Inactivation of group 2 σ factors upregulates production of transcription and translation machineries in the cyanobacterium Synechocystis sp. PCC 6803

We show that the formation of the RNAP holoenzyme with the primary σ factor SigA increases in the ΔsigBCDE strain of the cyanobacterium Synechocystis sp. PCC 6803 lacking all group 2 σ factors. The high RNAP-SigA holoenzyme content directly induces transcription of a particular set of housekeeping genes, including ones encoding transcription and translation machineries. In accordance with upregulated transcripts, ΔsigBCDE contain more RNAPs and ribosomal subunits than the control strain. Extra RNAPs are fully active, and the RNA content of ΔsigBCDE cells is almost tripled compared to that in the control strain. Although ΔsigBCDE cells produce extra rRNAs and ribosomal proteins, functional extra ribosomes are not formed, and translation activity and protein content remained similar in ΔsigBCDE as in the control strain. The arrangement of the RNA polymerase core genes together with the ribosomal protein genes might play a role in the co-regulation of transcription and translation machineries. Sequence logos were constructed to compare promoters of those housekeeping genes that directly react to the RNAP-SigA holoenzyme content and those ones that do not. Cyanobacterial strains with engineered transcription and translation machineries might provide solutions for construction of highly efficient production platforms for biotechnical applications in the future.

For transcription initiation, the RNAP core recruits one of the several σ factors to form a transcription initiation competent RNAP holoenzyme. All σ factors compete for the same RNAP core 8 . This competition is affected by numerous factors including the amount of each σ factor, affinity of different σ factors to the RNAP core, amounts and activities of anti-σ factors that prevent recruitment of a particular σ factor, and RNAP modifying factors like the small signaling molecule ppGpp that directly binds to the RNAP core and changes the recruitment efficiency of different σ factors and affect the promoter selectivity of the RNAP holoenzyme [8][9][10] . These processes are not yet well understood in cyanobacteria.
Cyanobacteria typically encode multiple group 2 σ factors that closely resemble the essential primary σ factor but are non-essential in standard growth conditions [11][12][13] . The primary σ factor is assumed to be mainly responsible for transcription of housekeeping genes during growth. In a particular stress condition, the recruitment of one or more of the group 2 σ factors increases. Accordingly cells lacking group 2 σ factors show acclimation defects in high salt 11,14,15 , bright light 12,16,17 or heat stresses [18][19][20] . In addition to stress responses, the group 2 σ factors control metabolic changes during dark-light transitions 17,[21][22][23][24] .
We have constructed a Synechocystis sp. PCC 6803 strain without any functional group 2 σ factors 13 . The ΔsigBCDE strain is vulnerable to all tested stress conditions but grows well in standard conditions 13 . However, even in the standard conditions almost 20% of genes are at least two-fold up or down regulated in ΔsigBCDE compared to the control strain 13 . Interestingly, many genes encoding subunits of the transcriptional or translational machineries are upregulated in the ΔsigBCDE mutant. In the present paper, we further investigated the regulation of transcriptional and translational machineries in cyanobacteria. The results indicate that in the absence of competition between SigA and group 2 σ factors, more RNAP-SigA holoenzyme is formed than in the control strain. Numerous RNAP-SigA holoenzymes enhance the transcription of a particular set of housekeeping genes including those encoding transcription and translation machineries, and ΔsigBCDE produces more RNAPs and ribosomal subunits than the control strain. The extra RNAPs are fully functional, and enhanced transcription leads to a high RNA content per cell whereas extra ribosomal subunits do not form translationally active ribosomes, and ΔsigBCDE cells show similar translation activity and protein content as the control strain. Co-regulation of transcription and translation machineries as well as promoter differences between subgroups of housekeeping genes will be discussed.

Results
Transcripts for transcription and translation machineries are abundant in the ΔsigBCDE strain. We have previously shown that the ΔsigBCDE strain that lacks all functional group 2 σ factors grows well in standard growth conditions 13 . DNA microarray analysis of the transcriptome of ΔsigBCDE in the standard growth conditions revealed that many genes encoding proteins for transcription and translation machineries were up-regulated in the ΔsigBCDE strain 13 (Supplementary Table S1). These comprise genes for RNAP core subunits, tRNA synthases, ribosomal protein subunits and translation initiation, elongation and termination factors. Unlike the RNAP core genes, the sigA gene, encoding the primary σ factor in Synechocystis, was down-regulated (Supplementary Table S1). The opposite behavior of genes encoding RNAP core and sigA promoted us to study the amount and activity of RNAP complexes in the mutant strain.

RNAP content and transcription activity are upregulated in ΔsigBCDE.
