Introduction

Cyanobacteria carry out oxygenic photosynthesis but sometimes experience microoxic conditions during the night. Subsequently, cyanobacteria perfrom fermentation, using glycolytic products as electron acceptors. (1). Cyanobacteria have multiple fermentation pathways according to the environment. For example, the freshwater living cyanobacterium Synechocystis sp. PCC 6803 (hereafter referred to as Synechocystis) produces acetate, lactate, dicarboxylic acids, and hydrogen (1, 2). Genetic manipulations on Synechocystis have demonstrated that modulating the expression of certain enzymes can alter fermentative metabolic flow (3, 4). This provides evidence that transcription regulates the fermentative pathway.

Hydrogen has attracted significant attention among other cyanobacterial bioproducts owing to its potential as a biofuel. Bidirectional hydrogenase is a key enzyme for H2 production from protons (5) and is commonly found in cyanobacteria (6). Cyanobacterial hydrogenase comprises five subunits (HoxEFUHY) containing nickel and Fe-S clusters (7). This enzyme can utilize NADH, reduced ferredoxin, and flavodoxin as substrates (8). Hydrogenase can also use excess electrons from the photosynthetic electron transport chain, but photosystem II-generated oxygen can inhibit its activity (9). Another metal-ion-harboring and O2-sensitive enzyme, pyruvate-ferredoxin oxidoreductase (PFOR), oxidizes pyruvate to produce acetyl- CoA during fermentation. PFOR works coordinatedly with hydrogenase to reduce ferredoxin, which hydrogenase uses as an electron donor (8, 10). Despite their O2 sensitivity, hydrogenase and PFOR work under mixotrophic (i.e., aerobic) conditions (11, 12). Therefore, uncontrolled expression of hox operon and nifJ may hamper metabolism under photosynthetic conditions or shorten the supply of inorganic cofactors. Thus, transcriptional regulation in response to the environment is essential for optimal energy cost performance.

Promoter recognition by RNA polymerases is an essential step in transcriptional regulation. Sigma factors, subunits of RNA polymerase, recognize core promoter sequences. Transcription factors can also bind to promoter regions to suppress or promote RNA polymerase transcription. As well as recruitment or blocking of RNA polymerase, some transcriptional factors, known as Nucleoid-associated proteins (NAPs), modulate chromosomal conformation to regulate transcription (13). NAPs are common in bacteria, but cyanobacterial NAPs remain unidentified, and higher order of DNA structure in cyanobacteria is yet to be shown. A recent study suggested that the manipulation of chromosomal supercoiling impacts transcriptional properties in cyanobacteria (14). There is room for consideration of NAPs modulating chromosomal conformation and regulating expression in cyanobacteria.

In Synechosysits, the coding genes of HoxEFUHY form a single operon (sll1220–1226), while PFOR is encoded in the nifJ (sll0741) gene. Both hox and nifJ operons are highly expressed under microoxic conditions (15). Genetic analysis has revealed that multiple global transcriptional regulators control hox and nifJ expression. Sigma factor SigE (Sll1689) promotes the expression of hox and nifJ operons (16, 17), while transcription factor cyAbrB2 (Sll0822) represses them (18, 19). Positive regulators for the hox operon include LexA (Sll1626) and cyAbrB1 (Sll0359) (20–22).

SigE, an alternative sigma factor, controls the expression of genes involved in glycogen catabolism and glycolysis, as well as PFOR/nifJ and hydrogenase (16). SigE shares a high amino acid sequence homology with the primary sigma factor SigA, which is responsible for transcribing housekeeping and photosynthetic genes (23). A ChIP-seq study revealed that, while most SigE-binding sites are the same as SigA, SigE exclusively occupies the promoters of glycogen catabolism and glycolysis (24).

CyAbrB2 and its homolog cyAbrB1 are transcription factors highly conserved in cyanobacteria. For example, cyAbrB homologs in Anabaena sp. PCC7120 is involved in heterocyst formation (25). CyAbrB2 in Synechocystis regulates the expression of several genes involved in carbon metabolism, nitrogen metabolism, and cell division (19, 26, 27). CyAbrB2 binds to the hox promoter in vitro and represses its expression in vivo (18). CyAbrB1, an essential gene, physically interacts with the cyAbrB2 protein (28) and binds the hox promotor region in vitro to promote its expression (20).

