Two proteins of carbon metabolism pathways, previously identified as induced in 4% salt, were also induced here, fructose-bisphosphate aldolase (1.52 and 1.55-fold; with probabilities of being up-regulated under high salt conditions (p₊) of 0.940, 0.965) and phosphoglycerate kinase (up to 3-fold; p₊ = 1.00, 1.00). Dihydrolipoamide acetyltransferase which is part of the pyruvate dehydrogenase complex (post-glycolysis) was also >2-fold up-regulated in high salt (p₊ = 0.624, 1). These changes imply increased requirement for energy as carbon is metabolised, as has been reported formerly 182526. As expected, a large increase (4.86 and 6.21-fold) in a compatible solute synthesis enzyme, glucosylglycerol (GG) phosphate synthase was observed, an essential response in salt acclimation. An increase in 4-alpha-glucanotransferase (2.78 and 3-fold; p₊ = 0.00, 1.00) has also been seen previously in the long-term 18, and may play a role in making carbon available for GG synthesis. This may also explain the 1.62 and 1.66-fold increase in glucose-1-phosphate adenylyltransferase (p₊ = 0.994, 0.994). These latter two enzymes are positioned close together in pathways involving the re-arrangement of carbon atoms in glycogen synthesis pathways, and the accumulation of glycogen in salt acclimated cells has been reported previously in Synechocystis 16.

The cellular response to salt stress implies an increase in energy conversion, however, it is important to note that not all enzymes involved in energy synthesis are necessarily directly related to energy supply in salt stress. A good example of this was reported by Suzuki et al. 27, where the kinetic activities of enzymes in the glycolytic pathway were investigated in the mangrove tree in response to long-term 150 mM salt stress. Fructose-6-phosphate transferase (p_- = 0.998, 0.00) was thought to play an important role in the logarithmic growth stage where biosynthesis is rapid, however, it is not thought to be directly related to energy supply for the removal of salt from the cytoplasm 27. The role of the PPP regulatory control protein OxPPCycle gene appears reduced here (1.52 and 1.69-fold). In support of overall reduced PPP requirement, the reduced expression (2.56 and 2.23-fold) of an enzyme responsible for the hydrolysis of pyrophosphate (PPi), which is formed principally as the product of the many biosynthetic reactions that utilise ATP, suggests an overall reduction in biosynthesis, and linking in with the findings above, these are most likely biosynthesis reactions associated with slowed growth rate in 6% salt.

Fifteen proteins, involved in photosynthesis and respiration, were differentially regulated, 13 of which decreased in expression with an increase in salt. As expected, eight are pigment proteins and therefore 6% salt leads to pigment loss in cells, as does 4% salt 18. In addition, long-term acclimation does not lead to recovery of these pigments. Cytochrome c550 or PsbV decreased 1.93 and 2.13-fold and is associated with the oxygen evolving complex in PSII. Reductions in oxygen evolving machinery have been seen previously at the transcript level in both Synechocystis cells (4% salt at 24 hrs) and in in vitro isolated thylakoid membranes from Synechococcus sp. PCC7942 cells 1028.

Perhaps the most surprising of the proteins identified with reduced expression was the electron carrier flavodoxin. This protein has previously been associated as an important alternative electron carrier implemented in salt adapted cells of Synechocystis (2 to 4% salt) 1318. The marked decrease of this protein here implies this role is no longer applicable in higher salt (6%). Its non-essential role has been demonstrated however, when a flavodoxin mutant displayed reduced cyclic electron flow around PSI, but was still capable of acclimation to 2% salt 13.

In PSII, one subunit, CP43 increased 1.82 and 1.80-fold in expression. The increase in PSII reaction protein was unexpected, as previous studies have shown PSII activity to decrease in salt shocked and adapted cells 1813 (although this particular protein has not previously been identified). It is possible that an initial decrease in levels of this protein may function as damage limitation during initial salt stress, and an increase in expression functions to replace damaged reaction centres when cells have adapted. However, oxidative stress is known to actually inhibit repair of salt-damaged PSII protein D1 in cyanobacteria 29. These results imply variable stoichiometry in the complexes.

Vipp1 is a protein previously identified as salt induced in Synechocystis, and is believed to transfer reaction centre proteins to the thylakoid membranes during stress, and therefore replacement of damaged subunits occurs 30. Increased expression of replacement proteins has been observed in light stressed cells previously 31. Finally, apocytochrome f precursor, part of the cytochrome b₆/f complex, which is involved in electron transfer between both photosystems, increased 1.93 and 1.69-fold in 6% salt. This implies increased electron transfer to PSI (non-cyclic photophosphorylation) and therefore enhanced energy synthesis.