To measure the amount of the RNAP in ΔsigBCDE, total proteins were isolated from cells grown in standard conditions and the amounts of RNAP subunits were detected by western blotting. The RNAP core proteins α, β, β′, γ and ω were approximately 2.5-fold more abundant in ΔsigBCDE than in the control strain (CS) ( Fig. 1A; original Western blots are shown in Supplementary Fig. S1). Similar amount of the ATP synthase β subunit in both strains confirmed equal loading of the samples in western blots. Thus increased transcripts of RNAP core genes were indeed used to produce more RNAPs. In addition, the SigA protein was 1.4-fold more abundant in ΔsigBCDE than in CS (Fig. 1A) although sigA transcripts were 6.1-fold down-regulated in ΔsigBCDE (Supplementary Table S1) indicating either post-transcriptional regulation of the sigA gene or higher stability the SigA protein in the mutant than in CS.
To analyze RNAP complexes, isolated protein samples were treated with ServaG blue and separated by blue native gel electrophoresis. ServaG blue binds to protein complexes and proteins, giving a negative charge without disturbing the subunit composition of the complex. After electrophoresis, proteins were transferred to a membrane and then the SigA factor and the α subunit were detected by western blotting. The antibody against the α subunit detected a wide area and a faint band (Fig. 1B). The upper part of the wide area contains the RNAP holoenzyme complex with the σ factor, as this part is also recognized by the SigA antibody (Fig. 1C). The lower part of the wide area contains the RNAP core complex without the σ factor (Fig. 1B,C). The faint band contains unassembled α subunits (Fig. 1C). The RNAP-SigA holoenzyme was more abundant in ΔsigBCDE than in the control strain while the amount of free SigA protein did not differ between the strains (Fig. 1C). Also the RNAP core complexes were more abundant in ΔsigBCDE than in CS (Fig. 1B).
The RNAP holoenzyme with the primary σ factor is assumed to transcribe housekeeping genes during active growth. Upregulation of RNAP-SigA holoenzyme in ΔsigBCDE might thus enhance transcription. To figure out if this is the case, the RNA content of cells of ΔsigBCDE and CS was analyzed. Cell counting with a flow cytometer confirmed that the cell number per OD 730 was similar in both strains (Fig. 1D). The total amount of RNA isolated from 1 ml of cell culture with optical density at 730 nm of 1.0 was approximately three-fold as high in ΔsigBCDE as in CS (Fig. 1E) showing that the extra RNAPs of ΔsigBCDE are active. The vast majority of all cellular RNAs are ribosomal RNAs, and analysis of RNA by agarose gel electrophoresis revealed a high rRNA content in ΔsigBCDE. Ribosomal RNA genes form two identical operons in Synechocystis 25 and the final rRNAs are maturated in a complex process including cleavages of the sugar-phosphate backbone and modification of the bases. Only mature rRNAs were detected (Fig. 1F) indicating that not only transcription of rRNA operons was upregulated in the mutant strains but these extra RNAs were efficiently processed to mature rRNAs. Upregulation of ribosomal subunits does not enhance translation in ΔsigBCDE. The synthesis rate of ribosomal proteins in E. coli depends on the rRNA content of the cells. We found that also in ΔsigBCDE the high rRNA content was accompanied with a high amount of the Rps1 (subunit of 30S ribosomal particle) and the Rpl1 (subunit of 50S ribosomal particle) proteins ( Fig. 2A: original Western blots are shown in Supplementary  Fig. S1). However, labelling newly synthesized protein with radioactive 35 S-methionine revealed that translation activity was similar in both strains (Fig. 2B) as was also the protein content of cells (Fig. 2C). Thus, an increased amount of ribosomal subunits did not accelerate translation in ΔsigBCDE.
To find reason(s) for the discrepancy between the ribosomal subunit content and translation activity we further analyzed the ribosomes of ΔsigBCDE. To that end, a soluble protein fraction containing ribosomes was isolated and separated by sucrose density gradient centrifugation. To keep 70S ribosomes intact, buffers were supplemented with 10 mM Mg 2+ . After the centrifugation, nineteen fractions were collected, and RNAs and the Rps1 and Rpl1 proteins were analyzed from each fraction. In CS, all ribosomal subunits (16S and 23S rRNAs and the Rps1 and Rpl1 proteins) were only detected in Fraction 19 (Fig. 3A,B) whereas in ΔsigBCDE all these ribosomal subunits were detected in Fractions 18 and 19 (Fig. 3C,D). Although the overall number of translating ribosomes  was similar in both strains, the results indicate that polysomes (Fraction 19) are more common in CS than in ΔsigBCDE while in ΔsigBCDE monosomes (Fraction 18) are also typical. As the amount of translating ribosomes was similar in both strains but many housekeeping genes produce more copies of mRNA in ΔsigBCDE than in CS, the number of ribosomes reading each mRNA molecule must be lower in ΔsigBCDE than in CS. Fractions 15-17 contained 16S rRNA and the Rps1 protein in both strains indicating that these fractions represent 30S ribosomal particles (Fig. 3). In addition, the Rps1 protein was detected in Fractions 1-11 in ΔsigBCDE but not in CS. As Fractions 1-11 did not contain full length 16S rRNA, they do not represent functional ribosomal particles. In the lightest fractions, Rps1 might be a free protein while in the other fractions it might form complexes with some other ribosomal proteins and partially degraded 16S rRNA, seen as shorter RNA species. Similarly, the Rpl1 proteins and shorter RNA fragments in Fractions 10-17 detected in the ΔsigBCDE strain might represent functionally incompetent 50S ribosomal particles (Fig. 3C,D).
