations and has biochemical parameters that uncouple the period from amplitude and midpoint? To do so we will make gross simplifications and ignore many of the known features of the system while emphasizing others. First, we focus on the insight that KaiC hexamers can exist in two allosterically distinct states: a “daytime” state characterized by weakened interactions between the CI and CII domains, an inability to bind KaiB, and exposed A-loops capable of KaiA binding, and a “nighttime” state where the CI and CII domains are stacked and the B-loops permit KaiB binding (Chang et al. 2012). We keep track of two dynamical variables, C, the fraction of KaiC hexamers that are in the daytime state, and P, the average level of Ser431 phosphorylation in a hexamer. For simplicity, we assume that each hexamer experiences the same phosphorylation level, i.e., we make a mean-field approximation that ignores variation between individual hexamers, and we ignore the role of Thr432 phosphorylation. Suppose that KaiC interconverts between the day and night states with rate constants α and β, and that the role of increased phosphorylation is to both accelerate conversion into the nighttime state and slow conversion into the daytime state (Fig. 2a). We codify this in the following assumption. Assumption 1 Phosphorylation acts linearly to modify both the forward and reverse rates. In a two-state equilibrium, this can be achieved by having a low barrier between states and raising and lowering the free energy of the two states respectively by the same amount. Importantly, this is a coarse-grained picture of a complex system and not a statement of thermodynamic equilibrium; among other things, interconversion between the day and night states involves ATP hydrolysis. dC ¼ ðα  aPÞð1  C Þ  ðβ þ aPÞC dt Here a sets the scale of how much the rate of KaiC interconversion is changed by a single phosphorylation. Taking another derivative, we get: d2 C dC dP a ¼  ðα þ β Þ dt dt dt 2 where the linearity of Assumption 1 causes the cross-terms to cancel. We now must write a differential equation describing the rate of change of phosphorylation. Assumption 2 Catalysis in a KaiC hexamer is sequential. The consequence of this assumption is that when a KaiC hexamer is in its kinase mode, only one subunit is available to receive the next phosphorylation (a similar rule applies for dephosphorylation). Thus, the rate of phosphorylation will (approximately) depend only on

100

M. J. Rust

A

B

KaiA

P f(C) P- P-

P- P-

P P- P-

P- P-

C

P

P

C

f(C)(1-C)

C

C

Fig. 2 A toy model for the KaiABC oscillator. (a) The core of the model treats the KaiC hexamer as a 2-state system, interconverting between a nighttime state (phosphatase active, capable of KaiB binding) and a daytime state (kinase active). The role of (Ser431) phosphorylation is to slow conversion into the daytime state and accelerate conversion into the nighttime state. (b) The role of KaiA in the toy model is to catalyze conversion into the daytime state. The nighttime state inactivates KaiA through KaiB binding and other mechanisms, leading to the action of KaiA depending on the balance between daytime and nighttime states, a process described by f(C). (c) Approximate forms for the reaction flux f(C)(1 - C) where f(C) is a Hill function. Blue and orange curves correspond to two different KaiA concentrations by changing the half-maximal constant in the Hill function (marked by a dashed line). Gray bar indicates the region where the slope of the function is large and positive—in the mechanical oscillator analogy, this is where energy is pumped into the system

the probability for a hexamer to be in its kinase mode and not the current level of phosphorylation. This can only be approximate because, if a hexamer becomes completely phosphorylated, further phosphorylation is impossible. Nevertheless, at intermediate levels of phosphorylation, we may write: dP  γ þ C  γ  ð1  C Þ ¼ γ ðC  C 0 Þ dt Here C0 represents the point at which the kinase and phosphatase activities of the two states are balanced and γ is the total catalytic rate of the CII domain. Plugging this into the above equation, we get:

Oscillation and Input Compensation in the Cyanobacterial Kai Proteins

101

d2 C dC  aγ ðC  C0 Þ ¼  ðα þ β Þ dt dt 2 This is formally analogous to a mass on a spring or ball oscillating in a potential qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2

Þ well subject to viscous drag. The frequency is aγ þ ðαþβ and the midpoint of the 4 oscillation is C0. Of course, as discussed above, circadian rhythms are not damped oscillators; they are self-sustaining oscillations with a fixed amplitude. To see how this toy model can be modified to generate stable oscillations, we must include the effect of KaiA and KaiB. Suppose that there is another KaiA-dependent process described by a rate f(C) that converts KaiC into the daytime state, and that this rate itself increases with C, the current fraction of KaiC molecules in the daytime state, i.e., a positive feedback loop. The model is:

dC ¼ ðα  aPÞð1  CÞ  ðβ þ aPÞC þ f ðCÞð1  C Þ dt Again taking a derivative: d2 C dC d dC  aγ ðC  C 0 Þ þ ½ f ðC Þð1  C Þ ¼  ðα þ β Þ dt dC dt dt 2 Let us consider a specific mechanism for f(C) in terms of KaiA and KaiB, which have not yet appeared in this toy model. Suppose that KaiA acts to promote the daytime state of C by binding to the A-loops and causing nucleotide exchange. When KaiC is in the nighttime state, it forms KaiB•KaiC complexes that can in turn trap and inhibit KaiA. If we (unrealistically) assume that the formation of these complexes is rapid, this can be considered simply as a KaiA-dependent rate to convert the nighttime state (1 - C) to the daytime state (C) that depends nonlinearly on C itself because the daytime state can reinforce itself by opposing KaiB binding and ensuring KaiA is free to act (Rust et al. 2007). To get the gist of how this model will behave, we can take f(C) to be a Hill n function CnCþK n. Here K is the fraction of KaiC in the daytime state needed to release enough KaiA to achieve a half-maximal effect on stimulating KaiC. It is easier to see d  the implications graphically (Fig. 2c). There will be a region around K where dC ½ f ðCÞð1  CÞ is large and positive. This is a positive feedback loop where the daytime state of KaiC is selfreinforcing. In the mechanical oscillator analogy, this corresponds to a region of the bowl where a force opposes viscous drag and pumps energy into the oscillator. Thus, a stable limit cycle emerges at the amplitude where the energy loss from friction over the entire cycle is balanced by the kick the system receives in the region d where dC ½ f ðC Þð1  C Þ is positive. Likely other positive feedback mechanisms contribute to what is lumped into f(C) in this model. For example, previous data

102

M. J. Rust

suggest that it is easier for KaiA to bind to the daytime form of KaiC where the A-loops are extended, thus K may itself be a function of C (Chang et al. 2012).

6 Period and Amplitude in the Model Despite how contrived this toy model is, the oscillations have some features that are similar to what have been observed in experiments, at least in appropriate parameter regimes. At the beginning of this chapter, I laid out the challenges of understanding why the Kai oscillator shows balanced, nearly sinusoidal oscillations even when perturbed, why changes in external conditions tend to alter amplitude much more than period, and why mutation of KaiC seems to be the only way to achieve large changes in period. We can now ask what the answers to these questions look like in the model. The similarity to a driven harmonic oscillator results in nearly sinusoidal waveforms. The period of oscillation can be relatively stable over the functional range of KaiA concentrations (Fig. 3)—this is because the period is primarily set by the catalytic rate constant γ describing the timescale of phosphorylation/dephosphorylation, and the effect of changing KaiA concentration largely alters where the “f(C) kick” occurs that gives the system a stable amplitude, by changing the constant K that describes how the amount of free KaiA depends on C. Changing this kick changes the amplitude of oscillations, analogous to how changing how close you stand to a swingset when pushing a swing changes the amplitude. Note that α, β, γ are set by KaiC catalytic rates, and if these rates are constant in the face of external fluctuations then the period of oscillations can be also held nearly constant. Indeed, experiments have shown that the total rate of ATP turnover is nearly unchanged at different temperatures, though the roles of the CI and CII domains have not been unambiguously dissected. The question of how the enzymology of KaiC enforces a temperature-independent catalytic rate remains an unsolved mystery. But we can now ask how temperature might couple to other aspects of the system to change amplitude and midpoint, while having little effect on period. One possibility is that temperature shifts the balance between kinase and phosphatase activities, in this model changing the feedback set point C0. This could potentially explain the observation that baseline KaiC phosphorylation increases, even in the absence of KaiA, as temperature is decreased. Another possibility is that temperature changes the effectiveness of KaiA, described in this model via the f(C) function. These possibilities could be looked for in future experiments.

Oscillation and Input Compensation in the Cyanobacterial Kai Proteins

B

vary [KaiA] (K)

P-KaiC / hexamer

vary catalytic rate ( )

P-KaiC / hexamer

A

103

time

time

C relative change

period amplitude

K Fig. 3 Numerical simulations of the toy model. (a) Simulations where the catalytic rate γ is increased or decreased by a factor of 2. (b) Simulations where the half-maximal constant K in the Hill function used for f(C) is varied. (c) Plot of normalized period and amplitude as the halfmaximal constant K is varied. K is the fraction of KaiC in the daytime state needed to release enough KaiA to achieve a half-maximal effect on stimulating KaiC. In all simulations, time is rescaled to make the period of the reference conditions exactly 24 h. α ¼ 5, β ¼ 4, γ ¼ 10, a ¼ 1, C 0 ¼ 5

C 0:5, f ðC Þ ¼ 40 C5 þ0:8 5 . The model output P is analogous to inhibitory Ser431 phosphorylation

7 Does KaiC Phosphorylation Implement Integral Feedback? Are the biochemical ingredients in this toy model plausible? The form of the dP/dt equation describing changing levels of phosphorylation where P itself does not appear on the right-hand side has a special structure sometimes called “integral feedback,” because integrating both sides reveals that P(t) follows the “error signal” between the activity state of KaiC and its set point C0. It is known from systems engineering that integral control is the only way to guarantee zero steady-state error in a system, and integral controllers are thought to appear in several biochemical contexts such as bacterial chemotaxis and developmental patterning (Barkai and Leibler 1997; Ben-Zvi and Barkai 2010). A key difference here is that, by coupling integral feedback to KaiA-dependent positive feedback, stable oscillations around the fixed point can occur (Fig. 4a).

104

M. J. Rust

A

B

KaiA

P- P-

P- P-

P dt

D P-KaiC / hexamer

C

P

C

time

Fig. 4 Mechanical analogy and attractor structure. (a) In the toy model, KaiC phosphorylation acts as an integral controller that is made unstable by positive feedback through KaiA interaction. (b) When an integral controller acts linearly on the system, it is formally equivalent to a damped harmonic oscillator where positive feedback plays the role of counteracting friction and pumping energy into the system. Depicted here as a parent (analogous to the KaiA-dependent positive feedback loop) pushing a child on a swing. The natural frequency of the swing is set by the timescale of the integral controller, in this case KaiC catalysis. (c) A possible qualitative phase portrait of this system. When the rising part of f(C) is localized away from the set point C0, a possible scenario is that the set point will be stable and sufficient close initial conditions will fail to oscillate (pink region). Initial conditions that allow far-enough excursions from the set point will feel the “kick” of positive feedback and approach a limit cycle (green region). (d) Numerical simulations show that these possibilities can be realized in the toy model. The different colored trajectories correspond to different initial phosphorylation states

As mentioned earlier, the absence of P from the right-hand side of the dP/dt equation implies that the number of sites available for (de)phosphorylation does not depend on the current level of phosphorylation P. One biochemical interpretation is that catalysis must proceed sequentially around a KaiC hexamer. Even in this case, the P-independence can only be approximate. For example, when a hexamer becomes completely phosphorylated, no further increase in P is possible and so the P-independence must break down. To our knowledge, there is currently no decisive evidence whether catalysis within the KaiC ring is sequential, though there is evidence that related enyzmes such as RecA and the F1 ATPase use sequential catalysis (Cox et al. 2005; Uchihashi et al. 2011).

Oscillation and Input Compensation in the Cyanobacterial Kai Proteins

105

Another possible biochemical interpretation is that only one of the KaiC subunits is selected, perhaps by conformational fluctuations, to be catalytically active at any time. Because phosphorylation and dephosphorylation occur at the same active site, this selected subunit would either be phosphorylated or dephosphorylated depending on the activity state of the hexamer. In this scenario as well, the dP/dt equation above could hold approximately. An important ingredient of the model is that while phosphorylation is correlated with KaiB binding, it is not, as has sometimes been assumed, the direct cause. Rather the role of phosphorylation is to bias interconversion between the daytime and nighttime states, an interconversion that is also being influenced by KaiA. This is a somewhat subtle conceptual difference, but it amounts to viewing the daytime and nighttime conformational interconversion of KaiC as a the central hub of the reaction network, and the other players—KaiA, phosphorylation status—only influence each other through it. There are some experimental data in support of this idea. For example, KaiC with phosphomimetic mutations binds strongly to KaiB, but large concentrations of KaiA can prevent the KaiB–KaiC interaction even though the protein cannot be further phosphorylated (Lin et al. 2014). The kinetics of KaiC phosphorylation is also quite different in this model. Phosphorylation time series in the literature have been described reasonably well by first-order kinetic models. In such models each equivalent phosphorylation site in a hexamer has an equal probability to be phosphorylated per unit time, so the rate of increase of phosphorylation slows as more sites are modified. In a sequential phosphorylation model with integral feedback, the phosphorylation rate instead slows over time because a more heavily phosphorylated enzyme alters the activity of the enzyme. Future experimental work that allows a more detailed view of the KaiC catalytic cycle and detailed kinetic studies, especially including phosphorylation site mutants, will hopefully shed light on these possibilities.

8 Coexistence of a Stable Fixed Point and a Limit Cycle The utility of toy models is that they can promote intuitive understanding and inspire new kinds of experiments. The picture that emerges from this model is analogous to KaiC as a child being pushed on a swing; the angle that the swing makes with the ground represents the balance the daytime and nighttime conformations of the KaiC hexamer. Stimulation by KaiA and the sequestration feedback on KaiA is analogous to a parent standing at a fixed position from the swing and providing a push every cycle (Fig. 4b). The usual language of biochemical oscillators arising from delayed negative feedback loops looks different here: integral feedback through KaiC phosphorylation creates an effective harmonic potential and stable oscillations arise from the addition of positive feedback through KaiA interaction. The combination of negative- and positive-feedback is a motif that can produce stable oscillations even if the negative feedback is not through an integral controller (Li et al. 2017); the observation here is that when integral feedback couples linearly to a positive

106

M. J. Rust

feedback system, the resulting harmonic-like oscillations naturally decouple period from amplitude in the biochemical parameters. Without KaiA interaction—without the parent pushing the swing—KaiC will settle into a steady state, but the equations suggest that damped oscillations on the way to steady state are possible. Elegant experiments from the Akiyama lab show that, when KaiC is freshly reloaded with ATP, the catalytic activity shows a damped oscillation on the way to steady state in the absence of KaiA and KaiB (Abe et al. 2015). The model suggests that these oscillations in ATPase activity following an ATP reload may be accompanied by a pulse of (KaiA-independent) phosphorylation that could be looked for experimentally. Finally, we consider a qualitative prediction about the attractor structure of the model. To return to the child-on-a-swing analogy: changing KaiA levels corresponds to moving the position of the parent relative to the swing and changing the strength of the push. At low KaiA levels, we can imagine the parent moving further and further from the resting position of the swing, and an interesting phenomenon becomes possible. If the swing is close to steady state, it will settle to the fixed point without being able to receive a push, and oscillations will die out. For a different set of initial conditions, for example, starting the swing far from steady state, repeated pushes will stabilize a high-amplitude oscillation. In other words, whether circadian rhythms will occur or not may depend on the initial conditions, especially as KaiA levels are lowered (Fig. 4c, d). This possibility could be probed experimentally through a phase mixing experiment where oscillating reactions in different phases are mixed together in various proportions—some mixtures that get sufficiently close to the fixed point would be expected to annihilate rhythms, a possibility sometimes called a “black hole” (Winfree 2001).

9 Conclusion Biochemical oscillations can occur whenever a negative feedback process acts with a delay. My hope here was to point out that the constraints on circadian oscillators are stronger than just this simple statement. Circadian oscillators must be able to entrain to cycling environments but also maintain the correct speed when environmental cues are absent. One possible way to achieve this can be described by visualizing how the orbit describing the state of the oscillator changes in the phase plane in response to the environment. If an input shifts and stretches the orbit while leaving the angular velocity mostly unchanged, it can act as an efficient entraining cue. This is hardly a new observation in circadian biology, but it leads to two obvious questions. Do the orbits of real circadian oscillators respond to the environment in this way? If so, how is this process implemented molecularly? Work on the Kai protein oscillator from multiple labs has answered the first question in the affirmative, both for temperature changes and for changes in the nucleotide composition of the buffer that simulate metabolic changes (Leypunskiy et al. 2017; Murayama et al. 2017; Phong et al. 2013; Rust et al. 2011; Yoshida et al. 2009).

Oscillation and Input Compensation in the Cyanobacterial Kai Proteins

107

To try to address the second question, I suggested a framework for a model of the Kai system that is based on much of the known biochemistry but makes some speculative and unusual assumptions that make it different from many models. In this model, in an appropriate parameter regime, oscillations naturally look sinusoidal, corresponding to nearly circular orbits with nearly constant angular velocity. The period is largely set by rates internal to KaiC, and the amplitude and midpoint can be adjusted by tweaking the internal set point of KaiC or the interaction strength with KaiA and KaiB. Manipulations that achieve this are candidates for being good entraining cues for the oscillator while maintaining a compensated period. If we view this model as a tool for generating hypotheses, it suggests that we look at any manipulation that changes amplitude much more than period and investigate how it might modulate the activity of KaiA. Notably, changing the ATP/ADP ratio can be understood as interfering with KaiA’s nucleotide activity and the impact of varying ATP/ADP on phosphorylation is greater when KaiA concentrations are low (Hong et al. 2020). The flipside of the ability to tune the amplitude through changing interaction with KaiA is the identification of period-determining parameters with KaiC itself. This suggests that KaiC is able to create a protected environment for slow biochemistry to occur that is remarkably shielded from external conditions. The ability to dramatically change the speed of the clock with single amino acid substitutions in KaiC suggests that its catalytically properties are quite structurally finetuned (Ito-Miwa et al. 2020). A biophysical explanation of precisely how this timescale is set and how the insulation from the environment is achieved is an exciting unsolved puzzle.

References Abe J, Hiyama TB, Mukaiyama A, Son S, Mori T, Saito S et al (2015) Circadian rhythms. Atomicscale origins of slowness in the cyanobacterial circadian clock. Science 349(6245):312–316. https://doi.org/10.1126/science.1261040 Barkai N, Leibler S (1997) Robustness in simple biochemical networks. Nature 387 (6636):913–917. https://doi.org/10.1038/43199 Ben-Zvi D, Barkai N (2010) Scaling of morphogen gradients by an expansion-repression integral feedback control. Proc Natl Acad Sci U S A 107(15):6924–6929. https://doi.org/10.1073/pnas. 0912734107 Chang YG, Tseng R, Kuo NW, LiWang A (2012) Rhythmic ring-ring stacking drives the circadian oscillator clockwise. Proc Natl Acad Sci U S A 109(42):16847–16851. https://doi.org/10.1073/ pnas.1211508109 Chang YG, Cohen SE, Phong C, Myers WK, Kim YI, Tseng R et al (2015) Circadian rhythms. A protein fold switch joins the circadian oscillator to clock output in cyanobacteria. Science 349 (6245):324–328. https://doi.org/10.1126/science.1260031 Cox JM, Tsodikov OV, Cox MM (2005) Organized unidirectional waves of ATP hydrolysis within a RecA filament. PLoS Biol 3(2):e52. https://doi.org/10.1371/journal.pbio.0030052 Egli M, Mori T, Pattanayek R, Xu Y, Qin X, Johnson CH (2012) Dephosphorylation of the core clock protein KaiC in the cyanobacterial KaiABC circadian oscillator proceeds via an ATP synthase mechanism. Biochemistry 51(8):1547–1558. https://doi.org/10.1021/bi201525n

108

M. J. Rust

Francois P, Despierre N, Siggia ED (2012) Adaptive temperature compensation in circadian oscillations. PLoS Comput Biol 8(7):e1002585. https://doi.org/10.1371/journal.pcbi.1002585 Hong L, Vani BP, Thiede EH, Rust MJ, Dinner AR (2018) Molecular dynamics simulations of nucleotide release from the circadian clock protein KaiC reveal atomic-resolution functional insights. Proc Natl Acad Sci U S A 115(49):E11475–E11484. https://doi.org/10.1073/pnas. 1812555115 Hong L, Lavrentovich DO, Chavan A, Leypunskiy E, Li E, Matthews C et al (2020) Bayesian modeling reveals metabolite-dependent ultrasensitivity in the cyanobacterial circadian clock. Mol Syst Biol 16(6):e9355. https://doi.org/10.15252/msb.20199355 Ishiura M, Kutsuna S, Aoki S, Iwasaki H, Andersson CR, Tanabe A et al (1998) Expression of a gene cluster kaiABC as a circadian feedback process in cyanobacteria. Science 281 (5382):1519–1523. https://doi.org/10.1126/science.281.5382.1519 Ito H, Mutsuda M, Murayama Y, Tomita J, Hosokawa N, Terauchi K et al (2009) Cyanobacterial daily life with Kai-based circadian and diurnal genome-wide transcriptional control in Synechococcus elongatus. Proc Natl Acad Sci U S A 106(33):14168–14173. https://doi.org/ 10.1073/pnas.0902587106 Ito-Miwa K, Furuike Y, Akiyama S, Kondo T (2020) Tuning the circadian period of cyanobacteria up to 6.6 days by the single amino acid substitutions in KaiC. Proc Natl Acad Sci U S A. https:// doi.org/10.1073/pnas.2005496117 Kim YI, Dong G, Carruthers CW Jr, Golden SS, LiWang A (2008) The day/night switch in KaiC, a central oscillator component of the circadian clock of cyanobacteria. Proc Natl Acad Sci U S A 105(35):12825–12830. https://doi.org/10.1073/pnas.0800526105 Leypunskiy E, Lin J, Yoo H, Lee U, Dinner AR, Rust MJ (2017) The cyanobacterial circadian clock follows midday in vivo and in vitro. elife 6:80. https://doi.org/10.7554/eLife.23539 Li Z, Liu S, Yang Q (2017) Incoherent inputs enhance the robustness of biological oscillators. Cell Syst 5(1):72–81 e74. https://doi.org/10.1016/j.cels.2017.06.013 Lin J, Chew J, Chockanathan U, Rust MJ (2014) Mixtures of opposing phosphorylations within hexamers precisely time feedback in the cyanobacterial circadian clock. Proc Natl Acad Sci U S A 111(37):E3937–E3945. https://doi.org/10.1073/pnas.1408692111 Murayama Y, Kori H, Oshima C, Kondo T, Iwasaki H, Ito H (2017) Low temperature nullifies the circadian clock in cyanobacteria through Hopf bifurcation. Proc Natl Acad Sci U S A 114 (22):5641–5646. https://doi.org/10.1073/pnas.1620378114 Nakajima M, Imai K, Ito H, Nishiwaki T, Murayama Y, Iwasaki H et al (2005) Reconstitution of circadian oscillation of cyanobacterial KaiC phosphorylation in vitro. Science 308 (5720):414–415. https://doi.org/10.1126/science.1108451 Nakajima M, Ito H, Kondo T (2010) In vitro regulation of circadian phosphorylation rhythm of cyanobacterial clock protein KaiC by KaiA and KaiB. FEBS Lett 584(5):898–902. https://doi. org/10.1016/j.febslet.2010.01.016 Nishiwaki T, Kondo T (2012) Circadian autodephosphorylation of cyanobacterial clock protein KaiC occurs via formation of ATP as intermediate. J Biol Chem 287(22):18030–18035. https:// doi.org/10.1074/jbc.M112.350660 Nishiwaki-Ohkawa T, Kitayama Y, Ochiai E, Kondo T (2014) Exchange of ADP with ATP in the CII ATPase domain promotes autophosphorylation of cyanobacterial clock protein KaiC. Proc Natl Acad Sci U S A 111(12):4455–4460. https://doi.org/10.1073/pnas.1319353111 Pattanayak GK, Lambert G, Bernat K, Rust MJ (2015) Controlling the cyanobacterial clock by synthetically rewiring metabolism. Cell Rep 13(11):2362–2367. https://doi.org/10.1016/j. celrep.2015.11.031 Phong C, Markson JS, Wilhoite CM, Rust MJ (2013) Robust and tunable circadian rhythms from differentially sensitive catalytic domains. Proc Natl Acad Sci U S A 110(3):1124–1129. https:// doi.org/10.1073/pnas.1212113110 Rust MJ, Markson JS, Lane WS, Fisher DS, O'Shea EK (2007) Ordered phosphorylation governs oscillation of a three-protein circadian clock. Science 318(5851):809–812. https://doi.org/10. 1126/science.1148596

Oscillation and Input Compensation in the Cyanobacterial Kai Proteins

109

Rust MJ, Golden SS, O'Shea EK (2011) Light-driven changes in energy metabolism directly entrain the cyanobacterial circadian oscillator. Science 331(6014):220–223. https://doi.org/10.1126/ science.1197243 Schmitz O, Katayama M, Williams SB, Kondo T, Golden SS (2000) CikA, a bacteriophytochrome that resets the cyanobacterial circadian clock. Science 289(5480):765–768. https://doi.org/10. 1126/science.289.5480.765 Snijder J, Schuller JM, Wiegard A, Lossl P, Schmelling N, Axmann IM et al (2017) Structures of the cyanobacterial circadian oscillator frozen in a fully assembled state. Science 355 (6330):1181–1184. https://doi.org/10.1126/science.aag3218 Taniguchi Y, Katayama M, Ito R, Takai N, Kondo T, Oyama T (2007) labA: a novel gene required for negative feedback regulation of the cyanobacterial circadian clock protein KaiC. Genes Dev 21(1):60–70. https://doi.org/10.1101/gad.1488107 Terauchi K, Kitayama Y, Nishiwaki T, Miwa K, Murayama Y, Oyama T, Kondo T (2007) ATPase activity of KaiC determines the basic timing for circadian clock of cyanobacteria. Proc Natl Acad Sci U S A 104(41):16377–16381. https://doi.org/10.1073/pnas.0706292104 Tseng R, Goularte NF, Chavan A, Luu J, Cohen SE, Chang YG et al (2017) Structural basis of the day-night transition in a bacterial circadian clock. Science 355(6330):1174–1180. https://doi. org/10.1126/science.aag2516 Uchihashi T, Iino R, Ando T, Noji H (2011) High-speed atomic force microscopy reveals rotary catalysis of rotorless F(1)-ATPase. Science 333(6043):755–758. https://doi.org/10.1126/ science.1205510 Winfree AT (2001) The geometry of biological time, 2nd edn. Springer, New York Wood TL, Bridwell-Rabb J, Kim YI, Gao T, Chang YG, LiWang A et al (2010) The KaiA protein of the cyanobacterial circadian oscillator is modulated by a redox-active cofactor. Proc Natl Acad Sci U S A 107(13):5804–5809. https://doi.org/10.1073/pnas.0910141107 Yoshida T, Murayama Y, Ito H, Kageyama H, Kondo T (2009) Nonparametric entrainment of the in vitro circadian phosphorylation rhythm of cyanobacterial KaiC by temperature cycle. Proc Natl Acad Sci U S A 106(5):1648–1653. https://doi.org/10.1073/pnas.0806741106

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria Maria Luísa Jabbur, Chi Zhao, and Carl Hirschie Johnson

Abstract Circadian clocks are ubiquitous throughout the Tree of Life, being present in organisms from bacteria to mammals. These clocks are generally thought to increase the fitness of organisms by allowing the anticipation of and preparation for the predictable daily changes associated with Earth’s rotation. Here we consider what it takes to show that clocks indeed increase fitness and adaptiveness, as well as what bacteria can tell us about the evolution of clocks. We give a historical panorama of the experimental approaches that attempted to demonstrate the adaptive value of circadian clocks and explore current and future experiments that could facilitate an understanding of the adaptiveness of clocks in both intra- and interspecific contexts. Finally, we explain how studying bacteria can generate a greater appreciation and understanding for the general principles of the adaptive value of circadian clocks not only towards the daily cycle but also for the seasonal cycle, from both a past and future evolutionary perspective.

1 Evolution of Circadian Clocks: What Can Bacterial Clocks Tell Us? In the last couple of decades, bacteria—especially cyanobacteria—have established themselves as an important tool to test hypothesis regarding the fitness and evolutionary aspects of circadian clocks. But what can bacteria teach us about the evolution of circadian clocks? Why do we care about them, when they are so different from eukaryotes, and there seems to be little conservation between their clocks both from a genetic and a mechanistic perspective? Complex eukaryotes—such as mammals—often have genetic/proteic/metabolic pathways that are elaborate, redundant, or cross-functional, which can complicate the testing of evolutionary hypotheses and easily confound our conclusions. M. L. Jabbur · C. Zhao · C. H. Johnson (*) Department of Biological Sciences, Vanderbilt University, Nashville, TN, USA e-mail: [email protected]; [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_7

111

112

M. L. Jabbur et al.

Consider, for example, the Bmal1 gene, which is a central transcriptional activator in the core clock mechanism of mammals. The behavior of processes in Bmal1knockout animals are often used as THE criterion for clock-regulation—if the process is arhythmic in Bmal1-knockout mice, then that process is often labeled as a clock-controlled system. However, when the Bmal1 gene is knocked out in mice, the resulting animals are not merely arhythmic; they are usually also sick in laboratory conditions (Kondratov et al. 2006; Somanath et al. 2011). One could assume that their arhythmicity leads to this sick phenotype, however these deleterious effects are not seen in equally arhythmic mice in which the double knockouts of Cry1/Cry2 or Per1/Per2 genes have been performed (Cry1/2 and Per1/2 are also key clock genes in mammals). The sickness observed in Bmal1-ko mice appears to result from the fact that Bmal1 is a major transcriptional activator linked to many processes. The knockout of Bmal1 disrupts many intricately interlinked pathways, not all of which need to be regulated in a rhythmic fashion. Conversely, the fact that Cry1/2-dko and Per1/2-dko mice are healthy in the lab does not mean that arhythmicity is irrelevant to them. It is likely that in a natural environment, which is more demanding than laboratory conditions, those mice would be at a disadvantage when compared to wild-type mice. As we will discuss throughout this chapter, although bacteria are by no means “simple,” their generally streamlined and compact genomes allow for precise manipulation of clock components and more direct measurements of fitness than is usually possible for eukaryotes. It is also easier to manipulate their genetic background and environmental conditions, allowing us to better assess the relationship between trait variation and fitness—ultimately leading to a broader understanding of the selective pressures that could have resulted in the evolution of circadian clocks, as well as the past, present, and future role that they might have as adaptive daily and seasonal timekeepers.

2 General Considerations Concerning the Evolutionary Significance of Clocks Nature is inherently cyclic. The certainty of a daily sunrise and sunset and the predictable number of hours in between have accompanied life from its very beginning. It is not surprising, therefore, that life has been incredibly successful at evolving mechanisms that allow organisms to take advantage of the predictability of the cycles that surround them. This is usually achieved through some kind of timekeeping device, which often takes the form of a circadian clock, a type of timer that is even more (phylogenetically) widespread than the phrase “circadian clocks are ubiquitous” is in chronobiology papers. Circadian clocks have been described throughout the entire “Tree of Life” and have been thoroughly studied in multiple model organisms such as mice, fruit flies, the fungus Neurospora crassa, the thale-cress plant Arabidopsis thaliana, and the

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria

113

cyanobacterium Synechococcus elongatus. Despite the fact that clocks are phylogenetically ubiquitous, there seems to be very little conservation of clock gene sequences. Cyanobacteria, for example, operate on a system based on kaiA, kaiB, and kaiC, fungi have frq (frequency) and wcc (white collar), and the main clock genes in plants are toc1 (timing of CAB expression 1) and cca1 (circadian clock associated 1); none of these genes are homologs of each other (Loudon 2012). Even among animals—which have had less time to diverge—there is diversity: while both invertebrates and vertebrates have homologs of Per (Period) with similar functions (Tei et al. 1997), and Drosophila’s Cyc (Cycle) is a homolog of mammalian Bmal1 (Brain and Muscle ARNT-Like 1) (Rutila et al. 1998), the Drosophila oscillator uses TIM (encoded by Timeless) as a co-repressor with PER while that of mammals enlisted CRY (Cryptochrome) for the co-repressor function instead. Not all insect clocks are like that of Drosophila, however. Honey bees and butterflies interestingly harbor a mammalian-like Cry, suggesting that their clocks and those of mammals might have evolved from a system that had both the Drosophila version of Tim and Cry and the mammalian version of Cry (Rubin et al. 2006; Zhu et al. 2008). Despite this diversity, however, all of these clocks appear to be formally similar as they abide by the three fundamental characteristics of circadian clocks: (i) an endogenous and self-sustained circa-24 h periodicity, (ii) entrainability, and (iii) temperature compensation. Most of them also function in a rather similar fashion, either being based on transcriptional-translational feedback loops and/or on a posttranslational oscillator. The ubiquity of these characteristics despite their phylogenetic diversity is remarkable, and it makes one think that the “blind watchmaker of evolution” is quite fond of making circadian clocks. But why exactly would this be the case? The fact that circadian clocks are so widespread in the Tree of Life hints at the likelihood that they are adaptive, but if they are truly so, then that idea begs at least three questions. What could have been the selective forces that caused the evolution of circadian clocks? Why do circadian clocks appear to be more common than other types of daily timers? How can we demonstrate whether circadian clocks are truly adaptive? Those questions have inspired multiple hypotheses. Before we delve into those hypotheses, however, it is important to clarify some relevant verbiage: when we discuss possible selective forces and adaptations, we are referring to the idea that, in a given environment, there is variation in the reproductive success (fitness) among its resident species. This variation could either be the result of random chance, or it could be directly tied to particular phenotypes. In a scenario in which certain characteristics of an environment make it so that a particular phenotype has a higher chance of reproducing relative to other possible phenotypes, one can say that particular phenotype is an “adaptation.” Adaptation is also thought of as being a process, namely an ongoing genetic evolutionary change driven by natural selection that favors some phenotypes over others. However, an important thing to consider is that an adaptation is beneficial in a specific context, and in that sense, it can only be presumed with certainty to be adaptive the first time it confers an advantage. While its persistence in a population over time could indicate that it is still adaptive—i.e., there is still selective pressure that favors the trait—a feature can persist for at least

114

M. L. Jabbur et al.

two other reasons: first, it could remain because the selective pressure has relaxed and now there is neither strong selection for nor against this trait, and thus it persists passively; or second, it could by itself no longer be adaptive but in the process of evolution, it could have linked itself to other features that are themselves adaptive. Throughout this chapter, when we discuss whether clocks are adaptive, we mean specifically whether there is still a selective force that favors their presence. This distinction is important because it keeps us from wandering into the dangerous territory of equating “just-so stories” with a demonstration of fitness (Gould and Lewontin 1979). Our noses, for example, are quite efficient at holding spectacles, allowing for an excellent placement of spectacles’ bridges and nose pads. Therefore, noses are certainly advantageous for bespectacled individuals who would not be able to see properly were their noses not stable platforms for their spectacles. Yet, it would be thoroughly preposterous to say that our noses are adaptations for holding spectacles. For a more bacteria-centric example, it is undoubtedly true that circadian clocks create temporal organization within an organism. For example, the temporal separation of mutually interfering processes, such as nitrogen fixation and photosynthesis in cyanobacteria intuitively seems adaptive (Mitsui et al. 1986). However, the presence of this separation per se is not a sign that it is in any way adaptive, even when it might seem to us that it should obviously be the case. This point will become particularly important when we discuss the third question mentioned, that is, how to demonstrate if circadian clocks are indeed adaptive.

3 How and Why Did Bacteria Evolve Circadian Timekeepers? Now that the relevant terms of adaptation and fitness have been defined, let us focus on the first question proposed: what could have been the selective forces that caused the evolution of circadian clocks? Multiple theories have been proposed to answer this question, the most notable of which is probably Pittendrigh’s “Escape from Light” Hypothesis (Pittendrigh 1965, 1993), which posits that clocks might have evolved as a mechanism to protect organisms from the damaging effects of light, both visible and ultraviolet (UV). Because many important molecules in cells can absorb even visible light (e.g., cytochromes and flavins), visible light can affect growth rates and metabolism (Robertson et al. 2013). UV radiation can have particularly detrimental effects on cells, one of which is increasing the rate of mutations in their DNA. The sensitivity to UV, however, is likely to be higher at specific phases of the cell cycle, particularly when the DNA is unwound for replication. For yeast, cells in G1/S phase are the most sensitive to UV damage, while those in late S/G2 phase are the least sensitive (Siede and Friedberg 1990). Additionally, for some eukaryotic microorganisms that experience a 24 h cycle of cell division, the bulk of DNA replication happens during the night (Edmunds 1988),

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria

surviving colonies (% control)

a.

115

100 80 60 40 20 D

N

SD

0

SD

SN

12

24

SN

36

48

circadian time (h) 100

% survivors (viable cells)

b.

80

60

40

20

D

N

SD

0

SD

SN

12

24

SN

36

48

circadian time (h) Fig. 1 (a) Circadian rhythm of UV-C sensitivity of the alga Chlamydomonas reinhardtii in LD and LL (adapted from Nikaido and Johnson 2000) and (b) the protozoan Euglena gracilis in DD (adapted from Bolige et al. 2005). D refers to day, N to night, SD and SN to subjective day and subjective night, respectively. Both species show a rhythm of UV-C sensitivity that is the highest during subjective night, in accordance with the “Escape from Light” hypothesis. LD ¼ light/dark cycles, LL ¼ constant light, and DD ¼ constant darkness

which would be expected in a scenario in which a clock’s function was to temporally restrict DNA replication to a time in the daily cycle in which it is least likely to experience UV radiation. In accordance with this correlation, Chlamydomonas reinhardtii, a eukaryotic alga, shows a daily and circadian rhythm of UV sensitivity and exhibits the lowest survival when exposed to UV at the beginning of the subjective night, which coincides with its nuclear division time (Fig. 1a; Nikaido and Johnson 2000). A similar phenomenon was observed in the protozoan Euglena gracilis (Bolige et al. 2005), as depicted in Fig. 1b. In cyanobacteria, there is another line of evolutionary evidence that supports the Escape from Light Hypothesis: the cyanobacterial circadian clock—possibly the earliest extant circadian clock to have

116

M. L. Jabbur et al.

evolved—is composed of the genes kaiA, kaiB, and kaiC. The oldest of those three genes is kaiC, and it appears to have derived from the RecA/DnaB superfamily of DNA recombinases and replication helicases. Therefore, kaiC’s ancestry suggests that its initial role might have been associated with DNA damage repair, replication, and/or RNA metabolism. Congruent with those observations, Simons (2009) proposed that kaiC, which is thought to have appeared 3.5 billion years ago (Dvornyk et al. 2003), could protect DNA from UV damage by promoting DNA compaction during the day. For modern cyanobacteria, however, DNA compaction occurs primarily during the night, rather than in the day (Smith and Williams 2006). While it is possible that this phase relationship was reversed for ancient cyanobacteria, it is currently unknown if that was indeed the case.

4 Self-Sustained Versus Damped Oscillators Versus Hourglass Timers An important caveat to note, however, is that merely “escaping from light” would not, by itself, necessarily select for a circadian clock: if all that is needed for optimal phasing is to time important events to when UV—or another daily rhythmic variable—is not present, then a simple hourglass (i.e., a timer that requires a “hard restart” to initiate) would allow anticipation of a specific daily phase. The daily cycle of light and dark (esp. dawn or dusk signals) would be potentially sufficient to restart this hourglass every day, and a self-sustained oscillator would not be necessary. Curiously, however, hourglass timers appear to be less common than circadian clocks as daily timekeepers and have been more commonly observed for developmental timing and aging (Rensing et al. 2001). That brings us to the second question posed before: why are circadian clocks overwhelmingly more common than other types of daily timekeepers? The answer to this question is not entirely certain, but two explanations are often offered. First, while an hourglass needs to be reinitiated every day, an oscillator can keep time in the absence of inputs from a zeitgeber for several cycles, and it could thus maintain its approximate phasing from prior entrainment despite environmental noise in the regularity of zeitgeber signals like cloudy days, shading, or aberrant temperatures (Pittendrigh 1981a). Second, the yearly variation in day/night length that comes with the changing seasons could be a selection factor for selfsustainability. In a scenario in which day length is not constant throughout the year, one would expect that the optimal phase angle of entrainment of a critical rhythmic factor to dawn or dusk would also not be constant. An oscillator, as compared with an hourglass, can increase the “flexibility” of phasing and thus allow for more precise fine-tuning to the environment (Johnson et al. 2017; Hut and Beersma 2011). This hypothesis is supported by in silico research showing that a consistent LD12:12 cycle does not evolve self-sustained oscillations in a gene network, while changing photoperiods, and especially changing photoperiods

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria

117

coupled with environmental noise (e.g., variations in the timing of light and/or temperature) are more likely to select for circadian oscillations (Troein et al. 2009). Yet, the general lack of organisms with hourglasses instead of clocks makes it hard to test the hypothesis that photoperiods played an important role in the evolution of self-sustained clocks. There are very few examples of such organisms, which might include Saccharomyces cerevisiae (Eelderink-Chen et al. 2010), aphids (Beer et al. 2017), and surface-cave fishes like Astyanax mexicanus (Beale et al. 2013). However, luckily for microbiologists, one incredibly widespread cyanobacterium seems to possess an hourglass: Prochlorococcus marinus. While it is notoriously difficult to grow in laboratory conditions, often requiring the presence of “helper” bacteria like Alteromonas sp. (Sher et al. 2011), Prochlorococcus is an interesting group as it lacks kaiA—apparently due to a secondary loss—but still retains the remaining two thirds of the canonical cyanobacterial circadian clock, kaiB and kaiC. While it is one of the most abundant genera of cyanobacteria in the ocean—along with Synechococcus—, and exhibits a remarkable diversity of ecotypes, capable of surviving in a wide range of light intensities, temperature, and nitrate concentrations (Johnson et al. 2006), Prochlorococcus is geographically restricted to latitudes between ~40 N and ~40 S, such that its abundance decreases sharply 30 north or south from the equator (Flombaum et al. 2013). These observations are consistent with the hypothesis that an hourglass might suffice in environments with little variation in day length, but larger variations would favor a circadian clock. Future studies involving this genus could help elucidate the mechanisms through which an hourglass timer phases to its environment and whether extreme annual variations in photoperiod would indeed jeopardize the optimal synchronization of its metabolism with the environment.

5 Testing Whether Clocks Are Adaptive We have laid out above the theoretical framework that chronobiologists have used in the past to understand the evolution of circadian clocks. Let us now address the next question: specifically, can we demonstrate that circadian clocks are truly adaptive, and if so, how? As mentioned before, in order to show that something is an adaptation, we need to first gather evidence that there is selection ongoing for that particular trait—i.e., that organisms possessing this trait have a higher reproductive success than their counterparts that lack it. Thus, to begin answering this question, we must first define which characteristics of the clock are most likely to be under selection. The usual “just-so story” for the advantage derived by having circadian clocks is that they enable an organism to anticipate temporal transitions in the environment (e.g., dawn or dusk) and prepare appropriately so that key behaviors or metabolic states are precisely initiated at optimal times of the day. Selection can directly act on this timing by enhancing the reproduction of individuals whose phase relationship is favorable, and thereby indirectly selecting for features of the clock that lead to the

118

M. L. Jabbur et al.

most optimal phase in that particular environment. Underlying features of the circadian clock that are mechanistic characteristics rather than overt phenotypes, such as temperature compensation, period length, entrainment mechanisms (e.g., the shape of the Phase Response Curve {PRC}), etc., can only be indirectly selected by the environment through direct selection on the actual phenotype displayed: the phase angle of entrainment. This argument highlights the difference between hypothesizing that “temperature compensation is a feature of the clock that emerged because clocks evolved in environments that experience a lot of temperature variation throughout the day and the year,” versus hypothesizing that “temperature compensation evolved because given temperature variation, an organism that does not compensate for temperature would have a less stable phase relationship with its environment so that it fails to time itself reliably and consequently experiences decrements in reproductive success.” The first hypothesis is incompatible with the idea of clocks as an adaptation because it does not present itself in the context of differential reproductive success, whereas the latter hypothesis does. The first hypothesis would be unable to differentiate between temperature variation in the environment selecting for temperature compensation versus temperature compensation being merely a possible scenario in this particular environment that emerged as a result of random drift. In essence, to demonstrate that clocks are an adaptation, we must start with hypotheses that examine a particular phenotype (e.g., phase angle) that can be altered by changes in a characteristic of the clock (in this case, temperature compensation) whereby an array of phenotypical variants (e.g., stable phasing versus unstable phasing) are generated by the clock variants that have different reproductive success in a particular environment. In the case of temperature compensation, we can imagine that if an organism is unable to accurately compensate for temperature variations (e.g., a warm day followed by a cold day), the free-running period of its clock (FRP) is going to vary depending on the environmental temperature. Stable entrainment is achieved when the environmental zeitgeber occurs cycle after cycle at the same circadian phase and the circadian clock’s entrained period matches that of the environment’s period “T” (following the equation FRP-T ¼ phase shift (Pittendrigh 1981b)). If the FRP is changing with temperature, the zeitgeber will not strike the Phase Response Curve (PRC) reliably at the same circadian phase, and therefore the phase angle (aka phase relationship) of the clock to the environment will vary from day to day. This variability is likely to be maladaptive. For example, a rodent that comes out of its burrow at a nonoptimal time is more likely to be picked off by a predator. The logic here is that poor temperature compensation could lead to variable FRPs that are in turn responsible for unreliable phasing relative to the environment, and voilà, lower fitness! (Note that this simplified exposition concerns temperature compensation of FRP; temperature cycles can also be a zeitgeber, and therefore the linkages between clocks and temperature can be complex.) Demonstrating the adaptive value of a circadian clock is not an easy endeavor. To understand why that is the case, it is important to think about this idea from a theoretical and a historical perspective. Theoretically, to demonstrate adaptiveness,

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria

119

we must show that a particular phenotype is more “fit” than its alternatives. Fitness is, by definition, a genotype’s (or sometimes an individual’s, or a phenotype’s) reproductive success as measured by the average per capita lifetime contribution of individuals of a certain genotype to future populations (Futuyma and Kirkpatrick 2017). One may be tempted, for example, to look at growth, longevity, or developmental rate under different conditions and equate whichever phenotype grew and developed faster or lived longer as the most “fit.” It is easy to assume that faster growth and longer lives might be correlated with a better exploration of an environment and a higher likelihood of being reproductively successful. Yet, neither of those variables are necessarily directly correlated to fitness (Endler 1986). For example, “Methuselah” might have been long-lived, but had he been sterile, his fitness would have been zero. A proper measurement of fitness, therefore, requires being able to directly quantify reproductive success, but that requires analysis of multiple generations, which is ofttimes labor intensive, costly, or simply not feasible. Historically, because of the difficulties associated with measuring reproductive success, almost all of the early experimental work done to establish whether circadian clocks were adaptive focused on these indirect but specious measurements of fitness, esp. growth, longevity, and survival. Examples are summarized in Table 1, and range from measurements of growth in plants (Ketellapper 1960), lifespan in the lab for animals (Pittendrigh and Minis 1972; von Saint Paul and Aschoff 1978; Hurd and Ralph 1998) to laborious and heroic measures of survival in the field (DeCoursey and Krulas 1998; DeCoursey et al. 2000). While all these studies collectively supported the researchers’ hypotheses that circadian clocks and some of their properties are adaptive, they were not definitive because they all lacked a direct measurement of fitness. This liability was ultimately rectified by choosing organisms and approaches that allowed fitness to be directly assessed, as in the pioneering studies with the tractable cyanobacteria that will be discussed in the next section.

6 Competition Experiments and Assessment of Fitness Given the above definitions of “fitness” and “adaptation,” as well as the experimental issues raised, the question remains: how does one actually show whether clocks are adaptive? The experiments necessary to achieve such a feat require an organism that allows us to (i) look at multi-generational data and (ii) measure reproductive success, which is not necessarily equal to survival. Once an appropriate organism has been identified, we can proceed with competition experiments similar to those depicted in Fig. 2, which are the “gold standard” among population biologists for tests of adaptive fitness. Namely, we would gather groups of individuals with different clock phenotypes and subject them to a selective environment in competition with each other. For example, a temperature compensation mutant strain could be competed against a wild-type strain, with both being exposed together to an environment with a daily

120

M. L. Jabbur et al.

Table 1 A selection of publications that attempted to demonstrate whether circadian clocks are adaptive Publication Ketellapper (1960) Pittendrigh and Minis (1972) von Saint Paul and Aschoff (1978) Hurd and Ralph (1998) Klarsfeld and Rouyer (1998) Ouyang et al. (1998)

Organism Tomatoes and peanuts Fruit flies

Conditions T-cycles at different temperatures T-cycles and LL

Proxy for fitness Growth (dry matter)

Blowflies

LL versus LD

Lifespan in lab

Golden hamsters Fruit flies

LD14:10, period mutants

Lifespan in lab Lifespan in lab

Cyanobacteria

LD12:12 vs LD8:8, period mutants T-cycles, period mutants

Lifespan in lab

Frequency in the population through generations Survival in the field

DeCoursey and Krulas (1998), DeCoursey et al. (2000) Woelfle et al. (2004)

Chipmunks

Field

Cyanobacteria

Dodd et al. (2005)

Thale-cress

LD12:12 versus LL, wildtype and damped/arhythmic mutants T-cycles, period mutants

Daan et al. (2011)

Mice

Spoelstra et al. (2016)

Mice

CLOCK PHENOTYPES

Seminatural, clock mutants with impaired rhythmicity and entrainment Seminatural, period mutants

SELECTIVE ENVIRONMENT

Frequency in the population through generations Chlorophyll content, carbon fixation, growth, and mortality Survival and frequency in the population through generations Frequency in the population through generations

POSSIBLE OUTCOMES orange wins, selection likely favors orange phenotype no clear winner, either weak or no selection blue wins, selection likely favors blue phenotype

Fig. 2 Schematic showing the general protocol and interpretations for a competition experiment designed to assess the adaptive value of a certain clock feature. Two clock phenotypes are subjected together in competition to a selective environment such as different light/dark cycles, constant conditions, different temperatures, etc. In this scenario, either one of the phenotypes wins (i.e., either becomes dominant or is fixed in the population), or there is no clear winner (i.e., the proportions of the two phenotypes in the population remain at about the same composition as at the beginning of the competition). See text for interpretations of the possible outcomes

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria

121

temperature cycle. Ideally, in this environment (or set of environments), the differences between the phenotypes result in differential reproductive success so that one phenotype eventually out competes the other in the population. If that is the case, we can conclude that there is likely selection for that phenotype and that its underlying basis—in this case, temperature compensation—is an adaptation (That is, if the mutant loses!). It is worth mentioning, however, that a phenotype can disappear from a population simply because of random drift, that is, chance events that can lead to differential reproductive success and a consequent loss of variation. For phenotypes based on genetic differences, in any given population—provided enough time—one allele is bound to become fixed regardless of whether there is the selection for it or not. In the case of no selection and merely random drift, the probability that an allele will be fixed in a given population is equal to its starting frequency, which in our conceptual experiment would be 50% (Futuyma and Kirkpatrick 2017). This means that, in the setup depicted in Fig. 2, there is a considerable chance that, for each of our experimental populations, either of our “winning” scenarios could be a false positive. This is a troubling observation, as it implies that in order to conclude anything about these experiments, we would need a rather high number of replicates. The key to counteracting the random drift problem lies in two properties of this process: (i) for any population, an allele is bound to be fixed given enough time, and (ii) the time it takes for an allele to become fixed is dependent on the population size. Operationally, multiple replicates of sufficiently large populations are needed to confirm that the winners in the competition are out-competing the losers due to a selective advantage and not due to drift. Taking into consideration the constraints discussed, cyanobacteria presented themselves as a promising candidate organism to address our question, and the first application of the competition assay for a potential fitness advantage of circadian clocks was reported in 1998 when Ouyang and colleagues used the cyanobacterium Synechococcus elongatus to test the hypothesis that consonance of the circadian period of the clock with the period of the environmental cycle is adaptive. Cyanobacteria fill all the requirements previously described: a short generation time, ease of genetic quantification (i.e., it is relatively easy to infer the percentage of an allele at any given time in a population), low cost (enables multiple replicates), and large populations so that random drift is unlikely. Those characteristics allowed for the first demonstration of the adaptive value of clocks that actually measured reproductive success as opposed to survival, growth, or another indirect measurement of fitness. Ouyang and co-workers employed a classic competition protocol similar to what had been previously used in studies such as those of Richard Lenski, who competed strains of Escherichia coli to assess fitness in his long-term experimental evolution experiments (see Lenski 1988; Lenski et al. 1991; Lenski and Travisano 1994 for some early examples). In the protocol of Ouyang and colleagues, two competing strains in axenic cultures are mixed in a 1:1 ratio. They are then grown together for several generations (Ouyang et al. 1998). Snapshots of the population composition are taken by plating the mixed cultures and assessing the phenotypes of the resulting

122

M. L. Jabbur et al.

Fig. 3 (a) A simplified experimental design for the competition experiments conducted by Ouyang et al. (1998) using different strains of the cyanobacterium Synechococcus elongatus. All strains grew at the same rate in pure-strain cultures, and this fitness test was performed in mixed cultures. Wild-type and period mutants were grown separately and then combined in a single tube to create mixed populations, with each strain starting at about 50% of the population. The mixed cultures were grown in either LD11:11 (T ¼ 22) or LD15:15 (T ¼ 30) and every few days a sample of the population was collected and plated. Once colonies grew on the plate, the clock phenotype of each colony was identified based on the colony’s bioluminescence profile and quantified to ascertain the population structure as the proportion of each strain. The cultures were diluted 10 approximately every week. (b) Results for the competitions between wild-type and two period-mutant strains. At the beginning of the competition, the cyanobacteria population was composed of approximately 50% of each strain. By day 27, however, for all competitions and LD cycles analyzed, the strains with periods closest to that of the environmental period outcompeted the other strain in each competition

colonies and what percentage of the population each phenotype now represents. Because the competition protocol assesses population composition over multiple generations, there is a direct proxy of reproductive success, as the proportion that a strain represents in a population is the result of the successful reproduction and survival of the offspring of that strain’s initial cells. For the cyanobacterial competition tests performed in our laboratory, the protocol is illustrated in Fig. 3a. All strains that we tested grew at the same rate in axenic cultures, and differential fitness was tested in cultures of two strains mixed together.

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria

123

We competed a wild-type strain (with FRP  25 h) interchangeably with two periodmutant strains: C22a (FRP  23 h) and C28a (FRP  30 h), both of which harbor point missense mutations in the kaiC gene. The goal of the experiment was to test whether the circadian period was adaptive, namely whether having a period whose value closely matches that of the environmental cycle is adaptive. This was achieved by competing the strains in different T-cycles whose frequency would resonate (i.e., have a closer period to) with the period of one strain’s FRP more so than the periods of other strains. To quantify each strain’s proportions in the mixed populations over time, either different antibiotic resistances were used, or the samples were plated, and the resulting colonies were assessed for their free-running period (Fig. 3a). We discovered that the winning strain was the one whose endogenous FRP was closest to the environmental period (¼ T), supporting the idea that there is the selection for this characteristic of the circadian clock (i.e., having a period close to the 24 h cycle of the day), and pointing towards its role as a likely adaptation (Fig. 3b). Our conclusion was that the circadian pacemaker in cyanobacteria confers a significant competitive advantage in light/dark cycles when the FRP of the clock is consonant with the environmental cycle so as to achieve an optimal phase relationship between the environmental cycle and the internal timekeeper (Ouyang et al. 1998). (Note that a FRP that is very close to 24 h may not always be the most adaptive for every species or situation. For example, the FRP of Neurospora in DD is about 21 h; this implies again that it is phase relationship and not FRP per se that is the target of selection, and that a short FRP and highly asymmetric PRC yields the most adaptive phase relationship for Neurospora.) The results of the competition experiment in cyanobacteria were particularly persuasive because mutant strains could outcompete wild-type strains; if wild-type always won, then it might be possible that the mutations hampered some non-clock process (like the case of the Bmal1-ko mice mentioned above). However, because mutant strains defeated wild-type in T-cycles whose period matched that of the mutant strains, the obvious conclusion was that this is a clock-specific phenomenon. Subsequent work showed that wild-type strains outcompete arhythmic or damped strains in LD conditions but not in the non-selective condition of continuous light (LL), suggesting that despite the clock’s self-sufficient internal temporal order, that particular ability does not increase a strain’s fitness in constant conditions (Woelfle et al. 2004). Therefore, the clock system in cyanobacteria is only beneficial in a rhythmic environment. The wild-type versus arhythmic mutant experiment was later extended to elucidate the mechanisms behind the enhancement of fitness conferred by the clock. Possible models tested were the “limiting resource model,” the “cell-tocell communication model,” and the “diffusible inhibitor model.” However, it is still unclear which of those mechanisms is the one responsible for the observed effects (Ma et al. 2013). However, an interesting mechanistic candidate was described in 2016 by Lambert and collaborators, who demonstrated that mismatches between the clock and the environment can lead to “catastrophic growth arrest”—i.e., growth that is interrupted even after at least 36 h of “recovery” in constant light—if darkness coincides with the cell’s subjective morning. Those authors concluded that the clock in cyanobacteria mediates a trade-off between growth and starvation tolerance in cycling environments such that the circadian system is only advantageous in

124

M. L. Jabbur et al.

regularly cycling environments with period lengths (T) sufficiently close to the circadian frequency (FRP) (Lambert et al. 2016). These results encouraged similar attempts in higher organisms such as the thalecress Arabidopsis (Dodd et al. 2005) and mice (Daan et al. 2011; Spoelstra et al. 2016). For Arabidopsis, it was observed that plants that matched their circadian period to the environmental period (e.g., long-period plants in a long T-cycle) had a higher ability to fix carbon, and both grew and survived better than plants whose circadian periods were “less appropriate” matches. However, in these experiments— and in plant experiments in general—it is not simple to look at multi-generational data, as the generation time is usually within the weeks or months range. Similar constraints are seen in the mice experiments, in which mutant and wild-type mice were competed against each other in seminatural light/dark conditions for 1–2 years. These experiments found parallel results to the cyanobacteria and plant work, namely that wild-type mice had higher fitness than clock mutants (whose mutations either affected their circadian period or impaired their rhythmicity and entrainment). It is also important to highlight that the mouse work had the advantage of being performed in seminatural conditions—which, hopefully, should be less “cushy” than lab conditions and allow for both higher intensity of selection and a more accurate estimate of what would actually happen in nature (Spoelstra et al. 2016). However, they do come at a cost: it is impossible to test different T-cycles in a natural environment, and thus there is no way to demonstrate that the mutant mice would have a higher fitness in an environment that resonates better with their clock, which would serve as a proof point that the results observed are caused by a mismatched clock rather than by another confounding factor. These criticisms are not to say, however, that bacteria are the only organisms in which the adaptive value of the clock can be properly measured. Bacteria also have shortcomings, such as the lack of sexual reproduction (in most cases), and the fact that it is not trivial to reproduce their natural conditions in the lab environment (e.g., variables like the high light intensity and spectrum of sunlight, rainfall, etc.) or raise them in seminatural or natural conditions without the risk of contamination. However, together with other model organisms, they help us build a broader and more concrete basis of our understanding of the evolution of circadian clocks.

7 “It Takes a Village:” Communities and Populations Twenty-two years after the first successful test of the fitness of a clock phenotype in Synechococcus elongatus, cyanobacteria remain a very useful tool for assessing whether a clock component is an adaptation to a certain environmental condition or not. The ease with which they can be genetically modified, coupled with the large number of characterized clock mutants, makes them a remarkable model organism for this type of experiment. However, there are limits to what has been accomplished, one of which is the fact that using Synechococcus alone limits the analysis to intraspecific interactions, in the context of a selective external environment which they are generally incapable of modifying. But, in a natural environment, species are

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria

125

not only competing against their conspecifics, but also against other species. Beneficial and/or competitive relationships can also be formed among bacteria inter- and intraspecifically, and it is possible that temporal mutualism—i.e., bacteria within a community exchanging beneficial metabolites in a rhythmic fashion—could play a role in the evolution of their timing systems (Sartor et al. 2019). In addition, for nonfree-living bacteria, the selecting environment is not unchangeable; the bacteria can affect the behavior of their host or of the bacteria that surround it and change the dynamics of the rhythmic environment to which they are subjected (This might arguably be also true for bacteria in some prey-predator relationships). Static snapshots of these types of interactions have already been described for higher taxa, and several examples of the role of the circadian clock in them are available: both the pollinator’s and the plant’s clocks play a role in the flower seeking behavior of tobacco hornworms and petunias (Fenske et al. 2018). Also, in the case of malaria, the host clock is important in determining the in-host replication of the parasites (O’Donnell et al. 2011); it was recently demonstrated that the malaria parasite itself has a circadian clock and that malaria infection can disrupt the circadian rhythm of its murine host (Rijo-Ferreira et al. 2020; Prior et al. 2019). Yet another example is green toad tadpoles that are preyed upon by dragonfly larvae and show a stronger response to the chemical cue produced by the larvae during the day—when the larvae are active—than during the night (Fraker 2008). Yet, it is hard to expand from those snapshots into a more encompassing panorama of the role of clocks in interspecific interactions, at least for the species mentioned. Hopefully, the principles learned from cyanobacteria can be applied to other bacteria and/or unicellular eukaryotes and further expand our understanding of clock evolution. Ultimately, we also hope that the cyanobacterial competition studies will guide future experiments involving higher organisms by defining the critical experiments to perform. Studies that employ multiple species, in both antagonistic and synergistic relationships, can shed light on how clocks have evolved and how they could influence the evolution of particular groups. One way of looking at this is through the perspective of the Red Queen Hypothesis (Van Valen 1973), which is an important tenet in evolutionary biology that proposes species must constantly evolve in order to survive the continuous “arms race” between themselves and the other species with which they interact. From this point of view, we could imagine that the evolutionary arms race between a predator and its prey, both of which have circadian clocks, could lead to alternating cycles as the predator tries to be active at the same time as the prey, and the prey, in turn, tries to avoid the predator’s active phase. An example of this coupling could be certain eukaryotic predators of cyanobacteria such as Paramecium or Amoeba, which may have circadian rhythms in their mobility and thus create a temporal selection factor on their prey (Hasegawa et al. 1984; Hasegawa and Tanakadate 1984; Simkovsky et al. 2012, 2016). On the other hand, it is possible that interactions among species could lead to the establishment of networks linking multiple bacterial communities (Liu et al. 2016) so that metabolic outputs/inputs could be rhythmically exchanged and interchanged from one to the other, for example, altering the cyclic properties of a secreted output metabolite that in turn modulates the rhythmicity of a different bacterium that uses the metabolite as an input.

126

M. L. Jabbur et al.

8 A Medically Important Community: The Mammalian Gut Microbiome Lastly, one interesting system that could provide some insights into host-symbiont interactions and interspecific competitions and networks is the gut microbiome. Many animals show rhythmic feeding as well as body temperature cycles that are controlled by their circadian clock. This means that the gut is subjected to a daily rhythm of temperature and nutrient availability that could act as a selective force on the organisms that inhabit it (Johnson et al. 2017). If a species could anticipate the time of feeding, for example, it might be more prepared to capture nutrients upon their arrival and have a higher growth rate (“reproductive success” for a microbe) than other species that lack this ability to anticipate. This would make the gut an interesting environment to search for rhythmic bacteria, and in fact, one enteric bacteria has been identified that apparently displays circadian rhythmicity (Paulose et al. 2016, and see Chapter “Circadian Organization of the Gut Commensal Bacterium Klebsiella aerogenes”). At the same time, it is possible that, for many species, foraging in the wild may occur in bouts throughout their active phase— which are likely to vary from day to day—and consequently selection within the microbiome might act to promote “preparation” to a broad interval of activity in general as opposed to a specific phase of the daily cycle. If so, a simpler type of timing system might be adequate. Interestingly, the gut microbiome can be quite resistant to colonization by conspecifics—at least in the case of Bacteroides, which is one of the predominant genera in mammals (Eckburg et al. 2005)—implying that there can be strong microbiotal interspecific competition (Lee et al. 2013). A strongly competitive environment could favor an adaptive feature that would otherwise be too costly in a more forgiving environment. However, interactions between species within the microbiome are often context-dependent and dynamic, so the intensity of selection for any given species within this environment is not easily determined (Coyte and Rakoff-Nahoum 2019). As discussed in other chapters of this book, the composition of the gut microbiome has been reported to oscillate in LD12:12 (Thaiss et al. 2014; Zarrinpar et al. 2014; Liang et al. 2015), and up to 17% of the OTUs (operational taxonomic units) cycle in LD conditions with food ad libitum (Zarrinpar et al. 2014). Moreover, disruption of the host’s circadian clock either by clock gene mutations or by constantly reversing the LD cycle have been shown to affect the gut microbiome composition (Voigt et al. 2014, 2016; Deaver et al. 2018); interestingly, this is associated with intestinal barrier dysfunctions (Summa et al. 2013; Deaver et al. 2018). Taken together, these observations suggest that the gut microbiome could be a fruitful environment for studying selective pressures related to circadian clocks, and our lab is currently attempting to understand how the rhythmic environment of the host’s alimentary canal could act as a selection factor on its microbiome. To do this, we are using wild-type and clock-mutant mice in light/dark (LD) and dim red light (RR) conditions. Both strains of mice have daily rhythms of feeding in LD, but only the wild-type mice feed rhythmically in RR, while the pattern of the feeding of

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria

127

the clock-mutant mice is highly disrupted in RR (Fig. 4a). Therefore, from a feedingcycle perspective, both wild-type and mutant mice present to the microbiome an equivalent and rhythmic selective environment in LD (and wild-type mice in RR), but in the mutant mice in RR, that selection is relaxed—or, possibly, turns into

a.

WT

mutant

b. day 183

day 239

PCoA1

PCoA1

Antibiotics

PCoA2

mutant

day 7

Antibiotics

WT

PCoA2

day -14

PCoA1

PCoA1

Fig. 4 (a) Diagram showing the general experimental design of the microbiome recovery from antibiotics. In LD conditions, wild-type (blue brains) and clock-mutant mice (orange brains) have rhythms of activity and feeding that peak during the night. However, in constant dim red light (RR), only wild-type behaves and feeds rhythmically. To assess the effect of rhythmic versus arhythmic feeding on the gut microbiome, we looked at the microbiome composition of wild-type and clockmutant mice during LD and RR after perturbing the microbiome with an antibiotic cocktail. (b) Principal component analysis (PCoA1 vs. PCoA2) of the gut microbiome communities showed that for both wild-type and mutant mice, the antibiotic treatment caused a stark shift in community, but only the wild-type mice were able to recover (partially in RR and almost completely once in LD), while the mutant mice remained with a distinct microbiome composition

128

M. L. Jabbur et al.

selection for bacteria that abhor rhythmic feeding. We subjected the mice to multiple weeks of LD12:12 followed by transfer to RR and at the same time to a 5-day antibiotic treatment to disrupt the gut microbiome. After several months in RR, all mice were returned to LD for several more months. Analysis of the changes in gut microbiome composition through time can give us candidate species or groups that respond to rhythmic food availability and thus possibly possess a daily timekeeper to anticipate those rhythms. Moreover, this experiment might yield insight towards how the microbiome as a whole functional unit is affected by the presence or absence of rhythmicity. So far, we have observed that antibiotic treatment drastically changes the microbiome diversity of both wild-type and clock-mutant mice, but that in the rhythmic wild-type, the microbiome can recover its initial diversity after several months, while the clock mutant fails to completely recover its initial diversity (Fig. 4b). Which species are driving these differences and do they have clocks? Time will tell.

9 Clocks Are Still Evolving! The process of adaptation/evolution allows an organism’s clock system to change so that it appropriately times itself to its environment, which is usually assumed to be the stable element in this relationship. But what happens when the environment itself changes drastically? This book chapter was written in 2020, and in a not so distant past, those who are today’s scientists were taught the distinction between climate and weather: “climate is what you expect, weather is what actually happens.” We were taught that “climate” spans multiple decades and summarizes averages and patterns, while “weather” is the more subjective day-to-day variation. That is to say, while this winter might be colder than last year’s, the world on average is still getting warmer. Today, that distinction is a bit less relevant than a few decades ago; observing climate change no longer requires painstaking record-keeping of weather over decades, but rather its fingerprint can now be experienced globally at the level of a single year, month, or even day (Sippel et al. 2020). The world is getting warmer, and even non-scientists are aware of it. For example, a few months before this chapter was written, Australia experienced its hottest day on record on December 16, 2019, with an average of 40.9  C across the entire country. A day later, on the 17th, that record was broken with a new average of 41.9  C (Australian Government Bureau of Meteorology 2020). How warm the world will actually get is still uncertain, as this will largely depend on current and future actions and how well we curb emissions of greenhouse gases. The current models suggested by the latest IPCC report (2014) work with scenarios (known as RCPs—Representative Concentration Pathways) that predict an average increase anywhere between 0.9–1.4  C by 2050 (RCP 2.6) and 1.0–3.7  C by 2100 (RCP 8.5), in comparison to the temperatures registered in the 1986–2005 period (RCP 2.6 and RCP 8.5 as mentioned in Fig. 6). However, temperatures have already risen (Fig. 5), and the effects of the current and future increases are many and broad;

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria

129

2014

1982 50

Feb

0 −50

50

May

Latitude

0 −50

Sea Surface Temperature

30 20 10

50

0 Aug

0 −50

50

Nov

0 −50 0

100

−160

−60

0

100

−160

−60

Longitude Fig. 5 Average sea surface temperature for different months in 1982 and 2014 derived from data provided by the NOAA/OAR/ESRL PSL, Boulder, Colorado, USA, OISST Version 2 (Reynolds et al. 2007). While there is a yearly variation, sea surface temperature is higher in 2014 than in 1982, particularly in the tropics and subtropics. The white dashed squares show areas in which there are the most noticeable differences between 2014 and 1982

with climate change, the frequency of droughts and floods increases, global warming leads to the melting of polar caps that raise sea levels, and the high levels of carbon dioxide acidify our oceans. These consequences will certainly change the spatial and temporal niches of myriad species, forcing them to adapt or become extinct. Temporal mismatches are a key idea when it comes to understanding the effects of climate change in phenology and chronobiology; the world is getting warmer, but the Earth is still rotating on its axis and orbiting the sun at essentially the same rates. Photoperiod is a very reliable predictor of future temperature—given the Earth’s thermal capacity, changes in photoperiod always anticipate changes in temperature. Day length-to-temperature relationships form ellipses as illustrated in Fig. 6, for

130

M. L. Jabbur et al.

a. Sea surface temperature

1982

2014 Latitude 15.5° 25.5° 35.5° 45.5° 55.5° 65.5°

20

10

0 5

10

15

20 5 Day length

10

15

20

Sea surface temperature

b. 17.5 year 15.0

1982 2014 2100 on RCP2.6 2100 on RCP8.5

12.5 10.0 7.5 10

12 Day length

14

Fig. 6 (a) The relationship between day length and average sea surface temperature for multiple latitudes (north of the equator) can be described as an ellipse in which changes in day length anticipate changes in temperature. The months of the year process counter-clockwise as indicated. The blue horizontal line indicates a temperature of ~17  C for latitude 45.5 , while the blue vertical line indicates the day length in which that temperature occurs (based on Hut et al. (2013) but plotting sea surface temperature instead of air temperature; the data on temperature comes from data provided by the NOAA/OAR/ESRL PSL, Boulder, Colorado, USA, OISST Version 2 (Reynolds et al. 2007); day length reflects that which happened at the middle of the month plotted, derived using the NOAA Solar Calculator). (b) Focusing on latitude 45.5 , we can see that the 2014 ellipse has slightly shifted up in relationship to 1982. Also plotted are model ellipses showing how the sea surface temperature versus day length relationship would change assuming the average predicted increase in sea surface temperature from RCP (Representative Concentration Pathways) 2.6 and 8.5 (IPCC 2019), two of the models suggested in the latest IPCC report (2014)

ocean surface temperature, and they highlight that the day length gets shorter before temperature drops and longer before it increases (in a hysteretic rather than a linear relationship). Because temperature has such a dramatic impact on the environment, changing thermal stress, metabolism, prey/predator relationships, rainfall, etc., being able to anticipate those changes and prepare for them in advance can be advantageous. Therefore, it is not surprising that many organisms have evolved mechanisms

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria

131

to exploit the predictability of photoperiodic changes. But if the relationship between photoperiod and temperature were to suddenly change, e.g., by shifting the day length-to-temperature association from the 1982 ellipse to that of 2014 (Fig. 6a), rigidly interpreted photoperiodic cues become less useful and potentially misleading. Imagine, for example, that a species has its peak reproduction at about 17  C (indicated by the blue horizontal line in Fig. 6a): at a latitude of 45.5 , that temperature was reached at a day length slightly under 15 h per day (i.e., late summer, in August) in 1982, while in 2014 it was reached when the days were slightly longer than 15 h (i.e., earlier in the summer, in July). If photoperiod is the cue that determines when that species will prepare for reproduction, this change will create a temporal mismatch in which the environment reaches 17  C long before the photoperiodic cue to initiate preparations for reproduction is triggered. This can influence fitness, as now the timing of reproduction no longer occurs at the optimal temperature. While most of the reports so far about the effects of temporal mismatch have focused on macroscopic eukaryotic examples, such as a change in vegetation and caterpillar phenology (Visser et al. 1998) or in bird migration (Mayor et al. 2017), bacteria are also likely to be affected by this phenomenon. Our lab has recently discovered that the cyanobacterium Synechococcus elongatus shows a clockdependent photoperiodic response to cold; cyanobacteria exposed to short days survive cold treatments much better than those exposed to long days. While we have observed this phenomenon in a freshwater lake strain, it is possible that oceanic strains also show a similar response. Both lake and oceanic strains have wide latitudinal distributions and are consequently exposed to a large range of photoperiods throughout the year. Therefore, they would also be subjected to the changes highlighted in Fig. 6, and it is likely that they would have to adapt in order to match the fast-changing hysteretic association between photoperiod and temperature. These factors suggest that cyanobacteria might be an excellent model organism to understand the effects of climate change on circadian clocks. Undeniably, for many kinds of experiments, examining cyanobacteria is a much easier task than studying macroscopic eukaryotes. Samples taken a year or even weeks apart can be separated by multiple generations and have a higher chance of accumulating genetic or other changes. At the same time, their short generation time means that the effects of increases in temperature and mismatches between photoperiod and temperature cycles could be experimentally tested in the lab at a much faster rate than that of climate change in the environment. While many bacterial species have a much faster generation rate than most of the eukaryotes that we often consider when studying climate change, bacterial clocks could still serve as an experimentally tractable model of the possible pressures and responses that eukaryotic clocks might experience when faced with temporal mismatches, especially if we normalize evolution rate to generation time. In addition to their potential usefulness in modeling the responses of eukaryotic organisms, bacteria are themselves crucially significant in terms of climate change because they play a pivotal role in maintaining oceans and other bodies of water, as well as directly affecting other organisms via symbiosis.

132

M. L. Jabbur et al.

Our environment is transmuting, and it will continue to do so at a swift rate. While we biologists cannot stop this transformation from happening, we should monitor it and attempt to predict how these changes will affect important biological phenomena such as photoperiodism and daily cycles and how this will in turn influence species’ chances of survival. Acknowledgments We dedicate this chapter to the memory of Dr. David McCauley, who generously advised and assisted us in our early analyses and experiments on adaptive fitness tests in cyanobacteria. Research on cyanobacteria in our laboratory has primarily been supported by grants from the National Institutes of Health (NIGMS GM067152 and GM107434). Finally, we thank other members of our laboratory and our collaborators for helpful discussions and emotional support.

