Circadian cycle mutations may cause adaptive changes in cyanobacteria

PCC 7942, a strain with a planktonic phenotype, incapable of forming biofilms, was chosen for a LTE experiment under serial passage (26). A master culture of PCC 7942 was retrieved from the Pasteur Cultures of Cyanobacteria (PCC) collection to serve as the initial population. During the evolution experiment, cells were grown in BG11 medium at 41 °C, under continuous illumination at 1,313 µmol photons m⁻² s⁻¹, and constant bubbling of 5% CO₂. Culture dilutions were performed every six to seven generations, starting with a 1:130 dilution that ensured an effective population size of at least 10⁹ cells. After 1,248 generations, we obtained an evolved strain, named C11, which reached saturation significantly faster than its ancestral counterpart. Analysis of the pigment content of C11 revealed a statistically significant increase in chlorophyll a (Chla), which was 14% increased in C11, compared to the wt (SI Appendix, Fig. S1↗). However, when we compared the colony-forming units (CFUs) obtained per unit of optical density measured at 720 nm unit (OD₇₂₀), no significant differences between the evolved and ancestral strains were observed (SI Appendix, Fig. S1↗). This allowed us to compare the growth rates of both strains by tracking OD₇₂₀ over time (Fig. 1). Growth curves in batch cultures showed that, under conditions of high light intensity (HL, 983 µmol photons m⁻² s⁻¹), high temperature (HT, 41 °C), and high CO₂ levels (HC, 3%), the generation time of C11 was 1.6 h, representing a 654% increase compared to the wild-type (wt) (Fig. 1 A, Left). This fitness gain, however, depended on the specific environmental conditions. Reducing the illumination to 120 µmol photons m⁻² s⁻¹ (LL) or lowering the temperature to 30 °C (LT) resulted in a decrease in the fitness of C11 (Fig. 1B). The effect of light intensity was most pronounced at HC, where a reduction from 983 to 120 µmol photons m⁻² s⁻¹ led to a drastic drop in the growth rate (Fig. 1A). Overall, carbon availability was the most critical environmental factor. When C11 was grown under atmospheric CO₂ concentrations (LC, 0.04%), its growth rate decreased sixfold, regardless of light intensity (Fig. 1 A, Right). As a result, at low CO₂ levels the wt grew faster than C11 in all conditions tested (SI Appendix, Fig. S2↗). Altogether, these results indicate that the adaptation to the physical environment of the experiment (HL, HT, HC), resulted in a maladaptive phenotype in different ecological conditions.

To identify the causes of adaptation, we sequenced the genome of C11 and compared it with its ancestral strain. A frozen sample of the evolving population obtained at generation 824 (G824) was also introduced in the analysis to follow evolutionary trajectories (Dataset S1↗). When the genomes of C11 and PCC 7942 were compared, we identified four mutations that were largely fixed (present in >80% of the reads) in C11 (Fig. 1C). A mutation in pilA1 (Synpcc7942_0048), a gene required for natural competence in cyanobacteria, was found in 88% of the C11 population. This mutation caused loss of transformability (SI Appendix, Fig. S3↗), a phenotype often observed in model cyanobacterial strains under laboratory cultivation (24, 27). Three additional mutations were fixed in the C11 population. The first one caused an early STOP codon in the gene Synpcc7942_0329, encoding a putative class I SAM-dependent methyl transferase (Refseq: WP_011377536). We named this mutation hMT, hypothetical Methyl Transferase. Homologs of this protein in Synechocystis sp. PCC 6803 and Synechococcus sp. PCC 7002 (CpcM) have been involved in the posttranslational methylation of phycobiliproteins, contributing to the efficiency of energy transfer in the photosynthetic chain (28). This mutation truncated the polypeptide at R71 which, given the length of the wt protein, is likely to result in loss of function. The second mutation (P_aroK) involves a C→T transition in the promoter region of aroK (Synpcc7942_0894, P_aroK), a gene encoding shikimate kinase, which catalyzes a key phosphorylation reaction in the chorismate synthesis pathway (29). Finally, the third mutation (sasAΔ30) introduces a premature STOP codon in sasA (Synpcc7942_2114), a gene involved in circadian regulation. SasA is a histidine-kinase that phosphorylates RpaA, the master regulator of circadian promoters (6, 30). The sasAΔ30 mutation in C11 truncates SasA at Q352, leading to a 30-amino acid deletion at the C-terminus, which contains the histidine kinase domain. Among these mutations, both P_aroK and sasAΔ30 were present in the population at G842 (Dataset S1↗), whereas hMT emerged later in the evolutionary process.

