From phenomenon to mechanism in the circadian clock of Synechococcus elongatus

Abstract

Creative and collaborative research in the early 21st century revealed the mechanistic secrets of the cyanobacterial circadian clock, a nanomachine whose function can be studied in vitro. The structural dynamics of the Synechococcus elongatus circadian oscillator, comprising KaiA, KaiB, and KaiC, establish a near 24 h cycle of activity of the global transcription regulator RpaA by rhythmically changing the interaction platform of kinases that control RpaA phosphorylation.

Near the end of the 20th century, a new model system burst onto the circadian biology stage, challenging long-standing assumptions about the cellular properties that should be required to drive a circadian clock. Reports of rhythmic metabolic and physiological phenomena in species of prokaryotic cyanobacteria had, by the mid-1980s, fulfilled the agreed-upon criteria for circadian designation¹: rhythms that are free-running with a periodicity of about 24 h, whose relative phase could be reset by environmental stimuli, and whose period is temperature compensated². Thus, the presumptions that an organism that reproduces faster than once a day wouldn’t have a circadian clock (the “circadian - infradian rule”³, or that one without cooperating nuclear and cytoplasmic compartments couldn’t generate 24 h rhythms, were demonstrated to be incorrect.

The adoption in the early 1990s of a genetically amenable cyanobacterial strain for circadian study, the Synechococcus elongatus isolate known as PCC 7942^4,5, enabled a genetic search to find mutants altered in circadian clock function, as had been pioneered by Konopka and Benzer (Drosophila)⁶, and Feldman and Hoyle (Neurospora)⁷. A fortuitous convergence in the development of luciferases as dynamic gene expression reporters in bacteria (Streptomyces)⁸, and as circadian reporters in Arabidopsis⁹, provided a foundation for an effective, automated, and high-throughput genetic screen in S. elongatus¹⁰. The extraordinary robustness of circadian rhythms of gene expression in this specific cyanobacterial strain, and what would later be revealed as its particularly uncomplicated clock mechanism, greased the wheels of discovery¹¹ (Fig. 1). By the year 2000, the cyanobacterial circadian clock mechanism had progressed from a completely unknown entity to a set of critical parts: the proteins KaiA, KaiB, and KaiC¹². Moreover, the S. elongatus system had already provided three major lessons for our broader understanding of circadian biology: (1) circadian rhythms of gene expression in S. elongatus run with a periodicity of about 24 h no matter how rapidly cells are dividing (demonstrated as fast as every 5 h), or whether they are not dividing at all¹³; (2) competitive co-culture of S. elongatus period mutants provided experimental confirmation that a circadian clock whose periodicity matches that of the environmental day provides an adaptive advantage¹⁴; and (3) the clock exerts global influence over expression of essentially the entire genome¹⁵.

Fig. 1. Three assays that powered analysis of the cyanobacterial clock mechanism.

Top: Bioluminescence from engineered strains of S. elongatus that carry luciferase reporter genes provides an automated in vivo assay for circadian rhythms of gene expression. These robust rhythms constitute a circadian phenotype that is amenable to genetic mutant screens, like that used to identify the kaiABC genes. The X axis at right is time in hours in constant conditions. Middle: Quantification of KaiC phosphorylation (yellow circles, marked S or T for Ser or Thr residues) in timed samples from in vitro oscillator reactions enables biochemical analysis of mechanistic aspects of the KaiABC circadian oscillation, providing a phase reference for clock progression. Highly stylized protein gel from indicated time points in hours depicts upper fully phosphorylated KaiC bands and lower non-phosphorylated bands. Actual data show additional bands of intermediate migration that represent specific phosphorylation states. Bottom: Fluorescence anisotropy measurements from a fluorescently tagged component allow real-time monitoring of component interactions in in vitro oscillator or full clock (KaiABC + CikA + SasA + RpaA + DNA) reactions, expanding analysis beyond the phosphorylation status of KaiC. When the fluorescent tag (yellow star) is on a DNA fragment that carries an RpaA-binding site, the reaction can capture the rhythmic binding of phosphorylated RpaA to its target, demonstrating rhythmic output of the clock at the level of gene regulation. The X axis at right is time in hours in constant conditions.