References Australian Government Bureau of Meteorology (2020) Special Climate Statement 73 – extreme heat and fire weather in December 2019 and January 2020, Australia Beale A, Guibal C, Tamai TK et al (2013) Circadian rhythms in Mexican blind cavefish Astyanax mexicanus in the lab and in the field. Nat Commun 4(1):1–10 Beer K, Joschinski J, Sastre A et al (2017) A damping circadian clock drives weak oscillations in metabolism and locomotor activity of aphids (Acyrthosiphon pisum). Sci Rep 7(1):1–9 Bolige A, Kiyota M, Goto K (2005) Circadian rhythms of resistance to UV-C and UV-B radiation in Euglena as related to ‘escape from light’ and ‘resistance to light’. J Photochem Photobiol B Biol 81(1):43–54 Coyte KZ, Rakoff-Nahoum S (2019) Understanding competition and cooperation within the mammalian gut microbiome. Curr Biol 29(11):R538–R544 Daan S, Spoelstra K, Albrecht U et al (2011) Lab mice in the field: unorthodox daily activity and effects of a dysfunctional circadian clock allele. J Biol Rhythm 26(2):118–129 Deaver JA, Eum SY, Toborek M (2018) Circadian disruption changes gut microbiome taxa and functional gene composition. Front Microbiol 9:737 DeCoursey PJ, Krulas JR (1998) Behavior of SCN-lesioned chipmunks in natural habitat: a pilot study. J Biol Rhythm 13(3):229–244 DeCoursey PJ, Walker JK, Smith SA (2000) A circadian pacemaker in free-living chipmunks: essential for survival? J Comp Physiol A 186(2):169–180 Dodd AN, Salathia N, Hall A et al (2005) Plant circadian clocks increase photosynthesis, growth, survival, and competitive advantage. Science 309(5734):630–633 Dvornyk V, Vinogradova O, Nevo E (2003) Origin and evolution of circadian clock genes in prokaryotes. Proc Natl Acad Sci 100(5):2495–2500 Eckburg PB, Bik EM, Bernstein CN et al (2005) Diversity of the human intestinal microbial flora. Science 308(5728):1635–1638 Edmunds LN Jr (1988) Cellular and molecular bases of biological clocks: models and mechanisms for circadian timekeeping. Springer, New York Eelderink-Chen Z, Mazzotta G, Sturre M et al (2010) A circadian clock in Saccharomyces cerevisiae. Proc Natl Acad Sci 107(5):2043–2047 Endler JA (1986) Natural selection in the wild. Princeton University Press, Princeton Fenske MP, Nguyen LP, Horn EK et al (2018) Circadian clocks of both plants and pollinators influence flower seeking behavior of the pollinator hawkmoth Manduca sexta. Sci Rep 8 (1):1–13

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria

133

Flombaum P, Gallegos JL, Gordillo RA et al (2013) Present and future global distributions of the marine Cyanobacteria Prochlorococcus and Synechococcus. Proc Natl Acad Sci 110 (24):9824–9829 Fraker ME (2008) The influence of the circadian rhythm of green frog (Rana clamitans) tadpoles on their antipredator behavior and the strength of the nonlethal effects of predators. Am Nat 171 (4):545–552 Futuyma D, Kirkpatrick M (2017) Genetic drift: evolution at random. In: Evolution: Futuyma D and Kirkpatrick M evolution, 4th edn. Sinauer, Sunderland, pp 165–189 Gould SJ, Lewontin RC (1979) The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme. Proc R Soc Lond Ser B Biol Sci 205(1161):581–598 Hasegawa K, Tanakadate A (1984) Circadian rhythm of locomotor behavior in a population of Paramecium multimicronucleatum: its characteristics as derived from circadian changes in the swimming speeds and the frequencies of avoiding response among individual cells. Photochem Photobiol 40:105–112 Hasegawa K, Katakura T, Tanakadate A (1984) Circadian rhythm in the locomotor behavior in a population of Paramecium multimicronucleatum. Biol Rhythm Res 15(1):45–56 Hurd MW, Ralph MR (1998) The significance of circadian organization for longevity in the golden hamster. J Biol Rhythm 13(5):430–436 Hut RA, Beersma DG (2011) Evolution of time-keeping mechanisms: early emergence and adaptation to photoperiod. Philos Trans R Soc B Biol Sci 366(1574):2141–2154 Hut RA, Paolucci S, Dor R et al (2013) Latitudinal clines: an evolutionary view on biological rhythms. Proc R Soc B Biol Sci 280(1765):20130433 IPCC (2019) Summary for Policymakers. In Pörtner H-O, Roberts DC, Masson-Delmotte V, et al (eds) IPCC special report on the ocean and cryosphere in a changing climate Johnson CH, Zhao C, Xu Y et al (2017) Timing the day: what makes bacterial clocks tick? Nat Rev Microbiol 15(4):232 Johnson ZI, Zinser ER, Coe A et al (2006) Niche partitioning among Prochlorococcus ecotypes along ocean-scale environmental gradients. Science 311(5768):1737–1740 Ketellapper HJ (1960) Interaction of endogenous and environmental periods in plant growth. Plant Physiol 35(2):238 Klarsfeld A, Rouyer F (1998) Effects of circadian mutations and LD periodicity on the life span of Drosophila melanogaster. J Biol Rhythm 13(6):471–478 Kondratov RV, Kondratova AA, Gorbacheva VY et al (2006) Early aging and age-related pathologies in mice deficient in BMAL1, the core component of the circadian clock. Genes Dev 20 (14):1868–1873 Lambert G, Chew J, Rust MJ (2016) Costs of clock-environment misalignment in individual cyanobacterial cells. Biophys J 111(4):883–891 Lee SM, Donaldson GP, Mikulski Z et al (2013) Bacterial colonization factors control specificity and stability of the gut microbiota. Nature 501(7467):426–429 Lenski RE (1988) Experimental studies of pleiotropy and epistasis in Escherichia coli. I. Variation in competitive fitness among mutants resistant to virus T4. Evolution 42(3):425–432 Lenski RE, Travisano M (1994) Dynamics of adaptation and diversification: a 10,000-generation experiment with bacterial populations. Proc Natl Acad Sci 91(15):6808–6814 Lenski RE, Rose MR, Simpson SC et al (1991) Long-term experimental evolution in Escherichia coli. I. Adaptation and divergence during 2,000 generations. Am Nat 138(6):1315–1341 Liang X, Bushman FD, FitzGerald GA (2015) Rhythmicity of the intestinal microbiota is regulated by gender and the host circadian clock. Proc Natl Acad Sci 112(33):10479–10484 Liu W, Røder HL, Madsen JS et al (2016) Interspecific bacterial interactions are reflected in multispecies biofilm spatial organization. Front Microbiol 7:1366 Loudon AS (2012) Circadian biology: a 2.5 billion year old clock. Curr Biol 22(14):R570–R571 Ma P, Woelfle MA, Johnson CH (2013) An evolutionary fitness enhancement conferred by the circadian system in cyanobacteria. Chaos, Solitons Fractals 50:65–74

134

M. L. Jabbur et al.

Mayor SJ, Guralnick RP, Tingley MW et al (2017) Increasing phenological asynchrony between spring green-up and arrival of migratory birds. Sci Rep 7(1):1–10 Mitsui A, Kumazawa S, Takahashi A et al (1986) Strategy by which nitrogen-fixing unicellular cyanobacteria grow photoautotrophically. Nature 323(6090):720–722 Nikaido SS, Johnson CH (2000) Daily and circadian variation in survival from ultraviolet radiation in Chlamydomonas reinhardtii. Photochem Photobiol 71(6):758–765 O’Donnell AJ, Schneider P, McWatters HG et al (2011) Fitness costs of disrupting circadian rhythms in malaria parasites. Proc R Soc B Biol Sci 278(1717):2429–2436 Ouyang Y, Andersson CR, Kondo T et al (1998) Resonating circadian clocks enhance fitness in cyanobacteria. Proc Natl Acad Sci 95(15):8660–8664 Paulose JK, Wright JM, Patel AG et al (2016) Human gut bacteria are sensitive to melatonin and express endogenous circadian rhythmicity. PLoS One 11(1):e0146643 Pittendrigh C S (1965) Biological clocks: The functions, ancient and modern, of circadian oscillations. In Science and the sixties, proceedings of the cloudcraft symposium, air force office of scientific research, pp 96–111 Pittendrigh CS (1981a) Circadian systems: general perspective. In: Aschoff J (ed) Handbook of Behavioral Neurobiology, vol 4: Biological rhythms. Springer, Boston, pp 57–80 Pittendrigh CS (1981b) Circadian systems: entrainment. In: Aschoff J (ed) Handbook of behavioral neurobiology, vol 4: Biological rhythms. Springer, Boston, pp 57–80 Pittendrigh CS (1993) Temporal organization: reflections of a Darwinian clock-watcher. Annu Rev Physiol 55(1):17–54 Pittendrigh CS, Minis DH (1972) Circadian systems: longevity as a function of circadian resonance in Drosophila melanogaster. Proc Natl Acad Sci 69(6):1537–1539 Prior KF, O’Donnell AJ, Rund SS et al (2019) Host circadian rhythms are disrupted during malaria infection in parasite genotype-specific manners. Sci Rep 9(1):1–12 Rensing L, Meyer-Grahle U, Ruoff P (2001) Biological timing and the clock metaphor: oscillatory and hourglass mechanisms. Chronobiol Int 18(3):329–369 Reynolds RW, Smith TM, Liu C, Chelton DB et al (2007) Daily high-resolution-blended analyses for sea surface temperature. J Clim 20:5473–5496 Rijo-Ferreira F, Acosta-Rodriguez VA, Abel JH et al (2020) The malaria parasite has an intrinsic clock. Science 368(6492):746–753 Robertson JB, Davis CR, Johnson CH (2013) Visible light alters yeast metabolic rhythms by inhibiting respiration. Proc Natl Acad Sci 110(52):21130–21135 Rubin EB, Shemesh Y, Cohen M et al (2006) Molecular and phylogenetic analyses reveal mammalian-like clockwork in the honey bee (Apis mellifera) and shed new light on the molecular evolution of the circadian clock. Genome Res 16(11):1352–1365 Rutila JE, Suri V, Le M et al (1998) CYCLE is a second bHLH-PAS clock protein essential for circadian rhythmicity and transcription of Drosophila period and timeless. Cell 93(5):805–814 Sartor F, Eelderink-Chen Z, Aronson B et al (2019) Are there circadian clocks in non-photosynthetic bacteria? Biology 8(2):41 Sher D, Thompson JW, Kashtan N et al (2011) Response of Prochlorococcus ecotypes to co-culture with diverse marine bacteria. ISME J 5(7):1125–1132 Siede W, Friedberg EC (1990) Influence of DNA repair deficiencies on the UV sensitivity of yeast cells in different cell cycle stages. Mutat Res Lett 245(4):287–292 Simkovsky R, Daniels EF, Tang K et al (2012) Impairment of O-antigen production confers resistance to grazing in a model amoeba–cyanobacterium predator–prey system. Proc Natl Acad Sci 109(41):16678–16683 Simkovsky R, Effner EE, Iglesias-Sánchez MJ et al (2016) Mutations in novel lipopolysaccharide biogenesis genes confer resistance to amoebal grazing in Synechococcus elongatus. Appl Environ Microbiol 82(9):2738–2750 Simons MJ (2009) The evolution of the cyanobacterial posttranslational clock from a primitive “phoscillator”. J Biol Rhythm 24(3):175–182

Insights into the Evolution of Circadian Clocks Gleaned from Bacteria

135

Sippel S, Meinshausen N, Fischer EM et al (2020) Climate change now detectable from any single day of weather at global scale. Nat Clim Chang 10(1):35–41 Smith RM, Williams SB (2006) Circadian rhythms in gene transcription imparted by chromosome compaction in the cyanobacterium Synechococcus elongatus. Proc Natl Acad Sci 103 (22):8564–8569 Somanath PR, Podrez EA, Chen J et al (2011) Deficiency in core circadian protein Bmal1 is associated with a prothrombotic and vascular phenotype. J Cell Physiol 226(1):132–140 Spoelstra K, Wikelski M, Daan S et al (2016) Natural selection against a circadian clock gene mutation in mice. Proc Natl Acad Sci 113(3):686–691 Summa KC, Voigt RM, Forsyth CB et al (2013) Disruption of the circadian clock in mice increases intestinal permeability and promotes alcohol-induced hepatic pathology and inflammation. PLoS One 8(6):e67102 Tei H, Okamura H, Shigeyoshi Y et al (1997) Circadian oscillation of a mammalian homologue of the Drosophila period gene. Nature 389(6650):512–516 Thaiss CA, Zeevi D, Levy M et al (2014) Transkingdom control of microbiota diurnal oscillations promotes metabolic homeostasis. Cell 159(3):514–529 Troein C, Locke JC, Turner MS et al (2009) Weather and seasons together demand complex biological clocks. Curr Biol 19(22):1961–1964 Van Valen L (1973) A new evolutionary law. Evol Theory 1:1–30 Visser ME, Noordwijk AV, Tinbergen JM et al (1998) Warmer springs lead to mistimed reproduction in great tits (Parus major). Proc R Soc Lond Ser B Biol Sci 265(1408):1867–1870 Voigt RM, Forsyth CB, Green SJ et al (2014) Circadian disorganization alters intestinal microbiota. PLoS One 9(5):e97500 Voigt RM, Summa KC, Forsyth CB et al (2016) The circadian clock mutation promotes intestinal dysbiosis. Alcohol Clin Exp Res 40(2):335–347 von Saint Paul U, Aschoff J (1978) Longevity among blowflies Phormia terraenovae R.D. kept in non-24-hour light-dark cycles. J Comp Physiol 127(3):191–195 Woelfle MA, Ouyang Y, Phanvijhitsiri K et al (2004) The adaptive value of circadian clocks: an experimental assessment in cyanobacteria. Curr Biol 14(16):1481–1486 Zarrinpar A, Chaix A, Yooseph S et al (2014) Diet and feeding pattern affect the diurnal dynamics of the gut microbiome. Cell Metab 20(6):1006–1017 Zhu H, Sauman I, Yuan Q et al (2008) Cryptochromes define a novel circadian clock mechanism in monarch butterflies that may underlie sun compass navigation. PLoS Biol 6(1):e4

Reasons for Seeking Information on the Molecular Structure and Dynamics of Circadian Clock Components in Cyanobacteria Shuji Akiyama

Abstract We have used cyanobacterium Synechococcus elongatus PCC 7942 as the model system and looked for the slow but temperature-compensated reaction functioning as the pacemaker of in vivo rhythms. The key reaction we focused was the ATP hydrolysis as slow as 12 ATP d
1 in the N-terminal half of the clock protein KaiC. This intra-molecular-scale slowness comes from structural regulations of steric hindrance, water molecules, and cis-to-trans peptide isomerization in KaiC, being related on one-to-one correspondence not only to the frequency of intermolecular-scale rhythm of KaiA/KaiB/KaiC oscillator, but also to the frequency of cellular-scale rhythms.

1 Introduction The circadian clock regulates biological rhythms in an approximately 24-hour cycle. This clock has been studied in a wide range of species, from bacteria to mammals, and the following three characteristics have been identified (Pittendrigh 1993): (1) self-sustained circadian oscillation under constant conditions, (2) temperature compensation of the period length, and (3) entrainment of the oscillator by external stimuli. One of the major goals of the study of circadian clock systems is to provide a model that coherently explains these three characteristics. Since the first isolation of Drosophila clock mutants (Konopka and Benzer 1971), scientists have collected a large volume of knowledge regarding circadian clocks. However, to what extent has such knowledge successfully elucidated the nature of the biological timekeeping mechanisms? In this review, I will revisit the scientific

S. Akiyama (*) Research Center of Integrative Molecular Systems (CIMoS), Institute for Molecular Science, National Institute of Natural Sciences, Myodaiji, Okazaki, Japan Department of Functional Molecular Science, SOKENDAI (The Graduate University for Advanced Studies), Myodaiji, Okazaki, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_8

137

138

S. Akiyama

significance of investigating the circadian clock system from structural, biophysical, and physicochemical perspectives, while referring to my own research work.

2 Narrowing a Research Question Following the cloning of a circadian clock gene cluster in cyanobacteria (Ishiura et al. 1998), scientists attempted to apply a transcriptional–translational oscillation (TTO) model to interpret findings. In 7 years of the report, however, this approach was questioned by the groundbreaking findings by Kondo and collaborators (Tomita et al. 2005; Nakajima et al. 2005). Those studies demonstrated that the circadian oscillation could be reconstituted by incubating the clock proteins KaiA, KaiB, and KaiC in the presence of ATP even in vitro (Nakajima et al. 2005). This proved to be a game-changer in the study of cyanobacterial circadian systems and shifted focus to the structure and interactions of the Kai proteins. Crystal structures of all three types of the Kai proteins were published in the same year of 2004 (Ye et al. 2004; Garces et al. 2004; Pattanayek et al. 2004), underscoring the heated interest in the scientific community at that time. Readers interested in an excursion into the individual protein structures are referred to other review papers (Swan et al. 2018; Akiyama 2012; Partch 2020) for more details. Other biophysical studies reported the low-resolution architecture of Kai protein complexes, which repeat time-dependent assembly and disassembly cycles, as determined by electron microscopy (Pattanayek et al. 2008), nuclear magnetic resonance (Vakonakis and LiWang 2004), and X-ray scattering techniques (Akiyama et al. 2008; Murayama et al. 2011; Pattanayek et al. 2011). I postulated that KaiB binds to the C-terminal domain of KaiC (Akiyama et al. 2008); however, other research groups showed experimentally that KaiB preferentially binds to the N-terminal domain of KaiC (Chang et al. 2012; Snijder et al. 2017; Tseng et al. 2017). I also confirmed this through our later study (Mukaiyama et al. 2018). These scientific efforts to date have deepened our understanding and grown consensus regarding the three-dimensional structures of the Kai protein complexes (Partch 2020). When I jumped into this field, I assumed that tracing the structural changes of the clock components along the circadian cycle would improve our understanding of the circadian events. Although phosphorylation-dependent structural changes of KaiC were successfully detected in the later study (Murayama et al. 2011), they did not appear to be complex enough to suggest satisfactory answers to simple but basic questions such as: “What mediates the slow turnover rate?” and “What determines the 24-hour cycle?”.

Reasons for Seeking Information on the Molecular Structure and Dynamics of. . .

139

3 Transmural Hierarchy Among the three Kai proteins, KaiC is the most important and is located at the core of the oscillator. In the presence of both KaiA and KaiB, KaiC reveals the rhythm of autophosphorylation and dephosphorylation; however, the period of the oscillation is correlated with the ATPase activity of KaiC measured in the absence of KaiA or KaiB (Terauchi et al. 2007; Abe et al. 2015). For example, when the ATPase activity of KaiC doubles as a result of amino acid substitutions, the frequencies of both the in vitro and in vivo rhythms also double. I call this feature a transmural hierarchy, in which the cycle frequency and even the temperature compensation both in vitro and in vivo are greatly affected by the function and structure of KaiC. To examine the functional mechanism of KaiC ATPase, we needed to investigate the process whereby nucleotide-free KaiC monomer (pre-steady state) selfassembles to form a KaiC hexamer (steady state) upon addition of ATP. However, we were faced with a serious problem: thermodynamically unstable monomers form large water-soluble aggregates in the absence of both ATP and ADP. We thus improved the existed protocol for monomerization (Nishiwaki and Kondo 2012), and then the resultant KaiC monomer was subjected to pre-steady-state analysis of KaiC ATPase activity (Mukaiyama et al. 2015; Abe et al. 2015). In the presence of both KaiA and KaiB, the ATPase activity of KaiC is activated and then inactivated in a cyclic manner with the period of 24 h (Fig. 1a, inset). Even without KaiA and KaiB, however, KaiC revealed a damping oscillation, driving the activity rapidly in one direction and then overcorrecting in the opposite direction to reach the steady-state activity (Fig. 1a) (Abe et al. 2015). I was surprised at this sign, because it could mean that KaiC possesses a restoring force to bring the system back toward equilibrium through the oscillation. Another important finding is that KaiC gets matured as temperature-compensated ATPase through this relaxation. The number of ATP molecules consumed for this maturation is simply calculated by estimating the area of this curve to be approximately 6 ATP. Taken together, there appears to exist the restoring force in KaiC causing the damping relaxation through which KaiC gets temperature compensated. Application of a control engineering approach to the time course of KaiC ATPase activity revealed that it has an intrinsic circadian periodicity of approximately 24 h. Assuming a transfer function of a quasi-second-order system, we analyzed the ATPase relaxation curve of KaiC-WT shown in Fig. 1a and obtained an undamped natural frequency (ω) of 0.91 events per day. In the case of short-period mutants (Fig. 1b), it decayed quickly with a significant down-shooting. On the other hand, in the case of long-period mutants, it approached slowly with a minimal downshooting. The ω values of these mutants together with KaiC-WT were plotted in Fig. 1c(i), where the horizontal and longitudinal axes give the ω value of KaiC alone and the frequency of in vivo bioluminescence rhythm, respectively. A fine correlation between the ω and frequency values both only in vitro and in vivo (Abe et al. 2015) suggests that the slow relaxation with some restoring force in KaiC is one of the determinants of the period length in the cyanobacterial circadian clock system.

140

S. Akiyama

Fig. 1 Transmural hierarchy in the cyanobacterial circadian clock system. (a) Pre-steady-state analysis of KaiC ATPase. The inset represents a time course of the ATPase activity of KaiC in the presence of KaiA and KaiB. (b) Pre-steady-state kinetics of the ATPase activity observed for shortand long-period KaiC mutants. The values in parentheses correspond to the in vitro periods. (c) Pairwise correlations (panels i to iii) among the ω value, the frequency of in vivo TTO cycle, and the steady-state ATPase activity of KaiC1

4 Structural Basis of Slowness in KaiC Figure 2 shows a zoomed-in view of the active site of C1 ATPase, where one ATP molecule is located in the interface between two subunits. ATP hydrolysis is an event to let a lytic water molecule attack the phosphorous atom of the terminal gamma-phosphate and then to cleave it apart from ATP to form ADP and inorganic phosphate (Fig. 2a). This SN2-type reaction proceeds effectively when the lytic water molecule approaches along the axis defined by O-Pγ bond (broken line in Fig. 2a). In the following, I will use a near-in-line position (cross in Fig. 2a) distant away by 3 Å

Reasons for Seeking Information on the Molecular Structure and Dynamics of. . .

141

Fig. 2 Atomic-scale regulation of slow but stable ATP hydrolysis events in KaiC as a basis of transmural hierarchy. (a) Schematic of ATP hydrolysis reaction. A cross represents a near-in-line position distant away by 3 Å from the phosphorous atom (Pγ) along the axis defined by O-Pγ bond (broken line). (b and c) Zoomed-in views of the active sites of KaiC in near and far configurations, respectively. (d and e) Views from inner-radius side in the near and far configurations, respectively. (f) van der Waals radius representations of atoms for terminal gamma-phosphate of ATP (red and orange spheres), carbonyl oxygen of Phe199 (red sphere), and nitrogen atom of Arg226 (blue sphere). The white cross indicates the near-in-line position. (g and h) Post- and pre-hydrolysis states of KaiC, respectively. Steric hindrance effects preventing water molecules from accessing the optimal position for hydrolytic attack (Fig. 2f) are canceled through radical structural changes accompanying cis-to-trans isomerization of a peptide bond between Asp145 and Ser146. (i) Conformational selection model for KaiB–KaiC interaction. KaiB exclusively selects the posthydrolysis KaiC upon binding to transmit the slow but stable timing cue from the intra- to intermolecular scales

as a reference point to measure the reactivity of water molecules found in the crystal structure. Two types of water molecules were identified just nearby the terminal γ-phosphate of ATP. In one configuration (near configuration, Fig. 2b), a potential lytic water molecule is placed at 3.8 Å distance away from the phosphorous atom with some angle (ϕ ¼ 26 ). In another structure (far configuration, Fig. 2c), the water molecule is located at 4.6 Å distance away from the phosphorous atom (Pγ) with a

142

S. Akiyama

larger angle (ϕ ¼ 39 ). It is fair to conclude that the water molecule in the near configuration is more-reactive than that in the far configuration, because it is positioned at a closer distance at a smaller angle. The interpretation will be influenced to some extent; however, if compared to other highly active ATPases (e.g., F1-ATPase, myosin, and kinesin); the lytic water molecules are positioned at much closer distances (2.3–3.8 Å) and at much smaller angles (ϕ ¼ 8–16 ). It is evident that the water molecules in both near and far configurations in KaiC are lessreactive than those examples (Abe et al. 2015). Figure 2d and e illustrate views from the inner-radius side of the ring, in which conformational change of an α7 helix indirectly regulates the position of the lytic water molecule. The N-terminus of the α7 helix is perfectly capped in the near configuration (Fig. 2d), while it is frayed in the far configuration (Fig. 2e). A residue responsible for this capped-frayed switching is Ser157. An S157P substitution, which stabilizes the capped form of the α7 helix in the near/more-active configuration of KaiC, resulted in an activated ATPase and an increased frequency of in vivo cycle. On the other hand, an S157C mutation stabilizing the frayed form in the far/ less-active configuration resulted in an inactivated ATPase and a decreased cycle frequency. The ATPase activity per hexamer is kept low and constant by asymmetrically arranging the more-reactive near and less-reactive far configurations in the hexamer as observed in our crystal structures (Abe et al. 2015). A structural comparison between pre- and post-hydrolysis states of KaiC reveals a substantial energy barrier preventing the rapid ATP hydrolysis. Figure 2f shows the active site drawn using the van der Waals radius representations of atoms for the terminal γ-phosphate of ATP, the carbonyl oxygen of Phe199, and the nitrogen atom of Arg226. The near-in-line position that is favorable for the SN2-type reaction is located along a white dotted axis but distant away by 3 Å from Pγ (cross in Fig. 2f). A water molecule may try to occupy this optimal position, but this is not an easy task because not enough space to go in. This can be easily understood by taking the size of the oxygen atom colored red into consideration. In fact, we confirmed this using molecular dynamics simulation and found that both Arg226 and Phe199 constitute a steric hindrance to prevent water molecules from occupying the optimal position for hydrolysis. This steric blocking thus needs to be removed to advance the reaction. We found the steric hindrance is canceled through a large conformational change interlocked with the peptide isomerization at Ser146. In ordinary case, the peptide bond of proteins adopts trans configuration, in which an amide hydrogen atom and carbonyl oxygen atom are located at two different sides (Fig. 2g). In cis configuration, they are located on the same side (Fig. 2h). It is fair to say that cis-to-trans isomerization separated by a substantial energy barrier amounting 16 kcal mol
1 is the time-consuming step (Abe et al. 2015). Taken all together, the ATP hydrolysis, involving access of a water molecule to the bound ATP and isomerization of the peptide, will require a much larger amount of free energy than for typical ATP hydrolysis.

Reasons for Seeking Information on the Molecular Structure and Dynamics of. . .

143

5 From Intra- to Inter-Molecular Scales To investigate a logic allowing properties, such as slowness and temperature compensation, to be transferred through upward causation from bottom-to-top in the spatiotemporal hierarchy (Fig. 1c), the time scale of the Kai-protein-complex formation was measured by using a fluorescent-probe approach and then compared to that of KaiC ATPase (Mukaiyama et al. 2018). A detailed simulation of reaction kinetics suggests that conformational selection of a post-hydrolysis C1-ring by KaiB is important as a key event to transmit the slow but stable timing cue from the intramolecular (KaiC ATPase) to inter-molecular (KaiB–KaiC complex) scales (Fig. 2i).

6 Concluding Remarks Fine pairwise correlations were confirmed among the ω value, the frequency of in vitro cycle, and the frequency of in vivo rhythm (Fig. 1c). This indicates that the molecular structure of KaiC possesses an intrinsic timekeeping ability and transfers it from the intra-molecular scale through the inter-molecular scale, finally up to the cellular scale. We have shown that the slow KaiC ATPase turnover is mediated by a series of chemical reactions involving ATP, water molecule, peptide isomerization, and other common biological substances and events. We believe our observation provides at least partial answers to the basic questions about the origin of slowness. Future studies on circadian clock systems will increasingly employ emerging techniques, including advanced physicochemical approaches, system automation, and faster and parallel analysis/sampling (Furuike et al. 2016; Ouyang et al. 2019). The recent dramatic improvement in the resolution of electron microscopy has enabled pseudo-atomic-level analysis of the physical and chemical properties of samples. Structural studies of large clock protein complexes will benefit from these advancements in microscopic resolution. A number of scientists are actively seeking to identify mammalian physicochemical oscillators, a functional counterpart of the Kai oscillator, and exploring the synergy between chronobiology, biophysics, and structural biology. Acknowledgments I would like to thank Drs. T. Kondo (Graduate School of Science, Nagoya University, Japan), E. Yamashita (Institute for Protein Research, Osaka University, Japan), A. Mukaiyama, Y. Furuike, D. Ouyang, D. Simon, T. Mori, and S. Saito (Institute for Molecular Science, NINS, Japan) for their contributions to the study of the cyanobacterial circadian clock system.

144

S. Akiyama

References Abe J, Hiyama TB, Mukaiyama A, Son S, Mori T, Saito S, Osako M, Wolanin J, Yamashita E, Kondo T, Akiyama S (2015) Atomic-scale origins of slowness in the cyanobacterial circadian clock. Science 349:312–316 Akiyama S (2012) Structural and dynamic aspects of protein clocks: how can they be so slow and stable? Cell Mol Life Sci 69:2147–2160 Akiyama S, Nohara A, Ito K, Maeda Y (2008) Assembly and disassembly dynamics of the cyanobacterial periodosome. Mol Cell 29:703–716 Chang YG, Tseng R, Kuo NW, LiWang A (2012) Rhythmic ring–ring stacking drives the circadian oscillator clockwise. Proc Natl Acad Sci USA 109:16847–16851 Furuike Y, Abe J, Mukaiyama A, Akiyama S (2016) Accelerating in vitro studies on circadian clock systems using an automated sampling device. Biophys Physicobiol 13:235–241 Garces RG, Wu N, Gillon W, Pai EF (2004) Anabaena circadian clock proteins KaiA and KaiB reveal a potential common binding site to their partner KaiC. EMBO J 23:1688–1698 Ishiura M, Kutsuna S, Aoki S, Iwasaki H, Andersson CR, Tanabe A, Golden SS, Johnson CH, Kondo T (1998) Expression of a gene cluster kaiABC as a circadian feedback process in cyanobacteria. Science 281:1519–1523 Konopka RJ, Benzer S (1971) Clock mutants of Drosophila melanogaster. Proc Natl Acad Sci USA 68:2112–2116 Mukaiyama A, Osako M, Hikima T, Kondo T, Akiyama S (2015) A protocol for preparing nucleotide-free KaiC monomer. Biophysics 11:79–84 Mukaiyama A, Furuike Y, Abe J, Koda S, Yamashita E, Kondo T, Akiyama S (2018) Conformational rearrangements of the C1 ring in KaiC measure the timing of assembly with KaiB. Sci Rep 8:8803 Murayama Y, Mukaiyama A, Imai K, Onoue Y, Tsunoda A, Nohara A, Ishida T, Maéda Y, Terauchi K, Kondo T, Akiyama S (2011) Tracking and visualizing the circadian ticking of the cyanobacterial clock protein KaiC in solution. EMBO J 30:68–78 Nakajima M, Imai K, Ito H, Nishiwaki T, Murayama Y, Iwasaki H, Oyarna T, Kondo T (2005) Reconstitution of circadian oscillation of cyanobacterial KaiC phosphorylation in vitro. Science 308:414–415 Nishiwaki T, Kondo T (2012) Circadian autodephosphorylation of cyanobacterial clock protein KaiC occurs via formation of ATP as intermediate. J Biol Chem 287:18030–18035 Ouyang D, Furuike Y, Mukaiyama A, Ito-Miwa K, Kondo T, Akiyama S (2019) Development and optimization of expression, purification, and ATPase assay of KaiC for medium-throughput screening of circadian clock mutants in cyanobacteria. Int J Mol Sci 20:2789–2800 Partch CL (2020) Orchestration of circadian timing by macromolecular protein assemblies. J Mol Biol. https://doi.org/10.1016/j.jmb.2019.12.046 Pattanayek R, Wang JM, Mori T, Xu Y, Johnson CH, Egli M (2004) Visualizing a circadian clock protein: crystal structure of KaiC and functional insights. Mol Cell 15:375–388 Pattanayek R, Williams DR, Pattanayek S, Mori T, Johnson CH, Stewart PL, Egli M (2008) Structural model of the circadian clock KaiB–KaiC complex and mechanism for modulation of KaiC phosphorylation. EMBO J 27:1767–1778 Pattanayek R, Williams DR, Rossi G, Weigand S, Mori T, Johnson CH, Stewart PL, Egli M (2011) Combined SAXS/EM based models of the S. elongatus post-translational circadian oscillator and its interactions with the output His-kinase SasA. PLoS One 6:e23697 Pittendrigh CS (1993) Temporal organization – reflections of a Darwinian clock-watcher. Annu Rev Physiol 55:16–54 Snijder J, Schuller JM, Wiegard A, Lössl P, Schmelling N, Axmann IM, Plitzko JM, Förster F, Heck AJR (2017) Structures of the cyanobacterial circadian oscillator frozen in a fully assembled state. Science 355:1181–1184 Swan JA, Golden S, LiWang A, Partch CL (2018) Structure, function, and mechanism of the core circadian clock in cyanobacteria. J Biol Chem 293:5026–5034

Reasons for Seeking Information on the Molecular Structure and Dynamics of. . .

145

Terauchi K, Kitayama Y, Nishiwaki T, Miwa K, Murayama Y, Oyama T, Kondo T (2007) ATPase activity of KaiC determines the basic timing for circadian clock of cyanobacteria. Proc Natl Acad Sci USA 104:16377–16381 Tomita J, Nakajima M, Kondo T, Iwasaki H (2005) No transcription-translation feedback in circadian rhythm of KaiC phosphorylation. Science 307:251–254 Tseng R, Goularte NF, Chavan A, Luu J, Cohen SE, Chang YG, Heisler J, Li S, Michael AK, Tripathi S, Golden SS, LiWang A, Partch CL (2017) Structural basis of the day-night transition in a bacterial circadian clock. Science 355:1174–1180 Vakonakis I, LiWang AC (2004) Structure of the C-terminal domain of the clock protein KaiA in complex with a KaiC-derived peptide: implications for KaiC regulation. Proc Natl Acad Sci USA 101:10925–10930 Ye S, Vakonakis I, Ioerger TR, LiWang AC, Sacchettini JC (2004) Crystal structure of circadian clock protein KaiA from Synechococcus elongatus. J Biol Chem 279:20511–20518

Single-Molecule Methods Applied to Circadian Proteins with Special Emphasis on Atomic Force Microscopy Tetsuya Mori and Takayuki Uchihashi

Abstract Single-molecule (SM) techniques have emerged as important tools to study biomolecules and chemical reactions in biophysics and biochemistry, providing detailed microscopic information about dynamic processes of molecules. In the field of circadian biology, the SM approach has rarely been taken to study the properties of clock components. In this chapter, we briefly overview the SM experimental tools available for studying circadian clocks and introduce our biophysical study on the in vitro KaiABC oscillator using high-speed atomic force microscopy (HS-AFM). Our HS-AFM observation, for the first time, visualized dynamic interactions of clock proteins working in real time at the single-molecule level. The KaiA dimer binds to the KaiC hexamer at the C-terminus of KaiC, and the affinity of the KaiA dimer binding to the KaiC hexamer depends on the phosphorylation state of the KaiC hexamer; KaiA has higher affinity for less phosphorylated KaiC hexamers. Mathematical modeling and simulations revealed that the phosphoform-dependent differential affinity (PDDA) supports rhythmicity over a broad range of Kai protein stoichiometries, serving as a buffer against noise. Additionally, we demonstrate real-time observations of the binding of fold-switched KaiB (fsKaiB) to the CI domain of KaiC hexamers and the sequestration of the KaiA dimer into the KaiB-KaiC complex.

T. Mori (*) Department of Biological Sciences, Vanderbilt University, Nashville, TN, USA e-mail: [email protected] T. Uchihashi Department of Physics and Institute for Glyco-core Research (iGCORE), Nagoya University, Nagoya, Aichi, Japan Exploratory Research Center on Life and Living Systems (ExCELLS), National Institutes of Natural Sciences, Okazaki, Aichi, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_9

147

148

T. Mori and T. Uchihashi

1 Introduction Circadian clocks are internal time-keeping systems that enable organisms to anticipate daily environmental changes and coordinate cellular processes with the external cycles. The systems exist ubiquitously in the Eukarya and certain bacterial phyla. Within the clock, a circadian oscillator is a self-sustained timekeeper that is entrainable by external time cues (inputs) and regulates a series of biochemical pathways and physiological processes (outputs). In eukaryotes, the circadian oscillator is thought to consist of a transcriptional-translational feedback loop (TTFL), in which expression of circadian clock genes is suppressed by their protein products on a daily basis (Hardin et al. 1990; Takahashi 2017). Within the TTFL, core clock genes and their gene products are modified throughout the cycle. The clock genes undergo epigenetic changes such as DNA methylation and histone acetylation, while their protein products receive numerous post-translational modifications such as phosphorylation, ubiquitination, methylation, and S-sulfenylation (Mehra et al. 2009; Takahashi 2017; Pei et al. 2019; Mauvoisin and Gachon 2020). Those modifications are tightly controlled and essential for the generation of a robust oscillation and the determination of ~24-h period by regulating biochemical activities of the clock components in the TTFL. In bacteria, especially cyanobacteria, a post-translational oscillator (PTO), composed of the core clock proteins KaiA, KaiB, and KaiC, is the central timekeeper, and TTFL exists as a slave oscillator for PTO (Tomita et al. 2005; Nakajima et al. 2005; Kitayama et al. 2008; Qin et al. 2010b; Teng et al. 2013). The phosphorylation state of KaiC feeds back to its autophosphorylation and autodephosphorylation activities through the regulation of physical interactions among KaiA, KaiB, and KaiC proteins (Kageyama et al. 2006; Mori et al. 2007). The rate of each chemical and physical step in the feedback loop determines the properties, including the period of ~22 h, of the bacterial posttranslational circadian oscillator. Over the past few decades, in the field of research on circadian clocks, there has been an enormous success in identifying clock genes and analyzing their functions in both eukaryotic and bacterial systems (Cohen and Golden 2015; Takahashi 2017; Dunlap and Loros 2017; Creux and Harmer 2019). The genetic and biochemical characteristics of clock genes and their protein products have been exclusively analyzed at ensemble levels. Methodologically, for instance, the rates of phosphorylation of clock proteins are measured by SDS-PAGE followed by Western blotting analysis using anti-phospho antibodies or quantifying the incorporation of radioactive phosphorus (32P). Protein-protein interactions are indirectly measured by co-immunoprecipitation or by microcalorimeter or surface plasmon resonance (SPR) in isolation. These conventional biochemical and biophysical measurements have provided indispensable information to understand the oscillatory dynamics of the circadian clocks. While the recent advances in X-ray crystallography, nuclear magnetic resonance (NMR), single-particle cryogenic electron microscopy (CryoEM), and small-angle X-ray scattering (SAXS) have allowed us to obtain valuable structure information of clock proteins and their interactions at atomic or sub-atomic

Single-Molecule Methods Applied to Circadian Proteins with Special. . .

149

resolution (Vakonakis et al. 2004a; Vakonakis and LiWang 2004; Uzumaki et al. 2004; Pattanayek et al. 2004; Akiyama et al. 2008; Murayama et al. 2011; Partch et al. 2014; Abe et al. 2015; Snijder et al. 2017; Tseng et al. 2017), the structures, however, are an average snapshot of near-homogeneous conformations of biomolecules and thus yield limited information on dynamic molecular processes. Biomolecules are heterogeneous even if their chemical compositions (e.g., amino acid sequence of protein) are identical. For example, a protein exists in multiple conformational states even in a constant environment. An individual molecule (e.g., single protein) exhibits large temporal fluctuations of its behavior. In the case of enzymatic reactions, there are statistical fluctuations of the reaction rate constant over a broad spectrum of time scales ranging from milliseconds to minutes. By conventional biochemical and biophysical methods, the net reaction rates measure an ensemble- and time-average in a molecular population. Single-molecule techniques are used to directly observe and manipulate the behavior of individual biomolecules (nucleic acids, proteins, small chemicals) and have been expanded since the 1990s (Peterman 2018). The observation of individual molecules enables the dynamics of molecular processes to be investigated in detail by characterizing the kinetics of each elementary step in a multistep reaction and the fluctuations (changes) of structural (conformational or energetic) states within a biomolecule. Single-molecule technologies have been used to study, for example, mechanochemical dynamics of motor proteins, structural changes in nucleic acids, protein folding, subunit assembly, protein-nucleic acid interactions, protein-protein interactions, and protein-ligand associations (Peterman 2018). However, it has rarely been applied to study the circadian clocks, except for measuring the activities of ion channels related to circadian rhythms. In this chapter, we briefly review applicable single-molecule techniques to the research on biological clocks and present our study of the KaiABC oscillator using high-speed atomic force microscopy (HS-AFM) (Mori et al. 2018).

2 Single-Molecule Techniques 2.1

Single-Channel Patch-Clamp Recording

Neher and Sakmann established a technique to record single-channel currents in biological membranes (Neher and Sakmann 1976). It was the first real-time observation of the activity of a single molecule and has been used widely for characterizing biophysical and pharmacological properties of ion channels. Specifically, it has been used to study ion channels related to circadian rhythms in animals (D’Souza and Dryer 1996; Pennartz et al. 2002; Kononenko et al. 2004).

150

2.2

T. Mori and T. Uchihashi

Single-Molecule Fluorescence Microscopy

An attempt to measure the activity of an enzyme in solution at the single-molecule level dates back to 1961 (Rotman 1961). Individual molecules of β-D-galactosidase were encapsulated in a silicone oil droplet with a fluorescent substrate (6HFG), and the increase of the reaction product (6HF) in each droplet was monitored by fluorometry. Since then, there had been many attempts to measure the activity of single biomolecules. In 1995, Xue and Yeung studied the difference in the chemical reactivity of individual molecules of lactate dehydrogenase (LDH-1) by using capillary electrophoresis and sensitive detection of laser-induced fluorescence (Xue and Yeung 1995). Remarkably they demonstrated that individual molecules of native LDH-1 exhibit different activities by a factor of up to four, suggesting structural variants (heterogeneity) of the enzyme with different chemical properties in the molecular population. In the same year, Funatsu et al. succeeded in measuring individual ATP turnovers by single myosin molecules by total internal reflection fluorescence (TIRF) microscopy using fluorescently-labeled ATP (Funatsu et al. 1995). The fluorescently (Cy5)-labeled myosin head S-1 fragments were immobilized on the surface of a quartz slide glass to maintain a distance of at least few micrometers from one another, and fluorescently (Cy3)-labeled ATP (Cy3-ATP) was applied to S-1 on the surface. Since an evanescent wave penetrates from the glass-water interface to a depth of about 100 nm, only S-1 molecules on the surface and nucleotides on or near the protein are selectively excited. This local illumination technique allows the researchers to reduce background fluorescence (noise) from freely moving Cy3-nucleotides and Raman scattering of water in solution and to visualize faint fluorescence signals from the single molecules. In their study, the association of single Cy3-ATP with individual myosin S-1 molecules and the disassociation of Cy3-ADP from the S-1 following ATP hydrolysis were directly observed in real time. TIRF has become one of the powerful and universal tools to monitor the events of single molecules. Besides TIRF, epi-fluorescence, confocal fluorescence, and other fluorescence methods have also been utilized for the single-molecule fluorescence imaging (Peterman 2018). The fluorescence method has some downsides. First, it is necessary to fluorescently label the target molecules. Either by chemical conjugation or genetic fusion, labeling biomolecules with a fluorophore could alter their biochemical and biophysics properties. Care must be taken to minimize the effect of labeling on the biochemical activity of biomolecules. This is especially important in labeling small chemicals with a fluorophore (for example, the dissociation constant may change significantly). Secondly, the biomolecules of interest have to be immobilized on the surface on a glass substrate. It not only precludes three-dimensional diffusion but also could result in the reduction of structural flexibility of the biomolecule. Uses of biotin-streptavidin anchors, flexible PEG linkers, and blocking proteins such as BSA and casein are known to reduce the nonspecific interaction of biomolecule with the supporting substrate. Encapsulations of individual biomolecules in

Single-Molecule Methods Applied to Circadian Proteins with Special. . .

151

dimyristoylphosphatidylcholine (DMPC) vesicle or femtoliter-scale PDMS reaction chambers have been developed to avoid direct immobilization while allowing a long-term observation of the molecules (Okumus et al. 2004; Rondelez et al. 2005). Fluorescence correlation spectroscopy (FCS) measures fluctuations of fluorescence intensity in a sub-femtoliter volume at the focal point to detect the diffusion time of biomolecules in solution (Murakami et al. 2016). It has been used as a bridge between conventional and single-molecule methods. Thirdly, the accessible concentrations of biomolecules for the single-molecule assay are typically limited to a range from nanomolar to picomolar and often have to be kept much lower than their physiological (in vivo) concentrations. To overcome this problem, techniques using FRET (Roy et al. 2008; Mazal and Haran 2019), quenching (dark quencher) (Le Reste et al. 2012), and zero-mode waveguide (Goldschen-Ohm et al. 2016; Goldschen-Ohm et al. 2017) have been exploited to reduce the high levels of background fluorescence from bulk excitation of fluorophores in the imaging field. Furthermore, combined with high-resolution optical or magnetic tweezers, singlemolecule force spectroscopy has enabled us to study the mechanical properties of biomolecules (DNA, RNA, proteins) and their interactions with other biomolecules by simultaneously manipulating and observing an individual biomolecule (Neuman and Nagy 2008).

2.3

Atomic Force Microscopy

Atomic force microscopy (AFM), which is one of scanning probe microscopy, allows us to visualize the surface topography of a sample at nanometer resolution by placing a sharp tip attached to the free end of a cantilever in contact with the sample surface (Binnig et al. 1986). It enables imaging of surface structures with high spatial resolution on the nanoscale and even atomic scale (Ohnesorge and Binnig 1993). As is obvious from its operation principle, AFM is applicable to an insulating specimen regardless of the observation environment (in vacuum, air, or liquid) (Drake et al. 2006). Therefore, since its invention, AFM has been applied to a wide variety of biological specimens from proteins, nucleic acids, and chromosomes to living cells in a liquid environment (Dufrêne et al. 2017). One of disadvantages of AFM is its slow imaging speed. It usually takes at least a few minutes to acquire one frame with conventional AFMs, limiting the application of AFM to static structures of biomolecules strongly adsorbed on a substrate. This limitation has been broken by the development of high-speed AFM (HS-AFM) in 2001 (Ando et al. 2001), and its technological progresses have enabled us to film structural dynamics of a single protein and protein-protein interactions with a nanometer scale on a time scale less than one second. So far HS-AFM has been applied to various proteins and has succeeded in visualizing several dynamic phenomena as such as conformational dynamics of motor proteins and membrane proteins, intermolecular interactions, diffusion, and assembly processes of proteins in a lipid membrane (Ando et al. 2014; Uchihashi and Scheuring 2018). This

152

T. Mori and T. Uchihashi

remarkable feature of HS-AFM would be very useful for single-molecule analysis of intermolecular interactions involved in cyanobacterial clock proteins KaiA, KaiB, and KaiC because their complex formation seems to be inherently inhomogeneous and dynamic. In this chapter, we describe recent studies for dynamic behaviors of Kai proteins analyzed by HS-AFM based on the published results (Mori et al. 2018) and the additional experiments.

2.3.1

HS-AFM Imaging

Among the several operating modes of AFM, the tapping mode is usually used to observe fragile samples that are weakly adsorbed to the substrate, such as biomolecules (Zhong et al. 1993). In the tapping mode, the cantilever is oscillated near the mechanical resonant frequency using a piezo actuator so that the tip lightly taps the sample surface as shown Fig. 1a. This tapping action largely reduces the frictional force during the raster scanning of the probe tip relative to the sample surface, realizing less damage to soft samples (Hansma et al. 1994). Changes in the oscillation amplitude of the cantilever caused by the change of surface height are detected by the optical lever deflection method (Fig. 1b). The sample stage is also moved in the Z direction (up and down) by PID (proportional-integral-differential) feedback control so that the oscillation amplitude of the cantilever (i.e., the force with which the tip intermittently contacts the sample) is kept constant during the XY scanning of the sample stage, which gives a height mapping corresponding to the surface topography by recording the PID signal (proportional to the amount of movement of the sample stage in the Z direction) at each pixel position in the XY direction. For fast scanning of AFM, all processes, including feedback loops such as cantilever response, amplitude detection, and moving the sample stage in the Z direction, must be performed quickly and accurately in addition to quick raster scanning of the sample stage in the XY direction (Ando et al. 2008). In particular, in order to perform minimally invasive imaging without impairing physiological functions of fragile proteins, it is essential to maintain a small loading force on the AFM probe during scanning. Therefore, fast response in the feedback loop, i.e., a high feedback bandwidth, is the most important factor in HS-AFM. For this we minimized the delays of all components involved in the feedback loop, the Z piezoactuator, the amplitude detector, and the PID control circuit (Fig. 1b) as well as to suppress the vibration of the sample stage caused by quick scanning. For example, the size of the cantilever used in HS-AFM is reduced to about 1/10 (6–10 μm long, 2 μm wide, and 90 nm thick) of that of a cantilever used in conventional AFM in order to increase the resonance frequency while keeping a small spring constant small in solution (Fig. 1c, left panel). As a result of the optimization of all the devices, the feedback bandwidth finally reached around 100 kHz, which allowed imaging at an imaging rate less than 100 ms/frame. Also, to gain high-resolution images, a carbon tip is deposited on the end of the miniaturized cantilever using the electron-beam deposition method (Fig. 1c, middle panel)

2 µm

X Scan & Probe trace

Probe

Laser

Protein

1 µm

Substrate

b

Control PC

Topography

X&Y

Focusing lens

Objective lens

Glass stage

Two-segmented Si PIN diode

Differential amp.

100 nm

Amplitude detector

fier HV ampli

Y Z

X

Laser diode

Polarizing beam

λ/4 plate

Piezoactuator

Cantilever

Scanner

Fig. 1 Principle of HS-AFM. (a) Operation principle of tapping-mode AFM. (b) Block diagram of HS-AFM. (c) A miniaturized cantilever for HS-AFM imaging (left: whole picture of the cantilever, middle: A carbon tip grown on the cantilever end with electron beam deposition. Right: the tip end after sharpening)

c

a

Single-Molecule Methods Applied to Circadian Proteins with Special. . . 153

154

T. Mori and T. Uchihashi

and is further sharpened to 4 nm in radius by plasma etching, which routinely gives us the spatial resolution of less than 2 nm (Fig. 1c, right panel).

3 Visualizing Circadian Clock Proteins by HS-AFM 3.1

Experimental Conditions for HS-AFM Imaging of Kai Proteins

To observe the intermolecular interactions between Kai proteins, KaiC hexamers were first immobilized to either bare mica or chemically modified mica with 3-aminopropyltriethoxy silane (APTES) glued on a cylindrical glass stage with a diameter of 1.5 mm and height of 2 mm. The APTES treatment was carried out by dropping 3 μL of APTES diluted to 0.1% with pure water onto the freshly cleaved mica substrate. After 3-min incubation, the substrate was washed thoroughly with 50 μL of pure water. KaiC immobilization onto the substrate was carried out by dropping 2 μL of sample solution and incubating for 5 min, followed by washing with an observation buffer to remove residual proteins. The chemical modification of the mica surface changes the adsorption orientation of the KaiC hexamer on the substrate, which allows us to control the orientation of the KaiC-hexamer ring exposed to the solution. Because KaiC has a peptide tentacle at the C-terminus (CII) but not at the N-terminus (CI) (Figs. 2a and b), the images observed with HS-AFM are expected to be different between the CI and CII ring of KaiC hexamer facing to the solution (Hayashi et al. 2003). To ascertain how much the observed images are likely to be different between the CI- and CII-surface, we constructed pseudo-AFM images for the both sides. Here, we constructed the pseudo-AFM images with a simple hard sphere model using an ideal conical tip (a radius of 0.5 nm and a cone angle of 10
) and the crystal structure of KaiC hexamer (PDB: 2GBL) (Pattanayek et al. 2006). The pseudo-AFM image of the CI surface has a central pore in the hexamer (Fig. 2d), while the pseudo-AFM image at the CI surface shows the protrusion because the presence of C-terminal tentacles covers the central pore (Fig. 2e). A real HS-AFM image obtained on the KaiC hexamer attached to bare mica shows hexagonal ring-like structures with the central pore, in good agreement with the pseudo-AFM image of the CI ring (Fig. 2g), indicating that the KaiC hexamer are adsorbed on the bare mica substrate with the CI ring facing upward. On the other hand, the KaiC hexamer on the APTES-mica surface had no obvious pore in the center and closely resembles the pseudo-AFM image of the CII side of KaiC hexamer. Moreover, for the KaiC-ΔC hexamer where the C-terminal tentacles are truncated, the pore in APTES-mica-immobilized KaiC becomes apparent again (Figs. 2c, f and i). Thus, we can control the orientation of the KaiC hexamer fixed on the substrate as follows: (1) on bare mica, the C-terminal side of the KaiC hexamer attached to the mica (CI end-up orientation) and (2) on APTESmica, the N-terminal side attached to the APTES-mica (CII end-up orientation). The

Single-Molecule Methods Applied to Circadian Proteins with Special. . .

(a)

155

(b)

(c)

CII ring

CII ring

CI ring

CI ring

AFM probe

CI ring

KaiC

CII ring bare mica

(d)

(g)

APTES-mica

APTES-mica

(e)

(h)

KaiC-ΔC

(f)

(i)

Fig. 2 HS-AFM images of KaiC hexamers immobilized on different substrates. (a–c) Side views of the KaiC hexamer on bare mica and APTES-mica and the AFM probe end. (d–f) Pseudo-AFM images of (d) CI and (e) CII end of the full-length KaiC and (f) CII ring of KaiC-ΔC hexamer that were simulated with a hard sphere model for the crystal structure (PDB code: 2GBL). Scale bar, 5 nm. (g–i) Typical HS-AFM images of the full-length KaiC immobilized on (g) bare mica and (h) APTES-mica, and the KaiC-ΔC hexamer immobilized on (i) APTES-mica. Scale bar, 15 nm. The insets show averaged HS-AFM images (scale bar, 5 nm). (Reproduced from Mori et al. 2018)

amino acid sequence of the KaiC hexamer CI-ring surface has negatively charged amino acid residues (aspartic acid and glutamate), while the CII-ring has positively charged amino acid residues (arginine and lysine) (Pattanayek et al. 2004). Because the bare mica surface is generally negatively charged, while the APTES-mica surface is positively charged by the amino group, the orientation of the KaiC hexamer adsorption on the substrate is as expected. For the observation of KaiC-KaiA interaction, the adsorption of KaiC hexamers on the substrate was first confirmed by HS-AFM observation, and then KaiA was added in the observation buffer during imaging. After the addition, the sample concentration in the observation buffer was mixed by pipetting the buffer solution several times. Then the binding and dissociation of KaiA on the KaiC hexamers on the substrate were monitored by HS-AFM. As for the observation of KaiC-KaiB interaction, due to a low affinity of KaiB for KaiC, a high concentration solution of KaiB was required for the HS-AFM observation of the KaiC-KaiB complex. Thus, after confirming the adsorption condition

156

T. Mori and T. Uchihashi

of KaiC to the substrate, the sample stage was pulled out of the observation buffer and then 2 μL of KaiB solution (the concentration is specifically described in the figure legends) was placed directly onto the substrate. After 30-min incubation, the sample was immersed in the observation buffer again without washing and then the HS-AFM observation was performed. For tapping-mode HS-AFM imaging, the miniaturized cantilever with resonant frequency of ~0.8 MHz, spring constant of 0.1 N/m, and quality factor ~ 2 in solution was used. The free oscillation amplitude of the cantilever was 1–2 nm, and the set point amplitude for feedback control was set at about 90% of the free amplitude to avoid significant disturbance of the interaction between the Kai proteins.

3.2 3.2.1

KaiA-KaiC Interaction KaiA Interaction Depends upon KaiC Phosphostatus

Previous experiments (Vakonakis and LiWang 2004) have shown that KaiA interacts with the C-terminal tentacles of KaiC. To confirm that HS-AFM is able to replicate these observations, the first experiment is to measure the interaction of KaiA on the CI and CII surface of wild-type KaiC. Native KaiA was added to the observation buffer at the final concentration of 1 μM while observing the CII ring surface of KaiCWT in an approximately 80% phosphorylated state. After a few seconds of the KaiA addition and pipetting of the buffer, several bright spots (approximately 3.5-nm high) appeared on the densely packed KaiC hexamers (Fig. 3a). These bright spots were observed only after the addition of KaiA, confirming that KaiA interacts with the C-terminus of KaiC. Moreover, we confirmed that non-KaiA proteins such as BSA, GFP, or GST did not interact with the KaiCWT hexamer (data not shown). The position of the KaiA spots appeared and disappeared over time, indicating that the interaction between KaiA and KaiCWT is inherently dynamic (Kageyama et al. 2006; Lin et al. 2014; Murakami et al. 2016; Qin et al. 2010a). On the other hand, KaiA had no interaction with KaiCWT fixed on the bare mica in the CI-up orientation, supporting KaiA binding to the C-terminal tentacles that are inaccessible in this orientation of KaiCWT (Fig. 3b). A hypothesis of differential affinity between KaiA and KaiC was proposed by van Zon et al. in their theoretical analysis of the in vitro oscillator (van Zon et al. 2007). In their prediction, the affinity of KaiA for KaiC is dependent on the phosphorylation state of KaiC though it lacks adequate experimental support as yet. Hereafter we refer to it as KaiC-phosphoform-dependent differential affinity (PDDA) of KaiA for KaiC. To test the PDDA model experimentally, we observed dynamic interactions of KaiA to the KaiCWT hexamers with different phosphor-states. We here used a KaiCWT that was 81% phosphorylated and a KaiCWT dephosphorylated to 27% by incubation at 30
C for 36 h. For both samples, native KaiA was added to the observation buffer with a final concentration of 0.4 μM on APTES-mica where the CII end and the C-terminal tentacles are up. The number of bound KaiA was clearly

Single-Molecule Methods Applied to Circadian Proteins with Special. . .

-KaiA

(a)

157

+KaiA 0s

4s

8s

12 s

CII

nm

10

CI

APTES-mica

5

0

(b)

0s

4s

8s

12 s

CI

bare mica

6 nm

CII

4 2 0

Fig. 3 Orientation-specific immobilization of KaiCWT and interaction with KaiA. (a) KaiCWT on bare mica before (middle) and after (right) the addition of KaiA. (b) KaiCWT on APTES-mica before (middle) and after (right) the addition of KaiA. For both observations, the final concentration of KaiA was 1 μM. Scale bar, 15 nm. (Reproduced from Mori et al. 2018)

different between hyperphosphorylated (81%) and hypophosphorylated (27%) KaiCWT (Fig. 4a), supporting that the affinity for KaiA depends on the phosphorylation status of KaiCWT. To determine the frequency of the binding event between KaiA and KaiC, the dwell times of the KaiA-bound state of KaiCWT were measured (Fig. 4b). KaiA remains bound dramatically longer to hypophosphorylated KaiCWT than to hyperphosphorylated KaiCWT; the bound-state dwell times τbound estimated by an exponential fitting for hypophosphorylated and hyperphosphorylated KaiCWT were 1.54  0.03 s and 0.38  0.01 s, respectively. We also measured the unboundstate dwell times (τunbound) from the interval of time in which KaiCWT hexamers remain unassociated between consecutive KaiA-binding events (Fig. 4c). Estimated τunbound for hyper- and hypophosphorylated KaiCWT were 7.62  0.12 s and 5.60  0.19 s, respectively. These results support that the affinity of KaiC is significantly modulated by the phosphorylation state of KaiC and that the phosphorylated state of KaiC has a lower affinity for KaiA, i.e., a shorter time of binding and a longer time of dissociation compared to the dephosphorylated state. KaiC has two autophosphorylation sites, S431 and T432, within CII, which proceed in the order of S/T!S/pT!pS/pT!pS/T (“p” means the site is phosphorylated), during the circadian rhythm (Nishiwaki et al. 2007; Rust et al. 2007). Thus, there are four possible substitutes of KaiC. The dephosphorylated states of serine

158

T. Mori and T. Uchihashi

(a) 1s

2s

3s

4s

0s

1.6 s

3.2 s

4.8 s

6.4 s

27%-phospho

81%-phospho

0s

(b)

81%-phospho, bound state

40

40 bound =

20

0.38 ± 0.01 s (n = 104)

Counts

30 Counts

27%-phospho, bound state

50

10

bound =

30

1.54 ± 0.03 s (n = 412)

20 10

0

0 0

2 4 6 8 Dwell time of bound state (s)

10

0

2 4 6 8 Dwell time of bound state (s)

10

(c) 81%-phospho, unbound state

Counts

8

unbound =

6

7.62 ± 0.25 s (n = 71)

27%-phospho, unbound state

75

unbound =

50 Counts

10

4

5.60 ± 0.09 s (n = 365)

25

2 0 0

10 20 30 40 Dwell time of unbound state (s)

50

0 0

10 20 30 40 Dwell time of unbound state (s)

50

Fig. 4 Interactions of KaiA to the hyper- and hypophosphorylated KaiCWT. (a) HS-AFM images of KaiC and KaiA interaction. The KaiCWT hexamers that were approximately 81% phosphorylated (upper panels) or approximately 27% phosphorylated (lower panels) were immobilized on APTESmica in the CII-end up (C-terminal-up) orientation and KaiA was added into the observation buffer at the final concentration of 0.4 μM. Images were acquired at an imaging rate of 1 s/frame (upper) and 0.8 s/frame (lower). Scale bar, 30 nm. (b, c) Dwell time analysis for KaiA-unbound (b) and KaiA-bound (c) states of hyperphosphorylated (81%; left) and hypo-phosphorylated (27%; right) KaiC. Time constants (τbound or τunbound) were determined by a single-exponential curve fit (black line) to histograms of dwell times. (Reproduced from Mori et al. 2018)

Single-Molecule Methods Applied to Circadian Proteins with Special. . .

159

(S) and threonine (T) are mimicked by alanine (A), while the phosphorylated states of S and T are mimicked by aspartic acid (D) and glutamate (E), respectively. Here, the four phosphorylation states of KaiC, substituting S431 and T432, are KaiC-AA, KaiC-AE, KaiC-DE, and KaiC-DA. Figure 5a shows HS-AFM images of the four KaiC phospho-mimics at 1-second intervals acquired at an imaging rate of 0.1 s/frame. The number of KaiA molecules bound to KaiC is clearly dependent on the KaiC phosphorylation mimics. Consistent with the PDDA hypothesis, the mimics of hypo-phosphorylated KaiC bind KaiA for a significantly longer time than the mimics of hyperphosphorylated KaiC (DE: τbound ¼ 0.26  0.05 s, DA: τbound ¼ 0.43  0.14 s, AE: τbound ¼ 1.00  0.15 s, AA: too slow to determine accurately, Fig. 5b). In KaiC-AA, which mimics the dephosphorylated state of both S431 and T432, the binding time was difficult to be estimated because KaiA binds to KaiC for a very long time with little dissociation. It should be noted that KaiC-pS/T and -S/pT, where each single site is phosphorylated, differ in their affinity with KaiA, with KaiC-S/pT remaining bound more than twice as long as KaiC-pS/T. This suggests that the phosphorylation of T432 affects affinity changes with KaiA more than that of S431. The interactions of KaiA with hypo- and hyperphosphorylated KaiCWT (Fig. 4) and hypo-phospho- and hyper-phospho-mimic KaiC derivatives (Fig. 5) convincingly demonstrate that the initiation of KaiA-KaiC binding is dependent on the phosphorylation state of the KaiC hexamer (τunbound) and that once KaiA and KaiC are bound, the binding is destabilized by increased phosphorylation of KaiC (τbound). Since it is known that KaiA binding to the C-terminal tentacles stimulates KaiC autokinase activity (Vakonakis and LiWang 2004; Kim et al. 2008), these differential KaiA-KaiC affinities (PDDA) should influence the KaiA-dependent difference in the phosphorylation rate of KaiC, depending on the phosphorylation state of KaiC (Lin et al. 2014).

3.2.2

Synchronous Oscillation of KaiA-KaiC Affinities with In Vitro Rhythm

To examine experimentally whether the KaiA-binding affinity to KaiCWT oscillates synchronously with the in vitro circadian oscillation, we examined the dynamic interaction of KaiA with KaiCWT samples that had been taken at different phases from an in vitro reaction. The circadian oscillation was reconstructed by mixing wild-type Kai proteins (KaiC: 3.43 μM, KaiA: 1.52 μM, KaiB: 4.27 μM, total volume 100μL) in vitro in the presence of 1 mM ATP. The samples were taken from the test tube every 3 hours during the circadian oscillation in a volume of 2 μL and adsorbed onto the APTES-mica substrate for the HS-AFM observation. The samples taken from the tube contained three types of proteins: KaiA, KaiB, and KaiC. When the samples were washed with the buffer after loading on APTES-mica, KaiA or KaiB were washed away, and only KaiCWT was observed to be adsorbed on the substrate with the CII surface facing upward. This is probably because most of KaiA were removed because of repeated binding/dissociation with KaiC. As for

160

T. Mori and T. Uchihashi

0s

1s

2s

3s

4s

5s

6s

0s

1s

2s

3s

4s

5s

6s

0s

1s

2s

3s

4s

5s

6s

0s

1s

2s

3s

4s

5s

6s

AA

AE

DA

DE

(a)

(b)

100

80

DA

τ bound = 0.26 ± 0.05 s (n = 211)

60

τ bound = 0.43 ± 0.14 s (n = 236)

60

Counts

Counts

80

DE

40

20

20 0

0 0

10

0

10

Counts

6 4

2 4 6 8 Dwell time of bound state (s)

10

AA

8

τ bound = 1.00 ± 0.15 s (n = 64)

8 Counts

2 4 6 8 Dwell time of bound state (s)

AE

10

τ bound = ND (n = 96)

6 4 2

2 0

40

0

2 4 6 8 Dwell time of bound state (s)

10

0

0

2 4 6 8 Dwell time of bound state (s)

10

Fig. 5 Modulation of the binding affinity of KaiA for various KaiC phospho-mimics. (a) Successive HS-AFM images showing interaction between the KaiAWT and the KaiC phosphomimics in the CII-end up orientation. The phosphomimics used were KaiC-DE (pS/pT mimic), KaiC-DA (pS/T mimic), KaiC-AE (S/pT mimic), and KaiC-AA (S/T mimic). Concentration of KaiA, 1 μM. Scale bar, 20 nm. (b) Dwell time analysis for KaiA-bound state of the phosphomimics of KaiC. The τbound value for AA could not be determined (ND). (Reproduced from Mori et al. 2018)

wild-type KaiB, the extent of KaiB binding to KaiC was quite small at the concentration of KaiB in the in vitro reaction, and as with KaiA, much of it was probably removed by washing. To the KaiC bound to the APTES-mica, free KaiA (final

Single-Molecule Methods Applied to Circadian Proteins with Special. . .

161

concentration 1.9 μM) was added to the observation buffer to observe transient KaiA-KaiCWT interaction (Fig. 6a). In parallel with HS-AFM observations, the phosphorylation states of KaiC taken from the same test tube were assessed by SDS-PAGE. We measured the binding times of KaiA to KaiCWT taken out every 3 h and made histograms and estimated the bound-state dwell times, τbound, for every 3 h (Fig. 6b). As shown in Fig. 6c, the affinity between KaiC and KaiA oscillated reproducibly in the opposite phase of the KaiC phosphorylation cycle. KaiCWT from the peak phase of the phosphorylation cycle showed a low mean affinity for KaiA (τbound ¼ 0.46  0.02 s at the 24-h time point) while KaiCWT at the lowest phosphorylated phase showed a higher mean affinity (τbound ¼ 0.77  0.05 s at 15 h) (Fig. 6b). These affinity values are averaged due to the different proportions of KaiCWT phosphoforms present at each stage of the in vitro cycle. There is a strong correlation between the KaiC phosphorylation state and the mean dwell time of the KaiA-binding to KaiC (Fig. 6d). That is, as predicted by the PDDA hypothesis, KaiA is bound longer to unphosphorylated KaiC than in phosphorylated KaiC.

3.2.3

Reinforcement of Oscillatory Resilience with PDDA

Our results clearly indicate the KaiC hexamer has differential affinities to KaiA depending on its phosphorylation states, i.e., PDDA (Mori et al. 2018). Because the binding of KaiA to the C-terminal tentacles of KaiC hexamer stimulates KaiC autokinase activity (Vakonakis and LiWang 2004; Pattanayek et al. 2006; Kim et al. 2008), it could be argued that the PDDA is a feedback control mechanism in which the affinity of KaiA to KaiC is negatively regulated by the KaiA-stimulating KaiC phosphorylation. The concept of differential affinity was proposed by van Zon et al. in their theoretical framework accounting for the in vitro oscillator (van Zon et al. 2007). Their mathematical model and numerical simulations predicted that the change of affinity of KaiC for KaiA is a mechanism for allowing individual KaiC hexamers to be synchronized, thereby stabilizing the circadian oscillator. To investigate the possible role of PDDA in the system, we constructed a stochastic model, in which experimentally known features of the KaiABC oscillator such as the sequential phosphorylation and dephosphorylation of KaiC (Nishiwaki et al. 2007; Rust et al. 2007), phosphoform transitions in each KaiC protomer (Kitayama et al. 2013; Lin et al. 2014), KaiB-KaiC interaction (Kageyama et al. 2006), KaiA sequestration (Qin et al. 2010a), and monomer exchange (Ito et al. 2007) are included. Our stochastic simulations consider hundreds of different hexameric states in the form of a simple mathematical matrix (including bound and unbound states of KaiC hexamer and four different phosphorylation states of each KaiC protomer) and also include intrinsic variability in the transition rates that are based on experimentally obtained reaction rates and dissociation constants (Rust

162

T. Mori and T. Uchihashi

(a) 1s

2s

3s

4s

0s

1s

2s

3s

4s

24h

15h

0s

(b) 15 h

50 40

τ bound = 0.46 ± 0.02 s (n = 212)

50 Counts

Counts

60

τ bound = 0.77 ± 0.05 s (n = 233)

30

24 h

70

20

40 30 20

10

10

0

0 0

2

4 6 8 10 Dwell time of bound state (s)

12

2

4 6 8 10 Dwell time of bound state (s)

12

(d) 1.0

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.0

0

12

24 36 Time (h)

48

0.0

0.9

Binding lifetime, τbound (s)

1.0

KaiC phosphorylation ratio

Binding lifetime, τbound (s)

(c)

0

0.8

r2 = 0.79

0.7 0.6 0.5 0.4 0.3 0.2 0.3

0.4 0.5 0.6 0.7 KaiC phosphorylation ratio

0.8

Fig. 6 Rhythmic change of KaiA’s affinity for KaiC over the cycle of the in vitro KaiABC oscillator. (a) HS-AFM images of the interaction between KaiAWT and KaiCWT at 15 h (trough of KaiC phosphorylation) and 24 h (peak of KaiC phosphorylation). Native KaiA (final concentration of 1.9 μM) was added to KaiC immobilized onto APTES-mica. All images were acquired at 0.1 s/frame and are shown every 1 s. Scale bar, 30 nm. (b) Dwell time analysis of KaiA-bound state lifetime at 15 h (left) and 24 h (right) in the in vitro phosphorylation cycle. (c) KaiA-bound state lifetime (τbound) depends on the phosphorylation status of KaiCWT over a course of the in vitro phosphorylation cycle. KaiA–KaiC binding lifetimes (τbound) were estimated from the bound state dwell time analysis (red). Parallel samples were collected, and the phosphorylation status of KaiC (blue) was analyzed by SDS-PAGE. (d) Correlation between the degree of KaiC phosphorylation and KaiA-bound state lifetime (τbound). (Reproduced from Mori et al. 2018)

Single-Molecule Methods Applied to Circadian Proteins with Special. . .

(b) 300 250 200

100 50 0

0

24

48

72

96

120 144 168 192

300 250

100 80 60 40 20 0

100 80 60 40 20 0

200

- PDDA [A2]/[C6]

0

24

48

72

3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.1

96

+ PDDA

0

24

(c)

150

48

72

96

Hours 50

100 50 0

0

24

48

72

96

120 144 168 192

300 250 200 150 100

+PDDA

30 20 10

50 0

-PDDA

40 Period (hours)

Ratio of [KaiA2] to [KaiC6] (%)

KaiC phosphorylation (%; -PDDA & +PDDA)

150

KaiC phosphorylation (%)

(a)

163

0

24

48

72

96

Hours

120 144 168 192

0 0

1

2 3 [KaiA2] / [KaiC6]

4

Fig. 7 PDDA enhances resilience to variations in Kai protein stoichiometry. (a) Simulated random fluctuations of KaiA concentration (green) on KaiC phosphorylation patterns +PDDA (red) versus PDDA (black). KaiC and KaiB concentrations were held constant as [KaiA] was varied. (b) KaiC phosphorylation as the ratio of concentrations of KaiA dimer to KaiC hexamer ([A2]/[C6]) was varied from the beginning of the time course. Simulations were performed without PDDA (PDDA, top) and with PDDA (+PDDA, bottom). (c) Range of allowed oscillatory regime (vertical dashed lines) and estimated period as the ratio of KaiA dimer to KaiC hexamer ([KaiA2]/[KaiC6]) was varied with (+PDDA, red) or without (PDDA, black) PDDA. Absence of oscillations is indicated as period ¼ 0. (Reproduced from Mori et al. 2018)

et al. 2007). The PDDA was implemented in our model and its implications on the oscillatory dynamics were investigated. We first examined whether PDDA has an effect on the synchronization of KaiC hexamers as predicted by van Zon et al. (van Zon et al. 2007). Unexpectedly, we found only a little effect of PDDA on the synchronization. Our simulations indicated that the magnitude of PDDA verified in our HS-AFM experiments (the affinity differs ~10 between hyper- and hypo-phosphorylated KaiC) is considerably lower than what is effective for the synchronization (100-1000 difference in affinity is required). So, what is the role of PDDA? In the course of our extensive simulations, we found another effect of PDDA that would be more significant under in vivo conditions. Cellular protein levels fluctuate with changes in environmental and physiological conditions. The concentrations of the clock proteins and the stoichiometric relationship between the clock proteins would also fluctuate in vivo. Figure 7a shows simulations when the concentration of KaiA is randomly

164

T. Mori and T. Uchihashi

varied over time. The simulated results show that the oscillations in the presence of PDDA (+PDDA, red) are more stable than those in the absence of PDDA (PDDA, black); under conditions with small fluctuations in [KaiA dimer]/[KaiC hexamer] ratio (1.7–2.3), the oscillations are maintained regardless PDDA, but under conditions with large fluctuations in [KaiA dimer]/[KaiC hexamer] ratio (1.5–3.0), KaiC phosphorylation becomes arhythmic without PDDA (PDDA) but maintains rhythmicity with PDDA (+PDDA). The simulations with different steady-state concentrations of KaiA are shown in Figs. 7b and c. The +PDDA and PDDA within the range of the ratio of [KaiA dimer]:[KaiC hexamer], which allows for rhythmic oscillations for both +PDDA and PDDA(0.3–2.6), with PDDA maintains a precise period (~22 h), but the ratio > 2 in the absence of PDDA the period gradually lengthens until it becomes arhythmic (period of ~35 hours at the ratio of 2.6). Those results suggest that PDDA may play a role in the resistance of the oscillation to internal and external noise in the bacterial cells living in diversified environments. Mechanistically, the transition (“switch”) from the autokinase to the autophosphatase mode in KaiC hexamer requires KaiA sequestration into ternary KaiABC complexes (Kim et al. 2008). KaiB slowly binds to the B-loop domain of KaiC hexamer in the pS/T state, and simultaneously KaiA docks onto the exposed ß2-sheet on fold-switched KaiB in the KaiBC complex to form the stable KaiABC complex (Chang et al. 2015; Snijder et al. 2017; Tseng et al. 2017). Our in silico analysis implies that PDDA facilitates the smooth transition of KaiC to the state that binds KaiB (pS/pT!pS/T) by preventing “overshot” or “undershot” of KaiC phosphorylation through the negative-feedback regulation of KaiA-stimulated autokinase activity, and thereby provides more resistance to noise. We further tested the prediction of the PDDA-enhanced oscillatory resilience in in silico and in vitro experiments. We altered the [KaiA dimer]:[KaiC hexamer] stoichiometry by adding KaiA to various final concentrations into in silico reactions in the middle of the dephosphorylation phase. PDDA enabled the system to tolerate a large range of single-step increases of [KaiA dimer]:[KaiC hexamer] ratio; the oscillation continues in the presence of PDDA (Fig. 8b, colored lines) but dampens rapidly in the absence of PDDA (Fig. 8b, black lines). Equivalently, to confirm the in silico prediction, we performed an in vitro experiment where the different amount of KaiA was added to an in vitro KaiABC reaction during the dephosphorylation phase as done in the simulations. Remarkably, we observed that the in vitro oscillation indeed continued after the addition of KaiA (Fig. 8a) as predicted from the simulation. Moreover, the parameters of the oscillation, such as amplitude, phase and period, predicted by the in silico +PDDA simulation, were observed similarly in the experiment in vitro. Our HS-AFM experiments revealed that the transient and rapid interactions between KaiA and KaiC generate association and dissociation rate constants that are dependent on the phosphorylation status of KaiC (Fig. 8c). Phosphorylation of clock proteins is an essential process in all the known circadian clock systems. Features of PDDA might be preserved in other clock systems as well. Additionally, the high-frequency events (~10,000 day1) we observed are considerably faster compared to the period of circadian oscillation (~1 day1). The timescale of many

Single-Molecule Methods Applied to Circadian Proteins with Special. . .

(a)

165

(b)

[A2]/[C6]

KaiC phosphorylation

1.33 →4.1 1.33 →3.0

1.33 → 2.4

1.0

1.33 →1.9

1.0

0.5

0.5 0.0

1.33 →1.33

0.0 0

12 24 36 48 60 72 84 96 108 120

0

12 24 36 48 60 72 84 96 108 120

Hours

Hours

(c) High frequency KaiA/KaiC binding (~10,000 day-1)

ON

ON

ON

ON

ON

OFF

OFF

OFF

OFF

OFF

S/T KaiA affinity to KaiC

S/pT

dwell time

pS/pT

pS/T

S/T

KaiC phosphorylation

KaiB affinity to KaiC

PDDA decreases KaiA affinity as KaiC phosphorylates

PDDA promotes KaiB binding ( less dependent on [KaiA] ) KaiA sequestration

Fig. 8 Testing of resilience enhancement by PDDA in silico and in vitro. (a) Experimental confirmation of the in silico prediction of the effects of an acute increase of KaiA concentration during the in vitro oscillation. A concentrated KaiA solution was added to a KaiABC cycling reaction at hour 32. The changes of KaiA concentration were from 1.53 μM to 1.47 (red), 2.08 (orange), 2.69 (green), 3.23 (blue), and 4.50 (magenta) μM, and KaiC concentration was changed from 3.43 to 3.32 μM in all the reactions. (b) In silico simulation of predictions for the experimental test in a. The increase of KaiA dimer to KaiC hexamer ratio ([A2]:[C6]) at hour 27 was from 1.33 to 1.33 (red), 1.9 (orange), 2.4 (green), 3.0 (blue), or 4.1 (magenta). Depicted are KaiC phosphorylation patterns resulting from step-increases of KaiA concentration during the dephosphorylation phase with PDDA (+PDDA, colored traces) were compared with those in the absence of PDDA (PDDA, black traces). The traces are vertically offset for as in (a). The [A2]/[C6] values at the far right apply to both the experimental data (a) and the simulation (b). (c) Model of PDDA’s action. (Reproduced from Mori et al. 2018)

166

T. Mori and T. Uchihashi

biochemical and biophysical reactions is of the sub-second or second order. The integration of such high-frequency events to the system might be a buffer for molecular noise (either extrinsic or intrinsic) and provides accuracy and stability in the “slow” dynamics of the circadian oscillator.

