Computational prediction of the osmoregulation network in Synechococcus sp. WH8102
© Mao et al. 2010
Received: 28 September 2009
Accepted: 10 May 2010
Published: 10 May 2010
Skip to main content
© Mao et al. 2010
Received: 28 September 2009
Accepted: 10 May 2010
Published: 10 May 2010
Osmotic stress is caused by sudden changes in the impermeable solute concentration around a cell, which induces instantaneous water flow in or out of the cell to balance the concentration. Very little is known about the detailed response mechanism to osmotic stress in marine Synechococcus, one of the major oxygenic phototrophic cyanobacterial genera that contribute greatly to the global CO2 fixation.
We present here a computational study of the osmoregulation network in response to hyperosmotic stress of Synechococcus sp strain WH8102 using comparative genome analyses and computational prediction. In this study, we identified the key transporters, synthetases, signal sensor proteins and transcriptional regulator proteins, and found experimentally that of these proteins, 15 genes showed significantly changed expression levels under a mild hyperosmotic stress.
From the predicted network model, we have made a number of interesting observations about WH8102. Specifically, we found that (i) the organism likely uses glycine betaine as the major osmolyte, and others such as glucosylglycerol, glucosylglycerate, trehalose, sucrose and arginine as the minor osmolytes, making it efficient and adaptable to its changing environment; and (ii) σ38, one of the seven types of σ factors, probably serves as a global regulator coordinating the osmoregulation network and the other relevant networks.
Osmotic stress refers to the stress on a cell induced by sudden changes in impermeable solute concentrations around a cell that affect the equilibrium with the solution inside the cell. When this happens, water molecules will move in and out of cells by diffusion via the lipid bilayer or the aquaporin channels to regain the equilibrium. The induced water flow across the cell membrane will tend to cause changes in the cell volume, specifically in the cytoplasmic volume, and will induce a number of cellular responses to maintain the homeostasis of the cell's water content . When the solute concentration inside a cell is higher than around the cell, i.e., when under hypoosmotic stress, water molecules will flow inwards, potentially causing animal cells to swell and increasing the turgor pressure in plant and bacterial cells. Alternatively when the impermeable solute concentration inside a cell is lower, i.e., when under hyperosmotic stress, water will flow outwards, hence shocking the cell. In this paper, we focus on the response system to the hyperosmotic stress caused by salt shock knowing that the general knowledge about the hypoosmotic-stress associated response system in prokaryotes is very limited. Throughout this paper osmoregulation refers to osmoregulation in response to the hyperosmotic stress caused by salt shock.
During evolution, all organisms have developed mechanisms to respond to osmotic stresses (or shocks) through tightly regulating a cell's osmolarity so it stays constant, a vitally important condition for cells to survive under changing environments. This regulation process is called osmoregulation. Prokaryotes are known to use two basic osmoregulation mechanisms: (i) the salt-in-cytoplasm mechanism involves adjusting the salt concentration in the cytoplasm according to the environmental osmolarity, and (ii) the organic-osmolyte mechanism involves accumulating uncharged, water-soluble organic compounds to maintain an osmotic equilibrium with the surrounding medium . Previous studies have found that most of the known prokaryotes use the second mechanism , that is to use organic osmolytes such as polyhydric alcohols, sugars, free amino acids and their derivatives, and combinations of urea and methylamines  to adjust a cell's osmolarity. It has been observed that the least salt-tolerant organisms tend to use disaccharides as the osmolytes, whereas the more halo-tolerant and halophilic species use sugar-polyols and nitrogen-containing solutes .
Typically when a cell is under hyperosmotic stress, Na+ and Cl- quickly move into the cell cytoplasm within several seconds; the excessive toxic Na+ are actively exported by Na+/H+ antiporters and the nontoxic K+ are also actively transported into the cell from the environment through its K+ antiporters or symporters to maintain the osmolarity needed by the cell in the first hour; and then transport some compatible osmolytes into the cell from the environment or synthesize them in the cell to replace the K+ surplus within the following several hours . To date, osmoregulation has been well studied in Bacillus subtilis, Escherichia coli and yeast, but very little is known about how osmoregulation works in marine cyanobacteria such as Synechococcus and Prochlorococcus. Synechococcus sp. strain WH8102 is a model organism for organisms that are known to play a key role in global carbon fixation. Since osmolytes represent an important fraction of the fixed carbon, understanding its allocation among different compounds is useful in building a predictive model of these microorganisms.