To test the impact of these mutations, we introduced them in the genome of PCC 7942, replacing wt alleles with their mutant counterparts. The growth rate of each individual mutant was measured under various environmental conditions and compared to the wt (Fig. 1B). The P_aroK mutation alone did not improve growth in any of the conditions tested. Both sasAΔ30 and hMT outperformed the WT under LTE conditions but grew significantly slower than C11. These results show that, in isolation, none of the mutations recapitulated the phenotype of C11. We thus introduced combinations of mutations in the genome of the wt. As shown in Fig. 1B, the combination of P_aroK and sasAΔ30 resulted in growth rates equivalent to that of C11. As it was the case for the C11 isolate, the growth advantage brought about by the P_aroK and sasAΔ30 mutations was only observed in HL, HT, and HC conditions. All our attempts at producing a triple mutant (hMT, P_aroK, sasAΔ30) under standard conditions for genetic transformation (31) were unsuccessful, and the hMT mutation proved lethal in HL and LT (Fig. 1B), a phenotype previously reported for its homolog in Synechocystis (28). These results thus indicate that the P_aroK and sasAΔ30 mutations are key for recapitulating the fast-growth phenotype of the evolved strain. A two-way ANOVA test for the main effects and interactions of environmental conditions and mutations revealed that C11, sasAΔ30, and the P_aroK/sasAΔ30 double mutant globally improved their growth rate as the light intensity increased. In contrast, increasing light and temperature led the wt and the P_aroKto lower growth rates, thus underscoring the specific adaptation of C11 to the LTE conditions (SI Appendix, Fig. S4 and Table S1↗).

To investigate the causes of the fast-growth phenotype, we conducted comparative transcriptomic analyses of C11 and the wt under high light and low light conditions (Dataset S2↗). We first identified genes that were differentially expressed in response to high-light intensity. For this purpose, we measured the transcription levels in the wt when grown at 983 µmol photons m⁻² s⁻¹ (HL), and 120 µmol photons m⁻² s⁻¹ (LL). As shown in the volcano plot in Fig. 2A (Left panel), among the genes with increased transcription, we identified well-known high-light-responsive genes, such as psbA2, which encodes the D1 protein of photosystem II. Genes downregulated by HL in the wt included the sigma factor sigC, several hydrogenase-related genes (hox, pnt, hyp), and genes involved in glycogen metabolism and the oxidative pentose phosphate (oxPP) pathway.

We then compared the transcriptomes of C11 and the wt in LL, and the volcano plot of this comparison (Fig. 2 A, Right panel) was surprisingly similar to the previous one (Fig. 2 A, Left panel). This indicated that C11 showed a transcriptional profile equivalent to that of the wt grown under HL. To clarify this observation, we also measured the transcriptomes of C11 and the wt in HL (SI Appendix, Fig. S5↗). A comparison of the transcription levels between the evolved and the ancestral strain under HL and LL revealed that genes downregulated by HL in the wt were constitutively repressed in C11 (Fig. 2 B–D). Similarly, genes upregulated by HL in the wt were also constitutively overexpressed in the evolved strain (SI Appendix, Fig. S6↗). Thus, the transcriptome of C11 mirrors that of the wt exposed to high-light intensities.