3.2.4

C-Terminal Tentacles of KaiC Hexamer Co-Operationally Bind to KaiA Dimer

KaiA is a domain-swapped, 2-fold-related homodimer and binds to KaiC hexamer to stimulate autokinase activity in the CII domain of KaiC (Vakonakis et al. 2004b; Garces et al. 2004; Uzumaki et al. 2004; Pattanayek et al. 2006; Kim et al. 2008). Biochemical and structural studies have revealed that the C-terminal α-helical bundle domains of KaiA dimers interact with the C-terminal tentacles of the KaiC hexamer. Each KaiA dimer can bind two KaiC tentacles (Egli and Johnson 2013; Pattanayek and Egli 2015). To investigate how a KaiA dimer interacts with the C-terminal tentacles of KaiC hexamer, we reconstituted KaiC hexamers having different numbers of the tentacles on its CII ring. We first dissociated the fulllength KaiC hexamer (KaiCAA) and the C-terminal truncated KaiC hexamer (KaiCAA-ΔC; lacking the residues 490–519) to monomers by removing ATP (Nishiwaki and Kondo 2012) and reconstituted hexamers by mixing the monomeric KaiCAA and KaiCAA-ΔC at different ratios of monomeric KaiCAA and KaiCAA-ΔC ([KaiC-AA:KaiC-AA-ΔC] ¼ [6:0], [5:1], [4:2], [3:3], [2:4], [1:5], [0.5:5.5], and [0:6]) in the presence of ATP. The reconstituted KaiC hexamers with different numbers of the tentacles were immobilized on the substrate, and the interactions with KaiA dimer were observed by HS-AFM. Figure 9a shows HS-AFM images capturing the interaction of those KaiC hexamers with KaiA. We observed a small fraction of KaiC hexamers transiently ( 2, meanwhile, damped oscillations occur when b = 1.8. (b, c) The dependency of amplitude and period on the parameter b (d–f) Morris-Lecar model reproduces neuronal oscillations. SNIC bifurcation occurs when the applied current I is approximately 40 and the value of other parameters are set as shown below. (d) Time evolution of V(t). (e, f) The dependency of amplitude and period on the parameter I. We used the following equations and values of parameters, CV = I
gL(V
EL)
gKn(V
EK)
gm(V)(V
ECa), n¯ = φ(n(V)
n)/τ(V), τ(V) = 1/cosh((V
V3)/2V4), m(V) = 1/2[1 + tanh ((V
V1)/V2)], n(V) = 1/2[1 + tanh((V
V3)/V4)], C = 20, φ = 0.067, gCa = 4, gK = 8, gL = 2, ECa = 120, EK =
84, EL =
60, V1 =
1.2, V2 = 18, V3 = 12, V4 = 17.4. The parameter values are adopted from (Moye and Diekman 2018)

that distribute across both self-sustained and driven oscillations. Then, the dependency of the oscillation period and amplitude on the control parameter can be thoroughly examined. In the past, the laboriousness of systematically changing the value of a parameter and the ambiguity of definition of oscillation amplitude have been obstacles to applying bifurcation theory to circadian rhythms unlike neural oscillations (Izhikevich 2007). However, the in vitro Kai oscillator enables us to overcome these difficulties. We shall see the first application of bifurcation theory to chronobiology in the next section.

Damped Oscillation in the Cyanobacterial Clock System

227

3 Low-Temperature MAKES In Vitro Rhythms Dampen Through Hopf Bifurcation Poikilotherms or plants generally lose their circadian rhythms under low temperature conditions. For example, circadian rhythms are not detected at 11.5  C in the marine dinoflagellate Lingulodinium polyedrum (formerly known as Gonyaulax, Hastings and Sweeney 1957; Njus et al. 1977; Roy et al. 2014). Cockroaches and fruit flies lose their circadian rhythms at around 20  C (Roberts 1962; Zimmerman 1969). Circadian rhythms in Neurospora are not observed at 10.5  C (Francis and Sargent 1979). The conditions at 4  C abolish circadian rhythms in some species of plants including Arabidopsis (Martino-Catt and Ort 1992; Ramos et al. 2005; Bieniawska et al. 2008). We built upon these preceding studies by focusing on the in vitro Kai oscillator (Murayama et al. 2017). We exposed the in vitro KaiABC clock to a range of temperatures. Bifurcation analyses can fit well to the reconstituted KaiC phosphorylation rhythm because the absolute oscillation amplitude of the rhythm can be directly measured with high precision, and changes of ambient temperature affect the dynamics of the Kai oscillator (Yoshida et al. 2009). We examined the period and amplitude of the in vitro KaiC phosphorylation rhythm from 17  C to 30  C. The oscillation period was relatively constant as the temperature diminished; on the other hand, the oscillation amplitude decreased monotonically (Fig. 2a, b). No oscillation in KaiC phosphorylation was observed below 19  C. These observations imply that the low temperature condition abolishes the reconstituted KaiC phosphorylation rhythm through Hopf bifurcation. Incidentally, Colin S. Pittendrigh came to a different conclusion for chillinginduced loss of rhythmicity (Pittendrigh 1976). He focused on the rhythms of some organisms after recovery from chilling and found they tend to begin from circadian time (CT) 12, the beginning of the subjective night (Pittendrigh 1976; Kondo and Tsudzuki 1980). He interpreted this result as the behavior of a “relaxation” type oscillator and considered that the system was stopped by low temperature at the beginning of the charge phase, which is reminiscent of the SNIC bifurcation. To check this hypothesis, we switched the temperature from low temperatures (4–16  C) to high temperatures (22–30  C). The rhythm resumed from CT18-CT1 when the temperature was increased from 4–16  C to 30  C and it resumed from CT22-CT3 when the temperature was increased from 16  C to 22–30  C. Unlike the general tendency, the phases of the in vitro clock under low temperature were not always arrested at CT12 but depended on the temperature before and after the stepup transition. The variety of the arrested phases can be explained as a temperaturedependent location of the stable fixed point created via Hopf bifurcation. As noted in the previous section, Hopf bifurcation implies the transition from a self-sustained to a damped oscillator. This interpretation predicts that a damped oscillation of KaiC phosphorylation would occur below the critical temperature. Consistent with that prediction, we found that the 30  C pulse followed by cold exposure yielded damped oscillations of KaiC phosphorylation, and the decay rate of

0.3 24

0.5

d

0 24

0 0.8

0 0.8

0 0.8

72

96

84

114

Incubation time (h)

54

Incubation time (h)

48

120

0

0.02

0.04

0.06

e

18

24

30

36

0

25

10

20

30

Temperature

20

40

0

0.1

0.2

50

30

Period of temperature cycle (h)

15

Critical temperature

c

0

24

0

0.04

f

0.3

0.3 0.5

0.5

48

72

96

10

20

30

40

50

Incubation time (h)

Period of temperature cycle (h)

3

Amplitude

120

Fig. 2 Damped oscillation under low temperature. (a) Low-temperature nullified the in vitro KaiC phosphorylation rhythm. KaiABC mixtures were incubated at 4  C for 24 h and then at the displayed temperature for 96 h. (b) Effect of temperature on period (pink) and amplitude (black) of the in vitro KaiC phosphorylation rhythm. The asymptotic standard error for the estimated period is displayed as an error bar. The blue line represents 19  C, which appears to be a critical temperature. (c) Damped oscillation of KaiC phosphorylation. Kai mixtures were incubated at constant temperatures of 16.7  C (black) and 18.7  C (blue). 30  C pulse for 12 h induced more evident damped oscillation. (d) Forced oscillation by cycles of temperatures lower than the critical temperature. The mixtures were incubated at 18.7  C for 24 h and then subjected to temperature cycles of 15 h 16.7  C and 15 h 18.7  C. (e) Resonance of the in vitro KaiC phosphorylation rhythms during the low-temperature cycles. The oscillation amplitude was measured by calculating the difference between the peak and trough level of phosphorylation ratio. The dotted line shows the averaged difference of KaiC phosphorylation ratio after 100 h at constant temperature of 16.7  C and 18.7  C shown in Fig. 2c. (f) Resonance of the Hatakeyama–Kaneko model. The parameter for temperature was periodically switched. The amplitude is defined as the difference between the maximum and minimum ratios of P-KaiC at steady state. Reprinted from Murayama et al. (2017)

Ratio of P-KaiC

0.8

Period (h) Amplitude

b

Ratio of P-KaiC Amplitude

a 228 H. Ito et al.

Damped Oscillation in the Cyanobacterial Clock System

229

the oscillation amplitude was higher at lower temperatures (Fig. 2c). The period of the damped oscillation was about 30 h at temperatures below the critical temperature. The fact that the oscillation periods just above and below the critical temperature do not change significantly supports the hypothesis of Hopf bifurcation rather than SNIC bifurcation.

4 Resonance of the Damped Oscillation of KaiC During Temperature Cycles The damped oscillation of KaiC phosphorylation can be modeled as a pendulum with friction. Such a pendulum can swing for a while after stimulation and continues to swing as long as a periodic external force is being applied. Even though the external force is small, the pendulum swings with a large amplitude when the period of the external force is close to the natural frequency of the pendulum. On the other hand, when the period of the external stimulus is not appropriate, the pendulum does not swing robustly. This phenomenon is known in physics as “resonance.” We reported the resonance of the KaiC phosphorylation rhythm with temperature cycles in Murayama et al. (2017). Temperature cycles of 16.7  C and 18.7  C were used as a weak external force. Under constant temperatures of 16.7  C or 18.7  C, the KaiC phosphorylation rhythm reaches an equilibrium state (Fig. 2c), and temperature cycles of 15 h at 16.7  C and of 15 h at 18.7  C (period ¼ 30 h) resulted in a forced oscillation of KaiC phosphorylation (Fig. 2d). It is possible that such forced oscillation of KaiC regulates the output pathway in vivo and functions as a semi-circadian clock even below the critical temperature. This hypothesis is analogous to the fact that no one can distinguish a pendulum with friction from a self-sustained oscillator while applying periodic external forces. Furthermore, it is possible that the amplitude of the cyanobacterial circadian rhythms can be maintained during winter at temperatures below the critical temperature by resonating with a daily variation, e.g., temperature or light. Moreover, we altered the period of the temperature cycle, hereafter denoted by T. Forced oscillations were observed when the temperature cycles were  24 h. The amplitudes of the KaiC phosphorylation rhythm were measured as a function of T. The amplitude of the forced oscillation was maximized when T ~ 30 h, which suggests temperature cycles caused resonance (Fig. 2e). We performed a numerical simulation to extract the essential conditions for resonance. The HatakeyamaKaneko model (Hatakeyama and Kaneko 2012) is one of the models that reproduces KaiC phosphorylation rhythms and explicitly incorporates the temperature dependence of the reaction rate. In fact, the model exhibits Hopf bifurcation when the temperature is diminished. In this model, the amplitude became higher as the frequency of the external force approached the approximate natural frequency of the Kai oscillator, meaning that this model reproduced the resonance (Fig. 2f).

230

H. Ito et al.

Moreover, a simpler model using the Stuart-Landau equation, reproduced the resonance, suggesting any damped oscillators undergoing Hopf bifurcation may generally exhibit resonance because the Stuart-Landau oscillator is a normal form, i.e., the representative of any oscillators near the Hopf bifurcation point (Guckenheimer and Holmes 1983). In addition, we numerically confirmed that a model in which oscillation disappears through SNIC bifurcation shows forced oscillations but does not resonate with a periodic stimulus. It should be noted that the concept of circadian resonance has been proposed in the context of enhanced growth and metabolic activity (Ouyang et al. 1998; Dodd et al. 2005; Lambert et al. 2016). These studies suggested a link between greater physiological advantage and matching periods between internal and external cycles. A possible explanation is that the biological clock could keep the most advantageous phase relationship to the daily environmental rhythms when the oscillation periods are close to each other. On the other hand, the enhancement of amplitude presented in this section may contribute another explanation for the advantage of circadian resonance.

5 Damped Oscillation in the Absence of KaiA Recently, we found another example of damped oscillations in Synechococcus, namely, when its kaiA gene was inactivated (Kawamoto et al. 2020). Since the first report of the kaiABC gene cluster in 1998 (Ishiura et al. 1998), it has been believed that all of the three kai genes are required for circadian oscillations. However, we recently found that the kaiA-inactivated strains (kaiA strains, hereafter) did not completely abolish the bioluminescence rhythm that monitors the kaiBC promoter activity in vivo but generated a faint damped rhythmicity (Fig. 3a). Since KaiA is important to activate the kaiBC promoter (Ishiura et al. 1998), the bioluminescence level is strikingly reduced in the kaiA strains. Thus, the peak-to-trough amplitude of the in vivo rhythm is strikingly lower than that of the wild-type strain (Fig. 3a). Because the amplitude of the residual rhythm was so low, it was interpreted to be arrhythmic, and therefore researchers including ourselves did not pay attention to the residual rhythms. Interestingly, when one of the coauthors (H. Iwasaki) reviewed old records of bioluminescence profiles in kaiA strains monitored 20 years ago, he found that the very weak, damped oscillation property was observable in these old data even though he had interpreted these data as indicating arrhythmia at that time. When either kaiB or kaiC was additionally nullified, the damped oscillation was abolished, indicating that the kaiA-less oscillation requires KaiB and KaiC functions. We employed the above-mentioned resonance experiment to verify the nature of the kaiA-less damped oscillation to test if forced external cycles with weak stimuli amplify the damped oscillation. We applied repetitive, weak (short 2 h) dark pulses with a period ranging from 16 to 32 h. As shown in Fig. 3b and c, external cycle (T) with a period of 24–26 was most effective to enhance amplitude of the

Damped Oscillation in the Cyanobacterial Clock System

a

1600

180

231

b

WT KaiA–

magnified

KaiA– 0 72

0 0

24

48

Hours in LL

c

Fig. 3 Damped oscillation in the absence of kaiA. (a) Bioluminescent profiles to monitor kaiBC promoter activity of the wild-type (WT, blue) and kaiA
(black and gray; magenta for magnified scale) strains in LL after two LD cycles. Bioluminescence is presented as counts per colony (c.p.c.). (b, c) Resonance of the kaiA-less damped oscillation in response to cyclic external cues. The bioluminescence profile in the kaiA
strain was monitored under LL, whereas the cells were exposed to four cycles of 2-h dark pulses (black bars) with a period of 24 h (b). After external cues (2-h dark pulses) were given for a period ranging from 16 to 32 h, the peak-to-trough ratio of the bioluminescence profile (as an index of amplitude) was calculated (c, n ¼ 3). Dot plots and error bars indicate mean and S.D., respectively. Reprinted from Kawamoto et al. (2020)

bioluminescence rhythm in the kaiA-less strain. This experiment established that the damped oscillation has its unique frequency which is close to that of the circadian oscillator. Although the mechanism to generate the kaiA-less oscillation still remains to be determined, several lines of evidence reported in Kawamoto et al. (2020) suggest that both TTFL and KaiB-KaiC physical interaction processes are involved. KaiA is known to activate the autophosphorylation activity of KaiC, and this function of KaiA is antagonized by KaiB. To test if the phosphorylation state of KaiC affects the kaiA-less rhythm, we substituted the phosphorylation sites (Ser-431 and Thr-432) of KaiC with glutamates (EE: mimicking hyperphosphorylated forms) in the kaiA-null background. This mutation is known to strikingly lengthen the period length with intact KaiA (Kitayama et al. 2008). By contrast, the kaiA-less rhythm was maintained in the kaiC[EE] background without lengthening its period length (even slightly shortened the period by 2 h to approximately 22 h). Introduction of other kaiC period mutations also had only minor effects on the period length

232

H. Ito et al.

+KaiA

Hourglass

Damped oscillator

Resonated Damped oscillator

–KaiA

Self-sustained oscillator

Fig. 4 Possible evolutionary process from hourglass to damped and self-sustained oscillators. Under LD conditions, an hourglass (timer) regulates diurnally or nocturnally driven passive fluctuations. A damped oscillator with natural frequency of the circadian range enables to regulate more coordinated temporal outputs. Especially under resonating conditions, the damped oscillator would regulate cellular functions, physiologically equivalent (or close) to the self-sustained oscillator. Examples of damped oscillations verified by resonance experiments include bioluminescence rhythm in the kaiA strain and in vitro KaiC phosphorylation cycle at a low temperature. Other candidates include KaiA-less, KaiBC-based systems in Prochlorococcus and Rhodopseudomonas to which resonance experiments should be addressed. Under annually changing photoperiodic conditions, a self-sustained oscillator with the intact KaiABC system is advantageous. Thus, evolution of the cyanobacteria-specific KaiA seems a key to transition from a damped oscillator to a self-sustained oscillator. Even after evolution of the self-sustained oscillator, some species in Prochlorococcus discharged the kaiA gene. It is possibly because kaiA-less damped oscillation is sufficient under relatively neutral (photoperiod-insensitive) light:dark conditions and to save energy for maintaining self-sustainability

of the kaiA-less oscillation. Therefore, the period length of the damped oscillation is unexpectedly robust against the previously known period altering mutations. kaiA is only found in cyanobacteria, while kaiB and kaiC genes are present not only in cyanobacteria but in other bacterial species, such as proteobacteria and Archaea. A phylogenetic tree analysis has proposed that kaiA is evolutionarily younger than kaiB and kaiC (Dvornyk et al. 2003). Our finding of the kaiA-less damped oscillation is consistent with an evolutionary model that a proto-circadian system evolved without KaiA as a KaiB–KaiC-based damped oscillatory system resonating to external cycles, and evolution of KaiA gave rise to the intact sustained oscillator (Johnson et al. 2017). As is mentioned in the chapters by Prof. Axmann and Prof. Johnson, the marine cyanobacteria Prochlorococcus marinus MED4 and PCC 9511 lack kaiA. The kaiA gene in these species was likely lost after completion of the intact kaiABC system. These species exhibit diurnal variations in transcription and propagation under light–dark (LD) cycles, while they fail to show sustained oscillation under continuous conditions (Holtzendorff et al. 2008). Moreover, the non-cyanobacterial purple bacterium Rhodopseudomonas palustris, harboring a kaiBC cluster without kaiA, also shows the diurnal but not free-running nitrogen fixation rhythm (Ma et al. 2016). Based on these observations, the authors have discussed the presence of a non-self-sustained timing system without kaiA functioning as a proto-clock system. In particular, Prof. Johnson discussed that Rhodopseudomonas may have a proto-clock system on kaiBC-based (kaiA-less) damped oscillation system (Johnson et al. 2017). Figure 4 represents a similar

Damped Oscillation in the Cyanobacterial Clock System

233

scheme of possible evolutionary transition of clock systems from the hourglass, the damped oscillator to the self-sustained oscillator. As mentioned above, we have established a protocol to verify damped oscillation by resonance experiments. Thus, we suggest to test if Prochlorococcus and/or Rhodopseudomonas show such a resonance effect to amplify its amplitude by external forcing cycles with a circadian range. Under resonated conditions, the damped oscillator would organize temporal cellular activities as does the sustained oscillator.

6 Evolution of Self-Sustained Oscillation We have posited that damped oscillators are an evolutionary precursor to selfsustained oscillators (Fig. 4). To theoretically examine the feasibility of this scenario, we adopted a gene regulatory network model for biological oscillators (Kobayashi et al. 2010). We randomly generated one-million networks with 5 nodes and 10 inhibitory regulations and classified them into three categories: self-sustained oscillators, damped oscillators, and non-oscillatory systems (hourglasses under LD cycles) (Fig. 5a; Seki & Ito, in preparation). We found that self-sustained oscillators were rare (1.3%) whereas damped oscillators were not uncommon (28.5%; Fig. 5b). Additional analyses indicated that a larger-dimensional system generated self-sustained oscillators more frequently but still uncommon in absolute terms (5.0% among the 10-node-20-regulation networks; Fig. 5b). Subsequently, by analogy to the emergence of the kaiA gene, we inserted a new node (node 6 in Fig. 5c) between two randomly chosen nodes (e.g., nodes 5 and 4; Fig. 5c). Six-node self-sustained oscillators were generated from 5-node damped oscillators twice as frequently as they are generated completely randomly, and when starting with hourglass timers, six-node self-sustained oscillators were generated only about one-half as frequently as they are generated randomly (2.2%; see the middle bar of Fig. 5b). In addition, only half of originally self-sustained oscillators retained their self-sustainability after the insertion of the sixth node, indicating fragility of selfsustained oscillators. Overall, these results suggest the possibility that the elaborate circadian oscillator possibly evolved from a damped oscillator.1 We suggest that a damped oscillator could regulate temporal cellular functions in a physiologically equivalent fashion (or close) to that of a self-sustained oscillator. If so, a new question is why most species have evolved to use self-sustained oscillators

1

Using the terms of bifurcation theory, conversion of a 5-node, non-self-sustained (or selfsustained) oscillator into a 6-node, self-sustained (or non-self-sustained) oscillator can be described as follows: We have a 6-node system and two control parameters that represent intensity of regulations from node 5 to 6 and from 6 to 4. Some type of bifurcation occurs as the parameter values are increased from 0 (no regulation) toward large values.

234

H. Ito et al.

b

a Randomly add 10 regulations

Put 5 nodes

Self-sustained Damped Hourglass

Classify expression pattern

1 5

1.3%

2.2%

28.5%

29.9%

2 4

5.0%

37.9%

3

Self-sustained

1 5 1

4

5

70.2%

2 3

57.1%

Damped

2 4

67.9%

1

3

5

2 4

5 nodes 6 nodes 10 nodes 10 regs 12 regs 20 regs

3

Hourglass

106 networks

c

Classified networks with 5 nodes and 10 regulations

Insert a new node between randomly chosed two nodes

1

1

5

5

2 4

6

3

4

5

2 4

5

6

3

2 4

3

Target node having an ideal profile

2 4

3

5

56.3%

3 1

1

Light 1

2.2% (chance level)

2

Self-sustained

d

Obtain percentage of self-sustained oscillators

4.4%

Damped

1

1 5

2 4

5

2 4

3

6

3

Hourglass

e

f Class I

1.3% Ratio of self-sustained oscillators

g Class II

Class I

Class II

Seasonal Hourglass

Self-sustained

Aseasonal

Trough at dusk

Trough at 12-h after dawn

Hourglass

Fig. 5 Simulations of a gene regulatory network model to consider evolution of self-sustained oscillation. (a) Schema of Seki & Ito (in preparation) simulations in which one million gene regulatory networks were randomly generated and classified into three categories according to their expression patterns: self-sustained oscillators, damped oscillators, and hourglasses. (b) Shares of self-sustained oscillators, damped oscillators, and hourglasses in the million networks. (c) Schema and results of simulations in which an additional node and regulations were inserted

Damped Oscillation in the Cyanobacterial Clock System

235

as their circadian systems? Self-sustained oscillators would generally drain more energy and resources for construction and maintenance (e.g., kaiA gene and proteins). We therefore expect that there must be a selective advantage for species to maintain a self-sustained circadian clock. Only a few theoretical studies have addressed the advantages and disadvantages of having self-sustained circadian oscillators. Pittayakanchit et al. (2018) suggested that self-sustained oscillators have an advantage over the non-self-sustained ones under noisy LD cycles in terms of synchrony of cellular clocks. Our study (Seki & Ito, in preparation) suggested that self-sustained oscillators are better at sensing seasonal change in day length. The major assumption in the study was that there is an ideal expression profile for one node (node 2 in Fig. 5d). The network model (Kobayashi et al. 2010) was revised so that it could handle the light-driven externally forced oscillation (Fig. 5d). After this revision, all randomly generated networks (including those showing self-sustained oscillations without light input) showed oscillations whose period length was the same as that of LD cycles. Then, we measured fitness of a network under given LD cycles by the closeness between an ideal profile under that LD cycle and the profile realized by node 2. Several types of ideal profiles were examined, and they were classified into two groups according to whether the form of profile depends on the day length (class I) or not (class II). An example of profiles belonging to each class is shown in Fig. 5e–f. Both examples share the property of having a peak at dawn, whereas the class I and class II examples have a trough at dusk and at 12 h after dawn, respectively. An instance of genes that show expression patterns independent from day length is GIGANTEA (GI) in plants, which is known to play an important role in day-length-dependent flowering events (Yeang 2013). The analyses further considered two types of environments: aseasonal and seasonal environments. Fitness of network under the aseasonal environment was calculated in 12L:12D cycle, whereas final fitness of network under the seasonal environment was defined as the mean of fitness calculated in several LD cycles (e.g., 8L:16D, 10L:14D, . . ., and 16L:8D) under a seasonal environment. It turned out that the network with the highest fitness among all the checked networks showed self-sustained oscillation in LL or DD only with class II ideal profile and under seasonal environments (the upper right cell in Fig. 5g; Seki & Ito, in preparation). This suggests that species may evolve to have a self-sustained oscillator when they need to detect change of the day length in a seasonal environment. The result may explain the evolutionary loss of self-sustained oscillation in Prochlorococcus marinus whose main habitats are rather aseasonal, tropical zones.

 ⁄ Fig. 5 (continued) between randomly chosen two nodes. (d) An example of networks involving regulation of light signal. (e, f) Examples of ideal profiles for the target node of class I and that of class II. (g) Profiles of the networks showing the highest fitness among all examined 10-node-20regulation in each condition. The profile of the target node is drawn with a thick blue line. Note that class I and class II are invariant under an aseasonal environment in which networks are exposed to only 12L:12D cycles (see middle panels of Fig. 5e–f)

236 Fig. 6 Self-sustained oscillation and damped oscillation in a parameter space of the circadian clock system

H. Ito et al.

Non-oscillatory system (Hourglass)

Damped oscillaion

Self-sustained oscillaion Chilling –kaiA Evolution?

7 Summary The circadian systems obtained self-sustainability during the process of evolution. We presented here two cases where the organisms lose self-sustainability. The selfsustained rhythm of the KaiC phosphorylation in a test tube turned to damped oscillation under chilly conditions. The absence of kaiA transformed the bonafide circadian rhythm of bioluminescence into a damped oscillation. These damped rhythms recovered their amplitude by resonating with repetitive external forcing, such as temperature cycles or LD cycles. Our numerical results suggest that although self-sustained oscillators are rare in a parameter space of the biological systems, seasonal variation in daylength pushes the clock system into a self-sustained regime through an evolutionary process (Fig. 6). There have not been many reports regarding the transition from self-sustained to damped oscillators in chronobiology. Even when considering damped oscillations, they have usually been interpreted as a desynchronization of self-sustained oscillators. One of the reasons might be chronobiologists’ tendency to focus on period and phase rather than amplitude. The definition of amplitude is mathematically concrete but subtle in a laboratory. For example, damping of oscillation amplitude in a living system in an artificial laboratory setting often happens due to diminishing biochemical activity of the output system of the clock, depletion of nutrients, overcrowding, etc. Or as noted above, desynchronization of self-sustained oscillation causes decrease of amplitude; thus amplitude depends on which level of organization we are examining. The preference for period and amplitude and avoidance of ambiguous amplitude has brought success to chronobiology. For instance, the theory of entrainment focuses on the dynamics of only phase. Mutagenetic screening based on period contributed to cloning of the clock genes. In turn, the phenomena involved in amplitude, such as damped oscillation, have attracted less attention. We believe that amplitude is a largely unexplored field in chronobiology. Moreover, the cyanobacterial system should provide an ideal tool to explore the field relating to amplitude because we can directly observe the behavior of the core circadian oscillator, the KaiC phosphorylation rhythm, enabling a much clearer definition of

Damped Oscillation in the Cyanobacterial Clock System

237

amplitude. It is our hope that this chapter may pave the way for novel discoveries in chronobiology based on amplitude. Acknowledgments We thank Takao Kondo, Kumiko Miwa, Isao Tokuda, Hiroshi Kori, and Chiaki Oshima for fruitful discussion and collaboration during the course of our studies. This study was supported in part by a Grant-in-Aid for Scientific Research (KAKENHI) from JSPS (Grant nos. 18K19349 and 23657138 to H. Iwasaki, 18H05474 to H. Ito, 16J40136 to Y.M.) and by the Education and Research Center for Mathematical and Data Science (Kyushu University to H. Ito).

References Bieniawska Z, Espinoza C, Schlereth A et al (2008) Disruption of the Arabidopsis circadian clock is responsible for extensive variation in the cold-responsive transcriptome. Plant Physiol 147:263–279 Bünning E (1931) Untersuchungen über die autonomen tagesperiodischen Bewungen der Primärblätter von Phaseolus multiflorus. Jahrb Wiss Bot 75:439–480 Bünning E (1960) Opening address: biological clocks. Cold Spring Harb Symp Quant Biol 25:1–9 Bünning E (1964) The physiological clock: endogenous diurnal rhythms and biological chronometry. Springer-Verlag, Berlin Bünning E (1975) Wilhelm Pfeffer: Apotheker, chemiker, botaniker, physiologe 1845–1920. English edition: Bünning E (1989). Ahead of his time: Wilhelm Pfeffer, early advances in plant biology (trans: Pfeffer HW), Carleton University Press, Ottawa Bünning E (1977) Fifty years of research in the wake of Wilhelm Pfeffer. Annu Rev Plant Physiol 28:1–23 Bünning E, Stern K (1929) Über die tagesperiodischen Bewegungen der Primärblätter von Phaseolus multiflorus I. Der Einfluss der Temperatur auf die Bewegungen. Ber d bot Ges 47:565–584 Darwin C, Darwin F (1880) The power of movement in plants. John Murray, London Dodd AN, Salathia N, Hall A et al (2005) Plant circadian clocks increase photosynthesis, growth, survival, and competitive advantage. Science 309:630–633 Dvornyk V, Vinogradova O, Nevo E (2003) Origin and evolution of circadian clock genes in prokaryotes. Proc Natl Acad Sci U S A 100:2495–2500 Francis CD, Sargent ML (1979) Effects of temperature perturbations on circadian conidiation in Neurospora. Plant Physiol 64:1000–1004 Gould PD, Domijan M, Greenwood M et al (2018) Coordination of robust single cell rhythms in the Arabidopsis circadian clock via spatial waves of gene expression. elife 7:e31700 Grobbelaar N, Huang T, Lin H et al (1986) Dinitrogen-fixing endogenous rhythm in Synechococcus RF-1. FEMS Microbiol Lett 37:173–177 Guckenheimer J, Holmes PJ (1983) Nonlinear oscillations, dynamical systems, and bifurcations of vector fields, applied mathematical sciences, vol 42. Springer-Verlag, New York Hastings JW, Sweeney BM (1957) On the mechanism of temperature independence in a biological clock. Proc Natl Acad Sci U S A 43:804–811 Hatakeyama TS, Kaneko K (2012) Generic temperature compensation of biological clocks by autonomous regulation of catalyst concentration. Proc Natl Acad Sci U S A 109:8109–8114 Holtzendorff J, Partensky F, Mella D et al (2008) Genome streamlining results in loss of robustness of the circadian clock in the marine cyanobacterium Prochlorococcus marinus PCC 9511. J Biol Rhythm 23:187–199 Ishiura M, Kutsuna S, Aoki S et al (1998) Expression of a gene cluster kaiABC as a circadian feedback process in cyanobacteria. Science 281:1519–1523

238

H. Ito et al.

Ito H, Kageyama H, Mutsuda M et al (2007) Autonomous synchronization of the circadian KaiC phosphorylation rhythm. Nat Struct Mol Biol 14:1084–1088 Izhikevich EM (2007) Dynamical systems in neuroscience: the geometry of excitability and bursting. MIT Press, Cambridge Johnson CH, Zhao C, Xu Y et al (2017) Timing the day: what makes bacterial clocks tick? Nat Rev Microbiol 15:232–242 Kawamoto N, Ito H, Tokuda IT et al (2020) Damped circadian oscillation in the absence of KaiA in Synechococcus. Nat Commun 11:2242 Kitayama Y, Nishiwaki T, Terauchi K et al (2008) Dual KaiC-based oscillations constitute the circadian system of cyanobacteria. Genes Dev 22:1513–1521 Kobayashi Y, Shibata T, Kuramoto Y et al (2010) Evolutionary design of oscillatory genetic networks. Eur Phys J B 76:167–178 Kondo T, Tsudzuki T (1980) Phase progress under low temperature treatment of the potassium uptake rhythm in a duckweed, Lemna gibba G3. Plant Cell Physiol 21:95–103 Kondo T, Strayer CA, Kulkarni RD et al (1993) Circadian rhythms in prokaryotes: luciferase as a reporter of circadian gene expression in cyanobacteria. Proc Natl Acad Sci U S A 90:5672–5676 Lambert G, Chew J, Rust MJ (2016) Costs of clock-environment misalignment in individual cyanobacterial cells. Biophys J 111:883–891 Ma P, Mori T, Zhao C et al (2016) Evolution of KaiC-dependent timekeepers: a proto-circadian timing mechanism confers adaptive fitness in the purple bacterium Rhodopseudomonas palustris. PLoS Genet 12:e1005922 Martino-Catt S, Ort DR (1992) Low temperature interrupts circadian regulation of transcriptional activity in chilling-sensitive plants. Proc Natl Acad Sci U S A 89:3731–3735 Mihalcescu I, Hsing W, Leibler S (2004) Resilient circadian oscillator revealed in individual cyanobacteria. Nature 430:81–85 Moye M, Diekman C (2018) Data assimilation methods for neuronal state and parameter estimation. J Math Neurosci 8:11 Murayama Y, Kori H, Oshima C et al (2017) Low temperature nullifies the circadian clock in cyanobacteria through Hopf bifurcation. Proc Natl Acad Sci U S A 114:5641–5646 Nagoshi E, Saini C, Bauer C et al (2004) Circadian gene expression in individual fibroblasts cellautonomous and self-sustained oscillators pass time to daughter cells. Cell 119:693–705 Nakajima M, Imai K, Ito H et al (2005) Reconstitution of circadian oscillation of cyanobacterial KaiC phosphorylation in vitro. Science 308:414–415 Njus D, McMurry L, Hastings JW (1977) Conditionality of circadian rhythmicity: synergistic action of light and temperature. J Comp Physiol 117:335–344 Ouyang Y, Andersson CR, Kondo T et al (1998) Resonating circadian clocks enhance fitness in cyanobacteria. Proc Natl Acad Sci U S A 95:8660–8664 Pfeffer W (1915) Beiträge zur Kenntnis der Entstehung der Schlafbewegungen. Ber. d. math.-phys. KI. d. Koenigl. Sächs. Gesellsch. d. Wissensch 34 (I–VI):1–154 Pittayakanchit W, Lu Z, Chew J et al (2018) Biophysical clocks face a trade-off between internal and external noise resistance. elife 7:e37624 Pittendrigh CS (1976) Circadian clocks: what are they? In: Hastings JW, Schweiger HG (eds) The molecular basis of circadian rhythms. Abakon Verlagsgesellschaft, Berlin, pp 11–48 Ramos A, Pérez-Solís E, Ibáñez C et al (2005) Winter disruption of the circadian clock in chestnut. Proc Natl Acad Sci U S A 102:7037–7042 Roberts SKDF (1962) Circadian activity rhythms in cockroaches. II Entrainment and phase shifting. J Cell Comp Physiol 59:175–186 Roy S, Letourneau L, Morse D (2014) Cold-induced cysts of the photosynthetic dinoflagellate Lingulodinium polyedrum have an arrested circadian bioluminescence rhythm and lower levels of protein phosphorylation. Plant Physiol 164:966–977 Strogatz SH (1994) Nonlinear dynamics and chaos: with applications to physics, biology, chemistry, and engineering. Addison-Wesley, Reading, MA Sweeney B (1987) Rhythmic phenomena in plants. Academic Press, San Diego, CA

Damped Oscillation in the Cyanobacterial Clock System

239

Tazawa H (2009) Biological clock coming out from beans: the story of Erwin Bünning. (in Japanese) Gakkai Shuppan, Tokyo Ukai H, Kobayashi TJ, Nagano M et al (2007) Melanopsin-dependent photo-perturbation reveals desynchronization underlying the singularity of mammalian circadian clocks. Nat Cell Biol 9:1327–1334 Welsh DK, Yoo SH, Liu AC et al (2004) Bioluminescence imaging of individual fibroblasts reveals persistent, independently phased circadian rhythms of clock gene expression. Curr Biol 14:2289–2295 Wever R (1965) Pendulum versus relaxation oscillation. In: Aschoff J (ed) Circadian clocks. NorthHolland Publication, Amsterdam, pp 74–83 Yeang HY (2013) Solar rhythm in the regulation of photoperiodic flowering of long-day and shortday plants. J Exp Bot 64:2643–2652 Yoshida T, Murayama Y, Ito H et al (2009) Nonparametric entrainment of the in vitro circadian phosphorylation rhythm of cyanobacterial KaiC by temperature cycle. Proc Natl Acad Sci U S A 106:1648–1653 Zimmerman WF (1969) On the absence of circadian rhythmicity in Drosophila pseudoobscura pupae. Biol Bull 136:494–500

Roles of Phosphorylation of KaiC in the Cyanobacterial Circadian Clock Taeko Nishiwaki-Ohkawa

Abstract Phosphorylation of circadian clock proteins is ubiquitously found in many organisms. The cyanobacterial circadian clock can be reconstituted in vitro by mixing three clock proteins, KaiA, KaiB, KaiC, and ATP, in which the phosphorylation state of KaiC exhibits robust circadian rhythm. In contrast to eukaryotic clock proteins, sometimes containing tens of phosphorylation sites, KaiC has only two phosphorylation sites, serine 431 and threonine 432, which enables the full understanding of the functions of clock protein phosphorylation in the circadian oscillation. Unusual properties of KaiC have been revealed with the use of the reconstituted circadian clock. KaiC has extremely slow but stable ATPase activity, which is essential for generating circadian oscillation. KaiC can phosphorylate and dephosphorylate itself by the phosphotransfer reaction between KaiC-bound adenine nucleotide and the hydroxyl group on Ser and Thr residues. KaiC phosphorylation has critical roles in sustaining robust oscillation of protein-based circadian clock, whose period is defined by ATPase activity of KaiC, by regulating the interaction among Kai proteins. KaiC phosphorylation also serves as one of the major output pathways connecting the protein-based oscillator and genome-wide transcription.

1 Introduction Posttranscriptional modification of clock proteins, especially phosphorylation, is important for the regulation of the circadian clock. Phosphorylation of the clock protein was first described in Drosophila PERIOD (PER) in 1994 (Edery et al. 1994). Since then, an increasing number of clock proteins in eukaryotes have been shown to be phosphorylated (Rutila et al. 1996; Lee et al. 1998; Garceau et al. 1997; T. Nishiwaki-Ohkawa (*) Laboratory of Animal Integrative Physiology, Graduate School of Bioagricultural Sciences, Nagoya University, Nagoya, Japan Division of Biological Science, Graduate School of Science, Nagoya University, Nagoya, Japan Institute of Transformative Bio-molecules (WPI-ITbM), Nagoya University, Nagoya, Japan e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_13

241

242

T. Nishiwaki-Ohkawa

Schafmeier et al. 2006; Lee et al. 2001; Camacho et al. 2001; Takano et al. 2000; Sanada et al. 2004; Harada et al. 2005). The mechanism of eukaryotic circadian clock is understood as a transcriptional-translational feedback loop (TTFL), in which the transcription of a clock gene is repressed by its own translation product (Takahashi 2017; McClung 2019; Loros 2020; Dubowy and Sehgal 2017). The phosphorylation levels of clock proteins often undergo daily oscillations, regulating their stability, interaction with other clock proteins, and nuclear translocation, which serves in TTFL as a process of period regulation (Hirano et al. 2016; Diernfellner and Brunner 2020; Crane and Young 2014). In eukaryotes, several protein kinases are involved in clock protein phosphorylation. Among these, Drosophila DOUBLETIME and its homologs, including mammalian casein kinase 1δ/ɛ (CK1δ/ɛ) and Neurospora CK1a, are chiefly responsible for regulation of circadian period (Yang et al. 2017). Cyanobacterial circadian study has lagged far behind that of eukaryotes. It was in 1998 when a gene cluster kaiABC, composed of kaiA and kaiBC operon, was identified as clock genes in Synechococcus elongatus PCC7942 (Ishiura et al. 1998). It was postulated that TTFL generates circadian oscillation in cyanobacteria as well since kaiBC operon transcription was activated and inhibited by KaiA and KaiC, respectively (Ishiura et al. 1998). TTFL had been thought to be the only mechanism to generate circadian oscillation in all organisms until 2005, when in vitro reconstitution of the circadian clock was reported by mixing KaiA, KaiB, KaiC, and ATP (Nakajima et al. 2005). This finding is extremely important because not only it is the first and so far the only example of reconstitution of circadian oscillation clearly independent of transcription and translation, but it also has accelerated interdisciplinary research of circadian oscillation involving researchers in another field such as physical and mathematical sciences. This chapter gives an overview of studies on KaiC phosphorylation in the past two decades, focusing on its roles in the protein-based oscillator composed of three Kai proteins and in the cyanobacterial whole circadian system. At the end of the chapter, future perspectives will be discussed.

2 Discovery of KaiC Phosphorylation All three kai genes are essential for the circadian clock since the disruption of any of these genes results in a loss of rhythmicity in promoter activities of kaiA, kaiBC, and psbAI, a photosynthesis-related gene; however, their mechanism of action was hard to imagine from the primary structures of their translation products (Ishiura et al. 1998). At a glance, KaiA and KaiB do not have any motifs that can be clues to study their functions. KaiC has no homology with eukaryotic clock proteins but comprises a duplicated structure with N-terminal and C-terminal halves called CI and CII domains, respectively, both of which have a phosphate-binding loop (P-loop), also known as Walker A motif (Ishiura et al. 1998), characteristic for NTPases (Walker et al. 1982; Yoshida and Amano 1995). KaiC displayed ATP-binding activity

Roles of Phosphorylation of KaiC in the Cyanobacterial Circadian Clock

243

in vitro as expected (Nishiwaki et al. 2000), and the CI domain had higher affinity for ATP than CII (Nishiwaki et al. 2000; Hayashi et al. 2003). On the contrary, autophosphorylation of KaiC was serendipitously discovered when recombinant KaiC was mixed with radioactive ATP (Nishiwaki et al. 2000). KaiC is also phosphorylated in a cyanobacterial cell, and the phosphorylation levels of KaiC exhibit circadian rhythm, which is higher in subjective nights and lower in subjective days (Iwasaki et al. 2002). Nonetheless, the relationship between in vitro autophosphorylation and in vivo phosphorylation was yet to be elucidated.

3 Relationship Between KaiC Phosphorylation and the Interaction Among Kai Proteins The physical association among three Kai proteins was first addressed by the yeast two-hybrid system, and it was found that KaiC interacts with both KaiA and KaiB (Iwasaki et al. 1999). Disruption and overexpression of kaiA in cyanobacteria decreases and increases KaiC phosphorylation (Iwasaki et al. 2002; Xu et al. 2003), respectively, whereas the KaiC phosphorylation level is increased in the kaiB-null strain (Kitayama et al. 2003; Xu et al. 2003), showing that KaiA and KaiB are the activator and the inhibitor of KaiC phosphorylation, respectively, in cyanobacterial cells. Interactions among Kai proteins also influence the phosphorylation state of KaiC in vitro (Iwasaki et al. 2002; Xu et al. 2003). Importantly, KaiC shows not only autophosphorylation but also autodephosphorylation activities (Kitayama et al. 2003; Xu et al. 2003). With solely KaiC, autodephosphorylation is favored over autophosphorylation (Iwasaki et al. 2002; Kitayama et al. 2003; Xu et al. 2003), whereas in the presence of KaiA, autophoshorylation is promoted. The effect of KaiB is somewhat complicated; with solely KaiB, there is no effect. In the presence of both KaiA and KaiB, the phosphorylation level of KaiC is initially increased with time; however, within a few hours, autodephosphorylation is favored over autophosphorylation, resulting in reduced phosphorylation levels (Kitayama et al. 2003). Contrarily, the phosphorylation state of KaiC also influences the interactions among Kai proteins. Amounts of KaiC-interacting KaiA and KaiB display circadian rhythm in cyanobacteria, both of which are increased in subjective nights when KaiC is highly phosphorylated (Kitayama et al. 2003), and hetero-multimeric protein complexes of about 600 kD, comprising KaiA, KaiB and KaiC, are formed in subjective nights (Kageyama et al. 2003). Even though these findings seem to contradict the fact that KaiA enhances KaiC phosphorylation, they explain the non-linear behavior of KaiC phosphorylation well in the presence of KaiA and KaiB if interpreted as follows: KaiA interacts with non-phosphorylated KaiC weakly and repeatedly to enhance autophosphorylation. KaiB binds specifically to phosphorylated KaiC and promotes autodephosphorylation by antagonizing KaiA

244

T. Nishiwaki-Ohkawa

through the recruitment of KaiA to KaiB-KaiC complex. This model formulates two ways of KaiA-KaiC binding, either acceleratory or inhibitory to autophosphorylation (Kitayama et al. 2003).

4 ATP-Binding Sites Located at the Subunit Interfaces of KaiC Hexamer KaiC exists as a homo-hexamer in the presence of ATP (Mori et al. 2002; Hayashi et al. 2003). X-ray crystallography revealed that the KaiC hexamer consists of CI and CII double rings with a total of 12 ATPγS, an analog of ATP, bound between individual subunits in both CI and CII halves (Pattanayek et al. 2004). The C-terminal tail of KaiC with around 25 amino acids in length protrudes from the CII ring (Pattanayek et al. 2006). P-loop constitutes a part of the ATP-binding pocket at the subunit boundaries of CI and CII. Catalytic carboxylates, glutamate 78 (E78) in CI and glutamate 319 (E319) in CII, are involved in the coordination of magnesium ions (Pattanayek et al. 2004, 2014). Mutations of E78 and E319 to glutamine (E78Q and E319Q) revealed that ATP binding to CI contributes more to hexamer formation than that to CII (Hayashi et al. 2004). Autophosphorylation sites were identified as serine 431 (S431) and threonine 432 (T432) using mass spectrometry (Nishiwaki et al. 2004). X-ray crystallography revealed that they are located in the subunit boundary of the CII ring within a 10 Å radius centered at the γ-phosphate of the nucleotide binding on the opposite subunit (Xu et al. 2004). Based on these findings, it was proposed that autophosphorylation and autodephosphorylation occur at the subunit interface of CII. This theory was later validated with the use of a hetero-hexamer formed by mixing KaiC monomer of non-phosphorylatable KaiC containing both S431 to alanine (S431A) and T432 to alanine (T432A) mutations (KaiC-AA) and that containing a catalytically inactive E318Q mutation (Kitayama et al. 2013). In a cyanobacterial strain expressing KaiC-AA, the phosphorylation of KaiC is completely abolished (Nishiwaki et al. 2004). Thus, it is most likely that phosphorylation of KaiC in cyanobacteria is also carried out by autophosphorylation/autodephosphorylation.

5 In Vitro Reconstitution of a Circadian Oscillator In 2005, a phenomenon that is contrary to TTFL was reported. In photoautotrophic Synechococcus, de novo transcription and translation are dramatically repressed under the constant dark (DD) condition (Doolittle 1979). Nonetheless, a robust circadian oscillation in KaiC phosphorylation is observed even in DD, showing that KaiC phosphorylation can persist even without transcription and translation (Tomita et al. 2005). In the same year, conclusive evidence was acquired.

Roles of Phosphorylation of KaiC in the Cyanobacterial Circadian Clock

245

Self-sustained circadian oscillation in KaiC phosphorylation was reconstituted in a test tube by mixing 1.2 μM KaiA, 3.5 μM KaiB, and 3.5 μM KaiC with 1 mM ATP (standard condition), clearly independent of TTFL (Nakajima et al. 2005). This truly protein-based oscillation is sustained for at least 10 days with no damping, is temperature compensated in the range between 25  C and 35  C, and can be entrained to temperature cycles, meeting all three criteria of circadian rhythms (Nakajima et al. 2005; Ito et al. 2007; Yoshida et al. 2009). Moreover, when mixing reconstituted oscillators with various phases, they are not averaged out to a constant phosphorylation level but are rapidly synchronized, resulting in a robust oscillation with a new phase whose amplitude is comparable to the original ones (Ito et al. 2007). This phenomenon is thought to be the basis for resilience of the cyanobacterial circadian clock against fluctuations originated from various cellular events including cell growth and cell division. Concentrations of KaiA and KaiB relative to KaiC influence the reconstituted oscillation in different manners; both period and amplitude are sensitive to KaiA concentration, whereas KaiB concentration does not influence them as long as it is higher than the threshold value below which there is oscillation (Nakajima et al. 2010). Time-dependent protein complex formation in vitro occurs in a similar manner with the cyanobacterial cell (Kageyama et al. 2006). Of note, by gel filtration chromatography, a peak containing KaiB about 25 kD in size is detected throughout a day (Kageyama et al. 2006), suggesting that a part of KaiB is not involved in oscillation under the standard condition. This finding is consistent with the existence of a threshold value in KaiB concentration and suggests that KaiB binds to KaiC in a specific stoichiometry, which is later validated by structural analyses. Since reconstitution of the circadian clock was achieved not based on any theoretical predictions but on the empirically determined composition of the reaction mixture, little was known as regards its mechanism. Researchers in different fields were interested in this mystery.

6 Sequential Phosphorylation of S431 and T432 Numerous proteins have multiple phosphorylation sites, often existing as clusters, that regulate a variety of cellular processes. If phosphorylation occurs on a residue in a cluster, it facilitates the phosphorylation of another site located nearby, possibly by increasing the effectiveness of the cluster as a substrate, resulting in all phosphorylation sites in the cluster to be sequentially phosphorylated (Salazar and Höfer 2009). Phosphorylation of mammalian PER2 by CK1 has been intensively studied, and it was found that the stability of PER2 is tightly regulated by sequential phosphorylation. (Zhou et al. 2015; Narasimamurthy et al. 2018; Philpott et al. 2020). KaiC has two phosphorylation sites S431 and T432, existing next to each other (Nishiwaki et al. 2004). With the use of mass spectrometry, high-resolution SDS-PAGE, and pulse labeling of KaiC in the reconstituted protein-based oscillator with [γ-32P] ATP, these sites were shown to be phosphorylated/dephosphorylated in

246

T. Nishiwaki-Ohkawa (ii) Phosphorylation of S431

P

P

P

S

T

T

S

(iii) Dephosphorylation of T432

P

S

P

P

T

S S

(i) Phosphorylation of T432

T

T

S

T

(iv) Dephosphorylation of S431

Fig. 1 Sequential autophosphorylation/autodephosphorylation of KaiC. Phosphorylation/dephosphorylation of KaiC proceeds in a fixed order as follows: (i) phosphorylation of T432, (ii) phosphorylation of S431 producing the double-phosphorylated form, (iii) dephosphorylation of T432, and (iv) dephosphorylation of S431. In every step, whichever reaction occurs, either autophosphorylation or autodephosphorylation, on one site is controlled by the phosphorylation state of the other

a fixed order as follows: (1) phosphorylation of T432, (2) phosphorylation of S431 producing the double-phosphorylated form, (3) dephosphorylation of T432, and (4) dephosphorylation of S431(Nishiwaki et al. 2007). This four-step reaction generates four forms of KaiC with different phosphorylation states as follows: T432-phosphorylated (SpT), T432- and S431-phosphorylated (pSpT), S431phosphorlated (pST), and non-phosphorylated (ST) (Nishiwaki et al. 2007; Rust et al. 2007) (Fig. 1). To reveal the underlying mechanism of this sequential reaction, recombinant KaiC containing a mutation mimicking constitutive phosphorylation on one of the two sites was generated. In KaiC-SE, mimicking the constitutive phosphorylation on T432, S431 is highly phosphorylated and cannot be dephosphorylated even in the absence of KaiA, suggesting that phosphorylated T432 facilitates and inhibits phosphorylation and dephosphorylation of S431, respectively. On the other hand, in KaiC-DT, mimicking constitutive phosphorylation on S431, the rate of phosphorylation on T432 dramatically decreased, and T432 dephosphorylation occurs even in the presence of KaiA (Nishiwaki et al. 2007). Likewise, in KaiC-AT, mimicking constitutive dephosphorylation of S431, facilitates phosphorylation of T432, whereas in KaiC-SA, mimicking constitutive dephosphorylation of T432, phosphorylation of S431 is reduced (Nishiwaki et al. 2007), showing that the phosphorylation state of one site influences the rate of phosphorylation/dephosphorylation reaction on the other, resulting in sequential phosphorylation/dephosphorylation of KaiC (Fig. 1). Circadian accumulation profiles of ST-, SpT-, pSpT-, and pST-KaiC were also reproduced by mathematical modeling constrained by the rate of interconversion among these forms (Rust et al. 2007). The interaction among KaiA, KaiB, and KaiC is influenced by the phosphorylation state of S431. In the reconstituted oscillator, the complex comprising three Kai

Roles of Phosphorylation of KaiC in the Cyanobacterial Circadian Clock

247

proteins is formed, which coincides with S431-phosphorlation of KaiC (Nishiwaki et al. 2007; Rust et al. 2007). Moreover, KaiC-DE, mimicking the doubly phosphorylated state, binds strongly to both KaiA and KaiB, whereas KaiC-AA, mimicking the non-phosphorylated state, exhibits a faint signal of KaiA binding (Nishiwaki et al. 2007). When the reconstituted oscillator was pulse-labeled with [γ-32P] ATP followed by gel-filtration chromatography, incorporation of phosphate was not detected from the large hetero-multimeric peak but from the KaiC homohexamer peak (Nishiwaki et al. 2007). From the findings obtained so far, an earlier model of how KaiA and KaiB regulate the phosphorylation state of KaiC was revised as follows: KaiA repeatedly interacts with KaiC hexamer to increase the phosphorylation of T432 and then S431. When S431 is sufficiently phosphorylated, KaiB interacts with KaiC, and KaiA is captured by the KaiB-KaiC complex, which interferes with KaiA to facilitate autophosphorylation of KaiC (Nishiwaki et al. 2007, Rust et al. 2007).

7 Discovery of ATPase Activity of KaiC Although KaiC exhibits an ATPase-like structure, ATPase activity of KaiC was reported much later than autophosphorylation. KaiC shows an extremely low but stable ATPase activity of 15 molecules of ATP/ KaiC monomer/day at 1 mM ATP, which is invariant against temperature within a physiological range, and in the presence of KaiA and KaiB, it displays a circadian rhythm (Terauchi et al. 2007). Although the accurate contribution of CI and CII domains to the total ATPase activity is yet to be elucidated, the contribution of CI seems to be higher than that of CII, since CII-truncated KaiC retains around 70% of the activity of the full length one (Terauchi et al. 2007). The structure of CI ATP-binding site was visualized by X-ray crystallography at a resolution better than 2.0 Å, and it was found that the position of the lytic water is unfavorable for in-line attack on the γ-phosphate of ATP, consistent with the slow ATPase activity (Abe et al. 2015). It is important to note that the ATPase activities of wild-type KaiC and five period-mutant proteins are directly proportional to their in vivo circadian frequencies (Terauchi et al. 2007). Hydrostatic pressure increases the frequency of the circadian, caused by pressureinduced acceleration of ATPase activity (Kitahara et al. 2019). Furthermore, KaiC-AA, in which S431 and T432 are substituted to alanine, has higher ATPase activity, whereas in phosphorylation-mimicking KaiC-DE, containing S431D and T432E double mutations, ATPase activity is decreased (Terauchi et al. 2007). Based on these findings, it is concluded that ATPase activity of KaiC serves as the pacemaker, responsible for period determination and temperature compensation, regulating the cyanobacterial circadian system, and that the autophosphorylation/ autodephosphorylation cycle is coupled to ATPase activity (Terauchi et al. 2007). How the ATPase activity of KaiC defines the period of protein-based oscillator, which is well compensated against temperature, is detailed in another chapter of this book.

248

T. Nishiwaki-Ohkawa

8 An Unusual Mechanism of KaiC Autodephosphorylation Various kinases, which transfer γ-phosphate of ATP to their substrate, such as sugars, nucleotides, and certain kinds of proteins, adopt P-loop NTPase fold (Leipe et al. 2003). Thus, it is not so strange that KaiC exhibits an autokinase activity. On the other hand, it is hard to explain why KaiC shows an autophosphatase activity since KaiC shows no homology with protein phosphatases that remove the phosphate moiety from phosphorylated Ser, Thr, and Tyr residues by hydrolysis (Shi 2009). When the autodephosphorylation reaction was followed using 32P-phosphorylated KaiC, transient formation of [32P] ATP was unexpectedly detected, preceding the accumulation of 32P inorganic phosphate (Nishiwaki and Kondo 2012). Kinetic analysis supports the following reaction scheme: a phosphate moiety on S431/T432 is transferred to KaiC-bound ADP to form ATP as an intermediate with a rate constant of 0.8 h 1 at 30  C, which can be regarded as a reversal of the autophosphorylation reaction, followed by the ATP hydrolysis with 1.2 h 1 generating inorganic phosphate, the final product (Nishiwaki and Kondo 2012). When KaiC is incubated in the presence of [14C]ADP, [14C] ATP is formed, showing that KaiC can produce ATP from ADP (Egli et al. 2012). These findings demonstrate that autophosphorylation and autodephosphorylation of KaiC are mediated by a reversible phosphotransfer reaction between KaiC-bound nucleotides and the hydroxyl group of S431/T432 catalyzed by the active site of ATP hydrolysis located in the subunit boundaries of CII (Nishiwaki and Kondo 2012; Egli et al. 2012). It can be predicted that the equilibrium of this inter-subunit phosphotransfer reaction is shifted periodically when phosphorylation levels of KaiC are oscillating, and KaiA and KaiB are supposed to be involved in this equilibrium shift. In general, the concentration of a reactant influences the direction of a reversible reaction. CII-bound ATP can be regarded as a phosphate donor of the forward reaction of autophosphorylation, and ADP a phosphate acceptor of the reverse reaction. To study whether KaiA and KaiB are involved in the regulation of KaiCbound nucleotide, the ratio of KaiC-bound ATP to the sum of bound ATP and bound ADP were addressed in the presence or absence of KaiA and KaiB. With solely KaiC, the ratio was about 0.3 at the steady state, whereas it was about 0.7 in the presence of KaiA (Nishiwaki-Ohkawa et al. 2014). Furthermore, with the use of α-[32P] ATP, KaiA facilitated the release of ADP, which was generated from KaiCbound ATP, and the incorporation of ATP from the outside of KaiC hexamer (Nishiwaki-Ohkawa et al. 2014). In the presence of both KaiA and KaiB, circadian rhythms in ATP incorporation, ratio of bound ATP to total bound nucleotides, and levels of KaiC phosphorylation were observed. The phase of the first one was advanced to the second one by several hours. The phase of KaiC phosphorylation is the latest among these (Nishiwaki-Ohkawa et al. 2014). These findings support the notion that KaiA and KaiB regulate the forward and the reverse reactions of KaiC autophosphorylation through the exchange of bound ADP to ATP, resulting in oscillation of KaiC phosphorylation levels (Nishiwaki-Ohkawa et al. 2014) (Fig. 2).

T

S

ATP

P

T P

S

ADP

ADP

Active KaiA

KaiA

CII

CI

KaiB

ATP

P

CII

CI

KaiABC complex

P

ATP

ATP hydrolysis

T

S

P

T

SP

TP

SP

Forward reaction of autophosphorylation

Inactive KaiA

KaiC

TP

S

P

P

ATP

Phosphate

P

ATP hydrolysis

ATP formation by reverse reaction of autophosphorylation

ADP

T

S P

ADP

ATP formation by reverse reaction of autophosphorylation

Fig. 2 Autophosphorylation and autodephosphorylation of KaiC mediated by the reversible phosphotransfer reaction between the hydroxyl group of S431/ T432 and KaiC-bound nucleotide. Autodephosphorylation of KaiC occurs through the reverse reaction of autophosphorylation, producing ATP as an intermediate, followed by its hydrolysis. ATP is the substrate of the forward reaction and ADP is that of the reverse reaction. This reversible phosphotransfer

ADP release

ATP P incorporation

T

S

Forward reaction of autophosphorylation

ADP release

ATP incorporation

Roles of Phosphorylation of KaiC in the Cyanobacterial Circadian Clock 249

250

T. Nishiwaki-Ohkawa

9 Structural Basis for Time-Specific Interactions Among Kai Proteins Structural analyses significantly contributed to reveal two ways of Kai protein interactions that drive the phosphotransfer reaction to either forward or reverse direction. KaiA is composed of an N-terminal pseudo receiver domain, which has a similar fold with that of the bacterial response regulators, and a C-terminal domain that contains the KaiC-binding site. These two domains are connected by a canonical linker, and two KaiA molecules form a domain-swapped dimer (Williams et al. 2002; Garces et al. 2004; Uzumaki et al. 2004; Ye et al. 2004; Vakonakis and LiWang 2004). Cryo-electron microscopy showed that KaiA interacts with C-terminal tails of KaiC protruding from CII ring when KaiA is promoting autophosphorylation (Pattanayek et al. 2006). NMR analysis was conducted on a complex between the C-terminal domain of KaiA and a KaiC peptide containing the C-terminal tail and A-loop, a segment located adjacent to the C-terminal tail. It is proposed that, in combination with X-ray crystallographic data of KaiC, the A-loop is in dynamic equilibrium in conformation between “exposed” and “buried.” When KaiA is binding to the C-terminal tail of KaiC, the exposed conformation is stabilized, resulting in the enhancement of the forward reaction of autophosphorylation. When the A-loop is in the buried conformation, the reverse reaction is favored (Kim et al. 2008). Atomic force microscopy visualized and quantified the reversible KaiA-KaiC interactions that stimulate the forward reaction of autophosphorylation. KaiA is transiently interacted with the C-terminal end of KaiC with a sub-second dwell time of bound state (Mori et al. 2018). The phosphorylation state of KaiC influences the rate constant of both association and dissociation of KaiA. As phosphorylation levels of KaiC increase, the affinity of KaiA for the C-terminus portion of KaiC decreases (Mori et al. 2018). KaiB adopts a thioredoxin-like βaβaββa fold in crystal, which is later called the ground state (gs), and forms either homo-dimer or homo-tetramer that consists of a dimer of dimer (Garces et al. 2004; Hitomi et al. 2005). NMR analysis revealed that the secondary structure of KaiB is reversibly switched in solution between the gs and βaβaββa fold called “fold-switched” (fs). The tetramer of KaiB in the gs state and the monomer of KaiB in the fs state are in equilibrium (Chang et al. 2015). It is suggested that phosphorylated S431 makes the structure of the CII ring of KaiC rigid and allosterically exposes KaiB binding sites on CI (Chang et al. 2012; Tseng et al. 2017). Six KaiB monomers in the fs state bind cooperatively to one KaiC

Fig. 2 (continued) reaction between the hydroxyl group of S431/T432 and KaiC-bound nucleotide is regulated by adenine nucleotide exchange activity of KaiA. When S431 is not phosphorylated, KaiA facilitates ADP release from the active site at the subunit boundary of CII and incorporation of ATP from the outside. Phosphorylation of S431 enhances formation of the KaiB-KaiC complex, which sequesters KaiA to inhibit its activity, retaining ADP at the active site

Roles of Phosphorylation of KaiC in the Cyanobacterial Circadian Clock

251

hexamer on the top surface of the CI ring and sequester six KaiA dimers (Chang et al. 2012; Chang et al. 2015; Tseng et al. 2017). This S431 phosphorylationdependent cooperativity was also proposed using mixed-hexamers composed of wild-type and mutant KaiC monomers containing phospho- and dephospho-mimetic substitutions. The phosphorylation state of a small fraction of subunits can modify the property of a hexamer as a whole, regulating Kai protein complex formation (Kitayama et al. 2013; Lin et al. 2014). Time-dependent assembly and disassembly of Kai protein complexes and their stoichiometry were tracked using native mass spectrometry. When phosphorylation levels of KaiC are low, most KaiC exists as a free hexamer, and a small fraction of KaiC interacts with KaiA to form the KaiC6A2 complex, promoting the forward reaction of autophosphorylation (Snijder et al. 2017). Six KaiB monomers are cooperatively bound to one KaiC hexamer containing phosphorylated S431, resulting in the KaiC6B6 complex (Snijder et al. 2017). Six KaiA dimers are then bound to KaiC6B6, resulting in the KaiC6B6A12 complex. Upon disassembly of KaiC6B6A12, KaiA2B1 is detected (Snijder et al. 2017). These structural studies give direct evidences for the conceptual model of S431 phosphorylation-dependent assembly and disassembly of Kai protein complexes that drives either the forward or reverse reaction of autophosphorylation (Fig. 2). To totally validate this model, the distribution of ATP and ADP over 12 ATP binding sites should be clarified in all possible complexes.

10

A Link Between KaiC Phosphorylation and Circadian Gene Expression

KaiC represses the activity of not only the kaiBC promoter but also promoters of various rhythmically transcribed genes (Nakahira et al. 2004; Xu et al. 2003). However, Kai proteins have no DNA binding motifs, suggesting that factors connecting Kai proteins and transcription should exist. The two-component system is an environmental signal transduction system in bacteria and plants composed of a sensory histidine kinase and its cognate response regulator. When signals are transduced into a histidine kinase, it autophosphorylates on its conserved histidine residue. The phosphate group on the histidine residue is then transferred to an aspartate residue on a response regulator, resulting in the transcriptional activation of target genes (Buschiazzo and Trajtenberg 2019). A sensory histidine kinase SasA was identified as a KaiC-interacting protein, which is composed of a KaiC-binding domain, homologous to KaiB, and a histidine kinase domain (Iwasaki et al. 2000). RpaA was identified as a cognate response regulator of SasA (Takai et al. 2006). Disruption of sasA and rpaA dramatically decreases the amplitude of transcriptional rhythms of clock genes and clock-controlled genes, suggesting that the SasA-RpaA system functions as a major output pathway of genome-wide circadian transcription (Iwasaki et al. 2000; Takai et al. 2006).

252

T. Nishiwaki-Ohkawa

Phospho-relay from SasA to RpaA is enhanced by KaiC in vitro. When SasA and RpaA are incubated in the presence of the reconstituted oscillator, SasA autophosphorylation and phosphotransfer from SasA to RpaA is detected, which becomes maximum when KaiC phosphorylation is increasing (Takai et al. 2006). The similar phosphorylation profile of RpaA is also observed in vivo; RpaA phosphorylation peaks at a few hours prior to the peak of KaiC phosphorylation level (Markson et al. 2013). These findings suggest that the phosphorylation of KaiC functions as an output signal from the oscillator composed of Kai proteins, and that the SasA-RpaA phospho-relay mediates the time signal to genome-wide transcription, although it is still controversial whether global transcriptional regulation is mediated by the direct binding of RpaA to the regulatory regions of genes or not (Markson et al. 2013; Hanaoka et al. 2012). In addition to these proteins, CikA and RpaB, a paralog of RpaA, are involved in the regulation of global transcription (Taniguchi et al. 2010; Hanaoka et al. 2012). CikA is a bacterial phytochrome originally identified as a circadian photoreceptor in cyanobacteria, which contains both histidine kinase and response regulator motifs (Schmitz et al. 2000). RpaB binds directly to transcriptional regulatory regions on the cyanobacterial genome (Hanaoka et al. 2012). Global transcriptional regulation by SasA is supported by structural analyses. SasA and KaiB share the binding sites on CI domain of KaiC, and their binding is regulated by the phosphorylation state of KaiC. When S431 is phosphorylated, KaiC-bound SasA is replaced by KaiB (Tseng et al. 2017), resulting in time-specific attenuation of SasA-RpaA phosphor-relay.

11

Multiple Output Systems of the Protein-Based Oscillator

In the cyanobacterial circadian system, KaiC phosphorylation had been considered to be a sole output from the protein-based oscillator composed of three Kai proteins; however, surprisingly, circadian oscillations of gene expression are also observed when the phosphorylation state of KaiC is arrested to the stable levels (Kitayama et al. 2008). In a KaiA-overexpressing strain, in which KaiC is constitutively hyperphosphorylated, the promoter activities of kaiBC and several clock-controlled genes exhibit robust circadian rhythm with a similar period and amplitude to those in the wild type, with the basal level of kaiBC expression being elevated (Kitayama et al. 2008). A mutant strain containing T432E and S432E double mutations, mimicking constitutive phosphorylation, or a T293H mutation in CII P-loop resulting in constitutively low levels in KaiC phosphorylation, also exhibits selfsustained oscillation of gene expression with a period longer than 24 h (Kitayama et al. 2008). Oscillations in these strains are more sensitive to low temperature than in the wild type, abolished at a temperature lower than 20  C. Furthermore, unstable damping oscillation of kaiBC promoter activity is observed in a kaiA-inactivated strain, in which the phosphorylation of KaiC is completely abolished (Kawamoto et al. 2020). This damping oscillation shows a period of about 24 h and can be

Roles of Phosphorylation of KaiC in the Cyanobacterial Circadian Clock

253

entrained to external light-dark cycles (Kawamoto et al. 2020). These findings propose that multiple output systems exist in the cyanobacterial circadian clock and that phosphorylation of KaiC is important for maintaining robust and precise circadian rhythmicity in global gene expression in cyanobacteria.

12

Perspective

In vitro reconstituted oscillator can be considered as a kind of a chemical oscillator in which the concentration of components changes periodically. Since the 1960s, mechanisms for chemical oscillations have been investigated in the field of non-linear physics (Tyson et al. 2008). To sustain chemical oscillation, it is necessary to supply energy to the system. The discovery of ATPase activity of KaiC provided a thermodynamical point of view with studies on the circadian clock mechanism. Based on data from structural and kinetic studies, a mathematical model was developed postulating that ATP hydrolysis provides energy for structural changes in KaiC to sustain oscillation and that the phosphorylation state of KaiC, regulated by ADP release from KaiC and ATP incorporation from the outside, determines the timing of the conformational changes (Paijmans et al. 2017). Most of the chemical oscillations studied so far show a period of seconds to several hours at the longest. The origin of the extremely long period of the KaiC phosphorylation rhythm, approximately one day, has not yet been clarified but has been ascribed to slow but stable ATPase hydrolysis of KaiC. To elucidate further the coupling between ATPase activity as the pacemaker and the phosphorylation cycle, it should be helpful to solve the structures of KaiC with different nucleotide and phosphorylation states and to examine the transition from one state to another. Recent advances in computational chemistry have allowed to simulate nucleotide release and its coupling to structural change in KaiC, which are too large in system size and too slow in timescale for conventional molecular dynamic simulation (Hong et al. 2018). It is expected that the mechanism of the cyanobacterial circadian system will be revealed along with advances in the research methodologies of the related fields. Acknowledgments I would like to thank Takao Kondo, Kazuki Terauchi, and Kumiko Miwa for fruitful discussions.

