Comparative Transcriptomics between Synechococcus PCC 7942 and Synechocystis PCC 6803 Provide Insights into Mechanisms of Stress Acclimation

Synechococcus sp. PCC 7942 and Synechocystis sp. PCC 6803 are model cyanobacteria from which the metabolism and adaptive responses of other cyanobacteria are inferred. Using stranded and 5′ enriched libraries, we measured the gene expression response of cells transferred from reference conditions to stress conditions of decreased inorganic carbon, increased salinity, increased pH, and decreased illumination at 1-h and 24-h after transfer. We found that the specific responses of the two strains were by no means identical. Transcriptome profiles allowed us to improve the structural annotation of the genome i.e. identify possible missed genes (including anti-sense), alter gene coordinates and determine transcriptional units (operons). Finally, we predicted associations between proteins of unknown function and biochemical pathways by revealing proteins of known functions that are co-regulated with the unknowns. Future studies of these model organisms will benefit from the cataloging of their responses to environmentally relevant stresses, and improvements in their genome annotations found here.


Introduction
Cyanobacteria are considered to be among the oldest organisms evolutionarily, given that putative microfossils attributed to cyanobacteria are 3.5 billion years old [1]. Synechocystis sp. strain 6803 (referred hereafter as Synechocystis) and Synechococcus sp. PCC 7942 (referred hereafter as Synechococcus) are two of the most well-studied cyanobacterial model organisms. Like other cyanobacteria, these organisms are capable of oxygenic photosynthesis. Most recently, they have attracted a great deal of attention because of their potential for photobiological production of biofuels and carbon sequestration [2,3]. Because of their great importance, the complete sequence of their genomes were among the first photosynthetic bacteria to be elucidated [4,5]. Despite their similarities, both strains exhibit a number of differences in their response to environmental stresses. For example, even though Synechocystis and Synechococcus are both freshwater unicellular strains, Synechocystis can tolerate elevated concentrations of NaCl better than Synechococcus. Furthermore, Synechococcus is an obligate photoautotroph [6], while Synechocystis is a facultative photoautotroph [7].
Data obtained with RNA-seq have been shown to have higher dynamic range than microarray chips and are useful for improving genome structural annotations [8,9,10]. In this study we used RNA-seq to characterize the transcriptomes of Synechocystis sp. PCC 6803 and Synechococcus sp. PCC 7942 under various stress conditions in order to reveal their transcriptional responses to environmental stress and to improve their genome annotations. The specific stresses we studied were increased salinity, increased pH, mild carbon limitation, decreased temperature, and decreased light. Although increased pH is considered a ''stress'', prior research has shown a coincidence of high pH with cyanobacterial blooms, as they draw down CO 2 during photosynthesis and shift the aquatic bicarbonate buffering system toward consumption of H + [11]. We also had a broader interest in the relative uses of respiration and photosynthesis for energy acquisition across all types of stress, given that cyanobacteria can acquire energy from either source.
Our main goals in this study were to: a) improve the structural annotation of the two genomes by correcting already annotated genes and identifying missed genes (including anti-sense) and transcriptional units (operons); b) improve the functional annotation of genes and pathways using gene co-expression information; and c) identify alterations in the short and long term expression patterns of genes in energy production pathways under various stress conditions.
The complete sequence and annotation of both strains is available through the Integrated Microbial Genomes (IMG) [12,13]. The expression data are also made available through IMG, and together with the transcriptome analysis presented in this paper, they serve as a valuable resource for further studies of the transcriptional mechanisms of these organisms.