Many of the genes transcriptionally altered in C11 are involved in key metabolic pathways. For this reason, we carried out a metabolomic comparison between C11 and the wt under LL conditions. Results, summarized in Fig. 3, and detailed in SI Appendix, Fig. S7↗ and Dataset S3↗, demonstrate that C11 exhibited increased levels of several intermediates of the Calvin–Benson–Bassham cycle (CBB), including fructose 6-phosphate and sedoheptulose 7-phosphate (Fig. 3B). Anaplerotic carbon fixation into oxaloacetate was also apparently enhanced in C11, as judged from higher oxaloacetate levels (Fig. 3B) along with increased transcription of the PEPC enzyme (Dataset S2↗). Taken together, these results pointed to an increased carbon fixation rate in C11, which would explain its faster growth rate. A global shift toward anabolic metabolism was also apparent in the expression of the two isoforms of fructose 1,6-bisphosphatase (fbp and fbpI). C11 showed a downregulation of fbp, and a moderate increase in fbpI, which is required to sustain photosynthesis and gluconeogenesis in Synechococcus (32). Similarly, the gap1 gene, encoding an NAD⁺-dependent glyceraldehyde-3-phosphate dehydrogenase, was downregulated in strain C11, while the expression of the gap2 gene, an NADP⁺-dependent isoform, remained unchanged (Dataset S2↗). In Synechocystis sp. PCC 6803 and other eubacteria, gap1 is involved in catabolism, whereas gap2 has anabolic functions (33).

Increased carbon fixation and rerouting of metabolism toward anabolism were accompanied by other shifts in central carbon and energy metabolism. In C11, we observed high levels of glucose 1-phosphate and glucose 6-phosphate (Fig. 3B). The evolved strain also presented higher glycogen levels than the wt strain under a high carbon atmosphere (Fig. 3C). These metabolomic changes correlated with the downregulation of genes involved in the oxPP pathway and glycogen mobilization (Fig. 2C), suggestive of a substantial rerouting of carbon and storage compound fluxes in the evolved strain. Interestingly, similar changes in glycogen metabolism and the oxPP pathway were observed in a ΔrpaA strain, linking perturbations in circadian regulation to the phenotype of C11.

Whole-genome sequencing revealed a sasAΔ30 mutation present in C11, which caused a deletion in the kinase domain of SasA. SasA is the protein that phosphorylates the circadian regulator RpaA to its active conformation. Since an active RpaA is required to sustain circadian transcription, the sasAΔ30 mutation could potentially disrupt the circadian rhythm of C11. To test this possibility, we measured the expression of sigC, a prototypical Class I gene (9, 34). For this purpose, a yellow fluorescent protein (YFP_LVA) was transcriptionally fused to the sigC promoter P_sigC (34). This transcriptional reporter was introduced into the NS1 neutral site of the wt, C11, sasAΔ30, P_aroK, and hMT strains. P_sigC expression profiles were obtained by measuring YFP levels in single cells using time-lapse microscopy (Fig. 4 and Movies S1↗–S4↗). In the wt, P_sigC showed the fluorescence profile expected of a prototypic Class I promoter (Fig. 4A), with a peak at dusk, a trough at dawn, and a circadian period of 23.8 ± 0 h (SI Appendix, Fig. S8↗). In contrast, P_sigC expression in C11 was arrhythmic (Fig. 4B). To identify the mutation driving this phenotype, we analyzed the expression profile of P_sigC in the different point mutants. P_sigC activity in the P_aroK mutant background showed a robust circadian behavior (Fig. 4D), with a period of 23.5 ± 0 h (SI Appendix, Fig. S9↗). The period in this mutant was thus similar to the wt, although the amplitude of the circadian fluctuation in P_sigC increased 3-fold. P_sigC circadian fluctuations in the hMT mutant maintained the amplitude of the wt but displayed a slightly shorter period of 21.4 ± 5.316 h (SI Appendix, Figs. S10 and S11↗). P_sigC activity in the sasAΔ30 mutant showed an arrhythmic phenotype, similar to C11 (Fig. 4C). Results thus indicated that this mutation is key to the loss of circadian expression in the evolved strain. The defect in C11 rhythmicity was confirmed using a P_psbA1-YFP construct, which yielded equivalent results in the first 48 h of the experiment. (SI Appendix, Fig. S12↗). Overall, results showed that C11 displayed an abnormal circadian cycle, with sasAΔ30 as the driving mutation.