To explore the dynamics of those transcription factors governing the expression of hox operon, we conducted a time-course analysis of the transcriptome and ChIP-seq of SigE and cyAbrB2. Our ChIP-seq and transcriptome analysis showed the NAPs-like nature of cyAbrB2, which prompted us to conduct chromosomal conformation capture assay. 3C analysis revealed the physical interaction between the hox promoter region and its downstream and upstream genomic region in the aerobic condition, and that the interaction decreased upon entry to the microoxic condition. Furthermore those interactions in the ΔcyabrB2 mutant were low both in aerobic and microoxic conditions. From those experiments, we propose that cyAbrB2 modulates chromosomal conformation like NAPs, allowing access to the SigE-containing RNA polymerase complex on the hox promoter, by which the hox operon achieves distinct expression dynamics. Chromosomal conformation of bacteria is a growing area of interest, and the findings of this study have brought insight into the transcriptional regulation of cyanobacteria.

Results

Transcriptomes on entry to dark microoxic conditions

To understand transcriptional regulation under microoxic conditions, we conducted a time-course transcriptome capturing aerobic and microoxic conditions at 1-, 2-, and 4-hours timepoints (Figure 1A). Gene set enrichment analysis based on KEGG pathway revealed that many biological pathways, including photosynthesis and respiration (oxidative phosphorylation), were downregulated under microoxic conditions compared to aerobic conditions (Figure 1B). Upregulated pathways included butanoate metabolism and two-component systems, but the actual cyanobacterial catabolic pathway differed slightly from that described in the KEGG pathway. Within 1 hour of switching from aerobic to microoxic conditions, the expression levels of 508 genes increased more than 2-fold. Furthermore, genes with increased expression levels were classified into four based on the time-course (Figure 1C). Of the 508 genes, 28 were termed “transiently upregulated genes” due to their decreased expression upon the comparison of 1- and 4- hours incubation under microoxic conditions (log2 fold change < −0.5), and 119 were termed “continuously upregulated genes.”, which continuously increased between 1- and 4-hours incubation under microoxic conditions (log2-fold change > 0.5). Other than 508 genes 2-fold upregulated within 1 hour, 28 genes showed more than 4-fold increment within 4 hours but not upregulated within 1 hour. We combined those “Late upregulated genes” with 508 genes and referred to as “All upregulated genes” in the subsequent analysis (Figure 1C). Mapping the classified genes to central carbon metabolism revealed that nifJ encoding PFOR and hox operon encoding a bidirectional hydrogenase complex were transiently upregulated (Figure 1D and Table 1). RTqPCR verified the transient expression of hoxF, hoxH, and nifJ (Figure S1).

Figure 1

Table 1

SigE and cyAbrB2 control the expression of transiently upregulated genes

The functional correlation between hydrogenase and PFOR, encoded by the hox operon and nifJ, suggests that transient upregulation has physiological significance. We focused on transiently upregulated genes and attempted to reveal the regulatory mechanism underlying transient upregulation. While SigE promotes the expression of hox and nifJ, cyAbrB2 represses them (16, 18, 19). We confirmed that the deletion of sigE and cyabrb2 (ΔsigE and Δcyabrb2, respectively) affected the expression of hoxF, hoxH, and nifJ by RT- qPCR (Figure S1). Thus, we conducted a time-course transcriptome analysis of ΔsigE and Δcyabrb2 under aerobic conditions and after 1- and 2-hours cultivation in dark microoxic conditions (Figures 2A and S2A).

Figure 2

The transcriptome data confirmed that SigE and cyAbrB2 regulated hox operon expression (Figure 2B). At each time point, we searched for differentially expressed genes (DEGs) between mutants and wildtype with a more than 2-fold expression change and a false discovery rate (FDR) less than 0.05. We found that deleting sigE or cyabrb2 preferentially affected the expression of transiently upregulated genes, not limited to hox and nifJ operons (Figures 2C and 2D). Interestingly, cyabrb2 deletion resulted in the upregulated expression of transient genes under aerobic conditions, in contrast to 1-hour cultivation under microoxic conditions (Figure 2C).