We have recently carried out a computational study aiming to predict the osmoregulation network in WH8102 by extending and applying a computational protocol for biological network prediction that our group previously developed . This prediction capability consists of three key steps for network prediction, namely (i) construction of template network models for related (model) organisms that have substantial experimental data and possibly known information about the target network, (ii) prediction of operons and functional relatedness among genes in the target genome, and (iii) mapping the template network models to the target genome through orthologous gene mapping that is consistent with the predicted operons and gene associations. This computational protocol has been used to predict the phosphorus assimilation network, the carbon fixation network and the nitrogen assimilation regulatory network in WH8102 [5–7]. By adapting this protocol to our target model, we have constructed a (partial) osmoregulation network for each of five selected organisms, i.e., Aphanothece halophytica (A. halo), Corynebacterium glutamicum ATCC 13032 (C. glut), Escherichia coli K12 (E. coli), Persephonella marina EX-H1 (P. mari), Synechocystis sp. PCC6803 (PCC6803), that have available experimental data related to osmoregulation; made computational prediction of operons as well as of gene functional relatedness in WH8102; and then predicted a model for osmoregulation in WH8102 through mapping the template models to WH8102 in conjunction and validating some of these predictions with experimental data.
Our study has led to a number of new discoveries about osmoregulation in WH8102, including identification of key transporters, synthetases, signal sensor proteins and transcription regulator proteins involved in WH8102 osmoregulation. Through analyses of the predicted regulatory network, we have gained a number of new insights about WH8102 : (i) WH8102 likely accumulates and uses glycine betaine as the major osmolyte, and glucosylglycerol, glucosylglycerate, trehalose, sucrose and arginine as the minor osmolytes; and (ii) σ38, one of the seven types of σ subunits of the RNA polymerase, probably serves as a global regulator in the osmoregulation network in WH8102. To the best of our knowledge, this is the first published study on construction of the osmoregulation network using computation methods for cyanobacteria or any prokaryote.
The osmoregulation process in prokaryotes typically consists of two components. First, under hyperosmotic stress, Na+ will be exported out of the cell and K+ will be transported into the cell as a transient response; and second, some compatible osmolytes will be transported from the environment into or synthesized inside the cell to replace the K+. The selection of the osmolytes depends on the duration of the osmotic stress and the availability of the substrates and osmolytes in the surroundings. Different species may prefer different osmolytes.
Components in the template models
Betaine synthetases with glycine as the
Two-component system that senses osmotic
Na+/H+ antiporter regulator
Predominant K+ channel
K+ channel playing a major role in neutral or slightly alkaline environments
High-affinity K+ channel playing a major role
under osmotic stress
Two component system activating Kdp
expression under osmotic stress
Betaine synthetases with proline as the
Trehalose synthetases with UDPG as the
Two-component system regulating OmpC and
OmpF under osmotic stress
RNA polymerase, Sigma 38 (Sigma S) factor
Predominant K+ transporter playing a major
role in K+ uptake under osmotic stress
High-affinity K+ channel playing a minor role
in K+ uptake under osmotic stress
Two component system activating Kdp expression under osmotic stress
RNA polymerase, Sigma 70 (sigma D) factor
A. halo has three genes encoding a Na+/H+ antiporter to export Na+ out of the cell  and two synthetases known to be involved in the betaine biosynthesis from glycine, compared to the widely used betaine biosynthesis pathway from choline .
E. coli has 31 genes known to be involved in the osmoregulation process, namely a two-component system for regulating two major porin-encoding genes , two Na+ antiporters  and three active K+ transport systems whose activation is determined by the environmental condition [3, 13, 14], and a number of transporters and synthetases for betaine , trehalose [15, 16], putrescine and spermidine , respectively. σ38, one of the seven types of σ subunits of RNA polymerase, is a master regulator in a complex regulatory network that governs the expression of many stationary-phase-inducible genes, and was recently proposed as a global regulator in the osmoregulation network [18, 19];
P. mari has two genes known for glucosylglycerate synthesis ;
PCC6803 has 18 genes known to be involved in osmoregulation process, including two K+ uptake systems in which Ktr plays a major role in K+ uptake under osmotic stress , a Na+/H+ antiporter  and a number of transporters and synthetases for glucosylglycerol [23, 24], sucrose  and arginine ; and
We have used these data as the templates and mapped them to WH8102, and built the initial target network in WH8102.