Deletions of sasA or rpaA have been shown to impair growth in light–dark cycles (LD) and cause darkness-induced lethality, a phenotype where cells die after being exposed to short periods of total darkness (6, 30, 35). We thus measured the growth rates of C11, the wt, and the individual mutants in diel cycles (12 h light/12 h darkness), varying the light intensity during daytime. Cycles of 12 h light/12 h darkness at 41 °C and LC were found to be lethal for the wt, thus measurements were performed at 30 °C. Under these conditions, when we grew cells in a LD cycle characterized by a moderate diurnal light intensity (120 µmol photons m⁻² s⁻¹), C11 and sasAΔ30 showed decreased growth rates after 72 h (Fig. 4G). Interestingly, the strain carrying the P_aroK mutation grew faster than the wt, and, when introduced into a sasAΔ30 background, this mutation restored the growth rate to wt levels. The advantage provided by the P_aroK mutation was most obvious when the LD conditions included a strong daylight intensity (983 µmol photons m⁻² s⁻¹). Under these conditions, even the wt arrested growth after 72 h, while the P_aroK mutant was able to keep growing at the fastest pace (Fig. 4H). Both C11 and the double mutant P_aroK/sasAΔ30 also suffered from arrested growth in these conditions.

To test whether the impaired growth of C11 in LD cycles was due to darkness-induced lethality, as in a ΔrpaA mutant (15), we measured cell viability after a period of total darkness. For this purpose, we grew cells under continuous light, exposing them to a period of total darkness of variable duration, and then measured the viability of the culture. As shown in Fig. 4E, we observed no darkness-induced lethality when cells were grown at very low light intensity (30 µmol photons m⁻² s⁻¹), as previously reported for the ΔrpaA strain (15). When cells were grown at 120 µmol photons m⁻² s⁻¹, however, C11 showed a sharp decrease in viability after 60 min of darkness (Fig. 4F). Our results thus indicated that C11 presents levels of darkness sensitivity similar to a ΔrpaA mutant.

C11 showed some hallmarks of a ΔrpaA strain, such as an arrhythmic expression in P_sigC and P_psbA1, darkness-induced lethality, and the repression of genes involved in glycogen metabolism and the oxPP pathway. For these reasons, we compared the transcriptomes of C11 (and each of the point mutants) with that of a ΔrpaA strain. We calculated the fold-change values with respect to the wt, and we plotted them against the fold-change values of a ΔrpaA strain (Fig. 5 A–D). In the case of the P_aroK mutant strain (Fig. 5C), values scattered along a flat line across the x-axis, indicating no correlation with ΔrpaA (Pearson's correlation coefficient, r = 0.06). However, a positive correlation was observed in C11 (r = 0.89) and sasAΔ30 (r = 0.71), as shown in Fig. 5 B and C. The transcription of genes highly repressed in these strains was also repressed in a ΔrpaA background. All 20 genes that showed the highest downregulation in C11 and sasAΔ30 were repressed by at least fivefold in ΔrpaA (Fig. 5E). Altogether, these results indicate that the sasAΔ30 mutation caused transcriptomic changes similar to that observed in a ΔrpaA mutant, suggesting a perturbed SasA-RpaA transduction pathway in sasAΔ30 and C11.

Intriguingly, UTEX 2973, a strain that, like C11, grows significantly faster under HL conditions, also harbors mutations in RpaA that are essential for its phenotype (25). We thus took advantage of the transcriptomic data available for UTEX 2973, and compared it, as in the previous case, with that of a ΔrpaA mutant. As shown in Fig. 5D, we found a weak positive correlation (r = 0.25), but significantly lower than the correlation levels observed between ΔrpaA and C11 (r = 0.89) or ΔrpaA and sasAΔ30 (r = 0.71). The correlation between UTEX 2973 and ΔrpaA was mostly driven by genes that are highly repressed in both strains with respect to the wt (upper right quadrant in panel D, Fig. 5). Indeed, a close inspection revealed a remarkable degree of overlap among the genes present in this upper right quadrant in the comparisons with C11, sasAΔ30, and UTEX 2973. This indicates that there is a common set of genes that are highly repressed in ΔrpaA, which are also downregulated in C11, sasAΔ30, and UTEX 2973 (Fig. 5E). These include the sigma factor sigC, the hox hydrogenases, and the gap1 glyceraldehyde dehydrogenase.

