Introduction

Histidine–aspartate phosphorelay systems, also known as two-component signal transduction systems (TCSTSs), are one mechanism that bacteria use to sense and respond to their environment. Arguably the most widespread means of signal transduction in bacteria, these systems have been shown to regulate a wide variety of cellular responses, including osmoregulation, photosynthesis, chemotaxis, sporulation, antibiotic production, and pathogenicity, in a number of different bacteria (Appleby et al. 1996; Quon et al. 1996; Fabret and Hoch 1998). TCSTSs are typically composed of two signaling proteins: a sensor histidine kinase (HK) and its cognate response regulator (RR) (Egger et al. 1997). A more complex version of this two-component phosphotransfer scheme involves multiple phosphotransfer steps and often more than two proteins. Most of the multiple-step phosphorelays involve hybrid-type HKs in which both histidine- and aspartate-containing domains are present within a single protein. The phosphoryl group is first transferred within the hybrid-type HKs from histidine to aspartate residue, then to an intermediate His-containing phosphotransfer (HPT) protein, and finally to a cytoplasmic RR. These expanded phosphorelay systems provide greater versatility in signaling strategies and a greater number of potential sites for regulation (Appleby et al. 1996). Based on the fact that most of the eukaryotic HKs were hybrid-type, it has been suggested that TCSTSs in prokaryotes generally use a simple phosphotransfer scheme, whereas multi-step phosphorelays using hybrid HKs dominate two-component signaling in eukaryotes (West and Stock 2001; Oka et al. 2002; Catlett et al. 2003; Zhang and Shi 2005) (Fig. 1).

Figure 1
figure 1

Domain structures of bacterial two-component signal transduction systems. (a) Typical two-component phosphotransfer system containing sensor HK and cytoplasmic RR. (b) A multi-component phosphorelay system containing hybrid-type HK and a His-containing phosphotransfer (HPt) protein.

Full size image

Sulfate-reducing bacteria (SRB) are a group of obligate anaerobic microorganisms that exist in diverse environments, display a spectrum of morphologies, and have distinct metabolic capabilities. Nevertheless, they all share the physiological activity of using sulfate ions, SO −24 , as the terminal electron acceptor in respiration and generating S−2 and HS- ions that form hydrogen sulfide (Voordouw 1995). In addition, some SRB strains are also able to enzymatically reduce many metals including chromium (VI), iron (III), and uranium (VI) by using them as an alternative terminal electron acceptor (King et al. 2000; Spear et al. 2000). The mechanisms involved in using organic acids as the electron donor for sulfate reduction and metal reduction have been extensively studied in the model SRB, Desulfovibrio vulgaris, for the past decade (Voordouw 1995). These studies showed that D. vulgaris belongs to the group of incompletely oxidizing sulfate reducers, where acetic acid is often an end product of the organic acid oxidation that provides electrons for sulfate reduction, and several pathways have been proposed for ATP synthesis (Heidelberg et al. 2004). However, how these processes are regulated is still unknown. In the recently completed D. vulgaris genome, a few dozen sensor HKs and RRs were identified (Heidelberg et al. 2004). Although little knowledge is so far available on physiological functions of these genes, the existence of such large numbers of TCSTS genes in the D. vulgaris genome clearly indicates the organism’s ability to sense a broad array of environmental signals.

In this study, phylogenetic and structural analyses of HKs and RRs were performed. Although bootstrap support was not strong because of the low ratio between the number of aligned residues and the number of operational taxonomic units (OTUs) (Koretke et al. 2000), phylogenetic trees constructed for TCSTSs from D. vulgaris, along with representatives from Escherichia coli and Bacillus subtilis, showed that most TCSTS components of D. vulgaris could be clustered with several subfamilies previously recognized in E. coli and B. subtilis. We also made the preliminary attempt to identify cognate HK-RR gene pairs based on the similar expression pattern across various conditions. D. vulgaris was subjected to various environmental and genetic alterations, and expression of all TCSTSs was measured by whole genome microarray. After verification using HK and RR components adjacent on the chromosome, the approach was applied to search for HK-RR pairs not linked on the chromosome and five sets of possible cognate HK-RR gene pairs were identified, which, although this still must be proved, constitutes a basis for further exploration of interaction and physiological function of TCSTSs in D. vulgaris.