Cyanobacterial growth and conditions
Synechococcus and Synechocystis were grown and sampled under an initial condition (reference), then subjected to five stress conditions, which were sampled at 1-hour and 24-hours (Table 1). To establish the reference condition, cells were grown in standard BG-11 medium in sterile glass flasks on a table top orbital shaker with cotton caps to allow for proper CO 2 incorporation. Except where noted, all cultures were grown at 30uC, 100 rpm, light 100 mE6m-26sec 21 , and 3% CO 2 . To establish the reference condition a 50 ml BG-11 flask was inoculated from a stock cyanobacterial culture and grown to late exponential state (D 730 ,8). Cells from this culture were then grown in a new 100 ml BG-11 flask containing media adjusted to pH 7.5 using 20 mM HEPES-KOH up to OD 730 = 2.0 (early to mid exponential phase), at which time the reference condition was defined and sampled. The following stresses were then applied as described below.
(i) Increased NaCl concentration: 200 mM for Synechococcus and 550 mM for Synechocystis. Based on different threshold point of salt tolerance between the two strains determined by previous studies [13,14,15,16,17,18]

RNA isolation -Sequencing
Ribosomal RNA was depleted from total RNA with MicrobExpress (Ambion/Life technologies). Purified mRNA was then fragmented using RNA fragmentation reagents (Ambion). Fragmented RNA served as the template for 1st strand cDNA synthesis. Reverse transcription was performed using SuperScript II Reverse Transcription (Invitrogen), with random hexamer as primer. This was followed by second strand synthesis using dNTP mix where dTTP is replaced by dUTP. Double stranded cDNA fragments were then blunt-ended, A tailed and ligated with Illumina adaptors. Digestion of dUTP was then performed using AmpErase UNG (Applied Biosystems).
Single Strand RNA adapter (59-ACACUCUUUCCCUACAC-GACGCUCUUCCGAUCU-39) was unidirectionally ligated to the 59 end of the mRNA using T4 RNA Ligase I (New England BioLabs). The first strand of cDNA was synthesized using a fusion primer (59-GTGACTGGAGTTCAGACGTGTGCTCTTCC-GATCTNN-NNN*N-39) and SuperScript II reverse transcriptase (Life technologies). After cleanup with Agencourt AMpure SPRI beads, 18 cycles of PCR reaction was carried out with KAPA HiFi DNA Polymerase and the Illumina TruSeq PCR primers. A gel size selection procedure was used to collect a library of products ranging from 200 to 700 bp in size. Sequencing data was generated from an Illumina GAII Sequencer. 59 enriched libraries were constructed according to [20] and sequence data was generated from an Illumina MiSeq benchtop sequencer.

Data quality and mapping
The quality of the sequenced reads was first assessed using the Bioconductor shortRead package [21]. The majority of the reads (.95%) were of high quality with the median Phred quality above 38 in all samples. Reads were aligned to the reference genome sequence or the corresponding strain provided by IMG [12] ( Table S1). The Burrows-Wheeler Alignment tool (BWA) [22] was used for the mapping, allowing default parameters and mapping to multiple locations. Artemis [23] and Gbrowse [24] were used to visualize the mapping results.
After the rRNA depletion step 66% of Synechocystis and 71% of Synechococcus reads mapped to ribosomal RNA genes and were not considered in downstream analysis (Table S1).

Gene expression analysis
Transcript abundance was normalized using RPKM (Read Per Kilobase of transcript per million Mapped reads) (Table S2) [25]. In order to determine whether biases in the expression profile of each gene along its length were consistent across all conditions we calculated the Spearman correlation between each pair of conditions for each gene. For the majority (.90%) of the genes the correlation was consistently above 0.5, indicating either absence of sequencing biases or consistency across all conditions. The remaining 10% of the genes were poorly expressed (RPKM,2.5*10 29 ). Gene set enrichment analysis was performed using the COG annotation provided by IMG. Statistical evaluation was performed using the Fisher's exact test with adjustment of the p-values for multiple comparisons using the Benjamini & Hochberg correction with FDR = 0.05.
Hierarchical clustering was performed using R hclust using Pearson correlation as distance metric and complete as the agglomeration method.
Differential gene expression analysis was performed using the R package edgeR [26]. P-values were adjusted with the Benjamini & Hochberg correction with FDR = 0.05. After selecting genes with an adjusted p-value,0.01 we added genes belonging to the same pathway if they exhibited similar logR 2 values or belonged to the same TU (as defined previously).