The first quarter of the 21st century would take these parts and assemble a fully functional clock that explains circadian timekeeping in cyanobacteria, in vivo and in vitro, from a steadily ticking oscillator to multiphasic gene expression. This explosion in understanding rested on concerted efforts in genetics, quantitative biochemistry, and structural biology by a small but mighty cohort of labs internationally, with a few extraordinary standouts in effort and insight. As a first critical development, the biochemical functions of the Kai proteins were revealed: KaiC autophosphorylates in vitro¹⁶, with KaiA stimulating KaiC autophosphorylation and KaiB opposing KaiA’s activity^17,18. By the early 2000s the databases of genome sequences had grown, and widespread conservation of kaiABC genes in diverse cyanobacteria¹⁹ both reinforced expectations of their adaptive significance and extended investigation of their products to thermophilic strains, facilitating structural studies. Early advances included the demonstration that KaiC forms homohexamers^20,21, and that the N-terminal domain of S. elongatus KaiA is related to the receiver domains of bacterial two-component signal transduction proteins¹⁷. Crystallographic and solution structures for all three Kai proteins rolled out in rapid succession^21–27. The monomers of the KaiC homohexamer are internally duplicated, such that the hexamerized C-terminal domains form a ring (CII) that sits on top of the similarly hexamerized N-terminal domain ring (CI)²⁵, resulting in what is affectionately referred to by the cyanobacterial clock community as a “double donut” (Fig. 2). KaiA was demonstrated to exert its stimulation of KaiC phosphorylation by binding specifically to a KaiC C-terminal peptide tail that extends from each CII domain^28,29, and specific autophosphorylation sites were identified on KaiC^30,31. Moreover, the phosphorylation of KaiC increases and decreases over the course of 24 h in vivo¹⁸. Determining just how that process occurs led to many startling discoveries.

Fig. 2. Major day-night conformational differences of each Kai oscillator components.

In each column, a specific Kai protein is highlighted to indicate its early day (dawn, upper) or early night (dusk, lower) configuration. KaiA (Left) assumes a form that is active in binding the C-terminal tails of the KaiC-CII ring, as well as an alternate form. During the day, the C-terminal tails of the KaiC-CII ring are available and bound by active KaiA. At night, KaiC (Right) is fully phosphorylated and assumes a compressed state, in which the C-terminal tails of KaiC recede inside the CII ring, making them less available for KaiA binding. Additionally, this night-time form of KaiC stabilizes the exposure of KaiB-binding sites, capturing the rare fold-switched form of KaiB (Middle), which forms a ring that in turn captures the alternate form of KaiA.

A bold and surprising development in 2005 caught the entire field of circadian biology, and indeed the greater biochemistry community, off guard, and turbo-charged subsequent investigation of the cyanobacterial clock mechanism. The circadian cycle of KaiC phosphorylation was reconstituted in a test tube with only KaiA, KaiB, KaiC, and ATP present as purified components³². Like the in vivo gene expression rhythms, the KaiC phosphorylation cycle – an in vitro oscillator – fulfills the criteria accepted for genuine circadian rhythms and recapitulates in vivo period phenotypes. This in vitro oscillator, replicated in several labs and self-sustained for more than 10 days^33,34, provided a defined and minimal system in which to probe biochemical steps of the clock mechanism. One feature that allows the in vitro oscillator system to persist so robustly in isolation is that the ATPase activity of KaiC is exceptionally low: on average 15 ATPs are hydrolyzed per day per KaiC monomer, thus not consuming the initial input of ATP nor creating inhibitory levels of ADP³⁵. The system also benefits from intrinsic mechanisms that provide synchronization of the phosphorylation status among KaiC hexamers, including a redistribution of monomers at a specific stage of the cycle³³. The in vitro oscillator enabled elucidation of the KaiC phosphorylation progression and the direct assessment of Kai mutant variants without the influence of any other cellular components (Fig. 1).

Reconstitution of the circadian KaiC phosphorylation cycle had many immediate implications. First, the purified Kai proteins complete multiple cycles in a sealed tube without the intervention of any other processes, ruling out the requirement for a transcription-translation feedback loop in the cyanobacterial clock, the leading mechanistic model for eukaryotic clocks³⁶. In vivo data had been pointing to this reality for some time. Although overexpression of KaiC inhibits expression of the kaiC gene in vivo, as seen with Drosophila Per and Neurospora Frq¹², other data suggested that this autoinhibition is not necessary for circadian timing. The clock runs robustly when the kai genes are expressed from heterologous promoters that lack potential autoregulatory elements³⁷, and under conditions in which transcription and translation are inhibited³⁸. Nonetheless, an entirely post-translational oscillator, so simple in constitution, was a surprise to all, as was the genuinely “slow biochemistry” of the 24 h cycle.

