Before the mid-1980s, chronobiologists thought that endogenous circadian rhythms were an exclusive property of eukaryotic organisms, and it became a dogma that prokaryotic organisms were either too “simple” or grew too rapidly to have evolved a bona fide circadian timekeeper (Johnson et al., 1996). That dogma cracked in 1986 with reports that the diazotrophic cyanobacterium Synechococcus RF-1 displayed daily rhythms of nitrogen fixation in LD cycles that persist in constant light (LL) (Grobbelaar et al., 1986; Huang et al., 1990; Chen et al., 1991). The salient properties of circadian rhythms–persistence, entrainment, and temperature compensation (Johnson et al., 2017) were established in these early studies, but progress on the genetics and mechanism of the cyanobacterial clockwork was hindered because no genetic tools were available for Synechococcus RF-1 (e.g., transformation, homologous recombination, luminescence/fluorescence reporters, etc.).

Several years after the reports of circadian phenomena in Synechococcus RF-1, we and our collaborators began the analysis of circadian rhythmicity in the cyanobacterium Synechococcus elongatus PCC 7942 (hereafter S. elongatus), which was a species for which genetic tools were available, and for which a luciferase reporter strain (P _psbAI ::luxAB) had already been generated for the analysis of light intensity regulation of gene expression (Kondo et al., 1993; Johnson & Xu, 2009). The availability of genetic tools and a robust luminescence reporter of rhythmic gene expression enabled spectacular progress on understanding the mechanism and adaptive significance of circadian rhythmicity in cyanobacteria, transforming S. elongatus into one of the best understood circadian model organisms (Johnson et al., 2017; Chavan et al., 2021; Ito-Miwa et al., 2021; Johnson & Rust, 2021).

On the other hand, the cyanobacterium Synechocystis sp. PCC 6803 (hereafter Synechocystis) has long been a model organism for photosynthesis research and biotechnological applications (Nixon et al., 1992; Pakrasi & Vermaas, 1992; Shen et al., 1993; Ikeuchi & Tabata, 2001; Jagannathan & Golbeck, 2009; Santos-Merino et al., 2019; Wang et al., 2021b) partially due to the early discovery of its natural competence for genetic transformation (Grigorieva & Shestakov, 1982). For example, Synechocystis has served as a model platform for the structural and functional characterization of photosynthetic mechanisms, especially of photosystem II (Nixon et al., 1992; Pakrasi & Vermaas, 1992; Larom et al., 2010). When the complete genome sequence of Synechocystis became available in 1996, it became among the first of all phototrophic organisms to have that wealth of genetic information (Kaneko et al., 1996; Nakamura et al., 1998). As a result of this genomic data, research productivity involving this strain dramatically accelerated. In particular, Synechocystis-derived strains were explored as microbial cell factories to produce a variety of fuels and chemicals (Wang, et al., 2021b), including precursors to (biodegradable) plastics, such as poly-3-hydroxybutyrate (Wu et al., 2001; Agarwal et al., 2022), 3-hydroxybutyrate (Wang et al., 2018b), 3-hydroxypropionate (Wang Y. et al., 2016), lactate (Varman et al., 2013; Angermayr et al., 2014), and ethylene (Ungerer et al., 2012; Zhu et al., 2015; Wang et al., 2021a).

However, in comparison with S. elongatus, Synechocystis has been understudied in terms of its circadian properties (Aoki & Onai, 2009; Schmelling et al., 2021). The analysis of circadian phenomena in Synechocystis may have been hampered by the fact that the P _dnaK ::luxAB reporter developed for Synechocystis (Aoki et al., 1995) was never as robust as the P _psbAI ::luxAB reporter that revolutionized circadian analyses in S. elongatus (Kondo et al., 1993; Kondo et al., 1994; Johnson & Xu, 2009; Johnson et al., 2017). Nevertheless, understanding and manipulating the circadian system of Synechocystis specifically would be beneficial for multiple reasons. For example, some strains of Synechocystis can grow photoheterotrophically by using glucose in the medium (Anderson & McIntosh, 1991), whereas S. elongatus is an obligate photoautotroph and must have light to grow. Therefore, photoheterotropic characteristics of Synechocystis could be useful in the analysis of circadian behavior in constant darkness using a luciferase reporter (Aoki et al., 1997), which is a capability that is not possible in the obligate photoautotroph S. elongatus. Moreover, in terms of biotechnological applications, we previously showed in S. elongatus that manipulating the circadian system can be used to enhance foreign gene expression (Xu et al., 2013b), and similar tools could be applied to the versatile biotech platform Synechocystis.