Annotation Improvement Techniques
Transcripts that mapped to intergenic regions were considered to confirm an unannotated gene if their coverage, defined as the number of reads mapped to any given nucleotide of this gene, was consistently 5 or more in at least 3 conditions, and the overall size of the transcript was greater than 200 bp. The latter threshold was selected to account for the exclusion of shorter transcripts during sequencing library preparation.
One five prime enriched library was used to identify putative starts of the transcripts on Synechococcus. This library was prepared from pooling RNA from all available samples of this strain. In order to identify the accuracy of the 59 signals we first analyzed their intensity on a set of high quality predicted CDSs. This set was comprised of genes that had been assigned to COGs and that had at least 150 bp downstream of their 39 end in order to avoid interference from trailing coverage of transcripts or operon structures. This allowed us to correlate the intensity of the 59 peak to the coverage of the CDS. In order to avoid 59 RNA-seq bias we also excluded the first 150 bps immediately after the 59 peak in the computation of the correlation. A threshold correlation of 0.6 or better using the Passing & Bablok regression was used to find all 59 peaks with similar correlation to their downstream transcript coverage. All the 59 signals that passed the previous filter were assigned to a CDS at most 200 bps downstream of the signal.
Transcriptional Units (TU) or operons were defined as collections of tandem CDSs with intergenic distance less than 150 bps, common expression profiles under different conditions (i.e. Pearson correlation between all conditions .0.6) and, in the case of Synechococcus, absence of 59 peaks assigned to genes inside the TU.

Whole genome expression
The total percentage of the two genomes, including intergenic and gene coding regions, covered by more than two RNA-seq reads from any condition was 98.6% for Synechococcus and 98.7% for Synechocystis. This can be attributed to (a) the high gene density (coding sequences cover 89.2% and 86.7% of the entire genome of Synechococcus and Synechocystis respectively). We observed that 99% of the coding bases were found to be covered by at least two reads, (b) expression of 59 or 39 UTRs or intergenic regions between genes belonging to operons, (c) the presence of transcripts in intergenic regions that have not been annotated. In any case, caution should be taken in interpreting these results, since they may also represent sequencing artifacts [27]. As described below, we tested these assumptions by examining the structural annotation of existing genes, specifically their 59 start sites. We also examined the validity of novel transcripts.

Structural annotation of existing genes
The 59 enriched libraries of Synechococcus revealed that 792 of 2612 CDSs did not have a reliable 59 signal. Of these 792, 522 were part of a predicted transcriptional unit (TU) and 270 were not. Those 270 CDSs ( Figure 1c) were poorly expressed under all experimental conditions (Table S3). Interestingly, these genes have similarities to other genes in other genomes and many of them have hits to protein families (pfam, COG) and functional annotations. The 59 enriched libraries also revealed 171 CDSs whose 59 peak was downstream of the annotated start codon ( Figure 1d) ( Table 2).
Overall, a total of 375 TUs were identified in Synechococcus with two or more predicted CDSs encompassing a total of 897 CDS (Figure 1e). Similarly, Synechocystis is predicted to contain 311 TUs comprised of 754 genes.

Novel transcripts
Using the coverage profiles, we looked for novel transcripts (i.e. transcripts not annotated in the current publicly available genome annotation) according to the method described in the materials and methods section. Due to size limitations imposed by the library preparation method, we restricted our prediction to transcripts of 200 bp or longer. We were able to predict 256 new transcripts in the Synechocystis and 127 in Synechococcus located in the intergenic regions -based on the current annotation. Approximately half of these transcripts returned significant hits (blast evalue ,10 25 ) when compared to the nr protein database (File S1), suggesting that these genes were missed in the current annotation of these organisms, while the other half most likely represents novel transcripts of unknown product.
Additional novel transcripts were detected on the antisense strands, 930 in Synechocystis and 573 in Synechococcus. In the case of Synechococcus we were also able to further support the detection of antisense transcripts by identifying 59 peaks supported by at least 5 reads, located in the antisense strand. Previous studies have detected expressed transcripts in the opposite (antisense) strand of protein coding genes in Bacteria and describe their role as a posttranscriptional mechanism to adjust mRNA levels [19,20,21]. However since our study was not designed to detect small size transcripts, we refrained from further exploring this prediction.