Genome-wide analysis of cyAbrB2, cyAbrB1, and SigE via ChIP-seq

To decipher the regulatory mechanism of transiently upregulated genes, we must first comprehend the fundamental features and functions of these transcriptional regulators. Therefore, a genome-wide survey of cyAbrB2 and SigE occupation (Figures S2 and S3) combined with transcriptome data was done. Specifically, we generated a Synechocystis strain in which cyAbrB2 was epitope-tagged and performed a ChIP-seq assay under aerobic and microoxic conditions (Figures S2B and S2C). SigE-tagged strains previously constructed and analyzed elsewhere were also employed (24). The primary sigma factor SigA was also analyzed to determine SigE-specific binding. In addition to cyAbrB2, we tagged and analysed cyAbrB1, which is the interactor of cyAbrB2 and positively regulates the hox operon.

CyAbrB2 binds to long tracts of the genomic region and suppresses genes in the binding region

The ChIP-seq data showed that cyAbrB2 bound to long tracts of the genomic region with lower GC content than the whole genome Synechocystis (Figures 3A and 3B). Vice versa, regions exhibiting lower GC contents showed a greater binding signal of cyAbrB2 (Figure 3C). This correlation was not a systematic bias of next-generation sequencing because the binding signals of SigE, SigA, and control showed no negative correlation to GC contents (Figure S4A). The binding regions of cyAbrB2 called by peak-caller covered 15.7% of the entire genome length. 805 of 3614 genes overlapped with cyAbrB2 binding regions, and almost half (399 of 805 genes) were entirely covered by cyAbrB2 binding regions. The cyAbrB2 binding regions included 80 of 125 insertion sequence elements (Figure 3D). Comparison with the transcriptome of ΔcyabrB2 revealed cyAbrB2 tended to suppress the genes overlapping with its binding regions under aerobic conditions (Figures 3A and 3E). A survey of the genomic localization of cyAbrB1 under aerobic conditions revealed that cyAbrB1 and cyAbrB2 shared similar binding patterns (Figures 3A and S4C). Due to the essentiality of cyAbrB1, we did not perform transcriptome analysis on a cyAbrB1- disrupted strain.

Figure 3

Localization of cyAbrB2 became blurry under the microoxic condition

When cells entered microoxic conditions, the relative ChIP-seq signals in the cyAbrB2 binding regions slightly declined (Figure S5A). However, the signal change did not necessarily indicate that cyAbrB2 binding declined under microoxic conditions because the relative signal was normalized by total reads of immunoprecipitation and input. Rather, the cyAbrB2 binding pattern became uniform when the cells entered microoxic conditions. Notably, the total quantities of precipitated DNA by tagged cyAbrB2 did not decrease, and qPCR confirmed that the cyAbrB2 binding signal increased in all positions tested (Figure S5B and S5C). The protein amount of cyAbrB2 was not altered on entry to the mircooxic condition (Figure S5D).

CyAbrB2 binds to transiently upregulated genes

The binding regions of cyAbrB2 overlapped 17 of 28 transiently upregulated genes, showing significant enrichment from all upregulated genes (Figure 3F). While cyAbrB2 covered the entire length of insertion sequences and unknown proteins, its binding positions on other transient genes were diverse (Figure 3G). Specifically, the hox and nifJ operons had two distinct binding regions located at the transcription start sites (TSSs) and middle of operons, the pntAB operon had two binding regions in the middle and downstream of the operon, and the nrtABCD operon had one binding region downstream of the operon (Figure 3H). Upon entry into microoxic conditions, the cyAbrB2 binding signal around the transiently upregulated genes became less specific, consistent with the general tendency.

Sigma factors SigE and SigA are excluded from cyAbrB2 binding regions regardless of GC contents

We searched for SigE and SigA binding sites under aerobic and microoxic conditions (Figures 4A and 4B, left and right). The SigE and SigA peaks identified in this study predominantly covered the previously identified peaks (Figure S2D), reproducing the previous study’s conclusion (24); i.e., SigE and the primary sigma factor SigA share localization on the promoters of housekeeping genes, but SigE exclusively binds to the promoters of its dependent genes. SigE and SigA binding peaks were significantly excluded from the cyAbrB2 binding regions (Figure 4C), and signals in cyAbrB2 binding regions were weaker than those in cyAbrB2-free regions (Figure 4D). As cyAbrB2 prefers AT-rich regions, we examined whether the peak summits of SigE and SigA were located in regions with high GC contents. Unexpectedly, no correlation was found between the GC content and binding signals of SigE and SigA (Figures S4A and S4B). Thus, SigA and SigE are excluded from cyAbrB2 binding regions regardless of GC contents.