To overcome the deficit of an inadequate circadian assay, we have developed a new luminescence reporter for Synechocystis that exhibits many of the advantageous properties of S. elongatus’ P _psbAI ::luxAB reporter in terms of brightness and excellent peak-to-trough amplitude. Surprisingly, our new reporter is based on using a promoter that is not endogenous to Synechocystis, but is derived from a chloroplast gene from a higher plant. We show here that the P _psbA ^Ah ::luxAB reporter as applied to Synechocystis identifies circadian properties that are equivalent to those that have been extensively characterized in S. elongatus and enabled the isolation of period mutants as well as rigorous tests of adaptive significance in Synechocystis.

A strong promoter system was developed independently for biotechnological purposes using Synechocystis that derived a hybrid promoter from the chloroplast psbA gene of the higher plant Amaranthus hybridus coupled to an optimized ribosome binding site (RBS) (Salis et al., 2009; Xiong et al., 2015). We tested whether this hybrid promoter would display circadian rhythms in Synechocystis by designing a bacterial luciferase reporter and recombining it into the genome. This hybrid promoter/reporter (P _psbA ^Ah ::luxAB) displays excellent circadian rhythms of luminescence (Figure 1). In comparison with the previously used P _dnaK ::luxAB reporter (Aoki et al., 1995), luminescence signals are 5–10 times brighter with the P _psbA ^Ah ::luxAB reporter and display robust circadian rhythms with a period τ) slightly longer than 24 h in constant light (LL) at 30°C (Figure 1B, Supplementary Figure S1). As in the case of S. elongatus, we take this luminescence rhythm to be a reporter of pervasively rhythmic gene expression in Synechocystis, which has also been assessed by microarray studies (Liu et al., 1995; Woelfle et al., 2007; Ito et al., 2009; Vijayan et al., 2009). An expanded scale is shown for the data of the P _dnaK ::luxAB reporter in Figure 1C, and this is among the most robustly rhythmic traces we have observed for this reporter, whereas the trace for the data of the P _psbA ^Ah ::luxAB reporter in Figure 1B is of average robustness. Clearly the quality of rhythms is excellent with the P _psbA ^Ah ::luxAB reporter, and its brightness is also an advantage because it is less demanding of the sensitivity of the monitoring instruments. In addition, the P _psbA ^Ah ::luxAB reporter exhibited robust rhythms over a broad range of temperatures, whereas the rhythms of P _dnaK ::luxAB reporter became poor or arhythmic at lower temperatures (see below).

We therefore applied the P _psbA ^Ah ::luxAB reporter to address a previously unresolved question, namely what is the role of the various kaiC genes in circadian rhythmicity in Synechocystis? In S. elongatus, there is a single kaiABC clock gene cluster which encodes the central core clock proteins mediating the clockwork (Ishiura et al., 1998; Johnson et al., 2017), but in Synechocystis, there are three kaiC homologs organized with and without kaiA and kaiB homologs (Figure 2A) (Aoki & Onai, 2009). When the presence of kaiC homologs among bacterial species is assessed globally, it is not unusual to find species with two or more kaiC homologs (Aoki & Onai, 2009; Schmelling et al., 2021). Since S. elongatus is able to elaborate a precise clockwork with only one kaiC homolog, what is the function of multiple kaiC genes in those species that harbor more than one copy? Synechocystis is an excellent test case, so we undertook to make null strains in which each kai cluster (and each kai gene individually) was knocked out to determine if its presence was necessary to enable robust rhythmicity of gene expression as reported by P _psbA ^Ah ::luxAB (see Supplementary Figure S2 for the genotyping of the knock-out strains). Figure 2 illustrates the results of the gene knockouts upon the luminescence rhythm, where Figure 2B shows the data for three representative wild-type (WT) colonies. As was found for the case of the kaiABC cluster in S. elongatus (Ishiura et al., 1998), knocking out the kaiAB1C1 cluster or the kaiB1 or kaiC1 genes individually led to immediate arhythmicity (Figure 2C). Of the kai homologs in Synechocystis, the kaiB1 and kaiC1 genes are the most similar to kaiB and kaiC from S. elongatus (GroupA kai homologs (Aoki & Onai, 2009).