References Abe J, Hiyama TB, Mukaiyama A et al (2015) Circadian rhythms. Atomic-scale origins of slowness in the cyanobacterial circadian clock. Science 349:312–316. https://doi.org/10.1126/science. 1261040

254

T. Nishiwaki-Ohkawa

Buschiazzo A, Trajtenberg F (2019) Two-component sensing and regulation: how do histidine kinases talk with response regulators at the molecular level? Annu Rev Microbiol 73:507–528. https://doi.org/10.1146/annurev-micro-091018-054627 Camacho F, Cilio M, Guo Y et al (2001) Human casein kinase id phosphorylation of human circadian clock proteins period 1 and 2. FEBS Lett 489:159–165. https://doi.org/10.1016/s00145793(00)02434-0 Chang YG, Tseng R, Kuo NW et al (2012) Rhythmic ring-ring stacking drives the circadian oscillator clockwise. Proc Natl Acad Sci U S A 109:16847–16851. https://doi.org/10.1073/ pnas.1211508109 Chang YG, Cohen SE, Phong C et al (2015) Circadian rhythms. A protein fold switch joins the circadian oscillator to clock output in cyanobacteria. Science 349:324–328. https://doi.org/10. 1126/science.1260031 Crane BR, Young MW (2014) Interactive features of proteins composing eukaryotic circadian clocks. Annu Rev Biochem 83:191–219. https://doi.org/10.1146/annurev-biochem-060713035644 Diernfellner ACR, Brunner M (2020) Phosphorylation timers in the Neurospora crassa circadian clock. J Mol Biol 432:3449–3465. https://doi.org/10.1016/j.jmb.2020.04.004 Doolittle WF (1979) The cyanobacterial genome, its expression, and the control of that expression. Adv Microb Physiol 20:1–102. https://doi.org/10.1016/s0065-2911(08)60206-4 Dubowy C, Sehgal A (2017) Circadian rhythms and sleep in Drosophila melanogaster. Genetics 205:1373–1397. https://doi.org/10.1534/genetics.115.185157 Edery I, Zwiebel LJ, Dembinska ME et al (1994) Temporal phosphorylation of the Drosophila period protein. Proc Natl Acad Sci U S A 91:2260–2264. https://doi.org/10.1073/pnas.91.6. 2260 Egli M, Mori T, Pattanayek R et al (2012) Dephosphorylation of the core clock protein KaiC in the cyanobacterial KaiABC circadian oscillator proceeds via an ATP synthase mechanism. Biochemistry 51:1547–1558. https://doi.org/10.1021/bi201525n Garceau NY, Liu Y, Loros JJ et al (1997) Alternative initiation of translation and time-specific phosphorylation yield multiple forms of the essential clock protein FREQUENCY. Cell 89:469–476. https://doi.org/10.1016/s0092-8674(00)80227-5 Garces RG, Wu N, Gillon W et al (2004) Anabaena circadian clock proteins KaiA and KaiB reveal a potential common binding site to their partner KaiC. EMBO J 23:1688–1698. https://doi.org/10. 1038/sj.emboj.7600190 Hanaoka M, Takai N, Hosokawa N et al (2012) RpaB, another response regulator operating circadian clock-dependent transcriptional regulation in Synechococcus elongatus PCC 7942. J Biol Chem 287:26321–26327. https://doi.org/10.1074/jbc.M111.338251 Harada Y, Sakai M, Kurabayashi N et al (2005) Ser-557-phosphorylated mCRY2 is degraded upon synergistic phosphorylation by glycogen synthase kinase-3b. J Biol Chem 280:31714–31721. https://doi.org/10.1074/jbc.M506225200 Hayashi F, Suzuki H, Iwase R et al (2003) ATP-induced hexameric ring structure of the cyanobacterial circadian clock protein KaiC. Genes Cells 8:287–296. https://doi.org/10.1046/ j.1365-2443.2003.00633.x Hayashi F, Itoh N, Uzumaki T et al (2004) Roles of two ATPase-motif-containing domains in cyanobacterial circadian clock protein KaiC. J Biol Chem 279:52331–52337. https://doi.org/10. 1074/jbc.M406604200 Hirano A, Fu YH, Ptáček LJ (2016) The intricate dance of post-translational modifications in the rhythm of life. Nat Struct Mol Biol 23:1053–1060. https://doi.org/10.1038/nsmb.3326 Hitomi K, Oyama T, Han S et al (2005) Tetrameric architecture of the circadian clock protein KaiB. A novel interface for intermolecular interactions and its impact on the circadian rhythm. J Biol Chem 280:19127–19135. https://doi.org/10.1074/jbc.M411284200 Hong L, Vani BP, Thiede EH et al (2018) Molecular dynamics simulations of nucleotide release from the circadian clock protein KaiC reveal atomic-resolution functional insights. Proc Natl Acad Sci U S A 115:E11475–E11484. https://doi.org/10.1073/pnas.1812555115

Roles of Phosphorylation of KaiC in the Cyanobacterial Circadian Clock

255

Ishiura M, Kutsuna S, Aoki S et al (1998) Expression of a gene cluster kaiABC as a circadian feedback process in cyanobacteria. Science 281:1519–1523. https://doi.org/10.1126/science. 281.5382.1519 Ito H, Kageyama H, Mutsuda M et al (2007) Autonomous synchronization of the circadian KaiC phosphorylation rhythm. Nat Struct Mol Biol 14:1084–1088. https://doi.org/10.1038/nsmb1312 Iwasaki H, Taniguchi Y, Ishiura M et al (1999) Physical interactions among circadian clock proteins KaiA, KaiB and KaiC in cyanobacteria. EMBO J 18:1137–1145. https://doi.org/10.1093/emboj/ 18.5.1137 Iwasaki H, Williams SB, Kitayama Y et al (2000) A KaiC-interacting sensory histidine kinase, SasA, necessary to sustain robust circadian oscillation in cyanobacteria. Cell 101:223–233. https://doi.org/10.1016/S0092-8674(00)80832-6 Iwasaki H, Nishiwaki T, Kitayama Y et al (2002) KaiA-stimulated KaiC phosphorylation in circadian timing loops in cyanobacteria. Proc Natl Acad Sci U S A 99:15788–15793. https:// doi.org/10.1073/pnas.222467299 Kageyama H, Kondo T, Iwasaki H (2003) Circadian formation of clock protein complexes by KaiA, KaiB, KaiC, and SasA in cyanobacteria. J Biol Chem 278:2388–2395. https://doi.org/10. 1074/jbc.M208899200 Kageyama H, Nishiwaki T, Nakajima M et al (2006) Cyanobacterial circadian pacemaker: Kai protein complex dynamics in the KaiC phosphorylation cycle in vitro. Mol Cell 23:161–171. https://doi.org/10.1016/j.molcel.2006.05.039 Kawamoto N, Ito H, Tokuda IT et al (2020) Damped circadian oscillation in the absence of KaiA in Synechococcus. Nat Commun 11(1):2242. Published 2020 May 7. https://doi.org/10.1038/ s41467-020-16087-x Kim YI, Dong G, Carruthers CW Jr et al (2008) The day/night switch in KaiC, a central oscillator component of the circadian clock of cyanobacteria. Proc Natl Acad Sci U S A 105:12825–12830. https://doi.org/10.1073/pnas.0800526105 Kitahara R, Oyama K, Kawamura T et al (2019) Pressure accelerates the circadian clock of cyanobacteria. Sci Rep 9:12395. https://doi.org/10.1038/s41598-019-48693-1 Kitayama Y, Iwasaki H, Nishiwaki T et al (2003) KaiB functions as an attenuator of KaiC phosphorylation in the cyanobacterial circadian clock system. EMBO J 22:2127–2134. https://doi.org/10.1093/emboj/cdg212 Kitayama Y, Nishiwaki T, Terauchi K et al (2008) Dual KaiC-based oscillations constitute the circadian system of cyanobacteria. Genes Dev 22(11):1513–1521. https://doi.org/10.1101/gad. 1661808 Kitayama Y, Nishiwaki-Ohkawa T, Sugisawa Y et al (2013) KaiC intersubunit communication facilitates robustness of circadian rhythms in cyanobacteria. Nat Commun 4:2897. https://doi. org/10.1038/ncomms3897 Lee C, Bae K, Edery I (1998) The Drosophila CLOCK protein undergoes daily rhythms in abundance, phosphorylation, and interactions with the PER-TIM complex. Neuron 21:857–867. https://doi.org/10.1016/s0896-6273(00)80601-7 Lee C, Etchegaray JP, Cagampang FR et al (2001) Posttranslational mechanisms regulate the mammalian circadian clock. Cell 107:855–867. https://doi.org/10.1016/s0092-8674(01) 00610-9 Leipe DD, Koonin EV, Aravind L (2003) Evolution and classification of P-loop kinases and related proteins. J Mol Biol 333:781–815. https://doi.org/10.1016/j.jmb.2003.08.040 Lin J, Chew J, Chockanathan U et al (2014) Mixtures of opposing phosphorylations within hexamers precisely time feedback in the cyanobacterial circadian clock. Proc Natl Acad Sci U S A 111:E3937–E3945. https://doi.org/10.1073/pnas.1408692111 Loros JJ (2020) Principles of the animal molecular clock learned from Neurospora. Eur J Neurosci 51:19–33. https://doi.org/10.1111/ejn.14354 Markson JS, Piechura JR, Puszynska AM et al (2013) Circadian control of global gene expression by the cyanobacterial master regulator RpaA. Cell 155:1396–1408. https://doi.org/10.1016/j. cell.2013.11.005

256

T. Nishiwaki-Ohkawa

McClung CR (2019) The plant circadian oscillator. Biology (Basel) 8:14. https://doi.org/10.3390/ biology8010014 Mori T, Saveliev SV, Xu Y et al (2002) Circadian clock protein KaiC forms ATP-dependent hexameric rings and binds DNA. Proc Natl Acad Sci U S A 99:17203–17208. https://doi.org/10. 1073/pnas.262578499 Mori T, Sugiyama S, Byrne M et al (2018) Revealing circadian mechanisms of integration and resilience by visualizing clock proteins working in real time. Nat Commun 9:3245. https://doi. org/10.1038/s41467-018-05438-4 Nakahira Y, Katayama M, Miyashita H et al (2004) Global gene repression by KaiC as a master process of prokaryotic circadian system. Proc Natl Acad Sci U S A 101:881–885. https://doi. org/10.1073/pnas.0307411100 Nakajima M, Imai K, Ito H et al (2005) Reconstitution of circadian oscillation of cyanobacterial KaiC phosphorylation in vitro. Science 308:414–415. https://doi.org/10.1126/science.1108451 Nakajima M, Ito H, Kondo T (2010) In vitro regulation of circadian phosphorylation rhythm of cyanobacterial clock protein KaiC by KaiA and KaiB. FEBS Lett 584:898–902. https://doi.org/ 10.1016/j.febslet.2010.01.016 Narasimamurthy R, Hunt SR, Lu Y et al (2018) CK1δ/ε protein kinase primes the PER2 circadian phosphoswitch. Proc Natl Acad Sci U S A 115:5986–5991. https://doi.org/10.1073/pnas. 1721076115 Nishiwaki T, Kondo T (2012) Circadian autodephosphorylation of cyanobacterial clock protein KaiC occurs via formation of ATP as intermediate. J Biol Chem 287:18030–18035. https://doi. org/10.1074/jbc.M112.350660 Nishiwaki T, Iwasaki H, Ishiura M et al (2000) Nucleotide binding and autophosphorylation of the clock protein KaiC as a circadian timing process of cyanobacteria. Proc Natl Acad Sci U S A 97:495–499. https://doi.org/10.1073/pnas.97.1.495 Nishiwaki T, Satomi Y, Nakajima M et al (2004) Role of KaiC phosphorylation in the circadian clock system of Synechococcus elongatus PCC 7942. Proc Natl Acad Sci U S A 101:13927–13932. https://doi.org/10.1073/pnas.0403906101 Nishiwaki T, Satomi Y, Kitayama Y et al (2007) A sequential program of dual phosphorylation of KaiC as a basis for circadian rhythm in cyanobacteria. EMBO J 26:4029–4037. https://doi.org/ 10.1038/sj.emboj.7601832 Nishiwaki-Ohkawa T, Kitayama Y, Ochiai E et al (2014) Exchange of ADP with ATP in the CII ATPase domain promotes autophosphorylation of cyanobacterial clock protein KaiC. Proc Natl Acad Sci U S A 111:4455–4460. https://doi.org/10.1073/pnas.1319353111 Paijmans J, Lubensky DK, Ten Wolde PR (2017) A thermodynamically consistent model of the post-translational Kai circadian clock. PLoS Comput Biol 13:e1005415. https://doi.org/10. 1371/journal.pcbi.1005415 Pattanayek R, Wang J, Mori T et al (2004) Visualizing a circadian clock protein: crystal structure of KaiC and functional insights. Mol Cell 15:375–388. https://doi.org/10.1016/j.molcel.2004.07. 013 Pattanayek R, Williams DR, Pattanayek S et al (2006) Analysis of KaiA-KaiC protein interactions in the cyano-bacterial circadian clock using hybrid structural methods. EMBO J 25:2017–2028. https://doi.org/10.1038/sj.emboj.7601086 Pattanayek R, Xu Y, Lamichhane A et al (2014) An arginine tetrad as mediator of input-dependent and input-independent ATPases in the clock protein KaiC. Acta Crystallogr D Biol Crystallogr 70:1375–1390. https://doi.org/10.1107/S1399004714003228 Philpott JM, Narasimamurthy R, Ricci CG et al (2020) Casein kinase 1 dynamics underlie substrate selectivity and the PER2 circadian phosphoswitch. elife 9:e52343. https://doi.org/10.7554/ eLife.52343 Rust MJ, Markson JS, Lane WS et al (2007) Ordered phosphorylation governs oscillation of a threeprotein circadian clock. Science 318:809–812. https://doi.org/10.1126/science.1148596

Roles of Phosphorylation of KaiC in the Cyanobacterial Circadian Clock

257

Rutila JE, Zeng H, Le M et al (1996) The timSL mutant of the Drosophila rhythm gene timeless manifests allele-specific interactions with period gene mutants. Neuron 17:921–929. https://doi. org/10.1016/s0896-6273(00)80223-8 Salazar C, Höfer T (2009) Multisite protein phosphorylation-from molecular mechanisms to kinetic models. FEBS J 276:3177–3198. https://doi.org/10.1111/j.1742-4658.2009.07027.x Sanada K, Harada Y, Sakai M et al (2004) Serine phosphorylation of mCRY1 and mCRY2 by mitogen-activated protein kinase. Genes Cells 9:697–708. https://doi.org/10.1111/j.1356-9597. 2004.00758.x Schafmeier T, Káldi K, Diernfellner A et al (2006) Phosphorylation-dependent maturation of Neurospora circadian clock protein from a nuclear repressor toward a cytoplasmic activator. Genes Dev 20:297–306. https://doi.org/10.1101/gad.360906 Schmitz O, Katayama M, Williams SB et al (2000) CikA, a bacteriophytochrome that resets the cyanobacterial circadian clock. Science 289:765–768. https://doi.org/10.1126/science.289. 5480.765 Shi Y (2009) Serine/threonine phosphatases: mechanism through structure. Cell 139:468–484. https://doi.org/10.1016/j.cell.2009.10.006 Snijder J, Schuller JM, Wiegard A et al (2017) Structures of the cyanobacterial circadian oscillator frozen in a fully assembled state. Science 355:1181–1184. https://doi.org/10.1126/science. aag3218 Takahashi JS (2017) Transcriptional architecture of the mammalian circadian clock. Nat Rev Genet 18:164–179. https://doi.org/10.1038/nrg.2016.150 Takai N, Nakajima M, Oyama T et al (2006) A KaiC-associating SasA-RpaA two-component regulatory system as a major circadian timing mediator in cyanobacteria. Proc Natl Acad Sci U S A 103:12109–12114. https://doi.org/10.1073/pnas.0602955103 Takano A, Shimizu K, Kani S et al (2000) Cloning and characterization of rat casein kinase 1e. FEBS Lett 477:106–112. https://doi.org/10.1016/s0014-5793(00)01755-5 Taniguchi Y, Takai N, Katayama M et al (2010) Three major output pathways from the KaiABCbased oscillator cooperate to generate robust circadian kaiBC expression in cyanobacteria. Proc Natl Acad Sci U S A 107:3263–3268. https://doi.org/10.1073/pnas.0909924107 Terauchi K, Kitayama Y, Nishiwaki T et al (2007) ATPase activity of KaiC determines the basic timing for circadian clock of cyanobacteria. Proc Natl Acad Sci U S A 104:16377–16381. https://doi.org/10.1073/pnas.0706292104 Tomita J, Nakajima M, Kondo T et al (2005) No transcription-translation feedback in circadian rhythm of KaiC phosphorylation. Science 307:251–254. https://doi.org/10.1126/science. 1102540 Tseng R, Goularte NF, Chavan A et al (2017) Structural basis of the day-night transition in a bacterial circadian clock. Science 355:1174–1180. https://doi.org/10.1126/science.aag2516 Tyson JJ, Albert R, Goldbeter A et al (2008) Biological switches and clocks. J R Soc Interface 5 (Suppl 1):S1–S8. https://doi.org/10.1098/rsif.2008.0179.focus Uzumaki T, Fujita M, Nakatsu T et al (2004) Crystal structure of the C-terminal clock-oscillator domain of the cyanobacterial KaiA protein. Nat Struct Mol Biol 11:623–631. https://doi.org/10. 1038/nsmb781 Vakonakis I, LiWang AC (2004) Structure of the C-terminal domain of the clock protein KaiA in complex with a KaiC-derived peptide: implications for KaiC regulation. Proc Natl Acad Sci U S A 101:10925–10930. https://doi.org/10.1073/pnas.0403037101 Walker JE, Saraste M, Runswick MJ et al (1982) Distantly related sequences in the alpha- and betasubunits of ATP synthase, myosin, kinases and other ATP-requiring enzymes and a common nucleotide binding fold. EMBO J 1:945–951 Williams SB, Vakonakis I, Golden SS et al (2002) Structure and function from the circadian clock protein KaiA of Synechococcus elongatus: a potential clock input mechanism. Proc Natl Acad Sci U S A 99:15357–15362. https://doi.org/10.1073/pnas.232517099

258

T. Nishiwaki-Ohkawa

Xu Y, Mori T, Johnson CH (2003) Cyanobacterial circadian clockwork: roles of KaiA, KaiB and the kaiBC promoter in regulating KaiC. EMBO J 22:2117–2126. https://doi.org/10.1093/emboj/ cdg168 Xu Y, Mori T, Pattanayek R et al (2004) Identification of key phosphorylation sites in the circadian clock protein KaiC by crystallographic and mutagenetic analyses. Proc Natl Acad Sci U S A 101:13933–13938. https://doi.org/10.1073/pnas.0404768101 Yang Y, Xu T, Zhang Y, Qin X (2017) Molecular basis for the regulation of the circadian clock kinases CK1δ and CK1ε. Cell Signal 31:58–65. https://doi.org/10.1016/j.cellsig.2016.12.010 Ye S, Vakonakis I, Ioerger TR et al (2004) Crystal structure of circadian clock protein KaiA from Synechococcus elongatus. J Biol Chem 279:20511–20518. https://doi.org/10.1074/jbc. M400077200 Yoshida M, Amano T (1995) A common topology of proteins catalyzing ATP-triggered reactions. FEBS Lett 359:1–5. https://doi.org/10.1016/0014-5793(94)01438-7 Yoshida T, Murayama Y, Ito H et al (2009) Nonparametric entrainment of the in vitro circadian phosphorylation rhythm of cyanobacterial KaiC by temperature cycle. Proc Natl Acad Sci U S A 106:1648–1653. https://doi.org/10.1073/pnas.0806741106 Zhou M, Kim JK, Eng GW et al (2015) A Period2 Phosphoswitch regulates and temperature compensates circadian period. Mol Cell 60:77–88. https://doi.org/10.1016/j.molcel.2015.08. 022

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for Biotechnological Applications Bo Wang, Jamey D. Young, and Yao Xu

Abstract Cyanobacteria have two unique characteristics that distinguish them from all other prokaryotes – being able to perform oxygenic photosynthesis and having a circadian clock. Because cyanobacteria have a relatively simple genetic background and usually grow much faster than higher plants, they have become promising microbial hosts to produce biofuels and chemicals directly from sunlight and CO2. Recent advances in synthetic biology and systems biology not only have provided impetus for reprogramming metabolic networks of cyanobacterial host cells, but also have greatly expanded our knowledge of the genetics, physiology, and regulatory mechanisms of cyanobacteria. The plasticity of cyanobacterial central metabolism allows researchers to reprogram the metabolic networks to enhance the production of chemicals, but the metabolic regulation that prevents maximizing chemical production is still poorly understood. The cyanobacterial circadian clock exerts global gene regulations and oscillates intracellular metabolism. Manipulation of the circadian clock genes has proven to be an effective strategy for modulating the expression patterns of both endogenous and exogenous genes, driving either constant up- or down-regulation, which might be leveraged to enhance the production of biofuels and chemicals in cyanobacteria.

1 Introduction Cyanobacteria are the only prokaryotes capable of performing oxygenic photosynthesis, a biological process also occurring in plants that fixes carbon dioxide and solar energy while releasing oxygen (Hamilton et al. 2016). Cyanobacteria typically

B. Wang (*) · J. D. Young Department of Chemical and Biomolecular Engineering, Vanderbilt University, Nashville, TN, USA e-mail: [email protected]; [email protected] Y. Xu (*) Department of Biological Sciences, Vanderbilt University, Nashville, TN, USA e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_14

259

260

B. Wang et al.

Table 1 Most commonly used model cyanobacterial strains

Strain Synechococcus PCC 7942 Synechococcus UTEX 2973 Synechococcus PCC 7002 Synechocystis PCC 6803

Chromosome 2.70 Mb

Nostoc PCC 7120

6.41 Mb

2.69 Mb 3.01 Mb 3.57 Mb

Endogenous plasmids 2 plasmids (7.8 & 46 kb) 2 plasmids (7.8 & 46 kb) 6 plasmids (4.8–186 kb) 7 plasmids (2.3–120 kb)

Physiological features Freshwater; unicellular Freshwater; unicellular Euryhaline; unicellular Freshwater; unicellular

6 plasmids (5.6–408 kb)

Freshwater; filamentous; diazotrophic

Optimum growth temperature 30–38
C

Core circadian clock genes kaiABC

38–42
C

kaiABC

30–38
C

kaiABC, kaiB2 kaiAB1C1, kaiC2B2, kaiB3, kaiC3 kaiABC, kaiB1

30–38
C 23–30
C

grow much faster than higher plants, have simpler genetic systems that are relatively easy to manipulate, and their cultivation does not require arable land. Many cyanobacterial species are also able to fix nitrogen from the atmosphere (Kasting and Siefert 2002). For decades, cyanobacteria have been widely used as genetic models for studying photosynthesis and nitrogen fixation (Zarzycki et al. 2012; Zehr 2011). More recently, due to concerns about increased global carbon dioxide emissions and the pressing need to develop renewable energy sources, researchers have developed cyanobacteria as chassis for building microbial cell factories that produce a variety of biofuels and commodity compounds, such as carbohydrates, lipids, proteins, alcohols, pigments, and precursors to pharmaceuticals (Wang et al. 2012; Lai and Lan 2015; Gao et al. 2016b; Zhou et al. 2016b; Knoot et al. 2018; Luan and Lu 2018). However, the best product titers achieved to date still lag far behind those of highly optimized yeast or E. coli strains. In comparison to these heterotrophs, slow carbon assimilation rate and poorly understood metabolic regulation hinder the rational improvement of chemical production in cyanobacteria. To date, hundreds of cyanobacterial genomes have been sequenced (http://genome.jgi. doe.gov/), several model cyanobacterial strains have been extensively investigated for their genetics and physiology, and an increasing number of strains have been genetically engineered for potential biotechnological applications. We depict some basic features of the commonly used model cyanobacterial strains in Table 1. Cyanobacteria are the only prokaryotes that have currently been shown to have a robust circadian clock, which makes them ideal models for studying the mechanism of circadian rhythms (Cohen and Golden 2015). A bonafide circadian rhythm has to satisfy three criteria: persistence under constant conditions with a period close to 24 h, entrainment, and temperature compensation. Since the first discovery of the circadian clock in the diazotrophic cyanobacterium Synechococcus sp. RF-1 in 1986 (Grobbelaar et al. 1986), tremendous efforts have been devoted to studying the

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

261

circadian clocks of cyanobacteria, and Synechococcus elongatus PCC 7942 has become the model cyanobacterial strain for which the mechanism of circadian oscillations is most vigorously studied (Cohen and Golden 2015; Johnson et al. 2017). It has been unveiled that the kaiA, kaiB, and kaiC genes encode the core components of the circadian clock, which underlies daily rhythms of major cellular events, such as the superhelical status of DNA, compaction of the chromosome, rhythmicity of gene expression profiles, patterns of global metabolism, natural competence of the cell, and timing of cell division (Cohen and Golden 2015; Johnson et al. 2017; Taton et al. 2020). While the circadian rhythm provides a fitness advantage for cyanobacteria to adapt to the natural light-dark diel habitats (Ouyang et al. 1998; Woelfle et al. 2004; Hellweger et al. 2020), manipulating the circadian clocks might also provide an intriguing strategy to optimize cyanobacteria as cell factories for producing biofuels and other renewable chemicals (Xu et al. 2013).

2 Model Cyanobacterial Strains Although there are a vast number of cyanobacteria on Earth, only a few model strains are well studied and each has its own characteristics (Table 1). Synechococcus elongatus PCC 7942 (hereafter Synechococcus PCC 7942) is a freshwater cyanobacterial strain that has been used extensively as a model to study circadian rhythms (Kondo et al. 1993; Cohen and Golden 2015). Synechocystis sp. PCC 6803 (hereafter Synechocystis PCC 6803) is a freshwater strain that has been commonly used as a genetic model to study photosynthesis, and its genome was the first to be sequenced among all cyanobacterial species (Kaneko et al. 1996). Synechococcus sp. PCC 7002 (hereafter Synechococcus PCC 7002) is a marine cyanobacterial strain that exhibits a fast doubling rate (2.6–4 h per generation) (Ludwig and Bryant 2012). Nostoc sp. PCC 7120 (hereafter Nostoc PCC 7120; formerly called Anabaena sp. PCC 7120) serves as a model filamentous cyanobacterial strain for studying cell differentiation and nitrogen fixation (Kaneko et al. 2001). Most recently, Synechococcus elongatus UTEX 2973 (hereafter Synechococcus UTEX 2973) has drawn considerable attention due to its unprecedented growth rate (as fast as 1.5 h per generation), which is the highest among all known cyanobacteria (Yu et al. 2015; Ungerer et al. 2018a). Notably, Synechococcus UTEX 2973 has a genome that is 99.8% identical to that of the widely studied Synechococcus PCC 7942 strain, with differences occurring at merely 55 genomic loci. Comparative genomics has revealed genetic modifications in three major genes that contribute to its fast growth phenotype. The first is a C252Y replacement in the alpha subunit of the ATP synthase, AtpA. The second is an E260D substitution in the NAD+ kinase PpnK. The third modification involves a 7-bp deletion at a genome region 102-bp upstream of the translational start codon of the rpaA gene, and two amino acid replacements, i.e., Q121R and E134K, in RpaA, the circadian clock master output regulator (Ungerer et al. 2018b).

262

B. Wang et al.

The genome sizes of these five model cyanobacterial strains range from 2.69 to 6.41 million base pairs (Mb), which is comparable to the genome size of E. coli (4.64 Mb for E. coli str. K-12 substr. MG1655). Additionally, while Synechococcus PCC 7942, Synechocystis PCC 6803, and Synechococcus PCC 7002 exhibit optimal growth in a temperature range of 30–38
C (Ludwig and Bryant 2012; Kuan et al. 2015; Yu et al. 2015; Zavřel et al. 2015; van Alphen et al. 2018), Synechococcus UTEX 2973 shows optimal growth in the temperature range of 38–42
C (Yu et al. 2015; Ungerer et al. 2018a). The filamentous Nostoc PCC 7120 strain grows optimally at 23–30
C (Klodawska et al. 2019). All the five model cyanobacterial strains have a kaiABC cluster on their genome. While Synechococcus PCC 7002 and Nostoc PCC 7120 each has an extra copy of kaiB on their genome, Synechocystis PCC 6803 has two extra copies of kaiB and kaiC (Wiegard et al. 2013) (Table 1). Although not discussed here, a few other cyanobacterial strains, such as Cyanothece sp. ATCC 51142 (Toepel et al. 2008; Stockel et al. 2008; Welsh et al. 2008; Vu et al. 2012; Alagesan et al. 2013; Liberton et al. 2019), have also been used by multiple research groups as models for studying important aspects of cyanobacteria.

3 Synthetic and Systems Biology in Cyanobacteria Since the onset of the post-genome era, synthetic biology toolkits for cyanobacteria have expanded dramatically. A few promoter libraries have since been developed and characterized in order to modulate the transcription of target genes of interest. A wide range of gene expression levels can be achieved using the constitutive promoters, which can be used to optimize the expression of heterologous pathways (Wang et al. 2012; Albers et al. 2015; Markley et al. 2015; Wang et al. 2018a; Sengupta et al. 2019; Ruffing et al. 2016; Liu and Pakrasi 2018). Among all these previously studied promoters, the E. coli σ70 Ptrc promoter provides the strongest constitutive expression when there is no lacI present in the cell (Wang et al. 2018a). The PcpcB promoter has also been reported to be very strong (Zhou et al. 2014), but its strength apparently varies depending on the downstream gene sequence and the culture condition (Wang et al. 2018a). In addition, multiple inducible expression systems have been developed for cyanobacteria (Gordon and Pfleger 2018; Behle et al. 2020), such as the Ni2+-inducible PnrsB promoter (Englund et al. 2016; Liu and Curtiss 2009), the Zn2+-inducible expression platform (Pérez et al. 2017), the anhydrotetracycline-inducible system (Higo et al. 2016; Zess et al. 2016), the IPTG-inducible LacI/Ptrc system (Albers et al. 2015; Markley et al. 2015; Geerts et al. 1995), the L-arabinose-inducible ParaBAD promoter (Immethun et al. 2017; Cao et al. 2017), the L-rhamnose-inducible PrhaBAD promoter (Kelly et al. 2018), and the Ni2+-inducible T7 expression system (Jin et al. 2019). The ribosome-binding site (RBS) is another important genetic element that controls expression of the downstream gene by affecting the translation initiation rate (TIR). Although the TIR of many genes in heterotrophs can be relatively accurately estimated by state-of-the-art prediction software (Salis et al. 2009), it turns out that the TIRs of genes expressed in

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

263

cyanobacteria are much harder to predict (Sebesta and Peebles 2020; Thiel et al. 2018; Wang et al. 2018a). Combinatorial design and testing of RBS sequences for the genes of interest is often necessary to complement software predictions. The use of riboswitches is an alternate strategy to control the expression of genes of interest. Riboswitches regulate gene expression through forming secondary structures at the 5’ ends of mRNA. Adjusting the concentration of an effector molecule in the culture medium can be used to toggle gene expression ON or OFF (Nudler and Mironov 2004). Although several cyanobacterial riboswitches have been reported so far (Sun et al. 2018), the theophylline-dependent synthetic riboswitch is most commonly used in cyanobacteria (Ma et al. 2014; Nakahira et al. 2013). Furthermore, several expression vectors (Huang et al. 2010; Xu et al. 2011; Chen et al. 2016; Bishé et al. 2019; Jin et al. 2018) and CRISPR/Cas systems (Gordon et al. 2016; Ungerer and Pakrasi 2016; Wendt et al. 2016; Niu et al. 2019; Li et al. 2016; Behler et al. 2018) have been developed to accelerate the cyanobacterial strain engineering process. We refer the readers to several excellent recent reviews for details (Sun et al. 2018; Santos-Merino et al. 2019; Sengupta et al. 2018). With the development of high-throughput sequencing and mass spectrometry technologies, systems biology in cyanobacteria has achieved significant progress. Applications of transcriptomics (Mitschke et al. 2011), proteomics (Battchikova et al. 2018), metabolomics (Schwarz et al. 2013), 13C-metabolic flux analysis (MFA) (Young et al. 2011), and genome-scale modeling (Hendry et al. 2019) in cyanobacteria have profoundly changed the landscape of cyanobacterial research. These technological advancements have greatly expanded the understanding of the genetics and physiology of cyanobacteria. For instance, it has been found through transcriptomics that synthesis of non-protein-coding RNA (npcRNA) occurs at regions of over 60% of all transcriptional start sites in Synechocystis 6803, indicating a major regulatory mechanism through npcRNA in this species (Mitschke et al. 2011). Another example is that through 13C-MFA the intracellular metabolic fluxes of a number of cyanobacterial strains grown under various conditions have been investigated (Young et al. 2011; You et al. 2014; Xiong et al. 2015; Sake et al. 2019; Nakajima et al. 2017; Yu et al. 2015; Yu King Hing et al. 2019; Abernathy et al. 2017, 2019), which has shed light on metabolic regulations, indicated potential metabolic bottlenecks, and successfully guided strain development efforts (Jazmin et al. 2017; Durall et al. 2020; Cheah et al. 2020).

4 Cyanobacterial Biofuels and Chemicals Researchers’ continuous efforts to adapt systems/synthetic biology tools for cyanobacteria have positioned these phototrophic hosts at the front line to curb the global warming and the energy sustainability issues facing modern human society. Many researchers have used cyanobacteria as models to study how to improve the photosynthesis efficiency and, thus, the yield of crop plants (Lin et al. 2014; Zarzycki et al. 2012), while enormous efforts have been made to transform

264

B. Wang et al.

cyanobacteria into cell factories that produce biofuels and chemicals directly from CO2 and sunlight (Luan and Lu 2018; Luan et al. 2020; Angermayr et al. 2015; Zhou et al. 2016b; Knoot et al. 2018; Lai and Lan 2015). In addition, cyanobacteria have also been engineered to produce non-carbon-based forms of energy, such as dihydrogen (H2) (Appel et al. 2020), electricity (Tschörtner et al. 2019; Saper et al. 2018), and nitrogen-rich compounds (Wang et al. 2019). Most products of interest are synthesized through tapping metabolic fluxes from central carbon intermediates (i.e., “hub” metabolites) or reducing power (such as NADPH for the case of H2) in cyanobacteria (Table 2). Since these hub metabolites are often essential to the cell, a typical first assumption is that the central carbon metabolism will readily adjust to meet the flux demands imposed by product formation (Xiong et al. 2017). Therefore, most efforts have been focused on increasing the driving force between the hub metabolite and the end product, which usually involves the design of metabolic pathways, comparison of isoenzymes, enhancing the gene expression levels, blocking competing pathways, and installing transporters (Angermayr et al. 2015; Knoot et al. 2018; Luan and Lu 2018; Sengupta et al. 2018; Sun et al. 2018; Santos-Merino et al. 2019). A number of enzymes or pathways from heterotrophic microorganisms have been deposited in public databases such as BRENDA (Placzek et al. 2016), MetaCyc (Caspi et al. 2015), and KEGG (Kanehisa et al. 2019), and some of these metabolic pathways could be directly introduced into cyanobacteria to couple with their host metabolism. Nevertheless, there are cases where the outcome is not straightforward. For instance, many metabolic pathways originating from anaerobic organisms consist of enzymes that are oxygen-sensitive, such as [FeFe]-hydrogenases and the butyryl-CoA dehydrogenase (in the n-butanol biosynthetic pathway), which conflicts with oxygenic photosynthesis. Another issue to consider is that while the redox cofactor NADH is more abundant in heterotrophic microbes, NADPH is the direct product of photosynthesis and is more abundant in cyanobacteria. Therefore, in order to tailor the metabolic pathway to suit the cyanobacterial host, isozymes using NADPH rather than NADH are preferable (Park and Choi 2017). Additionally, unlike heterotrophic microorganisms, cyanobacteria often lack transporters to import or export various kinds of compounds, probably because they typically live in nutrient-poor environments where availability of organic carbon is limited and export of metabolites would be disadvantageous. Therefore, transporters often need to be installed to excrete the end products (Niederholtmeyer et al. 2010; Ducat et al. 2012; Korosh et al. 2017). Here, we focus on reviewing recent progresses toward photosynthetic production of biofuels and chemicals synthesized from a few central metabolites, including sugar phosphates, dihydroxyacetone phosphate (DHAP), pyruvate, acetyl-CoA, TCA cycle metabolites, and amino acids (Fig. 1; Table 2). Technologies for the conversion of cyanobacterial biomass to useful products or feedstocks are also discussed.

Synechococcus UTEX 2973c Synechococcus PCC 7942 Synechococcus PCC 7002 Synechocystis PCC 6803 Synechocystis PCC 6803 Synechococcus PCC 7942 Synechococcus PCC 7942 Synechococcus PCC 7942 Synechocystis PCC 6803

Sucrose

Erythritol

Glycerol

E4P

DHAP

1,2-Propanediol

1,2-Propanediol

1,3-Propanediol

Glycerol

Mannitol

Glucose, fructose

Sucrose

Sucrose

Sucrose

Chassis strain Synechococcus PCC 7942 Synechocystis PCC 6803 Synechococcus PCC 7942 Synechococcus UTEX 2973

End product Sucrose

F6P

Hub metabolites F6P, G1P

0.90 g L1

0.15 g L1

1.22 g L1

1.24 g L1

1.316 g L1

0.256 g L1

1.1 g L1

0.045 g L1

3.34 g L1

8 g L1a, 3.4 g L1b

0.578 g L1

0.14 g L1

Titer 2.6 g L1

Productivity 0.866 g L1 day1 0.014 g L1 day1 0.193 g L1 day1 1.9 g L1 day1a, 1.1 g L1 day1b 0.853 g L1 day1 0.009 g L1 day1 0.15 g L1 day1 0.009 g L1 day1 0.077 g L1 day1 0.062 g L1 day1 0.061 g L1 day1 0.015 g L1 day1 0.056 g L1 day1

Table 2 Titers and productivities of biofuels and chemicals produced by cyanobacteria

Express gpd1, gpp2 (hor2), dhaB1, dhaB2, dhaB3, gdrA, gdrB, yqhD Express sADH (C. beijerinckii), yqhD, mgsA Express mgsA, yqhD, adh

Express gpp1

Express gpp2; salt stress

Express mtlD & mlp; and inactivate glgA1 & glgA2 Overexpress tm1254, gld1 and pkt

Overexpress inv, glf, galU; salt stress

Overexpress cscB; salt stress

Strategy Overexpress cscB; knockout invA, glgC; salt stress Overexpress sps, spp, ugp, cscB; knockout ggpS, ggtCD; salt stress Overexpress sps, glgC, cscB; knockout invA, glgC; salt stress Overexpress cscBa; overexpress cscB, sps, spp b

(continued)

Niederholtmeyer et al. (2010) Jacobsen and Frigaard (2014) van der Woude et al. (2016) Savakis et al. (2015) Wang et al. (2015) Hirokawa et al. (2016, 2017b) Li and Liao (2013) David et al. (2018)

Song et al. (2016)

Lin et al. (2020)

Qiao et al. (2018)

Note Ducat et al. (2012) Du et al. (2013)

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . . 265

Synechocystis PCC 6803 Synechococcus PCC 7002 Synechococcus PCC 7942 Synechococcus PCC 7942 Synechococcus PCC 7942 Synechocystis PCC 6803 Synechococcus PCC 7942

L-Lactate

Acetyl-CoA

Ethanol

Pyruvate

3Hydroxypropionate 3Hydroxypropionate 3-Hydroxybutyrate

2,3-Butanediol

Isobutyraldehyde

L-Lactate

L-Lactate

D-Lactate

D-Lactate

Isoprene

Chassis strain Synechococcus PCC 7942 Synechococcus PCC 7942 Synechocystis PCC 6803 Synechocystis PCC 6803 Synechococcus PCC 7942 Synechocystis PCC 6803

End product D-Lactate

GAP, pyruvate

Hub metabolites

Table 2 (continued)

1.22 g L1

0.837 g L1

0.659 g L1

2.38 g L1

1.1 g L1

0.80 g L1

0.837 g L1

1.8 g L1

0.829 g L1

1.14 g L1

5.5 g L1

1.26 g L1

Titer 1.23 g L1

0.060 g L1 day1 0.20 g L1 day1 0.138 g L1 day1 0.113 g L1 day1 0.041 g L1 day1 0.140 g L1 day1 0.044 g L1 day1

Productivity 0.051 g L1 day1 0.06 g L1 day1 0.21 g L1 day1 0.048 g L1 day1 0.083 g L1 day1 0.064 g L1 day1

Express NADPH-dependent mcr, pntAB, accBCAD, birA Express nphT7, phaA, phaB1, tesB, pptesB

Express NADPH-dependent mcr, msr

Express alsS, alsD, adh

Express NADPH-dependent ldh,; suppress glnA Express kivd, alsS-ilvC-ilvD

Express ldh, pk

Express ldh

Express NADPH-dependent ldhD, lldP

Express two copies of pdc, slr1192; knock out slr1993-slr1994 Express gldA101, sth

Express idi-GGGS-ispS, dxs, ispG

Strategy Express mgsA, lldP, gloAB

Wang et al. (2016) Ku and Lan (2018)

Angermayr and Hellingwerf (2013) Angermayr et al. (2014) Gordon et al. (2016) Atsumi et al. (2009) Oliver et al. (2013) Lan et al. (2015)

Varman et al. (2013) Li et al. (2015)

Note Hirokawa et al. (2017a) Gao et al. (2016a) Gao et al. (2012)

266 B. Wang et al.

Fumarate

p-Coumaric acid

3Hydroxypropionate Lysine

Guanidine

Fumarate

Tyrosine

Aspartate

Arginine

Synechocystis PCC 6803 Synechocystis PCC 6803 Synechococcus PCC 7942 Synechocystis PCC 6803 Synechocystis PCC 6803 Synechococcus PCC 7942 Synechococcus PCC 7002 Synechocystis PCC 6803

Synechocystis PCC 6803 Synechococcus PCC 7942 Synechocystis PCC 6803

0.587 g L1

0.37 g L1

0.186 g L1

0.083 g L1

0.13 g L1

0.43 g L1

–

–

4.8 g L1

0.404 g L1

1.84 g L1

0.171 g L1 day1 0.207 g L1 day1 0.054 g L1 day1 0.03 g L1 day1 0.021 g L1 day1 0.012 g L1 day1 0.07 g L1 day1 0.084 g L1 day1

0.184 g L1 day1 0.034 g L1 day1 0.171 g L1 day1

Express Ae_adc, Skpyd4, msr (M. sed), ppc, aspC Express jbjE (lysO), lysC [AK(T369I); feedback-resistant] Express 2 efe

Express sam8, knock out slr1573

ΔfumC, Δzwf

Express 3 efe, knock out sll1981 & slr0370 Express kgd, gabD, ppc, gltA

Express phaA, phaB1, tesB; knock out phaEC Express nphT7, phaB, phaJ, ter, pduP (O2tolerant), yqhD Express pduP-slr1192OP; ccr-phaJ-pkPa, nphT7-phaB(T173S)-ptaBs; knock out phaEC and ach Express 2 efe

Korosh et al. (2017) Wang et al. (2019)

Lan et al. (2015)

Xue et al. (2014)

Lan and Wei (2016) Du et al. (2019)

Ungerer et al. (2012) Zhu et al. (2015)

Liu et al. (2019)

Wang et al. (2018b) Lan et al. (2013)

b

With NaCl stress Without NaCl stress c The Synechococcus UTEX 2973 strain used in this study has a genotype slightly different from that of the strain from Pakrasi’s lab (Ungerer et al. 2019)

a

Succinate

Ethylene

Ethylene

Succinate

α-Ketoglutarate

1-Butanol

1-Butanol

3-Hydroxybutyrate

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . . 267

268

B. Wang et al.

Fig. 1 Biofuels and chemicals derived from central hub metabolites in cyanobacteria. G1P glucose-1-phosphate, G6P glucose-6-phosphate, F6P fructose-6-phosphate, FBP fructose-1,6bisphosphate, DHAP dihydroxyacetone phosphate, GAP glyceraldehyde phosphate, E4P erythrose-4-phosphate, S7P sedoheptulose-7-phosphate, R5P ribose-5-phosphate, X5P Xylulose5-phosphate, Ru5P ribulose-5-phosphate, RuBP ribulose-1,5-bisphosphate, 3PGA 3-phosphoglycerate, PEP phosphoenolpyruvate, DXP 1-deoxy-D-xylulose-5-phosphate, OAA oxaloacetate, CIT citrate, ICI isocitrate, AKG α-ketoglutarate, SUC succinate, FUM fumarate, MAL malate, PHB poly-3-hydroxybutyrate, 3HB 3-hydroxybutyrate, 3HP 3-hydroxypropionate

4.1

Derivatives from Sugar Phosphates

Mannitol has been photosynthetically produced from fructose-6-phosphate using an engineered Synechococcus PCC 7002 strain that overexpresses two heterologous enzymes: mannitol-1-phosphate dehydrogenase (MtlD; NADH dependent) and the mannitol-1-phosphatase (Mlp). Inactivation of glycogen biosynthesis through knocking out both glgA1 and glgA2 genes enhanced the mannitol production in this strain, leading to a maximum titer of 1.1 g L1 mannitol within 12 days of photoautotrophic cultivation. It is noteworthy that the mannitol-producing strain did not achieve chromosome segregation and was genetically unstable. The

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

269

biosynthesized mannitol was partially retained inside the cells, while over 70% was excreted out into the medium through an unknown mechanism (Jacobsen and Frigaard 2014). Erythritol, a non-caloric and non-cariogenic sweetener, has been biosynthesized through shunting carbon flux from erythrose-4-phosphate (E4P) in cyanobacteria. The biosynthesis was achieved by overexpressing two enzymes, i.e., the E4P phosphatase TM1254 and the erythrose reductase Gld1, in an engineered strain of Synechocystis PCC 6803. Overexpression of phosphoketolase (Pkt) slightly increased the productivity, resulting in titers up to 256 mg L1 erythritol excreted into the medium within 28 days (van der Woude et al. 2016). Sucrose accumulates in many wild-type cyanobacterial species, especially in freshwater strains, as a compatible solute to combat salt stress from the environment (Kolman et al. 2015; Kirsch et al. 2019; Klahn and Hagemann 2011). Sucrose biosynthesis branches off the glucose-1-phosphate and fructose-6-phosphate nodes (Fig. 1). Co-expression of invertase (InvA) and a hexose transporter (Glf) endowed cyanobacteria with the ability to secrete hexose out of the cell (Niederholtmeyer et al. 2010). Expression of a heterologous sucrose/H+ symporter, CscB, has proven effective for promoting the secretion of sucrose from the engineered Synechococcus PCC 7942 strain under salt-stressed conditions (Ducat et al. 2012). Inactivation of the invertase (encoded by invA) and blocking the glycogen biosynthesis pathway through deleting glgC further increased sucrose productivity. Up to 2.6 g L1 sucrose was secreted into the culture medium after three days of photoautotrophic cultivation with NaCl supplemented to a concentration of 150 mM (Ducat et al. 2012). Since then, Synechocystis PCC 6803 and Nostoc PCC 7120 have also been exploited for the production of sucrose (Du et al. 2013), and most recently the engineered cscB-expressing fast-growing Synechococcus UTEX 2973 strain was able to produce 8 g L1 sucrose within 5 days when grown photoautotrophically with 150 mM NaCl in the medium (Lin et al. 2020). This strain was further engineered to co-overexpress sucrose-phosphate synthase (SPS) and sucrosephosphate phosphatase (SPP) to redirect the carbon flux from glycogen biosynthesis to sucrose production, which led to the photosynthetic production of 3.4 g L1 sucrose in BG11 medium without addition of NaCl (Lin et al. 2020).

4.2

Derivatives from DHAP

Glycerol, 1,3-propanediol, and 1,2-propanediol have been produced from DHAP in cyanobacteria. Two enzymes are required to convert DHAP to glycerol, i.e., glycerol-3-phosphate dehydrogenase (G3PDH) and glycerol-3-phosphatase (GPP). Since Synechococcus PCC 7942 has a native G3PDH, expression of a single gpp1 gene from S. cerevisiae was sufficient for the engineered cyanobacterial strain to produce up to 1.24 g L1 glycerol in 20 days (Wang et al. 2015). Similarly, expression of a single gpp2 gene from S. cerevisiae in Synechocystis PCC 6803 led to production of 1.316 g L1 glycerol within 17 days in cultures supplemented

270

B. Wang et al.

with 200 mM NaCl (Savakis et al. 2015). Starting from glycerol, additional expression of glycerol dehydratase (encoded by dhaB1, dhaB2, dhaB3), an aldehyde reductase (YqhD), and the reactivating factor (encoded by gdrA, gdrB) led to the production of 1,3-propanediol to a titer of 1.22 g L1 within 20 days in engineered Synechococcus PCC 7942 (Hirokawa et al. 2016, 2017b). 1,2-Propanediol was produced in an engineered Synechococcus PCC 7942 strain after expressing a three-step heterologous pathway composed of methylglyoxal synthase (encoded by mgsA), an aldehyde reductase (encoded by yqhD) and a secondary alcohol dehydrogenase (encoded by adh). The titer of 1,2-propanediol reached 150 mg L1 after 10 days of cultivation (Li and Liao 2013). The same heterologous pathway was later expressed in Synechocystis PCC 6803, and the 1,2-propanediol titer reached up to 0.9 g L1 within 16 days under photoautotrophic culture conditions (David et al. 2018). Through co-expression of methylglyoxal synthase (encoded by mgsA), glyoxalase (encoded by gloAB) and the lactate transporter (encoded by lldP), D-lactate was successfully produced by redirecting flux from DHAP. Up to 1.23 g L1 D-lactate has been produced with the engineered Synechococcus PCC 7942 strain using this pathway (Hirokawa et al. 2017a).

4.3

Derivatives from Pyruvate

Starting from pyruvate, a number of compounds have been biosynthesized in cyanobacteria, including ethanol (Deng and Coleman 1999; Gao et al. 2012), D-lactate (Li et al. 2015; Hirokawa et al. 2017a; Varman et al. 2013), L-lactate (Angermayr et al. 2014; Gordon et al. 2016), isobutyraldehyde (Atsumi et al. 2009), isobutanol (Atsumi et al. 2009), 2,3-butanediol (Oliver et al. 2013; Savakis et al. 2013), 2-methyl-1-butanol (Shen and Liao 2012), isoprene (Gao et al. 2016a; Lindberg et al. 2010), and terpenoids (Lin and Pakrasi 2019) (Fig. 1). As an early effort, a synthetic pathway composed of a pyruvate decarboxylase (Pdc) and an alcohol dehydrogenase (Adh) was expressed in Synechococcus PCC 7942 via a shuttle vector, and cyanobacterial production of ethanol was thereafter demonstrated (Deng and Coleman 1999). Later, an operon composed of the pyruvate decarboxylase and an endogenous alcohol dehydrogenase (encoded by slr1192) was introduced into Synechocystis PCC 6803. With two copies of the operon inserted at two different loci of the chromosome, as well as inactivation of the endogenous polyhydroxybutyrate synthase (encoded by slr1993-slr1994), the endeavor led to production of 5.5 g L1 ethanol within 26 days (Gao et al. 2012). In order to realize photosynthetic production of isobutyraldehyde, Atsumi and colleagues constructed a hybrid pathway that consists of ketoacid decarboxylase (Kivd), acetolactate synthase (AlsS), acetohydroxy acid isomeroreductase (IlvC), and dihydroxy-acid dehydratase (IlvD). Co-overexpression of this synthetic pathway and the Rubisco enzyme (RbcLS) in Synechococcus PCC 7942 resulted in production of 1.1 g L1 isobutyraldehyde over 8 days. Additional co-overexpression of YqhD led to production of 450 mg L1 isobutanol in 6 days (Atsumi et al. 2009). 2,3-Butanediol has

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

271

been produced from pyruvate in both Synechocystis PCC 6803 and Synechococcus PCC 7942 (Savakis et al. 2013; Oliver et al. 2013), with a peak titer of 2.38 g L1 (R, R)-2,3-butanediol achieved with the engineered Synechococcus PCC 7942 after 21 days of photoautotrophic cultivation. The pathway consists of the acetoacetate synthase (AlsS), 2-acetolactate decarboxylase (AlsD), and a secondary alcohol dehydrogenase (Adh) (Oliver et al. 2013). In recent years, photosynthetic production of D- and L-lactate have been extensively studied, not only because they are important building blocks for synthesizing biocompatible and biodegradable plastics, but also because their biosynthesis from pyruvate necessitates only one single enzyme, lactate dehydrogenase (LDH). Varman and coworkers expressed a mutated glycerol dehydrogenase (NADHdependent; encoded by glydA101) that exhibits high D-lactate dehydrogenase activity in Synechocystis PCC 6803, and found that D-lactate biosynthesis occurred at the late photoautotrophic growth phase. Co-expression of a heterologous transhydrogenase (encoded by sth), which interconverts NADH and NADPH, significantly improved productivity, leading to photosynthetic production of 1.14 g L1 D-lactate in 24 days (Varman et al. 2013). Li et al. expressed an NADPH-dependent D-lactate dehydrogenase (LdhD) and co-expressed a lactate transporter (LldP) in Synechococcus PCC 7942, resulting in production of 829 mg L1 D-lactate within 10 days (Li et al. 2015). Angermayr and colleagues engineered Synechocystis PCC 6803 by overexpressing a heterologous LDH, and the resultant strain was able to produce 1.8 g L1 L-lactate in 4 weeks (Angermayr and Hellingwerf 2013). Expression of pyruvate kinase (Pyk) in addition to LDH resulted in increased carbon partitioning into L-lactate, up to 50% of fixed carbon, in the engineered Synechocystis PCC 6803 strain (Angermayr et al. 2014). In an engineered Synechococcus PCC 7002 strain, overexpression of an NADPH-dependent LDH in concert with repression of glnA using CRISPRi resulted in photosynthetic production of 800 mg L1 L-lactate within 4 days (Gordon et al. 2016). Terpenoids are widely used in pharmaceuticals, nutraceuticals, fragrances/ flavors, and industrial chemicals (Moser and Pichler 2019). In cyanobacteria, terpenoids play essential roles in light harvesting (chlorophyll and carotenoids), electron transfer (plastoquinone and ubiquinone), and photoprotection (carotenoids) (Lin and Pakrasi 2019). Cyanobacteria have an endogenous terpenoid biosynthetic pathway, the 2-C-methyl-D-erythritol 4-phosphate (MEP) pathway, through which terpenoids are biosynthesized from GAP and pyruvate (Fig. 1). However, compared to other chemicals produced from pyruvate by cyanobacteria, terpenoids yields had been about two orders of magnitude lower due to apparent endogenous regulation of the MEP pathway, until a recent study (Lin and Pakrasi 2019; Moser and Pichler 2019). Gao et al. expressed a plant-derived high-activity isoprene synthase (encoded by ispS), co-overexpressed the isopentenyl pyrophosphate isomerase (IDI) and the 1-deoxy-D-xylulose 5-phosphate synthase (DXS), and identified a pathway bottleneck, IspG, using kinetic flux profiling in Synechococcus PCC 7942. This combinatory approach successfully eliminated the bottlenecks in isoprene biosynthesis, which resulted in photosynthetic production of 1.26 g L1 isoprene within 21 days (Gao et al. 2016a).

272

4.4

B. Wang et al.

Derivatives from Acetyl-CoA

A variety of compounds of interest have been photosynthetically produced from acetyl-CoA, such as PHB (Kamravamanesh et al. 2019), 3-hydroxybutyrate (3HB) (Wang et al. 2018b), 3-hydroxypropionate (3HP) (Lan et al. 2015; Wang et al. 2016), ethanol (Lan et al. 2013), acetate (Hirokawa et al. 2020), acetone (Zhou et al. 2012; Lee et al. 2020), isopropanol (Hirokawa et al. 2020), 1-butanol (Lan et al. 2013; Liu et al. 2019), terpenoids (Lin and Pakrasi 2019), free fatty acids (Ruffing 2014; Liu et al. 2011), fatty alcohols (Yunus and Jones 2018; Tan et al. 2011), alkane (Wang et al. 2013b), and triacylglycerol (Aizouq et al. 2020). A few cases showed relatively high titers or productivities. Lan et al. expressed a monofunctional malonyl-CoA reductase (Mcr) in conjugation with a malonate semialdehyde reductase (Msr) in Synechococcus PCC 7942, and as a result, 3HP was synthesized to a titer of 665 mg L1 within 16 days (Lan et al. 2015). Wang et al. expressed a bifunctional malonyl-CoA reductase (Mcr) that serves as both alcohol and aldehyde dehydrogenases in Synechocystis PCC 6803 and achieved a 3HP titer of 689 mg L1 within 6 days. Additional expression of acetyl-CoA carboxylase (AccBCAD), biotinilase (BirA), and NAD(P) transhydrogenase (PntAB) further increased the 3HP productivity, reaching 837 mg L1 over 6 days (Wang et al. 2016). (R)-3Hydroxybutyrate (3HB) is biosynthesized from acetyl-CoA via three enzymatic steps, which are catalyzed by thiolase (PhaA), acetoacetyl-CoA reductase (PhaB), and thioesterase (TesB) (Wang et al. 2013a). The thiolase PhaA can be substituted by the combination of acetyl-CoA carboxylase (AccABCD) and acetoacetyl-CoA synthase (NphT7) (Ku and Lan 2018). Through identification and mitigation of the bottlenecks of this pathway in an engineered Synechocystis PCC 6803 strain, a recent study demonstrated photosynthetic production of 1.84 g L1 (R)-3HB within 10 days of photoautotrophic cultivation (Wang et al. 2018b). Starting from acetylCoA, biosynthesis of 1-butanol can be achieved via seven reactions carried out by acetyl-CoA carboxylase (AccABCD), acetoacetyl-CoA synthase (NphT7), acetoacetyl-CoA reductase (PhaB), enoyl-CoA hydratase (PhaJ), enoyl-CoA reductase (Ter or Ccr), CoA-acylating aldehyde dehydrogenase (PduP), and alcohol dehydrogenase (YqhD or Slr1192) under photo-oxygenic conditions (Lan et al. 2013; Liu et al. 2019). Most recently, Liu et al. systematically optimized the 1-butanol pathway gene expression levels, overexpressed phosphoketolase (Pkt) and phosphotransacetylase (Pta), and inactivated the acetyl-CoA hydrolase (Ach) and the PHB synthase (PhaEC). These efforts resulted in a maximal 1-butanol titer of 4.8 g L1 with the engineered Synechocystis PCC 6803 under photoautotrophic conditions. The average 1-butanol productivity during the 28-day cultivation period was 171 mg L1 day1, with a maximal rate of 302 mg L1 day1 (Liu et al. 2019).

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

4.5

273

Derivatives from TCA Cycle Metabolites

Succinate is a natural metabolite in cyanobacteria, which is usually not excreted into the culture medium unless under anaerobic conditions (Hasunuma et al. 2016). Overexpression of α-ketoglutarate decarboxylase and succinate semialdehyde dehydrogenase in Synechococcus PCC 7942 resulted in enhanced succinate production under phototrophic conditions, but the growth of the strain was severely inhibited, indicating a potential metabolic stress. Further overexpression of heterologous PEP carboxylase (PPC) and citrate synthase (GltA) restored the cell growth rate and dramatically enhanced succinate production. Up to 430 mg L1 succinate was produced after 8 days of cultivation (Lan and Wei 2016). Other TCA cycle metabolites, such as malate and fumarate, were also excreted by the engineered cyanobacterial strains under photoautotrophic conditions though at lower titers (Hu et al. 2018; Du et al. 2019). Interestingly, α-ketoglutarate, a key C/N signaling metabolite, has been excreted into the medium by ΔglgC cyanobacterial strains under nitrogen starvation conditions (Carrieri et al. 2012). Expression of one single enzyme, the ethylene-forming enzyme (EFE), enables cyanobacteria to produce ethylene directly from α-ketoglutarate (AKG). In the efeexpressing Synechococcus PCC 7942 strain, however, ethylene production was unstable and genetic investigation showed that the ethylene expression cassette was consistently disrupted after a few generations of cell division (Sakai et al. 1997; Takahama et al. 2003). The efe gene was later codon-optimized and introduced into Synechocystis 6803, which exhibited stable ethylene formation under photosynthetic conditions (Guerrero et al. 2012; Ungerer et al. 2012). Doubling the number of EFE expression cassettes on the chromosome resulted in 171 mg L1 day1 ethylene from cultures of engineered Synechocystis PCC 6803 (Ungerer et al. 2012). Expression of three copies of efe combined with inactivation of AKG decarboxylase (encoded by sll1981) and succinic semialdehyde dehydrogenase (encoded by slr0370) led to production of 207 mg L1 day1 ethylene in an engineered Synechocystis PCC 6803 strain (Zhu et al. 2015). Further optimization of the ribosome-binding site and the promoter for the efe gene led to remarkably high-level expression of EFE, up to 12.6% of total soluble protein, so that the expression of EFE was no longer the limiting step for ethylene production in the engineered Synechocystis (Xiong et al. 2015; Wang et al. 2018a). Most recently, it was found that co-overexpression of PPC was able to enhance the ethylene production in an efe-expressing Synechocystis strain, which almost doubled ethylene productivity under the examined culture condition (Durall et al. 2020).

4.6

Derivatives from Amino Acids

A few studies have demonstrated the production of compounds of interest directly from amino acids in cyanobacteria. Tyrosine, biosynthesized from PEP and E4P, has

274

B. Wang et al.

been successfully converted to the plant secondary metabolite p-coumaric acid in Synechocystis PCC 6803. This pathway was installed through expression of a heterologous tyrosine ammonia lyase and inactivation of the slr1573 gene that encodes a putative laccase, producing up to 82.6 mg L1 p-coumaric acid within 4 days (Xue et al. 2014). From a growth-coupled approach, 130 mg L1 fumarate was produced from aspartate after four days of cultivation of the ΔfumC-Δzwf strain derived from Synechocystis PCC 6803 (Du et al. 2019). Korosh et al. engineered Synechococcus PCC 7002 to express a feedback-resistant aspartate kinase [LysC (T369I)] and a lysine transporter (LysO), which rendered production of about 370 mg L1 L-lysine at a rate of 70 mg L1 day1 (Korosh et al. 2017). Arginine and aspartate form cyanophycin, the nitrogen reservoir in many cyanobacterial species, and arginine metabolism plays an essential role in nitrogen turnover in cyanobacteria (Zhang et al. 2018; Flores et al. 2019). Recently, the nitrogen-rich compound, guanidine, has been produced directly from arginine, catalyzed by the ethylene-forming enzyme, in an engineered Synechocystis strain under photoautotrophic conditions. Up to 587 mg L1 guanidine was excreted in to the medium within 7 days. Biosynthesis and excretion of guanidine was also demonstrated with an engineered efe-expressing Nostoc PCC 7120 strain under N2-fixing conditions (Wang et al. 2019).

4.7

Biomass Conversion

Rather than directly synthesizing products of interest under photo-oxygenic conditions, an alternate strategy is to produce cyanobacterial biomass as a feedstock for biosynthesis of other useful products. From this approach, Husunuma and colleagues were able to ferment Synechocystis 6803 biomass under dark anoxic conditions to produce various organic acids, including acetate, D-lactate, and succinate. Through genetic reprograming and optimization of fermentation conditions, the titers of each specific acid were increased significantly. Because this method can start with relatively high concentrations of biomass, it usually renders production of organic acids at titers much higher than those achieved under photo-oxygenic conditions. Since cyanobacterial cells usually have high protein content, ~50% of the dry cell weight (Broddrick et al. 2016), as well as many other essential nutrients, species like Spirulina can be used as dietary supplements for humans and animals (Godlewska et al. 2019). Alternatively, the cyanobacterial protein hydrolysate has been used as a feedstock for bioproduction of C4 and C5 alcohols (isobutanol, 2-methyl-1-butanol, and 3-methyl-1-butanol) using engineered E. coli strains that co-overexpress deaminase, keto acid decarboxylase, and alcohol reductase. About 4.0 g L1 biofuel composed of 50% isobutanol, 47% C5 alcohols, and 3% ethanol have been produced from starting biomass containing ~22 g L1 of amino acids, achieving 56% of the theoretical yield (Huo et al. 2011). In addition, cyanobacterial biomass slurry could be converted to biofuel intermediates (BFI), which are mostly free fatty acids and olefins, through hydrothermal liquefaction. In this case, it has been found that the

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

275

biomass composition exerts a large impact on the BFI yield. While the BFI yield is positively correlated to the lipid and medium-chain-length polyhydroxyalkanoate content, it is negatively correlated to the carbohydrate and polyhydroxybutyrate content of the cyanobacterial biomass (Dong et al. 2020).

5 Modification of Cyanobacterial Framework for Improved Performance Cyanobacteria have evolved to survive and thrive in their own ecological niches but not for the purpose of biofuel and chemical production in manmade settings. While efforts to redirect metabolic fluxes from hub metabolites toward products of interest via optimizing local pathways have been fruitful and enhanced photosynthesis has been observed in some cases (Oliver et al. 2013; Abramson et al. 2016; Gao et al. 2016a; Zhou et al. 2016a), limitations to the overall photosynthesis efficiency or the availability of central metabolites may become bottlenecks when productivities reach certain thresholds. Achieving further improvements may require refactoring the cellular framework by altering its intrinsic metabolic regulation in order to optimize cyanobacteria for enhanced bioproduction.

5.1

Enhancing Photosynthetic Efficiency

Although cyanobacteria exhibit higher photosynthetic efficiency (solar-to-biomass) than plants, typical photosynthetic efficiencies are still only about 3% (Blankenship et al. 2011; Claassens et al. 2016), and the best case has been assumed to be about 10% (Weyer et al. 2010; Markham et al. 2016). A vast majority of solar energy is lost due to the fact that photosynthesis only utilizes a portion of the incident solar radiation. The sub-spectral range of sunlight used by most cyanobacteria, microalgae and higher plants for photochemistry is 400–700 nm, called photosynthetically active radiation (PAR), which is about 50% of the full spectrum of solar energy (Weyer et al. 2010). For a long time, light with wavelengths >700 nm had been deemed as photosynthetically inactive (i.e., the “red limit”) until the discovery of chlorophyll d and f (Chen et al. 2010; Chen and Blankenship 2011). Both these two chlorophylls are isolated from certain cyanobacterial species grown under infrared light conditions, and they are able to absorb and use radiation at 700–750 nm for photosynthesis, with chlorophyll f the most red-shifted so far isolated from oxygenic phototrophs (Chen et al. 2010; Gan et al. 2014; Nurnberg et al. 2018). It has been reported that expanding the standard PAR to include the 700–750 nm infrared zone would increase the photon flux by 19% (Chen and Blankenship 2011).

276

B. Wang et al.

The capacity for energy absorption and utilization plays a key role in photosynthesis efficiency. The sunlight reaching the earth’s surface is quite dynamic during the day, and when the intensity exceeds the capacity of photosynthesis, excess electrons generated within the photosynthetic reaction center result in increased formation of damaging reactive oxygen species (ROS). Cyanobacteria have evolved non-photochemical quenching (NPQ) as a photo-protective mechanism to prevent damage caused by high light (Misumi et al. 2016). Instead of channeling the solar energy to charge separation, NPQ dissipates the absorbed excess solar energy as heat, thereby harmlessly grounding the excited state of chlorophyll to an inactive state. However, the loss of solar energy as heat during NPQ could be significant, up to 50% of the absorbed solar energy (Gorbunov et al. 2011), which could otherwise be used for photosynthetic production. Cyanobacterial strains that are able to redirect more solar energy for biomass production rather than dissipation under elevated light intensities are more desirable. Studies have found that some cyanobacterial strains are more tolerant to high light and exhibit fast growth under high-light conditions, which seems more promising in biotechnological applications. A few cyanobacterial strains with such a trait have been reported thus far (Ludwig and Bryant 2012; Yu et al. 2015; Jaiswal et al. 2018; Ungerer et al. 2018a; Włodarczyk et al. 2020). In a recent study, Ungerer et al. demonstrated that modification of three genes on the Synechococcus PCC 7942 genome accelerates the cell growth by approximately threefold, achieving a rate comparable to the fast-growing strain Synechococcus UTEX 2973 (Ungerer et al. 2018b). Among the three mutations, two exhibit independent beneficial effects toward accelerated cell growth, i.e., the C252Y replacement in the alpha subunit of the ATP synthase, AtpA, and the E260D substitution in the NAD+ kinase, PpnK. These two mutations result in higher enzyme activities toward ATP generation and NADPH production and, therefore, enable accelerated electron flow along the electron transport chain (Ungerer et al. 2018b). Alternatively, truncating the photosystem light-harvesting antenna has been proven a feasible strategy to minimize photodamage or energy loss due to NPQ (Melis 2009). Complete removal of phycocyanin from Synechocystis PCC 6803 (ΔcpcAB) substantially reduced the absorbance at 625 nm and, therefore, increased the saturation light intensity by two times. Under simulated bright sunlight (2000μE m2 s1) with a cell density of 0.5–1.0 g DCW L1, the engineered strain exhibited a higher growth rate and achieved 57% higher biomass accumulation compared to that of the wild-type strain (Kirst et al. 2014).

5.2

Improving Carbon Assimilation

Although cyanobacteria have evolved a potent carbon concentration mechanism to facilitate CO2 uptake and fixation from the relatively sparse CO2 in the atmosphere, the slow rate of CO2 assimilation is still one of the hurdles to overcome. Kamennaya et al. overexpressed the endogenous bicarbonate transporter, BicA, in Synechocystis PCC 6803 and demonstrated that the engineered strain grew nearly two times faster

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

277

than the wild-type when the cultures were aerated with air. When the cultures were grown under 0.5% or 5% CO2, the BicA overexpression had neutral or negative effect on the cell growth rate during the exponential growth phase but prolonged the active growth, which ultimately improved accumulation of biomass in the end (Kamennaya et al. 2015). The flux through the Calvin-Benson-Basham (CBB) cycle, which is the major pathway of carbon fixation in cyanobacteria, is also suboptimal for biotechnological applications. Ribulose-1,5-bisphosphate carboxylase/oxidase (Rubisco), which catalyzes the CO2 fixation reaction in the CBB cycle, also reacts with O2 to produce byproduct metabolites that feed into the photorespiration pathway. Although it is necessary to recycle these byproducts to avoid accumulation of toxic intermediates within the cell (Yu et al. 2018), many consider photorespiration to be a “wasteful” process because it consumes energy and leads to loss of CO2 that was previously fixed (Peterhansel et al. 2013). Furthermore, Rubisco exhibits extremely slow enzyme kinetics, and its carboxylation activity is believed to be the bottleneck for carbon fixation (Ducat and Silver 2012). In an effort to increase carbon fixation, Atsumi and colleagues overexpressed heterologous Rubisco-encoding genes, rbcLS, in an isobutyraldehyde-producing Synechococcus PCC 7942 strain, leading to an approximately two times higher isobutyraldehyde productivity (Atsumi et al. 2009). Nevertheless, overexpression of rbcLS in a 3HP-producing Synechocystis PCC 6803 strain did not result in elevated 3HP production (Wang et al. 2016), indicating that the Rubisco activity might not be limiting 3HP biosynthesis in that case. Besides overexpressing Rubisco, significant efforts have been invested in evolving more efficient Rubisco enzymes (Wilson and Whitney 2017; Satagopan et al. 2019). For example, it was found that the F140I mutation in the large subunit of Rubisco resulted in a threefold improvement in the carboxylation efficiency despite a 9% reduction in specificity toward CO2 versus O2. Replacing the wild-type Rubisco with the F140I mutant led to 25% lower level expression of Rubisco in Synechocystis, but the engineered strain exhibited an identical growth rate compared to the wild-type strain and its oxygen evolution rate was increased by 20% (Durao et al. 2015). Other CBB cycle enzymes have also been shown to limit CO2 fixation in cyanobacteria (Liang et al. 2018b). Similar to what has been found in higher plants, a recent study reported that expression of Rubisco, fructose-1,6/sedoheptulose-1,7biphosphatase (encoded by glpX), fructose-bisphosphate aldolase (encoded by fba), and transketolase (encoded by tkt) are the limiting factors for CO2 fixation in Synechocystis PCC 6803. Overexpression of these enzymes resulted in enhanced CO2 assimilation (Douchi et al. 2019), oxygen evolution, growth rate, and biomass accumulation (Liang and Lindblad 2016). When each of the aforementioned enzymes was overexpressed in an ethanol-producing Synechocystis strain (pdc+, adh+), the ethanol titer was increased by 37–69% and the total biomass (including ethanol and cell dry weight) was increased by 7.7–15.1%, respectively, compared to the parental ethanol-producing strain (Liang et al. 2018a).

278

5.3

B. Wang et al.

Rewiring the Central Carbon Metabolism

Central metabolites are not only raw materials for synthesis of essential components in the cell but also precursors for the production of useful biofuels and chemicals. Besides the CBB cycle, cyanobacterial central carbon metabolism also includes glycogen metabolism, Embden–Meyerhof–Parnas (EMP) pathway, Entner– Doudoroff (ED) pathway, the oxidative pentose phosphate (OPP) pathway, the phosphoketolase pathway, the photorespiration pathway, and the tricarboxylic acid (TCA) cycle (Xiong et al. 2017; Tang et al. 2013). The cyanobacterial central carbon metabolism exhibits high plasticity that could accommodate overproduction of a spectrum of compounds through flux redistribution (Fig. 1). However, artificially rewiring flux through central carbon metabolism has provided significant productivity improvement in some cases. For instance, by quantitatively profiling the central carbon metabolism of an isobutyraldehyde-producing Synechococcus strain using 13C-MFA, a recent study identified the metabolic flux from PEP to pyruvate as a potential bottleneck for the production of isobutyraldehyde. Overexpression of the pyruvate kinase (Pyk) or a three-step bypass shunt improved the isobutyraldehyde specific productivity by up to 60% (Jazmin et al. 2017), and downregulation of the PDH flux further improved specific productivity by about 60% (Cheah et al. 2020). Overexpression of Pyk also led to enhanced production of L-lactate from pyruvate in an LDH-overexpressing Synechocystis strain (Angermayr et al. 2014), indicating inefficient Pyk activity might be common in cyanobacteria. Downregulation of the PPC flux enabled increased carbon partitioning to isobutyraldehyde and L-lactate; nevertheless, it caused a severe growth defect (Angermayr et al. 2014; Cheah et al. 2020). In another example, the biosynthesis of acetyl-CoA was apparently limiting the downstream production of 1-butanol in cyanobacteria. Overexpression of a phosphoketolase redirected flux directly from the CBB cycle, which significantly augmented the flux toward acetyl-CoA, and thereby increased 1-butanol formation in Synechocystis (Anfelt et al. 2015; Liu et al. 2019). Overexpression of PDH has also proven an effective strategy to enhance the acetyl-CoA biosynthesis from pyruvate, which led to substantial improvement in acetate and isopropanol production in an engineered Synechococcus 7942 strain (Hirokawa et al. 2020). Additionally, enhanced flux toward the TCA cycle has been achieved through overexpressing PPC or citrate synthase in cyanobacteria (Lan and Wei 2016; Durall et al. 2020).

6 The Circadian Clock Regulates Gene Expression and Metabolism in Wild-Type Cyanobacteria The cyanobacterial circadian clock exerts global regulation of biological events and metabolism in the cell. Major cellular activities such as glycogen metabolism, nitrogen fixation, superhelical status of DNA, compaction of the chromosome, cell division, and natural competence are all under control of the cyanobacterial circadian

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

279

clock. The circadian rhythm not only is prominent when cells are subjected to lightdark cycles but also persists when cultures are grown under continuous light (Toepel et al. 2008; Ito et al. 2009; Vijayan et al. 2009; van Alphen and Hellingwerf 2015). The cyanobacterial circadian clock system is encoded by a cohort of circadian oscillator genes including kaiABC, sasA, labA, rpaA, rpaB and cikA, with kaiABC as its core timing mechanism (Swan et al. 2018; Cohen and Golden 2015; Johnson et al. 2017). Different strains may have various copy numbers of the kaiABC genes (Table 1). Most of the oscillating output genes are classified into two groups based on when they are expressed: class I and class II. The peak expression of class I genes is at subjective dusk, whereas class II genes are expressed with a peak at subjective dawn (Cohen and Golden 2015; Johnson et al. 2017). We herein mainly discuss how the circadian clock regulates central metabolism in the model strain Synechococcus PCC 7942 based on information gleaned from the literature.

6.1

The Circadian Clock Governs Oscillation of Glycogen Content

Glycogen is the main energy storage macromolecule in cyanobacteria, and its metabolism manifests high-amplitude rhythms, accumulating during the subjective day and breaking down during the subjective night, even under continuous-light conditions. It has been found that when the circadian clock in Synechococcus PCC 7942 is disrupted, e.g., in cases of ΔkaiBC or ΔcikA, the rhythmic metabolism of glycogen disappears. In the case of ΔcikA, the cells exhibit a dusk-like phenotype, overaccumulating glycogen in both subjective light and dark phases while maintaining a high energy charge (ratio of ATP vs. ATP+ADP). By contrast, ΔsasA cells maintain a lower glycogen content and lower energy charge than the wild-type cells, latching the metabolism in a constant dawn-like phenotype (Pattanayak et al. 2014). Transcriptional analyses revealed that the expression of genes associated with glycogen biosynthesis or degradation shows strong temporal segregation. Expression of the glycogen biosynthetic genes, i.e., pgm and glgC, peaks at dawn, whereas the expression of glycogen degradation genes, i.e., glgP, glgX and malQ, peaks at dusk (Diamond et al. 2015; Vijayan et al. 2009) (Fig. 2).

6.2

The Circadian Oscillator Regulates Global Gene Expression and Metabolism

A random promoter-trap analysis based on a bacterial luciferase reporter revealed that essentially all promoter activities are controlled by the circadian clock under continuous light (LL) conditions (Liu et al. 1995). Recent transcriptomics studies have corroborated that expression levels of most genes in Synechococcus PCC 7942

280

B. Wang et al.

Fig. 2 The circadian clock controls expression of genes associated with central metabolism in Synechococcus PCC 7942 under constant light conditions. Genes in red font are subjective dusk genes, and genes in green font are subjective dawn genes. Genes in black font are arrhythmic or the rhythmicity is unclear

are subject to circadian regulation (Ito et al. 2009; Vijayan et al. 2009). Most genes related to anabolic metabolism exhibit dawn gene expression profiles (expression levels peaking at dawn), whereas most genes related to catabolic metabolism are classified as dusk genes (expression levels peaking at dusk). Particularly, genes involved in the anabolic CBB cycle and glycogen biosynthesis pathway and most TCA cycle genes are dawn genes, whereas genes associated with the glycogen degradation pathway, oxidative pentose phosphate pathway, and hydrogen production are dusk genes (Ito et al. 2009; Vijayan et al. 2009) (Fig. 2). As gene expression levels influence the abundance of enzymes and, therefore, the biochemical reaction rates, it is logical to anticipate that the central metabolism is subject to regulation by the circadian oscillator in cyanobacteria. Manipulation of the circadian clock genes to enhance metabolic flux or favor biosynthesis of certain central metabolites that support biofuel and chemical production is one possible strategy to be tested in the future. In particular, because the circadian clock rhythmically turns promoters up

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

281

and down, this rhythmic regulation means that processes at the trough phase are not as active as they are at the peak phase. If the clock control of a target process can be latched at the peak phase continuously, then the process will have an integrated constitutive activity over the 24-h day that is much higher than if it displays a rhythm.

7 Global Complementary Regulation of Gene Expression Via Manipulation of the Clock The philosophical concepts yin and yang co-exist in nature. These two aspects are opposite to each other yet interdependent. Yin and yang are actually not opposites but are complementary forces that interact with each other to form a whole that is greater than either separate part. Yin and yang remain a balanced state with the two sides waning and waxing within an allowed range, and yin and yang can be interconverted under specific conditions. Interestingly, the global patterns of gene expression show that there is a complementary “yin-yang” regulation by KaiA versus KaiC (Fig. 3). Based on phase profiles, the complementary regulation by KaiA/KaiC indicates that the overexpression of kaiA mainly activates expression of dusk genes and represses expression of dawn genes, whereas kaiC overexpression complementarily activates dawn genes and represses dusk genes (Ito et al. 2009; Xu et al. 2013). The circadian rhythm of KaiC phosphorylation regulates the global patterns of gene expression via other output factors, such as SasA, CikA, RpaA, RpaB, etc., in Synechococcus PCC 7942 (Johnson et al. 2017; Swan et al. 2018). KaiC undergoes rhythms of phosphorylation that are regulated by interactions with KaiA and KaiB. The peak and trough levels of the KaiC phosphorylation rhythm can be mimicked by overexpression of KaiA or KaiC, respectively (Xu et al. 2013). Increased KaiA levels stimulate KaiC phosphorylation (Iwasaki et al. 2002) and inhibit KaiC dephosphorylation (Xu et al. 2003). Therefore, kaiA-OX vs. kaiC-OX inversely switch the KaiC phosphorylation status and the gene expression patterns between dusk (kaiA-OX) and dawn (kaiC-OX) phases, and continuous induction of kaiA or kaiC would confer constitutively increased transcriptional activity of either class I (dusk) or class II (dawn) genes (Xu et al. 2013; Ito et al. 2009).

8 Manipulation of the Circadian Clock for Enhancing Expression of Foreign Genes Due to the pervasive control of promoter activities by kaiA and kaiC expression levels in cyanobacteria, manipulation of these clock genes can be exploited to enhance the expression of genes of interest to near or above peak levels (Fig. 4).

Fig. 3 Complementary regulation of genome-wide gene expression by KaiA/KaiC in Synechococcus PCC 7942, adapted from (Xu et al. 2013). Spatial patterns of gene expression along the entire chromosome elicited by kaiA/kaiC overexpression are shown. The chromosome is represented in a linear format with the expression levels of each gene in response to kaiA-OX (red) or kaiC-OX (blue). Changes of gene expression are shown as the ratio of the transcript abundance in the presence of IPTG to that in the absence of IPTG in continuous light (LL). Each ratio was arranged in ascending order of Synpcc7942 gene number. The lower panel magnifies the region encompassing Synpcc7942_1565 to Synpcc7942_1583 where changes in expression levels regulated by kaiA-OX and kaiC-OX are denoted in red and blue, respectively. Increased gene expression in response to the indicated overexpression is classified as up-regulation, whereas decreased transcript levels are considered down-regulation

282 B. Wang et al.

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

283

Fig. 4 Prospect of enhancing bioproduction via manipulation of the biological clock. By reprograming the circadian clock, the oscillatory gene expression under control of the native circadian clock (depicted by the pink area) can be locked into a dusk or dawn phase to yield a greater accumulation of gene products in constant light over 24 or more hours (shown by the blue cross-hatched area). The figure is revised from (Xu et al. 2013)

Proof-of-principle studies have been accomplished in Synechococcus PCC 7942, in which manipulation of kaiA and kaiC expression levels has been used to optimize the expression of endogenous and foreign genes, including bacterial luciferase, [NiFe] hydrogenases for biohydrogen production, and human proinsulin (Xu et al. 2013). In the KaiA-knockout strain (ΔkaiA), the expression of the Vibrio harveyi luciferase driven by psbAI or kaiBC promoter was extremely weak and was arrhythmic. By contrast, the luciferase abundance and luminescence levels were significantly and/or constantly enhanced in other clock-manipulated strains, such as kaiA-OX, KaiC-Mut (unpublished mutated kaiC), M7942-1, and M7942-2 (unpublished two clockmanipulated mutants) (Fig. 5). As examples to illustrate how manipulation of clock genes can enhance production of foreign proteins in cyanobacteria, the expression of a heterologous [NiFe]-hydrogenase from Alteromonas macleodii Deep Ecotype and a fusion protein of human proinsulin (HPI) with a glutathione S-transferase (GST) tag were examined in the wild-type and kaiA-OX Synechococcus PCC 7942 (Xu et al. 2013). Expression of both exogenous proteins were dramatically enhanced in these clock-manipulated strains (Fig. 6).

9 Conclusions and Prospects Owing to the capability of oxygenic photosynthesis and a relatively simple genetic background, cyanobacteria have been developed as versatile microbial cell factories to directly convert CO2 and sunlight into a spectrum of biofuels and commodity products. Development of synthetic and systems biology tools have greatly accelerated the understanding and manipulation of cyanobacterial metabolism. Through expression and optimization of local biosynthetic pathways to directly tap fluxes from central metabolites, many chemicals have been photosynthetically produced at gram per liter levels. Concomitantly, the relatively low photosynthetic efficiency, slow CO2 fixation rate, and the intrinsic regulation of central metabolism are

284

B. Wang et al.

Fig. 5 Manipulation of the circadian clock for enhancing expression of bacterial luciferase in S. elongatus PCC 7942. (a) Enhanced or reduced luminescence expression levels in wild-type (WT) and different clock mutants KaiA-OX, KaiC-Mut, Out-M1, Out-M2, and ΔKaiA. For better comparison, linear scale (left) and log2 scale (right) are presented. (b) Enhanced accumulation of the psbAIp-driving Vibrio harveyi luciferase (Lux) by IPTG induction of kaiA. (c) Reduced or enhanced accumulation of the KaiBCp-driving luciferase LuxA in different strains. Panel B is adapted from (Xu et al. 2013)

Fig. 6 Manipulation of the circadian clock for enhancing the expression of potentially valuable foreign proteins in Synechococcus PCC 7942. (a) KaiA overexpression (kaiA-OX) enhances accumulation of human proinsulin protein (GST::HPI fusion protein). (b) KaiA enhances expression of heterologous hydrogenases. RC41 means expression of a [NiFe]-hydrogenase from Alteromonas macleodii. “nb” ¼ nonspecific bands. The figure is revised from (Xu et al. 2013)

constraining further improvement of the cyanobacterial productivity. Among the many native regulatory circuits is the circadian clock, which exerts pervasive control over the oscillatory gene expression and cellular events in cyanobacteria.

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

285

Manipulation of the clock genes has been effective in driving either constant up- or down-regulation of the expression of many endogenous genes in situ as well as foreign genes integrated into neutral sites on the genome. We envision that the circadian rhythms of different cyanobacterial strains may be reprogrammed by manipulating core clock genes, i.e., kaiABC, and/or auxiliary clock genes, to lock global gene expression into preferable patterns, which may improve 24/7 (7 days/ week schedule) production of biofuels and commodity chemicals. This tactic may also be applied to any other organism that has an endogenous circadian clock. Acknowledgments The authors would like to thank Dr. Carl Hirschie Johnson for comments on the manuscript. This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research, Genomic Science Program under award numbers DE-SC0019404 and DE-SC0019388, and NIH/NIGMS awards GM 067152 & GM 107434.