Figure 4

Dynamics of sigma factors upon exposure to the microoxic condition

When cells entered microoxic conditions, the binding signals of SigA and SigE were changed, although most of their peaks observed under aerobic conditions were present under microoxic conditions (Figure 4A). Next, we focused on sigma factor dynamics in transiently upregulated genes. SigE, but not SigA, binds at the TSS of pntAB under aerobic and microoxic conditions (Figure 4E top). SigE binding summits were not identified at the TSSs of the hox and nifJ operons under aerobic conditions. However, the SigE-specific binding summit appeared at the TSS of nifJ when cells entered microoxic conditions (Figure 4E middle). A bimodal peak of SigE was observed at the TSS of the hox operon, although the peak caller failed to recognize it as two peaks (Figure 4E bottom panel). One side of the bimodal peak marked with an arrow in Figure 4E lacked SigA binding. SigE binding without SigA on the promoters of hox, nifj, and pntAB is consistent with their SigE-dependent expression (Figure 2B).

CyAbrB2-dependent and independent chromatin conformation around hox operon

We have shown that cyAbrB2 broadly binds to AT-rich genomic regions, including insertion element sequences, and represses expression (Figures 3B and 3D). This is functionally similar to the NAPs (13), which makes us hypothesize that cyAbrB2 modulates chromosomal conformation. Therefore, we conducted the chromatin conformation capture (3C) assay against wildtype and cyabrb2Δ strain at aerobic and microoxic conditions (1hr and 4hr). qPCR was performed with unidirectional primer sets, where the genomic fragment containing hox operon (hereinafter hox fragment) was used as bait (Figure 5A). We then observed that the hox fragment interacted with its distal downstream (∼25 kbp) region and proximal upstream (∼10 kbp) region in the aerobic condition (Figure 5B and 5C, locus(c) and (j)). Those interactions were detected both in wildtype and Δcyabrb2. However, the interaction frequency between the hox fragment and its distal downstream was higher in wildtype than Δcyabrb2 (Figure S7 locus(c)).

Figure 5

CyabrB2 mediates chromatin conformation changes upon microoxic conditions

When the wildtype cells entered the microoxic condition, the hox fragment diminished its interaction with its distal downstream and proximal upstream region (Figure 5B locus(c) and (j)). By contrast, the interaction frequency in Δcyabrb2 mutant were low and unchanged in the aerobic and microoxic conditions. While the interaction scores exhibit considerable variability, the individual data over time demonstrate declining trends of the wildtype at locus (c) and (j) (Figure S7). Furthermore, both wildtype and Δcyabrb2 mutant showed increased interaction with the proximal downstream fragments on entry to the microoxic conditions (Figure 5B and 5C, locus (f),(g),(h)). In the wildtype cells, the interaction with locus (g) was transiently elevated at 1 hour of the microoxic condition but decreased to the aerobic level at 4 hours. In summary, 3C analysis demonstrated cyAbrB2-dependent and independent dynamics of chromosomal conformation around the hox operon in response to the microoxic condition (Figure 5D).

Discussion

Physiological significance of transient upregulation of hox and nifJ operons

The transcriptional upregulation of fermentative genes in response to the microoxic condition is adaptive for energy acquisition. Our time-course transcriptome showed upregulation of several genes involved in catabolism upon exposure to the microoxic condition. The transient upregulation of hox and nifJ operons is distinctive among them (Figure 1D).

The reason for transient upregulation may be the resource constraints of inorganic cofactors and the reusability of fermentative products. Hydrogenase and PFOR (the product of nifJ gene) have iron–sulfur clusters, and hydrogenase requires nickel for its activity (29, 30). Continuous expression of the hox operon should be futile under physiological conditions, as the overexpression of hoxEFUYH and hypABDCEF (hydrogenase maturation factors) does not yield a proportional increase in hydrogenase activity without an adequate nickel supply (31). Furthermore, hydrogenase-produced H2 is diffusive and thus difficult to store and reuse as reducing power when cells return to aerobic conditions. Unlike hydrogen fermentation, the reverse tricarboxylic acid (rTCA) cycle produces dicarboxylic acids, which are reusable for generating NADPH via the TCA cycle under aerobic conditions (32). However, adding bicarbonate to phosphoenolpyruvate (PEP) is a rate-limiting step in the rTCA cycle (33). Hydrogenase may favor fermentation initiation since it uses only protons as a substrate, and the rTCA cycle may become active subsequently. In fact, citH/mdh (sll0891) encoding a key enzyme of the rTCA cycle was classified as continuously upregulated genes in this study (Figure 1D).