During the assembly of 30S and 50S ribosomal particles, protein subunits are attached to the 16S and 23S rRNAs, respectively. Light complexes containing partially degraded rRNAs and part of ribosomal protein subunits are common in ΔsigBCDE but not in CS (Fig. 3). As all ribosomal proteins were not upregulated at the transcriptional level in ΔsigBCDE (Supplementary Table S1), it is possible that the assembly of extra ribosomal particles is never fully completed before the partially assembled ribosomes are degraded in ΔsigBCDE.

Arrangement of RNAP core genes in Synechocystis sp. PCC 6803 genome.
Our results indicate that the production of the main components of transcriptional and translational machineries were upregulated ΔsigBCDE. Interestingly, the gene encoding the α subunit of the RNAP core is surrounded by numerous genes encoding ribosomal proteins and thus arrangement of these genes into an operon could provide a basis for co-regulation of transcriptional and translational machineries. To detect if rpoA forms an operon with surrounding genes, DNA-free RNA was isolated and cDNA synthesis was performed using random hexanucleotide primers. Specific primer pairs recognizing two adjacent genes were then used in subsequent PCR reactions. Successful amplification of the DNA fragment in a PCR reaction indicates that the adjacent genes are in the same cDNA molecule thus belonging to the same operon.
The rpoA gene was found to belong to a large gene complex where 30 genes showed positive signals for an operon test ( Fig. 4A; original agarose gels are shown in Supplementary Fig. S2). In addition to the rpoA gene, this region comprises 8 genes for 30S ribosomal proteins, 17 genes for 50S ribosomal proteins, the infA gene encoding the translation initiation factor IF-1, the secY gene encoding a subunit of the protein translocase, the adk gene encoding adenylate kinase and the truA gene encoding the tRNA pseudouridine synthase (Fig. 4A). However, when expression levels of these 30 genes were compared, two genes, adk and infA, in the middle of the region, showed lower expression than the other genes (Supplementary Table S2). In an earlier transcriptome analysis, this region was divided to four transcription units rpl3-rpl15 (TU837), secY-adK (TU836), infA (Tu835), rpl36-rpl31 (TU833) 26 . Our RNA secondary structure analysis revealed possible transcription termination loops downstream of the rpl15 and secY genes. Because combining all these results turns out to be complicated, we next performed a Northern blot analysis.
Cells were treated with a transcription initiation inhibitor, rifampicin for 0, 5 or 15 min and transcripts were detected using the rpoA probe. The longest transcripts were circa 16 000 nt and could thus comprise genes from rpl3 to rpl31 (Fig. 4B). However, these long transcripts were not abundant. Instead, an approximately 4700 nt long transcript (this size would fit to the predicted TU833 from rpl36 to rpl31) were abundant together with 2700 nt, 2200 nt and 1500 nt long transcripts (Fig. 4B). The longest transcripts became more abundant when transcription initiation was prevented with rifampicin. Thus, actively transcribing cells might mainly produce the numerous short transcripts form rpl3-rpl31 while the full length rpl3-rpl31 transcripts are mainly produced only if transcription initiation inside the operon is not active. The longest transcript was more stable than the shorter ones (Fig. 4B). Our results indicate a complicated transcription pattern for the rpoA region including many internal transcription initiation and termination sites, possible regulation via antitermination process resulting in partially overlapping transcripts covering the whole region from rpl3 to rpl31.
The rpoC2 (the β′ subunit) and rpoB (the β subunit) genes form an operon together with the tatD gene (subunit of the twin-arginine protein export system) and the rps20 gene (a subunit of 30S ribosome), while the rpoC1 gene (the γ subunit) produces a monocistronic transcript (Fig. 4C,D). The rpoZ gene forms an operon with the sll1532 gene with an unknown function (Fig. 4E). Transcription of both the slr1634 gene and the rpoZ-sll1532 operon continues long downstream the actual coding regions so that a great part of this area is transcribed to both directions. The originally sequenced glucose tolerant Kazusa strain of Synechocystis sp. PCC 6803 contains a probable transposase gene slr1635 between the rpoZ and slr1636 genes, but slr1635 is missing from our CS and from many other sequenced Synechocystis strains 6 .