Surprisingly–and unlike the case for S. elongatus–knocking out the kaiA gene did not cause total arhythmicity (although the amplitude of the rhythm was reduced and the free-running period (FRP) was lengthened). In non-cyanobacterial species, there is precedence for the hypothesis that KaiB and KaiC can generate a daily timekeeping process in the absence of KaiA (Ma et al., 2016). Indeed, even in S. elongatus, the deletion of KaiA has been reported to allow the expression of a damped circadian oscillation under some conditions (Kawamoto et al., 2020). Another interpretation of the data in Figure 2C based on the known mechanism for the cyanobacterial clockwork of S. elongatus is that there might be another protein in Synechocystis that exhibits KaiA-like functionality. Based on the operon structure of kaiABC in S. elongatus (Ishiura et al., 1998), it is possible that the knockout of kaiA has disrupted the promoter region of kaiB1C1, but this is not likely because 1) the kaiA-ko strain is still rhythmic whereas the kaiB1-ko, kaiC1-ko, and kaiAB1C1-ko strains are all arhythmic, and 2) computational prediction of the kaiB1C1 promoter is most likely to occur in the 126bp intergenic region between kaiA and kaiB1 (Supplementary Figure S3).

Another surprise was that knocking out the GroupC kaiCB cluster had essentially no effect on the circadian rhythm (Figure 2D), despite the suggestion of circadian-like functions mediated by GroupC kaiB and kaiC in the purple bacterium Rhodopseudomonas (Ma et al., 2016). The final unexpected result was that knocking out kaiB3 and/or kaiC3 also abolished (or at least dramatically reduced the amplitude of) the rhythm even though kaiC2 is more similar to kaiC1 (and kaiC from S. elongatus) than is kaiC3 (Figure 2E). Therefore, in terms of the circadian gene expression rhythm reported by P _psbA ^Ah ::luxAB, kaiAB1C1 and kaiB3/kaiC3 appear to be important for rhythmicity, whereas kaiC2B2 is not. The adaptive fitness value of each of these kai clusters was assessed by competition experiments described below.

In addition to persisting free-running rhythmicity under constant conditions, the other two defining properties of circadian rhythms are “temperature compensation” of the FRP and entrainment by environmental cycles (usually light/dark signals; (Pittendrigh, 1960; Johnson et al., 2017), both of which were established for the S. elongatus rhythms (Kondo et al., 1993). Consistent with a previous report (Aoki et al., 1995), temperature compensation of the FRP for Synechocystis is clearly demonstrated by the near-independence of FRP over a temperature gradient of 25°C–35°C (Figure 3A). Over this temperature range, the FRP alters by only ∼2 h, leading to a Q₁₀ value of 1.08 (1.00 would be perfectly temperature independent), which is in the typical range for circadian clocks (Sweeney & Hastings, 1960). Panels B & C of Figure 3 underscore the value of using the P _psbA ^Ah ::luxAB reporter at 25°C and 35°C, where the amplitude of the P _psbA ^Ah ::luxAB rhythms are robust, whereas the amplitudes reported by P _dnaK ::luxAB are poor (and even somewhat bimodal) at the lower and higher temperatures. Entrainment to environmental cycles is another defining property, which can often be assessed by measuring the phase-dependent resetting by light or dark pulses that can be graphed as a Phase Response Curve (PRC) (Johnson, 1990). Because we are measuring Synechocystis rhythms in LL, we used dark pulses to elicit phase resetting. As expected, phase resetting to 6-h dark pulses exhibit a PRC with significant phase advances in the subjective day (CT 0-12) and phase delays in the subjective night (CT 12-24, see Figure 4), which is a commonly observed pattern for dark pulse PRCs from microorganisms to mammals (Boulos & Rusak, 1982; Johnson et al., 1989), and is appropriately in antiphase with the previously reported light-pulse PRC for Synechocystis (Aoki et al., 1997). Thus, all three defining properties of circadian pacemakers are satisfied by the rhythms reported by P _psbA ^Ah ::luxAB in Synechocystis.

In S. elongatus, the P _psbAI ::luxAB reporter enabled a mutant screening that identified a stable of FRP mutants that ultimately led to the identification of the kaiABC gene cluster (Kondo et al., 1994; Ishiura et al., 1998). The brightness and robustness of the P _psbA ^Ah ::luxAB reporter should make possible such a mutant screen in Synechocystis. However, given 1) the large quantity and variety of FRP mutants in S. elongatus for which the altered sequence is known and 2) the excellent sequence conservation between kaiABC in S. elongatus with that of kaiAB1C1 in Synechocystis, it might be possible to create the first FRP mutants in Synechocystis on the basis of the known mutations in S. elongatus. We set out to test that prediction by directed mutagenesis (Figure 5A), and indeed this prediction was upheld. Mutations of Synechocystis’ kaiA gene based on the work of Nishimura and coworkers with S. elongatus (Nishimura et al., 2002) led to the generation of three mutant strains with FRPs from ∼28 h to ∼23 h (kaiA ^D119E, kaiA ^E103K, kaiA ^F224S, Figures 5C–E). Moreover, based on the fantastic range of FRPs generated by mutations to a single residue (Tyrosine⁴⁰²) of S. elongatus’ kaiC that was reported from the Kondo lab (Ito-Miwa et al., 2020), we revealed three mutations in Y402 of Synechocystis’ kaiC1 that ranged from ∼23 h to ∼27 h (kaiC1 ^Y402F , kaiC1 ^Y402M , kaiC1 ^Y402W , Figures 5F–H). The similarity of the effect of these mutations in Synechocystis with those in S. elongatus confirms that the kaiAB1C1 gene is the master clock locus in Synechocystis. However, the results in Figure 2E implies an essential role for kaiC3 that could create a kaiC1-kaiC3 redundancy that might mask mutations in kaiC1, but at least so far, this has not been observed. It would probably be worthwhile to test the effects of the mutations shown in Figure 5 in the kaiB3/kaiC3 background, as well as testing point mutations in kaiC3.

