Circadian Clocks: Unexpected Biochemical Cogs

Abstract

A circadian oscillation can be reconstituted in vitro from three proteins that cycles with a period of ~24 h. Two recent studies provide surprising biochemical answers to why this remarkable oscillator has such a long time constant and how it can switch effortlessly between alternating enzymatic modes.

Circadian rhythms have three defining properties. The first is persistence in constant conditions with a period of ~24 h. The second is entrainment. The final characteristic is ‘temperature compensation’, namely that their periods are nearly the same at different constant temperatures. These three properties are difficult to explain by known biochemical reactions. The fascination of this phenomenon is to explain how a biochemical mechanism can keep time so precisely at such a long time constant (~24 h) and be temperature compensated [1]. A variety of mechanisms have been proposed to account for the long time constant of circadian oscillators, including time delays in transcription, mRNA processing, clock protein degradation, clock protein post-translational modifications (e.g., phosphorylation, ubiquitination, sumoylation, etc.), and/or nuclear transportation of clock proteins that negatively regulate their own expression in transcriptional/translational feedback loops (TTFLs). However, because all of those processes can occur within minutes and are temperature dependent, they do not provide a satisfying answer to circadian mechanism at a molecular and/or atomic level. A circadian oscillator reconstituted from only three purified proteins (KaiA, KaiB, and KaiC) is competent to maintain a temperature-compensated ~24 h period of KaiC phosphorylation in vitro [2]; this simple system should be ideal to answer the question of the biochemistry underlying the long circadian time constant in the absence of complicating factors like transcription, translation, degradation, nuclear translocation, and so on [3]. Recently, Abe et al. and Chang et al. revealed structural and biochemical bases of this in vitro oscillating system [4,5].

The KaiABC in vitro oscillator constitutes a core post-translational oscillator (PTO) of the cyanobacterial circadian system, which couples transcriptional (TTFL) and non-transcriptional (PTO) oscillators to achieve resilience and robustness [6]. The PTO can be reconstituted in vitro by incubating together the three proteins and Mg²⁺-ATP [2]. KaiC forms a hexamer from six monomeric KaiCs, each consisting of two similar but non-identical amino-terminal (CI) and carboxy-terminal (CII) domains [7]. Auto-phosphorylation and auto-dephosphorylation of KaiC occur at residues S431 and T432 in the CII domain [8,9]. KaiA binds to the carboxy-terminal tentacles of KaiCII, promoting KaiC phosphorylation (and inhibiting its dephosphorylation) in the day phase. KaiB binds to hyper-phosphorylated KaiC and sequesters KaiA, which inactivates the phosphorylation. Once KaiA is sequestered, KaiC auto-dephosphorylates. Therefore, KaiB acts as a key switch to flip KaiC from its phosphorylating state over to its dephosphorylating state. Note that KaiA and KaiB are regulators of KaiC’s intrinsic activities; neither KaiA nor KaiB have an independent kinase or phosphatase activity [3]. KaiC also has a remarkably low ATPase activity (~15 ATP hydrolyzed per KaiC monomer per day) and exhibits a circadian rhythm in this activity [10]. A correlation between the circadian period length of KaiC mutants in vivo and the ATPase activity of those mutated KaiCs in vitro has been demonstrated [4,10].