Components in the initial network model
When mapping the template networks onto WH8102, we noticed that only the Na+/H+ antiporter in PCC6803 maps to SYNW0157 (CPA2 family Na+/H+ antiporter) in WH8102, while Na+/H+ antiporters in E. coli and A.halo do not have any hits. The three K+ uptake systems in the template organisms can only be partially mapped to WH8102 (see Template networks); and only KtrBAE (SYNW2168-2169, 0663), is found in WH8102 through mapping, while the other two systems, Kdp and Kup, could not be mapped (see Table 2). This mapping result seems to make sense since (i) it has been reported that the Ktr system, not the Kdp system, plays a major role in the K+ uptake under osmotic stress in PCC6803; and (ii) the Kup system functions in a low pH environment in E. coli  while the living environment of WH8102 has a pH value 8.1, suggesting that Kup may not be useful for WH8102.
A number of key osmolyte accumulation systems have been found in WH8102 through mapping (see Table 2). Multiple transporters and synthetases for the major osmolyte betaine  are identified in WH8102 : BetT (SYNW0229), BetP (SYNW2494) and ProVWX (SYNW1915-1917) are probably used to uptake betaine from the environment when it is available; GsmT (SYNW1914) and SdmT (SYNW1913) are responsible for synthesizing betaine from glycine; and BetB (SYNW1956) are likely used to synthesize betaine from proline. In addition to the major osmolyte, SpsA (SYNW2520) and SpeA (SYNW2359) are two key enzymes in the sucrose synthesis pathway and arginine synthesis pathway respectively; GgtCDA (SYNW1283-1285) are likely used to uptake glucosylglycerol, trehalose and/or sucrose, and GgpSP (SYNW1281, 0860) are for glucosylglycerol biosynthesis; and GpgSP (SYNW2436, 2434) are used to synthesize glucosylglycerate.
The overall initial osmoregulation model can be summarized as follows: (i) under hyperosmotic stress, WH8102 first uptakes K+ possibly through the Ktr system, and then accumulates the major osmolyte betaine as well as some minor osmolytes such as glucosylglycerol, glucosylglycerate, trehalose, sucrose and arginine through the flexible osmolyte accumulation systems; and (ii) σ38 (SYNW1621) may serve as a global regulator to coordinate the K+ uptake and the osmolyte accumulation processes. We have noted that some key subunits of an osmoregulation network are missing from the initial model based on template mapping alone, such as GgtB (see Table 2), so additional information is needed to expand and refine the model.
We have expanded the initial network model through "guilt by association" based on co-location (operons), co-regulation and co-evolution information that we can calculate through comparative genomic analyses. The basic idea of such association-based prediction is that if protein A is in the initial model but B is not, we will consider adding B to the model if A and B are related based on the aforementioned "co-" analyses. Additional file 6 lists the expanded model of the osmoregulation network through execution of the following steps.
It is well known that genes in the same operon are functionally related, such as enzymes catalyzing subsequent steps in a metabolic pathway, or forming a protein complex . We added 24 new genes into the initial model based on operon predictions for WH8102  (see Additional file 6). Specifically, we added SYNW2165-2167 and 2170 since they share the same operon with SYNW2168-2169 (KtrBA), which could be candidates for the missing subunits of the potassium transporter complex Ktr (see Table 2), although additional experimental studies are needed to validate the prediction. SYNW1279-1280, 1282, 1286 are added to the model since they share the same operons with GgpS (SYNW1281) and GgtCDA (SYNW1283-1285), which are the WH8102 candidates for GgtB (b0529) in E. coli. We noted that this prediction is also supported by phylogenetic profile-based prediction (see Expansion based on phylogenetic profile analysis). SYNW0230-0233 and SYNW2495 are added since they share operons with BetT (SYNW0229) and ProP (SYNW2494), respectively, which are possible candidates for the missing betaine osmolyte transporter subunits. SYNW2435 is added since it shares the same operon with GpgP (SYNW2434), which might be involved in the glucosylglycerate synthesis pathway. SYNW0552 and SYNW2247-2250 are added since they share the same operon, respectively, with SYNW0552 and SYNW2246, which might be used for osmotic signal transduction. SYNW0664-0667 are added since they share the same operon with KtrE (SYNW0663); and SYNW0101 is added since it shares the same operon with RpoD (SYNW0102).