Promoter regions of differently regulated housekeeping genes. Our results revealed that housekeeping genes in Synechocystis can be divided to two categories, to those that are upregulated when RNAP-SigA content increases and to those that do not directly respond to RNAP-SigA content. We analyzed in silico the promoter regions of those operons/genes that were upregulated in ΔsigBCDE and those that were not affected or ΔsigBCDE (C,D) strains were fractionated with a 5-35% sucrose density gradient centrifugation, nineteen fractions were collected and either RNA or ribosomal proteins Rps1 and Rpl1 were detected from each fraction. RNA was isolated from the 300 μL sample of fraction, dissolved in 20 μL of water, and 15 μL of isolated RNA from each fraction of CS (A) or ΔsigBCDE (C) was loaded on a 1.2% agarose gel and stained with ethidium bromide. The DNA molecular marker was included to demonstrate separation of fragments. For western blots, a 24 μL sample of a fraction was solubilized and proteins were separated on SDS-PAGE and Rps1 and Rpl1 proteins were detected by specific antibodies by western blotting in the control (B) and ΔsigBCDE (D) strains.
were even down-regulated in ΔsigBCDE. Only genes/operons with experimentally verified transcription initiation sites were chosen 26,27 . The expression levels of the selected genes in ΔsigBCDE, in comparison to CS, are indicated in Supplementary Table S3. An extended −10 region characterized by TGN just upstream of the −10 region was detected in many genes upregulated in ΔsigBCDE but not in those ones showing similar or lower expression in ΔsigBCDE than in CS (Fig. 5) suggesting that an extended −10 region type promoter can be considered as a consensus promoter for genes that respond directly to an increased amount of the RNAP-SigA holoenzyme. Genes that were not up-regulated in ΔsigBCDE showed high variability in the −35 region and shared only the conserved −10 region with classical E. coli promoters (Fig. 5). To test which genes are transcribed in the same unit, DNA-free RNA was isolated from Synechocystis and used as a template for cDNA synthesis using random hexamer primers and reverse transcriptase. Specific primers were thereafter used to amplify the indicated regions in the genome by PCR, and the PCR products were separated with 0.8% agarose gels and stained with ethidium bromide. The left lane is always a reaction with reverse transcriptase and the right lane is a control reaction without the reverse transcriptase. (A) The rpoA operon. (B) The Northern blot analysis using the rpoA probe. The cells were treated with transcription initiation inhibitor, rifampicin, for 0, 5 or 15 min at the standard growth conditions, the RNAs were isolated with hot phenol method and separated with glyoxal-phosphate buffer system using 1.2% agarose gel and transferred to the membrane. The rpoA gene was amplified from the genomic DNA of Synechocystis with PCR, label with

Discussion
Regulation of the primary σ factor. The regulation of the formation of the RNAP-SigA holoenzyme is a complex process in Synechocystis. Transcriptional down-regulation of sigA in Synechocystis has been shown in many stress conditions like high temperature, high salt 21 or nitrogen deficiency 28 . Furthermore, the sigA transcripts diminish in the night and reappear again after the onset of illumination 21 . The group 2 σ factor SigB is required for rapid activation of sigA transcription after onset of light 21 suggesting that at least the RNAP-SigB holoenzyme, in addition to the RNAP-SigA holoenzyme, can recognize sigA prometer(s). In standard growth conditions the RNAP-SigB holoenzyme content is low 13 and thus the regulation of the sigA gene by the SigB factor is important only in special conditions like dark-light transitions. The sigA gene is located just upstream of the rrn operon 25 but not co-transcribed with the rrn operon 21 . Currently three different promoters have been identified for the sigA gene. The sigA gene has suggested to form a transcription unit with a non-coding RNA ncr1650 27 . A typical −10 region is found upstream of ncr1650 but the −35 region is unique (Supplementary Fig. S3). Imamura and co-workers 29 identified two other promoters closer to the initiation codon ( Supplementary Fig. S3). In addition to the transcriptional regulation, low content of the sigA mRNA in ΔsigBCDE but high abundance of the SigA protein indicates regulation at post-transcriptional level. The mechanism of this regulation remains to be solved, but one possibility is that extra SigA protein triggers down-regulation of sigA transcripts.