The contribution of Abe et al. [4] tackles the question of how this biochemical oscillator can be so slow and yet precise. On the basis of characterizations of KaiC mutants whose PTO periods in vitro correlate with gene expression (TTFL) rhythms in vivo, they suggest that the steady-state ATPase activity of the KaiCI domain constitutes the key rate-limiting step that controls the period of the oscillator [4,10]. KaiCI catalyzes ATP hydrolysis at a very slow rate of ~11 ATP hydrolyzed per day for each CI monomer, which is ~10⁶-fold lower than that of F₁-ATPase. This very slow and consistent rate — perhaps analogous to the sand running through an hourglass (Figure 1A) — has been hypothesized to be the most fundamental timekeeping reaction underlying circadian periodicity in cyanobacteria and that the transition of the conformational states during the ATPase cycle might exert a strain/tension on the structure of KaiC [10]. Abe et al. [4] use high-resolution X-ray crystallography of pre- and post-ATP hydrolysis states of KaiC to identify sources of the slow rate of ATPase activity in the CI domain. They discovered the lytic water molecule that may be involved in nucleophilic attack on the γ-phosphate of ATP is sequestrated away from the optimal ‘near-in-line’ position in the CI domain of KaiC. This position of the lytic water molecule is significantly less favorable for hydrolysis than the position of lytic water molecules in the catalytic sites of other ATPases with higher hydrolytic activity (such as kinesin, myosin and F₁-ATPase), and therefore results in a lower probability of ATP hydrolysis for KaiCI. A second source of the slowness of the ATPase reaction within the CI domain is due to a slow cis-to-trans isomerization of the peptide bond between residues D145 and S146 (pre-hydrolysis in cis configuration, post-hydrolysis in trans). The coupling of this cis-to-trans isomerization with hydrolysis raises the activation energy barrier of the reaction to 25–33 kcal mol⁻¹, which predicts a dramatic reduction in the ATP hydrolytic rate.

Figure 1. The KaiABC PTO: how a constant energy input can be converted into a clock.

(A) The intrinsic ATP hydrolytic activity of the CI domain of KaiC provides a constant energy input to the KaiABC oscillating system (PTO). This is the rate-limiting reaction whose slowness determines the long time constant of the circadian period [4]. This constant energy input functions as a timer, analogous to the flow of sand in an hourglass. KaiC’s structure is shown as an inset as a double domain hexamer with CI and CII domains labeled [7]. Carboxy-terminal tentacles are depicted as unstructured extensions from the CII domains. (B) Interactions of KaiA and KaiB provide phase-dependent feedback to the constant KaiCI timer. In particular, KaiB switches KaiC back and forth between phosphorylating and dephosphorylating modes, analogous to ‘flipping an hourglass’ (this analogy is inaccurate because flipping an hourglass is a transition from an inactive state to an active state, whereas KaiB binding/unbinding transits KaiC between two active but different modes). A dramatic conformational change to a distinct fold enables KaiB to bind KaiC [5]. KaiC is shown in hyper-phosphorylated form (red dots symbolize phosphorylation within CII domain) in complex with KaiB (green diamonds) and sequestered KaiA (pink dimer). (C) Modulation of the positive input (ATP hydrolysis) by negative feedback (KaiC phosphorylation and KaiA/KaiB interactions) generates a stable oscillation of both ATPase activity and KaiC phosphorylation status. KaiC phosphostatus serves as a phase marker of this oscillator.

Abe et al. [4] conclude that these features together with the conformational asymmetry of subunit arrangement within KaiC hexamer establish a remarkably slow but stable ATPase cycle in the CI domain of the KaiC hexamer [4]. These detailed analyses of the CI-ATPase and its coupling to the overall ATPase activity of full-length KaiC provide a spectacular explanation for how the precise and slow CI-ATPase determines the long time constant (period) of this circadian oscillation. However, other aspects of this study are less well developed. For example, the authors postulate that temperature compensation of CI-ATPase is due to an intrahexameric balance between prehydrolysis and posthydrolysis monomers, but provide no evidence in support of that contention. Moreover, while acknowledging the roles of KaiA and KaiB in harnessing the ATPase activity and regulating KaiC, the authors appear to downplay the key roles of the KaiA–KaiB interaction in establishing the resilient and sustained oscillation [3,11,12].