References Abernathy MH, Yu J, Ma F et al (2017) Deciphering cyanobacterial phenotypes for fast photoautotrophic growth via isotopically nonstationary metabolic flux analysis. Biotechnol Biofuels 10 (1):273. https://doi.org/10.1186/s13068-017-0958-y Abernathy MH, Czajka JJ, Allen DK et al (2019) Cyanobacterial carboxysome mutant analysis reveals the influence of enzyme compartmentalization on cellular metabolism and metabolic network rigidity. Metab Eng 54:222–231. https://doi.org/10.1016/j.ymben.2019.04.010 Abramson BW, Kachel B, Kramer DM et al (2016) Increased photochemical efficiency in Cyanobacteria via an engineered sucrose sink. Plant Cell Physiol 57(12):2451–2460. https:// doi.org/10.1093/pcp/pcw169 Aizouq M, Peisker H, Gutbrod K et al (2020) Triacylglycerol and phytyl ester synthesis in Synechocystis sp. PCC6803. Proc Nat Acad Sci 117(11):6216–6222. https://doi.org/10.1073/ pnas.1915930117 Alagesan S, Gaudana SB, Sinha A et al (2013) Metabolic flux analysis of Cyanothece sp. ATCC 51142 under mixotrophic conditions. Photosynth Res 118(1-2):191–198. https://doi.org/10. 1007/s11120-013-9911-5 Albers SC, Gallegos VA, Peebles CA (2015) Engineering of genetic control tools in Synechocystis sp. PCC 6803 using rational design techniques. J Biotechnol 216:36–46. https://doi.org/10. 1016/j.jbiotec.2015.09.042 Anfelt J, Kaczmarzyk D, Shabestary K et al (2015) Genetic and nutrient modulation of acetyl-CoA levels in Synechocystis for n-butanol production. Microbial Cell Factories 14(1):167. https:// doi.org/10.1186/s12934-015-0355-9 Angermayr SA, Hellingwerf KJ (2013) On the use of metabolic control analysis in the optimization of cyanobacterial biosolar cell factories. J Phys Chem B 117(38):11169–11175. https://doi.org/ 10.1021/jp4013152 Angermayr SA, van der Woude AD, Correddu D et al (2014) Exploring metabolic engineering design principles for the photosynthetic production of lactic acid by Synechocystis sp. PCC6803. Biotechnol Biofuels 7:99. https://doi.org/10.1186/1754-6834-7-99 Angermayr SA, Gorchs Rovira A, Hellingwerf KJ (2015) Metabolic engineering of cyanobacteria for the synthesis of commodity products. Trends Biotechnol 33(6):352–361. https://doi.org/10. 1016/j.tibtech.2015.03.009

286

B. Wang et al.

Appel J, Hueren V, Boehm M et al (2020) Cyanobacterial in vivo solar hydrogen production using a photosystem I–hydrogenase (PsaD-HoxYH) fusion complex. Nat Energy. https://doi.org/10. 1038/s41560-020-0609-6 Atsumi S, Higashide W, Liao JC (2009) Direct photosynthetic recycling of carbon dioxide to isobutyraldehyde. Nat Biotechnol 27(12):1177–1180. https://doi.org/10.1038/nbt.1586 Battchikova N, Muth-Pawlak D, Aro EM (2018) Proteomics of cyanobacteria: current horizons. Curr Opin Biotechnol 54:65–71. https://doi.org/10.1016/j.copbio.2018.02.012 Behle A, Saake P, Germann AT et al (2020) Comparative dose-response analysis of inducible promoters in cyanobacteria. ACS Synth Biol 9(4):843–855. https://doi.org/10.1021/acssynbio. 9b00505 Behler J, Vijay D, Hess WR et al (2018) CRISPR-based technologies for metabolic engineering in cyanobacteria. Trends Biotechnol 36(10):996–1010. https://doi.org/10.1016/j.tibtech.2018.05. 011 Bishé B, Taton A, Golden JW (2019) Modification of RSF1010-based broad-host-range plasmids for improved conjugation and cyanobacterial bioprospecting. iScience 20:216–228. https://doi. org/10.1016/j.isci.2019.09.002 Blankenship RE, Tiede DM, Barber J et al (2011) Comparing Photosynthetic and Photovoltaic Efficiencies and Recognizing the Potential for Improvement. Science 332(6031):805–809. https://doi.org/10.1126/science.1200165 Broddrick JT, Rubin BE, Welkie DG et al (2016) Unique attributes of cyanobacterial metabolism revealed by improved genome-scale metabolic modeling and essential gene analysis. Proc Natl Acad Sci 113(51):E8344–E8353. https://doi.org/10.1073/pnas.1613446113 Cao YQ, Li Q, Xia PF et al (2017) AraBAD based toolkit for gene expression and metabolic robustness improvement in Synechococcus elongatus. Sci Rep 7(1):18059. https://doi.org/10. 1038/s41598-017-17035-4 Carrieri D, Paddock T, Maness P-C et al (2012) Photo-catalytic conversion of carbon dioxide to organic acids by a recombinant cyanobacterium incapable of glycogen storage. Energy Environ Sci 5(11):9457–9461. https://doi.org/10.1039/C2EE23181F Caspi R, Billington R, Ferrer L et al (2015) The MetaCyc database of metabolic pathways and enzymes and the BioCyc collection of pathway/genome databases. Nucleic Acids Res 44(D1): D471–D480. https://doi.org/10.1093/nar/gkv1164 Cheah YE, Xu Y, Sacco SA et al (2020) Systematic identification and elimination of flux bottlenecks in the aldehyde production pathway of Synechococcus elongatus PCC 7942. Metab Eng 60:56–65. https://doi.org/10.1016/j.ymben.2020.03.007 Chen M, Blankenship RE (2011) Expanding the solar spectrum used by photosynthesis. Trends Plant Sci 16(8):427–431. https://doi.org/10.1016/j.tplants.2011.03.011 Chen M, Schliep M, Willows RD et al (2010) A red-shifted Chlorophyll. Science 329 (5997):1318–1319. https://doi.org/10.1126/science.1191127 Chen Y, Taton A, Go M et al (2016) Self-replicating shuttle vectors based on pANS, a small endogenous plasmid of the unicellular cyanobacterium Synechococcus elongatus PCC 7942. Microbiology (Reading, England) 162(12):2029–2041. https://doi.org/10.1099/mic.0.000377 Claassens NJ, Sousa DZ, dos Santos VAPM et al (2016) Harnessing the power of microbial autotrophy. Nat Rev Microbiol 14(11):692–706. https://doi.org/10.1038/nrmicro.2016.130 Cohen SE, Golden SS (2015) Circadian Rhythms in Cyanobacteria. Microbiol Mol Biol Rev 79 (4):373–385. https://doi.org/10.1128/MMBR.00036-15 David C, Schmid A, Adrian L et al (2018) Production of 1,2-propanediol in photoautotrophic Synechocystis is linked to glycogen turn-over. Biotechnol Bioeng 115(2):300–311. https://doi. org/10.1002/bit.26468 Deng MD, Coleman JR (1999) Ethanol synthesis by genetic engineering in cyanobacteria. Appl Environ Microbiol 65(2):523–528 Diamond S, Jun D, Rubin BE et al (2015) The circadian oscillator in Synechococcus elongatus controls metabolite partitioning during diurnal growth. Proc Natl Acad Sci 112(15):E1916– E1925. https://doi.org/10.1073/pnas.1504576112

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

287

Dong T, Wang B, Xiong W et al (2020) System-level optimization to improve biofuel potential via genetic engineering and hydrothermal liquefaction. ACS Sustain Chem Eng 8(7):2753–2762. https://doi.org/10.1021/acssuschemeng.9b06480 Douchi D, Liang F, Cano M et al (2019) Membrane-inlet mass spectrometry enables a quantitative understanding of inorganic carbon uptake flux and carbon concentrating mechanisms in metabolically engineered cyanobacteria. Front Microbiol 10:1356. https://doi.org/10.3389/fmicb. 2019.01356 Du W, Liang F, Duan Y et al (2013) Exploring the photosynthetic production capacity of sucrose by cyanobacteria. Metab Eng 19:17–25. https://doi.org/10.1016/j.ymben.2013.05.001 Du W, Jongbloets JA, Guillaume M et al (2019) Exploiting day- and night-time metabolism of Synechocystis sp. PCC 6803 for fitness-coupled fumarate production around the clock. ACS Synth Biol 8(10):2263–2269. https://doi.org/10.1021/acssynbio.9b00289 Ducat DC, Silver PA (2012) Improving carbon fixation pathways. Curr Opin Chem Biol 16 (3-4):337–344. https://doi.org/10.1016/j.cbpa.2012.05.002 Ducat DC, Avelar-Rivas JA, Way JC et al (2012) Rerouting carbon flux to enhance photosynthetic productivity. Appl Environ Microbiol 78(8):2660–2668. https://doi.org/10.1128/aem.07901-11 Durall C, Lindberg P, Yu J et al (2020) Increased ethylene production by overexpressing phosphoenolpyruvate carboxylase in the cyanobacterium Synechocystis PCC 6803. Biotechnol Biofuels 13:16. https://doi.org/10.1186/s13068-020-1653-y Durao P, Aigner H, Nagy P et al (2015) Opposing effects of folding and assembly chaperones on evolvability of Rubisco. Nat Chem Biol 11(2):148–155. https://doi.org/10.1038/nchembio.1715 Englund E, Liang F, Lindberg P (2016) Evaluation of promoters and ribosome binding sites for biotechnological applications in the unicellular cyanobacterium Synechocystis sp. PCC 6803. Sci Rep 6:36640. https://doi.org/10.1038/srep36640 Flores E, Arévalo S, Burnat M (2019) Cyanophycin and arginine metabolism in cyanobacteria. Algal Res 42:101577. https://doi.org/10.1016/j.algal.2019.101577 Gan F, Zhang S, Rockwell NC et al (2014) Extensive remodeling of a cyanobacterial photosynthetic apparatus in far-red light. Science 345(6202):1312–1317. https://doi.org/10.1126/science. 1256963 Gao Z, Zhao H, Li Z et al (2012) Photosynthetic production of ethanol from carbon dioxide in genetically engineered cyanobacteria. Energy Environ Sci 5(12):9857–9865. https://doi.org/10. 1039/C2EE22675H Gao X, Gao F, Liu D et al (2016a) Engineering the methylerythritol phosphate pathway in cyanobacteria for photosynthetic isoprene production from CO2. Energy Environ Sci 9 (4):1400–1411. https://doi.org/10.1039/C5EE03102H Gao X, Sun T, Pei G et al (2016b) Cyanobacterial chassis engineering for enhancing production of biofuels and chemicals. Appl Microbiol Biotechnol 100(8):3401–3413. https://doi.org/10.1007/ s00253-016-7374-2 Geerts D, Bovy A, de Vrieze G et al (1995) Inducible expression of heterologous genes targeted to a chromosomal platform in the cyanobacterium Synechococcus sp. PCC 7942. Microbiology 141 (Pt 4):831–841. https://doi.org/10.1099/13500872-141-4-831 Godlewska K, Michalak I, Pacyga P et al (2019) Potential applications of cyanobacteria: Spirulina platensis filtrates and homogenates in agriculture. World J Microbiol Biotechnol 35(6):80–80. https://doi.org/10.1007/s11274-019-2653-6 Gorbunov MY, Kuzminov FI, Fadeev VV et al (2011) A kinetic model of non-photochemical quenching in cyanobacteria. Biochim Biophys Acta 1807(12):1591–1599. https://doi.org/10. 1016/j.bbabio.2011.08.009 Gordon GC, Pfleger BF (2018) Regulatory tools for controlling gene expression in cyanobacteria. Adv Exp Med Biol 1080:281–315. https://doi.org/10.1007/978-981-13-0854-3_12 Gordon GC, Korosh TC, Cameron JC et al (2016) CRISPR interference as a titratable, trans-acting regulatory tool for metabolic engineering in the cyanobacterium Synechococcus sp. strain PCC 7002. Metab Eng 38:170–179. https://doi.org/10.1016/j.ymben.2016.07.007

288

B. Wang et al.

Grobbelaar N, Huang TC, Lin HY et al (1986) Dinitrogen-fixing endogenous rhythm in Synechococcus RF-1. FEMS Microbiol Lett 37(2):173–177 Guerrero F, Carbonell V, Cossu M et al (2012) Ethylene synthesis and regulated expression of recombinant protein in Synechocystis sp. PCC 6803. PLoS One 7(11):e50470. https://doi.org/ 10.1371/journal.pone.0050470 Hamilton TL, Bryant DA, Macalady JL (2016) The role of biology in planetary evolution: cyanobacterial primary production in low-oxygen Proterozoic oceans. Environ Microbiol 18 (2):325–340. https://doi.org/10.1111/1462-2920.13118 Hasunuma T, Matsuda M, Kondo A (2016) Improved sugar-free succinate production by Synechocystis sp. PCC 6803 following identification of the limiting steps in glycogen catabolism. Metab Eng Commun 3:130–141. https://doi.org/10.1016/j.meteno.2016.04.003 Hellweger FL, Jabbur ML, Johnson CH et al (2020) Circadian clock helps cyanobacteria manage energy in coastal and high latitude ocean. ISME J 14(2):560–568. https://doi.org/10.1038/ s41396-019-0547-0 Hendry JI, Bandyopadhyay A, Srinivasan S et al (2019) Metabolic model guided strain design of cyanobacteria. Curr Opin Biotechnol 64:17–23. https://doi.org/10.1016/j.copbio.2019.08.011 Higo A, Isu A, Fukaya Y et al (2016) Efficient gene induction and endogenous gene repression systems for the filamentous cyanobacterium Anabaena sp. PCC 7120. Plant Cell Physiol 57 (2):387–396. https://doi.org/10.1093/pcp/pcv202 Hirokawa Y, Maki Y, Tatsuke T et al (2016) Cyanobacterial production of 1,3-propanediol directly from carbon dioxide using a synthetic metabolic pathway. Metab Eng 34:97–103. https://doi. org/10.1016/j.ymben.2015.12.008 Hirokawa Y, Goto R, Umetani Y et al (2017a) Construction of a novel d-lactate producing pathway from dihydroxyacetone phosphate of the Calvin cycle in cyanobacterium, Synechococcus elongatus PCC 7942. J Biosci Bioeng 124(1):54–61. https://doi.org/10.1016/j.jbiosc.2017.02. 016 Hirokawa Y, Matsuo S, Hamada H et al (2017b) Metabolic engineering of Synechococcus elongatus PCC 7942 for improvement of 1,3-propanediol and glycerol production based on in silico simulation of metabolic flux distribution. Microb Cell Fact 16(1):212. https://doi.org/10. 1186/s12934-017-0824-4 Hirokawa Y, Kubo T, Soma Y et al (2020) Enhancement of acetyl-CoA flux for photosynthetic chemical production by pyruvate dehydrogenase complex overexpression in Synechococcus elongatus PCC 7942. Metab Eng 57:23–30. https://doi.org/10.1016/j.ymben.2019.07.012 Hu G, Zhou J, Chen X et al (2018) Engineering synergetic CO2-fixing pathways for malate production. Metab Eng 47:496–504. https://doi.org/10.1016/j.ymben.2018.05.007 Huang HH, Camsund D, Lindblad P et al (2010) Design and characterization of molecular tools for a Synthetic Biology approach towards developing cyanobacterial biotechnology. Nucleic Acids Res 38(8):2577–2593. https://doi.org/10.1093/nar/gkq164 Huo Y-X, Cho KM, Rivera JGL et al (2011) Conversion of proteins into biofuels by engineering nitrogen flux. Nat Biotechnol 29(4):346–351. https://doi.org/10.1038/nbt.1789 Immethun CM, DeLorenzo DM, Focht CM et al (2017) Physical, chemical, and metabolic state sensors expand the synthetic biology toolbox for Synechocystis sp. PCC 6803. Biotechnol Bioeng 114(7):1561–1569. https://doi.org/10.1002/bit.26275 Ito H, Mutsuda M, Murayama Y et al (2009) Cyanobacterial daily life with Kai-based circadian and diurnal genome-wide transcriptional control in Synechococcus elongatus. Proc Natl Acad Sci U S A 106(33):14168–14173. https://doi.org/10.1073/pnas.0902587106 Iwasaki H, Nishiwaki T, Kitayama Y et al (2002) KaiA-stimulated KaiC phosphorylation in circadian timing loops in cyanobacteria. Proc Natl Acad Sci U S A 99(24):15788–15793. https://doi.org/10.1073/pnas.222467299 Jacobsen JH, Frigaard NU (2014) Engineering of photosynthetic mannitol biosynthesis from CO2 in a cyanobacterium. Metab Eng 21:60–70. https://doi.org/10.1016/j.ymben.2013.11.004

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

289

Jaiswal D, Sengupta A, Sohoni S et al (2018) Genome features and biochemical characteristics of a robust, fast growing and naturally transformable cyanobacterium synechococcus elongatus PCC 11801 isolated from India. Sci Rep 8(1):16632. https://doi.org/10.1038/s41598-018-34872-z Jazmin LJ, Xu Y, Cheah YE et al (2017) Isotopically nonstationary 13C flux analysis of cyanobacterial isobutyraldehyde production. Metab Eng 42:9–18. https://doi.org/10.1016/j. ymben.2017.05.001 Jin H, Wang Y, Idoine A et al (2018) Construction of a shuttle vector using an endogenous plasmid from the cyanobacterium Synechocystis sp. PCC6803. Front Microbiol 9:1662. https://doi.org/ 10.3389/fmicb.2018.01662 Jin H, Lindblad P, Bhaya D (2019) Building an Inducible T7 RNA Polymerase/T7 Promoter Circuit in Synechocystis sp. PCC6803. ACS Synth Biol 8(4):655–660. https://doi.org/10.1021/ acssynbio.8b00515 Johnson CH, Zhao C, Xu Y et al (2017) Timing the day: what makes bacterial clocks tick? Nat Rev Microbiol 15(4):232–242. https://doi.org/10.1038/nrmicro.2016.196 Kamennaya NA, Ahn S, Park H et al (2015) Installing extra bicarbonate transporters in the cyanobacterium Synechocystis sp. PCC6803 enhances biomass production. Metab Eng 29:76–85. https://doi.org/10.1016/j.ymben.2015.03.002 Kamravamanesh D, Kiesenhofer D, Fluch S et al (2019) Scale-up challenges and requirement of technology-transfer for cyanobacterial poly (3-hydroxybutyrate) production in industrial scale. Int J Biobased Plastics 1(1):60–71. https://doi.org/10.1080/24759651.2019.1688604 Kanehisa M, Sato Y, Furumichi M et al (2019) New approach for understanding genome variations in KEGG. Nucleic Acids Res 47(D1):D590–D595. https://doi.org/10.1093/nar/gky962 Kaneko T, Sato S, Kotani H et al (1996) Sequence analysis of the genome of the unicellular cyanobacterium Synechocystis sp. strain PCC6803. II. Sequence determination of the entire genome and assignment of potential protein-coding regions. DNA Res 3(3):109–136. https:// doi.org/10.1093/dnares/3.3.109 Kaneko T, Nakamura Y, Wolk CP et al (2001) Complete genomic sequence of the filamentous nitrogen-fixing cyanobacterium Anabaena sp. strain PCC 7120. DNA Res 8(5):205–213.; 227–253. https://doi.org/10.1093/dnares/8.5.205 Kasting JF, Siefert JL (2002) Life and the Evolution of Earth’s Atmosphere. Science 296 (5570):1066–1068. https://doi.org/10.1126/science.1071184 Kelly CL, Taylor GM, Hitchcock A et al (2018) A Rhamnose-Inducible System for Precise and Temporal Control of Gene Expression in Cyanobacteria. ACS Synth Biol 7(4):1056–1066. https://doi.org/10.1021/acssynbio.7b00435 Kirsch F, Klahn S, Hagemann M (2019) Salt-Regulated Accumulation of the Compatible Solutes Sucrose and Glucosylglycerol in Cyanobacteria and Its Biotechnological Potential. Front Microbiol 10:2139. https://doi.org/10.3389/fmicb.2019.02139 Kirst H, Formighieri C, Melis A (2014) Maximizing photosynthetic efficiency and culture productivity in cyanobacteria upon minimizing the phycobilisome light-harvesting antenna size. Biochim Biophys Acta 1837(10):1653–1664. https://doi.org/10.1016/j.bbabio.2014.07.009 Klahn S, Hagemann M (2011) Compatible solute biosynthesis in cyanobacteria. Environ Microbiol 13(3):551–562. https://doi.org/10.1111/j.1462-2920.2010.02366.x Klodawska K, Bujas A, Turos-Cabal M et al (2019) Effect of growth temperature on biosynthesis and accumulation of carotenoids in cyanobacterium Anabaena sp. PCC 7120 under diazotrophic conditions. Microbiol Res 226:34–40. https://doi.org/10.1016/j.micres.2019.05.003 Knoot CJ, Ungerer J, Wangikar PP et al (2018) Cyanobacteria: Promising biocatalysts for sustainable chemical production. J Biol Chem 293(14):5044–5052. https://doi.org/10.1074/jbc.R117. 815886 Kolman MA, Nishi CN, Perez-Cenci M et al (2015) Sucrose in cyanobacteria: from a salt-response molecule to play a key role in nitrogen fixation. Life (Basel) 5(1):102–126. https://doi.org/10. 3390/life5010102

290

B. Wang et al.

Kondo T, Strayer CA, Kulkarni RD et al (1993) Circadian rhythms in prokaryotes: luciferase as a reporter of circadian gene expression in cyanobacteria. Proc Natl Acad Sci U S A 90 (12):5672–5676. https://doi.org/10.1073/pnas.90.12.5672 Korosh TC, Markley AL, Clark RL et al (2017) Engineering photosynthetic production of L-lysine. Metab Eng 44:273–283. https://doi.org/10.1016/j.ymben.2017.10.010 Ku JT, Lan EI (2018) A balanced ATP driving force module for enhancing photosynthetic biosynthesis of 3-hydroxybutyrate from CO2. Metab Eng 46:35–42. https://doi.org/10.1016/j. ymben.2018.02.004 Kuan D, Duff S, Posarac D et al (2015) Growth optimization of Synechococcus elongatus PCC7942 in lab flasks and a 2-D photobioreactor. Can J Chem Eng 93(4):640–647. https://doi.org/10. 1002/cjce.22154 Lai MC, Lan EI (2015) Advances in Metabolic Engineering of Cyanobacteria for Photosynthetic Biochemical Production. Metabolites 5(4):636–658. https://doi.org/10.3390/metabo5040636 Lan EI, Wei CT (2016) Metabolic engineering of cyanobacteria for the photosynthetic production of succinate. Metab Eng 38:483–493. https://doi.org/10.1016/j.ymben.2016.10.014 Lan EI, Ro SY, Liao JC (2013) Oxygen-tolerant coenzyme A-acylating aldehyde dehydrogenase facilitates efficient photosynthetic n-butanol biosynthesis in cyanobacteria. Energy Environ Sci 6(9):2672–2681. https://doi.org/10.1039/C3EE41405A Lan EI, Chuang DS, Shen CR et al (2015) Metabolic engineering of cyanobacteria for photosynthetic 3-hydroxypropionic acid production from CO2 using Synechococcus elongatus PCC 7942. Metab Eng 31:163–170. https://doi.org/10.1016/j.ymben.2015.08.002 Lee HJ, Son J, Sim SJ et al (2020) Metabolic rewiring of synthetic pyruvate dehydrogenase bypasses for acetone production in cyanobacteria. Plant Biotechnol J. https://doi.org/10.1111/ pbi.13342 Li H, Liao JC (2013) Engineering a cyanobacterium as the catalyst for the photosynthetic conversion of CO2 to 1,2-propanediol. Microb Cell Fact 12(1):4. https://doi.org/10.1186/1475-285912-4 Li C, Tao F, Ni J et al (2015) Enhancing the light-driven production of D-lactate by engineering cyanobacterium using a combinational strategy. Sci Rep 5:9777. https://doi.org/10.1038/ srep09777 Li H, Shen CR, Huang C-H et al (2016) CRISPR-Cas9 for the genome engineering of cyanobacteria and succinate production. Metab Eng 38:293–302. https://doi.org/10.1016/j.ymben.2016.09. 006 Liang F, Lindblad P (2016) Effects of overexpressing photosynthetic carbon flux control enzymes in the cyanobacterium Synechocystis PCC 6803. Metab Eng 38:56–64. https://doi.org/10.1016/ j.ymben.2016.06.005 Liang F, Englund E, Lindberg P et al (2018a) Engineered cyanobacteria with enhanced growth show increased ethanol production and higher biofuel to biomass ratio. Metab Eng 46:51–59. https://doi.org/10.1016/j.ymben.2018.02.006 Liang F, Lindberg P, Lindblad P (2018b) Engineering photoautotrophic carbon fixation for enhanced growth and productivity. Sustain Energy Fuels 2(12):2583–2600. https://doi.org/10. 1039/C8SE00281A Liberton M, Bandyopadhyay A, Pakrasi HB (2019) Enhanced Nitrogen Fixation in a glgXDeficient Strain of Cyanothece sp. Strain ATCC 51142, a Unicellular Nitrogen-Fixing Cyanobacterium. Appl Environ Microbiol 85(7):e02887-18. https://doi.org/10.1128/AEM.02887-18 Lin P-C, Pakrasi HB (2019) Engineering cyanobacteria for production of terpenoids. Planta 249 (1):145–154. https://doi.org/10.1007/s00425-018-3047-y Lin MT, Occhialini A, Andralojc PJ et al (2014) A faster Rubisco with potential to increase photosynthesis in crops. Nature 513(7519):547–550. https://doi.org/10.1038/nature13776 Lin PC, Zhang F, Pakrasi HB (2020) Enhanced production of sucrose in the fast-growing cyanobacterium Synechococcus elongatus UTEX 2973. Sci Rep 10(1):390. https://doi.org/10.1038/ s41598-019-57319-5

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

291

Lindberg P, Park S, Melis A (2010) Engineering a platform for photosynthetic isoprene production in cyanobacteria, using Synechocystis as the model organism. Metab Eng 12(1):70–79. https:// doi.org/10.1016/j.ymben.2009.10.001 Liu X, Curtiss R 3rd (2009) Nickel-inducible lysis system in Synechocystis sp. PCC 6803. Proc Natl Acad Sci U S A 106(51):21550–21554. https://doi.org/10.1073/pnas.0911953106 Liu D, Pakrasi HB (2018) Exploring native genetic elements as plug-in tools for synthetic biology in the cyanobacterium Synechocystis sp. PCC 6803. Microbial Cell Factories 17(1):48. https:// doi.org/10.1186/s12934-018-0897-8 Liu Y, Tsinoremas NF, Johnson CH et al (1995) Circadian orchestration of gene expression in cyanobacteria. Genes Dev 9(12):1469–1478. https://doi.org/10.1101/gad.9.12.1469 Liu X, Sheng J, Curtiss R 3rd (2011) Fatty acid production in genetically modified cyanobacteria. Proc Natl Acad Sci U S A 108(17):6899–6904. https://doi.org/10.1073/pnas.1103014108 Liu X, Miao R, Lindberg P et al (2019) Modular engineering for efficient photosynthetic biosynthesis of 1-butanol from CO2 in cyanobacteria. Energy Environ Sci 12(9):2765–2777. https:// doi.org/10.1039/C9EE01214A Luan G, Lu X (2018) Tailoring cyanobacterial cell factory for improved industrial properties. Biotechnol Adv 36(2):430–442. https://doi.org/10.1016/j.biotechadv.2018.01.005 Luan G, Zhang S, Lu X (2020) Engineering cyanobacteria chassis cells toward more efficient photosynthesis. Curr Opin Biotechnol 62:1–6. https://doi.org/10.1016/j.copbio.2019.07.004 Ludwig M, Bryant DA (2012) Synechococcus sp. strain PCC 7002 transcriptome: acclimation to temperature, salinity, oxidative stress, and mixotrophic growth conditions. Front Microbiol 3:354. https://doi.org/10.3389/fmicb.2012.00354 Ma AT, Schmidt CM, Golden JW (2014) Regulation of gene expression in diverse cyanobacterial species by using theophylline-responsive riboswitches. Appl Environ Microbiol 80 (21):6704–6713. https://doi.org/10.1128/aem.01697-14 Markham JN, Tao L, Davis R et al (2016) Techno-economic analysis of a conceptual biofuel production process from bioethylene produced by photosynthetic recombinant cyanobacteria. Green Chem 18(23):6266–6281. https://doi.org/10.1039/C6GC01083K Markley AL, Begemann MB, Clarke RE et al (2015) Synthetic biology toolbox for controlling gene expression in the cyanobacterium Synechococcus sp. strain PCC 7002. ACS Synth Biol 4 (5):595–603. https://doi.org/10.1021/sb500260k Melis A (2009) Solar energy conversion efficiencies in photosynthesis: minimizing the chlorophyll antennae to maximize efficiency. Plant Sci 177(4):272–280. https://doi.org/10.1016/j.plantsci. 2009.06.005 Misumi M, Katoh H, Tomo T et al (2016) Relationship between photochemical quenching and non-photochemical quenching in six species of cyanobacteria reveals species difference in redox state and species commonality in energy dissipation. Plant Cell Physiol 57 (7):1510–1517. https://doi.org/10.1093/pcp/pcv185 Mitschke J, Georg J, Scholz I et al (2011) An experimentally anchored map of transcriptional start sites in the model cyanobacterium Synechocystis sp. PCC6803. Proc Natl Acad Sci U S A 108 (5):2124–2129. https://doi.org/10.1073/pnas.1015154108 Moser S, Pichler H (2019) Identifying and engineering the ideal microbial terpenoid production host. Appl Microbiol Biotechnol 103(14):5501–5516. https://doi.org/10.1007/s00253-01909892-y Nakahira Y, Ogawa A, Asano H et al (2013) Theophylline-dependent riboswitch as a novel genetic tool for strict regulation of protein expression in cyanobacterium Synechococcus elongatus PCC 7942. Plant Cell Physiol 54(10):1724–1735. https://doi.org/10.1093/pcp/pct115 Nakajima T, Yoshikawa K, Toya Y et al (2017) Metabolic flux analysis of the Synechocystis sp. PCC 6803 DeltanrtABCD mutant reveals a mechanism for metabolic adaptation to nitrogenlimited conditions. Plant Cell Physiol 58(3):537–545. https://doi.org/10.1093/pcp/pcw233 Niederholtmeyer H, Wolfstadter BT, Savage DF et al (2010) Engineering cyanobacteria to synthesize and export hydrophilic products. Appl Environ Microbiol 76(11):3462–3466. https://doi. org/10.1128/AEM.00202-10

292

B. Wang et al.

Niu T-C, Lin G-M, Xie L-R et al (2019) Expanding the potential of CRISPR-Cpf1-based genome editing technology in the cyanobacterium Anabaena PCC 7120. ACS Synthetic Biol 8 (1):170–180. https://doi.org/10.1021/acssynbio.8b00437 Nudler E, Mironov AS (2004) The riboswitch control of bacterial metabolism. Trends Biochem Sci 29(1):11–17. https://doi.org/10.1016/j.tibs.2003.11.004 Nurnberg DJ, Morton J, Santabarbara S et al (2018) Photochemistry beyond the red limit in chlorophyll f-containing photosystems. Science 360(6394):1210–1213. https://doi.org/10. 1126/science.aar8313 Oliver JW, Machado IM, Yoneda H et al (2013) Cyanobacterial conversion of carbon dioxide to 2,3-butanediol. Proc Natl Acad Sci U S A 110(4):1249–1254. https://doi.org/10.1073/pnas. 1213024110 Ouyang Y, Andersson CR, Kondo T et al (1998) Resonating circadian clocks enhance fitness in cyanobacteria. Proc Natl Acad Sci U S A 95(15):8660–8664. https://doi.org/10.1073/pnas.95. 15.8660 Park J, Choi Y (2017) Cofactor engineering in cyanobacteria to overcome imbalance between NADPH and NADH: a mini review. Front Chem Sci Eng 11(1):66–71. https://doi.org/10.1007/ s11705-016-1591-1 Pattanayak GK, Phong C, Rust MJ (2014) Rhythms in energy storage control the ability of the cyanobacterial circadian clock to reset. Curr Biol 24(16):1934–1938. https://doi.org/10.1016/j. cub.2014.07.022 Pérez AA, Gajewski JP, Ferlez BH et al (2017) Zn2+-inducible expression platform for Synechococcus sp. strain PCC 7002 based on the smtA promoter/operator and smtB repressor. Appl Environ Microbiol 83(3):e02491–e02416. https://doi.org/10.1128/aem.02491-16 Peterhansel C, Krause K, Braun H-P et al (2013) Engineering photorespiration: current state and future possibilities. Plant Biol 15(4):754–758. https://doi.org/10.1111/j.1438-8677.2012. 00681.x Placzek S, Schomburg I, Chang A et al (2016) BRENDA in 2017: new perspectives and new tools in BRENDA. Nucleic Acids Res 45(D1):D380–D388. https://doi.org/10.1093/nar/gkw952 Qiao C, Duan Y, Zhang M et al (2018) Effects of reduced and enhanced glycogen pools on saltinduced sucrose production in a sucrose-secreting strain of Synechococcus elongatus PCC 7942. Appl Environ Microbiol 84(2):e02023–e02017. https://doi.org/10.1128/aem.02023-17 Ruffing AM (2014) Improved Free Fatty Acid Production in Cyanobacteria with Synechococcus sp. PCC 7002 as Host. Front Bioeng Biotechnol 2:17. https://doi.org/10.3389/fbioe.2014.00017 Ruffing AM, Jensen TJ, Strickland LM (2016) Genetic tools for advancement of Synechococcus sp. PCC 7002 as a cyanobacterial chassis. Microbial Cell Factories 15(1):190. https://doi.org/10. 1186/s12934-016-0584-6 Sakai M, Ogawa T, Matsuoka M et al (1997) Photosynthetic conversion of carbon dioxide to ethylene by the recombinant cyanobacterium, Synechococcus sp. PCC 7942, which harbors a gene for the ethylene-forming enzyme of Pseudomonas syringae. J Ferment Bioeng 84 (5):434–443. https://doi.org/10.1016/S0922-338X(97)82004-1 Sake CL, Metcalf AJ, Boyle NR (2019) The challenge and potential of photosynthesis: unique considerations for metabolic flux measurements in photosynthetic microorganisms. Biotechnol Lett 41(1):35–45. https://doi.org/10.1007/s10529-018-2622-4 Salis HM, Mirsky EA, Voigt CA (2009) Automated design of synthetic ribosome binding sites to control protein expression. Nat Biotechnol 27(10):946–950. https://doi.org/10.1038/nbt.1568 Santos-Merino M, Singh AK, Ducat DC (2019) New Applications of Synthetic Biology Tools for Cyanobacterial Metabolic Engineering. Front Bioeng Biotechnol 7:33. https://doi.org/10.3389/ fbioe.2019.00033 Saper G, Kallmann D, Conzuelo F et al (2018) Live cyanobacteria produce photocurrent and hydrogen using both the respiratory and photosynthetic systems. Nat Commun 9(1):2168. https://doi.org/10.1038/s41467-018-04613-x Satagopan S, Huening KA, Tabita FR (2019) Selection of Cyanobacterial (Synechococcus sp. Strain PCC 6301) RubisCO variants with improved functional properties that confer enhanced

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

293

CO2-dependent growth of Rhodobacter capsulatus, a Photosynthetic Bacterium. mBio 10(4): e01537-01519. https://doi.org/10.1128/mBio.01537-19 Savakis PE, Angermayr SA, Hellingwerf KJ (2013) Synthesis of 2,3-butanediol by Synechocystis sp. PCC6803 via heterologous expression of a catabolic pathway from lactic acid- and enterobacteria. Metab Eng 20:121–130. https://doi.org/10.1016/j.ymben.2013.09.008 Savakis P, Tan X, Du W et al (2015) Photosynthetic production of glycerol by a recombinant cyanobacterium. J Biotechnol 195:46–51. https://doi.org/10.1016/j.jbiotec.2014.12.015 Schwarz D, Orf I, Kopka J et al (2013) Recent applications of metabolomics toward cyanobacteria. Metabolites 3(1):72–100. https://doi.org/10.3390/metabo3010072 Sebesta J, Peebles CAM (2020) Improving heterologous protein expression in Synechocystis sp. PCC 6803 for alpha-bisabolene production. Metab Eng Commun 10:e00117. https://doi. org/10.1016/j.mec.2019.e00117 Sengupta A, Pakrasi HB, Wangikar PP (2018) Recent advances in synthetic biology of cyanobacteria. Appl Microbiol Biotechnol 102(13):5457–5471. https://doi.org/10.1007/ s00253-018-9046-x Sengupta A, Sunder AV, Sohoni SV et al (2019) Fine-tuning native promoters of Synechococcus elongatus PCC 7942 to develop a synthetic toolbox for heterologous protein expression. ACS Synth Biol 8(5):1219–1223. https://doi.org/10.1021/acssynbio.9b00066 Shen CR, Liao JC (2012) Photosynthetic production of 2-methyl-1-butanol from CO2 in cyanobacterium Synechococcus elongatus PCC7942 and characterization of the native acetohydroxyacid synthase. Energy Environ Sci 5(11):9574–9583. https://doi.org/10.1039/ C2EE23148D Song K, Tan X, Liang Y et al (2016) The potential of Synechococcus elongatus UTEX 2973 for sugar feedstock production. Appl Microbiol Biotechnol 100(18):7865–7875. https://doi.org/10. 1007/s00253-016-7510-z Stockel J, Welsh EA, Liberton M et al (2008) Global transcriptomic analysis of Cyanothece 51142 reveals robust diurnal oscillation of central metabolic processes. Proc Natl Acad Sci U S A 105 (16):6156–6161. https://doi.org/10.1073/pnas.0711068105 Sun T, Li S, Song X et al (2018) Toolboxes for cyanobacteria: Recent advances and future direction. Biotechnol Adv 36(4):1293–1307. https://doi.org/10.1016/j.biotechadv.2018.04.007 Swan JA, Golden SS, LiWang A et al (2018) Structure, function, and mechanism of the core circadian clock in cyanobacteria. J Biol Chem 293(14):5026–5034. https://doi.org/10.1074/jbc. TM117.001433 Takahama K, Matsuoka M, Nagahama K et al (2003) Construction and analysis of a recombinant cyanobacterium expressing a chromosomally inserted gene for an ethylene-forming enzyme at the psbAI locus. J Biosci Bioeng 95(3):302–305. https://doi.org/10.1016/S1389-1723(03) 80034-8 Tan X, Yao L, Gao Q et al (2011) Photosynthesis driven conversion of carbon dioxide to fatty alcohols and hydrocarbons in cyanobacteria. Metab Eng 13(2):169–176. https://doi.org/10. 1016/j.ymben.2011.01.001 Tang K-H, Feng X, Bandyopadhyay A et al (2013) Unique central carbon metabolic pathways and novel enzymes in phototrophic bacteria revealed by integrative genomics, 13C-based metabolomics and fluxomics. In: Kuang T, Lu C, Zhang L (eds) Photosynthesis research for food, fuel and the future. Springer, Berlin, Heidelberg, pp 339–343 Taton A, et al (2020) The circadian clock and darkness control natural competence in cyanobacteria. Nat Commun 11:1688. https://doi.org/10.1038/s41467-020-15384-9 Thiel K, Mulaku E, Dandapani H et al (2018) Translation efficiency of heterologous proteins is significantly affected by the genetic context of RBS sequences in engineered cyanobacterium Synechocystis sp. PCC 6803. Microbial Cell Factories 17(1):34. https://doi.org/10.1186/ s12934-018-0882-2 Toepel J, Welsh E, Summerfield TC et al (2008) Differential transcriptional analysis of the cyanobacterium Cyanothece sp. strain ATCC 51142 during light-dark and continuous-light growth. J Bacteriol 190(11):3904–3913. https://doi.org/10.1128/JB.00206-08

294

B. Wang et al.

Tschörtner J, Lai B, Krömer JO (2019) Biophotovoltaics: Green Power Generation From Sunlight and Water. Front Microbiol 10:866. https://doi.org/10.3389/fmicb.2019.00866 Ungerer J, Pakrasi HB (2016) Cpf1 is a versatile tool for CRISPR genome editing across diverse species of cyanobacteria. Sci Rep 6(1):39681. https://doi.org/10.1038/srep39681 Ungerer J, Tao L, Davis M et al (2012) Sustained photosynthetic conversion of CO2 to ethylene in recombinant cyanobacterium Synechocystis 6803. Energy Environ Sci 5(10):8998–9006. https://doi.org/10.1039/C2EE22555G Ungerer J, Lin PC, Chen HY et al (2018a) Adjustments to photosystem stoichiometry and electron transfer proteins are key to the remarkably fast growth of the cyanobacterium synechococcus elongatus UTEX 2973. mBio 9(1):e02327-17. https://doi.org/10.1128/mBio.02327-17 Ungerer J, Wendt KE, Hendry JI et al (2018b) Comparative genomics reveals the molecular determinants of rapid growth of the cyanobacterium Synechococcus elongatus UTEX 2973. Proc Natl Acad Sci U S A 115(50):E11761–E11770. https://doi.org/10.1073/pnas.1814912115 Ungerer J, Wendt KE, Hendry JI et al (2019) Reply to Zhou and Li: Plasticity of the genomic haplotype of Synechococcus elongatus leads to rapid strain adaptation under laboratory conditions. Proc Natl Acad Sci 116(10):3946–3947. https://doi.org/10.1073/pnas.1900792116 van Alphen P, Hellingwerf KJ (2015) Sustained circadian rhythms in continuous light in Synechocystis sp. pcc6803 growing in a well-controlled photobioreactor. PLoS One 10(6): e0127715. https://doi.org/10.1371/journal.pone.0127715 van Alphen P, Abedini Najafabadi H, Branco Dos Santos F et al (2018) Increasing the photoautotrophic growth rate of Synechocystis sp. PCC 6803 by identifying the limitations of its cultivation. Biotechnol J 13(8):e1700764. https://doi.org/10.1002/biot.201700764 van der Woude AD, Perez Gallego R, Vreugdenhil A et al (2016) Genetic engineering of Synechocystis PCC6803 for the photoautotrophic production of the sweetener erythritol. Microb Cell Fact 15:60. https://doi.org/10.1186/s12934-016-0458-y Varman AM, Yu Y, You L et al (2013) Photoautotrophic production of D-lactic acid in an engineered cyanobacterium. Microb Cell Fact 12:117. https://doi.org/10.1186/1475-2859-12117 Vijayan V, Zuzow R, O'Shea EK (2009) Oscillations in supercoiling drive circadian gene expression in cyanobacteria. Proc Natl Acad Sci U S A 106(52):22564–22568. https://doi.org/10. 1073/pnas.0912673106 Vu TT, Stolyar SM, Pinchuk GE et al (2012) Genome-scale modeling of light-driven reductant partitioning and carbon fluxes in diazotrophic unicellular cyanobacterium Cyanothece sp. ATCC 51142. PLoS Comput Biol 8(4):e1002460. https://doi.org/10.1371/journal.pcbi. 1002460 Wang B, Wang J, Zhang W et al (2012) Application of synthetic biology in cyanobacteria and algae. Front Microbiol 3:344. https://doi.org/10.3389/fmicb.2012.00344 Wang B, Pugh S, Nielsen DR et al (2013a) Engineering cyanobacteria for photosynthetic production of 3-hydroxybutyrate directly from CO2. Metab Eng 16:68–77. https://doi.org/10.1016/j. ymben.2013.01.001 Wang W, Liu X, Lu X (2013b) Engineering cyanobacteria to improve photosynthetic production of alka(e)nes. Biotechnol Biofuels 6(1):69. https://doi.org/10.1186/1754-6834-6-69 Wang Y, Tao F, Ni J et al (2015) Production of C3 platform chemicals from CO2 by genetically engineered cyanobacteria. Green Chem 17:3100–3110. https://doi.org/10.1039/C5GC00129C Wang Y, Sun T, Gao X et al (2016) Biosynthesis of platform chemical 3-hydroxypropionic acid (3-HP) directly from CO2 in cyanobacterium Synechocystis sp. PCC 6803. Metab Eng 34:60–70. https://doi.org/10.1016/j.ymben.2015.10.008 Wang B, Eckert C, Maness PC et al (2018a) A Genetic Toolbox for Modulating the Expression of Heterologous Genes in the Cyanobacterium Synechocystis sp. PCC 6803. ACS Synth Biol 7 (1):276–286. https://doi.org/10.1021/acssynbio.7b00297 Wang B, Xiong W, Yu J et al (2018b) Unlocking the photobiological conversion of CO2 to (R)-3hydroxybutyrate in cyanobacteria. Green Chem 20(16):3772–3782. https://doi.org/10.1039/ C8GC01208C

Reprogramming Metabolic Networks and Manipulating Circadian Clocks for. . .

295

Wang B, Dong T, Myrlie A et al (2019) Photosynthetic production of the nitrogen-rich compound guanidine. Green Chemistry 21(11):2928–2937. https://doi.org/10.1039/C9GC01003C Welsh EA, Liberton M, Stockel J et al (2008) The genome of Cyanothece 51142, a unicellular diazotrophic cyanobacterium important in the marine nitrogen cycle. Proc Natl Acad Sci U S A 105(39):15094–15099. https://doi.org/10.1073/pnas.0805418105 Wendt KE, Ungerer J, Cobb RE et al (2016) CRISPR/Cas9 mediated targeted mutagenesis of the fast growing cyanobacterium Synechococcus elongatus UTEX 2973. Microb Cell Fact 15 (1):115. https://doi.org/10.1186/s12934-016-0514-7 Weyer KM, Bush DR, Darzins A et al (2010) Theoretical maximum Algal oil production. BioEnergy Res 3(2):204–213. https://doi.org/10.1007/s12155-009-9046-x Wiegard A, Dorrich AK, Deinzer HT et al (2013) Biochemical analysis of three putative KaiC clock proteins from Synechocystis sp. PCC 6803 suggests their functional divergence. Microbiology 159(Pt 5):948–958. https://doi.org/10.1099/mic.0.065425-0 Wilson RH, Whitney SM (2017) Improving CO2 fixation by Enhancing Rubisco performance. In: Alcalde M (ed) Directed enzyme evolution: advances and applications. Springer International Publishing, Cham, pp 101–126. https://doi.org/10.1007/978-3-319-50413-1_4 Włodarczyk A, Selão TT, Norling B et al (2020) Newly discovered Synechococcus sp. PCC 11901 is a robust cyanobacterial strain for high biomass production. Commun Biol 3(1):215. https:// doi.org/10.1038/s42003-020-0910-8 Woelfle MA, Ouyang Y, Phanvijhitsiri K et al (2004) The adaptive value of circadian clocks: an experimental assessment in cyanobacteria. Curr Biol 14(16):1481–1486. https://doi.org/10. 1016/j.cub.2004.08.023 Xiong W, Morgan JA, Ungerer J et al (2015) The plasticity of cyanobacterial metabolism supports direct CO2 conversion to ethylene. Nature Plants 1(5):15053. https://doi.org/10.1038/nplants. 2015.53 Xiong W, Cano M, Wang B et al (2017) The plasticity of cyanobacterial carbon metabolism. Curr Opin Chem Biol 41:12–19. https://doi.org/10.1016/j.cbpa.2017.09.004 Xu Y, Mori T, Johnson CH (2003) Cyanobacterial circadian clockwork: roles of KaiA, KaiB and the kaiBC promoter in regulating KaiC. EMBO J 22(9):2117–2126. https://doi.org/10.1093/ emboj/cdg168 Xu Y, Alvey RM, Byrne PO et al (2011) Expression of genes in cyanobacteria: adaptation of endogenous plasmids as platforms for high-level gene expression in Synechococcus sp. PCC 7002. Methods Mol Biol 684:273–293. https://doi.org/10.1007/978-1-60761-925-3_21 Xu Y, Weyman PD, Umetani M et al (2013) Circadian yin-yang regulation and its manipulation to globally reprogram gene expression. Current Biol: CB 23(23):2365–2374. https://doi.org/10. 1016/j.cub.2013.10.011 Xue Y, Zhang Y, Cheng D et al (2014) Genetically engineering Synechocystis sp. Pasteur Culture Collection 6803 for the sustainable production of the plant secondary metabolite p-coumaric acid. Proc Natl Acad Sci U S A 111(26):9449–9454. https://doi.org/10.1073/pnas.1323725111 You L, Berla B, He L et al (2014) 13C-MFA delineates the photomixotrophic metabolism of Synechocystis sp. PCC 6803 under light- and carbon-sufficient conditions. Biotechnol J 9 (5):684–692. https://doi.org/10.1002/biot.201300477 Young JD, Shastri AA, Stephanopoulos G et al (2011) Mapping photoautotrophic metabolism with isotopically nonstationary (13)C flux analysis. Metab Eng 13(6):656–665. https://doi.org/10. 1016/j.ymben.2011.08.002 Yu King Hing N, Liang F, Lindblad P et al (2019) Combining isotopically non-stationary metabolic flux analysis with proteomics to unravel the regulation of the Calvin-Benson-Bassham cycle in Synechocystis sp. PCC 6803. Metab Eng 56:77–84. https://doi.org/10.1016/j.ymben.2019.08. 014 Yu J, Liberton M, Cliften PF et al (2015) Synechococcus elongatus UTEX 2973, a fast growing cyanobacterial chassis for biosynthesis using light and CO(2). Sci Rep 5:8132. https://doi.org/ 10.1038/srep08132

296

B. Wang et al.

Yu H, Li X, Duchoud F et al (2018) Augmenting the Calvin–Benson–Bassham cycle by a synthetic malyl-CoA-glycerate carbon fixation pathway. Nature Communications 9(1):2008. https://doi. org/10.1038/s41467-018-04417-z Yunus IS, Jones PR (2018) Photosynthesis-dependent biosynthesis of medium chain-length fatty acids and alcohols. Metab Eng 49:59–68. https://doi.org/10.1016/j.ymben.2018.07.015 Zarzycki J, Axen SD, Kinney JN et al (2012) Cyanobacterial-based approaches to improving photosynthesis in plants. J Exp Bot 64(3):787–798. https://doi.org/10.1093/jxb/ers294 Zavřel T, Sinetova MA, Búzová D et al (2015) Characterization of a model cyanobacterium Synechocystis sp. PCC 6803 autotrophic growth in a flat-panel photobioreactor. Eng Life Sci 15(1):122–132. https://doi.org/10.1002/elsc.201300165 Zehr JP (2011) Nitrogen fixation by marine cyanobacteria. Trends Microbiol 19(4):162–173. https://doi.org/10.1016/j.tim.2010.12.004 Zess EK, Begemann MB, Pfleger BF (2016) Construction of new synthetic biology tools for the control of gene expression in the cyanobacterium Synechococcus sp. strain PCC 7002. Biotechnol Bioeng 113(2):424–432. https://doi.org/10.1002/bit.25713 Zhang H, Liu Y, Nie X et al (2018) The cyanobacterial ornithine-ammonia cycle involves an arginine dihydrolase. Nat Chem Biol 14(6):575–581. https://doi.org/10.1038/s41589-018-0038z Zhou J, Zhang H, Zhang Y et al (2012) Designing and creating a modularized synthetic pathway in cyanobacterium Synechocystis enables production of acetone from carbon dioxide. Metab Eng 14(4):394–400. https://doi.org/10.1016/j.ymben.2012.03.005 Zhou J, Zhang H, Meng H et al (2014) Discovery of a super-strong promoter enables efficient production of heterologous proteins in cyanobacteria. Sci Rep 4:4500. https://doi.org/10.1038/ srep04500 Zhou J, Zhang F, Meng H et al (2016a) Introducing extra NADPH consumption ability significantly increases the photosynthetic efficiency and biomass production of cyanobacteria. Metab Eng 38:217–227. https://doi.org/10.1016/j.ymben.2016.08.002 Zhou J, Zhu T, Cai Z et al (2016b) From cyanochemicals to cyanofactories: a review and perspective. Microbial Cell Factories 15(1):2. https://doi.org/10.1186/s12934-015-0405-3 Zhu T, Xie X, Li Z et al (2015) Enhancing photosynthetic production of ethylene in genetically engineered Synechocystis sp. PCC 6803. Green Chem 17(1):421–434. https://doi.org/10.1039/ C4GC01730G

Insights from Mathematical Modeling/ Simulations of the In Vitro KaiABC Clock Mark Byrne

Abstract This chapter discusses mathematical modeling of the in vitro KaiABC clock, focusing on various potential insights into mechanisms of synchrony and attempts by different groups to reproduce various experimental properties of observed clock dynamics. It is separated into a discussion of modeling the dynamics of population phosphoform state transitions to generate limit cycles, and in simulating the dynamics of KaiC hexamers and protein-protein interactions, including allosteric transitions. The latter models include both phenomenological simplified state transition models between various hexameric complexes and protein-protein interactions, and direct numerical simulations of biochemical reactions which incorporate thousands of combinatorial states and their transitions, including the effects of stochasticity (noise) on the simulated dynamics.

1 Introduction In this chapter, the aim is to provide a relatively concise update of mathematical modeling and apparent insights into the functional aspects of the in vitro cyanobacterial KaiABC clock since an initial modeling overview was published in a similar venue in 2009 (Byrne 2009). For the reader interested in delving into the details of various mathematical models in the literature (usually described in detail in the supplemental methods of various papers), relatively recent published mathematical models of the KaiABC oscillator, both constrained by experiments and reproducing a variety of experimental results on the system, have been published (see, for example, Lin et al. 2014; Chang et al. 2015; Paijmans et al. 2017; Mori et al. 2018; Das et al. 2018; Chew et al. 2018). In particular, the recent modeling work of Paijmans et al. (2017) comprehensively includes CI and CII domain-specific ATP-ADP dynamics coupled to phosphoform transitions in a thermodynamically consistent manner and accounts for hexameric heterogeneity and allosteric state

M. Byrne (*) Department of Chemistry, Physics and Engineering, Spring Hill College, Mobile, AL, USA e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_15

297

298

M. Byrne

Fig. 1 A schematic allosteric model for the KaiABC oscillator (courtesy of T. Mori, adapted from Mori et al. 2018). Rapid KaiA-KaiC interactions phosphorylate hexamers, while phosphoform differential affinity results in lower KaiA association at the end of the phosphorylation phase, promoting the D ! S ( pS/pT ! pS/T ) transition. KaiC hexamers switch to a KaiB-binding competent state (rate limited by CI ATP-ase activity) based on hexameric pS/T and pS/pT distributions in the CII domain, and KaiB-KaiC complexes sequester KaiA, autocatalytically generating population dephosphorylation

switching in a hexamer population. A recent model I worked on (Mori et al. 2018) examined the potential role of phosphoform-dependent differential affinity (PDDA), with reduced KaiA association as a hexamer is phosphorylated, constrained by partial reaction data for phosphoform transitions, and included an examination of allosteric hexamer state switching and KaiB-KaiC interactions on oscillator robustness (e.g., see Figs. 1, 2 and 5 below, adapted from the supplementary modeling section of Mori et al. 2018). To keep this chapter self-contained, only a brief recapitulation of the structure and aims of math modeling/simulation applied to this system is included. I recognize there may be a tendency among experimentalists to be (rightfully?) suspicious of, and potentially minimize, the relevance of mathematical modeling to investigating biological systems! However, presented with interesting, non-trivial dynamics for a biochemical subsystem, cartoon descriptions, while certainly useful (see Fig. 1), are certainly not convincing explanations of a mechanism to a physicist and a mathematical model with quantitative predictions is preferable to schematic mechanisms or verbal descriptions of physical processes (even if the model has limited quantitative experimental data based on measured reaction rates, etc.) Mathematical models of the KaiABC system span from more phenomenological (with a lowerdimensional state space, lower-dimensional parameter space, and use of effective parameters) to complex and intricate, accounting for part of the state space of

Insights from Mathematical Modeling/Simulations of the In Vitro KaiABC Clock

299

Fig. 2 (a) Stochastic and deterministic simulations of phospho-state transitions in a hexamer population for the phosphorylation phase (KaiA+KaiC alone, standard concentrations) and dephosphorylation phase initiated at 20 h (KaiC alone) with linear interconversion rates of phosphoforms Eq. (1a, 1b, 1c and 1d) in each phase taken from Rust et al. 2007. An estimate of hexamer population synchrony using a Euclidean metric and phosphoforms is shown in orange. (b) Stochastic simulations of oscillating reactions of KaiC phosphoforms from a hybrid model with phosphoform differential affinity (PDDA) using a phosphoform-dependent allosteric switch for KaiB binding with heterogeneous KaiC hexamers, and subsequent KaiA sequestration in complexes (Figures adapted from Mori et al. 2018). The purple trace shows the initial KaiA dimer to KaiC hexamer ratio

combinatorial permutations of molecular interactions that occur, and including various constraints from experiments probing different aspects of the system’s dynamics (partial reactions, stoichiometry, ATP/ADP variation, temperature compensation, phase response characteristics, etc.). As a potential motivation for model construction, it should probably be noted that even if there were a first principles, non-relativistic quantum simulation of the molecular interactions of these proteins (with ATP&ADP, including water and buffers, etc.), this method of investigation would only provide structural and thermodynamic insights on conformational changes on approximately nanosecond timescales, which is ~1014 of a circadian period (24 h ~ 105 s). A central role of the mathematical models and simulations is to phenomenologically bridge this large hierarchy (~14 orders of magnitude temporal gap) between the natural rapid molecular dynamics timescales and the circadian timescale, hopefully providing a rational basis for various proposed oscillatory mechanisms that satisfy the defining characteristics of a circadian oscillator (capable of self-sustained oscillations without external input, entrainable by external cues, yet compensated for temperature or other external perturbations). Furthermore it may be worth noting that any proposed mechanism for the functioning and robustness of a circadian clock mechanism may be non-oscillatory and missing key feedback properties unless a formal mathematical model has been constructed that can reproduce many of the observed properties of the system. It may be amusing to note this general “hierarchy problem” in biological systems is very loosely analogous to the “hierarchy” problem in particle physics where the electroweak and Planck scale are separated by a similar vast chasm (and the mechanisms for stabilizing the weak scale and the cosmological constant scale are not currently known). Fortunately the study

300

M. Byrne

of mathematical models of oscillators and circadian oscillators is well-established, and the general principles are known for creating general, biochemical oscillations (limit cycle and damped oscillators, etc.) using standard chemical kinetics and nonlinear feedback (positive and/or negative) of one “state variable” ¼ molecular component or complex, on another (e.g., Novák and Tyson 2008). For example, negative feedback repression in the standard transcription-translation feedback loop (TTFL), with time delays and nonlinearity, naturally yields limit-cycle oscillatory behavior that can reproduce many of the properties of mammalian circadian oscillators (e.g., Goldbeter 1995; Leloup and Goldbeter 2003; Forger and Peskin 2003). Since the KaiABC clock is the only known circadian clock that can be reconstituted from simple components outside cells (Tomita et al. 2005; Nakajima et al. 2005), there have been a plethora of mathematical modeling papers (and mixed experiment/theory papers) quantitatively exploring various aspects of this system. A non-exhaustive list of KaiABC mathematical modeling publications includes Emberly and Wingreen 2006; Mehra et al. 2006; Van Zon et al. 2007; Clodong et al. 2007; Mori et al. 2007; Miyoshi et al. 2007; Rust et al. 2007; Yoda et al. 2007; Eguchi et al. 2008; Brettschneider et al. 2010; Zwicker et al. 2010; Qin et al. 2010b; Rust et al. 2011; Ma and Ranganathan 2012; Lin et al. 2014; Paijmans et al. 2016a, b; Paijmans et al. 2017; Mori et al. 2018; and Das et al. 2018. Notably, within 2 years after the original experimental discovery papers of the KaiABC in-vitro oscillator, the population phosphoform cycle was elucidated (Nishiwaki et al. 2007; Rust et al. 2007) and a relatively simple and experimentally constrained mathematical model of the clock dynamics for the phosphoforms was published in (Rust et al. 2007). The data from both groups and mathematical modeling directly supported phase-dependent KaiA sequestration as necessary for sustained oscillations. Several authors provided data and various models further supporting negative feedback repression of KaiC kinase activity (by sequestration of KaiC’s autokinase effector, KaiA, in concert with KaiB binding) as being central to generating limit cycle oscillations in KaiC phosphorylation (Van Zon et al. 2007; Clodong et al. 2007; Brettschneider et al. 2010). In tandem with the kinase activity in the CII domain, a rate-limiting ATP-ase activity of KaiC in the CI domain appears to regulate the characteristic circadian timescale setting the clock period (Murakami et al. 2008; Terauchi et al. 2007). Thus, the average time-dependent KaiC kinase reaction, in a population of KaiC molecules, appears to be regulated (via proteinprotein interactions and ATP/ADP activity) via a nonlinear, autocatalytic negative feedback loop among the participating proteins and small molecules. It would be surprising if other auto-catalytic feedback oscillators were not generally present post-translationally in different biochemical systems that mediate various cellular functions.

Insights from Mathematical Modeling/Simulations of the In Vitro KaiABC Clock

301

2 Insights from Simplified Phosphoform Dynamic Models In this section, classes of oscillatory models based on varying degrees of simplification of the complex molecular dynamics that actually occurs during the in vitro reaction of the three proteins, ATP, ADP, and buffer molecules are described. The goal of such models is to verify that putative proposed mechanisms are consistent with oscillatory dynamics seen in experiments to reasonably reproduce various experimental results and to predict the effect of various perturbations on the system. Since each KaiC monomer may be unphosphorylated (U ), phosphorylated on S431 (S) or T432 (T), or both (D), one approach to modeling considers only the dynamics of a population of monomeric KaiC states U, T, S, D without considering directly the hexameric nature of KaiC and potential allosteric effects that operate in heterogeneous hexamers. In this approach one simply writes down reasonably motivated differential equations for the putative state transitions, in which the transition rates are some effective functions of KaiA and KaiB (and ADP and ATP, etc., as desired). At the population level, the state transitions follow the approximate sequential (cyclic) sequence U ! T ! D ! S ! U (Nishiwaki et al. 2007; Rust et al. 2007); more realistically, accounting for stochasticity, both forward and reverse reactions occur for the transitions with inter and intra-hexameric variation (e.g., see Fig. 2). The phosphoform ODE method tracks only the population average of each of the phosphoform abundances such that the models are attempting to describe the dynamics assuming a well-mixed monomer pool at each timestep (so these models may already implicitly include the exchange of monomers among hexamers, Kageyama et al. 2006). This phenomenological approach was successfully carried out by Rust et al. (2007) using a constrained model with partial reaction data and by measuring the effective rate of the kinase reaction, U ! T, as a function of KaiA. Thus the transition rates are assumed to be nonlinear functions of KaiA, with free KaiA proteins globally regulating the switch (mediated by KaiB) between population phosphorylation and dephosphorylation phases. Writing differential equations for the state transitions and denoting the phosphoform transitions rates as kij, indicating a transition from state i to state j: dU=dt ¼ kUT U  k US U þ k TU T þ kSU S

ð1aÞ

dT=dt ¼ k UT U þ kDT D  k TU T  kTD T

ð1bÞ

dD=dt ¼ k TD T þ kSD S  kDT D  k DS D

ð1cÞ

dS=dt ¼ kUS U þ kDS D  kSU S  kSD S

ð1dÞ

where the state transitions U $ D and S $ T are not included in the above, and the system can be reduced to three differential equations since the sum of derivatives is zero, or U + T + D + S is a constant as determined by the initial conditions. Generally the effective transition rates between phosphoforms could be complicated functions of KaiA and KaiB (and ADP/ATP). Fortunately a fit to the approximate U ! T transition suggested hyperbolic dependence on KaiA in KaiA-KaiC partial reactions,

302

M. Byrne

and KaiB appears to have only a small effect on KaiC phosphoform transitions alone (KaiB-KaiC partial reactions) so the transition rates could be approximated as being solely KaiA dependent: ki j ¼ ki j 0 þ kij A f ð½AÞ

ð2Þ

where ki j0 indicates the basal rate (KaiC-only) and kijA indicates the effect of KaiA on each relevant phosphoform transition rate. The formulation of the state transitions above Eq. (2) allows for both positive and negative regulation of transition rates (by KaiA). By modifying only the kinase rates by including a simple ATP/ADPdependent regulatory term, kijA ! kijA [ATP]/([ATP] + Kr[ADP]) with relative affinity Kr ~ 1, Rust et al. (2011) showed that this can nicely reproduce typical experimental phase resetting characteristics (entrainment) of the in vitro clock that correlates with in vivo dark pulses. The form of Eq. (2) unfortunately introduces a large parameter space of rate constants that is generally under-determined by data fitting, even with various partial reaction measurements; it is also not clear that the effective phospho-transition rates should be regulated by KaiA in the same hyperbolic manner (and for state transitions that are unaffected by global free KaiA, kijA ¼ 0, for some i and j). Notably, the U ! T data suggested a hyperbolic regulatory function for modification of the phosphorylation rate and a half-maximal effective concentration of K1/2 ~ 0.43μM: f ð½AÞ ¼ ½A= ½A þ K 1=2




ð3Þ

Since (free) KaiA is time-dependent, its sequestration can drive rapid loss of the T phosphoform population and promotes the D ! S transition. A simple and effective assumption was that KaiA is phenomenologically sequestered correlated with the fraction of S states (pS/T protomers) instantaneously present (and experimentally, the S state abundance is correlated with the extent of KaiB binding to KaiC; in later published extended models, alternative state variables were considered as the sequestering states). A simple sequestration model for KaiA is. ½A ¼ ½A0   nS

ðnon‐negativeÞ

ð4Þ

with n representing the relative number of S phospho-states (protomers) that sequester one KaiA dimer; it would likely be more realistic to write a differential equation for KaiA sequestration as a function of the effective S phosphoform population to explore the rapidity and non-linearity of this important process, schematically kA + S ! (AkS) ! (AkU ) ! kA + U, where (AkS) represents a sequestering state with multiple (k) KaiA that are subsequently released with the transition to the unphosphorylated state. In any case, this proposed negative feedback Eq. (4) is the essential non-linearity proposed in these models that permit phosphoform oscillations and suggests KaiA sequestration as the primary mechanism for oscillations (the model structure in Eq. (2) permits both positive and negative feedback by KaiA on

Insights from Mathematical Modeling/Simulations of the In Vitro KaiABC Clock

303

phosphoform transitions). Essentially, formation of the S states auto-catalytically depopulates the T states, driving the system to a “dephosphorylating” phase. These phosphoform models were subsequently modified and expanded to include ATP/ADP regulation of the phospho-transition rates as described above (Rust et al. 2011), a kinase-independent approximately constant CI-domain ATP-ase rate (Phong et al. 2013) and incorporating a slow KaiB fold-switch mechanism for KaiBKaiC complex formation (Chang et al. 2015); the basic model structure means it is convenient to simultaneously retain the rate constants and transitions that fit experimental partial reaction data and expand the phosphoform state space to include new relevant “effective” states associated with temporal delays from ATP-ase activity (e.g., S ! S* with S* representing a modified sequestration state), or select KaiBassociated phosphoform pools (e.g., including S ! BS*, D ! BD* states), or for incorporating different allosteric forms of KaiB implicated in KaiA sequestration. The proposed central mechanism (selective sequestration) can be investigated in even simpler contexts for the creation of limit cycles in a manner that may be useful for investigating other biochemical oscillators. For the 4-state system it is easily verified by numerical integration that it is possible to reproduce circadian oscillations by simulating the (overly) simplified “cyclic” sequence directly (U ! T ! D ! S ! U ) incorporating Eq. (3 and 4) including the KaiA regulation of only the state transition U!T, using constant transition rates for the other rates (without special fine-tuning of parameters): dU=dt ¼ k UT U þ kSU S

ð5aÞ

dT=dt ¼ k UT U  kTD T

ð5bÞ

dD=dt ¼ kTD T  k DS D

ð5cÞ

dS=dt ¼ k DS D  k SU S

ð5dÞ

For example, approximately circadian limit cycles are generated using kUT ¼ f ([A]) 1.0 h1, [A]0 /[C0] ¼ 0.5 (dimer/hexamer), K1/2/[C0] ¼ 0.5, n ¼ 3, kSU ¼ kTD ¼ kDS ¼ 0.15 h1. Of course, there are a number of experimental constraints from the actual dynamics that overly simplified mathematical models are not able to reproduce for any of their parameter choices (these oversimplified models are generally over-constrained by data). Nonetheless, simplified limit cycle “toy models” such as these may be useful for researchers exploring variations in model designs attempting to reproduce qualitative trends in experimental data such as stoichiometric variation of proteins, temperature, and metabolic perturbations (pulsatile or systemic), and for examining the corresponding phase response characteristics of a system. For example, by having transition rates that depend solely on ratios of protein concentrations, robustness of oscillations to overall concentration scaling (Kageyama et al. 2006) can be maintained essentially by construction. Along these lines, the effect of scaling all concentrations ([A], U, T, D, S) to the initial (hexameric) KaiC concentration ([C0]) indicates that the system dynamics in state

304

M. Byrne

transition models in Eq. (1a, 1b, 1c and 1d) varies solely from concentration changes by changing the dimensionless regulatory constant, K1/2/[C0]. The effective fitted rates for the phosphoform transitions with a characteristic circadian timescale (~h1) are clearly integrating an intricate molecular dynamics taking place between the proteins (KaiA differential affinity), nucleotide exchange, and ATP hydrolysis, and protein allosteric transitions. In the path to yet further simplification of the core sequestration mechanism, simple classes of just two coupled differential equations can yield limit cycle oscillations for the fractional site occupancy of regulatory sites (xj) in the protein population (2 regulatory sites/4 protein states) in the continuum, well-mixed approximation (see Figs. 3 and 4 adapted from Byrne 2020, for example). For example, it was surprising that the following systems of simple coupled ODEs generate limit cycles with selective sequestration for reasonable parameter ranges: dx1 =dt ¼ f ð½AÞ  k1 ð1  x1 Þ  k 1 x1 dx2 =dt ¼ k 1 x1  k 2 x2

ð6Þ

or dx1 =dt

¼ f ð½AÞ  k1 ð1  x1 Þ  k 1 x1

dx2 =dt ¼ f ð½AÞ  k2 ð1  x2 Þ  k 2 x2

ð7Þ

Equation (6) is supposed to represent a very simplified schematic version of a transferase mechanism shifting occupancy from one residue to another, while Eq. (7) is more similar to simplified KaiABC models with similar regulatory functions of the regulatory rates on site occupancy. In the above the rates k+/ are fixed constants (for a given protein) and f is a regulatory function of the effector protein (A). If A is sequestered by one of the four protein states (Cs), [A] ¼ [A0]n[Xs] where [Xs] is the sequestering state. The fraction of proteins in the sequestering state can be approximately computed from x1 and x2 using the product of probabilities of occupancy of each site; a convenient choice is the fraction of the protein population with site 1 unoccupied and site 2 occupied (analogous to the S state for KaiC): [Xs] ¼ (1x1) x2. With sequestration, various regulatory functions of occupancy (linear, MichealisMenten type, Hill type, etc.) can generate limit cycles. Thus f ¼ [A]/([A] + K), as the data for KaiA suggests, generates limit cycles, and surprisingly, even linear regulatory functions ( f ¼ [A]) can generate limit cycles! This simple abstraction of the sequestration proposal has the advantage that the system can be non-dimensionalized and the parameter space for oscillations completely explored (see, for example, Fig. 3d). By scanning the parameter space and numerically integrating, a fast-slow separation for regulation of the two sites is generally required for sustained oscillatory dynamics in these simple models (in the KaiC system, the regulation of T432 appears to occur approximately five times faster than S431). So here we have an example in which mathematical modeling from the KaiABC system has suggested a general mechanism for limit cycles, where one of the protein states

Insights from Mathematical Modeling/Simulations of the In Vitro KaiABC Clock

305

Fig. 3 Abstracted sequestration mechanism for a “clock” protein with two modification sites that sequesters an effector protein regulating site occupancy (adapted from Byrne 2020). (a) Some example designs that generate limit cycle oscillations in site occupancy (example differential equations for regulation of protein site occupancy are in the text, Eqs. 6 and 7) (b and c). Sample oscillations in % site occupancy for the two regulatory sites and four protein states. (d) Oscillatory parameter space and dimensionless period (colormap) from the simple sequestration model of Eq. (6)

selectively sequesters an effector protein; the general mechanism is broadly generalizable to a protein with multiple regulatory sites; an interesting question is whether this type of mechanism is employed in other biochemical systems? Furthermore in the simple models above Eqs. (6 and 7) the system’s period is primarily set by a slow transition rate (or rates) in a power-law manner, and these slow rates can be biochemically buffered from the external “fast” rate so that both entrainment and temperature compensation are possible, even in the very minimal and simplified models given above. The KaiA sequestration feedback mechanism was previously suggested as necessary for sustained oscillations and employed in several earlier mathematical models of the clock (Van Zon et al. 2007; Clodong et al. 2007; Brettschneider et al. 2010); general phosphatase and kinase designs for circadian oscillations with selective sequestration by a four-state protein were explored systematically in detail using mass-action models in the comprehensive oscillatory design study by Jolley et al. (2012). In the KaiABC system, some surprising findings include the near temperature independence of the CI ATP-ase activity (Terauchi et al. 2007; Abe et al. 2015) and the near period-independence of the oscillator for a sustained reduction in the [ADP]/

306

M. Byrne

Fig. 4 Panels a–c. Even very simple selective sequestration models can be entrained and temperature compensated. (a and b) Simulated approximate period invariance as a “fast” modification rate (k+1) on one residue is varied (such as by ADP/ATP ratio) while a “slow” internal rate is essentially buffered from external perturbations (panel B shows sample oscillatory traces as k+1 is varied). (c) Simulated entrainment of the simple sequestration design oscillator (unperturbed period of 21.77 h.) to external continuous low-amplitude sinusoidal driving of the modification rate of the site 1 residue (external driving rhythms of 19 and 24 h. on the unperturbed oscillator are shown). (d) Sample vector field and nullclines for Eq. (6) with linear regulation by free [A] using the rate parameters from Fig. 3b. Details of simulations and sample phase response curves for pulsatile (“step down”) perturbations are in Byrne (2020)

[ATP] ratio (and a corresponding decrease in the kinase rates) (Phong et al. 2013). These experimental results can be (and have been) interpreted and simulated in these phosphoform state transition models. Returning to a four-state system for the KaiC phosphoforms, the dependence of the transition rates between population phosphoform abundances are generally some complicated functions of KaiA, KaiB, and ATP and ADP, and can be encoded in the (generally nonconstant) transition rates (rate km,j for the state transition, m ! j) between states which can be formalized (X1 ¼ U, X2 ¼ T, X3 ¼ D, X4 ¼ S): dX j =dt ¼ Σ m km,j X m  Σ n k


j,n

Xj

ð8Þ

where the state transition rates are generally (unknown) functions of initially exogenous nucleotides, protein concentrations, etc.:

Insights from Mathematical Modeling/Simulations of the In Vitro KaiABC Clock

k

j,m

¼k

j,m ð½ATP, ADP, ½A, ½B, ½C ,

307

. . .Þ

ð9Þ

Equation (1a, 1b, 1c and 1d) is a simplified version of (6). Experimentally motivated differential equations for the proteins and nucleotides then feedback into the state transition rates (see Paijmans et al. 2017, for an example). Models of this type can be further generalized to account for delays between state transitions due to presumed rate-limiting steps, such as ATPase activity in the CI domain of KaiC which may be necessary for stable KaiB-KaiC complex formation (Phong et al. 2013). This may involve further enlarging the state space to include additional transitions (such as schematically, S!S* and S*!U, where S* might represent a pool of KaiC in a KaiB-competent binding state associated with KaiA sequestration: [Afree] ¼ [A0]nS*). In this formulation, all of the intricate structural dynamics implicated in CI- and CII-dependent nucleotide exchange, ATP hydrolysis, and protein-protein interactions are encoded in the rate dependencies of the transition functions (kj,m) that determine the oscillator period and the effect of various perturbations on the clock. It is reasonable to approximate the transition rates such that they depend on concentrations ratios as suggested by various experiments (Kageayama et al. 2006): k

j,m

¼k

j,m ð½ADP=½ATP, ½Afree =½C , ½B=½C ,

. . .Þ

ð10Þ

A further assumption is that the transition rates may be approximated as a product of functions: k

j,m

¼ f

j,m ð½ADP=½ATPÞ

g

j,m ð½Afree =½C Þh j,m ð½Bfree =½C ,

. . .Þ

ð11Þ

Since KaiB seems to have little effect on the population phosphoform transition rates (from partial reactions): k

~

j,m f j,m ð½ADP=½ATPÞg j,m ð½Afree =½C Þ

ð12Þ

One complication is that the dependence of these phospho-transition rates as a function of KaiA concentration is generally associated with a varying affinity of hexamers with variable phosphoform composition of KaiC for KaiA, such that the KaiA dissociation probability increases as hexamers become hyper-phosphorylated (e.g., Mori et al. 2018). Thus in these state transition state models, different assumptions are used to approximate the KaiA dependence of the effective phosphotransition rates; for example, in Paijmans et al. 2017, a proposed free energy landscape for the CII domain phosphorylation is used to estimate these rates. While phosphoform state transition models are useful for gaining insights into clock dynamics, models and simulations of hexamer states allow for a detailed exploration of protein-protein interactions and hexamer heterogeneity on clock dynamics.