In ΔsigBCDE, a high amount of SigA leads to extra efficient formation of the RNAP-SigA holoenzyme. However, the recruitment of SigA by the RNAP core does not depend only on the abundance of SigA. The Synechocystis strain missing the non-essential ω subunit of the RNAP core recruits less SigA than CS although the SigA protein is available 6 . During nitrogen starvation, the RNAP-SigA holoenzyme content decreases more prominently than the SigA protein content and simultaneously the recruitment of group 2 σ factors increases 28 . In E. coli, unfavorable conditions induce production of the small signaling molecule ppGpp that binds to the RNAP core and renders the core less favorable to recruitment of σ 70 and more favorable to the group 2 σ factor σ S 9 . In addition, stationary phase induces production of the Rsd protein that functions as an anti-σ 70 factor in E. coli, thereby decreasing the formation of the RNAP-σ 70 holoenzyme 30 . Furthermore, binding of the RNAP-σ 70 holoenzyme to the small 6S RNA molecule prevents active transcription in E. coli until favorable conditions are resumed (for 6S RNA regulation, see review 31 ). All together, these regulatory mechanisms lower the transcription activity of housekeeping genes in E. coli a lot without large changes in the σ 70 content. Unlike in E. coli, the SigA content of cells drops drastically in the stationary phase in Synechocystis, whereas 6S RNA content increases only slightly 28 , suggesting that the dose of the primary σ factor might play a more direct role in controlling the formation of the RNAP holoenzyme in Synechocystis than in E. coli.

RNAP-SigA holoenzyme and growth rate.
In eubacteria, all σ factors compete for the same RNAP core to form a transcription initiation competent RNAP-holoenzyme, different σ factors having different affinity to the core 8 . In ΔsigBCDE, an elevated dose of SigA factor and a low amount of competitors result in formation of a high number of RNAP-SigA holoenzymes, over expression of many housekeeping genes and accumulation of extra RNAs. However, the growth rate of the ΔsigBCDE is not elevated 13 , indicating that overall transcriptional enhancement does not induce faster growth and on the other hand that accumulation of extra RNA is not harmful for Synechocystis cells. However, when the growth rate of Synechocystis changes due to changed environmental conditions, the growth rate and formation of RNAP-SigA holoenzymes often correlate. For example, in high CO 2 the formation of RNAP-SigA holoenzymes is abundant and cells grow fast 7 whereas nitrogen depletion causes cessation of growth and minimal formation of RNAP-SigA holoenzymes 28,32 . E. coli cells produce different amounts of RNAP depending on growth medium, and the RNAP level has been shown to be just above the level needed for maximal growth in the particular growth medium 33 . In E. coli, similarly as in Synechocystis, simultaneous overproduction of RNAP core and the primary σ 70 factor enhances transcription of many housekeeping genes but does not enhance growth 34 . Selection of promoters by RNAP-SigA holoenzyme. In the ΔsigBCDE strain, a high content of RNAP-SigA holoenzyme induces expression of particular housekeeping genes 13 . We selected a collection of upregulated and non-upregulated housekeeping genes in ΔsigBCDE and compared their promoters. Upregulated Synechocystis promoters often resembled E. coli promoters with an extended −10 region 35 . The promoters of non-upregulated genes only contain a typical −10 region while the rest of the promoter region varies from gene to gene. Thus promoters recognized by RNAP-SigA holoenzyme in Synechocystis contain a clear −10 region but otherwise those promoters show high variability, suggesting that housekeeping genes are regulated in a complicated manner that requires other factors in addition to the RNAP holoenzyme. Increased availability of RNAP-SigA holoenzymes directly enhanced the transcription efficiency of genes with extended −10 region promoters in Synechocystis, while cyanobacterial-specific promoters of the photosynthetic machinery did not respond to the overdose of RNAP-SigA holoenzyme. Thus, our results suggest that activation of photosynthesis related genes requires more specific regulators than just a high content of RNAP-SigA holoenzyme. Demand for mechanisms strictly regulating expression of photosynthetic genes is obvious, as imbalances in the function of photosynthetic reactions potentially induces production of harmful reactive oxygen species in cyanobacteria 36 . We have previously shown that in the ΔrpoZ strain, a low RNAP-SigA holoenzyme content lowers especially the transcription of genes encoding the carbon concentrating and carbon fixating machineries while expression of many other housekeeping genes remain at the normal level 6 . Patterns of gene expression in the ΔsigBCDE and ΔrpoZ strains show that different functional groups of housekeeping genes respond differently to the dose of the RNAP-SigA holoenzyme. This might be a common phenomenon in bacteria, as a recent study showed that the dose sensitivity of the RNAP-σ S holoenzyme correlates with the functional category of the σ S regulated genes in E. coli 37 . Regulation of translation machinery. The ΔsigBCDE strain produces more rRNAs than CS. In E. coli activity of rRNA operons correlates with growth rate 38 and the presence of thirteen highly active rRNA operons in Vibrio natriegens has been use to explain the rapid growth of this bacterium (doubling time of only 10 min) 39 . High rRNA content but normal growth of ΔsigBCDE shows that growth rate and rRNA content do not directly correlate in Synechocystis.