Comparison of new features to previous studies
Our results were compared to other studies that report new features of cyanobacterial genomes [2,31]. The predictions that are demonstrated in this study are similar to those reported in [31] but are not widely comparable against [2].
Compared to [31] TU predictions for Synechococcus completely agree in 183 cases and exhibit difference of at least one gene in 142 of our predictions. 243 TUs were not found in our study and we predicted an additional 50. TUs that were not predicted in our study typically either have low expression of intergenic regions between genes, or significant 59 peaks were detected upstream of the CDSs, or the genes show different expression profiles in the studied conditions. The ncRNA genes predicted in [31] were compared against our combined novel transcript predictions resulting in identifying 301 common genes (out of the 700 of our predictions) with at least a 40% overlap in their sequence. Similar comparisons were accomplished between our and [2] predictions for Synechocystis, with an overlap of 324 out of 1185.
Comparison to [2] was limited to predicted transcripts with length greater than 200 bp, a limitation presented by the library preparation method used in our study. In that case the 46 out of 82 long asRNA and ncRNA in [2] are also present in our dataset.
Transcripts predicted by these studies but not in ours were manually inspected to identify the cause of the discrepancy. In all cases inspected, the reason for this discrepancy is due to either the Example of a CDS (Synpcc7942_0801) with evidence of structural misannotation since the 59 peak is located 60 bps after the annotated start site. e. Two annotated genes with contiguous expression and a peak at the 59 of the first gene suggesting that both CDSs are expressed together as one transcript forming an operon. doi:10.1371/journal.pone.0109738.g001 lack of significant 59 peak or expression levels below the thresholds set. These differences can be attributed to the different sequencing technologies and protocols that were applied and the amount of data collected. Further investigation of these candidate genes can help verify their role and understand the biology of the organisms.

Gene expression and stress condition clustering
Unsupervised hierarchical clustering reveals the similarities and differences in stress response for the two study organisms ( Figure 2). The strongest similarities with respect to time were carbon depletion at 1-h and 24-h in Synechococcus (Figure 2A) and reduced light at 1-h and 24-h in Synechocystis ( Figure 2B). These time similarities indicate that the respective responses for the two organisms manifest themselves within 1-h and do not change significantly for the next 24-h. In addition, the greater distance between the reference condition and carbon stress cluster in Synechococcus and the distance between the reference condition and reduced light cluster in Synechocystis suggests that the immediate carbon stress response in Synechococcus is stronger than the immediate reduced light response in Synechocystis. The same kind of analysis leads to the conclusion that the pH response does in Synechocystis occurs largely within an hour. Clearly, different stresses elicit similar time responses in the gene expression patterns of the two organisms.
Unsupervised clustering can be used to reveal gene relationships based on the assumption that genes with similar expression profiles are functionally linked [32]. Based on this approach we were able to identify 6 proteins of unknown function (annotated as hypothetical proteins) in the current annotation, in Synechocystis and 1 in Synechococcus, which exhibited high correlation (Pearson correlation .0.9) to proteins of known function (Table S4). The six proteins in Synechocystis are smaller proteins (,145 amino acids) related to energy and nitrate metabolism. In one case (ssl5045), transcriptomics confirmed the existence of the small gene corresponding to a protein of 60 amino acids, which is located in the intergenic space in the July 2012 draft update of the genome. The Synechococcus protein was co-regulated with annotated twitching motility proteins. It needs to be noted that these associations do not reveal the exact function of these proteins, but rather indicate a common regulation and/or a functional correlation among them, and require additional experiments to verify or reject these hypotheses.