Mechanisms for transient expression mediated by SigE and cyAbrB2

SigE and cyAbrB2 can independently contribute to the transient transcriptional upregulation. This is evident as the single mutants, ΔsigE or Δcyabrb2, maintained transient expression of hoxF and nifJ (Figure S1). Here we discuss the cyAbrB2-mediated mechanism, then that of SigE will be discussed.

Three types of the localization pattern and function of cyAbrB2

Herein, we classified three types of binding patterns for cyAbrB2 based on the ChIP-seq data of cyAbrB2. The first is that cyAbrB2 binds a long DNA tract covering the entire gene or operon, represented by the insertion sequence elements. CyAbrB2 suppresses expression in this pattern, as shown in Figures 3D and 3E. In the second pattern, cyAbrB2 binds on promoter regions, such as hox operon and nifJ. Binding on those promoters fluctuates in response to environmental changes, thus regulating expression. This pattern also applies to the promoter of sbtA (Na⁺/HCO3⁻ symporter), as the previous study reported that cyAbrB2 is bound to the sbtA promoter in a CO2 concentration-dependent manner (27). The last one is cyAbrB2 binding in the middle or downstream of operons. The middle of hox, pntAB, and nifJ operons and the downstream of nrt operon are the cases (Figure 3G). Our data show that genes in the same operon separated by the cyAbrB2 binding region behave differently. In particular, pntB is classified as the transiently upregulated gene, while pntA is not, despite being in the same operon. This might be explained by the recent study which reported that cyAbrB2 affects the stability of mRNA transcribed from its binding gene (34). The cyAbrB2-mediated stability of mRNA may also account for the decrease in transcript from transient upregulated genes at 4 hours of cultivation. Hereafter, we will focus on the mechanism of the second pattern, regulation by cyAbrB2 on the promoter.

Insights into transcriptional regulation of hox operon and nifJ

Genome-wide analysis indicates that the cyAbrB2-bound region blocks SigE and SigA (Figures 4C and 4D). This is presumably because sigma factors recognize the promoter as a large complex of RNA polymerase. CyAbrB2 binds to the hox and nifJ promoter region and may inhibit access to RNA polymerase complex under aerobic conditions. When cells entered microoxic conditions, the binding pattern of cyAbrB2 on the hox and nifJ promoters got less specific, and SigE binding peaks on those promoters became prominent (Figure 4E). Notably, cyAbrB2 ChIP efficiency at the hox promoter is higher in the microoxic condition than in the aerobic condition (Figure S5C). Hence, the mechanism by which cyAbrB2 inhibits SigE-containing RNAP in the aerobic condition is not solely based on exclusion by occupancy. Rather, it is plausible that chromosomal conformation change governed by cyAbrB2 provides SigE- containing RNAP with access to the promoter region (Figure 5D).

Chromosomal conformation in hox region changes according to the environmental change

3C analysis demonstrated the interaction between the hox promoter and its distal genomic region, which decreased on entry to the microoxic condition. CyAbrB2 mediates the interaction between the hox fragment and the distal downstream region. The observed conformational change correlates with the expression of the hox operon, though further investigations are needed to establish causality (Figures 5B and S7, locus (f), (g), and (j)). A recent study demonstrated that manipulating the expression of topoisomerase, which influence chromosomal conformational change through supercoiling, affect transcriptional properties in cyanobacteria (14). Moreover, Song et al. pointed out that DNA looping may inhibit transcription in cyanobacteria (34) because artificial DNA looping by the LacI repressor of E. coli can repress cyanobacterial transcription (35). Thus, we infer conformation change detected by the present 3C experiment regulates expression of hox operon.