NADH dehydrogenase subunit I is part of the larger enzyme complex, and involved in oxidative phosphorylation in respiration. Despite its 1.59 and 1.75-fold decrease in expression in high salt, other NADH dehydrogenase subunits were identified and not differentially expressed, for example, beta subunit (1.01 and 0.71-fold change), suggesting a stoichiometric change in this complex. A DNA microarray study found transcript levels for subunits of this complex to increase immediately in response to 0.02 M (~1.2%) salt, but decrease in response to 1.0 M (6%) salt.

MMG analysis was used to build a systems picture of the metabolic response to salt stress. The proteomic data was mapped onto the KEGG metabolic network (accessed 13^th August 2008) using a Perl script and the probabilities were computed using the R module 'MMG' (http://cran.r-project.org/, R v2.8.1, MMG v1.4). The parameter σ was set to 0.2, by fitting the central part of distribution of the 115:116 log₁₀ ratios to a Gaussian distribution. The parameter α was set to 1, as advised in the original paper 32. MMG helps to interpret the proteomic data in light of the metabolic network's topology. This allows the identification of regulated enzymes, even when they have not been quantified or when they were quantified and did not exhibit a marked abundance change, which still could be biologically significant.

MMG's 1/λ statistics give insights into the degree of regulation: for down-regulated proteins, 1/λ = 0.73 (SD 0.10), while, for the up-regulated proteins, 1/λ = 0.61 (SD 0.09). This indicates that more down regulated processes can be detected by MMG.

The networks identified by MMG using the dataset resulting from Workflow 2 are presented in Figure 1 (up regulated in high salt conditions) and Figure 2 (down regulated in high salt conditions). They were obtained by selecting proteins with a probability of being up-regulated, p₊, greater than 0.45 (for the up-regulated network) and a probability of being down-regulated, p_-, greater than 0.45 (for the down-regulated network). The nodes in Figure 1 and 2, represent the single-unit and multiple-unit enzymes in Synechocystis' metabolic network, as per the KEGG database. Two enzymes are connected when a product of the reaction catalysed by one enzyme, is a reactant of the reaction catalysed by the other enzyme. Edges are weighted during the MMG analysis, and the weights are inversely proportional to the frequency of the compound within the metabolic network in order to give less prominence to currency metabolites such as ATP, ADP, NADPH, etc.

Among the proteins found to be down-regulated by MMG, the 6-phosphogluconolactonase (p_- = 0.5) was found. It complements the pathway made of quantified enzymes (namely, glucose-6-phosphate 1-dehydrogenase, 6-phosphogluconate dehydrogenase, and ribose-5-phosphate isomerase A/B) from the oxidative phase of the pentose phosphate pathway, whereby β-D-glucose-6-phosphate is turned into D-ribose-5-phosphate. The non-oxidative phase, on the other hand, of the pentose phosphate pathway is found to be up-regulated. The reasons for this discord in both stages in the PPP require further investigation, but the requirement of increased carbon fixation may be an important factor.

Among the proteins found to be up-regulated by MMG, two enzymes are identified that are related to glycogen synthesis and degradation: the glycogen synthase, the glycogen phosphorylase, and the glycogen branching enzyme. Since the glycogen synthase and the glycogen phosphorylase function in opposition, the combined effect of their up-regulation is unclear. However, the relationship between carbon requirement for glucosylglycerol synthesis and carbon excess being converted to a carbon store in the form of glycogen, is likely to be determined by adaptation time, where the need for compatible solute diminishes.

Due to the connection between high salt response and oxidative stress, this study prompts the question of the existence of producing and consuming pathways of NAD(P)H. Whereas superoxide dismutase is manifestly up-regulated in both Workflows (2.83 and 2.83 in Workflow 1, 2.44 and 2.36 in Workflow 2), further detoxification of H₂O₂ is not clearly up-regulated. Peroxidase is indeed found to be down-regulated (0.58 and 0.50 in Workflow 2), whereas the catalase is only mildly up-regulated (1.06 and 1.21 in Workflow 1, 1.45 and 1.39 in Workflow 2). Down-regulation of the oxidative phase of the pentose phosphate pathway could mean the depletion of reducing power, which could be compensated by photosynthesis. Figure 3 summarises the metabolic changes that occur in cells when adapted to 6% salt.