Differential gene expression
Pairwise comparisons between the reference and stress conditions, as well as between the 1-h and 24-h conditions revealed a relatively small number of genes that exhibit either increased or decreased transcription levels (an arbitrary threshold of logR 2 greater than 2 was selected). The genes and their classification to COG functional groups are shown in detail in 2.
In order to further associate the number of up/down-regulated genes to known pathways we performed gene set enrichment analysis using the COG functional classification. This analysis also revealed a different pattern of response between the two organisms.
Some responses to alkaline pH stress that we observed in both Synechococcus and Synechocystis are consistent with previous studies [11]. Alkaline pH in freshwater systems is thought to be caused by high levels of photosynthesis, which drive down dissolved CO 2 levels. Lowered CO 2 shifts the bicarbonate buffering system in the direction of consuming hydrogen, resulting in raised pH, as the hydrogen combines with bicarbonate to form CO 2 and water, thus maintaining equilibrium. Cyanobacteria, specifically Synechocystis, have been shown to acclimate in alkaline pH by spending energy to acquire and concentrate scarce CO 2 and bicarbonate, allowing them to grow and outcompete other photosynthesizers at high pH. Consistent with these studies, we found that Synechococcus gene sets related to cell wall biosynthesis and cell motility/signal transduction were enriched with upregulated genes under alkaline pH stress at 1-h and that gene sets for energy production and translation were upregulated at 24-h. We also found upregulation of ribosomal genes and translation in Synechocystis after 1-h. Certainly energy production, motility and translation are indicators of high metabolic activity, such as occurs during the exponential growth phase of blooms. We speculate that cell wall biosynthesis reflects either increased growth or physical restructuring of cell walls to tolerate alkaline conditions [11].
Other results of gene set enrichment analysis are itemized below.
N In Synechococcus elevated NaCl after 1-h causes upregulation of genes related to signal transduction and downregulation of nutrient transport (particularly nitrate and phosphate). After 24-h translation is upregulated, particularly the operon Synpcc7942_2203 -Synpcc7942_2235, and genes related to transcription, replication and signal transduction are downregulated. In Synechocystis, elevated NaCl after 1-h causes upregulation of translation and replication, downregulation of genes classified in cell wall biogenesis, and downregulation of translation after 24-h. These results, however, are not directly comparable since the two organisms exhibit different levels of tolerance for NaCl. N In Synechococcus CO 2 depletion after 1-h results in higher expression of energy production genes and intracellular trafficking genes, and lower expression of genes related to signal transduction. After 24-h energy conversion related genes remain enriched. In Synechocystis CO 2 depletion after 24-h results in lower expression of genes involved in energy production, basic metabolism and inorganic transport.
N In Synechococcus lower temperature after 1-h increases expression levels of genes related to chaperones and heat shock proteins, and decreases expression of signal transduction related genes. After 24-h the transcription levels of energy production and translation related genes are elevated while transcription and signal transduction are lowered. In Synechocystis lower temperature after 1-h upregulates genes related to replication, which remain elevated after 24-h. Translation appears to be elevated after 24-h as well. Interestingly transposases are also elevated both after 1-h and 24-h. N In Synechococcus lowered light after 1-h caused increased expression levels of genes related to energy production, and decreases the expression levels of genes related to nutrient transport (phosphate, nitrate, sulphate). After 24-h, lowered light increases expression of genes associated with post-translational modifications, and decreases signal transduction mechanisms. In Synechocystis lowered light after 1-h and 24-h decreases expression of genes for metabolic pathways and protein translation.
Overall, many stress conditions affected expression of protein translation genes. Expression of genes associated with protein translation has been shown to vary under stress conditions [33] although not to the degree reported here. Increased expression of ribosomal proteins and ATP synthase, has been associated with favorable growth conditions, including light-versus-dark transition and log-phase growth versus stationary-phase growth [11]. This could be a reason for their increased expression in several of the 24-h conditions, particularly in the case of Synechococcus. In the case of salt stress it has been suggested that high ionic strength might destabilize ribosomes, requiring replacement of ribosomal proteins to maintain ribosomal activity [34].