A possible mechanism for cyAbrB2-mediated chromosomal conformation

The hox promoter region possesses the cyAbrB2 binding region, and locus (c) is adjacent to the cyAbrB2 binding region. Consequently, cyAbrB2 may mediate chromosomal conformation by bridging its binding regions, as other NAPs do. However, no biochemical data mentioned the DNA bridging function of cyAbrB2 in the previous studies (18, 26, 34). CyAbrB1, the homolog of cyAbrB2, may cooperatively work, as cyAbrB1 directly interacts with cyAbrB2 (28), and their distribution is similar (Figure 3A). In the microoxic condition, cyAbrB2 binding became blurry, which might reduce the specific interaction between the hox region and locus (c). On the other hand, cyAbrB2 indirectly mediates interaction with its proximal upstream region (Figure 5B, locus (j)), as this region is devoid of cyAbrB2 binding (Figure 5A locus (i) to (l)).

Generality for chromosomal conformation in cyanobacteria

Our 3C analysis revealed that local chromosomal conformation changes upon entry to the microoxic conditions (Figure 5D). As cyAbrB2 occupies about 15% of the entire genome and globally regulates gene expression, cyAbrB2 likely governs the whole chromosomal conformation (Figure 5E). Furthermore, the conformational changes occur in both AbrB2-dependent and AbrB2-independent manners. Cyanobacteria face various environmental changes besides the microoxic condition, and there are potential NAPs in cyanobacteria yet to be characterized. It is speculated that conformational change of the entire chromosome occurs to deal with many environmental stresses.

The sigE-mediated mechanism for the transient expression

One possible SigE-mediated mechanism for transient expression is the post-transcriptional activation and degradation of SigE in the dark; i.e., SigE is sequestered by binding to the H subunit of Magnesium- chelatase under light conditions and released under dark (36), enabling acute transcription of hox operon and nifJ. Transcripts of sigE were continuously downregulated in our time-course transcriptome, while sigB (sll0306) and sigC (sll0184) were classified as continuous upregulated genes (Supplemental Table S1). It is possible that upregulated SigB and SigC outcompete SigE in prolonged incubation under microoxic conditions. Finally, SigE is degraded under dark within 24 hours (37).

Another reason for the microoxic specific expression may exist in the sequence of the hox promoter. We previously determined the consensus sequence of -10 element for SigE regulon in the aerobic condition as “TANNNT,” where N is rich in cytosine (24). The -10 sequence of the hox promoter “TAACAA” (22) deviates from the consensus, and no hexamer precisely fitting the consensus are found in the nifJ promoter. This deviation can inhibit SigE from binding during aerobic conditions, aside from cyAbrB2-mediated inhibition. Under the microoxic condition, transcription factors LexA (22) and Rre34 (15) may aid SigE binding to the promoter of hox and nifJ, respectively. Overall, multiple environmental signals are integrated into the hox and nifJ promoter through the cyAbrB2 and SigE dynamics.

Materials and Methods

Bacterial strains and plasmids

The glucose-tolerant strain of Synechocystis sp. PCC 6803 (38) was used as a wildtype strain in this study. The sigE (sll1689)-disrupted strain (G50) was constructed in a previous study (16). Disruption and epitope- tagging of cyabrb1(sll0359) and cyabrb2(sll0822) were performed by homologous double recombination between the genome and PCR fragment (38). The resulting transformants were selected using three passages on BG-11 plates containing 5 µg/mL kanamycin. Genomic PCR was used to confirm the insertion of epitope tag fragments and gene disruption (Figure S2A). Supplemental Tables S2, S3, and S4 contain the cyanobacterial strains, oligonucleotides, and plasmids used in this study.

Aerobic and microoxic culture conditions

For aerobic conditions, cells were harvested after 24 hours cultivation in HEPES-buffered BG-110 medium (39), which was buffered with 20 mM HEPES-KOH (pH 7.8) containing 5 mM NH4Cl under continuous exposure to white light (40 µmol m⁻²s⁻¹) and bubbled with air containing 1% CO2 (final OD730 = 1.4–1.8). For the dark microoxic culture, the aerobic culture cell was concentrated to an OD730 of 20 with the centrifuge and resuspended in the culture medium. The concentrated cultures were poured into vials, bubbled with N2 gas, and sealed. The sealed vials were shaded and shaken at 30°C for the described times.

Antibodies and immunoblotting

Sample preparation for immunoblotting was performed as previously described (24), and FLAG-tagged proteins were detected by alkaline-phosphatase-conjugated anti-FLAG IgG (A9469, Sigma Aldrich, St. Louis, MO) and 1-Step NBT/BCIP substrate solution (Thermo Fisher Scientific, Waltham, MA).