Both in vivo and in vitro, the 24 h cycle of KaiC phosphorylation steps through an ordered sequence of phosphorylation on threonine 432, then serine 431, dephosphorylation of threonine, then serine, and back to a non-phosphorylated state^39,40 (Fig. 3). However, the wealth of structures of the individual proteins, and exquisite in vitro oscillator assays with wild-type and mutant proteins, did not readily reveal how the Kai proteins fit together, or how the different phosphorylation states define time of day. Two major findings pushed past experimental and conceptual bottlenecks. Protein dynamics measurements revealed changes in stiffness of KaiC, and stacking of the CII ring on the CI ring as a result of phosphorylation, that weren’t evident in crystal structures of mimics of the different phosphorylation states^41,42. Stiffness and stacking changes cause asymmetries between the two rings, and differences in the looseness or tightness of each ring. The fully phosphorylated state of KaiC (biochemical “dusk”) has the CII ring tight and stacked on CI, arresting a state of CI after it has hydrolyzed ATP, in which the binding sites for KaiB are exposed (Fig. 2). The ability to detect this binding required another breakthrough: the discovery that the form of KaiB that is competent for KaiC binding is an entirely different fold, with different underlying secondary structure, than the KaiB tetramer whose structure had been determined by several labs⁴³. Unexpectedly, KaiB is among a currently small known group of metamorphic proteins, meaning that it reversibly switches between fundamentally different native folds. KaiB rarely monomerizes, unfolds, and refolds to an unstable active state; this active state is stabilized only if the CI ring is receptive to capture it (Fig. 2). The discovery of the fold-switched KaiB active state enabled the stabilization of that state via mutagenesis, which fostered binding of KaiB to KaiC. The creation of binding-competent KaiB facilitated the assembly of the long-sought Kai complexes^43,44. KaiB forms yet another ring, bound to KaiC-CI, that forms an early night-time complex (Fig. 2), explaining an earlier finding that in vivo, KaiA and KaiB are both physically associated with KaiC specifically at night⁴⁵.

Fig. 3. The Kai-based oscillator functions to control the activity of a global transcription factor, RpaA, by regulating the kinases that activate it.

The phosphorylation cycle of KaiC alters the structure of the hexamer to rhythmically favor the docking of either SasA (daytime) or CikA (nighttime). Top: The unbound state of CikA is an RpaA kinase, whereas SasA is inactive unless bound to KaiC. The daytime non-phosphorylated KaiC hexamer favors binding of SasA, activating its kinase activity. The combined actions of these kinases increase RpaA phosphorylation, peaking at dusk. When the fully phosphorylated state of KaiC hydrolyzes ATP and captures fold-switched KaiB, a newly formed KaiB ring captures both the alternative conformation of KaiA and the kinase CikA, activating CikA’s RpaA phosphatase activity. Thus, after dusk, RpaA progressively dephosphorylates. RpaA binds its DNA recognition site only when phosphorylated, so its activity peaks at dusk. Lower right: For promoters of dusk-peaking genes, RpaA binding attracts RNA polymerase and activates them; For promoters of dawn-peaking genes, RpaA binding blocks RNA polymerase and represses them. Lower left: Thus, a single phase of RpaA phosphorylation, regulated by the KaiABC complex, results in two opposite phases of gene expression. In vivo, additional cellular and environmental factors act on promoters, creating intermediate phases of gene expression.