Photosynthetic and respiratory energy generation
Photosynthetic electron transport and respiratory electron transport are alternative biochemical pathways that ultimately create a proton gradient that energizes F 0 F 1 ATPase synthase. The thylakoid membranes of cyanobacteria are characterized by the simultaneous presence and intersection of the photosynthetic and respiratory electron transport chains, where several components are partially utilized in both processes [35][36][37]. Despite the spatial overlap of photosynthesis and respiration, the two electron chains offer alternative ways to adjust to environmental stresses. More precisely, this spatial overlap of the two electron transport chains in the thylakoids is more apparent in Synechocystis, while in Synechococcus the separation of the respiratory chain in the cytoplasmic membrane and the photosynthetic chain in the thylakoids is more defined [38]. Our data (Table S5) suggest that under certain conditions, photosynthetic and respiratory chain components were highly regulated ( Figure 3A).
Transcriptional responses for the two electron transport chains under all stress conditions could be grouped into three major categories: 1) downregulation of photosynthetic electron transport chain genes and upregulation of the respiratory chain genes, 2) upregulation of the respiratory transport chain genes solely, 3) concurrent downregulation of genes from both electron transport chains, and 4) downregulation of the photosynthetic electron transport chain genes solely (Figure 3). The model in Figure 3 presents graphically how we speculate that these transcriptional alterations may impact the actual energy flow of the two strains in our selected conditions. Further experimental verification is required to establish this model.

Upregulation of respiration and downregulation of photosynthetic related genes
Relative to reference conditions, stress conditions of low temperature, alkaline pH, and low light favored the upregulation of the respiratory-related genes in Synechococcus ( Figure 3B, Table S5). For the last two conditions, this is summarized in the upregulation of certain gene components of the NDH-2 complex, NDH-1 dehydrogenase complex, and the two available terminal oxidases, the aa 3type cytochrome c oxidase (CoxBAC) and the cytochrome bd quinol oxidase (CydAB). In contrast, the transcript levels of the photosynthetic antenna and the photosystem I (PSI) related genes were reduced. In both stress conditions these alterations were evident both after 1-h and 24-h of exposure. Despite the fact that we found a similar pattern for low temperature in Synechococcus cells, this applies only for the first hour of exposure in the corresponding stress conditions, while after 24-h only respiration remained enhanced (see Respiratory energy upregulation).
The only example in Synechocystis where upregulation of the respiration-related genes prevailed over the photosynthetic ones is low temperature (24-h). More precisely, certain components of the NDH-1 dehydrogenase complex and the NDH-2 complex were induced, while the PSI and the antenna reduced.

Upregulation of respiratory-related genes
Synechococcus cells after 24-h in low temperature exhibit an upregulation of certain components of the NDH-1 dehydrogenase complex and the NDH-2 dehydrogenase (see Table S5). This follows an enhanced respiratory and reduced photosynthetic  (24-h)). For simplicity, the two electron transport chains are represented in two different membrane sections, even though respiration is met in the cytoplasmic and both respiration and photosynthesis are intertwined in the thylakoid membrane. The intersection of the two pathways in the thylakoids is more apparent in Synechocystis than Synechococcus, where those two processes are more separated between the cytoplasmic and the thylakoid membrane. This representation only serves the portrayal of the transcriptional changes and their putative affect on energy flow and not the actual distribution of the two chains between the thylakoid and the cytoplasmic membrane. Abbreviations: Temperature (Tm). doi:10.1371/journal.pone.0109738.g003 transcriptional profile in the first hour of exposure. However, in carbon depletion we monitored elevated transcriptional levels of respiratory components only in the first hour of exposure, which is supported by an induction of the NDH-1 complex (but not the Nhd3 and nhF3 components), terminal oxidases, the cytochrome b6f complex and the F 0 F 1 -ATP synthase ( Figure 3C).