RNA isolation

Total RNA was isolated with ISOGEN (Nippon gene, Tokyo, Japan) following the manufacturer’s instructions and stored at −80°C until use. The extracted RNA was treated with TURBO DNase (Thermo Fisher Scientific) for 1 hour at 37°C to remove any genomic DNA contamination. We confirmed that the A260/A280 of the extracted RNA was >1.9 by NanoDrop Lite (Thermo Fisher Scientific). We prepared triplicates for each timepoint for the RNA-seq library. RT-qPCR was performed as discribed elsewhere (37).

ChIP assay

Two biological replicates were used for each ChIP-seq experiment, and one untagged control ChIP was performed. ChIP and qPCR analyses were performed using the modified version of a previous method (24). FLAG-tagged proteins were immunoprecipitated with FLAG-M2 antibody (F1804 Sigma-Aldrich) conjugated to protein G dynabeads (Thermo Fisher Scientific).

Library preparation and next-generation sequencing

For the ChIP-seq library, input and immunoprecipitated DNA were prepared into multiplexed libraries using NEBNext Ultra II DNA Library Prep Kit for Illumina (New England Biolabs, Ipswich, MA). For the RNA-seq library, isolated RNA samples were deprived of ribosomal RNA with Illumina Ribo-Zero Plus rRNA Depletion Kit (Illumina, San Diego, CA) and processed into a cDNA library for Illumina with the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (New England Biolabs). Dual-index primers were conjugated with NEBNext Multiplex Oligos for Illumina (Set1, New England Biolabs). We pooled all libraries, and the multiplexed libraries were dispatched to Macrogen Japan Inc. and subjected to paired-end sequencing with HiSeqX. Adapter trimming and quality filtering of raw sequence reads were conducted with fastp (ver. 0.21.0) (40) under default conditions. The paired-end sequences were mapped onto the Synechocystis genome (ASM972v1) using Bowtie2 (41) (ver. 2.4.5 paired-end). Supplementary Table S5 contains the read counts that passed via fastp quality control and were mapped by Bowtie2.

RNA-seq analysis

Mapped reads were counted by HT-seq count (ver. 2.0.2) (42) for the GFF file of ASM972v1, with the reverse-strandedness option. EdgeR package (ver. 3.40.1) (43) was used to perform the differential expression analysis. Fold changes in expression and FDR were used for gene classification. Supplemental Table S6-S8 contains fold change in gene expression calculated by edgeR.

Genome-wide analyses

Peaks were called using the MACS3 program (ver. 3.0.0b1) (44). For paired-end reads for SigE, SigA, and untagged control ChIP, narrow peaks were called with <1e−20 of the q-value cut-off and “--call- summits” options. The peak summits from two replicates and the untagged control were merged if summits were located within 40 bp of each other. Peak summits identified in both replicates but not in the control, were considered for further analysis. The midpoint of the peak summits for the two merged replicates was further analyzed.

Broad peak calling methods were applied to paired-end reads for cyAbrB2, cyAbrB1, and untagged control ChIP using the “–broad” option, with a q-value cut-off of < 0.05 and a q-value broad cut-off of < 0.05. The intersection of broad peaks from two replicates, excluding those called by the control, was used in subsequent analyses.

The positions of the TSS, including internal start sites, were obtained as reported by (45). The read count, merging, and the intersection of the binding region were calculated using BEDTools (ver. 2.30.0) (46). Supplemental Tables S9–S15 contain SigA and SigE peaks and the broad binding regions of cyAbrB2 and cyAbrB1, respectively.

Binding signals in every 100 bp bin for scatter plots were calculated as (IP read counts within 100 bp window) / (input read counts within 100 bp window) * (total input read counts/total IP read counts). GC contents were calculated within 500 bp in 100 bp sliding windows by seqkit (ver. 2.3.0) (47).