Two major regulatory signals, the concentration of initiator trinucleotides (ATP for six operons and GTP for one) and the amount of the starvation induced ppGpp molecule, set the expression level of the rRNA operons in E. coli 40 . The rrn operons of E. coli contain two promoters P1 and P2; P2 provides rather constant expression while P1 is highly induced or repressed depending on conditions 38 . The high transcription efficiency from the P1 promoters has been explained by their unique features that lead to formation of a scrunched open complex, thereby reducing abortive initiation 41 . The two identical rrn operons of Synechocystis have GTP as an initiator nucleotide and they contain only a single promoter resembling the σ 70 consensus promoter with neither an extended −10 region nor a discriminator sequence typical for the P1 promoter ( Supplementary Fig. S4), suggesting that cyanobacterial and E. coli rrn operons are regulated differently. Furthermore, a Blast search revealed that Synechocystis lacks close homologs of the HN-S, DksA or Fis proteins regulating rRNA operons in E. coli 38,40,42 . In Mycobacterium tuberculosis the CarD protein replaces the function of the DksA protein 43 but close homologs of this protein are missing in Synechocystis as well.
Synechocystis does contain Nus proteins that regulate Rho-dependent attenuation of the rrn operon in E. coli 38 , but the Rho factor is missing and thus an E. coli type attenuation mechanism cannot function in cyanobacteria. Nus proteins in Synechocystis might be involved in rRNA processing in the same way as was recently reported in E. coli 44 . At least rRNAs are efficiently processed in ΔsigBCDE, as only 16S, 23S and 5S rRNAs were detected and according to microarray data 13 transcripts of many Nus proteins and rRNA processing enzymes were up-regulated in ΔsigBCDE.
Ribosomal RNA content regulates the production of ribosomal proteins with a feedback mechanism in which ribosomal proteins bind strongly to rRNA and weakly to their own mRNA 45 . In accordance with that, the high rRNA content of ΔsigBCDE is accompanied with high ribosomal protein content in the mutant strain. Further studies are required to reveal the actual mechanism regulating ribosome production in Synechocystis.

Co-regulation of transcription and translation machineries.
Our results show that the production of both transcription and translation machineries are similarly up-regulated in the ΔsigBCDE strain. Interestingly, both the rpoA operon and the rpoB-rpoC2 operon contain ribosomal protein genes in addition to RNAP genes. Actually the long rpoA cluster in Synechocystis comprises three ribosomal protein operons of E. coli (Supplementary Fig. S5). Gene order near the rpoA gene seems to be conserved. Analysis of circa 5000 bacterial genomes reveals that in eubacteria, the arrangement of rps13, rps11, and rpl17 genes next to the rpoA gene is highly conserved, and in many cases also rps4 is part of the gene cluster (Supplemental Table S5). In Synechocystis, the rps4 gene forms a monocistronic transcript in a separate location, and the rpoA operon encodes numerous other ribosomal genes in addition to the rps13, rps11, rpl17 genes. High conservation of rps13-rps11-(rps4)-rpoA-rpl17 gene cluster suggests that the rpoA operon might play a central role for simultaneous production of transcriptional and translational machineries. The biogenesis of the RNAP core starts with the formation of the α dimer 46 but the actual roles of Rps13, Rps11 and Rpl17 proteins in the formation of functional ribosomes during the complicated assembly of ribosomes (for a review see 47 ) remains to be elucidated.

Conclusions
Our findings about cyanobacterial regulation of transcriptional and translational machineries are summarized in Fig. 6. Inactivation of group 2 σ factors increases the content of the principal σ factor SigA and the formation of the RNAP-SigA holoenzyme, which leads to an increase in the transcription of many housekeeping genes including genes for RNAP itself and the translational machinery. Extra transcripts of the RNAP subunits are used to produce high amounts of transcriptionally active RNAP, which enhances transcription and triples the RNA content of the cell. Ribosomal subunits are also produced in excess in ΔsigBCDE, but translation activity and the total protein content of the control and ΔsigBCDE strains are similar. The amount of functional ribosomes in ΔsigBCDE remains similar as in the control strain, pointing to post-translational regulation of the number of active ribosomes. A possibility to engineer a cyanobacterial strain with high transcription and translation capacity without a high growth rate might offer new solutions for the use of cyanobacteria for efficient production of valuable compounds in future.