Since C11 exhibited traits similar to a ΔrpaA mutant, we studied gene expression levels across its entire genome in light–dark cycles. Under these conditions, the transcriptome of Synechococcus experiences the effects of the endogenous circadian cycle, as well as the impact of external fluctuations in light intensity during the day/night phases. These measurements are thus intended to identify the actual changes in expression experienced by C11, in comparison to the wt, when grown in physiological diel conditions. For this purpose, cells were grown in cycles of 12 h light/12 h darkness for four days, and total RNA was extracted at dusk of the third day, and the dawn and dusk of the fourth. We then performed a transcriptomic analysis, comparing gene expression levels at these three points (Dataset S4↗). In the wt, the expression of a majority of genes (2,159) exhibited a peak during dusk, thus behaving as Class I (Fig. 6A). A second set (428) showed a peak in transcription during dawn, characteristic of Class II genes. Finally, a total of 128 genes peaked neither at dusk nor dawn and were considered noncircadian. We restricted our analysis to genes with a strong circadian character, selecting those in which the amplitude of the fluctuation was higher than four SD of the average (P-value < 0.001). Using this criterion, we identified a total of 479 representative genes belonging to Class I and 43 genes belonging to Class II (Fig. 6B). When we analyzed these genes in C11, we found profound changes in their expression profiles. The most apparent was observed in Class II genes, which radically altered their phase. When we compared their profiles in the wt and C11 at low light intensities, we detected that most of the genes in this class have switched their phase (Fig. 6 A, Right panel). Instead of exhibiting a peak at dawn, as they do in the wt, Class II genes in C11 showed higher expression levels at dusk. This inversion of phase was observed in the entire class (Fig. 6D), affecting canonical genes such as purF (9) (Fig. 6D). As shown in the figure, this effect was caused by a repression of nighttime transcription, which pushed expression levels at dawn below those observed at dusk.

In C11, Class I genes also experienced decreased transcription during nighttime. However, since these genes are naturally downregulated during the night, the phase of this class was generally unaffected, and the most significant impact was an increase in the amplitude of the circadian fluctuation. We nevertheless identified a subset of Class I genes in which the downregulation of transcription also occurred during daytime, leading to the loss of their circadian character. Such was the case of sigC (Fig. 6C), confirming the results obtained in time-lapse experiments (Fig. 4B). Other classical Class I genes, such as kaiA or kaiC were, however, unaffected and maintained the phase and amplitude of their circadian fluctuation (Fig. 6C). These results thus demonstrate that, under physiological diel conditions, the circadian cycle is largely retained in C11, but the phasing and amplitude of many genes (especially of Class II) is significantly altered.

The circadian alteration in C11, together with the constitutive repression of genes downregulated by high light, suggested that light intensity during the day could somehow influence the phase and amplitude of the circadian rhythm in the wt. To study this possibility, we analyzed the impact of high intensity illumination during daytime in the circadian cycle of PCC 7942. We repeated the previous experiment, but in this case cells were grown in diel cycles in which the light intensity during daytime was increased to 983 µmol photons m⁻² s⁻¹. Results, shown in Fig. 6 A and B, indicated a global nighttime repression of all Class II genes, which again led to the reversal of their circadian phase. For Class I genes, however, expression levels during the night were elevated, leading to shallower fluctuations. In sigC, and the subset of genes that were downregulated and arrhythmic in C11, the circadian cycle was also abated. Thus, C11 presented some circadian perturbations that were also observed in wt cells exposed to high intensity illumination. The switching of Class II genes to a Class I phase, and the downregulation and loss of rhythmicity in sigC were the most conspicuous features.