To test if the various kai clusters in Synechocystis influence the fitness of the cells in LL or LD, we performed mixed-strain competitions modeled upon those we pioneered for assessing the adaptive significance of circadian periodicity in S. elongatus (Figure 6; (Woelfle et al., 2004). First, we assessed the single-strain growth rates of the WT, kaiAB1C1-ko, kaiC2B2-ko, and kaiB3/kaiC3-ko strains individually in LL, which is the non-selective environmental condition (Figure 6B). In LL, only the kaiB3/kaiC3-ko strain appears to grow significantly slower than the other three strains. In mixed cultures, WT outcompetes kaiAB1C1-ko even though there was not a significant difference in their growth rates in pure cultures (Figure 6C). The rhythm of the kaiB3/kaiC3-ko strain was severely disrupted (Figure 2E), and its growth was slower which apparently leads to out-competition by WT in mixed cultures (Figures 6B,C). In the clock-selective condition of LD 12:12, the kaiAB1C1-ko has a significantly lower growth rate. And both of the strains whose rhythms are disrupted (Figures 2C,E), namely kaiAB1C1-ko and kaiB3/kaiC3-ko, are out-competed by WT in LD (Figure 6E). Interestingly, rhythms in kaiC2B2-ko seem to be largely unaffected (Figure 2D), and the growth rate and competitive ability of this strain appears to be relatively equivalent to WT in both LL and LD (Figure 6).

Clearly, circadian rhythmicity in Synechocystis is robust (Figures 1–5) and adaptive (Figure 2; Figure 6). In hindsight, it was fortuitous that the first cyanobacterial species to be analyzed genetically for circadian rhythmicity was S. elongatus rather than Synechocystis (Kondo et al., 1993; Kondo et al., 1994; Ishiura et al., 1998). First, S. elongatus has a single kaiABC cluster (Ishiura et al., 1998), whereas Synechocystis and many other cyanobacterial species have multiple copies/clusters of kai genes that might have complicated the initial analysis of the genetic basis of the circadian clockwork (Aoki & Onai, 2009; Schmelling et al., 2021). Second, the first luciferase reporter created in S. elongatus for other experimental goals, the P _psbAI ::luxAB reporter, turned out to exhibit excellent high-amplitude rhythms (Kondo et al., 1993) that were optimal for enlistment to the high-throughput screen for clock mutations (Kondo et al., 1994) that ultimately led to the discovery of the kaiABC clock gene cluster (Ishiura et al., 1998). However, P _psbAI ::luxAB is not uniquely high amplitude in S. elongatus–in fact, in contrast to Synechocystis, most promoters from S. elongatus exhibit high amplitude rhythms (Liu et al., 1995). Therefore, serendipitously S. elongatus turned out to be an optimal species to begin circadian analyses in cyanobacteria.