Materials and Methods
Strains and growth conditions. The glucose-tolerant strain of Synechocystis sp. PCC 6803 (CS) and ΔsigBCDE were grown in BG-11 medium buffered with 20 mM Hepes, pH 7.5, in normal air at 32 °C under constant illumination at the photosynthetic photon flux density of 40 µmol m −2 s −1 as described earlier 13 . The BG-11 agar plates of ΔsigBCDE were supplemented with appropriate antibiotics but for the experiments, liquid cultures were grown without antibiotics.
Cell content of the culture. The cell content was detected by measuring OD 730 , dense cultures were diluted to OD 730 = 0.4 before measurements. To determine the cell number of the CS and ΔsigBCDE cultures, the OD 730 was set to 0.6 and cells were calculated with a flow cytometer BD LSRRortessa (BD Biosciences).

Analyses of RNA polymerase complexes.
To analyze the amounts of the RNAP subunits, 30 mL of culture (OD 730 was 1.0) grown in standard conditions was harvested at 4 °C by centrifugation at 4500 g for 6 min. Then total proteins were isolated as described earlier 12 and protein concentration was measured with Bio-Rad DC protein assay kit. Then 15 µg of proteins were solubilized and separated with 10% NEXT GEL ® SDS-PAGE (Amresco), transferred to Immobilon-P membrane (Millipore) with Trans-Blot ® (BioRad) according to the manufacturer's instructions. The RNAP core proteins and the primary σ factor SigA were detected by Western blotting as described 13 . Immunoblots were quantified using a FluorChem image analyzer (Alpha Innotech Corp). As a loading control, ATPase β subunit was detected with a specific antibody (Agrisera). Three independent biological replicates were analyzed.
The RNAP complexes were analyzed by blue native gel electrophoresis. Cells were grown and harvested as describe above. The cell pellet was washed with BN isolation buffer (50 mM TrisHCl pH 7.5, 150 mM NaCl, 10 mM NaEDTA pH 8.0, supplemented with Protease Inhibitor Cocktail (Roche)). Then 180 µL of BN isolation buffer and the pellet volume of acid-washed 150-212 µm glass beads (Sigma) were added, and cells were broken by vortexing 8 × 1 min at 4 °C. Cell debris and membranes were removed by two centrifugation steps, first at 10 000 g for 5 min and then 18 000 g for 15 min.
To separate protein complexes, 60 μg of soluble proteins were solubilized with BN solubilization buffer (Serva G Blue 5 mg mL −1 , 2% glycerol, 90 mM sucrose, 2.5 mM BisTris) for 5 min on ice, and separated using 5-12% gradient gels (acrylamide:bis-acrylamide 32:1) with 4% of stacking gel using Mini protean gel system (BioRad) at 4 °C. The anode buffer was 50 mM BisTris pH 7.0 and the cathode buffer 50 mM Tricine, 15 mM BisTris pH 7.0 supplemented with 0.01% Serva G; the voltage was increased after 30 min of run from 75 V to 100 V and again after 30 min to 150 V. Then after 30 min, the cathode buffer was changed to a similar buffer without Serva G and the run was continued at 200 V until the blue color ran out of the gel. The gel was soaked in a transfer buffer (48 mM Tris, 39 mM glycine, 0.0375% SDS and 20% methanol) for 30 min and thereafter protein transfer and immunodetection was performed as described above. Two independent biological replicates were analyzed.
RNA content of cells. Total RNAs was isolated from 1-ml cell culture (OD 730 was 1) with the hot phenol method as described earlier 48 , and any DNA contamination was removed with TURBO DNA-free TM kit (Ambion). The RNA content was measured with a spectrophotometer at 260 nm (Biodrop). To visualize RNA, a sample of isolated RNA from the same amount of the cells was separated on 1.2% agarose gel and stained with ethidium bromide. The DNA microarray data 13 is available in GEO accession GSE69981.
Operon analysis. DNA-free RNA was extracted from CS as described earlier 6 . Reverse transcription (RT) was performed for 900 ng RNA using SuperScript III kit (Invitrogen) according to the manufacturer's instructions. Then 2 µL of RT reaction was used in sequent PCR amplification with Phusion HeatShock II (Thermo Scientific). The PCR primers are listed in Supplemental Table S5. To detect PCR products, 5 µL of the PCR products were separated on 1% agarose gel and stained with ethidium bromide.