S. elongatus PCC 7942 (PCC 7942) and all derived strains were routinely grown and maintained in liquid BG11 medium supplemented with appropriate antibiotics and 10 mM bicarbonate, when CO₂ concentrations above atmospheric levels were used. Antibiotics used for selecting PCC 7942 were neomycin at 5 or 25 µg mL⁻¹ (Neo5 or Neo25), spectinomycin at 10 or 20 µg mL⁻¹ (Sp10 or Sp20) and streptomycin at 10 or 50 µg mL⁻¹ (Sm10 or Sm50). Except otherwise stated, cells were grown under a continuous light flux of 60 µmol photons m⁻² s⁻¹ from white fluorescence lamps, in a Sanyo Plant Growth Chamber, at 30 °C and constant ambient air supply. Experiments with different light intensities, diel cycles, and CO₂ concentrations were performed in a MC 1000-OD Multicultivator (Photon Systems Instruments), equipped with programmable temperature, CO₂, and light intensity (cold white LED). Doubling times were determined from growth curves performed in a MC 1000-OD Multicultivator (Photon Systems Instruments). Growth was estimated from OD₇₂₀ and OD₆₈₀ measurements obtained every 10 min by the MC 1000-OD Multicultivator. The effective growth rate (K′) was obtained by the slope of the Ln (OD₇₂₀) in the early exponential phase of each growth curve. The doubling time was calculated as Ln2/K′.

For the evolution experiment, a single colony of PCC 7942 was grown in liquid BG11 medium in 250 mL glass tubes with a continuous light flux of 1,313 µmol photons m⁻² s⁻¹ at 41 °C and 5% of CO₂. A control population was grown in parallel, at 25 °C, atmospheric CO₂ levels and a continuous light flux of 65 µmol photons m⁻² s⁻¹. Cells were grown under serial passage in 200 mL of BG11 supplemented with 10 mM sodium bicarbonate. Iterations started from cultures at OD₇₂₀ = 0.01 which were left to grow for 48 h. At the end of the iteration the OD₇₂₀ was measured, and appropriate dilutions were performed to restart growth at OD₇₂₀ = 0.01. The number of generations in each iteration n was calculated as n = Log₂(OD_f) - Log₂(OD_i) where OD_f and OD_i correspond to the final and initial OD₇₂₀ of the iteration, respectively. Contamination was ruled out in each iteration by plating 100 μL of the culture on LB agar plates and letting them grow at 37 °C for 48 h. Whenever contamination was detected, the culture was discarded, and the LTE was resumed from the previous, uncontaminated batch. Whole genome sequencing was performed at generations 0, 824 (37 passages), and 1,284 (54 passages) to track the emergence of mutations during the experiment.

Genetic engineering in PCC 7942 was performed through natural transformation (31). For each transformation, approximately 4 × 10⁹ cells, equivalent to 10 μg of chlorophyll, were mixed with 500 ng of DNA, and incubated in the dark for 16 h at 30 °C and 100 rpm. Transformation mixtures were then deposited onto a 0.45 μm nitrocellulose filter (Millipore) and incubated for 24 h on top of BG11 plates supplemented with appropriate antibiotics, at 30 °C and continuous light. The filter was transferred to new plates with antibiotics every 24 to 72 h. Transformation colonies appeared after 7 to 14 d. Segregation of mutants was achieved by repeatedly streaking individual transformant colonies on selective plates. Mutant genotypes were confirmed by PCR and DNA sequencing using specific primers.