Genome extraction, digestion, and ligation for 3C assay

A 3C assay was conducted based on the previous prokaryotic Hi-C experiment (48, 49), with certain steps modified. To begin, Synechocystis were fixed with 2.5% formaldehyde for 15 minutes at room temperature. Fixation was terminated by adding a final concentration of 0.5M of glycine, and cells were stored at -80°C until use. Fixed cells were disrupted using glass beads and shake master NEO (Bio Medical Science, Tokyo, Japan), following the previous study’s instructions for preparing cell lysate for ChIP. The lysates were incubated with buffer containing 1mM Tris-HCl (pH7.5), 0.1mM EDTA, and 0.5% SDS for 10min at room temperature, and 1% Triton X-100 quenched SDS. Genomes in the cell lysate were digested by 600U/mL of HindIII (Takara Bio, Shiga, Japan) for 4 hours at 37°C, and RNA in the lysate was simultaneously removed by 50 µg/mL of RNaseA (Nippon genetics, Tokyo, Japan). The digestion was terminated by adding 1% SDS and 22 mM EDTA. The fill-in reaction and biotin labeling steps were omitted from the procedure. The digested genomes were diluted by ligation buffer containing 1% triton-X100 to the final concentration of approximately 1µg/mL and incubated for 10 min at room temperature. Ligation was performed with 2 U/mL of T4 DNA ligase (Nippon Gene) overnight at 16°C. Crosslinking was reversed under 65°C for 4 hours in the presence of 2.5 mg/mL of proteinase K (Kanto Chemical, Tokyo, Japan), and DNA was purified with the phenol-chloroform method and ethanol precipitation method.

Preparation of calibration samples for 3C qPCR

Based on a previous study, calibration samples for possible ligated pairs were prepared in parallel with 3C ligation (50). In brief, the purified genome of Synechocystis was digested by HindIII, and DNA was purified with the phenol-chloroform and ethanol precipitation. Purified DNA was dissolved into the ligation buffer at a concentration of about 600ng/µL and ligated with 2 U/mL of T4 DNA ligase at 16°C overnight. As a result, DNA ∼600-fold condensed DNA than 3C samples were ligated.

Quantification of crosslinking frequency for 3C assay

Before the real-time PCR assay, we confirmed that each primer set amplified single bands in a ligation- dependent manner by GoTaq Hot Start Green Master Mix (Promega, Madison, WI) (Figure S6). Real-time PCR was performed with StepOnePlus (Applied Biosystems, Foster City, CA) and Fast SYBR Green Master Mix (Thermo Fisher Scientific) according to the manufacturer’s instructions. Interaction frequency was calculated by ΔΔCt method using dilution series of calibration samples described above. We confirmed each primer set amplified DNA fragment with a unique Tm value. The amount of the bait fragment containing hox operon were quantified and used as an internal control. Supplemental Table S4 contains the list of primers used in the 3C quantification. Interaction frequency for each primer position was calculated as the relative abundance of ligated fragments against the calibration samples and normalized among samples by internal control.

Statistical analysis

Statistical analyses were performed with R version 4.2.2 (51). The “fisher.test” function was used for Fisher’s exact test, and p-values < 0.05 were denoted as asterisks in the figure. Multiple comparisons of Fisher’s exact test were conducted using “fisher. Multcomp” function in the RVAideMemoire package (52), where p-values were adjusted by the “fdr” method and FDRs <0.05 are shown in the figures. Multiple comparisons of the Wilcoxon-rank test were conducted by “pairwise.wilcox.test,” and p-values were adjusted by the “fdr” method. Adjusted p-values < 0.05 are shown in the figure. The correlation coefficient was calculated with the “cor” function. Gene set enrichment analysis (GSEA) was performed by culsterPlofiler package (53) in R with p-value cut-off of 0.05. The enriched pathways detected by GSEA are listed in Supplemental Table S16.

Manuscript writing

We employed GPT-3.5 to refine the expression of the manuscripts.

Accession Numbers

Raw ChIP sequencing and RNA sequencing reads were deposited in the Sequence Read Archive (accession ID: PRJNA956842).

Declaration of interest

The authors declare no competing interests.

Acknowledgements

This study was supported by the following grants to T.O.: Grant-in-Aid for Scientific Research (B) (grant no. 20H02905), JST-ALCA of the Japan Science and Technology Agency (grant number JPMJAL1306), and the Asahi Glass Foundation. We thank Dr. Kohki Yoshimoto for providing laboratory instruments and Ms. Kaori Iwazumi for the support of bacterial culture and the medium preparation.

Author contributions

R.K. and T.O. designed the study; R.K. conducted the experiment; R.K. analyzed the data; R.K. and T.O. wrote the paper.