σ38 has been suggested as a global regulator in the osmotic control of gene expression in E. coli , so we added additional genes based on orthology mapping of the σ38 regulon of E. coli to WH8102. We first tested and validated our regulon mapping method in PCC6803 (see Materials and Methods) since there are multiple microarray datasets for this organism. We first collected 103 genes of the σ38 regulon in E. coli from RegulonDB 6.3 , and mapped 49 genes to their orthologs in PCC6803 (see Additional file 7). To assess the reliability of our regulon mapping, we collected microarray data of PCC6803 under hyperosmotic stress from the public domain (see Additional file 8). We found that 11 of the 49 mapped genes showed more than two-fold change in their expression levels. This result has a significant P-value 0.099 (see Materials and Methods), suggesting that our prediction is statistically significant. The possible reason for the P-value being not any lower could be that (i) the effect of the osmotic stress is rather local; or (ii) the time points at which the expression data are collected do not coincide well with the timing of the relevant response process.
Having demonstrated the reliability of the σ38 mapping on PCC6803, we then applied the same mapping to WH810, and mapped 41 genes onto WH8102 (see Additional file 9). We noticed that some genes in the initial network model, e.g., GgpS (SYNW1281), ProVWX (SYNW1915-1918), BetB (SYNW1956) and ProP (SYNW2494), are predicted as targets of σ38, indicating that our mapping functions properly.
Overall, 86 new genes are added to the initial osmoregulation network model, 24 based on co-operon information, 15 based on protein-protein interactions, 41 based on σ38 regulon, and 13 based on predicted co-evolutionary relationships, among which some genes are added by more than one method. For some of these newly added genes, we have predicted their possible functional roles in osmoregulation but for others, all we can say is that we believe that they are involved in osmoregulation in WH8102 but we do not have further information about their functional roles. Experimental studies are clearly needed to elucidate the detailed functions of these genes.
To validate the components in the predicted osmoregulation network and refine it based on the validation results, we have checked our predictions against (i) the published literature related to osmoregulation in WH8102; (ii) whole-genome microarray gene expression data of both WH8102 and PCC6803. Our WH8102 data is collected under mild salt shock (see Materials and Methods), and the PCC6803 data under multiple conditions were collected from public databases; and (iii) protein domain architecture information from public databases.
Our predicted osmoregulation model is highly consistent with the published literature. Lu et al. showed experimentally that WH8102 synthesizes betaine from glycine using SYNW1913-1914 (SdmT and GsmT) , both genes in our initial model. A recent review by Scanlan et al. summarized 15 WH8102 genes known to be involved in the biosynthesis and uptake of betaine and glucosylglycerol , all of which are in our model. Notably, BetAB is considered missing across all cyanobacteria genomes except for Trichodesmium erythraeum and Crocosphaera watsonii, according to this review. We found a homolog (SYNW1956) in WH8102 with an E-value 1.67 × 10-23, and included this homolog in our in our model based on the significance of the E-value. Clearly further experiments are needed to study its possible role relevant to osmotic stress.
Validation of the added genes based on different information sources. The P-value is calculated using hyper-geometric distribution.