Northern blot analysis. The cell culture (30 mL; OD 730 was 1) was supplemented with rifampicin (500 μg mL −1 ) and cells were collected by centrifugation after 0, 5 or 15 min of incubation in the standard growth conditions. RNA was isolated with the hot-phenol method 48 . Samples containing 8 μg of total RNA were denatured with the glyoxal system, separated on 1.2% agarose gel in phosphate buffer and transferred to a Hybond-N membrane 48 . The rpoA gene was amplified from genomic DNA using specific primers (Supplementary Table S5) and labelled with α-32 P-dCTP 10 mCi mL −1 (Perkin Elmer) with the Prime-a-gen labelling system (Promega) according to the manufacturer's instructions. After prehybridization in 6xSSC, 1xDenhardt's, 0.1% SDS, 100 μg mL −1 Herring sperm DNAS for 1 h at 67 °C, denatured probe was added and hybridized overnight at 67 °C. Then membrane was washed once with 2xSCC, 0.1% SDS at 60 °C for 15 min and once with 1xSCC, 0.1% SDS at 60 °C for 15 min and autoradiographed.
Promoter analysis. Two groups of promoters of highly expressed genes were selected for sequence analysis.
The first group comprises genes up-regulated in ΔsigBCDE and the second group consists of genes that either show similar expression in ΔsigBCDE and CS or are down-regulated in ΔsigBCDE (Supplementary Table S3). Each of the selected genes contained a well-defined transcriptional start site, obtained from previous studies 26,27 and the sequence 1-60 nt upstream of each start site was retrieved. For modeling of the −10 region, the sequences were aligned, maximizing the information content in a 6-nt window ending 8 nt upstream of the transcription start site. In the optimization of the alignment, each sequence was allowed to move by 1 nt to the left and to the right. The −35 motif was modeled by aligning the sequences initially according to the −10 region, and then maximizing the information content of a 6-nt window located 15-17 nt upstream of the −10 region. The alignments were optimized with custom software used earlier in 6 . Ribosome analysis. Cells (100 mL, OD 730 was 1) were collected from standard conditions by centrifugation and ribosomes were isolated at 4 °C. The cell pellet was rapidly washed with 20 mM Na-EDTA, 20 mM Tris-HCl pH 8 and then with isolation buffer (20 mM NH 4 Cl, 10 mM MgCl 2 , 5 mM dithiothreitol, 20 mM Tris-HCl pH 8.0, and protease inhibitor cocktail (Roche)). The pellet was resuspended in 2 mL of isolation buffer and the cells were broken using a French press (Constant Systems Ltd) treating the samples twice at 2.7 kPa. Cell debris was removed by centrifugation at 10 000 g for 10 min, and the supernatant was centrifuged again at 30 000 g for 30 min. The supernatant was collected and protein concentration was determined. Two mg of proteins were loaded onto the top of 5% to 45% linear sucrose density gradient in isolation buffer and the samples were centrifuged in TH-641 rotor (Sorval) at 107000 g for 18 h at 4 °C. After centrifugation, 19 fractions, 620 μL of each, were collected. Three independent biological replicates were analyzed.
Ribosomal proteins were analyzed by western blotting. A 24-μL sample of each fraction was solubilized and proteins were separated on 12% SDS-PAGE gels with 4% stacking gel (acrylamide: bis-acrylamide 37.5:1) using Protean ® II (BioRad). Proteins were transfer to the membrane and immunodetected as described above. The Rps1 (AS08309) and Rpl1 (AS111738) antibodies were purchased from Agrisera.
Total RNAs were isolated from each fraction. An aliquot of 300 μL of a fraction was mixed with 200 μL of RNA isolation buffer (20 mM Na-acetate pH 4.5, 2% SDS, 250 mM Na-EDTA) and treated with one unit of DNaseI (Life Technologies) for 5 min and then extracted once with phenol:chlororoform (1:1) and once with chloroform. RNAs were precipitated at −20 °C overnight after addition of LiCl to a final concentration of 0.725 M and 2. volumes of ethanol. RNAs were collected by centrifugation at 18 000 g for 15 min and the RNA pellet was washed once with 70% ethanol and dissolved into 20 μL of water. Fifteen μL samples were separated on 1.2% agarose gels in TAE buffer and stained with ethidium bromide.
Detection of newly synthesized proteins. Cells were grown in standard growth conditions. Five μL of radioactive 35 S-methionine (35S L-Met 185 MBq, Perkin Elmer) was added to 20 mL of cell culture (OD 730 was 1) and cells were labelled for 10 or 30 min under the standard growth conditions. Cold L-Met was added to a final concentration 0.4 mg mL −1 , samples were rapidly cooled and cells were harvested by centrifugation at 4500 g for 6 min at 4 °C and total proteins were isolated as described 15 . Proteins (60 μg) were separated by 10% NEXT GEL SDS-PAGE, transferred onto membrane and autoradiographed. Two independent biological replicates were performed.