Simultaneous downregulation of photosynthetic and respiratory-related genes
Synechocystis cells 1-h after addition of 550 mM of NaCl exhibited a broad modification of expression levels for the majority of the genes that participate in the two electron transport chains. More precisely, 16 of the 32 components of the NDH-1 complex and all the three terminal oxidases (CydAB, CoxBAC and the quinol oxidase or QOX) were downregulated by 2-to 8-fold, while only 6 components were similarly upregulated. Interestingly, similar downregulation was also measured for almost all the components of the photosynthetic apparatus and certain components of the b6f cytochrome. Since both electron transport chains were downregulated, it is not surprising that the components of the F 0 F 1 ATP synthase were also downregulated. However, transcript levels returned to normal 24-h after exposure to elevated NaCl levels, evidently showing an acclimation of Synechocystis cells to rapidly respond to elevated salinity. Our data are in agreement with previous microarray and proteomic data that subjected Synechocystis cells to salt stress for short and long periods [24,28,29,30].
A similar, but less intense response was seen in Synechocystis for low light conditions. A 2-to 4.6-fold downregulation was seen for 43 to 47 components of the photosynthetic and respiratory electron transport chains for both time points, with only 2 genes upregulated (one of which degrades antenna protein). Synechocystis sp. PCC 6803 cells in low light seemed to be overall less affected compared to Synechococcus sp. PCC 7942, something that coincides with the photo-dependence differences these two strains exhibit, reported elsewhere [6,7]. The fact that simultaneous downregulation of genes of the photosynthetic and respiratory electron transport chains was exclusively seen in Synechocystis sp. PCC 6803 cells may imply that this strain is either more adapted to survive with an impaired energy production and/or that employs other catabolic pathways to compensate for the energy loss ( Figure 3D).

Downregulation of photosynthetic-related genes
Ignoring upregulation of ''respiratory'' genes ndhD3 and ndhF3 (which actually participate in inorganic carbon transport) and the petA and petC1 genes (which are part of the shared cytochrome b6f complex) there was a general downregulation of ''photosynthetic'' genes in Synechocystis after 24 hours of exposure to Carbon limitation and 1-h to low temperature ( Figure 3E), but never in Synechococcus. Photosynthesis provides the force for the essential carbon fixation pathway in photosynthetic organisms. The loss of the photosynthetic capacity as a result of the absence of inorganic carbon has been documented before in Synechocystis [31,32]. Apparently, Synechococcus presents a different response to our tested conditions from Synechocystis at the indicated time points.
Overall, it is worth mentioning that Synechocystis cells often responded to stress with strong downregulation (up to 20-fold decrease) of F 0 F 1 -ATP synthase. This argues either for a lower energy requirement of the corresponding strain under stress conditions, or alternative pathways of energy acquisition.