In this study, relative quantitations show that several of these proteins are required in elevated amounts in Synechocystis cells adapted (for exactly 9 days) to 6% salinity. A 2.24 and 1.90-fold induction was seen in heat shock protein 90 (Hsp90), and this particular stress protein has not been identified as salt induced in previous proteomics experiments in the short or long-term 19. Hsp90 contributes to an array of cellular processes, including protein degradation, protein folding and even signal transduction 1933. Its gene transcript levels have been seen to increase in response to ~3% salt 17 and 4% salt 10 in the short-term. In the latter study by Marin et al. 10, transcript levels were not induced after 5 days.

Anti-oxidative enzyme superoxide dismutase was up-regulated over 2-fold in all experiments (Additional file 3: Table S2), and this response has also been seen in the short-term and after 5 days at the protein level 18. A similar scenario was seen for molecular chaperone DnaK, although fold increases were less than for superoxide dismutase. Although the induction in DnaK in response to high salt (4% and 6%) has been seen previously at the protein level 1819, elevated transcript levels have not been observed in the long-term 10. It is not possible to extrapolate whether this gene is transcribed in the long-term, and therefore dnaK was a candidate for the RT-qPCR analysis as discussed below. The stress chaperones, 16.6 kDa small heat shock protein and GroES (co-chaperonin) were both expressed higher in 6% salt (1.75/2.10-fold and 1.59/1.83-fold, respectively), however, previous studies have identified transcript and protein levels to be salt responsive in the short-term only 1018. Not all stress proteins were expressed in higher amounts in 6% salt and heat shock protein, GrpE, was reduced by 1.66 and 1.72-fold.

In contrast to the results observed in Synechocystis cells adapted to 4% salt (over 6 to 8 days), four nutrient binding proteins experienced large reductions in expression in 6% salt. Relative levels of iron transport protein (1.75 and 1.99-fold), periplasmic phosphate binding protein (10.07 and 10.38-fold), periplasmic iron binding protein (9.28 and 11.08-fold) and a bicarbonate transporter (2.09 and 3.62-fold) were all significantly diminished. Moreover, bacterioferritin, which is responsible for storage of as much as 50% of cellular iron, was 1.61 and 1.72-fold reduced. It is not certain why cells reduce expression of transporter proteins in high salt (6%). Reduced growth and overall protein synthesis may mean fewer nutrients are required, although further work needs to be done to elucidate the specific reasons for this unexpected response.

Three transport proteins did increase in expression in 6% salt. An ABC-1 like transporter increased 4.27 and 4.93-fold, and is considered to be a putative ubiquinone biosynthesis protein and therefore could be implemented in energy generation. This protein, or its coding gene, has not previously been identified as differentially regulated during salt stress. Another ABC transporter also increased 2.35 and 2.66-fold. Finally, a 1.83 and 1.52-fold increase in nitrate transport protein was observed. Interestingly, the uptake of combined nitrogen has previously been suggested to inhibit Na⁺ ion influx, and therefore a mechanism for the protection against salt stress in the freshwater and brackish water cyanobacteria, Anabaena sp. strain L-31 and Anabaena torulosa 34. This response however, has not previously been reported in Synechocystis.

Overall, 26 proteins involved in transcription and translation were identified as differentially expressed in this study, and 13 of these proteins experienced over a 2-fold change (in at least one replicate).

A 2.91 and 3.17-fold increase in elongation factor EF-G may occur for synthesis of so called 'stress proteins' required for cell survival. Cellular stresses like salt stress which lead to global repression of translation, are often accompanied by increased translation of selected proteins that are required for cell survival. It is possible that increased levels of mRNA coding for stress-proteins, compete more successfully for protein synthesis ability in salt stressed cells, than transiently expressed mRNA.

Proteins were differentially expressed in this long-term adaptation study which had only previously been shown to change in the short-term (at transcript and protein level 1018), for example, a 2.21 and 2.22-fold increase in RNA-binding protein. It is possible that this protein assists in stabilising nucleic acid structures that are susceptible to damage caused by changes in water potential and ionic strength. Another protein that assists in tackling the detrimental effects of salt in the long-term is peptidyl-prolyl cis-trans isomerise (2.07 and 2.40-fold increase). This enzyme accelerates protein folding 35. Finally, the 1.83 and 2.32-fold induction of glutamyl-tRNA synthetase (p₊ = 1, 0.99394) has been observed previously in the long-term 18.