The importance of this KaiABC complex is best understood by incorporating additional key elements of the clock. Insights into the physiological and genetic functions of the circadian clock in S. elongatus were emerging in parallel with the biochemical progress. Two histidine protein kinases were defined genetically that affect circadian gene expression, CikA and SasA^46,47 (Fig. 3). These proteins would later be revealed as clock components that engage physically with the Kai oscillator to drive temporal gene regulation by controlling the phosphorylation state, and thus DNA-binding affinity, of a response-regulator type transcription factor, RpaA⁴⁸. These components defined an “output pathway” from the oscillator, which turned out to be more direct than expected. The kinase SasA was initially discovered in part because of its sequence similarity to KaiB, but with some differences in secondary and three-dimensional structure that were puzzling⁴⁷. These differences were reconciled when the fold-switched structure of KaiB was revealed to match the N-terminal domain of SasA^43,47. Indeed, the two proteins bind both competitively and cooperatively to the same sites on KaiC, but with different kinetics and inter-subunit interactions^43,44. The transition from a SasA-dominated daytime association with KaiC, to the night-time KaiB ring on KaiC, is a critical step of circadian timekeeping (Fig. 2). Once the KaiB ring forms on KaiC-CI, it provides a platform for binding two other key components: KaiA, in a conformation that is not competent to stimulate KaiC phosphorylation on KaiC-CII (Figs. 2,3), and the other kinase, CikA (Fig. 3). Just as SasA and KaiB share structural similarity and bind the same site on KaiC, CikA and KaiA also bind to the same site on the KaiB ring⁴⁴.

Two additional findings connect the Kai oscillator to the robust rhythms of gene expression that formed the phenotype on which the genetic hunt for the cyanobacterial circadian clock mechanism was based. The response-regulator transcription factor RpaA is the target of the kinases SasA and CikA⁴⁹. The activities of those kinases on RpaA are altered by their engagement with the Kai complex (Fig. 3): SasA has little kinase activity unless it binds to KaiC, which occurs during the day, when KaiC phosphorylation is rising; CikA may also phosphorylate RpaA in the day, when it is not on the oscillator^50,51, but activates a phosphatase activity against RpaA when it docks on the KaiB ring at night^49–51. Thus, the ultimate function of the Kai oscillator is to rhythmically alter the phosphorylation state of RpaA by providing different binding platforms to alternatively accelerate the kinase and phosphatase functions that operate on RpaA during the 24 h cycle. Like the in vitro oscillator, this tight regulatory loop can be reconstituted in vitro, with RpaA cyclically binding and releasing a fragment of DNA that carries a recognition site from one of its target genes⁵¹ (Fig. 1). Moreover, if RpaA binding is used to block progression of a T7-RNA polymerase, rhythmic transcription itself can be recapitulated in a test tube with the established properties of a circadian rhythm⁵².

The direct contact of SasA and CikA with the Kai complex, and their cooperative and competitive binding with oscillator components, intimately connects these kinases with timekeeping^44,51. The presence of SasA or CikA alters circadian period in vivo and in vitro, and each can even rescue rhythms when its KaiB or KaiA alter-ego (respectively) is limiting by enabling progression to the next physical state of KaiC⁵¹. Thus, the components of the output pathway are also part of an extended Kai-SasA-CikA oscillator, blurring the identities of these divisions of the clock. The input “pathways” also reside within the oscillator itself. Rather than gathering environmental input from sensory photoreceptors, the cyanobacterial clock components directly bind metabolites that are proxies for light and darkness in this photosynthetic organism. CikA and KaiA bind oxidized but not reduced quinones^53,54, electron carriers of photosynthetic electron transport whose oxidized status spikes upon the onset of darkness. Additionally, the time stamp of KaiC phosphorylation is altered by the ratio of ATP and ADP, which varies with photosynthetic activity⁵⁵. Moreover, KaiB activity is intrinsically temperature responsive⁵⁶ and the ATPase of KaiC is temperature invariant³⁵, contributing to temperature entrainment and compensation.

The aggregation of all clock functions into this small assemblage of proteins (Fig. 3) describes a clock that is shockingly “mechanical” in its mechanism⁵¹. Indeed, single particle analysis models of some components closely resemble gears^44,57. In vitro, the system can perform functions from resettable timekeeping⁵⁸ to circadian transcription⁵² with only a five-protein clock, as CikA can provide both kinase and phosphatase functions when isolated from post-translational modification that occurs in vivo⁵⁰. But the cellular environment does complicate and enrich the timing functions, with KaiC showing time-specific localization⁵⁹, a phase separation that requires threshold protein concentrations⁶⁰, influence of proteases^61,62, and additional autoregulation through what can be considered a transcription-translation feedback loop^63,64. The cellular milieu confers the ability to express more than the two phases, dawn and dusk, of the direct targets of RpaA^15,48,65, and the ability to prepare for seasonal changes by entrainment that tracks midday^66,67.