308

M. Byrne

3 Hexamer Models and Allosteric Transitions Several simplified and more complex hybrid hexamer models that include phosphoform state transitions have been published (e.g., Van Zon et al. 2007; Eguchi et al. 2008; Qin et al. 2010a; Lin et al. 2014; Paijmans et al. 2017; Mori et al. 2018; Das et al. 2018). One class of models enumerates differential equations for the various phosphorylation transitions among various states: [C]j, [AC]k, [BC]l, [ABC]m and potentially includes active (Cj*) and inactive (Cj) states of KaiC analogous to the “tense” and “relaxed” states in the allosteric MYC model (Monod et al. 1965). The number of differential equations and complexes rapidly rises to account for the combinatorial complexity of potential states; for example, inclusion of the phosphoform distribution for KaiC-only states, Cj (S, T, D) with S, T and D taking on values 0–6 constrained by 0  S + T+ D  6 and U ¼ 6(S + T + D). In the above the states, A, B, and C are assumed in their native form in solution, assumed to be dimer, tetramer (or dimer), and hexamer, respectively. Another class of models uses Monte Carlo (pseudo-random number generation) to simulate the state transitions probabilistically for both monomer transitions and hexamer interactions. In exploring the system dynamics, both types of models were constructed (including hybrid versions); the Monte-Carlo models are perhaps more intuitive since one can directly visualize the system dynamics, and the statistical information on the hexamer populations are directly available for investigating hexamer-hexamer synchrony mechanisms. For Monte-Carlo simulations in this system, I have separated loops over potential monomer reactions and hexamer transitions in the code and used a matrix representation for hexamer states (using rows for hexamers and columns for phosphorylation sites, hexamer binding status, etc.) For potential reactions within a particular class (such as monomer phosphoform or hexamer reactions), with transition probabilities per unit time, pj, timesteps are chosen so that ∑pj < < 1 and a uniform pseudo-random number on the interval [0,1] determines which particular reaction can occur (in a reaction class) on an interval of width pj, with the probability of no state transition during a time-step, (1  ∑ pj) > 0. These stochastic methods (Gillespie 1977) reproduce the ODE methods for partial reactions (albeit with higher computational expense) as a simple check on the algorithms for simulating transitions and are useful for studying hexamer variability and synchronization across a molecular population. Figure 2 includes deterministic simulations for the phosphorylation and dephosphorylation phases from linear interconversion rates between phospho-states Eq. (1a, 1b, 1c and 1d) along with stochastic simulations of partial reactions of KaiC and KaiA+KaiC, and the subsequent effect of KaiB inclusion inducing oscillations with allosteric switching of KaiC to a KaiB-binding competent state (figures adapted from the supplementary modeling section of Mori et al. 2018). As is well known, basic cyclic oscillatory “schemes” with constant rates do not undergo sustained oscillations, e.g., the highly simplified model (Qin et al. 2010b), C0 ! AC0 ! AC1 ! . . . ! ACN ! ABCN ! ABCN1 . . .!ABC0 ! A + B + C0 undergoes damped oscillations without another synchrony mechanism (even though

Insights from Mathematical Modeling/Simulations of the In Vitro KaiABC Clock

309

there is clearly explicit KaiA sequestration during the “dephosphorylation” phase, with the phase-transition step indicated by ACN!ABCN). Nonlinearities due to KaiA-KaiC differential affinity (Van Zon et al. 2007) and phase-dependent monomer exchange (Mori et al. 2007; Yoda et al. 2007) were also proposed as hexamerhexamer synchronizing mechanisms; more recent models use KaiC hexamer stateswitching (Paijmans et al. 2017, Mori et al. 2018) to a KaiB-competent binding state, which can be schematically indicated by (a) Cm ! Cm* ! BC*m and subsequent KaiA sequestration, and (b) A + BC*m ! ABC*m based on an individual hexamer’s phosphoform composition and nucleotide distribution. Roughly speaking the conformational switch (Cm ! Cm*) is potentially reversible and rate limited by the CI ATP-ase rate (Phong et al. 2013; Oyama et al. 2016); the CI ATP-ase rate can be explicitly modeled or incorporated into an effective KaiB binding rate. The subsequent KaiB binding step appears to be regulated by a slow fold-switching rate of KaiB, Chang et al. 2015, (B ! B*), that can be incorporated into an effective slow “binding” rate of KaiB to KaiC*, B + Cm* ! BCm* ! B*C*m. From our simulations (Mori et al. 2018) the KaiB binding-competent hexameric state, Cm*, includes a variety of mixed hexamer configurations with a phosphoform composition likely with some minimum number of pS/T (“S”) protomers, and/or possibly a minimum number of pS/pT (“D”) protomers. In contrast to previous models that the author constructed, in these hexameric states ({Cm*, BCm*, ABCm*}) the phosphoform transitions in these hexamers are influenced by global KaiA concentrations in the same manner as the {Cj, ACj} states; thus, while the BC*-associated states are capable of sequestering KaiA, they do not dephosphorylate until sufficient global free KaiA is sequestered and the hexamer population synchronously dephosphorylates from auto-catalysis. This allosteric switch was previously investigated using the constrained phosphoform transition rates published in Rust et al. 2007, and a stochastic matrix model for simulating the state transitions to a KaiB-binding competent state (Byrne 2014). For the partial reactions, the simulations suggested a typical hexamer with eight phosphates bound in the average state at the end of the phosphorylation phase; in simulations the average KaiC hexamer in the population at the “end” of the phosphorylation phase has an average monomer subunit distribution consisting of 3D, 2T, and either 1U or 1S state. By simulating the state-transition for different pS/T protomer numbers per hexamer, simulations suggested the heterogeneous hexamer pool requires at least two pS/T protomers per hexamer to initiate KaiB binding and subsequent KaiA sequestration (this is assuming typical “slow” effective scaled KaiB binding rates of order the circadian timescale, ~0.1 h.1). Allowing monomer shuffling among all hexamers indicates that pS/T protomers from different hexamers could allow state transitions to a KaiB-binding competent state (C ! C*) to occur faster in the hexamer population than without (suggesting a role of monomer exchange in speeding up the autocatalytic process by synchronizing sequestration). Simulations also showed for a mixture of cycling hexamers, one pool approximately in the middle of the phosphorylating phase and one pool approximately in the middle of the de-phosphorylating phase (with the same stoichiometry); mixing these pools resulted in the general dominance of the dephosphorylating pool, similar to that seen

310

M. Byrne

Fig. 5 Adapted from Mori et al. 2018. (a). Simulated phosphoform-dependent differential affinity (PDDA) in a hexamer population by variation of the scaled (dimensionless) dissociation constant (kd ¼ kA/(k+A[C0]) in a 1:1 binding model of KaiA dimer to KaiC hexamer using the pseudosteady state approximation for the kinase-promoting KaiA-KaiC complex; the oscillations in affinity correspond to the phosphoform oscillations shown in Fig. 2b. (b) Investigating the robustness of oscillations to initial KaiA concentrations in allosteric simulations PDDA (where the -PDDA simulations, black, allowed the D!S transition). Further details can be found in the supplementary modeling section of Mori et al. 2018

in experiments (Kitayama et al. 2013). This does not include bound ATP and ADP-bound nucleotides in these matrix simulations, and external ATP and ADP, with ATP hydrolysis in CI and re-phosphorylation of ADP during dephosphorylation reactions as seen in experiments (Nishiwaki and Kondo 2012; Egli et al. 2012); it would likely be useful to do so, and perhaps future work will result in revisiting and revising simulations to include ATP/ADP energetics and comparison with various perturbation experiments to probe thermodynamic aspects of the system dynamics in more detail. In a more recent study (Mori et al. 2018), phosphoform differential affinity (PDDA) was simulated using a pseudo-steady state assumption for the phosphorylation phase KaiA-KaiC interaction (i.e., rapid KaiA-KaiC on-off rates relative to the circadian timescale during the phosphorylation phase prior to sequestration) to assign a KaiA-binding probability for individual hexamers based on their phosphoform composition (Fig. 5). Phenomenologically, we assumed a linear model for the KaiA dissociation probability with varying weights for each T, S, and D protomers (roughly in the ratio 1:2:4 based on electron microscopy of KaiAKaiC bound-state lifetime data). Constant phosphoform transition rates were used with KaiA bound (approximately reproducing experimental partial reaction data) and the phosphoform de-phosphorylation rates adapted from Rust et al. (2007) for non-KaiA bound hexamers. This study (and Paijmans et al. 2017) suggests that the phosphoform transition rates for the KaiA-KaiC partial reaction (absent KaiA sequestration) vary as the hexamers are phosphorylated, such that the fitted average phosphoform transition rates are a consequence PDDA of KaiA during the phosphorylation phase. One of the roles of PDDA seems to be promoting the

Insights from Mathematical Modeling/Simulations of the In Vitro KaiABC Clock

311

dephosphorylation transition, D ! S, enhancing sequestration, and simulations indicated that promoting the transition was useful for extending the range of KaiA concentrations that generates sustained oscillations. Notably PDDA could be responsible for sufficient D ! S transitions in the presence of free KaiA and may be necessary for oscillations; on the other hand if sufficient D ! S transitions occur with KaiA transiently bound, then PDDA does not seem to be a necessary condition for oscillations. To summarize, many research groups have gained substantial insights from constructing and simulating both simple and more complex mathematical models of this circadian oscillator. Much remains to be learned about the intracellular coupling of this post-translational oscillator (PTO) to transcription-translation feedback (TTFL); some earlier work from mathematical modeling attempts at PTO-TTFL coupling include Qin et al. (2010b); Zwicker et al. (2010); and Paijmans et al. (2016a, b). Even more exciting would be finding other purely post-translational oscillators that can be reconstituted in vitro, exploring the extent to which lessons learned from the remarkable KaiABC protein-based oscillator are applicable more generally across diverse biological systems.

References Abe J, Hiyama TB, Mukaiyama A, Son S, Mori T, Saito S et al (2015) Atomic-scale origins of slowness in the cyanobacterial circadian clock. Science 349(6245):312–316 Brettschneider C, Rose RJ, Hertel S, Axmann IM, Heck AJR, Kollmann M (2010) A sequestration feedback determines dynamics and temperature entrainment of the KaiABC circadian clock. Mol Syst Biol 6(389):1–10 Byrne M (2009) Mathematical modeling of the in vitro cyanobacterial circadian oscillator. In: Ditty JL, Mackey SR, Johnson CH (eds) Bacterial circadian programs, vol 16. Springer, Berlin, pp 283–300 Byrne M (2014) Computational modeling of protein interactions and phosphoform kinetics in the KaiABC cyanobacterial circadian clock. arXiv:1405.3586 Byrne M (2020) Simple post-translational circadian clock models from selective sequestration bioRxiv 2020.02.21.958827; https://doi.org/10.1101/2020.02.21.958827 Chang Yg, Cohen SE, Phong C, Myers WK, Kim Yi, Tseng R, et al. (2015) A protein fold switch joins the circadian oscillator to clock output in cyanobacteria. Science; 349(April):324–328. https://doi.org/10.1126/science.1260031 PMID: 26113641 Chew J, Leypunskiy E, Lin J, Murugan A, Rust MJ (2018) High protein copy number is required to suppress stochasticity in the cyanobacterial circadian clock. Nat Commun 9:3004 Clodong S, DuÈhring U, Kronk L, Wilde A, Axmann I, Herzel H et al (2007) Functioning and robustness of a bacterial circadian clock. Mol Syst Biol 3:1–9. https://doi.org/10.1038/ msb4100128 Das S, Terada TP, Sasai M (2018) Single-molecular and ensemble-level oscillations of cyanobacterial circadian clock. Biophys Physicobiol 15:136–150. https://doi.org/10.2142/ biophysico.15.0_136. ECollection 2018. PMID: 29955565 Egli M, Mori T, Pattanayek R, Xu Y, Qin X, Johnson CH. (2012) Dephosphorylation of the core clock protein KaiC in the cyanobacterial KaiABC circadian oscillator proceeds via an ATP synthase mechanism. Biochemistry: 51(8):1547–1558. https://doi.org/10.1021/bi201525n PMID: 22304631

312

M. Byrne

Eguchi K, Yoda M, Terada TP, Sasai M (2008) Mechanism of robust circadian oscillation of KaiC phosphorylation in vitro. Biophys J 95:1773–1784. 18502804. https://doi.org/10.1529/ biophysj.107.127555 Emberly E, Wingreen NS (2006) Hourglass model for a protein-based circadian oscillator. Phys Rev Lett. 96:038303. https://doi.org/10.1103/PhysRevLett.96.038303 PMID: 16486780 Forger DB, Peskin CS (2003) A detailed predictive model of the mammalian circadian clock. Proc Natl Acad Sci 100(25):14806–14811. https://doi.org/10.1073/pnas.2036281100 Gillespie DT (1977) Exact stochastic simulation of coupled chemical reactions. J Phys Chemf 81 (25):2340–2361. https://doi.org/10.1021/j100540a008 Goldbeter A (1995) A model for circadian oscillations in the Drosophila period protein (PER). Proc Biol Sci 261(1362):319–324. https://doi.org/10.1098/rspb.1995.0153 Jolley CC, Ode KL, Ueda HR (2012) (2012) A design principle for a posttranslational biochemical oscillator. Cell Rep 2(4):938–950. https://doi.org/10.1016/j.celrep.2012.09.006 Kageyama H, Nishiwaki T, Nakajima M, Iwasaki H, Oyama T et al (2006) Cyanobacterial circadian pacemaker: Kai protein complex dynamics in the KaiC phosphorylation cycle in vitro. Mol Cell 23:161–171 Kitayama Y, Nishiwaki-Ohkawa T, Sugisawa Y et al (2013) KaiC intersubunit communication facilitates robustness of circadian rhythms in cyanobacteria. Nat Commun 4:2897. https://doi. org/10.1038/ncomms3897 Leloup J-C, Goldbeter A (2003) Toward a detailed computational model for the mammalian circadian clock. Proc Natl Acad Sci 100(12):7051–7056. https://doi.org/10.1073/pnas. 1132112100 Lin J, Chew J, Chockanathan U, Rust MJ (2014) Mixtures of opposing phosphorylations within hexamers precisely time feedback in the cyanobacterial circadian clock. Proc Natl Acad Sci. 111 (37):E3937-E3945. https://doi.org/10.1073/pnas.1408692111 PMID: 25197081 Ma L, Ranganathan R (2012) Quantifying the rhythm of KaiB-C interaction for in vitro cyanobacterial circadian clock. PLoS One 7(8):e42581. https://doi.org/10.1371/journal.pone. 0042581 PMID: 22900029 Mehra A, Hong CI, Shi M, Loros JJ, Dunlap JC, Ruoff P (2006) Circadian rhythmicity by autocatalysis. PLoS Comput Biol 2(7):1–8. https://doi.org/10.1371/journal.pcbi.0020096 Miyoshi F, Nakayama Y, Kaizu K, Iwasaki H, Tomita M (2007) A mathematical model for the kai-protein–based chemical oscillator and clock gene expression rhythms in cyanobacteria. J Biol Rhythm 22(1):69–80. https://doi.org/10.1177/0748730406295749. Monod J, Wyman J, Changeux JP (1965) On the nature of allosteric transitions: A plausible model. J Mol Biol. 12(1):88–118. https://doi.org/10.1016/S0022-2836(65)80285-6 PMID: 14343300 Mori T, Williams DR, Byrne MO, Qin X, Egli M, Mchaourab HS et al (2007) Elucidating the ticking of an in vitro circadian clockwork. PLoS Biol 5(4):841–853. https://doi.org/10.1371/ journal.pbio.0050093 Mori T, Sugiyama S, Byrne M, et al (2018) Revealing circadian mechanisms of integration and resilience by visualizing clock proteins working in real time. Nat Commun 9:3245. https://doi. org/10.1038/s41467-018-05438-4 Murakami R, Miyake A, Iwase R, Hayashi F, Uzumaki T, Ishiura M (2008) ATPase activity and its temperature compensation of the cyanobacterial clock protein KaiC. Genes Cells 13 (4):387–395. https://doi.org/10.1111/j.1365-2443.2008.01174.x PMID: 18363969 Nakajima M, Imai K, Ito H, Nishiwaki T, Murayama Y, Iwasaki H, et al (2005) Reconstitution of circadian oscillation of cyanobacterial KaiC phosphorylation in vitro. Science 308 (5720):414–415. https://doi.org/10.1126/science.1108451 PMID: 15831759 Nishiwaki T, Kondo T (2012) Circadian autodephosphorylation of cyanobacterial clock protein KaiC occurs via formation of ATP as intermediate. J Biol Chem. 287(22):18030–18035. https:// doi.org/10.1074/jbc.M112.350660 PMID: 22493509 Nishiwaki T, Satomi Y, Nakajima M, Lee C, Kiyohara R, Kageyama H et al (2004) Role of KaiC phosphorylation in the circadian clock system of Synechococcus elongatus PCC 7942. Proc Natl Acad Sci 101(38):13927–13932. https://doi.org/10.1073/pnas.0403906101

Insights from Mathematical Modeling/Simulations of the In Vitro KaiABC Clock

313

Nishiwaki T, Satomi Y, Kitayama Y, Terauchi K, Kiyohara R, Takao T, et al (2007) A sequential program of dual phosphorylation of KaiC as a basis for circadian rhythm in cyanobacteria. EMBO J. 26:4029–4037. https://doi.org/10.1038/sj.emboj.7601832 PMID: 17717528 Novák B, Tyson J (2008) Design principles of biochemical oscillators. Nat Rev Mol Cell Biol 9:981–991. https://doi.org/10.1038/nrm2530 Oyama K, Azai C, Nakamura K, Tanaka S, Terauchi K (2016) Conversion between two conformational states of KaiC is induced by ATP hydrolysis as a trigger for cyanobacterial circadian oscillation. Sci Rep 6(32443):1–11 Paijmans J, Bosman M, ten Wolde PR, Lubensky DK (2016a) Discrete gene replication events drive coupling between the cell cycle and circadian clocks. Proc Natl Acad Sci. 113 (22):4063–4068. https://doi.org/10.1073/pnas.1507291113 PMID: 27035936 Paijmans J, Lubensky DK, ten Wolde PR (2016b) Period robustness and entrainability under changing nucleotide concentrations in the post-translational Kai circadian clock. arXiv:1612.08305 Paijmans J, Lubensky DK, ten Wolde PR (2017) A thermodynamically consistent model of the post-translational Kai circadian clock. PLoS Comput Biol 13(3):e1005415. https://doi.org/10. 1371/journal.pcbi.1005415 Phong J, Markson S, Wilhoite CM, Rust MJ (2013) Robust and tunable circadian rhythms from differentially sensitive catalytic domains. Proc Natl Acad Sci U S A 110:1124–1129 Qin X, Byrne M, Mori T, Zou P, Williams DR, Mchaourab H, Johnson CH (2010a) Intermolecular associations determine the dynamics of the circadian KaiABC oscillator. Proc Nat Acad Sci 107 (33):14805–14810 Qin X, Byrne M, Xu Y, Mori T, Johnson CH. (2010b) Coupling of a core post-translational pacemaker to a slave transcription/translation feedback loop in a circadian system. PLoS Biol. 8(6):e1000394. https://doi.org/10.1371/journal.pbio.1000394 PMID: 20563306 Rust MJ, Markson JS, Lane WS, Fisher DS, O'Shea EK (2007) Ordered phosphorylation governs oscillation of a three-protein circadian clock. Science 318(5851):809–812 Rust MJ, Golden SS, O’Shea EK. (2011) Light-driven changes in energy metabolism directly entrain the cyanobacterial circadian oscillator. Science 331(6014):220–223. https://doi.org/10. 1126/science.1197243 PMID: 21233390 Terauchi K, Kitayama Y, Nishiwaki T, Miwa K, Murayama Y, Oyama T, et al (2007) ATPase activity of KaiC determines the basic timing for circadian clock of cyanobacteria. Proc Natl Acad Sci. 104(41):16377–16381. https://doi.org/10.1073/pnas.0706292104 PMID: 17901204 Tomita J, Nakajima M, Kondo T, Iwasaki H (2005) No transcription-translation feedback in circadian rhythm of KaiC phosphorylation. Science 307(5707):251–254. https://doi.org/10. 1126/science.1102540 Van Zon JS, Lubensky DK, Altena PRH, ten Wolde PR (2007) An allosteric model of circadian KaiC phosphorylation. Proc Natl Acad Sci. 104(18):7420–7425. https://doi.org/10.1073/pnas. 0608665104 PMID: 17460047 Yoda M, Eguchi K, Terada TP, Sasai M (2007) Monomer-shuffling and allosteric transition in KaiC circadian oscillation. PLoS One 5(5):1–8 Zwicker D, Lubensky DK, ten Wolde PR (2010) Robust circadian clocks from coupled proteinmodification and transcription-translation cycles. Proc Natl Acad Sci. 107(52):22540–22545. https://doi.org/10.1073/pnas.1007613107 PMID: 21149676

Part II

Circadian Phenomena in Microbiomes/ Populations and Bacteria Besides Cyanobacteria

Basic Biology of Rhythms and the Microbiome Melina Heinemann, Karina Ratiner, and Eran Elinav

Abstract The mammalian microbiome undergoes diurnal oscillations in composition and function throughout a 24-period that are regulated by host clock and nutritional signals. These diurnal oscillations, in turn, impact the host’s transcriptome and multiple other physiologic functions. Emerging evidence has begun to uncover the molecular mechanisms that underlie the coordinated metaorganismal diurnal rhythmicity with crucial implications for homeostasis and disease. Herein, we highlight current mechanistic understanding by which the diurnally oscillating commensal microbiota is regulated by the host and its environment, and how commensals diurnally modulate host biology in health and disease. Finally, current challenges, open questions, and perspectives in this exciting new field of chronobiology are discussed.

1 Introduction 1.1

Circadian Rhythms in Mammals

Created by the Earth’s rotation in relation to the sun, most living beings on Earth experience a daily light/dark cycle (Kaczmarek et al. 2017). Living organisms are subject to an evolutionary pressure to develop regulatory clock networks to adapt to the circadian nature of their environment (Troein et al. 2009). Circadian rhythms are

Melina Heinemann and Karina Ratiner contributed equally with all other contributors. M. Heinemann · K. Ratiner Immunology Department, Weizmann Institute of Science, Rehovot, Israel E. Elinav (*) Immunology Department, Weizmann Institute of Science, Rehovot, Israel Cancer-Microbiome Division Deutsches Krebsforschungszentrum (DKFZ), Heidelberg, Germany e-mail: [email protected]; [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_16

317

318

M. Heinemann et al.

cycles of physiology and behavior driven by an endogenous oscillator set to a period of approximately 24 hours (Panda et al. 2002). Evolving mechanisms of diurnal homeostasis allow organisms to accommodate to rhythmic environmental challenges and, thus, provide them with a competitive survival advantage (Dodd et al. 2005; Hellweger 2010). Diurnal rhythms of varying degree of complexity are exhibited by the majority of organisms including animals, bacteria, fungi, and plants (Hastings et al. 2007). For many years, circadian clocks were conceptualized in terms of feedback loops of transcription and translation. However, more recent studies have unraveled a wealth of posttranslational circadian oscillators (Brown et al. 2012). Going beyond the “clock” metaphor, more recent research has revealed that unlike a conventional clockwork, the circadian systems of many organisms are dynamic and highly adaptive (Roenneberg and Merrow 2005). Nevertheless, certain biological rhythms persist in the absence of environmental signals (such as light and temperature) (Panda et al. 2002). A mammalian 24-h circle is typically comprised of one active and feeding as well as one resting and fasting phase. In humans, the light phase represents the active phase, while in mice, arguably the most commonly used animal model in biomedical research, the dark phase constitutes the active phase. Growing evidence suggests that the host’s circadian rhythms, diet, and commensal microbiota are closely interconnected (Kaczmarek et al. 2017). In mammals, the central oscillator, located in the suprachiasmatic nucleus (SCN) of the hypothalamus, generates a circadian rhythm and synchronizes the periphery (Hastings et al. 2003). The SCN is normally synchronized to solar time by retinal afferents from intrinsically photoreceptive retinal ganglion cells (Mohawk et al. 2012). Peripheral tissues generate oscillations using circuitry based on clock proteins that can function cell-autonomously to generate transcriptional rhythms (Kaczmarek et al. 2017). In mice, up to 45% of the transcriptome adheres to an approximately 24-h oscillation pattern (Zarrinpar et al. 2016). The underlying mechanism in both neurons and peripheral cells of mammals is a transcriptional-translational feedback loop oscillating with a periodicity of 24 h. It comprises the proteins CLOCK and BMAL1 that activate transcription of genes encoding the repressors PERIOD (PER) and CRYPTOCHROME (CRY). The products of these genes can form the PER/CRY repressive complex that can translocate into the nucleus and inhibit CLOCK/BMAL1 transcription activity, subsequently resulting in PER and CRY gene repression (Green et al. 2008). Figure 1 outlines the interactions between environmental, host-related, and microbial factors in circadian regulation and its impact on physiological outcomes

1.2

Diurnal Rhythms of the Mammalian Microbiota

The human body is colonized by a diverse community of microorganisms, especially in the gut. Circadian rhythms in prokaryotes are far less understood compared to those in mammals and have primarily been investigated in light-responsive cyanobacteria (Nobs et al. 2019). Though the prokaryotic microorganisms largely

Basic Biology of Rhythms and the Microbiome

319

Environmental changes

Clock Bmal1 Cry Per

SCN Circadian clocks

Timing and composition of food

Gut microbiome

Body weight

Blood pressure

Glucose tolerance

Hepatic metabolism

Fig. 1 Circadian rhythm in host-diet-microbiome interactions. The circadian clock oscillates with a periodicity of 24 h in neurons of the suprachiasmatic nucleus (SCN) and peripheral cells and is entrained by diurnal environmental changes. It leads to circadian oscillations in the composition and function of the gut microbiome, which in turn modulates the circadian rhythm and associated physiological functions of the host. The timing and composition of food intake is another major factor interacting with both host rhythms and circadian microbiome oscillations

constituting the mammalian gut microbiome are not exposed to light, the daily environmental changes in the intestine lead to various diurnal oscillations in the commensal microbial community (Nobs et al. 2019; Saran et al. 2020). The gut microbiome’s taxonomic and functional genomic composition is highly dynamic, exhibiting cyclical fluctuations (Thaiss et al. 2014, 2016; Zarrinpar et al. 2014; Liang et al. 2015; Deaver et al. 2018; Wu et al. 2018; Godinho-Silva et al. 2019; Leach et al. 2019). In mice, the proportion of species belonging to the Firmicutes phylum was reported to peak during nocturnal feeding and to reach its nadir during daytime

320

M. Heinemann et al.

fasting. Contrary, the phyla Bacteroidetes and Verrucomicrobia species peak during daytime fasting and bottom out during nocturnal feeding (Zarrinpar et al. 2014). Moreover, the absolute number of fecal bacteria also seems to be subject to diurnal oscillations. A distinction must be drawn regarding the circadian variation of the relative versus the absolute abundance of a bacterial clade. As such, no circadian oscillation was demonstrated for the relative abundance of the phylum Proteobacteria, whereas the inferred absolute abundance oscillated during the light-dark cycle (Liang et al. 2015). The bacterial microbiome features rhythmic patterns of localization and metabolite secretion in the colon. The abundance of commensal bacterial genes implicated in chemotaxis and flagellar assembly was shown to reach their peak at the end of the resting (light) phase in mice (Thaiss et al. 2016). This may drive bacterial penetration into the intestinal mucus layer in order to utilize mucus as a nutritional source when food intake is reduced. The interconnection between anabolic and catabolic metabolism and circadian clocks is required to synchronize the energy turnover with diurnal variations in nutrient supply. Feeding in restricted periods of the light-dark cycle dissociates peripheral clocks from the central master zeitgeber. The relationship between circadian genes and feeding rhythms seems to be interdependent. Interestingly, mice with genetically altered circadian clock genes not only lose many host transcriptome oscillations but also their usual nocturnal feeding behavior. However, restricting access to food to the dark cycle partly rescues behavioral and transcriptional patterns in these mice, strongly suggesting that many circadian processes are driven by feeding rhythms independently of the central master clock (Hughes et al. 2009; Vollmers et al. 2009). Indeed, the microbiome’s circadian rhythmicity and host metabolism are intertwined. Various microbiome-derived metabolites oscillate diurnally including secondary bile acids (Zhang et al. 2011; Joyce et al. 2014). Plasma bile acids and key genes in bile acid biosynthesis are regulated by both the host’s hepatic molecular clock and by food intake (Eggink et al. 2017). Alteration of microbiota-derived short-chain fatty acids (SCFAs) under high-fat diet directly modulates circadian clock gene expression in hepatocytes and other peripheral tissues (Leone et al. 2015; Parkar et al. 2019). Knockout of the important peripheral circadian rhythm orchestrating genes Per1 and Per2 leads to oscillatory loss in many bacterial operational taxonomic units (OTUs) under ad libitum feeding (Fig. 2a). In the mouse gut, about 23% of identified KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways and the respective underlying bacterial gene abundances display oscillations within a 24-h cycle. Among these, levels of genes encoding for pathways involved in energy metabolism, DNA repair, and cell growth show a peak during the dark (active) phase, whereas those involved in detoxification, motility, and environmental sensing reach the peak during the light (resting) phase in mice. The diurnal fluctuation of KEGG pathways in microbiota from wild-type mice is absent in mice with knockout of Per1 and Per2 (Fig. 2b). In humans, 10% of bacterial OTUs and around 20% of KEGG pathways show circadian oscillation patterns. Mice deficient in circadian rhythmicity because of knockout of the Per1 and Per2 genes feature lower alpha biodiversity. The loss of bacterial circadian oscillations in arrhythmic Per1/

Fig. 2 Loss of taxonomic and functional microbial diurnal oscillations in mice with Per1/2 knockout. (a) Taxonomic diurnal oscillation: compared to wild-type mice, mice with knockout of Per1 and Per2 show oscillatory loss in many bacterial operational taxonomic units. (b) Functional diurnal oscillation: wild-type mice show diurnal fluctuation of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways in microbiota, which is absent in mice with knockout of Per1 and Per2. Samples were taken at time points of changing light conditions (Zeitgeber times [ZT] 12 and 0, i.e., “dusk” and “dawn”) and the midpoint of the dark and light phases (ZT 18 and 6, respectively) (Thaiss et al. 2014)

Basic Biology of Rhythms and the Microbiome 321

322

M. Heinemann et al.

Fig. 3 Manifestations of diurnal rhythms in the mammalian microbiota. The intestinal microbiome displays cyclical rhythms over a 24-hour period in terms of taxonomic composition, biogeographical localization (rhythmic bacterial adherence to the mucosa and penetration into the mucus layer), production of various metabolites (such as short-chain fatty acids, carbohydrates, etc.), and functional activities with some genes being more expressed during the nighttime and others during daytime

2-deficient mice was rescued by introducing time-restricted feeding (Thaiss et al. 2014). Therefore, timing of feeding is a dominant orchestrator of temporal microbiome composition and functional dynamics. Although it is now evident that commensal microbial signaling affects maintenance of gut homeostasis and circadian control of intestinal and various extra-intestinal functions, it is still largely unclear through which mechanisms commensal bacteria and host tissues communicate and how the microbiome may take advantage of host circadian functions to maintain its own homeostasis. Moreover, since circadian phenomena of the host’s physiology can be habituated by daily cycles of restricted feeding (Stephan 1984), it remains to be clarified whether the restoration of microbial circadian rhythmicity is explained by direct influence of feeding, restoration of the host’s circadian clock, or

Basic Biology of Rhythms and the Microbiome

323

a combination of both. Figure 3 summarizes relevant aspects of diurnal fluctuation in microbiota.

2 Circadian System in Host-Microbiome Interactions 2.1

Host Factors Shaping Microbiota Rhythms

In addition to studies showing that the ablation of the molecular clock Per genes causes changes in the commensal microbiota composition (Thaiss et al. 2014; Zarrinpar et al. 2016), microbial configuration was also shown to be influenced by Bmal1-dependent forward limb of the clock signaling pathway (Liang et al. 2015). Interestingly, the impact of host factors on microbial circadian rhythmicity shows a sex-dimorphic pattern. Although both male and female mice display circadian behavior and physiology, circadian oscillation in females is more pronounced than in male animals. Deficiency of Bmal1 not only abrogated circadian behavior of the fecal microbiome in both sexes, but the resulting shifts in the microbiome configuration showed similarly intriguing sex-specific patterns (Liang et al. 2015). The absence of the microbiota in germ-free mice levels hepatic rhythmic and sex-dimorphic gene expression and metabolism. Additionally, there is evidence that sex-specific diurnal rhythms of gene expression are driven by microbial metabolites (Weger et al. 2019). Exposure to unnatural light cycles are increasingly common in modern society due to availability of electricity, long-distance travel, and shift work. In mice, perturbation of the physiological cycle through constant light or dark exposure may lead to altered abundance of several taxonomic groups in the intestinal microbiome and altered levels of certain microbiome-derived plasma metabolites subject to regulation by the microbiome, such as tryptophan (Kim et al. 2019). The induction of jet lag leads to deregulated microbiota diurnal fluctuations and altered microbiota composition (dysbiosis), driven by impaired feeding rhythmicity, in both humans and mice (Thaiss et al. 2014). Although there is growing concern that altered circadian rhythms, sleep deprivation, and related stressors may adversely impact human gut microbiota with significant health implications, evidence for this is still very scarce (Zhang et al. 2017; Karl et al. 2018) and warrants further research. At least one species commonly colonizing the human gut, Enterobacter aerogenes, may be sensitive to the host-derived neuro-hormone melatonin, strongly suggesting the existence of autochthonous clocks in some commensal bacterial cells synchronizing with host circadian regulators (Paulose and Cassone 2016). The host’s innate immune arm utilizes a wealth of antibacterial polypeptides, known as defensins, to regulate commensal microbiota and to combat invading pathogens. Mouse enteric defensins, also known as cryptdins, are produced and secreted constitutively but are increasingly expressed upon enteric infection or inflammation. An analysis of expression patterns of cryptdin 1 and cryptdin 4 around the circadian cycle in mice revealed a circadian oscillation of these defensins with a

324

M. Heinemann et al.

peak at the end of the dark phase (Froy et al. 2005). In mice, the circadian regulator Arntl has been shown to be a key regulator in intestinal group 3 innate lymphoid cells (ILC3s) and contributing to regulation of the gut microbiome by ILC3s in a circadian rhythm-dependent manner (Godinho-Silva et al. 2019). Immune system parameters change according to the time of day, and disruption of circadian rhythms has been linked to inflammatory pathologies (Curtis et al. 2014). As the host’s immune system is now known to be a major regulator of microbiome homeostasis (and vice versa), there is potentially an enormous wealth of circadian immunological mechanisms regulating the commensal microbiota, which are yet to be discovered.

2.2

The Influence of Microbiota on Host Rhythms and Metabolism

The oscillating bacterial adherence to the colonic mucus layer may regulate the circadian changes in the colonocyte epigenome and transcriptome (Thaiss et al. 2016). The host’s metabolism, which is increasingly recognized to be shaped by both host-intrinsic factors and the microbiome, is entwined with circadian regulation. The integrity of the circadian clock of intestinal epithelial cells (IECs) is required to regulate the dialog between IECs and the microbiota. This is especially mediated through a rhythmic expression of Toll-like receptors (TLRs) by IECs. The importance of this crosstalk is highlighted by the finding that microbiome signaling deficiencies induce a prediabetic syndrome due to ileal corticosterone overproduction, which manifests as a consequence of IEC clock disruption (Mukherji et al. 2013). The liver is a key organ orchestrating metabolic homeostasis and maybe a primary target impacted by gut microbiome-secreted factors influxing through the portal venous system. Indeed, the microbiome is required for integration of liver clock oscillations to regulate gene expression for optimal liver function. This has implications for the regulation of diverse metabolite levels, including glucose, cholesterol, free fatty acids, bilirubin, and lactate. Moreover, the hepatic clock-microbiome cross-talk influences xenobiotic metabolism, protein turnover, and redox balance (Montagner et al. 2016). Circadian transcriptomic changes in the liver may be regulated by timely fluctuations of microbiota-derived metabolites, including lipids, amino acids, carbohydrates, vitamins, nucleotides, and xenobiotics (Thaiss et al. 2016). The microbiome-derived SCFA butyrate may function as a histone deacetylase inhibitor in the liver, hereby exerting epigenetic control of host circadian rhythms (Leone et al. 2015). Preventive efficacy of dietary fiber intake and microbiome-derived acetate against hypertension may be partly mediated by their influence on the expression of circadian genes in the heart and the kidney (Marques et al. 2017). Unconjugated bile acids, generated through bile salt hydrolases activity of the gut microbiota, are potentially chronobiological regulators of host circadian gene expression. This

Basic Biology of Rhythms and the Microbiome

325

represents an additional potential mechanism for microbe-host crosstalk regulating host circadian gene expression (Govindarajan et al. 2016). Moreover, the microbiota modulates the diurnal variation in hepatic drug detoxification and hepatotoxicity. Acetaminophen, a potentially hepatotoxic drug undergoing hepatic metabolism, was demonstrated to exert hepatotoxicity with differing severity depending on the time of the day. This diurnal phenotype is co-regulated by the microbiome as germ-free or antibiotic-treated mice do not feature diurnal variation in acetaminophen-induced liver injury (Thaiss et al. 2016). Diurnal homeostasis is intertwined with weight maintenance and glucose tolerance, which is partially mediated by the microbiome. Jet-lag challenge induces alterations in the microbiome configuration in both humans and mice. This may confer an increased risk for metabolic disease, since obesity and glucose intolerance are transferable by fecal microbiota transplantation from jet-lagged humans and mice to germ-free mice (Thaiss et al. 2014). Although it is increasingly clear that circadian clocks are key orchestrators of immune responses (Scheiermann et al. 2018), understanding how circadian rhythmicity of the microbiota may impact host immunity is only at its very beginning. Some clues come from studies with bacterial pathogens. For example, in murine infection with the enteric pathogen Salmonella typhimurium, the efficacy of colonization by the pathogen and the magnitude of the host’s inflammatory response depend on the time of day of pathogen exposure (Bellet et al. 2013). Despite these advances, the understanding of the diurnal mechanisms underlying regulation of host biology by the microbiota remains largely unknown and warrant additional research.

2.3

Perspectives and Challenges

In recent years, major progress has been achieved in unraveling the interplay between the host’s and the microbiota’s circadian rhythms, related environmental cues, and their concerted impact on host physiology. However, in this relatively young field, many questions on the circadian regulation of the microbiome’s composition and function and associated host-microbiota mutualism remain hitherto unanswered. It is clear that more mechanistic evidence is required for a proper understanding of circadian phenomena at the host-microbiome interface. Therefore, the field presents exciting opportunities for future discoveries. Inquiries into the basic biology of circadian microbiota regulation hold great potential to deepen the understanding of inflammatory and metabolic disorders. Many fundamental questions in this area of research need to be addressed. What are the mechanisms regulating diurnal oscillations in commensal intestinal bacteria apart from feeding rhythms? Which autochthonous cellular clockworks may exist in commensal bacteria? What is the role of host-derived regulators of neural, hormonal, or immunologic origin? Is there a direct relationship between the central circadian regulator in the SCN and gut microbiota circadian rhythms? Which mechanisms

326

M. Heinemann et al.

underlie this potential relationship? Through which mechanisms are host peripheral clocks and dependent biological processes synchronized with bacterial oscillations? What is the role of circadian phenomena regarding non-bacterial members of the commensal microbiota? It will be crucial to disentangle the role of circadian control of immune responses by the host and modulation of host immunity by microbiota rhythms, e.g., by employing animal models with knockouts for specific genes involved in circadian regulation. It is challenging to directly demonstrate causal links between diurnal changes in the microbiome and host biology in humans. Nevertheless, in addition to elaborated basic research using advanced animal models, more comprehensive observational human studies are required with sufficiently deep analysis of metagenome and transcriptome profiles to confirm that circadian alterations associated with modern life style, such as represented by shift work, truly impact the human microbiome in a way significant for human health and disease. This is of high public health relevance as it is estimated that currently up to 30% of the working population perform shift work and about one-third of adults sleep less than 6 h per night (Liang and FitzGerald 2017). Going one step further, the recent emergence of microbiota-targeted therapies (Skelly et al. 2019) may inspire novel chronopharmacological approaches targeting the host’s peripheral clocks and microbiome rhythms in order to treat inflammatory or metabolic diseases.

References Bellet MM, Deriu E, Liu JZ et al (2013) Circadian clock regulates the host response to Salmonella. PNAS 110:9897–9902. https://doi.org/10.1073/pnas.1120636110 Brown SA, Kowalska E, Dallmann R (2012) (re)inventing the circadian feedback loop. Dev Cell 22:477–487. https://doi.org/10.1016/j.devcel.2012.02.007 Curtis AM, Bellet MM, Sassone-Corsi P, O’Neill LAJ (2014) Circadian clock proteins and immunity. Immunity 40:178–186. https://doi.org/10.1016/j.immuni.2014.02.002 Deaver JA, Eum SY, Toborek M (2018) Circadian disruption changes gut microbiome taxa and functional gene composition. Front Microbiol 9:737. https://doi.org/10.3389/fmicb.2018.00737 Dodd AN, Salathia N, Hall A et al (2005) Plant circadian clocks increase photosynthesis, growth, survival, and competitive advantage. Science (80-) 309:630–633. https://doi.org/10.1126/ science.1115581 Eggink HM, Oosterman JE, de Goede P et al (2017) Complex interaction between circadian rhythm and diet on bile acid homeostasis in male rats. Chronobiol Int 34:1339–1353. https://doi.org/10. 1080/07420528.2017.1363226 Froy O, Chapnik N, Miskin R (2005) Mouse intestinal cryptdins exhibit circadian oscillation. FASEB J 19:1920–1922. https://doi.org/10.1096/fj.05-4216fje Godinho-Silva C, Domingues RG, Rendas M et al (2019) Light-entrained and brain-tuned circadian circuits regulate ILC3s and gut homeostasis. Nature 574:254–258. https://doi.org/10.1038/ s41586-019-1579-3 Govindarajan K, MacSharry J, Casey PG et al (2016) Unconjugated bile acids influence expression of circadian genes: a potential mechanism for microbe-host crosstalk. PLoS One 11:e0167319. https://doi.org/10.1371/journal.pone.0167319 Green CB, Takahashi JS, Bass J (2008) The meter of metabolism. Cell 134:728–742. https://doi. org/10.1016/j.cell.2008.08.022

Basic Biology of Rhythms and the Microbiome

327

Hastings MH, Reddy AB, Maywood ES (2003) A clockwork web: circadian timing in brain and periphery, in health and disease. Nat Rev Neurosci 4:649–661. https://doi.org/10.1038/nrn1177 Hastings M, O’Neill JS, Maywood ES (2007) Circadian clocks: regulators of endocrine and metabolic rhythms. J Endocrinol 195:187–198. https://doi.org/10.1677/JOE-07-0378 Hellweger FL (2010) Resonating circadian clocks enhance fitness in cyanobacteria in silico. Ecol Model 221:1620–1629. https://doi.org/10.1016/j.ecolmodel.2010.03.015 Hughes ME, DiTacchio L, Hayes KR et al (2009) Harmonics of circadian gene transcription in mammals. PLoS Genet 5:e1000442. https://doi.org/10.1371/journal.pgen.1000442 Joyce SA, MacSharry J, Casey PG et al (2014) Regulation of host weight gain and lipid metabolism by bacterial bile acid modification in the gut. PNAS 111:7421–7426. https://doi.org/10.1073/ pnas.1323599111 Kaczmarek JL, Thompson S, Holscher HD (2017) Complex interactions of circadian rhythms, eating behaviors, and the gastrointestinal microbiota and their potential impact on health. Nutr Rev 75:673–682. https://doi.org/10.1093/nutrit/nux036 Karl PJ, Hatch AM, Arcidiacono SM et al (2018) Effects of psychological, environmental and physical stressors on the gut microbiota. Front Microbiol 9:2013. https://doi.org/10.3389/fmicb. 2018.02013 Kim YM, Snijders AM, Brislawn CJ et al (2019) Light-stress influences the composition of the murine gut microbiome, memory function, and plasma Metabolome. Front Mol Biosci 6:108. https://doi.org/10.3389/fmolb.2019.00108 Leach WB, Carrier TJ, Reitzel AM (2019) Diel patterning in the bacterial community associated with the sea anemone Nematostella vectensis. Ecol Evol 9:9935–9947. https://doi.org/10.1002/ ece3.5534 Leone V, Gibbons SM, Martinez K et al (2015) Effects of diurnal variation of gut microbes and high-fat feeding on host circadian clock function and metabolism. Cell Host Microbe 17:681–689. https://doi.org/10.1016/j.chom.2015.03.006 Liang X, FitzGerald GA (2017) Timing the microbes: the circadian rhythm of the gut microbiome. J Biol Rhythm 32:505–515. https://doi.org/10.1177/0748730417729066 Liang X, Bushman FD, FitzGerald GA (2015) Rhythmicity of the intestinal microbiota is regulated by gender and the host circadian clock. PNAS 112:10479–10484. https://doi.org/10.1073/pnas. 1501305112 Marques FZ, Nelson E, Chu PY et al (2017) High-fiber diet and acetate supplementation change the gut microbiota and prevent the development of hypertension and heart failure in hypertensive mice. Circulation 135:964–977. https://doi.org/10.1161/CIRCULATIONAHA.116.024545 Mohawk JA, Green CB, Takahashi JS (2012) Central and peripheral circadian clocks in mammals. Annu Rev Neurosci 35:445–462. https://doi.org/10.1146/annurev-neuro-060909-153128 Montagner A, Korecka A, Polizzi A et al (2016) Hepatic circadian clock oscillators and nuclear receptors integrate microbiome-derived signals. Sci Rep 6:20127. https://doi.org/10.1038/ srep20127 Mukherji A, Kobiita A, Ye T, Chambon P (2013) Homeostasis in intestinal epithelium is orchestrated by the circadian clock and microbiota cues transduced by TLRs. Cell 153:812–827. https://doi.org/10.1016/j.cell.2013.04.020 Nobs SP, Tuganbaev T, Elinav E (2019) Microbiome diurnal rhythmicity and its impact on host physiology and disease risk. EMBO Rep 20:e47129. https://doi.org/10.15252/embr.201847129 Panda S, Antoch MP, Miller BH et al (2002) Coordinated transcription of key pathways in the mouse by the circadian clock. Cell 109:307–320. https://doi.org/10.1016/S0092-8674(02) 00722-5 Parkar SG, Kalsbeek A, Cheeseman JF (2019) Potential role for the gut microbiota in modulating host circadian rhythms and metabolic health. Microorganisms 7:41. https://doi.org/10.3390/ microorganisms7020041 Paulose JK, Cassone VM (2016) The melatonin-sensitive circadian clock of the enteric bacterium Enterobacter aerogenes. Gut Microbes 7:424–427. https://doi.org/10.1080/19490976.2016. 1208892

328

M. Heinemann et al.

Roenneberg T, Merrow M (2005) Circadian clocks - the fall and rise of physiology. Nat Rev Mol Cell Biol 6:965–971. https://doi.org/10.1038/nrm1766 Saran AR, Dave S, Zarrinpar A (2020) Circadian rhythms in the pathogenesis and treatment of fatty liver disease. Gastroenterology 158:1948–1966.e1. https://doi.org/10.1053/j.gastro.2020.01. 050 Scheiermann C, Gibbs J, Ince L, Loudon A (2018) Clocking in to immunity. Nat Rev Immunol 18:423–437. https://doi.org/10.1038/s41577-018-0008-4 Skelly AN, Sato Y, Kearney S, Honda K (2019) Mining the microbiota for microbial and metabolite-based immunotherapies. Nat Rev Immunol 19:305–323. https://doi.org/10.1038/ s41577-019-0144-5 Stephan FK (1984) Phase shifts of circadian rhythms in activity entrained to food access. Physiol Behav 32:663–671. https://doi.org/10.1016/0031-9384(84)90323-8 Thaiss CA, Zeevi D, Levy M et al (2014) Transkingdom control of microbiota diurnal oscillations promotes metabolic homeostasis. Cell 159:514–529. https://doi.org/10.1016/j.cell.2014.09.048 Thaiss CA, Levy M, Korem T et al (2016) 03 microbiota diurnal rhythmicity programs host Transcriptome oscillations. Cell 167:1495–1510.e12. https://doi.org/10.1016/j.cell.2016.11.003 Troein C, Locke JCW, Turner MS, Millar AJ (2009) Weather and seasons together demand complex biological clocks. Curr Biol 19:1961–1964. https://doi.org/10.1016/j.cub.2009.09.024 Vollmers C, Gill S, DiTacchio L et al (2009) Time of feeding and the intrinsic circadian clock drive rhythms in hepatic gene expression. PNAS 106:21453–21458. https://doi.org/10.1073/pnas. 0909591106 Weger BD, Gobet C, Yeung J et al (2019) The mouse microbiome is required for sex-specific diurnal rhythms of gene expression and metabolism. Cell Metab 29:362–382.e8. https://doi.org/ 10.1016/j.cmet.2018.09.023 Wu G, Tang W, He Y et al (2018) Light exposure influences the diurnal oscillation of gut microbiota in mice. Biochem Biophys Res Commun 501:16–23. https://doi.org/10.1016/j. bbrc.2018.04.095 Zarrinpar A, Chaix A, Yooseph S, Panda S (2014) Diet and feeding pattern affect the diurnal dynamics of the gut microbiome. Cell Metab 20:1006–1017. https://doi.org/10.1016/j.cmet. 2014.11.008 Zarrinpar A, Chaix A, Panda S (2016) Daily eating patterns and their impact on health and disease. Trends Endocrinol Metab 27:69–83. https://doi.org/10.1016/j.tem.2015.11.007 Zhang YKJ, Guo GL, Klaassen CD (2011) Diurnal variations of mouse plasma and hepatic bile acid concentrations as well as expression of biosynthetic enzymes and transporters. PLoS One 6: e16683. https://doi.org/10.1371/journal.pone.0016683 Zhang SL, Bai L, Goel N et al (2017) Human and rat gut microbiome composition is maintained following sleep restriction. PNAS 114:E1564–E1571. https://doi.org/10.1073/pnas. 1620673114

Disease Implications of the Circadian Clocks and Microbiota Interface Laura Tran, Christopher B. Forsyth, Faraz Bishehsari, Robin M. Voigt, Ali Keshavarzian, and Garth R. Swanson

Abstract Circadian rhythms are closely tied to and regulate a variety of host physiologic functions (e.g., sleep-wake, immune function, metabolism). Emerging evidence has shown that there is an interplay between host circadian rhythms and the gut microbiota. Host circadian rhythm impacts the structure and function of gut microbial community, while in turn the microbiota regulates the host circadian clock and metabolism. Disruption of host rhythms can have detrimental effects on health and have been implicated in metabolic syndrome, gastrointestinal and metabolic diseases, and cancer. In this chapter, we will cover the importance of the bi-directional relationship between circadian rhythms and the intestinal microbiota on host health and disease.

1 Circadian Rhythms Circadian rhythms are endogenous 24-h oscillations in cellular function that are conserved across multiple kingdoms including bacteria, plants, and animals including mammals underscoring their critical importance. In complex animals, circadian rhythms regulate multiple physiologic functions including various cellular processes, immune function, and energy metabolism (Panda 2016; Panda et al. 2002). On the molecular level, circadian rhythms are the consequence of a transcriptional-translational loop that takes approximately 24 h to complete called the molecular clock. The core molecular clock is comprised of many proteins including Clock (CLK), Brain and Muscle ARNT-Like 1 (BMAL1), Period (PER), and Cryptochrome (CRY). CLK and BMAL initiate transcription of what are known as clock-controlled genes including Per and Cry. Once PER and CRY have been produced in sufficient quantity, end product inhibition occurs in which PER/CRY prevent further CLOCK/BMAL-mediated transcription (Van Der Horst

L. Tran · C. B. Forsyth · F. Bishehsari · R. M. Voigt · A. Keshavarzian (*) · G. R. Swanson Rush Center for Integrated Microbiome and Chronobiology Research, Rush University Medical Center, Chicago, IL, USA e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_17

329

330

L. Tran et al.

et al. 1999). This loop is then fine-tuned by additional factors such as Rev-erb, SIRT, and others (Bass and Takahashi 2010; Grimaldi et al. 2009). This loop takes approximately 24 h to complete, and this cyclic control is thought to contribute to the nearly 30% of transcription that exhibit circadian oscillations (Konturek et al. 2011). This molecular clock is present in nearly every cell in the body. There is a hierarchy to circadian rhythms in the body. Circadian rhythms are driven by the “central clock” found in the suprachiasmatic nucleus (SCN) within the hypothalamus of the brain which regulates “peripheral clocks” that reside in various cells and tissues throughout the body including the heart (Young et al. 2014), lungs (Sukumaran et al. 2011), kidney (Yoo et al., 2004), liver (Kornmann et al. 2007), and the gastrointestinal (GI) tract (Hoogerwerf et al. 2007). Circadian rhythms can be entrained, or synchronized, to various environmental cues which are known as zeitgebers (i.e., “time givers”). One such zeitgeber is light which robustly entrains central circadian rhythms in the SCN which then regulates the subordinate peripheral rhythms. In the absence of the SCN, periphery rhythms gradually became desynchronized from one another (Yoo et al. 2004). However, signals other than those from the SNC can also entrain the peripheral clocks. For example, time of eating (e.g., exposure to nutrients) can robustly entrain rhythms in the intestine and liver (Mendoza 2007). Circadian rhythms exist to align and optimize biological functions with regular and predictable oscillations in the 24 h day. For example, our Neanderthal ancestors slept at night when there is no sunlight to permit daily activities. Our body’s circadian rhythm cues us when it is time to sleep and time to wake up. Thus, these physiological functions are very different when you are awake and asleep because most people do not eat when they are sleeping. In addition, there are subtle nuances in circadian rhythms that appear to be biologically meaningful. Based on the 24 h solar day, human circadian rhythms are approximately 24 h in duration. However, some individuals have rhythms that are slightly shorter or longer than this average. This variation in clock speed results in individual variation in the preferred timing of rest and activity, known as the chronotype. Circadian rhythms guide nearly every aspect of biology, physiology, and behavior, and it should not be surprising that altering optimal function of these rhythms can have negative effects on health. In this chapter, we will mechanistically discuss how perturbations in circadian rhythms and chronotype may contribute to disease and the important role of the intestinal microbiota in mediating metabolic dysfunction and cancer.

2 Circadian Disruption Circadian rhythm disruption occurs when there is a mismatch between the internal (central and peripheral) circadian clocks and external environmental cues that results in negative health consequences. A few aspects of the modern “Western” lifestyle that commonly disrupt circadian rhythms include shift work (Schernhammer et al. 2003), increased night time light exposure (e.g., use of light emitting devices while

Disease Implications of the Circadian Clocks and Microbiota Interface

331

in bed, bright street lights) (Fonken et al. 2010), social jet lag (different sleep patterns on work days and free days) (Roenneberg et al. 2012), irregular eating patterns (e.g., eating at different times each day, eating late at night, eating large meals close to biological rest time) (Asher and Sassone-Corsi 2015; Bishehsari et al. 2020), and diet composition (Leone et al. 2015; Zarrinpar et al. 2014). While all of these lifestyle features contribute to circadian misalignment, the most impactful and dramatic effects are the consequence of shift work. The prevalence of shift work has steadily increased over the last decade. According to the U.S. Bureau of Labor Statistics, in 2017–2018, 16% of the work force works a non-daytime schedule (e.g., 6% work evenings, 4% work night shifts, and the remaining 6% work a rotating shift, split shift, or an irregular schedule). This means that as many as 52 million people in the United States may suffer from chronic circadian misalignment. Disruption of circadian rhythms is associated with a broad range of negative health consequences (Fig. 1). For example, a number of studies show that circadian rhythm disruption places individuals at a higher risk of developing cancer (e.g., breast, colorectal) (Hansen and Lassen 2012; Innominato et al. 2014), cardiovascular disease (Ruger and Scheer 2009), gastrointestinal disease (Konturek et al. 2011), and metabolic disease (e.g., diabetes, obesity) (Brown 2016). Of these, metabolic syndrome (a cluster of conditions including abdominal obesity, hypercholesterolemia, high blood pressure) is frequently observed in shift workers including high levels of triglycerides, low levels of high-density lipoprotein (HDL)-cholesterol (Karlsson et al. 2001), high resting heart rate, and high blood pressure (Roeser et al. 2012) compared to non-shift workers. Mechanistically, this may be because circadian rhythms regulate various hormones, such as ghrelin and leptin, that are crucial in controlling appetite and modulating risk of obesity (Markwald and Wright 2012). Chronotype, i.e., the entrained phase of the circadian rhythm, may be predictive of the ability to tolerate and adapt to shift work. A majority of individuals (60%) fall in the intermediate of the chronotype spectrum between the extremes of early risers (morning/early chronotype) to late night owls (evening/late chronotype). It is generally well accepted that circadian disruption negatively impacts health, but chronotype may subtly alter the types and consequences of circadian disruption. Individuals with a later chronotype have negative health consequences following circadian disruption including features of metabolic syndrome including poor glycemic control in those with type 2 diabetes (Reutrakul et al. 2013; Vetter et al. 2015) and cancer (Hansen and Lassen 2012; Papantoniou et al. 2015). Those individuals with an early chronotype are also negatively impacted by circadian disruption. For example, night shift workers with an early chronotype suffer from a higher risk of cancer than their non-night shift worker counterparts (Dickerman et al. 2016). Interestingly, individuals with a morning chronotype who perform night shift may actually experience greater circadian misalignment than their late chronotype counterparts because of a mismatch between their chronotype and work schedule as they will revert back to an early schedule during non-work days (Juda et al. 2013; Wittmann et al. 2006). Therefore, a mismatch between one’s chronotype and work schedule on either extreme can have detrimental health effects because the individual’s preferences for sleep and activity are at odds with their schedule on work and

Fig. 1 Central and peripheral circadian clocks. Light-dark cycles regulate the central circadian clock located in the suprachiasmatic nucleus (SCN) within the hypothalamus. Time of eating (i.e., timing of nutrient availability) regulates the peripheral circadian clocks/peripheral oscillator in the intestine and liver. Alterations in light-dark cycles or time of eating can disrupt central and peripheral circadian rhythms, respectively. Intestinal microbiota are also impacted by circadian disruption, and this bi-directional relationship subsequently influences host health and disease

332 L. Tran et al.

Disease Implications of the Circadian Clocks and Microbiota Interface

333

free days. However, there are differing associations with chronotype and cancer risk emphasizing the complexity of this association. It is important to note that chronotype is not always reported to be associated with cancer risk (Ramin et al. 2013), emphasizing the need for more research on this topic with a detailed assessment of chronotype, work schedule, and social schedule. Because circadian misalignment is so prevalent in modern Western societies, it is of high value to understand the mechanisms contributing to disease and techniques. Studying circadian rhythms in humans is challenging due to differences in genetics and environment that may influence results; these clinical studies have been supplemented by studies using animal models including both genetic and environmental manipulations. Multiple approaches can be used to study the impact of circadian disruption and/or circadian misalignment on metabolism and inflammation. Genetic disruption directly influences the function of the molecular clock, and environmental approaches mimic conditions that may be experienced by humans. Both approaches show that manipulating circadian rhythms can negatively impact health. Genetic manipulation of the circadian clock molecular machinery is widely used to study how the molecular clock influences biology and physiology. Studies have demonstrated that altering the function of the molecular clock has a negative impact on metabolic function and promotes inflammation-mediated disease. For example, mice that have a mutation in the core molecular circadian Clock gene (homozygous Clock mutant mice) display a greatly attenuated diurnal feeding rhythm and demonstrate features of metabolic syndrome. Specifically, Clock mutant mice exhibit pathologies related to dysregulated metabolism, such as obesity, hyperleptinemia, hyperlipidemia, hepatic steatosis, hyperglycemia, and hypoinsulinemia (Turek et al. 2005). Similarly, the ablation of other components of the molecular circadian clock has demonstrated the interconnectedness between the circadian clock and metabolism (e.g., obesity and diabetes). Bmal1 knockout mice display dysregulated glucose homeostasis characterized by decreased gluconeogenesis and an increased insulin sensitivity (Marcheva et al. 2010), while Rev-erbα and Rev-erbβ double-knockout mice display deregulated lipid metabolism (Cho et al. 2012). Environmental factors such as sleep disturbance, timing of food consumption, and diet composition can be manipulated in the lab to influence central and peripheral circadian rhythms. Chronic sleep disruption or sleep fragmentation (e.g., sleep apnea in humans) can have a significant impact on body mass and insulin sensitivity. In one study, a month of sleep fragmentation increased food intake, visceral fat mass, systemic insulin resistance, and altered the intestinal microbial community in mice (Poroyko et al. 2016). Timing of food consumption (e.g., eating during the “wrong” time of day) can also affect metabolism. Studies have shown that feeding at the “wrong” circadian time can lead to weight gain. Mice fed a high-fat diet during their biological rest time (i.e., the light phase) gained 2.5 times more weight than those eating during their normal nocturnal hours (Arble et al. 2009). In addition, diet composition can alter diurnal oscillations of gut microbiota and microbial metabolites in mice leading to dysregulated metabolism. Mice consuming a high-fat diet had disrupted circadian rhythms and, consequently, developed high-fat induced obesity

334

L. Tran et al.

(Leone et al. 2015). These are just a few examples of the compelling data demonstrating the critical link between circadian homeostasis and optimal metabolic function. Both genetic and environmental models used to study the impact of circadian disruption on host biological functions are associated with the structure and function of the gut microbiota. Taken together, there is ample evidence that the Western lifestyle promotes circadian misalignment causing adverse health consequences.

3 Implications of Circadian Disruption and the Microbiota There is compelling evidence that disruption of circadian homeostasis promotes metabolic dysfunction and cancer. Mechanistically, this may occur simply by circadian rhythm disruption changing gene expression (e.g., change in the function of immune cells which are robustly regulated by circadian rhythms), but another possibility that is gaining steam involves changes in the intestinal microbiome. Several studies by our group and others have demonstrated that disruption of circadian rhythms changes the intestinal microbiome to increase the relative abundance of pro-inflammatory intestinal bacterial and decrease anti-inflammatory intestinal bacterial abundance. This condition called microbial dysbiosis. It is plausible that this is the mechanism by which circadian disruption promotes cancer and metabolic dysfunction by changing the microbiota composition because altered microbiota composition (dysbiosis) is one of the key mechanisms regulating inflammation and metabolism (Bishehsari and Keshavarzian 2019). The potential role of the intestinal microbiota in promoting diseases and disorders associated with circadian disruption will be discussed further in the next sections.

3.1

Intestinal Microbiota

The human gastrointestinal tract is a dynamic environment that is home to microbial communities including bacteria, archaea, fungi, and viruses, among others. Traditional culture-based methods have not been wholly successful in identifying the vast number of species within the microbiota, only encapsulating less than 30% of the microbiota present (Schwabe and Jobin 2013). In recent years, efforts to characterize the bacterial microbiota have been performed using next-generation sequencing with computational analysis of 16S ribosomal RNA sequencing (Vivarelli et al. 2019) to identify the microbes taxonomically and whole-genome shotgun (WGS) analysis of body-site specific microbial community DNA through amplified bacterial nucleic acid extracted from stool or biopsies (Kho and Lal 2018). This has greatly improved our understanding of the composition of microbial communities across the human population and the functional roles the microbiota play. Thus, most research to date has focused on the bacterial component which is altered by host factors such as

Disease Implications of the Circadian Clocks and Microbiota Interface

335

circadian rhythms and sleep as well as diet and time of food consumption (Murakami et al. 2016). The dominant bacterial phyla in the human intestine are Firmicutes, Bacteroidetes, Actinobacteria, Proteobacteria, Fusobacteria, and Verrucomicrobia (Arumugam et al. 2011). The microbiota influence host cellular function and can impact host health and disease. Alterations in microbiota communities are associated with numerous diseases including obesity, metabolic syndrome, and inflammation-mediated diseases such as cancer (just to name a few). For example, a higher Firmicutes to Bacteroidetes ratio is associated with obesity in both animal and human studies (Ley et al. 2005, 2006). Specifically, elevated levels of Firmicutes such as Ruminococcaceae and depleted levels of Bacteroidetes such as Bacteroidaceae and Bacteroides are observed in obese children (Riva et al. 2017). These changes may be the consequence of diseases or may have contributed to promoting the development or progression of various pathologies and diseases. Thus, it is possible that conditions that promote a pro-inflammatory intestinal microbiome may contribute to the development of diseases and conditions such as metabolic syndrome and cancer (discussed below). The microbiota influences the physiology of the host. The bacteria in the intestine are very metabolically active (O’Hara and Shanahan 2006), and this is one way that the bacteria can impact the host. For example, bacteria ferment dietary fiber resulting in the production of short-chain fatty acids (SCFA) (Venegas et al. 2019). The SCFAs (especially acetate, propionate, and butyrate) play multiple roles including improving intestinal barrier function and influence immune function with the net result of reducing intestinal and systemic inflammation (Arun et al. 2019). The ratio of Firmicutes (which produce butyrate and are generally are considered to be antiinflammatory) to Bacteroidetes (which produce acetate/propionate and generally are considered to be pro-inflammatory) can be considered an indirect estimate of gut microbial health (Fig. 2). The intestinal microbiota exhibits daily diurnal oscillations in both community structure and metabolic function (Liang et al. 2015). There is evidence that these bacterial oscillations depend on host circadian rhythms both directly and indirectly through circadian regulation of the timing of food intake. For instance, mice that have ad libitum access to a high-fat diet lack rhythmic oscillation in the relative abundance of Bacteroidetes and Firmicutes. However, restricting food consumption to only the active phase in mice can restore diurnal oscillations. Firmicutes abundance peaks during times of food consumption and decreases during fasting, while Bacteroidetes and Verrucomicrobia peak during fasting (Zarrinpar et al. 2014). The diurnal oscillations in the bacterial community can reciprocally influence host daily oscillations including immune function (Rosselot et al. 2016) and gene transcription (Mu et al. 2016; Thaiss et al. 2016). Disruption of circadian rhythms and circadian misalignment promote pro-inflammatory shifts in the intestinal microbiota that are sufficient to promote host dysfunction. A study by our group shows that circadian misalignment through repeated phase shifting of light/dark cycles alters the gut microbiota in mice that were fed a high-fat, high-sugar diet. Specifically, circadian disruption decreases the abundance of Bacteroidetes and increases Firmicutes (Voigt et al. 2014). These

336

L. Tran et al.

Fig. 2 The phyla Firmicutes and Bacteroidetes represent the majority of gut microbiota; therefore, the optimal proportion of Firmicutes and Bacteroides is important in maintaining host health. Low-diversity dysbiosis may be driven by many factors that differ by disease context

changes in the microbiota appear to be biologically relevant. Transplantation of stool from circadian-disrupted mice into germ-free mice leads to significant metabolic dysfunction in the recipient including increasing body adiposity (Thaiss et al. 2014). Beyond the role of the central clock, altering circadian rhythms in the periphery also impacts the intestinal microbiota. Disruption of circadian rhythms in the intestine and liver via altering time of food availability also influences the intestinal microbiota. Alterations in the microbiota by wrong-time eating can be characterized as a pro-inflammatory milieu including reduced abundance of bacteria that produce putative beneficial products such as anti-inflammatory SCFA (Bishehsari et al. 2020) and an increase in pathobionts. Mice with homogenous ClockΔ19/Δ19 mutations also exhibited intestinal dysbiosis, with a lower taxonomic diversity compared to wildtype counterparts fed an alcohol-containing diet (Voigt et al. 2016). Another study subjected mice to an abnormal light/dark cycles (weekly 12 hour phase shifts of the light/dark cycles instead of a constant 12:12 light/dark cycle) and observed similar trends in dysfunction of the intestinal barrier (Summa et al. 2013). Subsequent changes in the structure and function of the gut microbiota included an increase in Ruminococcus torques, a bacterial species known to decrease gut barrier integrity, and a decrease in Lactobacillus johnsonii, a bacterium that helps maintain the intestinal epithelial cell layer (Liu et al. 2015) after circadian rhythm disruption (Deaver et al. 2018). Our currently unpublished data suggest that wrong-time eating of a non-obesogenic diet causes gut dysbiosis, characterized by decreased SCFAproducing bacteria, leading to increased gut leakiness, and predisposition to weight gain. Studies have observed changes in gene expression for a number of pathways as genes involved in pathways promoting host-beneficial immune responses, cellular

Disease Implications of the Circadian Clocks and Microbiota Interface

337

processes, and energy metabolism were downregulated, while genes involved in the synthesis and transportation of LPS were upregulated in mice with disrupted circadian cycles (Deaver et al. 2018; Voigt et al. 2016). Emerging evidence supports that gut microbiota also impact circadian clock genes of the host, suggesting a complex bi-directional relationship between the microbiota and the host. Both germ-free mice (lack a microbiome) and antibiotictreated mice (reduced intestinal bacterial biomass) have disrupted circadian clock gene expression including Bmal1, Cry1, Per1, and Per2 in intestinal epithelial cells and hepatocytes (Leone et al. 2015; Mukherji et al. 2013). High-fat diets drive intestinal dysbiosis that mediate peroxisome proliferative activated receptor-γ (PPARγ) to affect hepatic diurnal rhythmicity and induced an increase in fat depots (Murakami et al. 2016). Depletion of the intestinal microbiota (antibiotic-treated mice) reversed the effects of PPARγ-driven reprogramming of the host liver clock and reverted the fat depot phenotype. In addition, the intestinal microbiota is known to impact energy storage and adiposity, and one study found that body composition is regulated through epithelial cell circadian transcription factor NFIL3 (Wang et al. 2017). These findings provide insight into how the circadian clock and the gut microbiota exist within a bi-directional relationship that impacts host health and disease.

3.2

Metabolic Syndrome and Dietary Impact

Host metabolism is tightly interlocked with circadian rhythm and food intake, such as diet and food timing, which affect peripheral circadian clocks in the liver and GI tract. Diet and time of eating, via changes in nutrient availability (David et al. 2014) or peripheral circadian clock (Longo and Panda 2016; Voigt et al. 2013), regulate a variety of cellular functions (Kaczmarek et al. 2017). Energy harvest, cell growth, and DNA repair are predominant during the periods of nutrient availability, while detoxification is predominant during periods of fasting (Thaiss et al. 2014). These host-associated cues could impact gut microbial composition as well. For instance, mice that have ad libitum access to a high-fat diet lack rhythmic oscillation in the relative abundance of Bacteroidetes and Firmicutes. However, feeding restricted to the active dark phase in mice can restore cyclical rhythms. Firmicutes peaked during feeding and decreased during daytime fasting, while Bacteroidetes and Verrucomicrobia peaked during daytime fasting (Zarrinpar et al. 2014). Similar rhythmicity was observed in male and female ad libitum-fed mice on 12/12 light/ dark cycles, with more pronounced rhythmicity in female mice (Liang et al. 2015). However, dietary changes can alter the gut microbiota and circadian signaling. Diets that are high fiber-low fat are known to confer health benefits, while diets that are more “Westernized” (low fiber-high fat) induce microbial changes that are distinct and negatively impact health (Parks et al. 2013). Diet changes the microbiota; thus it could be that these diet-induced microbiota changes alter the host circadian rhythm leading to changes in food intake. Mice fed a high-fat diet

338

L. Tran et al.

consume more food during the light (inactive, rest) period than their non-high-fat diet fed counterparts (Kohsaka et al. 2007). Another study has shown that circadian misalignment through repeated phase shifting only altered the gut microbiota in mice that were fed a high-fat, high-sugar diet but not in the mice that were fed the regular chow diet. A high-fat, high-sugar diet acts as a dietary stressor and combined with circadian misalignment led to significantly altered gut microbiota, shifting with a decrease in abundance of Bacteroidetes and an increase in Firmicutes (Voigt et al. 2014). Interestingly, it is possible to transfer the effects of circadian disrupted microbiota. Germ-free mice fail to gain weight on a high-fat diet (Leone et al. 2015); however, transplantation of circadian shifted microbiota into germ-free mice led to significant metabolic disorder by increasing body adiposity (Thaiss et al. 2014). Emerging evidence supports that gut microbiota also impacts circadian clock genes in a bi-directional relationship.

3.3

Gastrointestinal (GI) Tract

Cells in the GI tract express circadian clock genes (Pardini et al. 2005), and GI function demonstrates clear diurnal variations including gastric emptying, colonic motility, epithelial cell proliferation (Marra et al. 1994), mucosal permeability (Forsyth et al. 2015), and gut immunity (Bando and Colonna 2016). For example, gastric emptying rates have significantly longer half-times in the evening compared to the morning (Goo et al. 1987), and colonic motility is minimal prior to meals and during sleep (Narducci et al. 1987). Circadian misalignment disrupts these functions and renders the host susceptible to diseases including those in the GI tract. For example, circadian disruption increases intestinal permeability (i.e., “gut leakiness”) as a function of impaired intestinal barrier function (Forsyth et al. 2015; Maury et al. 2010; Summa et al. 2013). The barrier restricts pro-inflammatory luminal contents (e.g., lipopolysaccharide, LPS) to the intestine and is critical to prevent inflammation that can occur if the barrier is not intact. Tight junction proteins, such as occludin and claudin-1, maintain the integrity of the intestinal barrier against bacterial products and decreased expression of these tight junction proteins (or inappropriate localization) allow for the translocation of microbes and microbial products to reach the intestinal mucosa and the systemic circulation. Tight junction proteins are under circadian control (Oh-oka et al. 2014) and indeed diurnal fluctuations can be observed in intestinal barrier integrity (Voigt et al. 2018). Genetic disruption of mice with homogenous ClockΔ19/Δ19 mutations and environmental disruption through chronic phase shifts of the light/dark cycle have been shown to significantly increase intestinal permeability. In the presence of an external stressor such as chronic alcohol exposure (Keshavarzian et al. 2009), both methods of circadian disruption displayed gut leakiness and promoted alcohol-induced intestinal permeability (Summa et al. 2013).

Disease Implications of the Circadian Clocks and Microbiota Interface

339

Toll-like receptors (TLRs) are a part of the innate immune system and recognize a wide variety of molecules derived from the microbiota and are expressed by intestinal epithelial cells and mucosal-associated immune cells, both important in maintaining intestinal barrier function (Chassaing et al. 2014; Shibolet and Podolsky 2007). Therefore, it is not surprising that TLR signaling in the gut is involved in both maintaining intestinal homeostasis and the induction of inflammatory responses. For instance, the circadian rhythm of TLR expression regulates the intestinal microbiome and confers homeostasis (Mukherji et al. 2013). Members of the TLR family play a role in the regulation of intestinal immunity. TLR2 dimerizes with TLR1 or TLR6 to recognize bacterial cell wall components such as lipoproteins. TLR4 recognizes lipopolysaccharides (LPS). LPS is a component of Gram-negative bacterial cell walls that promotes inflammation and cause tissue injury, contributing to dysfunction of the intestinal barrier. Thus, increased abundance of pro-inflammatory bacteria in the intestine coupled with intestinal barrier dysfunction resulting from microbiota circadian disruption can promote inflammation-mediated diseases.

3.4

Cancer

Disrupted circadian rhythms are associated with cancer development, exacerbation of disease progression, and predictors of poor survival in cancer patients. One of the many cellular processes regulated by circadian rhythms is the cell cycle (Feillet et al. 2015). Dysfunction of the cell cycle can have numerous detrimental effects including unchecked cell proliferation and DNA damage leading to the formation of tumors. In a chronic jet lag model, repeated 8 h advances of the light/dark cycle in mice with Glasgow osteosarcomas have increased tumor growth compared to noncircadian-disrupted controls (Filipski et al. 2004). Similarly, disruption of the central circadian clock in tumor-bearing mice leads to accelerated tumor growth (Filipski et al. 2006). Several studies have shown that both genes regulating the cell cycle and circadian clock genes are dysregulated in cancer cells (Soták et al. 2014). For instance, disrupted expression of the circadian clock gene Per2 results in significant downstream effects on DNA damage responses, such as cell cycle and apoptotic targets, suggesting that Per2 has a tumor suppressive role (Mitchell and Engelbrecht 2015). As will be discussed in more detail, recent studies highlight the possible role of microbiota in mediating the effect of circadian disruption on colon carcinogenesis. For example, circadian disruption, either by shift in light/dark cycle or wrong-time eating (i.e., during the rest period), can worsen colonic inflammation induced by alcohol consumption and exacerbate colon polyposis and carcinoma development, at least in part by negatively impacting the intestinal microbiota (Bishehsari et al. 2016, 2020).

340

3.4.1

L. Tran et al.

Breast Cancer

Circadian clock gene expression is present within the breast epithelium and other peripheral tissues (Yoo et al. 2004). Indeed, expression of the Per gene correlates with genes implicated in breast cancer such as the estrogen receptor (Blakeman et al. 2016). Additionally, melatonin (produced by the pineal gland) is required for the circadian regulation of sleep and may represent a unique risk factor for breast cancer. For instance, studies have shown that melatonin inhibits the development and growth of breast cancer, and removal of the pineal gland in rodents or exposure to constant light (which suppresses melatonin production) stimulates tumorigenesis (Blask et al. 2005a, b). Similarly, increased exposure to light at night promotes breast cancer growth, which may have to do with altered tumor metabolism. Aerobic glycolysis (characteristic in cancer cells and termed the “Warburg effect”) demonstrates a daily rhythmic pattern. Glucose uptake and lactate production were highly expressed during the daytime, corresponding with an increase in tumor cell proliferation and survival. Expression of these processes was repressed during the dark phase (Blask et al. 2014). Circadian-disrupted conditions, such as light at night, promotes the continuation of stimulatory tumor growth processes observed in the daytime. In human studies, night shift workers have an increased risk of developing breast cancer (Hansen and Lassen 2012), especially with prolonged exposure to night shift work (e.g., 20 years or shorter periods with consecutive shifts) (Hansen 2017). Almost 80% of all breast cancers are estrogen receptor-positive. We already mentioned that the estrogen receptors are under circadian regulation, but the gut microbiota also has a prominent role in regulating steroid hormone metabolism (i.e., estrogen). Therefore, perturbation of the microbiota (via circadian misalignment) would directly affect systemic levels of estrogen and estrogen metabolites and may influence the pathogenesis of breast cancer. In addition, bacterial β-glucuronidase may be important. Bacterial β-glucuronidase de-conjugates estrogen, which consequently allows free, active estrogen to be reabsorbed. A healthy microbiome produces enough β-glucuronidase to maintain estrogen homeostasis. However, breast tumors and breast cancer cell lines express high levels of β-glucuronidases (Raftogianis et al. 2000). β-glucuronidase activity is altered when intestinal dysbiosis occurs. Increased abundance of several β-glucuronidase bacteria in the Clostridia and Ruminococcaceae families may drive estrogen metabolism (Rea et al. 2018) and increase the risk of estrogen-receptor positive breast cancer. High fecal microbial diversity has been associated with a high ratio of estrogen metabolites to parent estrogen (Fuhrman et al. 2014). Higher ratios of hydroxylated metabolites to parent estrogens have been associated with a reduced risk of postmenopausal breast cancer, suggesting that postmenopausal breast cancer risk may be reduced for women who have high intestinal microbial diversity. Additionally, it was shown that a decrease in stool bacterial diversity leads to estrogen excretion and an elevated risk of breast cancer. As circadian misalignment can alter microbial communities and function, it is plausible that circadian-derived dysbiosis can impact breast cancer risk.

Disease Implications of the Circadian Clocks and Microbiota Interface

3.4.2

341

Colorectal Cancer

Rotating night shifts are associated with an increased risk of colorectal cancer, with higher risk in workers who have been working night shifts for longer periods of time (Schernhammer et al. 2003). In addition to this central mechanism, wrong time eating, characterized by eating during the inactive rest period, causes circadian disruption in mice and subsequently promotes pro-tumorigenic intestinal mucosal inflammation. Wrong-time eating promotes increased alcohol-induced burden of polyps, predisposing circadian disrupted mice to alcohol-induced colon cancer (Bishehsari et al. 2020). Mechanistically, this could be due to direct effects of circadian disruption in intestinal epithelial cells but could also be due to circadian disruption effects on the intestinal microbiota. Microbial dysbiosis can mediate colon carcinogenesis by decreasing the abundance of butyrate-producing bacteria. Indeed, a high-fiber treatment to increase the abundance of butyrate-producing bacteria reduces colon tumorigenesis (Bishehsari et al. 2018). In humans, circadian misalignment in night shift workers causes increased susceptibility to the injurious effects of other insults such as alcohol-induced intestinal barrier dysfunction compared to day shift workers (Swanson et al. 2016). Alcohol is associated with tissue injury and causes alterations in the gut microbiota composition and function, such as decreased abundance in Bacteroidetes and increased abundance of Proteobacteria (Mutlu et al. 2012, 2009). These changes, mediated in part by gut-derived plasma SCFAs, were found to have a diurnal rhythm that is impacted by the central clock of the host. Moderate alcohol consumption suppresses plasma SCFAs, which was associated with increased colonic permeability (Swanson et al. 2020). Impaired intestinal barrier function allows for bacterial translocation across the intestinal mucosa, initiating inflammation. Indeed, inflammation, inflammasome activation (via TLR) (Elinav et al. 2013), and activation of the NF-κB pathway resulting in pro-inflammatory cytokine production (e.g., IL-6) (Karin and Greten 2005) are implicated in colorectal carcinogenesis. Thus, the microbiota (and its antigens) are important drivers of immune responses that can trigger carcinogenesis. This is demonstrated in germ-free mouse models of intestinal tumorigenesis as germ-free mice had reduced tumor load compared to conventionally raised mice (Vannucci et al. 2009). The final mechanism that we will discuss is melatonin production. It was already mentioned that circadian misalignment is associated with low levels of melatonin production. Clinical studies report low melatonin levels in colorectal cancer patients compared to healthy control subjects (Vician et al. 1999). Melatonin positively correlates with better intestinal barrier function (Swanson et al. 2015); therefore, reduced melatonin as a result of circadian dysregulation may be a mechanism promoting colorectal cancer. Current studies have also suggested bi-directional associations between the gut microbiota and melatonin concentrations that potentially influence the circadian-metabolic axis (Fig. 3) (Hardeland et al. 2011; Paulose et al. 2016). Indeed, the interplay between the biological clock and gut microbiota has a crucial role in maintaining host homeostasis.