This shotgun study identified several proteins that could collectively play an important role to assist a cell's survival in high salt. Cell envelope enzyme UDP-N-acetylmuramoylalanyl-D-glutamyl-2,6-diamino-pimelate-D-alanyl-D-alanine ligase (p₊ = NA, 0.9918) is thought to regulate cell shape by controlling the cell wall structure 36. The cell wall in Synechocystis cells and all bacteria protects the cell from the environment by providing rigidity to counteract internal osmotic pressure. This enzyme is specifically involved in peptidoglycan synthesis, and it is this that provides strength. It is probable that the change in external salt concentration means this enzyme is required in larger amounts to produce more peptidoglycan and help maintain the overall shape of the cell, and confidence in this possibility is increased because of the 2.57 and 2.72-fold increase in enzyme glutamate--ammonia ligase (p₊ = NA, 0.9965), which functions in the same pathway. Both proteins have not previously been identified in the salt response, however, genes in the same pathway have been shown to be induced in heat shocked Synechocystis cells 37, and an increase in peptidoglycan layers for protection was supposed.

Glycerol-3-phosphate dehydrogenase was also identified as salt induced (1.95 and 1.52-fold; p₊ = 0.998, 0.97375) for the first time using proteomics, and is thought to play a role in compatible solute accumulation. This enzyme coding gene, has been induced in the short-term and long-term 10, and its transcript levels were also investigated.

Anti-sigma B factor antagonist was 2.70 and 2.30-fold less expressed in high salt, and despite not being identified previously at the protein or transcript level, this change can be expected. This protein is involved in terminating transcription of the general stress regulon and therefore is not required in 'stressed' conditions. Its mode of action has been studied in Bacillus subtilis, and it is assumed that a reduced amount of anti-sigma B factor makes the stress response process more efficient in adapted cells.

Currently, understanding of the Synechocystis systems response to salt cannot be achieved using MMG for proteins that are not mapped to metabolic pathways. Therefore, an extra level of investigation was included for a selection of differentially expressed non-metabolic proteins by performing qRT-PCR on their respective transcripts. Many proteins acknowledged to play a role in salt acclimation have been discovered previously in the long-term, short-term and at both the protein and transcript level, for example, superoxide dismutase (Additional file 3: Table S2. However, there were also non-metabolic proteins identified in this study, which had opposite fold changes to previous studies. A selection of non metabolic protein coding genes were therefore analysed at the transcript level based on the following premises

i) Proteins not previously associated with salt response (novel study).

ii) Proteins that were controlled in the opposite direction to previous studies (contradictory study).

iii) Proteins or genes previously associated with salt response (confirmatory study).

iv) Proteins or genes with differences between short-term and long-term responses (shock vs. adaptation study)

In total, 14 protein-coding genes were quantified across the time-series and similar to the iTRAQ data, were only treated as significant if both biological replicates exceeded the 1.5-fold threshold for extra stringency. Figures 4a-n show the gene expression changes (dotted bars) after incubation in 6% salt, after 2 hours, 24 hours and 9 days. The dashed bars signify iTRAQ results after only 9 days. The values are all relative to time = 0 or 0% salt, and therefore the first value is recorded as 1 in all cases.

RT-qPCR was performed on four protein-coding genes that have not previously been associated with a salt response (Figure 4a-d). Anti-sigma B factor antagonist has not previously been identified as an essential salt response protein, and Figure 4a illustrates how the gene transcript levels (icfG) decrease after only 2 hours in 6% salt, and remains approximately at this level until cells are fully acclimated. This is the first time to our knowledge that this protein has been identified in a proteomics salt stress study in Synechocystis, so it is not possible to ascertain whether its reduced expression is related to the extent of salt stress.

Figure 4b shows the regulation of the gene encoding the ribosome releasing factor and the corresponding protein. The results between the differential expression based on gene and protein level on day 9 correlate well with reduced expression. An immediate increase in transcript (frr) level occurs after 2 hours of salt stress. This may be due to cells initiating the transcription and translation of general and salt stress specific proteins crucial for long-term adaptation.

The first example of poor correlation between transcript and protein quantitation occurs with an ABC-1 like protein (UbiB) (Figure 4c). It was induced 4.27 and 4.93-fold in high salt and therefore postulated to play a significant role in salt acclimation. Transcript data shows an immediate reduction in expression (2 hours). After a recovery to approximately initial levels (24 hours), the expression of the gene after 9 days is again reduced and this final quantitation contrasts vastly with protein levels.

ATP-dependent Clp protease proteolytic subunit is involved in degrading protein aggregates during stress and can act as a molecular chaperone 38. It would therefore be expected to increase in expression in stressed cells, however, in 6% salt it decreased 2.38 and 2.39-fold. Figure 4d shows that transcripts levels for this gene (clpP) also decrease immediately in salt-stressed cells, and remain reduced after 9 days, supporting the assumption that this subunit does not play a role in acclimation to 6% salt.