The extraordinary mechanism of the S. elongatus circadian clock has unfolded in exquisite detail over three decades. At its heart is a controlled riot of allostery, with the physical interactions of each protein nudging the system toward the next conformational status, and back reactions thermodynamically suppressed^42,44,68. The changes that occur in KaiC are as much up-and-down, donut-to-donut, as they are side-to-side, with almost every part of the KaiC protein participating in the daily cycle^42,44,57. This fuller picture of subtle molecular changes in internal positions and stiffness explains why the initial period mutations of the genetic screen were scattered throughout the kaiC gene rather than defining an “active site” as might have been expected¹⁰. This view also explains why a phosphorylation cycle on the CII domain—the only feature of the clock complex that could be tracked for many years—has a big effect on the CI ring, on which the other components depend for interaction⁵⁷. Rather than shifting gears as the role of the phosphorylation cycle was initially envisioned, a more accurate analogy of KaiC is of a docking platform whose piers and slips are rendered available or inaccessible to other components by an ebbing and flowing tide of phosphorylation-mediated changes⁴⁴.

Each of the oscillator proteins has two major conformations that are dramatically different in activity (Fig. 2): KaiA in its CII-tail-binding, KaiC-stimulating (daytime) form, vs its alternate form that sleeps through the night, attached to KaiB at the opposite end of the double donut; KaiB biding most of its time in a tetrameric and apparently inactive ground state, vs. its active fold-switched monomeric state that forms the night-time hub on KaiC; and KaiC itself, whose overall structure is quite different pre- and post-hydrolysis of ATP in the CI ring. The rate of this ATPase function has been demonstrated to regulate the pace of the oscillator, and to be fundamentally temperature compensated³⁵. The unusually slow ATPase of KaiC-CI is so defining of the cyanobacterial clock that the discovery of a protein in clock complexes from mammalian and other eukaryotic cells that carries KaiC-like ATPase motifs waved a flag to demand investigation. Indeed, the RUVBL2 protein in eukaryotic lineages has a slow ATPase, and mutagenic alteration of this activity alters the period of the mammalian clock⁶⁹. The extent of functional equivalency of RUVBL2 to the cyanobacterial post-translational oscillator remains under investigation.

With 30 years of hindsight, the decision to look for circadian rhythms in S. elongatus and to develop it as a model for circadian investigation was a stroke of outstanding fortune for discovering the clock mechanism. Although cyanobacterial species throughout the phylum have kai genes¹⁹, many species are genetically amenable⁷⁰, and diverse clades display circadian physiology^1,71–73, the convergence of robust rhythms with an accessible phenotype for screening and extensive genetic tools is still maximal in the strain S. elongatus PCC 7942. Some species have multiple sets or partial sets of kai genes, complicating the assignment of a gene to a specific function or knowing whether a particular paralog has a clock function at all⁷⁴. Some species are missing kaiA or have a truncated version of the protein⁷⁵. Some have rhythmic processes that damp quickly under free-running conditions⁷⁶. These differences that would have hampered the initial probe of clock mechanism now provide a wealth of opportunities to investigate a spectrum of biological timers that may stray significantly from the agreed-upon criteria for circadian oscillators. In S. elongatus, the clock seems to rule all of its biology. The entire genome is expressed rhythmically¹⁵, and our lab’s survey of insertion mutants for every locus in the genome⁷⁷ indicated that even major assaults on metabolism and physiology that render cells “sickly” have little effect on rhythmicity of gene expression. Additionally, the clock’s disregard for cell division rate reflects a dogged determination in this species to run its processes “like clockwork”¹³. Notably, this indifference isn’t bidirectional, as the cell division process cares very much what time of day it is, with a non-permissible window early in the night^78–80. A robust clock may confer benefits, including tolerance for varied photoperiods^81,82. However, S. elongatus, with its streamlined genome and growth habit as an obligately photosynthetic unicellular species that does not fix nitrogen, might be “special.” Some species need their clock to mesh well with other priorities of biology⁷⁶ such as heterotrophy⁷³, multicellularity, and differentiation of cell types⁷¹. Given that the same alternation of day and night that applied the selective pressure for circadian clocks to evolve in the first place also provided daily entrainment, a damped clock or hourglass may also be of great value to prokaryotes, especially in the presence of other species⁸³. These diverse demands may explain why other phyla encode variations on the extraordinarily precise, robust, and efficient circadian clock of S. elongatus. These muddier, but potentially richer, complexities make up an exciting frontier for prokaryotic circadian biology⁸⁴.

Author contributions

S.S.G. wrote the manuscript.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The author declares no competing interests.