Number of recruited genes
Differentially-expressed genes (false discovery rate ≤ 0.01)
5.24 × 10-7
1.44 × 10-2
3.44 × 10-5
We also used conservation information of protein domains to validate the gene pairs predicted by P-MAP mapping with one gene from a template genome and the other being the mapped gene in WH8102. The consideration is that true orthologous genes across two (related) genomes should in general have the same domain architecture. For each gene in our initial network model as well as genes mapped from the σ 38 regulon and its orthologous gene from the corresponding reference genome, we calculated a conservation score for their protein-domain architectures, having the score ranging from 0 to 1, with 1 representing two domain architectures being identical and 0 for being totally different (see Materials and Methods). For 63 such gene pairs, 56 pairs have domain-conservation scores almost being 1, indicating that our orthology mapping works properly. seven gene pairs did not have good domain conservation scores, possibly indicating that these gene pairs may not be correctly mapped, which include the following: SYNW2169 (KtrA) does not have PFPF02080, the TrkA C-terminal domain; SYNW1915 (ProV) does not have PF00571, the CBS domain; SYNW2520 (SpsA) does not have PF00862, the sucrose synthase domain; SYNW0944 (idcC) does not have the PF03709, the Orn/Lys/Arg decarboxylase, N-terminal domain; SYNW1884 (SohB) does not have PF08496, the peptidase family S49 N-terminal; SYNW2260 (NhaR) does not have PF03466, the LysR substrate binding domain; and SYNW1759 (TalA) has two additional tandem PF00036, the EF hand domain.
Na + export system : SYNW0157 functions as a Na+/H+ antiporter to export the excessive Na+ out of the cell under hyperosmotic stresses;
K + uptake systems : SYNW0663, SYNW2168-2169 consist of a Ktr system and may play a major role in active K+ uptake under hyperosmotic stresses;
Osmolyte accumulation systems : SYNW0229 (BetT), SYNW1915-1917 (ProVWX) and SYNW2494 (ProP) are used to uptake the major osmolyte betaine if available in the environment, while SYNW1913-1914 (SdmT and GsmT) are used to synthesize betaine from glycine when needed; SYNW2436 (GpgS) and SYNW2434 (GpgP) are used to synthesize glucosyl glycerate or mannosylglycerate from glucose; SYNW1281 (GgpS) and SYNW0860 (GgpP) are used to synthesize glucosylglycerol from glucose, while SYNW1283-1287 (GgtABCD) uptake glucosylglycerol when available in the environment; SYNW2359 (SpeA) is for synthesizing the minor osmolyte arginine; and SYNW2520 (Sps) for synthesizing the minor osmolyte sucrose;
Global regulator : SYNW1621 (σ38) functions as a global transcription regulator possibly to bridge the K+ uptake and osmolyte accumulation processes and to coordinate osmoregulation with other biological processes; and
Two-component signal transduction systems : SYNW0807-0808 (EnvZ and OmpR) regulate water across the outer membrane and gene responses of some osmoregulatory elements; and SYNW0551 and SYNW2246 may be responsible for sensing external osmotic stress and/or activating a number of genes relevant to osmoregulation, which suggests other member genes should also be relevant to osmosensing but further experiments are needed to derive more information about this.
From our predicted model, we found that (i) WH8102 likely accumulates and uses glycine betaine as the major osmolyte, and glucosylglycerol, glucosylglycerate, trehalose, sucrose and arginine as the minor osmolytes; and (ii) σ38, one of the seven types of σ subunits of the RNA polymerase, probably serves as a global regulator in the osmoregulation network in WH8102. We believe that this working model provides useful information to experimental biologists in their research design for further studying the osmoregulation process in WH8102. To the best of our knowledge, this model represents the first published study on construction of the osmoregulation network using computation methods for cyanobacteria or any prokaryote.
All the used genome sequences and annotations were retrieved from NCBI FTP (ftp://ftp.ncbi.nih.gov/genomes/Bacteria, 05-02-2008). The σ38 regulon of E. coli are downloaded from RegulonDB 6.3 .
We used the P-MAP program  with E-value ≤ 10-6, in conjunction with BLAST  with E-value ≤ 10-20, to map template networks to WH8102. The mapping results from different organisms were then merged, which gives the initial network model. When different functional roles are assigned to the same gene or one functional role is assigned to different genes based on different templates, we resolved the conflict by assessing the information of evolutionary distance of organism, the conservation of protein domain architectures, and available gene expression values (see below).
We downloaded the predicted operons of WH8102 from the DOOR database  and the protein-protein interactions from http://www.cs.uncc.edu/~zcsu/pathways/nitrogen/nitrogen.
with , d H is the Hamming distance between the phylogenetic profiles of g i and g j, α is a weighting factor (the default is α = 2), and h is the number of genomes that have orthologs of both g i and g j.