Carboxysome Operons
In Synechocystis, genes encoding CcmK1, CcmK2 and CcmL, which belong to the structural components that comprise the exterior of the capsule, are in the same positional cluster with the ccmM and ccmN genes that comprise the carboxysomal interior. Our data indicate that this cluster behaves as an operon since all components responded similarly to all tested conditions (see M&M regarding the definition of TUs). It is worth mentioning that the separate ccmK3/ccmK4 cluster and the ccmP gene, encoding an outer shell facet protein, followed the trend of the CcmK1/K2/L/ M/N cluster with the exception of 24-h in alkaline pH and 1-h of elevated salinity. Under these conditions, ccmP expression was unchanged while the other genes were all downregulated. The expression of the second outer shell facet CcmO component was mainly unaffected except for an upshift observed after 1-h in elevated salinity.
In Synechococcus the CcmK2/L/M/N cluster is adjacent to ccmO and two rbcL,S (RuBisCo) genes. However, the ccmO and the rbc genes seem to be transcripted independently from the CcmK2/L/M/N cluster. More precisely, the rbc genes exhibit a common pattern between them, placing them in their own operon, while ccmO is not coordinately regulated with either the CcmK2/ L/M/N cluster or the rbc operon. Therefore, ccmO must form its own operon (Table S5). These observations are also seen in our 59 data for Synechococcus.
Unlike in Synechocystis, we saw evidence in Synechococcus that it may alter carboxysome composition in response to different environmental conditions. After 1-h in low light, ccmK3 and ccmK4 in Synechococcus were downregulated while the CcmK2/ L/M/N cluster was upregulated. Also, in most tested conditions ccmK3 and ccmK4 transcript levels in Synechococcus remained stable and did not follow the observed changes of the CcmK2/L/ M/N cluster. Both of these Synechococcus responses in carboxysome structural protein gene expression were not seen in Synechocystis.

Carboxysome gene expression
Overall, in Synechocystis the majority of the carboxysomal components were significantly repressed under several conditions, namely low light (1-h and 24-h), salinity stress (1-h), pH (24-h), and carbon limitation (24-h). On the other hand, in Synechococcus the ccmK2/L/M/N carboxysomal component had a trend to upregulation under alkaline pH (24-h) and low light (1-h), and slightly in elevated salinity (24-h) and carbon limitation (1-h and 24-h). Also, in Synechococcus the RuBisCo components also exhibited downregulation trends, under low temperature (1-h), alkaline pH (1-h) and low light (24-h) conditions, and were unchanged under all other conditions. These data indicate differences between the two strains in terms of protein and/or organelle stability under the tested conditions. The NdhD3/F3 complex and the SbtA transporter were the most upregulated inorganic carbon transport systems for both strains, although the specific conditions causing upregulation varied. In Synechocystis upregulation occurred under low temperature (1-h), alkaline pH (1-h) and 24-h after carbon limitation. In Synechococcus upregulation occurred under low light (1-h), low temperature (24-h) and alkaline pH (1-h and 24-h) conditions. The fact that no induction of the major inorganic carbon transport systems was observed after carbon limitation in Synechococcus, could be due to the fact that either the carbon response is not yet fully activated in the tested time window (1-h and 24-h), or that they responded sometime between the 1-h to 24-h [39].

Conclusions
In this report we have employed an Illumina RNA-seq study for an in depth analysis of the transcriptome of the two cyanobacteria Synechococcus and Synechocystis. Using this data we were able to verify and improve the structural annotation of both genomes by detecting new transcripts located in the intergenic regions. Additionally gene annotations in Synechoccocus were improved based on 59 enriched libraries evidence, and transcriptional units (operons) were identified containing two genes on average. The identification of genes with similar expression profiles under the conditions examined allowed us to link genes with undetermined function to annotated ones and their corresponding pathways.
Overall, comparative analysis of the response of the two organisms revealed similarities but also significant changes in the way they acclimate to external stress at two specified time points representing a short (1-h) and long term acclimation (24-h). In Synechococcus the majority of the stress conditions resulted in the upregulation of the respiratory genes and the downregulation of the photosynthetic ones. This indicates that the loss of energy production that results from a sensitive photosynthetic electron transport chain is compensated by the induction of the respiratory one and suggests a more robust interaction between them in this organism. On the other hand, in Synechocystis there were cases that genes of both electron transport chains were downregulated. The carboxysome, a crucial organelle for carbon fixation, exhibits differences between the two strains. Synechococcus cells likely enhance the carboxysomal capsule in a number of environmental stresses, while Synechocystis does not.
We believe that this work will provide a valuable reference for further studies of cyanobacterial transcription and acclimation to environmental stress.   File S1 Sequence similarity results based on blast for newly identified transcribed sequences located in intergenic regions.

Supporting Information
(ZIP)