The aspects of KaiC interaction with KaiA and KaiB that are under-represented in the Abe et al. [4] contribution are the very features that are studied by Chang et al. [5] and these authors also have a delightfully unexpected story to tell. The role of KaiA in facilitating KaiC’s auto-phosphorylation has been well studied, but KaiB has been the ‘dark horse’ of the Kai protein triumvirate — little has been known about the mechanism by which KaiB binding switches KaiC’s activity between auto-kinase and auto-phosphatase modes. Chang et al. [5] finally enlighten KaiB’s pivotal action. A previously inexplicable curiosity about KaiB is that its primary structure (amino acid sequence) indicates that it is a thioredoxin-like protein, and yet the crystal structure of KaiB has a 3-D fold that is unique and distinctly different from that of thioredoxin [13,14]. NMR analyses of free KaiB and KaiC-bound KaiB revealed significantly different secondary structures, namely that free KaiB exhibits a secondary structure like that found in crystals (‘ground-state’ or gsKaiB), but KaiB bound to KaiC is indeed in the expected thioredoxin-like fold (‘fold-switched’ or fsKaiB)! A KaiB mutant (G88A/D90R) preferentially adopts the fsKaiB configuration and binds to KaiC much more rapidly than native KaiB, which is largely in the gsKaiB configuration. Based on these data, Chang et al. [5] claim that KaiB is an unusual ‘metamorphic protein’ that can interconvert between distinct folds under native conditions [15]. The slow/rare transitions of KaiB between the inactive free form (gsKaiB) and the KaiC-binding active form (fsKaiB) can provide a small time delay (a few hours), but despite the authors’ contention, this is not a major contributor to the long time constant of circadian period as compared with the analysis of ATPase activity by Abe et al. [4]. Nevertheless, the gsKaiB/fsKaiB transitions are a remarkable mechanism to explain how KaiB switches KaiC back and forth between phosphorylating and dephosphorylating phases, crudely analogous to ‘flipping an hourglass’ to restart the flow of sand in a new phase (Figure 1B). Hyper-phosphorylated KaiC binds fsKaiB, and the KaiC/fsKaiB complex sequesters and inactivates KaiA. This switches KaiC to the dephosphorylating mode, and KaiC slowly dephosphorylates. Once KaiC is hypo-phosphorylated, fsKaiB disentangles from KaiC, reconverts to gsKaiB and releases KaiA. KaiA is then free to re-engage hypo-phosphorylated KaiC and in conjunction with KaiC’s intrinsic ATP hydrolysis, KaiA facilitates KaiC phosphorylation and the cycle begins anew [3]. Therefore, KaiB binding (as fsKaiB) and unbinding is the critical switch of KaiC’s kinase/phosphatase modes [5].

In addition to shining light on the mechanism of KaiC switching within the central oscillator, the discovery of fsKaiB illuminated the biochemistry of clock output as well [5]. The PTO regulates gene expression (TTFL) via the two-component regulatory system protein SasA [16] that interacts with phospho-KaiC and controls the response regulator/transcriptional factor RpaA [17]. The amino terminus of SasA has a thioredoxin-like fold [18] and SasA has been known to compete with KaiB for binding KaiC [19,20], but previously we could not explain why the thioredoxin-like SasA would compete with the non-thioredoxin-like KaiB. Chang et al. [5] can now explain this enigma — it is the fsKaiB configuration that competes with SasA for binding to KaiC. Therefore, fsKaiB not only switches KaiC activity modes, it also chaperones SasA’s association with KaiC, thereby regulating the timing of clock-controlled transcriptional outputs.

Together, these two papers address key questions about circadian biochemistry using the model in vitro system from cyanobacteria. How can such a long time constant be achieved? How can a continuous reaction (e.g., ATP hydrolysis) be converted into a biochemical oscillator by judicious phase switching? The data of Abe et al. [4] indicate that the CI-ATPase provides a constant slow rate of input that explains the long time constant and, in conjunction with KaiA, promotes KaiC phosphorylation (‘Timer/Energy’, Figure 1A). The KaiA/KaiB interactions lead to negative feedback regulation of this input (‘Phase-dependent feedback’, Figure 1B), and Chang et al. [5] contend that gsKaiB/fsKaiB transitions explain the switching of this feedback as well as phase-dependent control of clock output. The coupling of positive input and negative feedback creates the oscillation (‘Clock’, Figure 1C), where KaiC’s phosphostatus is the phase-marker of the pacemaker.