RT-qPCR was performed on three protein-coding genes that were controlled in the opposite direction to previous studies (Figure 4e-g).

Glutathione peroxidase is an enzyme whose principal function is to protect cells against damage caused by endogenously-formed hydroxyperoxides (p_- = NA, 0.99985). This type of cell damage is often a result of secondary stress associated with ROS 39 A 3-fold increase in this protein was observed previously in 4% salt stressed cells 18, as well as a transient increase in transcript levels 10. In agreement with these previous studies, the RT-qPCR results also showed a transient increase in transcript (gshP) levels (Figure 4e). It seems likely that after the initial response to oxidative stress, cells that are acclimated no longer require this stress enzyme and therefore there is a decrease in abundance in this iTRAQ study.

A very large decrease in periplasmic phosphate binding protein was observed in the iTRAQ study (10.07 and 10.38-fold). An ABC-transporter which binds phosphate was identified as newly induced in cells adapted to 4% salt 30, and therefore implies that the requirement for phosphate changes significantly in cells adapted to higher salt concentrations (6% salt). Figure 4f shows that transcript levels (pstS) for this protein are only transiently decreased, and in contrast to protein levels, it recovers when approaching adaptation (9 days). This effect of higher salt concentrations on transport proteins is demonstrated further where bicarbonate transporter protein is reduced 2.09 and 3.62-fold in 6% salt. Transcript (cmpA) levels are also stably reduced according to RT-qPCR results (Figure 4g). As with phosphate binding protein discussed above, previous studies show the protein to be newly induced in 4% salt 30, and transcript levels to be over 5-fold induced after 6 hours 10.

RT-qPCR was performed on four protein coding genes which have previously associated with salt response (Figure 4h-k).

The gene encoding for the circadian clock protein (kaiC) is known to increase in 4% salt in the long-term (5 days) 10, but has not been quantified at the protein level formerly. iTRAQ results show a 3.23 and 3.58-fold increase after 9 days, but in contrast to the study by Marin et al. 10, transcript levels quantified here were induced in the short-term too (Figure 4h). Fold changes of both transcript and proteins were very similar after 9 days.

Two further gene transcripts, dnaK and sodA, were quantified for confirmatory purposes, and code for stress chaperone DnaK and anti-oxidative enzyme superoxide dismutase, respectively. DnaK has previously been induced during short and long-term responses to 4% salt 18, and its induction in iTRAQ concurs. Also, the immediate response at transcript level to ~3% salt 17 was seen using RT-qPCR at 24 hours (Figure 4i). However, in this study, its transcript and protein levels were in complete disagreement after 9 days, and this could indicate post-transcriptional control. The same comparison with superoxide dismutase and SOS regulatory protein (LexA protein) produced more complementary results (Figures 3j and 3k).

RT-qPCR was performed on three protein-coding genes where discrepancies arose in regard to short-term and long-term responses between previous work with microarrays, 2DE gels and iTRAQ data presented here (Figure 4l-n).

Hypothetical protein sll1106 does not currently have a putative function, but control of its gene transcription was shown to be 12.6-fold enhanced in ~3% salt using a DNA microarray in the short-term 17, The corresponding protein was quantified by iTRAQ and found to decrease 1.57 and 1.51-fold after 9 days, and these results pose the question of whether this protein is only important for a short-term response to salt stress. RT-qPCR results seem to suggest this is the case, with an increase in transcript level only seen after 24 hours (Figure 4l). Relative changes in transcript and protein levels were in agreement after 9 days. Cell division protein FtsZ was observed to be 1.72 and 1.62-fold reduced in expression in high salt using iTRAQ, and this was expected as growth rate slows. Despite this, Kanesaki et al. 17 reported a short-term increase in this protein in ~3% salt. RT-qPCR and iTRAQ results contrasted completely at 9 days, with transcript levels reduced and protein levels induced, and therefore post-transcriptional control may be involved in cell division (Figure 4m). Cell growth in coordination with cell division is a process thought to be under control of translation regulation 40. The initiation of translation is a complex process, and is the main target for post-translational control 40. Finally, protein and transcript levels of stress co-chaperonin GroES have been seen to be induced only in the short-term previously 1018 whereas iTRAQ results show a 2.61 and 2.93-fold increase after 9 days. Analysis of RT-qPCR data shows a general repression in transcription of the groES gene, despite a brief return to 'normal' levels after 24 hours (Figure 4n). Again, the possibility of post-transcriptional control in expression arises to explain its accumulation after 9 days.