Fig. 3 (a) Melatonin is secreted by the pineal gland. When darkness is detected by the retina, the pineal gland is stimulated to produce melatonin. Melatonin production is inhibited when light is detected. The melatonin binds to melatonin receptors and activates Gi and Gq proteins that in turn inhibit the adenylate cyclase/ cAMP pathway and activate the phospholipase C/IP3 pathway, respectively. The downstream effectors, including protein kinases, CREB, and MSK1, regulate the expression of Clock genes and thus maintain the circadian rhythm. (b) Melatonin accumulates in the gastrointestinal (GI) tract independent of production from the pineal gland. This melatonin is produced from dietary protein in large amounts within the intestine from 5HT (serotonin) and can enter the bloodstream. Once in the bloodstream, melatonin acts as an endocrine hormone and controls biological functions with circadian rhythms (e.g., the sleep-wake cycle)

342 L. Tran et al.

Disease Implications of the Circadian Clocks and Microbiota Interface

343

4 Conclusion Circadian rhythms orchestrate numerous physiologic functions to maintain homeostasis. External factors such as light and time of food consumption and other factors can impact circadian rhythmicity, which in turn leads to altered microbial community structure and function. It is well established that circadian rhythms of the host influence diurnal oscillations in microbiota community structure and function, and recently emerging studies also demonstrate that the microbiota can significantly impact the circadian clock of the host, thus establishing that a bi-directional relationship is critical to maintain host health. Thus, it is critical to further elucidate the various mechanisms through which the microbiota impact host metabolism and circadian rhythmicity. Understanding these relationships has the potential to develop microbial-derived therapies to ameliorate the effects of disrupted circadian pathologies. Acknowledgments We would like to thank Mr. and Mrs. Larry Field, Mr. and Mrs. Glass, Mr. Keehn, and the Alvin Baum Family fund for their philanthropic funding. Funding A.K. is supported by the National Institute on Alcohol Abuse and Alcoholism of the National Institutes of Health in part by grant R24AA026801 (A. Keshavarzian) and R01AA023417 (A. Keshavarzian, K. Khazaie). F.B. is supported by NIH grant AA025387, Brinson foundation, and Rush Translational Sciences Consortium/Swim Across America Organization grant.

References Arble DM, Bass J, Laposky AD, Vitaterna MH, Turek FW (2009) Circadian timing of food intake contributes to weight gain. Obesity 17(11):2100–2102. https://doi.org/10.1038/oby.2009.264 Arumugam M, Raes J, Pelletier E, Le Paslier D, Yamada T, Mende DR et al (2011) Enterotypes of the human gut microbiome. Nature 473(7346):174–180. https://doi.org/10.1038/nature09944 Arun KB, Madhavan A, Reshmitha TR, Thomas S, Nisha P (2019) Short chain fatty acids enriched fermentation metabolites of soluble dietary fibre from Musa paradisiaca drives HT29 colon cancer cells to apoptosis. PLoS One 14(5):e0216604. https://doi.org/10.1371/journal.pone. 0216604 Asher G, Sassone-Corsi P (2015, March 26) Time for food: the intimate interplay between nutrition, metabolism, and the circadian clock. Cell 161:84–92. https://doi.org/10.1016/j.cell.2015.03.015 Bando JK, Colonna M (2016, June 21) Innate lymphoid cell function in the context of adaptive immunity. Nat Immunol 17:783–789. https://doi.org/10.1038/ni.3484 Bass J, Takahashi JS (2010) Circadian integration of metabolism and energetics. Science (New York, N.Y.) 330(6009):1349–1354. https://doi.org/10.1126/science.1195027 Bishehsari F, Keshavarzian A (2019, September 27) Microbes help to track time. Science 365:1379–1380. https://doi.org/10.1126/science.aaz0224 Bishehsari F, Saadalla A, Khazaie K, Engen PA, Voigt RM, Shetuni BB et al (2016) Light/dark shifting promotes alcohol-induced colon carcinogenesis: possible role of intestinal inflammatory milieu and microbiota. Int J Mol Sci 17(12):2017. https://doi.org/10.3390/ijms17122017 Bishehsari F, Engen P, Preite N, Tuncil Y, Naqib A, Shaikh M et al (2018) Dietary Fiber treatment corrects the composition of gut microbiota, promotes SCFA production, and suppresses Colon carcinogenesis. Genes 9(2):102. https://doi.org/10.3390/genes9020102

344

L. Tran et al.

Bishehsari F, Engen PA, Voigt RM, Swanson G, Shaikh M, Wilber S et al (2020) Abnormal eating patterns cause circadian disruption and promote alcohol-associated colon carcinogenesis. CMGH 9(2):219–237. https://doi.org/10.1016/j.jcmgh.2019.10.011 Blakeman V, Williams JL, Meng QJ, Streuli CH (2016, September 2) Circadian clocks and breast cancer. Breast Cancer Res 18:89. https://doi.org/10.1186/s13058-016-0743-z Blask DE, Dauchy RT, Sauer LA (2005a, July) Putting cancer to sleep at night: the neuroendocrine/ circadian melatonin signal. Endocrine 27:179–188. https://doi.org/10.1385/ENDO:27:2:179 Blask D, Sauer L, Dauchy R (2005b) Melatonin as a Chronobiotic / anticancer agent: cellular, biochemical, and molecular mechanisms of action and their implications for circadian-based Cancer therapy. Curr Top Med Chem 2(2):113–132. https://doi.org/10.2174/ 1568026023394407 Blask DE, Dauchy RT, Dauchy EM, Mao L, Hill SM, Greene MW et al (2014) Light exposure at night disrupts host/cancer circadian regulatory dynamics: impact on the Warburg effect, lipid signaling and tumor growth prevention. PLoS One 9(8):e102776. https://doi.org/10.1371/ journal.pone.0102776 Brown SA (2016, June 1) Circadian metabolism: from mechanisms to metabolomics and medicine. Trends Endocrinol Metab 27:415–426. https://doi.org/10.1016/j.tem.2016.03.015 Chassaing B, Kumar M, Baker MT, Singh V, Vijay-Kumar M (2014, September 1) Mammalian gut immunity. Biom J 37:246–258. https://doi.org/10.4103/2319-4170.130922 Cho H, Zhao X, Hatori M, Yu RT, Barish GD, Lam MT et al (2012) Regulation of circadian behaviour and metabolism by REV-ERB-α and REV-ERB-β. Nature 485(7396):123–127. https://doi.org/10.1038/nature11048 David LA, Maurice CF, Carmody RN, Gootenberg DB, Button JE, Wolfe BE et al (2014) Diet rapidly and reproducibly alters the human gut microbiome. Nature 505(7484):559–563. https:// doi.org/10.1038/nature12820 Deaver JA, Eum SY, Toborek M (2018) Circadian disruption changes gut microbiome taxa and functional gene composition. Front Microbiol 9:737. https://doi.org/10.3389/fmicb.2018.00737 Dickerman BA, Markt SC, Koskenvuo M, Hublin C, Pukkala E, Mucci LA, Kaprio J (2016) Sleep disruption, chronotype, shift work, and prostate cancer risk and mortality: a 30-year prospective cohort study of Finnish twins. Cancer Causes Control 27(11):1361–1370. https://doi.org/10. 1007/s10552-016-0815-5 Elinav E, Nowarski R, Thaiss CA, Hu B, Jin C, Flavell RA (2013) Inflammation-induced cancer: crosstalk between tumours, immune cells and microorganisms. Nat Rev Cancer 13:759–771. https://doi.org/10.1038/nrc3611 Feillet C, van der Horst GTJ, Levi F, Rand DA, Delaunay F (2015) Coupling between the circadian clock and cell cycle oscillators: implication for healthy cells and malignant growth. Front Neurol 6(May):96. https://doi.org/10.3389/fneur.2015.00096 Filipski E, Delaunay F, King VM, Wu MW, Claustrat B, Gréchez-Cassiau A et al (2004) Effects of chronic jet lag on tumor progression in mice. Cancer Res 64(21):7879–7885. https://doi.org/10. 1158/0008-5472.CAN-04-0674 Filipski E, Li XM, Lévi F (2006, May) Disruption of circadian coordination and malignant growth. Cancer Causes Control 17:509–514. https://doi.org/10.1007/s10552-005-9007-4 Fonken LK, Workman JL, Walton JC, Weil ZM, Morris JS, Haim A, Nelson RJ (2010) Light at night increases body mass by shifting the time of food intake. Proc Natl Acad Sci U S A 107 (43):18664–18669. https://doi.org/10.1073/pnas.1008734107 Forsyth CB, Voigt RM, Burgess HJ, Swanson GR, Keshavarzian A (2015, June 1) Circadian rhythms, alcohol and gut interactions. Alcohol 49:389–398. https://doi.org/10.1016/j.alcohol. 2014.07.021 Fuhrman BJ, Feigelson HS, Flores R, Gail MH, Xu X, Ravel J, Goedert JJ (2014) Associations of the fecal microbiome with urinary estrogens and estrogen metabolites in postmenopausal women. J Clin Endocrinol Metab 99(12):4632–4640. https://doi.org/10.1210/jc.2014-2222

Disease Implications of the Circadian Clocks and Microbiota Interface

345

Goo RH, Moore JG, Greenberg E, Alazraki NP (1987) Circadian variation in gastric emptying of meals in humans. Gastroenterology 93(3):515–518. https://doi.org/10.5555/URI: PII:0016508587909139 Grimaldi B, Nakahata Y, Kaluzova M, Masubuchi S, Sassone-Corsi P (2009, January 1) Chromatin remodeling, metabolism and circadian clocks: The interplay of CLOCK and SIRT1. Int J Biochem Cell Biol 41:81–86. https://doi.org/10.1016/j.biocel.2008.08.035 Hansen J (2017, September 1) Night shift work and risk of breast Cancer. Current Environ Health Rep 4:325–339. https://doi.org/10.1007/s40572-017-0155-y Hansen J, Lassen CF (2012) Nested case-control study of night shift work and breast cancer risk among women in the Danish military. Occup Environ Med 69(8):551–556. https://doi.org/10. 1136/oemed-2011-100240 Hardeland R, Cardinali DP, Srinivasan V, Spence DW, Brown GM, Pandi-Perumal SR (2011, March 1) Melatonin-A pleiotropic, orchestrating regulator molecule. Progress Neurobiol 93:350–384. https://doi.org/10.1016/j.pneurobio.2010.12.004 Hoogerwerf WA, Hellmich HL, Cornélissen G, Halberg F, Shahinian VB, Bostwick J et al (2007) Clock gene expression in the murine gastrointestinal tract: endogenous rhythmicity and effects of a feeding regimen. Gastroenterology 133(4):1250–1260. https://doi.org/10.1053/j.gastro. 2007.07.009 Innominato PF, Roche VP, Palesh OG, Ulusakarya A, Spiegel D, Lévi FA (2014) The circadian timing system in clinical oncology. Ann Med 46(4):191–207. https://doi.org/10.3109/ 07853890.2014.916990 Juda M, Vetter C, Roenneberg T (2013) Chronotype modulates sleep duration, sleep quality, and social jet lag in shift-workers. J Biol Rhythm 28(2):141–151. https://doi.org/10.1177/ 0748730412475042 Kaczmarek JL, Musaad SM, Holscher HD (2017) Time of day and eating behaviors are associated with the composition and function of the human gastrointestinal microbiota. Am J Clin Nutr 106 (5):ajcn156380. https://doi.org/10.3945/ajcn.117.156380 Karin M, Greten FR (2005, October) NF-κB: linking inflammation and immunity to cancer development and progression. Nat Rev Immunol 5:749–759. https://doi.org/10.1038/nri1703 Karlsson B, Knutsson A, Lindahl B (2001) Is there an association between shift work and having a metabolic syndrome? Results from a population based study of 27 485 people. Occup Environ Med 58(11):747–752. https://doi.org/10.1136/oem.58.11.747 Keshavarzian A, Farhadi A, Forsyth CB, Rangan J, Jakate S, Shaikh M et al (2009) Evidence that chronic alcohol exposure promotes intestinal oxidative stress, intestinal hyperpermeability and endotoxemia prior to development of alcoholic steatohepatitis in rats. J Hepatol 50(3):538–547. https://doi.org/10.1016/j.jhep.2008.10.028 Kho ZY, Lal SK (2018) The human gut microbiome – a potential controller of wellness and disease. Front Microbiol 9:1835. https://doi.org/10.3389/fmicb.2018.01835 Kohsaka A, Laposky AD, Ramsey KM, Estrada C, Joshu C, Kobayashi Y et al (2007) High-fat diet disrupts behavioral and molecular circadian rhythms in mice. Cell Metab 6(5):414–421. https:// doi.org/10.1016/j.cmet.2007.09.006 Konturek PC, Brzozowski T, Konturek SJ (2011, April) Gut clock: implication of circadian rhythms in the gastrointestinal tract. J Physiol Pharmacol 62:139–150 Kornmann B, Schaad O, Bujard H, Takahashi JS, Schibler U (2007) System-driven and oscillatordependent circadian transcription in mice with a conditionally active liver clock. PLoS Biol 5 (2):0179–0189. https://doi.org/10.1371/journal.pbio.0050034 Leone V, Gibbons SM, Martinez K, Hutchison AL, Huang EY, Cham CM et al (2015) Effects of diurnal variation of gut microbes and high-fat feeding on host circadian clock function and metabolism. Cell Host Microbe 17(5):681–689. https://doi.org/10.1016/j.chom.2015.03.006 Ley RE, Bäckhed F, Turnbaugh P, Lozupone CA, Knight RD, Gordon JI (2005) Obesity alters gut microbial ecology. Proc Natl Acad Sci U S A 102(31):11070–11075. https://doi.org/10.1073/ pnas.0504978102

346

L. Tran et al.

Ley RE, Turnbaugh PJ, Klein S, Gordon JI (2006) Microbial ecology: human gut microbes associated with obesity. Nature 444(7122):1022–1023. https://doi.org/10.1038/4441022a Liang X, Bushman FD, FitzGerald GA (2015) Rhythmicity of the intestinal microbiota is regulated by gender and the host circadian clock. Proc Natl Acad Sci U S A 112(33):10479–10484. https://doi.org/10.1073/pnas.1501305112 Liu HY, Roos S, Jonsson H, Ahl D, Dicksved J, Lindberg JE, Lundh T (2015) Effects of Lactobacillus johnsonii and Lactobacillus reuteri on gut barrier function and heat shock proteins in intestinal porcine epithelial cells. Physiol Rep 3(4):e12355. https://doi.org/10.14814/ phy2.12355 Longo VD, Panda S (2016) Fasting, circadian rhythms, and time-restricted feeding in healthy lifespan. Cell Metabolism 23:1048–1059. https://doi.org/10.1016/j.cmet.2016.06.001 Marcheva B, Ramsey KM, Buhr ED, Kobayashi Y, Su H, Ko CH et al (2010) Disruption of the CLOCK components CLOCK and BMAL1 leads to hypoinsulinaemia and diabetes. Nature 466 (7306):627–631. https://doi.org/10.1038/nature09253 Markwald RR, Wright KP (2012) Circadian misalignment and sleep disruption in shift work: implications for fatigue and risk of weight gain and obesity. In: Sleep loss and obesity: intersecting epidemics. Springer, New York, pp 101–118. https://doi.org/10.1007/978-1-46143492-4_8 Marra G, Anti M, Percesepe A, Armelao F, Ficarelli R, Coco C et al (1994) Circadian variations of epithelial cell proliferation in human rectal crypts. Gastroenterology 106(4):982–987. https:// doi.org/10.1016/0016-5085(94)90757-9 Maury E, Ramsey KM, Bass J (2010, February) Circadian rhythms and metabolic syndrome: from experimental genetics to human disease. Circ Res 106:447–462. https://doi.org/10.1161/ CIRCRESAHA.109.208355 Mendoza J (2007) Circadian clocks: setting time by food. J Neuroendocrinol 19(2):127–137. https://doi.org/10.1111/j.1365-2826.2006.01510.x Mitchell MI, Engelbrecht A-M (2015) Circadian rhythms and breast Cancer: the role of Per2 in doxorubicin-induced cell death. J Toxicol 2015:392360. https://doi.org/10.1155/2015/392360 Mu C, Yang Y, Zhu W (2016March 17) Gut microbiota: the brain peacekeeper. Front Microbiol 7:345. https://doi.org/10.3389/fmicb.2016.00345 Mukherji A, Kobiita A, Ye T, Chambon P (2013) Homeostasis in intestinal epithelium is orchestrated by the circadian clock and microbiota cues transduced by TLRs. Cell 153(4):812–827. https://doi.org/10.1016/j.cell.2013.04.020 Murakami M, Tognini P, Liu Y, Eckel-Mahan KL, Baldi P, Sassone-Corsi P (2016) Gut microbiota directs PPAR γ-driven reprogramming of the liver circadian clock by nutritional challenge. EMBO Rep 17(9):1292–1303. https://doi.org/10.15252/embr.201642463 Mutlu E, Keshavarzian A, Engen P, Forsyth CB, Sikaroodi M, Gillevet P (2009) Intestinal dysbiosis: a possible mechanism of alcohol-induced endotoxemia and alcoholic steatohepatitis in rats. Alcohol Clin Exp Res 33(10):1836–1846. https://doi.org/10.1111/j.1530-0277.2009. 01022.x Mutlu EA, Gillevet PM, Rangwala H, Sikaroodi M, Naqvi A, Engen PA et al (2012) Colonic microbiome is altered in alcoholism. Am J Physiol Gastrointest Liver Physiol 302(9):G966. https://doi.org/10.1152/ajpgi.00380.2011 Narducci F, Bassotti G, Gaburri M, Morelli A (1987) Twenty four hour manometric recording of colonic motor activity in healthy man. Gut 28(1):17–25. https://doi.org/10.1136/gut.28.1.17 O’Hara AM, Shanahan F (2006) The gut flora as a forgotten organ. EMBO Rep 7(7):688–693. https://doi.org/10.1038/sj.embor.7400731 Oh-oka K, Kono H, Ishimaru K, Miyake K, Kubota T, Ogawa H et al (2014) Expressions of tight junction proteins occludin and claudin-1 are under the circadian control in the mouse large intestine: implications in intestinal permeability and susceptibility to colitis. PLoS One 9(5): e98016. https://doi.org/10.1371/journal.pone.0098016 Panda S (2016, November 25) Circadian physiology of metabolism. Science 354:1008–1015. https://doi.org/10.1126/science.aah4967

Disease Implications of the Circadian Clocks and Microbiota Interface

347

Panda S, Hogenesch JB, Kay SA (2002, May 16) Circadian rhythms from flies to human. Nature 417:329–335. https://doi.org/10.1038/417329a Papantoniou K, Castaño-Vinyals G, Espinosa A, Aragonés N, Pérez-Gómez B, Burgos J et al (2015) Night shift work, chronotype and prostate cancer risk in the MCC-Spain case-control study. Int J Cancer 137(5):1147–1157. https://doi.org/10.1002/ijc.29400 Pardini L, Kaeffer B, Trubuil A, Bourreille A, Galmiche JP (2005) Human intestinal circadian clock: expression of clock genes in colonocytes lining the crypt. Chronobiol Int 22(6):951–961. https://doi.org/10.1080/07420520500395011 Parks BW, Nam E, Org E, Kostem E, Norheim F, Hui ST et al (2013) Genetic control of obesity and gut microbiota composition in response to high-fat, high-sucrose diet in mice. Cell Metab 17 (1):141–152. https://doi.org/10.1016/j.cmet.2012.12.007 Paulose JK, Wright JM, Patel AG, Cassone VM (2016) Human gut Bacteria are sensitive to melatonin and express endogenous circadian rhythmicity. PLoS One 11(1):e0146643. https:// doi.org/10.1371/journal.pone.0146643 Poroyko VA, Carreras A, Khalyfa A, Khalyfa AA, Leone V, Peris E et al (2016) Chronic sleep disruption alters gut microbiota, Induces systemic and adipose tissue inflammation and insulin resistance in mice. Sci Rep 6:35405. https://doi.org/10.1038/srep35405 Raftogianis R, Creveling C, Weinshilboum R, Weisz J (2000) Chapter 6: estrogen metabolism by conjugation. JNCI Monographs 2000(27):113–124. https://doi.org/10.1093/oxfordjournals. jncimonographs.a024234 Ramin C, Devore EE, Pierre-Paul J, Duffy JF, Hankinson SE, Schernhammer ES (2013) Chronotype and breast cancer risk in a cohort of US nurses. Chronobiol Int 30(9):1181–1186. https://doi.org/10.3109/07420528.2013.809359 Rea D, Coppola G, Palma G, Barbieri A, Luciano A, Del Prete P et al (2018) Microbiota effects on cancer: from risks to therapies. Oncotarget 9(25):17915–17927. https://doi.org/10.18632/ oncotarget.24681 Reutrakul S, Hood MM, Crowley SJ, Morgan MK, Teodori M, Knutson KL, Van Cauter E (2013) Chronotype is independently associated with glycemic control in type 2 diabetes. Diabetes Care 36(9):2523–2529. https://doi.org/10.2337/dc12-2697 Riva A, Borgo F, Lassandro C, Verduci E, Morace G, Borghi E, Berry D (2017) Pediatric obesity is associated with an altered gut microbiota and discordant shifts in Firmicutes populations. Environ Microbiol 19(1):95–105. https://doi.org/10.1111/1462-2920.13463 Roenneberg T, Allebrandt KV, Merrow M, Vetter C (2012) Social jetlag and obesity. Curr Biol 22 (10):939–943. https://doi.org/10.1016/j.cub.2012.03.038 Roeser K, Obergfell F, Meule A, Vögele C, Schlarb AA, Kübler A (2012) Of larks and hearts – morningness/eveningness, heart rate variability and cardiovascular stress response at different times of day. Physiol Behav 106(2):151–157. https://doi.org/10.1016/j.physbeh.2012.01.023 Rosselot AE, Hong CI, Moore SR (2016, January 1) Rhythm and bugs: circadian clocks, gut microbiota, and enteric infections. Curr Opin Gastroenterol 32:7–11. https://doi.org/10.1097/ MOG.0000000000000227 Ruger M, Scheer FAJL (2009, December) Effects of circadian disruption on the cardiometabolic system. Rev Endocr Metab Disord 10:245–260. https://doi.org/10.1007/s11154-009-9122-8 Schernhammer ES, Laden F, Speizer FE, Willet WC, Hunter DJ, Kawachi I et al (2003) Night-shift work and risk of colorectal cancer in the nurses’ health study. J Natl Cancer Inst 95 (11):825–828. https://doi.org/10.1093/jnci/95.11.825 Schwabe RF, Jobin C (2013) The microbiome and cancer. Nat Rev Cancer 13(11):800–812. https:// doi.org/10.1038/nrc3610 Shibolet O, Podolsky DK (2007) TLRs in the gut. IV Negative regulation of Toll-like receptors and intestinal homeostasis: addition by subtraction. Am J Physiol Gastrointest Liver Physiol 292(6): G1469–G1473. https://doi.org/10.1152/ajpgi.00531.2006 Soták M, Sumová A, Pácha J (2014) Cross-talk between the circadian clock and the cell cycle in cancer. Ann Med 46(4):221–232. https://doi.org/10.3109/07853890.2014.892296

348

L. Tran et al.

Sukumaran S, Jusko WJ, DuBois DC, Almon RR (2011) Light-dark oscillations in the lung transcriptome: implications for lung homeostasis, repair, metabolism, disease, and drug action. J Appl Physiol 110(6):1732–1747. https://doi.org/10.1152/japplphysiol.00079.2011 Summa KC, Voigt RM, Forsyth CB, Shaikh M, Cavanaugh K, Tang Y et al (2013) Disruption of the circadian clock in mice increases intestinal permeability and promotes alcohol-induced hepatic pathology and inflammation. PLoS One 8(6):e67102. https://doi.org/10.1371/journal. pone.0067102 Swanson GR, Gorenz A, Shaikh M, Desai V, Forsyth C, Fogg L et al (2015) Decreased melatonin secretion is associated with increased intestinal permeability and marker of endotoxemia in alcoholics. Am J Physiol Gastrointest Liver Physiol 308(12):1004–1011. https://doi.org/10. 1152/ajpgi.00002.2015 Swanson GR, Gorenz A, Shaikh M, Desai V, Kaminsky T, van Den Berg J et al (2016) Night workers with circadian misalignment are susceptible to alcohol-induced intestinal hyperpermeability with social drinking. Am J Physiol Gastrointest Liver Physiol 311(1): G192–G201. https://doi.org/10.1152/ajpgi.00087.2016 Swanson GR, Siskin J, Gorenz A, Shaikh M, Raeisi S, Fogg L et al (2020) Disrupted diurnal oscillation of gut-derived short chain fatty acids in shift workers drinking alcohol: possible mechanism for loss of resiliency of intestinal barrier in disrupted circadian host. Transl Res 221:97–109. https://doi.org/10.1016/j.trsl.2020.04.004 Thaiss CA, Zeevi D, Levy M, Zilberman-Schapira G, Suez J, Tengeler AC et al (2014) Transkingdom control of microbiota diurnal oscillations promotes metabolic homeostasis. Cell 159(3):514–529. https://doi.org/10.1016/j.cell.2014.09.048 Thaiss CA, Levy M, Korem T, Dohnalová L, Shapiro H, Jaitin DA et al (2016) Microbiota diurnal rhythmicity programs host transcriptome oscillations. Cell 167(6):1495–1510.e12. https://doi. org/10.1016/j.cell.2016.11.003 Turek FW, Joshu C, Kohsaka A, Lin E, Ivanova G, McDearmon E et al (2005) Obesity and metabolic syndrome in circadian clock mutant nice. Science 308(5724):1043–1045. https://doi. org/10.1126/science.1108750 Van Der Horst GTJ, Muijtjens M, Kobayashi K, Takano R, Kanno SI, Takao M et al (1999) Mammalian Cry1 and Cry2 are essential for maintenance of circadian rhythms. Nature 398 (6728):627–630. https://doi.org/10.1038/19323 Vannucci L, Stepankova R, Grobarova V, Kozakova H, Rossmann P, Klimesova K et al (2009, December) Colorectal carcinoma: importance of colonic environment for anti-cancer response and systemic immunity. J Immunotoxicol 6:217–226. https://doi.org/10.3109/ 15476910903334343 Venegas DP, De La Fuente MK, Landskron G, González MJ, Quera R, Dijkstra G et al (2019) Short chain fatty acids (SCFAs)mediated gut epithelial and immune regulation and its relevance for inflammatory bowel diseases. Front Immunol 10:277. https://doi.org/10.3389/fimmu.2019. 00277 Vetter C, Devore EE, Ramin CA, Speizer FE, Willett WC, Schernhammer ES (2015) Mismatch of sleep and work timing and risk of type 2 diabetes. Diabetes Care 38(9):1707–1713. https://doi. org/10.2337/dc15-0302 Vician M, Zeman M, Herichová I, Juráni M, Blažček P, Matis P (1999) Melatonin content in plasma and large intestine of patients with colorectal carcinoma before and after surgery. J Pineal Res 27 (3):164–169. https://doi.org/10.1111/j.1600-079X.1999.tb00612.x Vivarelli S, Salemi R, Candido S, Falzone L, Santagati M, Stefani S et al (2019) Gut microbiota and Cancer: from pathogenesis to therapy. Cancers 11(1):38. https://doi.org/10.3390/ cancers11010038 Voigt RM, Forsyth CB, Keshavarzian A (2013) Circadian disruption: potential implications in inflammatory and metabolic diseases associated with alcohol. Alcohol Res: Current Rev 35 (1):87–96. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/24313168

Disease Implications of the Circadian Clocks and Microbiota Interface

349

Voigt RM, Forsyth CB, Green SJ, Mutlu E, Engen P, Vitaterna MH et al (2014) Circadian disorganization alters intestinal microbiota. PLoS One 9(5):e97500. https://doi.org/10.1371/ journal.pone.0097500 Voigt RM, Summa KC, Forsyth CB, Green SJ, Engen P, Naqib A et al (2016) The circadian clock mutation promotes intestinal Dysbiosis. Alcohol Clin Exp Res 40(2):335–347. https://doi.org/ 10.1111/acer.12943 Voigt RM, Forsyth CB, Shaikh M, Zhang L, Raeisi S, Aloman C et al (2018) Diurnal variations in intestinal barrier integrity and liver pathology in mice: implications for alcohol binge. Am J Physiol Gastrointest Liver Physiol 314(1):G131–G141. https://doi.org/10.1152/ajpgi.00103. 2017 Wang Y, Kuang Z, Yu X, Ruhn KA, Kubo M, Hooper LV (2017) The intestinal microbiota regulates body composition through NFIL3 and the circadian clock. Science 357 (6354):912–916. https://doi.org/10.1126/science.aan0677 Wittmann M, Dinich J, Merrow M, Roenneberg T (2006) Social jetlag: misalignment of biological and social time. Chronobiol Int 23(1–2):497–509. https://doi.org/10.1080/07420520500545979 Yoo SH, Yamazaki S, Lowrey PL, Shimomura K, Ko CH, Buhr ED et al (2004) PERIOD2:: LUCIFERASE real-time reporting of circadian dynamics reveals persistent circadian oscillations in mouse peripheral tissues. Proc Natl Acad Sci U S A 101(15):5339–5346. https://doi.org/ 10.1073/pnas.0308709101 Young ME, Brewer RA, Peliciari-Garcia RA, Collins HE, He L, Birky TL et al (2014) Cardiomyocyte-specific BMAL1 plays critical roles in metabolism, signaling, and maintenance of contractile function of the heart. J Biol Rhythm 29(4):257–276. https://doi.org/10.1177/ 0748730414543141 Zarrinpar A, Chaix A, Yooseph S, Panda S (2014) Diet and feeding pattern affect the diurnal dynamics of the gut microbiome. Cell Metab 20(6):1006–1017. https://doi.org/10.1016/j.cmet. 2014.11.008

Circadian Organization of the Gut Commensal Bacterium Klebsiella aerogenes Kinga B. Graniczkowska and Vincent M. Cassone

Abstract While the expression of circadian rhythmicity is nearly universal among eukaryotic organisms, demonstration of this phenomenon in prokaryotes has been largely restricted to photosynthetic cyanobacteria until very recently. Growing interest in gastrointestinal microbiomes has revealed a complex temporal relationship between the clock of gastrointestinal track and the microbiome within. We have discovered that at least one member of the gut microbiome, Klebsiella (neé Enterobacter) aerogenes, responds to the indoleamine hormone melatonin, secreted by the gastrointestinal system itself, in a specific, dose-dependent fashion such that melatonin increases bacterial motility. Further research revealed that K. aerogenes also express a circadian rhythm in motility and gene expression that is temperature compensated if maintained in constant temperatures ranging from 27
C to 40
C. Although rhythmicity is unaltered by changes in constant temperature, cycles of ambient temperature entrain circadian clock in K. aerogenes. Circadian rhythms in these bacteria rapidly decrease in amplitude following exposure to temperature cycles. The mechanisms of this damping are discussed.

1 Introduction Circadian clocks are mechanisms that evolved in most organisms to anticipate daily changes in the environment such as the light-dark cycle or daily temperature fluctuations and to orchestrate complex processes within the organism. The daily oscillations generated by these clocks regulate metabolic and behavioral changes in anticipation of 24 h changes in the environment. They share common characteristics among all organisms studied thus far. These include the following: (1) Circadian rhythms are endogenously generated such that rhythms persist when organisms are placed in constant environmental conditions such as constant temperature, light, and/or darkness (Dunlap et al. 2004; Bell-Pedersen et al. 2005). Typically, these

K. B. Graniczkowska · V. M. Cassone (*) Department of Biology, University of Kentucky, Lexington, KY, USA e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_18

351

352

K. B. Graniczkowska and V. M. Cassone

rhythms express periods of close to but rarely exactly 24 h (Johnson et al. 2001). (2) The periods of circadian rhythms vary very little in response to changes in constant ambient temperature (TA) and are considered temperature insensitive or temperature compensated at TA’s that support life. (3) Circadian rhythms can be entrained to cyclical patterns of environmental change. These are typically patterns of light and dark (LD) but can also be patterns of changing TA such as cycles of high TA and low TA. While the expression of circadian rhythms is almost universal among eukaryotic organisms, the numbers of prokaryotes that have been shown to express them is vanishingly small, even though the genetic diversity and ecological distribution of prokaryotes far exceed those of eukaryotes (Hug et al. 2016). For many years scientists believed that bacteria do not possess a circadian clock due to their short generation time (Johnson et al. 1996), which is usually shorter than the 24 h day. However, the discovery of a robust circadian clock in the cyanobacteria S. elongatus dispelled this notion. As is discussed in detail elsewhere (Cohen and Golden 2015), the cyanobacterial core oscillator comprises three proteins (KaiA, KaiB, and KaiC), which are responsible for driving circadian rhythms of gene expression, cell division, photosynthesis, and all aspects of cellular metabolism (Cohen and Golden 2015; Mackey et al. 2008). Homologs of KaiB and KaiC proteins can be found in many other microorganisms, including other cyanobacteria, Proteobacteria and Archaea, but KaiA homologues have only been identified in Cyanobacteria thus far (Schmelling et al. 2017). Circadian growth rhythms have been reported in other bacteria, which do not express the kaiABC operon, such as Escherichia coli (Halberg and Conner 1961), Klebsiella pneumoniae (Sturtevant 1973), and Pseudomonas putida (Soriano et al. 2010); however, robust and persistent circadian rhythms have never been characterized in these microorganisms. Human gut bacteria may not be subjected to the most common zeitgeber, the LD cycle. However, they still encounter daily fluctuations in body temperature (Aschoff 1967), nutrient availability, gastrointestinal hormones, and levels of host-delivered antimicrobial peptides and gut mucosal antibodies (Liang and FitzGerald 2017). Recent studies have shown that intestinal microbiota exhibits population level rhythmicity. Several research groups have described daily rhythmic changes in abundance and functional capacities of intestinal commensal bacteria in both humans and mice (Thaiss et al. 2014; Zarrinpar et al. 2014). According to these studies, 15–17% of identified OTUs (operational taxonomic units) exhibit time-ofday-dependent differences in their relative abundance. In addition, total biomass of murine gut bacteria and their taxonomic abundance exhibit daily rhythmicity (Liang et al. 2015). At least some of these patterns are regulated by the host’s circadian clock. Disruption of the host circadian clock mechanism by Bmal-1 or Per1/2 deletion leads to arrhythmicity of the intestinal microbial community, suggesting that this oscillation is regulated by the host. Daily oscillation of microbial communities requires an intact circadian clock in the host (Zarrinpar et al. 2014; Thaiss et al. 2014). Interactions between the host circadian oscillator and microbiota are mutual. The host receives and integrates light signals from the environment by the circadian

Circadian Organization of the Gut Commensal Bacterium Klebsiella aerogenes

353

master pacemaker, the suprachiasmatic nuclei (SCN), which regulates the peripheral clocks in the intestine and its gut microbial composition. Another apparent circadian output is rhythmic secretion of pineal melatonin, which is also present throughout the gastrointestinal system (Chen et al. 2011) and can be measured in the feces of laboratory mice; however, this rhythm diminished with age (Paulose et al. 2019a).

1.1

Klebsiella aerogenes

Klebsiella (née Enterobacter) aerogenes a gram-negative, flagellated bacterium, belonging to the Family Enterobacteriaceae. It is a facultative anaerobic bacillus that possesses class 1 fimbriae (Hart 2006). This bacterium has both commercial and biomedical significance. It is an outstanding hydrogen producer (Asadi and Zilouei 2017) and is capable of synthesizing an important chemical intermediate 2,3-butanediol (Jung et al. 2012). K. aerogenes can be found in many environments such as soil, water, and the bovine rumen (Szczerba et al. 2020), but, for the purposes of this discussion, it is also a minor component of the human gut commensal microbiota, with up to 107 cfu/g (Gillespie and Hawkey 2006). Even though K. aerogenes’ presence in the human gut is not harmful for healthy individuals, its increased abundance has been connected to atherosclerotic cardiovascular disease (ACVD) (Jie et al. 2017). Additionally, K. aerogenes is a nosocomial and opportunistic pathogen, being dangerous to immunocompromised patients and capable of causing urinary and respiratory tract infections in hospitalized patients. K. aerogenes accounts for 6–17% of all nosocomial urinary tract infections (UTI) (Surve and Bagde 2011; Jiménez-Guerra et al. 2020). These infections originate from the gastrointestinal tracts of patients and the hands of hospital staff members, along with unsanitary medical equipment. K. aerogenes contains several virulence factors such as pili, which are non-flagellar projections from the cell surface. These structures allow cells to adhere to the host’s mucosal and epithelial layers of the urogenital, respiratory, and gastrointestinal tracts (Podschun and Ullmann 1998). Other pathogenicity factors include production of iron chelating siderophores (Hartmann et al. 2008) and the expression of urease (Collins and D’Orazio 1993), which allows them to induce infection. Additionally, this microbe has the potential to develop antibiotic resistance due to its inducible chromosomal ampC betalactamase gene and the ability to modify non-specific porins to allow for active efflux of various drugs from cells (Chevalier et al. 2004).

2 Characteristics of Circadian Rhythm in K. aerogenes To address the hypothesis that gut bacteria have adapted to melatonin signaling from the intestine, a computational comparison of the proteome from the human gut microbiome database and a binding site of human melatonin receptor was

354

K. B. Graniczkowska and V. M. Cassone

Fig. 1 Circadian rhythms in K. aerogenes. (a) Morphological differences in macrocolony formation of K. aerogenes grown in motility assay with 1 nM or 0 nM melatonin; (b) pictures of one representative macrocolony captured at 24 h intervals grown with 0 nM melatonin on semisolid eosin methylene blue (EMB) media

performed. The results indicated 24–42% identity to the sequence of the human melatonin receptor and several proteins produced by the gut microbiome (Paulose et al. 2016). Among these, the highest resemblance was found in the mntH gene (manganese transport, H-dependent (Helmann 2013)) from K. aerogenes. A search for sequences that resembled cyanobacterial kaiABC complexes revealed several similarities. Dephospho coenzyme A kinase (coaE) resembles KaiC from S. elongatus. However, motif analyses did not uncover functional similarities other than the fact that both KaiC and CoaE exhibit kinase activity. Further analysis of the K. aerogenes genome sequence (Shin et al. 2012) and genome sequencing of our lab strain (not published) showed that the circadian clock of this bacterium is likely driven by a different mechanism since kaiABC sequences are not present in this bacterium. Subsequent motility assay experiments with K. aerogenes grown in different concentrations of melatonin added to eosin-methylene blue (EMB) agar plates with a 50% reduction in agar to facilitate motility, ranging from 0 nM to 1 nM, revealed morphological differences among groups (Fig. 1a). Bacteria cultivated on plates with melatonin produced much greater diameters of macrocolonies compared to controls in a dose-dependent fashion. These bacteria produced a new outer ring each day of the culture with new microorganisms growing on the outer edge of the microcolony but also on the top of older bacteria (Fig. 1b). When the number of rings was calculated and divided by time of incubation in 1 nM melatonin’s presence, the period of swarming behavior was 25.1  1.4 h.

Circadian Organization of the Gut Commensal Bacterium Klebsiella aerogenes

355

Fig. 2 Bioluminescence reporter. (a) Plasmid map: this plasmid encodes the luciferase cassette and tetracycline resistance gene. (b) Plasmid stability: K. aerogenes was cultivated in the motility assay as in (Paulose et al. 2016). Every 24 h samples were collected and serially diluted for the colony forming unit (CFU) count; CFUs were compared between plates with no antibiotics and those with tetracycline. Differences indicated how many bacteria lost the plasmid over time. (c) Bioluminescence rhythm of K. aerogenes with the pmotA::luxCDABE promoter grown in the motility assay at 37 or 40
C, spiked with 1 nM melatonin or control conditions corresponding to 0 nM melatonin. Each colorful trace represents one plate kept in the Lumicycle32 system. Adapted from Paulose et al. (2016)

To further test this phenomenon, K. aerogenes was transformed with a plasmid containing a bacterial luciferase cassette from Photorabdus luminesens, which is commonly employed in cyanobacteria as a real-time reporter of gene expression (Kondo et al. 1993). The plasmid was modified to contain the promoter of the motility gene motA regulating the luciferase operon expression (pmotA:: luxCDABE), resulting in a real-time reporter of gene expression (Fig. 2a). MotA is

356

K. B. Graniczkowska and V. M. Cassone

a component of the flagellar motor, providing energy for flagellar rotation and, consequently, bacterial motility (Blair and Berg 1990). K. aerogenes maintained a free running circadian period of gene expression in vitro. Measurement of bioluminescence with a Lumicycle32 photomultiplier system revealed temperature compensated circadian rhythms in 31–44% of cultures. These cultures were rhythmic in temperatures ranging from 27
C to those corresponding to human body temperatures (TB) of 34
C, 37
C, and even 40
C (Paulose et al. 2016). Additionally, cultures grown in the presence of 1 nM melatonin were more synchronized than the control plates (Fig. 2c). It is important to highlight that other indole molecules such as of tryptophan, serotonin, or N-acetylserotonin did not affect K. aerogenes motility or the bioluminescence signal. Thus, the circadian clock of K. aerogenes exhibits two canonical characteristics of circadian clocks in most organisms. It expresses a circadian rhythm that persists in constant environmental conditions with a period of close to but not exactly 24 h. Secondly, the K. aerogenes circadian rhythm is temperature compensated such that periods were invariant in constant TA from 27
C to 40
C (Q10 ¼ 0.96).

3 Temperature Entrainment of the K. aerogenes Circadian Clock Further experiments were conducted to characterize the remaining canonical property of this circadian clock – the ability to entrain to cyclical environmental cycles. Because there is no evidence that these bacteria were sensitive to ambient light, we focused on cyclical patterns of TA. It has been known for some time that the core body temperature (TB) of homeothermic animals such as birds and mammals exhibits pronounced circadian patterns of TB, including that of humans and mice (Refinetti and Menaker 1992). In humans, the average normal TB range is 37.5
C during the day to 36.5
C during the night, revealing a daily amplitude of 1
C (Aschoff 1967). In mice, the amplitude can be as high as 4
C (Gordon 2017). The amplitude of this rhythm is affected by health (fever) and chronotype. We, therefore, hypothesized that daily patterns of TA simulating the daily patterns of TB may act as a zeitgeber for entraining the K. aerogenes clock. Bacterial cultures were exposed to various periods (T-cycles) of changing TA with cycles of 1
C (35
C–36
C) or 3
C (34
C–37
C) in amplitude (Paulose et al. 2019b). Cultures were maintained for 5 days in several T cycles of different periods: 22 h (11 h high temperature, 11 h low temperature), 24 h (12 h high temperature, 12 h low temperature), and 28 h (14 h high temperature, 14 h low temperature). They were then released into constant low temperature (34
C or 35
C). Cultures entrained to all T cycles with a stable phase relationship (Ψ) to the rise in TA. When cultures were released into constant low temperature, rhythms persisted with a Ψ that corresponded to the time of high temperature but with a significantly lower amplitude. These data showed that the circadian clock of K. aerogenes was

Circadian Organization of the Gut Commensal Bacterium Klebsiella aerogenes

357

Fig. 3 Entrainment of the K. aerogenes circadian clock by temperature cycles of 1
C (35
C– 36
C) or 3
C (34
C–37
C) to short day of 22 h, standard day of 24 h, and long day of 28 h. Adapted from Paulose et al. (2019b)

capable of entraining to a wide array of T cycles and at two different amplitudes (Fig. 3). To determine the nature of temperature entrainment in this microorganism, a phase response curve (PRC) to pulses of 1 h pulses of 3
C temperature increase (34
C–37
C) at different times of the day circadian was established. Temperature pulses resulted in phase (ϕ)-dependent changes in ϕ(Δϕ) of the bioluminescence rhythms of these microorganism (Fig. 4a), revealing a high-amplitude PRC with Type-0 phase resetting. In this case, the phase reference (ϕR) was the acrophase, or peak, of the bioluminescence signal. Calculation of a phase transition curve (PTC),

358

K. B. Graniczkowska and V. M. Cassone

Fig. 4 K. aerogenes exhibit “type 0” phase resetting behavior in response to temperature and spatial differences in the expression of the bioluminescence reporter. (a) PRC generated by plotting the magnitude of phase shift across 10 phases. Each data point represents a single plate, with multiple experiments comprising the entire dataset. CT6 here is defined as the pre-pulse peak of bioluminescence. A phase transition curve (b) was generated by plotting the pre-pulse phase against the post-pulse phase from the PRC. Linear regression analysis is represented by the solid line. The dashed line indicates a slope equal to 1. (c) Bioluminescence recorded from a 24-well plate containing cultures of K. aerogenes: initially a high-amplitude rhythm that damps over time. One representative well (d) is shown with inset images of the quantified signal at 4 h intervals after baseline correction. Adapted from Paulose et al. (2019b)

in which the new ϕ was plotted on the y axis and the old ϕ was plotted on the x axis, revealed a Type-0 PTC in that the slope of the resultant linear regression was 0.19 with an r2 value of 0.11 (Fig. 4b). Next, microorganisms expressing the bioluminescence reporter were cultivated and recorded using a Perkin Elmer in vivo imaging system (IVIS), which employs a CCD camera system. This instrument allowed us to characterize the spatial distribution of the bioluminescence signal in the motility assay. A high amplitude signal of bioluminescence was recorded in constant temperature for 2.5 days with a freerunning period of 25.4  0.4 h. These results were comparable with our previous reports of data recorded with the Lumicycle (with t ¼ 24.5  0.5 h) (Paulose et al.

Circadian Organization of the Gut Commensal Bacterium Klebsiella aerogenes

359

2016). The bioluminescence signal originated from the center, where bacteria were inoculated and with time proceeded toward the periphery (Fig. 4d).

4 Discussion K. aerogenes, as a human gut commensal bacterium, lives in rhythmic conditions of fluctuating core body temperature, daily changes in the concentration of hostdelivered antimicrobial peptides, gut mucosal antibodies, nutrient availability, and shifting composition of commensal gut bacteria community. Therefore, it is understandable that this bacterium over a course of coevolution with humans adapted to its host circadian rhythmicity and developed its own mechanism to anticipate diurnal changes in its environment. The understanding of this phenomenon will provide important insights into the complexity of microbiome-host interactions and human health. The fact that a commensal gut bacterium has its own endogenous clock adds to the hierarchical organization of human circadian rhythms. We have proposed (Paulose and Cassone 2016) that circadian organization resembles a meta-organism of circadian clocks within circadian clocks: the circadian oscillators in the brain entrain circadian oscillators in the periphery (Malloy et al. 2012; Hoogerwerf 2006), such as in the intestines, which subsequently affects bacterial circadian clocks. Even though our recent findings show that K. aerogenes possess a daily pattern of motility, suggesting a vigorous circadian behavior, further assessment of rhythmicity of other genes is required and more importantly identification of their core oscillator is crucial at this point, which is a focus of our studies. K. aerogenes colonies exhibit rhythmic motility patterns (Fig. 1a, b) and demonstrate circadian expression of motA gene as recorded by the bacterial luciferase reporter (Fig. 2c) (Paulose et al. 2016). K. aerogenes maintain a circadian period of gene expression not only during temperature entrainment but also during the free running conditions. This observation is in accordance with the metagenomic analysis reported by Thaiss et al. (2014) indicating daily oscillation in the functional composition of microbial genes (microbiome) with the antiphase fluctuations of distinct gene groups in murine microbiota. Specifically, Thaiss et al. (2014) demonstrated higher expression of genes involved in energy metabolism, DNA repair, and cell growth during the active phase (darkness for nocturnal mice) and upregulation of environmental sensing, detoxification, and flagella assembly genes during the resting phase (light). Upregulation of motility genes in the resting phase of the host reported by Thaiss et al. (2014) can be explained by the need to seek nutrient sources and mucus penetration (Thaiss et al. 2014, 2016). Therefore, K. aerogenes, like other commensal bacteria, regulates its life cycle, using a circadian clock, to maximize the benefits of interaction with the host and optimize own fitness by expressing a diurnal pattern of motility. Clear morphological changes visible with a naked eye in the appearance of growing colonies of K. aerogenes (Fig. 1a) indicate that these bacteria move at a certain time of the circadian day, but they can also regulate its cell division in a

360

K. B. Graniczkowska and V. M. Cassone

circadian fashion. Similarly, cyanobacteria control their cell division cycle in a circadian fashion, and even rapidly growing cultures of S. elongatus exhibit rhythmicity as a population even though the generation time in cyanobacteria is 8–12 h long. In contrast, K. aerogenes divide every 30  0.3 min in exponential growth phase (Kelly and Rahn 1932). Nevertheless, we still observe a circadian pattern of gene expression at a population level recorded by bioluminescence reporters. Study of the cyanobacteria circadian clock on a single-cell resolution revealed that individual bacterial cells are rhythmic but this synchrony is lost after 166  100 days (Mihalcescu et al. 2004), which, in the case of K. aerogenes, when taking into account drastic differences in the length of the generation time would correspond to only 2.75–5.5 days. As we pointed out earlier, the K. aerogenes bioluminescence rhythm has a very high amplitude during the temperature entrainment (Paulose et al. 2019b), and upon releasing to free running conditions the rhythm persists for at least 4 days. However, the amplitude of this signal greatly decreases in constant conditions. This is in accordance with Sturtevant’s work on continuously cultured K. pneumoniae. When these bacteria were exposed to the 12:12 light dark cycle, cultures displayed rhythmic changes in the optical density with a period of 24.1 h. However, when bacteria were grown in constant dim light or darkness this rhythm decreased in amplitude and damped out (Sturtevant 1973). There are several possible explanations why the signal amplitude reduces with time in cultured K. aerogenes. First, the oscillator itself within each bacterium could be damping, or that the individual bacteria could be losing the synchrony between each other in the culture. While this is possible, we find very little change in the period or waveform shape, suggesting the decrease in amplitude is not a form of damping. At this point, however, without identifying the core oscillator we cannot determine whether the oscillator is damping or not. Another possibility is that these microorganisms are losing the plasmid carrying the luciferase reporter. Establishing the synchrony on a single-cell resolution in bacteria, which in the case of K. aerogenes is only 1.0 μm wide by 3 μm long, is very challenging, especially since these are fast-growing, motile microorganisms. Even so, we have tested the stability of the plasmid, and plasmid loss was not significant during the 6 days of the motility assay (Fig. 2b). Therefore, we excluded plasmid loss as a reason for signal decrease. Finally, light production generated by the bacterial luciferase depends on oxygen (Botella et al. 2012), and our current culture conditions do not supply extra oxygen after plates are inoculated and sealed to prevent media from drying out; thus after several days O2 is depleted, which may contribute to decrease in amplitude. The next important characteristic of the circadian rhythmicity in K. aerogenes is its sensitivity to melatonin. Many studies provide evidence for the melatonin production by gut tissue in different animals (Muñoz-Pérez et al. 2016; Paulose et al. 2019a; Reiter et al. 2011; Chen et al. 2011), and at least in young laboratory mice, the concentration of this hormone in feces shows diurnal rhythmicity (Paulose et al. 2019a). The melatonin concentration found in the gut is 400 times higher than in the pineal gland and 10–100 times higher than in the serum, and increased secretion coincidences with food consumption (Acuña-Castroviejo et al. 2014). Additionally, many edible plants contain this hormone (Dubbels et al. 1995). A recent study

Circadian Organization of the Gut Commensal Bacterium Klebsiella aerogenes

361

showed that melatonin suppresses dysbiosis of the gut microbiome through a Tolllike receptor (TLR) 4 signal pathway (Kim et al. 2020). Others have shown that melatonin reduces body weight by modulation of the gut microbiota composition, particularly by changing the ratio of Firmicutes to Bacteroidetes and by increasing the number of Akkermansia, which are associated with lean individuals (Xu et al. 2017). In the case of K. aerogenes rhythmicity, we observed that melatonin synchronizes the phase of these bacteria even across different culture plates. This could be happening by affecting an element of two component systems, which are important in regulation of S. elongatus circadian rhythmicity, or other transcription regulators, but this hypothesis still requires empirical research. We need to highlight that in our experiments melatonin was constantly present in the agar media. Therefore, we have also hypothesized that rhythmic administration of melatonin could entrain K. aerogenes cultures. These studies are ongoing. We identified melatonin as a novel source of host-commensal microbiota communication within the gut and potential zeitgeber for the whole microbiota community, or at least some members of this group. The human gut microbiota is a very dynamic environment that has gained a lot of interest in the past decade due to its complexity and importance for human health. This dynamic ecological niche is prone to increased diversity through horizontal gene transfer (HGT) even among unrelated species. It has been reported that human-associated microorganisms, including gut bacteria, experience 25-fold more HGT than bacteria residing in other ecologically diverse environments (Smillie et al. 2011). Therefore, it is very likely that human gut microbiota contains more bacteria expressing its own circadian clock. Discovery of the circadian clock and melatonin sensitivity in human commensal bacterium is an important milestone in the chronobiology and human microbiome field, and these phenomena are likely not restricted to one species. Further investigation of the circadian rhythmicity in K. aerogenes and identification of its core oscillator still need to be elucidated. Nevertheless, we have enough evidence to believe that this bacterium has a robust clock mechanism. Acknowledgments The research described in this paper was supported by NIH R01 GM118541, HHMI Sustaining Excellence Award 52008116, and KSEF-3783-020. We would like to thank Charles Cassone, Dr. Stephen Ferguson, Sydney Ellen Gooding, Kaylynne Glover, and Clifford Harpole for helpful discussions.

References Acuña-Castroviejo D, Escames G, Venegas C, Díaz-Casado ME, Lima-Cabello E, López LC, Rosales-Corral S, Tan DX, Reiter RJ (2014) Extrapineal melatonin: sources, regulation, and potential functions. Cell Mol Life Sci 71(16):2997–3025. https://doi.org/10.1007/s00018-0141579-2 Asadi N, Zilouei H (2017) Optimization of organosolv pretreatment of rice straw for enhanced biohydrogen production using Enterobacter aerogenes. Bioresour Technol 227:335–344. https://doi.org/10.1016/j.biortech.2016.12.073

362

K. B. Graniczkowska and V. M. Cassone

Aschoff J (1967) Human circadian rhythms in activity, body temperature and other functions. Life Sci Space Res 5:159–173 Bell-Pedersen D, Cassone VM, Earnest DJ, Golden SS, Hardin PE, Thomas TL, Zoran MJ (2005) Circadian rhythms from multiple oscillators: lessons from diverse organisms. Nat Rev Genet 6 (7):544–556 Blair DF, Berg HC (1990) The MotA protein of E. coli is a proton-conducting component of the flagellar motor. Cell 60(3):439–449 Botella E, Noone D, Salzberg LI, Hokamp K, Devine SK, Fogg M, Wilkinson AJ, Devine KM (2012) High-resolution temporal analysis of global promoter activity in Bacillus subtilis. In: Methods in microbiology, vol 39. Elsevier, Amsterdam, pp 1–26 Chen CQ, Fichna J, Bashashati M, Li YY, Storr M (2011) Distribution, function and physiological role of melatonin in the lower gut. World J Gastroenterol 17(34):3888–3898. https://doi.org/10. 3748/wjg.v17.i34.3888 Chevalier J, Bredin J, Mahamoud A, Malléa M, Barbe J, Pagès J-M (2004) Inhibitors of antibiotic efflux in resistant Enterobacter aerogenes and Klebsiella pneumoniae strains. Antimicrob Agents Chemother 48(3):1043–1046 Cohen SE, Golden SS (2015) Circadian rhythms in cyanobacteria. Microbiol Mol Biol Rev 79 (4):373–385. https://doi.org/10.1128/mmbr.00036-15 Collins CM, D'Orazio SE (1993) Bacterial ureases: structure, regulation of expression and role in pathogenesis. Mol Microbiol 9(5):907–913 Dubbels R, Reiter RJ, Klenke E, Goebel A, Schnakenberg E, Ehlers C, Schiwara H, Schloot W (1995) Melatonin in edible plants identified by radioimmunoassay and by high performance liquid chromatography-mass spectrometry. J Pineal Res 18(1):28–31 Dunlap JC, Loros JJ, DeCoursey PJ (2004) Chronobiology: biological timekeeping. Sinauer Associates, Sunderland, MA Gillespie S, Hawkey PM (2006) Principles and practice of clinical bacteriology. John Wiley & Sons, Hoboken, NJ Gordon CJ (2017) The mouse thermoregulatory system: its impact on translating biomedical data to humans. Physiol Behav 179:55–66. https://doi.org/10.1016/j.physbeh.2017.05.026 Halberg F, Conner RL (1961) Circadian organization and microbiology: variance spectra and a periodogram on behavior of Escherichia coli growing in fluid culture. Proc Minn Acad Sci 29:227–239 Hart CA (2006) Klebsiella, citrobacter, enterobacter and serratia spp. In: Principles and practice of clinical bacteriology. John Wiley and Sons, England, UK, pp 377–386 Hartmann H, Eltzschig HK, Wurz H, Hantke K, Rakin A, Yazdi AS, Matteoli G, Bohn E, Autenrieth IB, Karhausen J (2008) Hypoxia-independent activation of HIF-1 by enterobacteriaceae and their siderophores. Gastroenterology 134(3):756–767.e756 Helmann JD (2013) Regulation of bacterial MntH genes. Madame Curie Bioscience Database Available from: https://www.ncbi.nlm.nih.gov/booksNBK6015/ Accessed 17 May 2020 Hoogerwerf WA (2006) Biologic clocks and the gut. Curr Gastroenterol Rep 8(5):353–359. https:// doi.org/10.1007/s11894-006-0019-3 Hug LA, Baker BJ, Anantharaman K, Brown CT, Probst AJ, Castelle CJ, Butterfield CN, Hernsdorf AW, Amano Y, Ise K, Suzuki Y, Dudek N, Relman DA, Finstad KM, Amundson R, Thomas BC, Banfield JF (2016) A new view of the tree of life. Nat Microbiol 1(5):16048. https://doi.org/ 10.1038/nmicrobiol.2016.48 Jie Z, Xia H, Zhong SL, Feng Q, Li S, Liang S, Zhong H, Liu Z, Gao Y, Zhao H, Zhang D, Su Z, Fang Z, Lan Z, Li J, Xiao L, Li J, Li R, Li X, Li F, Ren H, Huang Y, Peng Y, Li G, Wen B, Dong B, Chen JY, Geng QS, Zhang ZW, Yang H, Wang J, Wang J, Zhang X, Madsen L, Brix S, Ning G, Xu X, Liu X, Hou Y, Jia H, He K, Kristiansen K (2017) The gut microbiome in atherosclerotic cardiovascular disease. Nat Commun 8(1):845. https://doi.org/10.1038/s41467017-00900-1 Jiménez-Guerra G, Borrego-Jiménez J, Gutiérrez-Soto B, Expósito-Ruiz M, Navarro-Marí JM, Gutiérrez-Fernández J (2020) Susceptibility evolution to antibiotics of Enterobacter cloacae,

Circadian Organization of the Gut Commensal Bacterium Klebsiella aerogenes

363

Morganella morganii, Klebsiella aerogenes and Citrobacter freundii involved in urinary tract infections: an 11-year epidemiological surveillance study. Enferm Infecc Microbiol Clin 38 (4):166-169. https://doi.org/10.1016/j.eimc.2019.07.010 Johnson CH, Golden SS, Ishiura M, Kondo T (1996) Circadian clocks in prokaryotes. Mol Microbiol 21(1):5–11. https://doi.org/10.1046/j.1365-2958.1996.00613.x Johnson CH, Golden SS, Kondo T (2001) Circadian rhythms in cyanobacteria. In: eLS. John Wiley & Sons, Chichester. https://doi.org/10.1038/npg.els.0000389 Jung M-Y, Ng CY, Song H, Lee J, Oh M-K (2012) Deletion of lactate dehydrogenase in Enterobacter aerogenes to enhance 2,3-butanediol production. Appl Microbiol Biotechnol 95 (2):461–469. https://doi.org/10.1007/s00253-012-3883-9 Kelly C, Rahn O (1932) The growth rate of individual bacterial cells. J Bacteriol 23(2):147 Kim SW, Kim S, Son M, Cheon JH, Park YS (2020) Melatonin controls microbiota in colitis by goblet cell differentiation and antimicrobial peptide production through toll-like receptor 4 signalling. Sci Rep 10(1):2232. https://doi.org/10.1038/s41598-020-59314-7 Kondo T, Strayer CA, Kulkarni RD, Taylor W, Ishiura M, Golden SS, Johnson CH (1993) Circadian rhythms in prokaryotes: luciferase as a reporter of circadian gene expression in cyanobacteria. Proc Natl Acad Sci U S A 90(12):5672–5676. https://doi.org/10.1073/pnas.90. 12.5672 Liang X, FitzGerald GA (2017) Timing the microbes: the circadian rhythm of the gut microbiome. J Biol Rhythm 32(6):505–515. https://doi.org/10.1177/0748730417729066 Liang X, Bushman FD, FitzGerald GA (2015) Rhythmicity of the intestinal microbiota is regulated by gender and the host circadian clock. Proc Natl Acad Sci 112(33):10479–10484. https://doi. org/10.1073/pnas.1501305112 Mackey SR, Choi J-S, Kitayama Y, Iwasaki H, Dong G, Golden SS (2008) Proteins found in a CikA interaction assay link the circadian clock, metabolism, and cell division in Synechococcus elongatus. J Bacteriol 190(10):3738–3746 Malloy JN, Paulose JK, Li Y, Cassone VM (2012) Circadian rhythms of gastrointestinal function are regulated by both central and peripheral oscillators. Am J Physiol Gastrointest Liver Physiol 303(4):G461–G473. https://doi.org/10.1152/ajpgi.00369.2011 Mihalcescu I, Hsing W, Leibler S (2004) Resilient circadian oscillator revealed in individual cyanobacteria. Nature 430(6995):81–85 Muñoz-Pérez JL, López-Patiño MA, Álvarez-Otero R, Gesto M, Soengas JL, Míguez JM (2016) Characterization of melatonin synthesis in the gastrointestinal tract of rainbow trout (Oncorhynchus mykiss): distribution, relation with serotonin, daily rhythms and photoperiod regulation. J Comp Physiol B 186(4):471–484. https://doi.org/10.1007/s00360-016-0966-4 Paulose JK, Cassone VM (2016) The melatonin-sensitive circadian clock of the enteric bacterium Enterobacter aerogenes. Gut Microbes 7(5):424–427. https://doi.org/10.1080/19490976.2016. 1208892 Paulose JK, Wright JM, Patel AG, Cassone VM (2016) Human gut Bacteria are sensitive to melatonin and express endogenous circadian rhythmicity. PLoS One 11(1):e0146643. https:// doi.org/10.1371/journal.pone.0146643 Paulose JK, Cassone CV, Cassone VM (2019a) Aging, melatonin biosynthesis, and circadian clockworks in the gastrointestinal system of the laboratory mouse. Physiol Genomics 51 (1):1–9. https://doi.org/10.1152/physiolgenomics.00095.2018 Paulose JK, Cassone CV, Graniczkowska KB, Cassone VM (2019b) Entrainment of the circadian clock of the enteric bacterium Klebsiella aerogenes by temperature cycles. iScience 19:1202–1213. https://doi.org/10.1016/j.isci.2019.09.007 Podschun R, Ullmann U (1998) Klebsiella spp. as nosocomial pathogens: epidemiology, taxonomy, typing methods, and pathogenicity factors. Clin Microbiol Rev 11(4):589–603 Refinetti R, Menaker M (1992) The circadian rhythm of body temperature. Physiol Behav 51 (3):613–637 Reiter RJ, Rosales-Corral S, Coto-Montes A, Boga JA, Tan DX, Davis JM, Konturek PC, Konturek SJ, Brzozowski T (2011) The photoperiod, circadian regulation and chronodisruption: the

364

K. B. Graniczkowska and V. M. Cassone

requisite interplay between the suprachiasmatic nuclei and the pineal and gut melatonin. J Physiol Pharmacol 62(3):269–274 Schmelling NM, Lehmann R, Chaudhury P, Beck C, Albers S-V, Axmann IM, Wiegard A (2017) Minimal tool set for a prokaryotic circadian clock. BMC Evol Biol 17(1):169–169. https://doi. org/10.1186/s12862-017-0999-7 Shin SH, Kim S, Kim JY, Lee S, Um Y, Oh M-K, Kim Y-R, Lee J, Yang K-S (2012) Complete genome sequence of Enterobacter aerogenes KCTC 2190. J Bacteriol 194(9):2373-2374. https://doi.org/10.1128/jb.00028-12 Smillie CS, Smith MB, Friedman J, Cordero OX, David LA, Alm EJ (2011) Ecology drives a global network of gene exchange connecting the human microbiome. Nature 480(7376):241–244 Soriano MI, Roibás B, García AB, Espinosa-Urgel M (2010) Evidence of circadian rhythms in non-photosynthetic bacteria? J Circadian Rhythms 8:8–8. https://doi.org/10.1186/1740-3391-88 Sturtevant RP (1973) Circadian variability in Klebsiella demonstrated by cosinor analysis. Int J Chronobiol 1(2):141–146 Surve N, Bagde U (2011) Effect of sulphaphenazole on pathogenic microorganism Klebsiella aerogenes. Int J Biol 3(2):106 Szczerba H, Komoń-Janczara E, Krawczyk M, Dudziak K, Nowak A, Kuzdraliński A, Waśko A, Targoński Z (2020) Genome analysis of a wild rumen bacterium Enterobacter aerogenes LU2 a novel bio-based succinic acid producer. Sci Rep 10(1):1986. https://doi.org/10.1038/s41598020-58929-0 Thaiss CA, Zeevi D, Levy M, Zilberman-Schapira G, Suez J, Tengeler AC, Abramson L, Katz MN, Korem T, Zmora N (2014) Transkingdom control of microbiota diurnal oscillations promotes metabolic homeostasis. Cell 159(3):514–529 Thaiss CA, Levy M, Korem T, Dohnalova L, Shapiro H, Jaitin DA, David E, Winter DR, GuryBenAri M, Tatirovsky E, Tuganbaev T, Federici S, Zmora N, Zeevi D, Dori-Bachash M, Pevsner-Fischer M, Kartvelishvily E, Brandis A, Harmelin A, Shibolet O, Halpern Z, Honda K, Amit I, Segal E, Elinav E (2016) Microbiota diurnal rhythmicity programs host transcriptome oscillations. Cell 167(6):1495–1510.e1412. https://doi.org/10.1016/j.cell.2016. 11.003 Xu P, Wang J, Hong F, Wang S, Jin X, Xue T, Jia L, Zhai Y (2017) Melatonin prevents obesity through modulation of gut microbiota in mice. J Pineal Res 62(4):e12399. https://doi.org/10. 1111/jpi.12399 Zarrinpar A, Chaix A, Yooseph S, Panda S (2014) Diet and feeding pattern affect the diurnal dynamics of the gut microbiome. Cell Metab 20(6):1006–1017. https://doi.org/10.1016/j.cmet. 2014.11.008

Daily Rhythmicity in Coastal Microbial Mats Christine Hörnlein and Henk Bolhuis

Abstract To understand the ecological role of circadian rhythms in prokaryotes, we often move from an initial observation in nature to an in-depth analysis of isolated species under laboratory conditions. The other way around is less common: newly found insights on circadian control of microbial physiology and metabolic interactions are rarely tested in the original ecosystem. This is partly due to the general inability to use laboratory-type analytical tools and controlled conditions in the environment. Thanks to novel developments in high-throughput DNA and RNA sequencing and meta-proteomic approaches, we now have tools to determine the extant activity of individual genes and expressed proteins in the natural community. We apply this approach to photosynthetic microbial mats, unique millimeter-scale ecosystems that are dominated by circadian clock-controlled Cyanobacteria. The rhythmic activities of cyanobacteria related to photosynthesis and dinitrogen fixation have been well documented. Therefore, microbial mats form ideal model systems that are easily accessible, well-characterized, have a constant community composition in the short term, and can be mimicked under laboratory conditions. Here we will provide an overview on these diverse microbial ecosystems and describe our studies on rhythmic processes in coastal microbial mats.

1 Introduction Biological timing is an important process in many organisms in response to the natural cycles of day and night or, more correct, light and darkness. The nearly 24 h (circadian) oscillations in biological processes are essential for many organisms that rely on processes that depend on light and on its absence. The most notable of these processes is of course photosynthesis, but also food gathering and resting periods are important examples (for extensive reviews on this topic see among others: Ishida

C. Hörnlein · H. Bolhuis (*) Department of Marine Microbiology and Biogeochemistry, Royal Netherlands Institute for Sea Research (NIOZ), Texel, The Netherlands e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_19

365

366

C. Hörnlein and H. Bolhuis

et al. 1999; Wulund and Reddy 2015; Huang 2018). The biological clock affects many organisms and many more biological processes in plants and animals. Groundbreaking research to unravel the mechanisms of biological rhythms led to the 2017 Nobel Prize in Physiology and Medicine that was awarded to Jeffery C. Hall and Michael Rosbash of Brandeis University and Michael Young of Rockefeller University (https://www.nobelprize.org/prizes/medicine/2017/press-release/). One of their major findings was the so-called transcriptional-translational feedback loop (TTFL) model for the generation of an autonomous oscillator with a period of approximately 24 h (Hardin et al. 1990). This model requires a nucleus where transcription takes places whereas translation occurred outside the nucleus and, therefore, excluded Bacteria and Archaea from having such a circadian clock (Wulund and Reddy 2015). In fact, it was long thought that Bacteria and Archaea lacked a circadian rhythm at all (Mori et al. 1996; Golden et al. 1997). In addition to the absence of a nucleus, many bacteria have short generation times ranging from 20 minutes to several hours, shorter than a day. Remarkably, this idea was also picked up by social scientists who were interested with the conceptions of time and were especially concerned with the microbial way of telling time (Schrader 2020; Schrader 2021). How could an organism with a life span shorter than one day maintain a 24-h rhythm? Cell division would have to be independent of the circadian rhythm, while a mother cell would have to transfer the rhythmic state to her daughter cells. Rhythmic processes in Cyanobacteria were known for several decades before identifying these as circadian rhythms. These daily rhythms are found in natural cyanobacterial populations, and many of these observations are related to oxygenic photosynthesis and dinitrogen (N2) fixation. These two processes are, in principle, incompatible since the key enzyme for dinitrogen fixation, nitrogenase encoded by the three structural genes nifHDK, is irreversibly inhibited by oxygen (Robson and Postgate 1980). Especially in cyanobacteria-dominated biofilms, oxygen can reach supersaturating concentrations (Revsbech et al. 1983). Cyanobacteria evolved several adaptations to separate the process of dinitrogen fixation from photosynthesis. These adaptations include compartmentalization (spatial separation), motility between oxic and anoxic zones, and temporal separation by only fixing nitrogen when oxygen production is decreasing, which occurs between sunset and sunrise, or is completely ceased at night (Berman-Frank et al. 2003). Initial descriptions of diurnal patterns in nitrogenase activity and photosynthesis in cyanobacterial cultures (Millineaux et al. 1981; Mitsui et al. 1986; Grobbelaar et al. 1986) and Cyanobacteria-dominated communities (Horne and Goldman 1972; Bebout et al. 1987; Griffiths et al. 1987; Villbrandt et al. 1991) were not discussed in the light of a potential circadian clock mechanism and were mainly seen as a result of direct light/ dark driven activation of the responsible genes. Several criteria must be met for a rhythmic pattern to be under what is considered “true circadian control.” First, the observed rhythm should have an endogenous freerunning period that lasts approximately 24 h and must persist for several days under constant conditions (e.g. constant darkness or constant light). Second, it should be possible to entrain these rhythms. Hence, shifts in rhythms should eventually lead to

Daily Rhythmicity in Coastal Microbial Mats

367

a reset and resynchronization of the clock. Finally, the rhythms should exhibit temperature compensation, i.e., its activity should not be affected by changes in temperature. Normally, enzymes change kinetics based on temperature, but this is not the case with the clock proteins that keep a steady 24 h periodicity during temperature variations. Confirmation of the first criterion, a free-running period, was found for cyanobacterial nitrogenase activities in cultures of Oscillatoria sp. (Stal and Krumbein 1985) and Synechococcus (Grobbelaar et al. 1986) that were kept at, respectively, constant light or constant darkness. The unicellular cyanobacterium Synechococcus RF-1 maintained rhythmic nitrogenase activity for more than 4 days in the dark (Mitsui et al. 1986; Grobbelaar et al. 1986). In a follow-up study, Huang and Chow (1986) continued the analysis of Synechococcus RF-1 and showed that also the two other criteria – entrainment of the rhythm and temperature compensation were – met by this strain. Subsequently, genetic evidence was found for circadian-controlled gene expression in cyanobacteria using luciferase-coupled reporter genes (Kondo et al. 1994), which finally led to the identification of the molecular background of the cyanobacterial clock driven by the three key genes, kaiC, kaiA, and kaiB, encoding, respectively, for the key regulator KaiC that interacts with the transcription/translation machinery depending on its phosphorylated state, KaiA that enhances the rhythmic autokinase activity of KaiC, and KaiB that moderates KaiC’s autophosphatase activity. Further details about the molecular mechanism of cyanobacterial circadian rhythms are broadly covered in other chapters in this book. In this chapter we will focus on daily rhythmic patterns in natural and artificial coastal microbial mats. After introducing the system, we discuss two of our key publications addressing the potential circadian control in a natural system and the effects of different cultivation conditions on circadian-controlled gene expression.

2 Microbial Mats 2.1

Structure and Occurrence

One of the earliest documentations of a characteristic laminated microbial mat is a drawing (Fig. 1) published in 1813 in the Danish Flora (Flora Danica, Royal Danish Library). Similar microbial mats located on the North Sea beach of the Dutch barrier island of Schiermonnikoog (De Wit et al. 1989) and the German island Mellum (Stal et al. 1984; Gerdes et al. 1985) have been extensively studied for over 30 years. Benthic microorganisms often develop into a biofilm and may form macroscopic structures such as a microbial mat (Stal 2012). Microbial mats are immobilized phototrophic laminated ecosystems that are highly complex and contain a large diversity of microorganisms. The biochemical cycles of the essential elements, carbon, nitrogen, and sulfur, are largely closed in the microbial mats, which are, therefore, nearly self-sustainable (Berlanga and Guerrero 2016).

368

C. Hörnlein and H. Bolhuis

Fig. 1 Early drawing of a microbial mat. This drawing (1813) was printed in the Flora Danica and depicts a layered microbial mat and bundles of a filamentous cyanobacterial species, most likely Coleofasciculus chthonoplastes (formerly known as Microcoleus chthonoplastes). Image is published with the permission of the Royal Danish Library, Copenhagen, Fol. Top. Bot. Danmark:Flora Danica, Image title: Conferva Chthonoplastes Mert, Image ID: floradanica_1485.tif

Laminated microbial mats are considered to represent modern analogues to fossilized stromatolites, which are the oldest identified ecosystem dating back almost 3.5 billion years (Byerly et al. 1986). Microbial mats occur globally and often grow under extreme environmental conditions such as in deserts (Rasuk et al. 2016), hypersaline (Gerdes et al. 1993; Caumette et al. 1994) and alkaline environments (Schultze-Lam et al. 1996), hot springs (van der Meer et al. 2007), and in coastal and intertidal habitats (Nicholson et al. 1987; Stal 2012). In the latter habitats, microbial mats may be exposed to desiccation and to fluctuations of temperature, salinity, oxygen, and sulfide. Despite the fact that these mats are only a few millimeters thick, a wide range of physicochemical gradients are formed by incident light, temperature, and salinity, and by the metabolic activity of the microbial mat inhabitants. Along these gradients different functional groups arrange themselves in distinct layers (Jorgensen et al. 1983) displaying a colorful banding which has been named “Farbstreifen-Sandwatt” by early microbial mat researchers (Schulz 1936; Schulz and Meyer 1939). The uppermost brown/green layer contains diatoms and cyanobacteria and under certain conditions a purple layer of phototrophic sulfur bacteria develops right below the cyanobacteria (Fig. 2). The sulfide-smelling, anoxic, lower part of a mat is black due to the anaerobic reduction of ferric iron (Fe(III)) by sulfur-reducing bacteria. Driven by the daily cycle of light and dark, these bacterial consortia of functional groups recycle the elements within the mat

Daily Rhythmicity in Coastal Microbial Mats

369

Fig. 2 Simplified scheme of a microbial mat. Visualization of the microbial mat layers, their dominant functional groups, and the major pathways of energy and nutrient flow. Note that this scheme depicts the situation during the day where photosynthesis only oxygenates the top 2 millimeters of the mat. From sunset on, the mat becomes oxygen depleted and processes like fermentation and dinitrogen fixation will take place

during a 24 h cycle producing continuously shifting gradients of oxygen and sulfide. During the day, oxygenic photosynthesis saturates the mat with oxygen, and excess fixed inorganic carbon is exuded as low- and high-molecular weight compounds. These compounds are degraded by specialized chemoheterotrophic bacteria using oxygen as electron acceptor which, together with dark respiration by phototrophic organisms, depletes the mat of oxygen during the night. End oxidation of the organic matter is done by sulfate- and sulfur-reducing bacteria while producing sulfide. Sulfide is oxidized by the purple sulfur bacteria and other auto- and chemoheterotrophic bacteria (Fig. 2). Cyanobacteria are the major primary producers and initiate coastal microbial mat formation. Their ability to perform oxygenic photosynthesis and fixation of atmospheric dinitrogen (N2) allows them to colonize the nutrient-poor and nitrogendepleted environments in which microbial mats prosper (Stal and Noffke 2011). Through photosynthesis, Cyanobacteria use solar energy to convert CO2 and H2O into sugars and O2 and create an electron gradient that can drive a variety of transport processes while generating ATP via a proton ATPase (Lassen et al. 1994). A large part of the produced sugars cannot be used for growth directly because of other limiting nutrients and is exuded in the form of extracellular polymeric substances (EPS). This EPS forms the basis of the microbial mat matrix in which sand grains and cells are embedded and held together in an approximately 2–3 mm stable layer. The resulting firm structure promotes sediment stabilization and enables salt marsh formation, thereby contributing to natural coastal protection (Yallop et al. 1994). This coherent microbial mat structure can sometimes be lifted from the beach

370

C. Hörnlein and H. Bolhuis

Fig. 3 Coastal microbial mat. This picture, taken at the southern North Sea beach of Schiermonnikoog, reveals a typical coastal microbial mat with its signature “doormat-like” structure

sediment as if it were a doormat (Stal et al. 1985) (Fig. 3). Through photosynthesis and fermentation, the mat community is supplied with low- and high-molecular weight organic carbon, which can be degraded by anaerobic and chemoheterotrophic bacteria (Miyatake et al. 2014). At night, when oxygenic photosynthesis stops, oxygen is quickly depleted by the microbial community including the cyanobacteria. Sulfate- and sulfur-reducing bacteria end-oxidize the organic matter while producing sulfide (van Gemerden 1993). During the day, amongst others, sulfate-oxidizing bacteria such as anoxygenic, purple sulfur bacteria use sulfide as electron donor for photosynthesis and will oxidize sulfide to sulfur and eventually to sulfate (Stal and Noffke 2011).