Using this measure, we can define for each WH8102 gene its functionally most related genes. For each gene g in our initial model, we recruit all genes into the model, which are more closely related to g than any gene already in the initial model.
We used the Pfam database  as the source of protein domain definitions, and the InterproScan program  to identify all protein domains in the template genomes and the target genome. To measure the similarity between two given domain architectures, we implemented a domain-architecture similarity score, which was originally defined by Lin et al  as a linear combination of three different indices: the Jaccard index, which measures the number of common domains that the two architectures contain; the Goodman-Kruska γ function, which estimates the similarity of the arrangement of the distinct domains shared by the two architectures; and the domain duplicate index, which assess the similarity among the duplicated domains in the two architectures. The related scripts can be downloaded from http://csbl.bmb.uga.edu/~xizeng/research/osmoregulation.
We implemented a script based on BioRuby http://bioruby.org to calculate the average fold-change of expression levels for each gene under two conditions if microarray data is available, otherwise we have collected genes from published papers, which were found to be differentially expressed (see Additional file 12). We have used fold-change = 2.0 and 0.5 as the cutoffs for calling a gene up- or down-regulated, respectively, and no change for values in-between.
Microarray data was obtained using a whole-genome microarray, design of which was described previously . Cultures of Synechococcus sp. strain WH8102 were grown in synthetic ocean water-based media with supplemented nutrients as described previously . This medium is 75% seawater salinity. Triplicate cultures were grown to mid-exponential at which point RNA was extracted from half of each culture and designated as control RNA; and 0.125 M NaCl was added to the other half of each culture bringing its salinity up to approximately that of seawater. After two hours RNA was extracted from the remaining half of the culture. RNA was harvested using a Trizol-based method and purified with a Qiagen RNeasy kit according to the manufacturer's specifications. Microarray hybridization was performed as described previously . Briefly, an indirect labeling method was used to label cDNA with Cy3 (control samples) or Cy5 (treated samples), samples were then pooled and hybridized to the same array. This was done in duplicate for each biological replicate. Additionally, reverse labelling was performed for each biological replicate, resulting in three total technical replicates per biological replicate. Following hybridization, slides were promptly scanned at a 10-μm resolution using an Axon 4000B scanner with GenePix 4.0 software. Processing of the TIFF images from hybridized arrays was performed using TIGR-Spotfinder http://www.tigr.org/software, and the datasets normalized by applying the LOWESS algorithm, using block mode and a smooth parameter of 0.33, available in the TIGR-MIDAS package http://www.tigr.org/software. Statistical analysis was performed on the mean of log2-transformed signal ratios of the replicate spots using the Statistical Analysis of Microarrays (SAM) algorithms  with a false discovery rate (FDR) of less than 1%.
where N is the number of all genes in the genome, and A is the number of differentially expressed genes under hyperosmotic stress; K is the number of genes in our network; and V is the number of predicted genes in our network that are differentially expressed under hyperosmotic stress.
Pathway enrichment analysis can usually give biological insights about the genes showing differential expressions under two conditions [43, 44]. We used the KOBAS program  to identify significantly enriched pathways by genes in our predicted osmoregulation network against all known genes in WH8102 as the background. KOBAS maps interested genes onto known KEGG pathway  using BLAST with E-value ≤ 1e-5 and rank ≤ 10, and then uses hyper-geometric distribution to calculate the statistical significance of each pathway populated by the interested genes with respect to all the genes encoded in the whole genome as the background. The relevant formulas have been well described in previous work .
This work was supported in part by the National Science Foundation (DBI-0542119 to YX and BP, DEB-0830024) and the DOE BioEnergy Science Center grant. We thank all the members of the comparative genomics group of the CSBL Lab at UGA, especially Dr. Fenglou Mao for the help with the P-MAP program and operon prediction. We thank Mr. Bingqiang Liu for helpful discussions about orthology mapping, and Dr. Huiling Chen for discussions about structure prediction, Dr. Xiaojia Tang for discussions about phylogenetic profile analysis. We also thank Dr. Zhengchang Su and Ms. Shan Li of the University of North Carolina at Charlotte for providing programs for phylogenetic analysis and data published in the previous work.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.