At first, it might seem surprising that a promoter from a very different organism exhibits a better rhythm than the endogenous promoters that have been tested in Synechocystis. The example of S. elongatus, however, again exemplifies this principle; the conII promoter from E. coli expresses a beautiful rhythm in S. elongatus when coupled to a luxAB reporter (Xu, et al., 2013b). In S. elongatus, this result has been interpreted to mean that the entire chromosome is undergoing circadian cycles of supercoiling/compaction that confer circadian rhythmicity on all promoters, even heterologous promoters (Woelfle et al., 2007; Xu et al., 2013b). The PpsbA ^Ah promoter used for driving the expression of luxAB in this study is among the strongest constitutive promoters characterized in cyanobacteria so far. It is a hybrid of the core promoter region of a chloroplast psbA gene (from Amaranthus hybridus) and an artificial RBS generated using the RBS Calculator (Salis et al., 2009; Xiong et al., 2015). This promoter-RBS hybrid, sometimes with subtle modifications, exhibited superior performance in expressing a variety of genes, such as the ethylene-forming enzyme (Efe), catalase (KatE), and two synthetic hydrogenase operons, rendering up to 12% of the total soluble protein as the target enzyme in Synechocystis (Xiong et al., 2015; Wang et al., 2018a). In addition, the same promoter was also able to drive strong expression of Efe and a limonene synthase in S. elongatus, and this capability was used to eliminate pathway bottlenecks (Wang X. et al., 2016; Wang et al., 2021a). Investigating the behaviors of biotechnology-relevant strong promoters, such as the herein studied PpsbA ^Ah promoter, in response to the circadian rhythm regulation in Synechocystis could provide important information for guiding the synthetic biology and metabolic engineering efforts in Synechocystis in the future.

As we previously found for S. elongatus (Woelfle et al., 2004), the rhythmic WT strain of Synechocystis was able to outcompete arhythmic strains in LD (Figure 6). This experimental result clearly indicates the adaptiveness of appropriately aligning an internal oscillator with the external environmental cycle, and therefore that “desynchrony” of these two rhythms hampers competitiveness. Interestingly, arhythmic strains are able to compete successfully with WT in the non-selective LL conditions for S. elongatus (Woelfle et al., 2004), whereas WT Synechocystis is also dominant in LL (Figure 6), implying a fitness advantage for temporal organization in Synechocystis even when the environment is not rhythmic. On the basis of these knockout studies, the kaiC2B2 cluster appears to not contribute to rhythmicity nor to fitness (Figure 2; Figure 6). This result is unexpected, given that GroupC kai genes do contribute to fitness in rhythmic environments in Rhodopseudomonas (Ma et al., 2016). Interestingly, we found that knocking out kaiB3 and/or kaiC3 drastically attenuated the circadian rhythm in Synechocystis, a phenotype that was unexpected when compared to the findings from a previous report in which ΔkaiC3 mutant did not exhibit any apparent growth defect relative to the WT Synechocystis strain under photoautotrophic and photomixotrophic conditions during spot assays (Dörrich et al., 2014). In the current study, we not only found that growth of the ΔkaiB3C3 mutant was significantly impaired (Figure 6), but also that the circadian rhythmicity of Synechocystis cells was eliminated when either kaiB3 or kaiC3 was knocked out (Figure 2). Our results hence provided sufficient evidence that Group B kai genes, i.e., kaiB3 and kaiC3, are essential to the circadian clock of Synechocystis.

In this study, we developed a new methodology for measuring the proportions of two strains in mixed cultures based on QPCR rather than on our previous method of differential antibiotic sensitivity and colony-forming unit (CFU) counting (Woelfle et al., 2004). An advantage of our new methodology is that both strains in a mixed culture can harbor the same antibiotic resistance, thereby avoiding the complication of whether having different antibiotic resistance genes in the two strains are having a fitness effect (we always undertake our competition assays in the total absence of antibiotics, but the resistance genes themselves might have an impact on intracellular metabolism). A potential disadvantage of the new QPCR method is that we are measuring chromosome copy numbers rather than cell numbers or CFUs. WT Synechocystis cells have an average number of chromosome copies of three to six per cell (Pope et al., 2020). However, if the number of chromosome copies has been changed in any of the kai knockout strains, then the absolute proportion of the cells might be biased. We consider this to be a minor liability because even if the two tested strains have different numbers of chromosomes/cell, the relative proportions and their change over time will still be valid. And for the competition assay, the most important factor to be measured is the change over time of the proportion of one strain to the other. Therefore, we believe that the advantages of the new method outweigh its potential disadvantages.

As stated above, S. elongatus has functioned as an excellent system that provided superb insights into circadian mechanisms and adaptiveness, but its “simple” kaiABC genetic background makes it impossible to study the orchestration of Kai proteins from different Kai Gene Groups. On the other hand, the more complicated kai genetics of Synechocystis may provide a more realistic view of redundant genetics in a prokaryote. Moreover, the mechanism of the GroupB and GroupC Kai proteins may be most easily addressed in Synechocystis and subsequently those insights can be applied to the analysis of other species of bacteria that harbor those other versions of the kai genes. In that sense, Synechocystis may be a more productive “springboard” towards understanding circadian rhythmicity in non-cyanobacterial species than has been S. elongatus. Therefore, Synechocystis may be the model system that enlarges the scope of our comprehensive understanding and appreciation of daily timekeeping mechanisms in bacteria.