2.2 2.2.1

Microbial Mats of the Southern North Sea Species Composition and Activity and Its Seasonal and Spatial Variation

Spatial and temporal variation of the coastal microbial mat community of the Dutch barrier island of Schiermonnikoog has been repeatedly assessed using various tools. The microbial community composition was determined by means of microscopy, lipid biomarkers, stable carbon and nitrogen isotopes, high-throughput amplicon sequencing of 16S ribosomal RNA and their encoding genes, and by other metagenomics/transcriptomics approaches (Dijkman et al. 2010; Bolhuis and Stal 2011; Cardoso et al. 2017, 2019; Hörnlein et al. 2018). The biodiversity of the coastal microbial mats is among the highest reported for a benthic microbial ecosystem (Bolhuis and Stal 2011) and is dominated by Proteobacteria, Cyanobacteria, and Bacteroidetes (Bolhuis and Stal 2011; Bauersachs et al. 2011; Cardoso et al. 2019). Molecular analysis involved both DNA- and RNA-based studies. In contrast to the short-lived RNA fraction that includes mostly molecules from active growing cells, the more resilient DNA fraction may also consist of dormant cells and

Daily Rhythmicity in Coastal Microbial Mats

371

extracellular DNA and may represent the resident fraction. DNA and RNA were extracted from microbial mat samples taken at a 4 h interval in a 24 h period. Amplicon sequencing of the 16S ribosomal RNA gene revealed considerable differences in community composition between the “resident” (DNA) and “active” (RNA) fraction. However, no significant variation in community composition was observed between the different time points during the 24 h of sampling (Cardoso et al. 2017). Cyanobacteria were highly represented in the active fraction, contributing up to 94% of the rRNA sequences but accounted for only 60% of the resident DNA sequences. Additional 16S-rRNA-based studies describing the spatial and seasonal variation in microbial community composition revealed a consistent numerical dominance of the phyla Proteobacteria, Cyanobacteria, and Bacteroidetes throughout the year. However, significant spatial and temporal differences were found at a finer taxonomic resolution (genus and species level) (Cardoso et al. 2019). Cardoso et al. (2019) reported that richness and diversity of the microbial mats increased from an initial pioneer population dominated by the alphaproteobacterial genus Loktanella and the cyanobacterial genus Nodularia to a more diverse climax population dominated by the cyanobacterial genera of Coleofasciculus and Nodosilinea. Bauersachs et al. (2011) used a combination of microscopy, lipid biomarkers, stable carbon and nitrogen isotopes, and 16S rRNA gene libraries to study the diazotrophic community of the North sea coastal microbial mats and confirmed the dominance of Cyanobacteria along the littoral gradient and observed a shift from more nitrogen utilization in the intertidal zone to more dinitrogen fixation in the supratidal zone (Bauersachs et al. 2011). Cyanobacteria may be the dominant dinitrogen fixing organisms in coastal microbial mats but the ability is also known from mat-dwelling gammaproteobacteria (Severin et al. 2010).

2.2.2

Rhythm on the Beach

Daily rhythmicity in cyanobacterial processes such as dinitrogen fixation, photosynthesis, respiration, fermentation, and migration were examined in a plethora of studies around the world (Richardson and Castenholz 1987; Happey-Wood and Jones 1988; Stal and Moezelaar 1997; Berman-Frank et al. 2003; Jonkers et al. 2003; Mitbavkar and Anil 2004; Steunou et al. 2008). One of the first observations describing a rhythmic process in a coastal microbial mat involved the daily patterns in dinitrogen fixation with highest activities at sunset and sunrise (Stal et al., 1984). Subsequent analysis provided one of the first examples of potential circadian control in Cyanobacteria (Stal and Krumbein 1985). These authors performed nitrogenase activity assays with cultures of the non-heterocystous cyanobacterium Oscillatoria sp. strain 23 growing under different light regimes. When transferring the cultures growing under a 16–8 h light dark to a 20–4 h light dark regime, they noticed that nitrogenase activity was initiated 4 h before the light was turned off, at the same time that previously the dark period would have started. In continuous light, nitrogenase activity continued with a free-running rhythmic period for several days thereby adhering to the first criterion of true circadian rhythmicity. In principle, this

372

C. Hörnlein and H. Bolhuis

observation is counterintuitive since light would result in photosynthesis and hence in nitrogenase inhibiting oxygen production. However, aerobic dinitrogen fixation in Oscillatoria sp. strain 23 in the light period is possible due to quick cessation of photosynthetic oxygen production during dinitrogen fixation (Stal and Krumbein 1987). Despite the observed free-running periodicity of dinitrogen fixation, these and other expression patterns were often considered a consequence of an on/off switch regulated by environmental cues. Potential involvement of a circadian clock was largely left out of the discussion since true circadian control is difficult to prove in a natural system. Moreover, dinitrogen fixation can be carried out by a plethora of Bacteria (sulfate-reducing deltaproteobacteria, some genera of Chromatiales and Chloroflexi, and some Cyanobacteria) and is also widespread in methanogenic Archaea and some species of anaerobic methane-oxidizing Euryarchaea (Offre et al. 2013). Therefore, despite the abundance of Cyanobacteria in microbial mats it is difficult to establish the actual responsible organism for the observed activities, especially for dinitrogen fixation (Severin et al. 2010). Furthermore, several taxonomic distinct groups of mat microorganisms are also involved in the major rhythmic process of photosynthesis. Examples of other photosynthetic organisms are the eukaryotic brown micro-algae (Bacillariophyceae, diatoms) and anoxygenic photosynthetic bacteria such as the purple sulfur bacteria (Chromatiales) and the green sulfur bacteria (Chloroflexi). Circadian control in prokaryotes is potentially not limited to Cyanobacteria given the widespread occurrence of kaiB and kaiC homologs in bacterial phyla such as Proteobacteria, Bacteroidetes, Chloroflexi, and also in Archaea (Dvornyk et al. 2003; Loza-Correa et al. 2010). Moreover, circadian-like rhythmicity could also be regulated by other potential endogenous rhythmicity inducers in prokaryotes (Min et al. 2005; Whitehead et al. 2009; Edgar et al. 2012; Ma et al. 2016). Circadian-like rhythmicity was suspected for bacterial populations in the open ocean that displayed a daily periodicity in their transcriptome-derived expression patterns (Ottesen et al. 2013). Indeed, significant daily periodicity was not only found in the cyanobacterial fraction that is dominated by the unicellular species Prochlorococcus MED4, but also in a variety of heterotrophic bacteria. For example, expression of genes derived from the alphaproteobacterium Roseobacter oscillated strongly during the day, especially in genes involved in bacteriochlorophyll-associated aerobic anoxygenic photosynthesis (Ottesen et al. 2013). The observed population-specific timing of peak transcript expression results in a daily variety of metabolic gene suites that were coined as multispecies waves of gene transcription and may influence both the tempo and mode of matter and energy transformation in the sea. Moreover, true circadian control in Procholorococcus is debatable since several known strains only contain kaiBC and do not possess the kaiA gene (Holtzendorff et al. 2008; Chew et al. 2018). Nevertheless, Procholorococcus marinus strain PCC 9511 shows daily rhythmicity but does not possess the free running characteristics, and oscillation is damped very quickly compared to closely related kaiABC containing cyanobacterial strains (Holtzendorff et al. 2008). Gene expression patterns in microbial assemblages of surface water at the Hawaiian Ocean Time-Series (Poretsky et al. 2009) revealed day/night correlated rhythmicity in cyanobacterial and alphaproteobacterial genes. Rhythmic interactions

Daily Rhythmicity in Coastal Microbial Mats

373

between Cyanobacteria and other bacteria were also observed in a hot spring microbial mat (van der Meer et al. 2005). Through the incorporation of [13C]labelled bicarbonate, it was shown that inorganic carbon was taken up by Cyanobacteria during the day and incorporated into glycogen. At sunset and during the night, the cyanobacteria ferment the glycogen and excrete the end-products acetate, formate, ethanol, and lactate that are metabolized by phototrophic green non-sulfur bacteria such as Chloroflexi. The observed daily rhythmicity in bacteria other than Cyanobacteria and the presence of kai gene homologs or other endogenous regulators may hint to a yet unexplored circadian clock-like feature. However, we developed an alternative and postulated the Choirmaster-Choir theory to explain the 24 h rhythmicity found throughout the members of a microbial mats (Hörnlein et al. 2018). This will be discussed in more detail below.

3 Choirmaster-Choir Theory The Choirmaster-Choir theory entails that microorganisms possessing a circadian clock (the Choirmaster) can induce community-wide rhythmic gene expression to organisms that by themselves are not rhythmic (the Choir). Fitting within this theory we formulated three possibilities: 1. Only Cyanobacteria and algae in the microbial mats have a fully functional circadian clock and direct their gene expression patterns to other mat members through the rhythmic release of photosynthates and other metabolites. 2. Other microbial mat members have their own molecular clock that is only entrained by the rhythmic release of metabolites from phototrophic microorganisms. 3. Other microbial mat members have their own clock that is entrained by a cocktail of zeitgebers such as light, temperature, and rhythmic release of photosynthate and metabolites by their neighbors.

3.1

The Choirmaster

In order to test our theory, we examined the natural 24 h periodicity in gene expression in a coastal microbial mat (Hörnlein et al. 2018). Samples were taken at 4 h intervals and subjected to metatranscriptome analysis via high-throughput RNA sequencing. It was found that 7% of the overall expressed genes revealed a significant rhythmic expression pattern. Suspected Choirmasters are the photoautotrophic cyanobacteria that are abundantly present and possess a circadian clock. Assembled and identified rhythmic genes of cyanobacterial origin were mainly involved in processes such as protein-

374

C. Hörnlein and H. Bolhuis

and DNA metabolism and in vitamin and pigment biosynthesis. Specific mapping of the transcriptome sequence pool to genomes of the dominant cyanobacteria in the mats, Lyngbya aestuarii and Coleofasciculus chthonoplastes, revealed further circadian-like rhythmicity of genes involved in carbohydrate turn-over, stress response, photosynthesis, and dinitrogen fixation. In a different experiment we applied quantitative PCR analysis targeting the L. aestuarii psbA and nifH genes involved in, respectively, photosynthesis and dinitrogen fixation (Hörnlein et al. 2020). Both genes displayed a distinct day/night separation in their peak expression (Fig. 4). The day-night separation of photosynthesis and dinitrogen fixation results in a separation in the release of excess metabolites. Moezelaar and colleagues (Moezelaar et al. 1996) showed that photosynthates, mostly polysaccharides, were mainly released during the day while nitrogen-containing compounds (e.g., NH3) and small-molecular-weight fermentation products such as acetate, lactate, and ethanol were released during the night. These fermentation products are essential substrates for various bacteria such as obligate anaerobic sulfate-reducing bacteria (Hansen 1994). Hence, the rhythmic release of these compounds feeds the microbial community at designated times during a day. Diatoms also contain a circadian clock (Ashworth et al. 2013) and are present in significant numbers in these mats (Bolhuis et al. 2013). They may, therefore, contribute to the rhythmic release of photosynthates during the day (Smith and Underwood 2001) and fermentation products like acetate, ethanol, and formate or glycerol during dark fermentation (Bourke et al. 2017).

3.2

The Choir

Similar to what has been found in the open ocean (Poretsky et al. 2009; Ottesen et al. 2013) and in hot spring microbial mats (van der Meer et al. 2005), rhythmic gene expression in the natural coastal microbial mats was not limited to Cyanobacteria but was also found in members of the Proteobacteria and Bacteroidetes (Hörnlein et al. 2018). Similar to what was observed in the open ocean (Ottesen et al. 2013), a member of the genus Roseobacter (potentially R. denitrificans) displayed a distinct circadian-like rhythmicity in the microbial mats in genes involved in protein metabolism, carbohydrate turnover, and photosynthesis. In fact, Roseobacter displayed a higher fraction of rhythmic genes than the circadian-clock-controlled C. chthonoplastes (Hörnlein et al. 2018). The rhythmic expression of non-cyanobacterial genes involved in carbohydrate turnover, utilization of polysaccharide, and fermentation products during the night are in sync with production of these substrates by the phototrophic organisms. This indicates a tight coupling in metabolism between the different mat members. A similar tight coupling was proposed for the previously mentioned interactions between Cyanobacteria and Chloroflexi (van der Meer et al. 2005) but was also

Daily Rhythmicity in Coastal Microbial Mats

375

Fig. 4 Gene expression patterns in L. aestuarii. Gene expression of psbA and nifH was measured in a natural coastal microbial mat (green line) (Hörnlein et al. 2018) and in laboratory cultures growing under different growth conditions and light regimes (Hörnlein et al. 2020). L. aestuarii was grown in liquid mono-culture, liquid co-culture (grown together with C. chthonoplastes), biofilm monoculture, and biofilm co-culture) (Hörnlein et al. 2020). The gradient plots show the expression of the genes during the light (light-blue) and dark (black) phases. The lower light-blue plots indicate data retrieved from cultures grown under continuous illumination. Error bars display standard deviations of biological triplicates (n ¼ 3). The samples were taken at 4 h intervals during a 24 h period

found for daily rhythmicity in gene expression in the animal gut microbiome as a response to the host feeding schedule (Trinder et al. 2015; Paulose et al. 2016). The tight interactions, daily rhythms, and diversity of the microbial mat led us to see the microbial mat as a single entity. This complex organism needs to have its

376

C. Hörnlein and H. Bolhuis

multiple subsystems cooperatively entrained to let the corpus (microbial mat) function optimally thereby following in the footsteps of Hans van Gemerden (1993), who called the intertwined metabolisms of the microbial mat community “a joint venture.”

4 Mimicking Microbial Mats A wealth of environmental and molecular information can be obtained by studying a natural microbial mat through metagenomics and metatranscriptomics. However, its enormous taxonomic diversity and functional complexity prohibits in-depth analysis of several aspects of species interactions. Not surprisingly most circadian-clockrelated studies are performed with pure cultures of single isolates of cyanobacteria. The organisms of choice are easily maintained, often unicellular, have a high growth rate, and are genetically easily accessible for the construction of dedicated gene knockout mutants, genetic recombination, and gene reporter studies. Unfortunately, most of the microbial mat cyanobacteria are difficult to isolate in pure culture, may change in genetic makeup upon prolonged propagation under laboratory conditions, and genetic tools are often not available. Moreover, the dominant species Lyngbya aestuarii and Coleofasciculus chthonoplastes are filamentous, of which the latter forms tight bundles packed in a polysaccharide sheet (Fig. 1). Their growth in a wellmixed aqueous growth medium has probably little relevance for growth in their natural complex ecosystem consisting of multiple other species and interactions therewith and its geochemical complexity. For example, Staphylococcus aureus and Vibrio cholerae changed their gene regulation patterns and metabolic activity when grown in the increased complexity of the growth environment (Resch et al. 2005; Moorthy and Watnick 2005). Daily rhythmic gene expression patterns in Prochlorococcus MED4 obtained from the open ocean transcriptome were also compared to a transcriptome obtained from a single MED4 isolate in culture (Ottesen et al. 2013). A fast majority of genes revealed similar expression levels in culture and nature. However, especially around noon, significant differences between culture and nature were observed in especially photosynthesis genes. Moreover, a large number of the genes that peak around noon could not be functionally annotated and probably encode novel, environmental specific proteins from novel ecotypes of the same strain. Another problem with analyzing natural samples is the high species diversity. Even at high sequencing depth, the high diversity in the coastal microbial mat resulted in a limited metatranscriptome that only uncovered the proverbial tip of the iceberg. Moreover, sampling biases in RNA extraction, RNAseq library preparation, and downstream analysis (e.g., recruitment analysis and species determination) introduce insecurities in the dataset that make predictions about the validity of the Choirmaster-Choir theory difficult. Our solution to these limitations is the construction of artificial microbial mats that can be analyzed under laboratory conditions. In contrast to the liquid-grown

Daily Rhythmicity in Coastal Microbial Mats

377

monocultures, we attempt to maintain the environmental complexity of growth in a substrate matrix (glass beads or sand) and reduce the species complexity. We are currently working with two different approaches, the so-called minimal microbial mats, consisting of random diluted low diversity communities and synthetic microbial mats consisting of a simple setup in which isolates representing different functional groups are combined. In our latest publication, we assess the impact of “added environmental complexity” on the expression of cyanobacterial circadian clock genes and genes under circadian control (Hörnlein et al. 2020). Two competing cyanobacterial species, L. aestuarii and C. chthonoplastes, were grown in mono- or co-culture, in liquid medium or as a biofilm on a glass bead support (ø 0.1 mm) to mimic the sandy matrix. The cultures were furthermore exposed to a 16–8 h light dark regime and subsequently incubated under free-running conditions at constant illumination. Gene expression levels under these conditions were assessed for both strains in triplicate by quantitative PCR. The selected genes consisted of the three circadian clock genes kaiABC and the circadian-clock-regulated genes psbA (photosynthesis), nifH (dinitrogen fixation), and ftsZ (cell division). In addition we looked at the universal circadian oscillator encoding gene prx (peroxiredoxin) (Edgar et al. 2012). Especially in C. chthonoplastes, expression of most genes is influenced by the imposed differences in growth conditions (Hörnlein et al. 2020). In L. aestuarii a free-running expression pattern was observed under continuous illumination for kaiABC, psbA, and nifH under nearly all growth conditions. In L. aestuarii also, a clear dark/light separation between the peak expression of psbA (day) and nifH (night) was observed and followed the expected patterns consistent with photosynthesis during the day and dinitrogen fixation in the dark. Most significant differences in expression patterns were observed when grown as biofilms. Although both cyanobacteria are functionally similar, they displayed a different 24 h transcriptional patterns in response to the treatments; apparently their circadian clocks adapted to their different life strategies. We concluded from this study that the laboratory-derived circadian clock expression patterns might be insufficient predictors for expression patterns in the field.

5 Outlook Cyclic processes are ubiquitous in microbial ecosystems but not all are necessarily controlled by a circadian clock mechanism. Cyclic processes can be a consequence of a fine-tuned “switch” reacting to subtle changes in environmental signals and gradients such as light, oxygen concentrations, and redox states. Absence of kaiABC homologs in non-cyanobacterial prokaryotes does not exclude a potential clock-like regulation with free running properties. The universal circadian oscillator, peroxiredoxin, appears to be a good candidate to fulfil this role and is present in each domain of life. Several other observations of daily cycles can be found in complex microbial ecosystems and in non-cyanobacterial species. For example, the

378

C. Hörnlein and H. Bolhuis

vertical migration of a white filamentous bacterial species in a coastal microbial mat followed a daily cyclic pattern as was established by time-lapse photography of a cross section of a microbial mat (Fig. 5). The white filaments contain clear intracellular inclusions that resemble elementary sulfur storage vacuoles (Preisler et al. 2007) characteristic of the sulfur-oxidizing bacterium Beggiatoa. Vertical migration is well known and has been previously described for Beggiatoa filaments in intact hypersaline microbial mats (Garcia-Pichel et al. 1994). What started as an art project in our laboratory to visualize the formation of the colored layering in microbial mats by time lapse photography led to some unexpected insight in the daily vertical migration and node formation of the potential Beggiatoa species. The filaments are found to be fully distributed over the 2 mm of the water column on top of the microbial mat during the light period but accumulate at the water surface in the dark, thereby forming dense nodes of entangled filaments (Fig. 5). Potentially this is an example of the Choirmaster-Choir theory where Beggiatoa cells react to the circadian clock-controlled production of oxygen by the Cyanobacteria during the light period and migrate to the surface at night for their oxygen demand. Another intriguing potential rhythmic feature is found on top of hypersalinesubmerged microbial mats that grow at approximately 10 cm under the brine water surface (Gerdes et al. 1993). These microbial mats are found in concentrator ponds of the solar salterns in Guerande, France. They are rich in cyanobacteria and are dominated by the cosmopolitan genus Coleofasciculus. These mats furthermore contains a large abundance of diatoms and bacteria involved in the sulfur cycle and in complex carbohydrate degradation. At a salinity of 10–15% (3–5 times higher salinity than seawater), intriguing ring structures are formed on top of the mats and are locally known as fairy rings (Fig. 6). These structures appear during the summer period and an additional, outward directed ring is formed each day. These rings resemble circadian-induced concentric growth rings found in, for example, the yeast Aureobasidium pullulans I (Franco et al. 2017). However, the fairy rings of the hypersaline mats of Guerande appear not to consist of actively growing cells but consists of a local EPS build-up on top of the mat. The formation of these EPS circles can be inhibited by blocking the sunlight and hence are formed by a rhythmic, lightdependent process. Potential causative agents can be anything from viruses, fungi, to toxin-producing bacteria. We are currently conducting a metagenome analysis of the rings to establish differences between the rings and the unaffected mats. Finally, our major challenge for the near future will be to find clear evidence for true circadian control of cyanobacterial gene expression in natural microbial mats and transfer of rhythms to non-circadian controlled microorganisms according to our Choirmaster-Choir theory. We are approaching this by using artificial microbial mat systems that have the full geochemical, structural, and functional complexity of a natural microbial mat but can be manipulated in the lab. These artificial mats have been developed and optimized to form the natural vertical gradients of light, oxygen, pH, and sulfide, which resulted in the characteristic colorful layering (manuscript in preparation). Metagenomic analysis confirmed the major functional diversity in relation to the carbon, nitrogen, and sulfur cycling as found in the natural ecosystem. The artificial mats are currently being tested at different levels of species diversity

Fig. 5 Rhythmic vertical migration in response to a light dark cycle. Frames from a time-lapse movie were taken at 4 h intervals and indicate the vertical movement of a white filamentous bacterium. This bacterium, identified as a Beggiatoa species, uses the full depth of the 2.5-millimeter water column on top of the microbial mat during the light period. When the light is off, Beggiatoa moves to the top 1 millimeter and forms intricate nodes. Red bars indicate the depth of the water column occupied by Beggiatoa filaments. The time is indicated in each frame in white (light period) or black (dark period). Each frame covers a dimension of approximately 12 by 10 millimeters of the top layer of a microbial mat. Images with permission and copyright © 2019 Wim van Egmond

Daily Rhythmicity in Coastal Microbial Mats 379

380

C. Hörnlein and H. Bolhuis

Fig. 6 Fairy rings on hypersaline microbial mats of Guerande, France. New concentric rings are formed outwards every day with a width of approximately 1 centimeter. These rings grow for several weeks and can cover a surface with a diameter of up to 1 meter. The mats are visually dominated by bundles of Coleofasciculus chthonoplastes and by yet unidentified diatom species. The salinity of the overlying water is in the order of 10–15% w.v. (100–150 gr/ kg)

(minimal mats). We have furthermore established cryopreservation protocols for long-term storage of microbial mat samples with minimal los of community structure. This allows us to establish replicate experiments and repeat experiments with different techniques. In the near future, these artificial mats will be manipulated in the lab to test the basic requirement for true circadian control of the cyanobacterial fraction: the free running oscillations, entrainment at different rhythms of dark and light, and to establish temperature compensation and its effect on the non-cyanobacterial population. Acknowledgments The authors especially want to thank and dedicate this chapter to Prof. Lucas Stal for his enormous contribution to understanding the ecophysiology of coastal microbial mats and for being one of the first to observe and document rhythmic processes of dinitrogen fixation in coastal microbial mats. This research was funded by a research program entitled “Metabolic networks and signalling in a synthetic microbial mat community,” awarded to H. Bolhuis by the Earth and Life Sciences program (ALW) of the Netherlands Organization of Scientific Research (NWO) under project number: 821.01.013. Further support was obtained from the European FP7 project MaCuMBA (Project number: 311975) of the European Commission, and additional fieldwork support was granted by the Academy Ecology Fund of the Royal Netherlands Academy of Arts and Sciences (KNAW). The work performed by Christine Hörnlein is part of her PhD project to be defended at the University of Amsterdam (Promotor prof. Lucas Stal). The research was performed at the Institute of Marine Microbiology of the Yerseke location of the Royal Netherlands Institute for Sea Research (Royal NIOZ). We are grateful to numerous lab members, technicians, and students for their support and contribution to this work.

Daily Rhythmicity in Coastal Microbial Mats

381

References Ashworth J, Coesel S, Lee A et al (2013) Genome-wide diel growth state transitions in the diatom Thalassiosira pseudonana. Proc Natl Acad Sci 110:7518–7523. https://doi.org/10.1073/pnas. 1300962110 Bauersachs T, Compaoré J, Severin I et al (2011) Diazotrophic microbial community of coastal microbial mats of the southern North Sea. Geobiology 9:349–359. https://doi.org/10.1111/j. 1472-4669.2011.00280.x Bebout BM, Paerl HW, Crocker KM, Prufert LE (1987) Diel interactions of oxygenic photosynthesis and N(2) fixation (acetylene reduction) in a marine microbial mat community. Appl Environ Microbiol 53:2353–2362 Berlanga M, Guerrero R (2016) Living together in biofilms: the microbial cell factory and its biotechnological implications. Microb Cell Factories 15:165 Berman-Frank I, Lundgren P, Falkowski P (2003) Nitrogen fixation and photosynthetic oxygen evolution in cyanobacteria. Res Microbiol 154:157–164. https://doi.org/10.1016/S0923-2508 (03)00029-9 Bolhuis H, Stal LJ (2011) Analysis of bacterial and archaeal diversity in coastal microbial mats using massive parallel 16S rRNA gene tag sequencing. ISME J 5:1701–1712 Bolhuis H, Fillinger L, Stal LJ (2013) Coastal microbial mat diversity along a natural salinity gradient. PLoS One 8:e63166. https://doi.org/10.1371/journal.pone.0063166 Bourke MF, Marriott PJ, Glud RN et al (2017) Metabolism in anoxic permeable sediments is dominated by eukaryotic dark fermentation. Nat Geosci 10:30–35. https://doi.org/10.1038/ ngeo2843 Byerly GR, Lower DR, Walsh MM (1986) Stromatolites from the 3,300-3,500-Myr Swaziland supergroup, Barberton Mountain land, South Africa. Nature 319:489–491. https://doi.org/10. 1038/319489a0 Cardoso DC, Sandionigi A, Cretoiu MS et al (2017) Comparison of the active and resident community of a coastal microbial mat. Sci Rep 7:1–10. https://doi.org/10.1038/s41598-01703095-z Cardoso DC, Cretoiu MS, Stal LJ, Bolhuis H (2019) Seasonal development of a coastal microbial mat. Sci Rep 9:1–14. https://doi.org/10.1038/s41598-019-45490-8 Caumette P, Matheron R, Raymond N, Relexans J-C (1994) Microbial mats in the hypersaline ponds of Mediterranean salterns (Salins-de-Giraud, France). FEMS Microbiol Ecol 13:273–286. https://doi.org/10.1111/j.1574-6941.1994.tb00074.x Chew J, Leypunskiy E, Lin J, Murugan A, Rust MJ (2018) High protein copy number is required to suppress stochasticity in the cyanobacterial circadian clock. Nat Commun 9:1–10. https://doi. org/10.1038/s41467-018-05109-4 De Wit R, Jonkers HM, Van Den Ende FP, Van Gemerden H (1989) In situ fluctuations of oxygen and sulphide in marine microbial sediment ecosystems. Netherlands J Sea Res 23:271–281. https://doi.org/10.1016/0077-7579(89)90048-3 Dijkman NA, Boschker HTS, Stal LJ, Kromkamp JC (2010) Composition and heterogeneity of the microbial community in a coastal microbial mat as revealed by the analysis of pigments and phospholipid-derived fatty acids. J Sea Res 63:62–70. https://doi.org/10.1016/j.seares.2009.10. 002 Dvornyk V, Vinogradova O, Nevo E (2003) Origin and evolution of circadian clock genes in prokaryotes. Proc Natl Acad Sci U S A 100:2495–2500. https://doi.org/10.1073/pnas. 0130099100 Edgar RS, Green EW, Zhao Y et al (2012) Peroxiredoxins are conserved markers of circadian rhythms. Nature 485:459–464. https://doi.org/10.1038/nature11088 Franco DL, Canessa P, Bellora N et al (2017) Spontaneous circadian rhythms in a cold-adapted natural isolate of Aureobasidium pullulans. Sci Rep 7:1–12. https://doi.org/10.1038/s41598017-14085-6

382

C. Hörnlein and H. Bolhuis

Garcia-Pichel F, Mechling M, Castenholz RW (1994) Diel migrations of microorganisms within a benthic, hypersaline mat community. Appl Environ Microbiol 60:1500–1511. https://doi.org/ 10.1128/AEM.60.5.1500-1511 Gerdes G, Krumbein WE, Reineck HE (1985) The depositional record of sandy, versicolored tidal flats (Mellum Island, southern North Sea). J Sediment Petrol 55:265–278 Gerdes G, Claes M, Dunajtschik-Piewak K et al (1993) Contribution of microbial mats to sedimentary surface structures. Facies 29:61–74. https://doi.org/10.1007/BF02536918 Golden SS, Ishiura M, Johnson CH, Kondo T (1997) Cyanobacterial circadian rhythms. Annu Rev Plant Physiol Plant Mol Biol 48:327–354. https://doi.org/10.1146/annurev.arplant.48.1.327 Griffiths MSH, Gallon JR, Chaplin ALANE (1987) The diurnal pattern of dinitrogen fixation by cyanobacteria in situ. New Phytol 107:649–657. https://doi.org/10.1111/j.1469-8137.1987. tb00903.x Grobbelaar N, Huang TC, Lin HY, Chow TJ (1986) Dinitrogen-fixing endogenous rhythm in Synechococcus RF-1. FEMS Microbiol Lett 37:173–177. https://doi.org/10.1111/j.1574-6968. 1986.tb01788.x Hansen TA (1994) Metabolism of sulfate-reducing prokaryotes. Antonie Van Leeuwenhoek 66:165–185. https://doi.org/10.1007/BF00871638 Happey-Wood CM, Jones P (1988) Rhythms of vertical migration and motility in intertidal benthic diatoms with particular reference to Pleurosigma angulatum. Diatom Res 3:83–93. https://doi. org/10.1080/0269249X.1988.9705018 Hardin PE, Hall JC, Rosbash M (1990) Feedback of the Drosophila period gene product on circadian cycling of its messenger RNA levels. Nature 343:536–540. https://doi.org/10.1038/ 343536a0 Holtzendorff J, Partensky F, Mella D, Lennon JF, Hess WR, Garczarek L (2008) Genome streamlining results in loss of robustness of the circadian clock in the marine cyanobacterium Prochlorococcus marinus PCC 9511. J Biol Rhythm 23:187–199. https://doi.org/10.1177/ 0748730408316040 Horne AJ, Goldman CR (1972) Nitrogen fixation in clear lake, California. I. Seasonal variation and the role of heterocysts. Limnol Oceanogr 17:678–692. https://doi.org/10.4319/lo.1972.17.5. 0678 Hörnlein C, Confurius-Guns V, Stal LJ, Bolhuis H (2018) Daily rhythmicity in coastal microbial mats. NPJ Biofilms Microbiomes 4:11. https://doi.org/10.1038/s41522-018-0054-5 Hörnlein C, Confurius-Guns V, Grego M, Stal LJ, Bolhuis H (2020) Circadian clock-controlled gene expression in co-cultured, mat-forming cyanobacteria. Sci Rep 10:14095 Huang RC (2018) The discoveries of molecular mechanisms for the circadian rhythm: the 2017 Nobel prize in physiology or medicine. Biom J 41:5–8 Huang T-C, Chow T-J (1986) New type of N 2 -fixing unicellular cyanobacterium (blue-green alga). FEMS Microbiol Lett 36:109–110. https://doi.org/10.1111/j.1574-6968.1986.tb01677.x Ishida N, Kaneko M, Allada R (1999) Biological clocks. Proc Natl Acad Sci U S A 96:8819–8820. https://doi.org/10.1073/pnas.96.16.8819 Jonkers HM, Ludwig R, De Wit R et al (2003) Structural and functional analysis of a microbial mat ecosystem from a unique permanent hypersaline inland lake: “La Salada de Chiprana” (NE Spain). FEMS Microbiol Ecol 44:175–189. https://doi.org/10.1016/S0168-6496(02) 00464-6 Jorgensen BB, Revsbech NP, Cohen Y (1983) Photosynthesis and structure of benthic microbial mats: microelectrode and SEM studies of four cyanobacterial communities1. Limnol Oceanogr 28:1075–1093. https://doi.org/10.4319/lo.1983.28.6.1075 Kondo T, Tsinoremas NF, Golden SS, Johnson CH, Kutsuna S, Ishiura M (1994) Circadian clock mutants of cyanobacteria. Science 266:1233–1236. https://doi.org/10.1126/science.7973706 Lassen C, Ploug H, Kühl M et al (1994) Oxygenic photosynthesis and light distribution in marine microbial mats. In: Microbial mats. Springer, Berlin, Heidelberg, pp 305–310 Loza-Correa M, Gomez-Valero L, Buchrieser C (2010) Circadian clock proteins in prokaryotes: hidden rhythms? Front Microbiol 1:130. https://doi.org/10.3389/fmicb.2010.00130

Daily Rhythmicity in Coastal Microbial Mats

383

Ma P, Mori T, Zhao C et al (2016) Evolution of KaiC-dependent timekeepers: a proto-circadian timing mechanism confers adaptive fitness in the purple bacterium Rhodopseudomonas palustris. PLoS Genet 12:e1005922. https://doi.org/10.1371/journal.pgen.1005922 Min H, Guo H, Xiong J (2005) Rhythmic gene expression in a purple photosynthetic bacterium, Rhodobacter sphaeroides. FEBS Lett 579:808–812. https://doi.org/10.1016/j.febslet.2005.01. 003 Mitbavkar S, Anil AC (2004) Vertical migratory rhythms of benthic diatoms in a tropical intertidal sand flat: influence of irradiance and tides. Mar Biol 145:9–20. https://doi.org/10.1007/s00227004-1300-3 Mitsui A, Kumazawa S, Takahashi A et al (1986) Strategy by which nitrogen-fixing unicellular cyanobacteria grow photoautotrophically. Nature 323:720–722. https://doi.org/10.1038/ 323720a0 Miyatake T, Moerdijk-Poortvliet TCW, Stal LJ, Boschker HTS (2014) Tracing carbon flow from microphytobenthos to major bacterial groups in an intertidal marine sediment by using an in situ 13C pulse-chase method. Limnol Oceanogr 59:1275–1287. https://doi.org/10.4319/lo.2014.59. 4.1275 Moezelaar R, Bijvank SM, Stal LJ (1996) Fermentation and sulfur reduction in the mat-building cyanobacterium Microcoleus chthonoplastes. Appl Environ Microbiol 62:1752–1758 Moorthy S, Watnick PI (2005) Identification of novel stage-specific genetic requirements through whole genome transcription profiling of Vibrio cholerae biofilm development. Mol Microbiol 57:1623–1635. https://doi.org/10.1111/j.1365-2958.2005.04797.x Mori T, Binder B, Johnson CH (1996) Circadian gating of cell division in cyanobacteria growing with average doubling times of less than 24 hours. Proc Natl Acad Sci U S A 93:10183–10188 Mullineaux PM, Gallon JR, Chaplin AE (1981) Acetylene reduction (nitrogen fixation) by cyanobacteria grown under alternating light-dark cycles. FEMS Microbiol Lett 10:245–247. https://doi.org/10.1111/j.1574-6968.1981.tb06249.x Nicholson JAM, Stolz JF, Pierson BK (1987) Structure of a microbiol mat at Great Sippewissett Marsh, Cape Cod, Massachusetts. FEMS Microbiol Lett 45:343–364. https://doi.org/10.1016/ 0378-1097(87)90021-8 Offre P, Spang A, Schleper C (2013) Archaea in biogeochemical cycles. Annu Rev Microbiol 67:437–457. https://doi.org/10.1146/annurev-micro-092412-155614 Ottesen EA, Young CR, Eppley JM et al (2013) Pattern and synchrony of gene expression among sympatric marine microbial populations. Proc Natl Acad Sci U S A 110:E488–E497. https://doi. org/10.1073/pnas.1222099110 Paulose JK, Wright JM, Patel AG, Cassone VM (2016) Human gut bacteria are sensitive to melatonin and express endogenous circadian rhythmicity. PLoS One 11:e0146643. https://doi. org/10.1371/journal.pone.0146643 Poretsky RS, Hewson I, Sun S et al (2009) Comparative day/night metatranscriptomic analysis of microbial communities in the North Pacific subtropical gyre. Environ Microbiol 11:1358–1375. https://doi.org/10.1111/j.1462-2920.2008.01863.x Preisler A, de Beer D, Lichtschlag A, Lavik G, Boetius A, Jørgensen BB (2007) Biological and chemical sulfide oxidation in a Beggiatoa inhabited marine sediment. ISME J 1:341–353 Rasuk MC, Fernández AB, Kurth D et al (2016) Bacterial diversity in microbial mats and sediments from the Atacama Desert. Microb Ecol 71:44–56. https://doi.org/10.1007/s00248-015-0649-9 Resch A, Rosenstein R, Nerz C, Götz F (2005) Differential gene expression profiling of Staphylococcus aureus cultivated under biofilm and planktonic conditions. Appl Environ Microbiol 71:2663–2676. https://doi.org/10.1128/AEM.71.5.2663-2676.2005 Revsbech NP, Jorgensen BB, Blackburn TH, Cohen Y (1983) Microelectrode studies of the photosynthesis and O2, H2S, and pH profiles of a microbial mat. Limnol Oceanogr 28:1062–1074. https://doi.org/10.4319/lo.1983.28.6.1062 Richardson LL, Castenholz RW (1987) Diel vertical movements of the cyanobacterium Oscillatoria terebriformis in a sulfide-rich hot spring microbial mat. Appl Environ Microbiol 53:2142–2150

384

C. Hörnlein and H. Bolhuis

Robson RL, Postgate JR (1980) Oxygen and hydrogen in biological nitrogen fixation. Annu Rev Microbiol 34:183–207. https://doi.org/10.1146/annurev.mi.34.100180.001151 Schrader A (2020) Marine Microbiopolitics: haunted microbes before the law. In: Braverman I, Johnson E (eds) Blue legalities. Duke University Press, London, pp 255–273 Schrader A (2021, In press) Reading science, caring with microbes in Methods and genealogies of new materialisms, eds. Iris van der Tuin and Felicity Coleman, Edinburg Press Schultze-Lam S, Ferris FG, Sherwood-Lollar B, Gerits JP (1996) Ultrastructure and seasonal growth patterns of microbial mats in a temperate climate saline-alkaline Lake: Goodenough lake, British Columbia, Canada. Can J Microbiol 42:147–161. https://doi.org/10.1139/m96-023 Schulz E (1936) Das farbstreifen-sandwatt und seine fauna, eine ökologische-biozönologische untersuchung an der nordsee. Kiel Meeresforsch:1–20 Schulz E, Meyer H (1939) Weitere Untersuchungen über das Farbstreifen-Sandwatt. Kieler Meeresforsch 1:321–336 Severin I, Acinas SG, Stal LJ (2010) Diversity of nitrogen-fixing bacteria in cyanobacterial mats. FEMS Microbiol Ecol 73:514–525. https://doi.org/10.1111/j.1574-6941.2010.00925.x Smith DJ, Underwood GJC (2001) The production of extracellular carbohydrates by estuarine benthic diatoms: the effects of growth phase and light and dark treatment. J Phycol 36:321–333. https://doi.org/10.1046/j.1529-8817.2000.99148.x Stal LJ, Noffke N (2011) Microbial mats. In: Gargaud M et al (eds) Encyclopedia of astrobiology. Springer, Berlin. https://doi.org/10.1007/978-3-642-11274-4_986 Stal LJ (2012) Cyanobacterial mats and stromatolites. In: Ecology of cyanobacteria II: their diversity in space and time. Kluwer Academic Publishers, Dordrecht, pp 65–125 Stal LJ, Krumbein WE (1985) Nitrogenase activity in the non-heterocystous cyanobacterium Oscillatoria sp. grown under alternating light-dark cycles. Arch Microbiol 143:67–71. https:// doi.org/10.1007/BF00414770 Stal LJ, Krumbein WE (1987) Temporal separation of nitrogen fixation and photosynthesis in the filamentous, non-heterocystous cyanobacterium Oscillatoria sp. Arch Microbiol 149:76–80. https://doi.org/10.1007/BF00423140 Stal LJ, Moezelaar R (1997) Fermentation in cyanobacteria. FEMS Microbiol Rev 21:179–211. https://doi.org/10.1111/j.1574-6976.1997.tb00350.x Stal LJ, Grossberger S, Krumbein WE, Division G (1984) Nitrogen fixation associated with the cyanobacterial mat of a marine laminated microbial ecosystem. Mar Biol 82:217–224 Stal LJ, Gemerden H, Krumbein WE (1985) Structure and development of a benthic marine microbial mat. FEMS Microbiol Lett 31:111–125. https://doi.org/10.1111/j.1574-6968.1985. tb01138.x Steunou A-S, Jensen SI, Brecht E et al (2008) Regulation of nif gene expression and the energetics of N2 fixation over the diel cycle in a hot spring microbial mat. ISME J 2:364–378. https://doi. org/10.1038/ismej.2007.117 Trinder M, Bisanz JE, Burton JP, Reid G (2015) Bacteria need “sleep” too?: microbiome circadian rhythmicity, metabolic disease, and beyond. Univ Tor Med J 92:52–55 van der Meer MTJ, Schouten S, Bateson MM et al (2005) Diel variations in carbon metabolism by green nonsulfur-like bacteria in alkaline siliceous hot spring microbial mats from Yellowstone National Park. Appl Environ Microbiol 71:3978–3986. https://doi.org/10.1128/AEM.71.7. 3978-3986.2005 van der Meer MTJ, Schouten S, Damsté JSS, Ward DM (2007) Impact of carbon metabolism on 13 C signatures of cyanobacteria and green non-sulfur-like bacteria inhabiting a microbial mat from an alkaline siliceous hot spring in Yellowstone National Park (USA). Environ Microbiol 9:482–491. https://doi.org/10.1111/j.1462-2920.2006.01165.x van Gemerden H (1993) Microbial mats: a joint venture. Mar Geol 113:3–25. https://doi.org/10. 1016/0025-3227(93)90146-M

Daily Rhythmicity in Coastal Microbial Mats

385

Villbrandt M, Krumbein WE, Stal LJ (1991) Diurnal and seasonal variations of nitrogen fixation and photosynthesis in cyanobacterial mats. Plant Soil 137:13–16. https://doi.org/10.1007/ BF02187426 Whitehead K, Pan M, Masumura KI et al (2009) Diurnally entrained anticipatory behavior in archaea. PLoS One 4:e5485. https://doi.org/10.1371/journal.pone.0005485 Wulund L, Reddy AB (2015) A brief history of circadian time: the emergence of redox oscillations as a novel component of biological rhythms. Perspect Sci 6:27–37. https://doi.org/10.1016/j. pisc.2015.08.002 Yallop ML, de Winder B, Paterson DM, Stal LJ (1994) Comparative structure, primary production and biogenic stabilization of cohesive and non-cohesive marine sediments inhabited by microphytobenthos. Estuar Coast Shelf Sci 39:565–582. https://doi.org/10.1016/S0272-7714 (06)80010-7

Daily and Seasonal Rhythms of Marine Phages of Cyanobacteria Gur Hevroni and Alon Philosof

Abstract Microbial communities are responsible for much of the primary and secondary production in the oceans and contribute to major global biogeochemical cycles. It is estimated that up to 50% of primary production in the ocean is made by marine phytoplankton. Since their discovery more than 30 years ago, the abundant marine unicellular Cyanobacteria of two genera Synechococcus and Prochlorococcus have been established as key members of marine microbial communities and as significant contributors to oceanic primary production. Marine cyanobacteria gene expression and cell division have been shown to exhibit a diel cycle, and their light-driven oscillations have been linked to community-wide diel cycles of gene expression, including those of heterotrophic members of the microbial communities. A growing body of evidence shows that cyanophages (viruses infecting cyanobacteria) are an important component of microbial assemblages in the oceans and that they show diel patterns of expression that are similar to those of their hosts. In this chapter we provide an overview of marine cyanobacteria and their phages and summarize current knowledge on their diel and seasonal patterns of abundance and activity. Finally, we discuss the ecological significance of these temporal patterns.

1 Cyanobacteria and Their Phages 1.1

Ecology of Cyanobacteria

The marine unicellular cyanobacteria Synechococcus (Johnson and Sieburth 1979; Waterbury et al. 1979) and Prochlorococcus (Chisholm et al. 1992; Partensky G. Hevroni (*) Department of Plant and Environmental Sciences, Weizmann Institute of Science, Rehovot, Israel A. Philosof (*) Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA, USA © The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 C. H. Johnson, M. J. Rust (eds.), Circadian Rhythms in Bacteria and Microbiomes, https://doi.org/10.1007/978-3-030-72158-9_20

387

388

G. Hevroni and A. Philosof

et al. 1999b) have attracted great interest since their discovery. As the most abundant photosynthetic microorganisms in the oceans (Ferris and Palenik 1998), they are estimated to contribute approximately 25% of the net global primary production (Flombaum et al. 2013). Prochlorococcus is overall more abundant and while these two genera are frequently found to co-occur, their global distribution patterns are different and they appear to inhabit specific niches (Flombaum et al. 2013; Kent et al. 2019). Prochlorococcus, the most abundant phototroph in the oceans, is found primarily in warm oligotrophic ocean gyres, between the 45 N and 40 S latitudes, (Partensky et al. 1999b) and is more abundant in the Pacific Ocean than the Atlantic Ocean (Kashtan et al. 2017). In addition, Prochlorococcus can be found as deep as 150–200 m below sea surface with phylogenetically distinct ecotypes (high light and low light adapted) occupying distinct niches in the water column (Rocap et al. 2003; West et al. 2001; Kashtan et al. 2014, 2017; Huang et al. 2012). In contrast, the closely related marine Synechococcus covers a wider global distribution and is mostly found in the upper layer of the euphotic zone. It is more abundant in nutrient-rich waters and it inhabits diverse regions including polar, tropical, coastal and temperate oceans (Partensky et al. 1999a; Huang et al. 2012; Paulsen et al. 2016). Overall, it has been shown that the regional distributions of Prochlorococcus and Synechococcus are controlled mainly by temperature, where the latter has a lower temperature boundary (Flombaum et al. 2013; Farrant et al. 2016). Furthermore, the light intensity and wavelength as well as iron availability appear to be fundamental in shaping niches for both genera (Grébert et al. 2018; Farrant et al. 2016). Despite being phylogenetically very close, Synechococcus is phylogenetically more diverse with several subclusters, of which the largest consists at least 10 distinct clades (Scanlan et al. 2009; Urbach and Chisholm 1998; Farrant et al. 2016). In contrast, the genus Prochlorococcus consists of seven clades, which correspond to the aforementioned ecotypes (Farrant et al. 2016; Scanlan et al. 2009). The Prochlorococcus genome has undergone a process of streamlining, which has been suggested to be the main reason for its widespread abundance (Ma et al. 2018; Scanlan et al. 2009; García-Fernández et al. 2004; Giovannoni et al. 2014; Kettler et al. 2007; Delmont and Eren 2018). Synechococcus and Prochlorococcus differ in their size (the former being larger on average) and in their light-harvesting complexes. In Synechococcus, the antennae are phycobilisomes rich in phycoerythrobilin (PEB) and phycourobilin (PUB). In contrast, Prochlorococcus lacks organized phycobilisomes and its antennae are comprised of thylakoid membrane proteins (PCB) that bind divinyl chl a, divinyl and monovinyl chl b and absorb light better in the blue wavelengths rather than green (Ting et al. 2002). While Prochlorococcus and Synechococcus require low levels of phosphorus (P) (Bertilsson et al. 2003), Prochlorococcus genomes from poor P environments are enriched in genes for P acquisition, phosphite and phosphonate utilization (Feingersch et al. 2012; Martiny et al. 2009). Neither genera can fix nitrogen; however, Synechococcus carries genes for nitrate assimilation (Scanlan et al. 2009). In Prochlorococcus however, only one high-light adapted clade and one low-light adapted clade have been found to have the potential to utilize nitrate

Daily and Seasonal Rhythms of Marine Phages of Cyanobacteria

389

(Berube et al. 2015). Furthermore, mixotrophy in the form of glucose and organic carbon (del Carmen Muñoz-Marín et al. 2013; Yelton et al. 2016) as well as amino acids uptake (Gomez-Pereira et al. 2013; Zubkov et al. 2003) has been observed to be globally widespread in diverse phylogenetic groups of Prochlorococcus and Synechococcus (Yelton et al. 2016), suggesting more complex lifestyles than originally expected. In addition, dimethylsulphoniopropionate (DMSP) is also consumed by both and Synechococcus has been shown to increase assimilation in the light (Malmstrom et al. 2005; Vila-Costa et al. 2006). Autophototrophic bacteria and heterotrophic bacteria are hypothesized to exchange public goods, or molecules for collective benefit (such as glycolate and pyruvate exerted by Prochlorococcus and malate by SAR11) and may have coevolved with symbiotic relations (Braakman et al. 2017). In addition, it has long been observed in cultures in the laboratory that Prochlorococcus benefits from the presence of some heterotrophs (Rippka et al. 2000; Morris et al. 2008; RothRosenberg et al. 2020a). One example is that Prochlorococcus, which is sensitive to hydrogen peroxide, relies on a heterotroph for H2O2 scavenging as Prochlorococcus lacks the genes for degrading it (Ma et al. 2018; Morris et al. 2011, 2012). In turn, marine Prochlorococcus and Synechococcus contribute to the dissolved organic matter (DOM) pool in the oceans (Becker et al. 2014; RothRosenberg et al. 2020b; Zheng et al. 2020), and some environmentally abundant strains carry genes for synthesizing secondary-metabolites (such as lantipeptides) which are expected to impact their immediate heterotrophic community (Li et al. 2010). Diel rhythms of gene expression of heterotrophic bacteria in the oceans have been suggested to be timed by the diel metabolic activity of photoautotrophs, of which a major proportion are Cyanobacteria (Poretsky et al. 2009; Ottesen et al. 2014; Aylward et al. 2015).

1.2

Diel and Seasonal Patterns in Marine Cyanobacteria

Diel light/dark synchronization has been observed in both Prochlorococcus and Synechococcus (Partensky et al. 1999b; Mella-Flores et al. 2012; Sweeney and Borgese 1989). The Prochlorococcus cell cycle has distinct S, G1 and G2 phases (Partensky et al. 1999b), and its S phase of DNA replication starts at dusk and the cell division takes place early at night (Vaulot et al. 1995; Jacquet et al. 2001). In contrast, (Ribalet et al. 2015) reported consistent Prochlorococcus cell division taking place at dusk and peak cell mortality at night. It has also been observed that regardless of light or temperature conditions, certain cell conditions are also required for replication initiation (Holtzendorff et al. 2008). Increased amino acids uptake has been shown to increase during the S stage, both in culture and in environmental samples (Mary et al. 2008). While Prochlorococcus cells were expected to undergo a single division per day (Vaulot et al. 1995), additional observation suggested ultraradian growth (Shalapyonok et al. 1998) in Prochlorococcus. In any case, Prochlorococcus distribution in the euphotic zone has been linked to differences

390

G. Hevroni and A. Philosof

in optimal light for growth in different ecotypes, with high-light adapted strains showing the highest relative growth rates (Moore et al. 1998; Moore and Chisholm 1999). A similar observation in Synechococcus WH7803 suggested that the circadian clock is also temperature compensated (Sweeney and Borgese 1989). In addition, UV was shown to have a negative effect on cell division (Jacquet et al. 1998). In Synechococcus WH8101, WH8102, WH8103 and WH7805, a biomodal intra-cellular DNA abundance has been observed (Holtzendorff et al. 2008; Binder and Chisholm 1995) and was linked to slow growth. In contrast, in Synechococcus WH7803, an asynchronized multimodal phenotype with asynchronous replication has been observed (Binder and Chisholm 1995). Diel oscillations of the elemental composition of Synechococcus WH8102 grown in a chemostat have been reported with increased C and N quota, relating to increased growth rate, when the cells were grown under light conditions. In contrast, P content remained mostly unchanged (Lopez et al. 2016). In addition, both genera have been shown to exhibit diel patterns of chl fluorescence with onset during the light hours. In Prochlorococcus chl decreases with cell division in the dusk, while in Synechococcus the decrease starts earlier, in the light (Jacquet et al. 2001). Synechococcus has been observed to show diel transcriptional activity in metatranscriptomic data from coastal and open ocean samples where cell division-related transcripts peaked at early night (Kolody et al. 2019; Ottesen et al. 2013). In addition, energy metabolism genes, involved in ATP synthesis and carbon fixation, peaked in the morning (Ottesen et al. 2013). The molecular clock genes, KaiA and KaiC, also showed diel oscillations in expression in the same study, while KaiB did not, perhaps due to low read coverage. Prochlorococcus showed diel patterns of gene expression, with a significant number of transcripts peaking at midday (Ottesen et al. 2014). In open ocean samples, these diel expression patterns have been shown to be central to metabolic networks of the entire microbial community (Aylward et al. 2015). Prochlorococcus diel transcription patterns have been detected throughout the euphotic zone, with levels decreasing as light intensity drops (Vislova et al. 2019). These networks appear to be structured around photoautotrophic diel cycles as transcripts from heterotrophic members of the microbial community also showed diel oscillations, potentially timed by the release of DOM by photoautotrophs (Aylward et al. 2015). Both Prochlorococcus and Synechococcus show seasonal changes in abundance in diverse oceanic regions (Giovannoni and Vergin 2012) as well as in cell division rates (Hunter-Cevera et al. 2014). It was hypothesized that this seasonality is driven primarily by water temperature and day length (Hunter-Cevera et al. 2016, 2020; Lambert et al. 2019). Synechococcus has been shown to be negatively correlated with high chl levels that indicate high eukaryotic phytoplankton abundance (Robidart et al. 2012; Paerl et al. 2012). In the Gulf of Aqaba, seasonal mixing leads to winter/spring eukaryotic phytoplankton bloom followed by Synechococcus blooms. These blooms are then followed by summer water column stratification that leads to Prochlorococcus taking dominance, which are then replaced again by Synechococcus as the water undergoes mixing again towards the winter (Lindell and Post 1995; Al-Najjar et al. 2007). Similarly, in the long-term Bermuda Atlantic

Daily and Seasonal Rhythms of Marine Phages of Cyanobacteria

391

Time-series Study (BATS), Synechococcus and Prochlorococcus also show seasonal variation that is more pronounced in the upper euphotic zone. In the summer, when the water is stratified, Prochlorococcus is the dominant photosynthetic group with peaks at deep chlorophyll maximum (DCM), while Synechococcus does not show an autumn peak (Fuhrman and Steele 2008). In the San Pedro Ocean Timeseries (SPOT), Synechococcus was present year round with an increase in abundance following upwelling and increased productivity (Chow et al. 2013). Both genera show a summer peak in abundance (Needham and Fuhrman 2016). At this site, higher seasonal signals have been observed at the surface rather than in deeper regions of the photic water column (Cram et al. 2015). In contrast, in the Hawaii Ocean Time-series (HOT), Prochlorococcus is abundant throughout the year showing only weak annual variance (Malmstrom et al. 2010). Synechococcus has also been shown to undergo seasonal clade succession in coastal samples from Southern California (Tai and Palenik 2009), Chesapeake Bay (Cai et al. 2010) and the northeast USA (Hunter-Cevera et al. 2020).

1.3

Genomics of the Molecular Clock in Marine Cyanobacteria

The well-studied molecular clock system characterized in the freshwater cyanobacterium Synechococcus elongatus has been established as a model system for circadian clocks in cyanobacteria over the past 30 years (Johnson et al. 2017). This three-gene protein oscillator, KaiA (found exclusively in Cyanobacteria), KaiB and KaiC, (Axmann et al. 2014), times a vast range of cellular processes including gene expression and cell cycle (Ishiura et al. 1998; Cohen and Golden 2015; Kondo et al. 1994; Mori et al. 1996). The molecular clock is suggested to confer a fitness benefit to marine Prochlorococcus and Synechococcus in the oceans, albeit at low latitudes this benefit is expected to be weaker (Hellweger et al. 2019). kaiC, the core of this system, undergoes intrinsic circadian (~24 h) phosphorylation cycles and is regulated by KaiA and KaiB (Xu et al. 2003; Nishiwaki et al. 2004). Environmental entraining input to this system is not directly via photoreceptors but rather through the sensing of the cell’s redox state, which is a function of photosynthetic activity (Cohen and Golden 2015; Wood et al. 2010; Rust et al. 2011). While marine Synechococcus harbours all three kai genes (Axmann et al. 2014), in Prochlorococcus, the molecular clock differs significantly from the model system (Cohen and Golden 2015) and to date, all published genomes lack kaiA but harbour both kaiB and kaiC (see next paragraph). Additionally, the input factors PrkE, Pex, CikA and the output factors LabA and LalA are missing from both marine Prochlorococcus and Synechococcus (Schmelling et al. 2017). Finally, a single input factor, LdpA, an output factor SasA and its response regulator RpaA are present in both Prochlorococcus MED4 and Synechococcus WH7803 (Axmann et al. 2009, 2014).

392

G. Hevroni and A. Philosof

It has been suggested that Prochlorococcus once possessed the KaiA gene, but it was lost during genome streamlining in a step-wise deletion manner (Holtzendorff et al. 2008; Axmann et al. 2009; Dufresne et al. 2005). In Prochlorococcus MED4, growth and the majority of genes show diurnal cycles of expression with peaks at dawn and dusk (Waldbauer et al. 2012; Zinser et al. 2009). In addition, a divergence in both timing and magnitude between the mRNA levels of a significant number of genes and their protein levels has also been reported (Waldbauer et al. 2012). When moved from dark-light cycles to constant light conditions, the diurnal rhythms of Prochlorococcus disappeared, and it has been suggested that this could be due, in part, to the lack of CikA (Holtzendorff et al. 2008; Schmelling et al. 2017). kaiC has been shown (in vitro) to be an ATPase and autokinase with KaiB influencing ATPase activity (Terauchi et al. 2007), suggesting that regulation of ATPase activity is the main function of this simple two-gene system in Prochlorococcus MED4 (Axmann et al. 2009). The resulting timing mechanism in Prochlorococcus is thus not robust and has been likened to an hour-glass as it runs for one diurnal cycle that has to be reset daily by environmental signals (Axmann et al. 2009; Holtzendorff et al. 2008).

1.4

Cyanophages: Viruses of Cyanobacteria

Bacteriophages, viruses that infect and eventually lyse bacteria, have a vast influence on the fate of bacterial primary and secondary production and on the population dynamics and diversity of bacterial populations (Suttle 2007; Fuhrman 1999; Zimmerman et al. 2019; Rodriguez-Brito et al. 2010; Ahlgren et al. 2019; Lara et al. 2017). In recent years, accumulating evidence has shown that marine viruses comprise the largest reservoir of genetic diversity in the oceans and are major players in oceanic processes and biogeochemical cycles (Brum et al. 2015; Paez-Espino et al. 2016; Zimmerman et al. 2019; Gregory et al. 2019; Dion et al. 2020) as well as an important component in carbon export networks (Guidi et al. 2016; Zimmerman et al. 2019). Marine cyanophages, bacteriophages infecting marine Synechococcus (Suttle and Chan 1993; Waterbury and Valois 1993; Wilson et al. 1993) and Prochlorococcus (Sullivan et al. 2003), were discovered 30 years after the discovery of phages that infect freshwater Cyanobacteria (Safferman and Morris 1963; Safferman et al. 1972). Their discovery led to a surge in research interest and to the establishment of many cyanophage-host systems in the decades that followed (Suttle 2000). To date, all described cyanophages belong to the Caudovirales (tailed) dsDNA viruses of the families T4-like Myoviridae, T7-like Podoviridae and lambda-like Siphoviridae (reviewed in (Mann and Clokie 2012)). Cyanomyophages appear to have overall a broader host range compared to podo- and sipho-cyanophages (Sullivan et al. 2003; Dekel-Bird et al. 2014). The ecological interest in studying cyanophages has been spurred by observations that virus-like particles (VLPs) are highly abundant in the marine environment and can exceed bacterial cells up to a 100-fold, with a median of a 10 to 1 ratio (Parikka

Daily and Seasonal Rhythms of Marine Phages of Cyanobacteria

393

et al. 2017; Knowles et al. 2016; Wigington et al. 2016). Cyanophages often dominate environmental samples (Zhao et al. 2013; Philosof et al. 2017), and it is estimated that a significant proportion, up to 30%, of marine cyanobacteria are killed by phages daily (Suttle and Chan 1994; Fuhrman 1999). Cyanophages can affect the structure of the microbial community in the ocean (Fuhrman et al. 2015) by driving diversification of Cyanobacteria (Mühling et al. 2005; Avrani et al. 2011; Kashtan et al. 2014; Marston et al. 2012). In addition, cyanophage-induced lysis of cyanobacteria can lead to a ‘viral shunt’ or the release of DOM that becomes available to other microorganisms instead of being consumed by higher trophic levels (Fuhrman 1999; Zhang et al. 2014). It has been suggested that the nature of the DOM released due to viral lysis can also have an effect on the microbial community since it can be utilized by specific microbial groups (Zhao et al. 2019; Shelford et al. 2012) as they are rich in P and N (Jover et al. 2014). In addition, cyanophage-induced lysis of Cyanobacteria can promote sinking of DOM in a particulate form (Fuchsman et al. 2019; Breitbart et al. 2018). Phage infection is not at a steady-state as infection rates grow in conjunction with host abundance, and the most affected taxa appear to be also the most abundant (Suttle 2007; Rodriguez-Brito et al. 2010; Thingstad 2000; Aylward et al. 2017) (Fig. 1a). Furthermore, infection was hypothesized to relate to the trophic condition of the host (Weitz and Dushoff 2008). The discovery of genes in the genomes of marine cyanophages, which are of cyanobacterial origin, encoding core photosystem II (PSII) subunits, suggested a unique interaction with the host metabolic machinery and hinted at the ecological importance of these phages (Mann et al. 2003a, b; Sullivan et al. 2003; Lindell et al. 2004). Over the last decade, a growing repertoire of additional bacterial homologous, auxiliary metabolic genes (AMGs) have been found in genomes of cyanophages (Puxty et al. 2015; Philosof et al. 2011; Crummett (a)

(b)

Fig. 1 (a) Schematic representation of predator-prey like temporal oscillations in abundance of Cyanobacteria and viral communities where there is a lagged peak in phage abundance following a peak in host abundance. (b) Conceptual rank abundance curve of a viral community at any given time, as predicted by the Bank model which describes a heavy-tailed distribution of low abundance taxa B and a small group of highly abundant taxa A. The former can act as a seed bank with seasonal changes in abundance rank of taxa from this group to the active one (Breitbart and Rohwer 2005; Breitbart et al. 2018)

394

G. Hevroni and A. Philosof

et al. 2016). Among these are cassettes of photosystem I (PSI) genes (Fridman et al. 2017; Roitman et al. 2015; Hevroni et al. 2014), electron transport genes (Sharon et al. 2011; Alperovitch-Lavy et al. 2011; Millard et al. 2009), phosphate acquisition genes (Kelly et al. 2013; Zeng and Chisholm 2012), highlight inducible protein genes (Lindell et al. 2004; Clokie and Mann 2006), lipid desaturase genes (Roitman et al. 2018) and nucleic acid and carbon metabolism genes (Enav et al. 2014; Thompson et al. 2011; Sullivan et al. 2005). Interestingly, these AMGs often differ from the hosts’ versions in sequence, and cyanophages introduce innovations into these AMGs such as gene fusions or rearrangements into operons (Philosof et al. 2011; Alperovitch-Lavy et al. 2011; Crummett et al. 2016; Thompson et al. 2011; Zeng and Chisholm 2012).

1.5

Ecology of Cyanophage-Cyanobacteria Dynamics

Viruses depend solely on their hosts for reproduction, and their abundance in the environment can be described as a power law function of the microbial cell abundance (Wigington et al. 2016). The coexistence of hosts and their lytic phages (Waterbury and Valois 1993) has been explained by a co-evolutionary arms race model (Bohannan and Lenski 2000). In this model, there are cycles of hosts developing resistance while the phages acquire mutations to overcome it. These dynamics often last only a limited number of cycles (Woolhouse et al. 2002; Avrani et al. 2012). In addition, a Red Queen scenario can then develop (Van Valen 1973) in which the host and the phage undergo cycles of co-evolution, adaptation and counter adaptation, but essentially remain at the same evolutionary place relative to one another (Woolhouse and Webster 2000; Paterson et al. 2010). Some cyanophages (generalists) can infect multiple cyanobacteria hosts, while others (specialists) only infect a specific host (Sullivan et al. 2003; Dekel-Bird et al. 2014). Cyanomyophages often have a broader host range compared to podo- and sipho-cyanophages (Sullivan et al. 2003; Dekel-Bird et al. 2014; Enav et al. 2012). Having also a wider range of G + C content and carrying tRNA genes, these cyanomyophages can adapt to the host codon usage during infection (Enav et al. 2012; Bailly-Bechet et al. 2007). In contrast cyanopodophages have G + C and codon usage that is adjusted to their hosts (Enav et al. 2012). These two types have demonstrated different evolutionary effects on the hosts (Avrani et al. 2011; Marston et al. 2012; Lennon et al. 2007; Avrani and Lindell 2015; Fedida and Lindell 2017) and the phages (Enav et al. 2018; Zborowsky and Lindell 2019; Schwartz and Lindell 2017). A generalist cyanomyophage (syn9) was shown to have similar infection cycles and temporal gene expression patterns (divided into early, middle and late genes) when infecting three different Synechococcus hosts (Doron et al. 2016). Surprisingly, the three hosts’ gene expression profiles were different from one another, suggesting different mechanisms to combat the infection. In addition, phages differ in their life-history traits, that is, in the length of the latent period (the time from infection to lysis), burst size and adsorption rate and other infection characteristics (Zborowsky and Lindell

Daily and Seasonal Rhythms of Marine Phages of Cyanobacteria

395

2019; Lindell et al. 2007; Sullivan et al. 2003; Suttle and Chan 1994). The fitness costs and mutations associated with developing host resistance to infection, along with seasonal changes in the abundance of potential Cyanobacteria hosts (due to changes in environmental conditions), are also expected to promote compositional changes in the cyanophage community (Rodriguez-Valera et al. 2009; Schwartz and Lindell 2017). This phenomenon has been generalized by (Thingstad 2000), suggesting that bacterial species abundance is controlled in a top-down manner, where the most abundant species will be lysed by phages at a higher rate, lowering the host abundance (“kill-the-winner”) (Thingstad 2000) (Fig. 1a). The resulting rank-abundance curve is heavy-tailed and can be viewed as a seed-bank scenario in which a small group of active phage taxa is largely outnumbered by the inactive phage taxa in the seed-bank (Breitbart and Rohwer 2005) (Fig. 1b).

2 Temporal Abundance Patterns of Marine Cyanophages Interactions between cyanophages and their cyanobacteria hosts are governed by various factors. Key among them are the abundance of the host which determines the rate of encounter (Wilcox and Fuhrman 1994) and the host physiological state that affects the intra-cellular progression of the infection cycle (Weitz and Dushoff 2008). As expected, in phages infecting photosynthetic microorganisms, these factors are also affected by the host diel cycles and light availability (Clokie and Mann 2006). As described earlier (Sect. 1.2), cell cycle and gene expression show diel oscillations in marine Synechococcus and Prochlorococcus. In the latter, an hourglass type of diel oscillator damps under constant conditions, implying its reliance on light/dark cycles for its ecology. Reports on the effects of light on the infection cycle of cyanophages span more than four decades (Cseke and Farkas 1979; Clokie and Mann 2006), but only recently has the ecological importance of these patterns has started to become clear. In this part we discuss how changes in light, on a diel and seasonal time-scale, translate into changes in cyanophage abundance. In addition, we discuss how seasonal changes in the abundance of Cyanobacteria affect cyanophage abundance.

2.1

The Effect of Light on Diel Oscillations of Cyanophage Infection

Early in cyanophage research, it was known that the total amount of cyanophage AS-1 attached to the freshwater cyanobacterium Synechococcus elongatus PCC 7942 was twice as high in illuminated cultures compared to those grown in the dark (Cseke and Farkas 1979). Similar observations were made on the attachment of cyanophage S-PM2, in a wavelength-dependent manner, to the marine

396

G. Hevroni and A. Philosof

Synechococcus sp. WH7803, while another group of cyanophages were found to be able to adsorb to the cell independently of light (Jia et al. 2010). The survival of free VLPs outside of the host cell is also affected by light. UV light was found to decrease infectivity and promote virus decay in the marine environment (Mojica and Brussaard 2014; Suttle and Chan 1994; Suttle and Chen 1992; Weinbauer et al. 1999), with seasonal increase in the ability to tolerate high UV reported for a cyanomyophage infecting a marine Synechococcus (Garza and Suttle 1998). The relationship between cyanophages and light gained new interest in the early 2000s following the discovery of genes in numerous cyanophage genomes that are homologous to the cyanobacterial psbA and psbD genes which encode the D1 and D2 proteins of the photosynthetic reaction centres (photosystem II) (Mann et al. 2003a, b; Lindell et al. 2004; Millard et al. 2004; Sullivan et al. 2006). Phylogenetic analysis of the cyanophage photosynthetic genes suggest they were horizontally acquired from their cyanobacterial host on several occasions throughout their mutual evolution (Zeidner et al. 2003, 2005; Lindell et al. 2004; Millard et al. 2004; Sullivan et al. 2006). During infection, translation and transcription of host genes decrease, to various degrees, as phage gene expression levels increase (Rohwer and Thurber 2009; Doron et al. 2016; Lindell et al. 2007). Several key studies have provided evidence that cyanophages that carry these genes actively maintain energy production during infection to support the completion of the infection cycle. Cyanopodophage PSSP7 PSII subunit D1 has been shown to be expressed during infection of Prochlorococcus MED4, most likely to replace the labile, and rapidly turned-over, host D1 subunit (Lindell et al. 2005; Thompson et al. 2011). Furthermore, initial evidence also suggests cyclic photosynthesis is enhanced in Prochlorococcus cells infected with a phage carrying a unique 7 gene cassette of different PSI genes (Fridman et al. 2017; Roitman et al. 2015). In addition, cyanophages carrying a Calvin cycle inhibitor and pentose phosphate pathway (PPP) genes were shown to reduce Calvin cycle activity and increase NADPH/NADP ratio during infection (Thompson et al. 2011). During infection of Synechococcus with two different cyanophages, measurement of photosynthetic yield was shown to increase 2 h post infection (Puxty et al. 2018) and CO2 fixation was inhibited while photosynthetic electron transport remained active (Puxty et al. 2016). Higher rates of N acquisition from the extracellular environment were measured, during infection of Synechococcus WH8102, as the infection progressed. In addition, higher incorporation of extracellular N into psbA in high-light compared to medium- and low-light conditions suggest faster turnover rates or overall more synthesis of this protein (Waldbauer et al. 2019). Furthermore, the long-standing use of photosynthesis inhibitors such as DCMU (3-(3,4-dichlorophenyl)-1,1-dimethylurea) has provided additional proof of the link between the photosynthetic activity of a cyanobacterial host and the outcome of infection by a cyanophage under laboratory conditions (Sherman and Haselkorn 1971; Benson and Martin 1981; Lindell et al. 2005; Liu et al. 2019). Cyanophage AS-1 infection of the freshwater Synechococcus elongatus PCC 7942 was reported to correlate with light intensity and to occur in a diel pattern

Daily and Seasonal Rhythms of Marine Phages of Cyanobacteria

397

(Kao et al. 2005). However, only minor effects on cyanophage progeny production were observed when mutating the host’s circadian regulatory proteins. This suggested that the diel periodicity of infection in this system was driven by lightdependent changes in the physiology of the host cell, rather than the circadian cycle. In a marine system of Synechococcus and the cyanophage S-PM2, adsorption was found to be dependent on light but not on photosynthetic activity of the host, or its circadian rhythm (Jia et al. 2010). Light was also found to increase cyanophage production (Lindell et al. 2005; Thompson et al. 2011, 2016) with high light intensities resulting in an earlier burst event in Synechococcus (Puxty et al. 2018). Recent evidence suggests a variety of diel light-dependent life history traits. Three cyanophages (P-HM2, P-SSM2, and P-SSP7) infecting Prochlorococcus were found by Liu and co-workers (Liu et al. 2019) to have life history traits that range from phages that do not adsorb to their host at all in the dark, to those that adsorb in the dark but do not transcribe viral genes, to those that both adsorb and transcribe viral genes in the dark. Interestingly, in the case of the latter group, the timing of progeny release was similar in both light- and dark-grown cultures. However, the dark cultures produced a significantly smaller burst size, and the gene expression patterns of the three cyanophage groups showed an increase in transcript abundance in light and decrease in dark (Liu et al. 2019). Using P-HM2 and P-SSP7 to model the interaction with Prochlorococcus, Demory and co-workers showed light-dependent infection dynamics for P-HM2 (higher adsorption in the light) but not for P-SSP7 whose infection dynamics were explained by light-induced host growth rate only (Demory et al. 2020). The ecological mechanisms that drive the evolution of these life history traits remain unknown. The prevalence of psbA and to a lesser extent psbD genes on the majority of the sequenced cyanophage genomes (Puxty et al. 2015; Crummett et al. 2016; Sullivan et al. 2006) along with their evolutionary history hints at their importance to the ecology and evolution of cyanophages. Cyanopodophages were found to follow the depth distribution of their hosts in the Red Sea, and clade B, which infects both marine cyanobacteria genera and harbours psbA, was 20 times more abundant than clade A that lack this gene (Baran et al. 2018). In addition, a transcriptome of cyanomyophages during infection was found to be well structured (Doron et al. 2016; Fedida and Lindell 2017; Thompson et al. 2016) with psbA, psbD expressing with the middle cluster of genes (Doron et al. 2016; Fedida and Lindell 2017; Thompson et al. 2016; Lin et al. 2016). A higher expression of AMGs in light compared to dark conditions (particularly photosynthesis, electron transport and PPP genes) was also reported (Thompson et al. 2016; Puxty et al. 2018). Few environmental studies have looked at the diel patterns of marine cyanophages. Oscillations in abundance of cyanophages as free virions in surface marine samples have been found to be minimal throughout the day apart from a peak after midnight (Clokie et al. 2006). A peak in Prochlorococcus mortality at midnight (Ribalet et al. 2015) provided support to this diel signal. In addition, the diel oscillations of VLPs were indirectly linked to Synechococcus diel patterns with a peak in late afternoon and a minimum at midnight (Bettarel et al. 2002). However, only recently, with the growing use of metagenomics and metatranscriptomics, has

398

G. Hevroni and A. Philosof

Cyanobacteria Photosynthesis

Translation

Cyanophage 0

2

4

Afternoon

Morning

6

8

10

DNA Replication

12

14

16

Dusk

18

20

22

Time of Day

Fig. 2 Schematic representation of diel gene expression patterns of cyanobacteria and cyanophages described in recent environmental studies. Cyanobacteria showing increased translation-related genes expressed before dawn, photosynthesis throughout the day, and DNA replication activities increased at dusk (Ottesen et al. 2014; Aylward et al. 2015). Peaks in transcriptional activity of Cyanophage genes were observed in the morning, afternoon and dusk (Aylward et al. 2017; Yoshida et al. 2018; Hevroni et al. 2020)

evidence for diel patterns of marine cyanophages started to emerge. Diel oscillations in gene expression of natural assemblages of marine cyanophages have been reported in diverse oceanic regions (Aylward et al. 2017; Yoshida et al. 2018; Kolody et al. 2019; Hevroni et al. 2020) (Fig. 2). In a metatranscriptome collected every 4 h over an 8-day period at the North Pacific Subtropical Gyre (NPSG), cyanophages – mostly putatively infecting Prochlorococcus – were shown to have an overall oscillating diel expression pattern peaking in the afternoon, with structural genes and AMGs such as psbA, helicases and a few other genes among them (Aylward et al. 2017). A reanalysis of the NPSG metatranscriptome (Aylward et al. 2017) has shown that the gene expression of cyanomyophages infecting Prochlorococcus could be divided into three separate clusters of early, middle and late genes (Doron et al. 2016; Thompson et al. 2016), suggesting similar synchronization in the environment (Chen and Zeng 2020) (Fig. 2). Similarly, diel periodicity of gene expression observed in Synechococcus phages was synchronized with the host transcriptome in seawater samples collected off the coast of Northern California (Kolody et al. 2019). Additionally, an increase in cyanophage abundance was observed following an afternoon-dusk transcriptional peak in a time-delayed manner, suggesting that progeny release was correlated with viral transcription (Yoshida et al. 2018). Concurrent diel metagenomics and metatranscriptomics samples collected every 2 h from the Gulf of Aqaba (Red Sea) showed a significant increase in the gene expression of cyanobacteria and cyanophages during daytime (Fig. 3a and b). Furthermore, different cyanophage families showed different timing of peak in transcription, with cyanopodophages peaking at 10:00, cyanosiphophages at 12:00 and cyanomyophages at 18:00 (Hevroni et al. 2020). The diel pattern of cyanophage gene expression (including both AMGs and structural genes) clustered together with cyanobacterial genes that are mostly related to energy production (e.g. photosynthesis related functions)

****

(b)

****

(c)

Fig. 3 Light-induced diel gene expression pattern in cyanobacteria and cyanophages. (a) Light-dark expression of selected genes in cyanobacteria (t-statistic 121.85, p-value