The gene replacement method used to insert the mutations from C11 in the wt strain was based on Matsuoka et al. (59). Briefly, a streptomycin resistant wt strain (MSM1) was obtained by naturally transforming PCC 7942 with the K43R recessive allele of the ribosomal protein S12 (rps12-R43) and selection of SmR resistant transformants. This SmR strain was used for the introduction of the individual mutations of C11 in the wt background. A set of two plasmids was designed to insert each mutation. The first one contained a kanamycin/neomycin resistance cassette and the Synechocystis sp. PCC 6803 wt version of the rps12 gene cloned under the promoter of PCC 7942 psbA1 gene. This gene was flanked by the regions immediately downstream or upstream the gene of interest. Transformants were selected on Neo25 plates and confirmed by their Neo^r/Sm^s phenotype and PCR. To introduce the desired mutation, a second plasmid containing the mutated gene of interest and its flanking regions was introduced through natural transformation. These transformants were then selected through their Neo^s/Sm^r phenotype. Transformants were checked through PCR and Sanger sequencing using appropriate primers. The complete list of mutant strains and plasmids used for their construction can be observed in Supporting information (SI Appendix, Tables S2 and S3↗).

Total RNA extraction protocol was an adaptation from Hein et al. (60). 30 mL of a culture at OD₇₂₀ ~ 0.6 were collected by centrifugation at 4,000×g for 10 min at RT. The pellet was resuspended in 1.5 mL of Trizol (Ambion-USA) and the suspension was divided in two precooled RNAse-free Eppendorf tubes and immediately frozen in liquid nitrogen and stored at −80 °C. To extract the RNA, the samples were incubated at 65 °C for 15 min, being vortexed several times in the process. 525 µL of chloroform:isoamiloalcohol (24:1) were added. After a 10 min incubation at RT, mixing gently several times, the samples were centrifuged at 6,000×g, 3 min at RT. The aqueous phase (470 µL) was transferred to new RNAse-free tubes and 470 µL of chloroform:isoamiloalcohol (24:1) were added, repeating the previous steps. The aqueous phase (370 µL) was transferred again to a new tube, mixed with the other half it was previously separated from. One volume (740 µL) of isopropanol was added, gently mixed, and left precipitating overnight at −80 °C. The RNA was pulled down by 12,000×g centrifugation at 4 °C, for 30 min. The pellet was washed three times with 200 µL of ethanol 75% and left to dry. The clean RNA was resuspended in 54 µL of RNAse-free distilled water. 2 µL per lane were loaded in a 2% agarose gel to value sample's integrity, and 2 µL were used to measure purity and concentration in a Nanodrop (Thermo). The remaining 50 µL were subjected to a DNAse digestion protocol (Invitrogen) and an additional cleaning step running the final volume through an RNA extraction kit column (Qiagen).

The sequencing reads from each sample were trimmed for Illumina adapters remnants and low-quality regions (<Q25) using Trim-Galore (v0.6.2) (61). Surviving high-quality reads were posteriorly aligned against three bacterial rRNA databases (5S, 16S and 23S) from SILVA (https://www.arb-silva.de/↗) and Rfam (https://rfam.org/↗) with SortMeRNA to filter out rRNA fragments (v4.3.6) (62). The processed reads were aligned against the complete genome sequence of S. elongatus, strain PCC 7942 (GenBank accession number: GCF_000012525.1) using Hisat (v2.2.1) (63), processing the intermediate alignment files and merging the technical replicates with Samtools (v1.6) (64). Individual sample gene expression levels were quantified with featureCounts (v2.0.1) (65). Differential expression of mRNA transcripts was computed with DESeq2 (v1.34.0) (66), using a Padj value < 10⁻⁸ as statistical cutoff. In parallel, normalized expression values were calculated using StringTie (v2.2.1) (67) in Transcripts Per Million (TPMs).

Statistical analyses were performed using R-Studio (version 2023.03.0 Build 386, R version 4.2.1) and GraphPad Prism (version 8.3.0). Shapiro–Wilk normality test, followed by a one-way ANOVA and Tukey's multiple comparison test were used to determine similarities between each strain growth rate for each condition tested. Growth rates grouped by the same letter do not differ statistically with a P-value > 0.05. Unpaired Student's t test was used to evaluate differences between two strains, wt and C11, in relation to transcriptomic and metabolic data.

SI Appendix, SI Methods↗, details materials and methods for the determination of pigment and glycogen content, as well as protocols and procedures employed in metabolomic analyses and time-lapse microscopy.