Materials and Methods

Strains, Media, and Culture Conditions

Desulfovibrio vulgaris DSM 644 was obtained from the Deutsche Sammlung für Mikroorganismen und Zellkulturen (DSMZ) and grown under strictly anaerobic conditions as described previously (Keon et al. 1997). Cells were grown at 30°C in minimal medium, which contains per liter: 1.0 g yeast extract, 6.6 g (NH4)2SO4, 1.7 g NaCl, 0.85 g KH2PO4, 0.34 g MgCl2 · 6H2O, 0.85 g NaHCO3, 0.26 g CaCl2 · 2H2O, and 1.0 % (v/v) trace elements solution (per 1: 12.8 g nitrilotriacetic acid, 0.1 g FeCl2 · 4H2O, 0.3 g MnCl2 · 4H2O, 0.2 g CoCl2 · 6H2O, 0.2 g ZnCl2, 50 mg Na2MoO4 · 2H2O, and 20 mg H3BO3; pH 6.5 with NaOH). Lactate (38 mM) or formate (38 mM) was used as the energy and carbon source. Growth experiments were performed at 30°C in 140-ml serum bottles containing 70 ml of medium, with head-space filled with a gas mixture of 10% (v/v) CO2 and 90 % (v/v) N2. After inoculation (10% volume) from a preculture grown under the same conditions, cultures were incubated for 2–3 days until they reached exponential phase (OD590 of 0.4 and 0.2 for lactate and formate, respectively) or stationary phase (OD590 of 0.65 and 0.35 for lactate and formate, respectively), as determined in a Shimadzu BioSpec 1601 Analyzer (Kyoto, Japan). Cells were collected by centrifugation at room temperature under anaerobic conditions and stored at –80°C. For RNA profiling, each sample used was a pool of three individual biological replicates.

The D. vulgaris wild type (wt) and mutants used in this study are listed in Table 1. Cultures for microarray analysis were grown on lactate-based medium (unless noted) to exponential (E) or stationary (S) phase under the following conditions: a) wt on formate to E and S; b) wt to E and S; c) wt to E and then subjected to heat shock at 45°C for 1 h; d) wt to E and then subjected to oxygen exposure by continuous bubbling of the culture with air for 1 h; e) wt to E and then N2 in the headspace was changed to H2 (85% hydrogen) and incubated for 0.5 and for 3 h; and f) mutants ADH, RK2, H100, and H801 grown individually to E.

Table 1 D. vulgaris strains used in the study
Full size table

Genomic Analyses of Putative HKs and RRs in D. vulgaris

Deduced protein sequences of TCSTSs from E. coli spp. were selected as the source of query sequences, because the biochemical and genetic characteristics of their TCSTS pathways are well known (Stock et al. 1989). Protein sequences of 30 HKs and 21 RRs were extracted from the E. coli genome. The sequences were subjected to domain identification for histidine kinase (phosphoacceptor) (HisKA) and histidine kinase-like ATPases (HATPase_C) domains, and signal receiver domain using the molecular architecture research tools provided by SMART (http://www.smart. embl-heidelberg.de/ ) with E value < 0.01 (Letunic et al. 2002). The conserved HisKA-HAPTase_C and receiver domains identified were then used to search D. vulgaris genome using PSI-BLAST. The homologous sequences of D. vulgaris with E value less than 0.01 were extracted and subjected to further confirmation by domain identification. Protein sequences of putative HisKA and HATPase_C, and signal receiver domain with E value < 0.01 were extracted.

Phylogenetic Analyses of Putative HKs and RRs in D. vulgaris

Deduced protein sequences of the homologous HisKA-HATPase_C domains of the histidine kinase (with average length of 220–240 amino acids), and the homologous receiver domains of response regulator proteins (with average length of 110 amino acids) were aligned using default parameters of the CLUSTALW program (Version 1.82) (Thompson et al. 1994) available from EMBL-European Bioinformatics Institute (http://www.ebi.ac.uk/clustalw/ ) (Protein Gap Open Penalty = 10.0, Protein Gap Extension Penalty = 0.2, Protein matrix = Gonnet). Multiple alignments were modified after visual examination, and a minimum number of gaps were inserted manually in order to produce what looked like a reasonable alignment. Alignments were subjected to two phylogenetic tree-building methods: Neighbor Joining (Saitou and Nei 1987) and Maximum Parsimony implemented in the program PAUP, Version 4.0 (Phylogenetic Analysis using Parsimony, beta version, Sinauer Associates, Sunderland, MA) and 1000 replications of bootstrap sampling were performed. In the NJ analyses, the number of characters re-sampled in each replicate is 484 for HKs and 195 for RRs, respectively.

RNA Isolation and Labeling

Frozen D. vulgaris cell pellets (150–200 μl) were used for RNA isolation. RNA isolation and labeling were performed as described previously (Zhang et al. 2005). The concentration and purity of the RNA was determined using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies), and the integrity was verified by gel electrophoresis. The quality of total RNA obtained was checked using the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). Biotin labeling of RNA was carried out as described previously (Zhang et al. 2005). cRNA was purified using Qiagen RNeasy RNA purification columns according to the manufacturer’s instructions (Qiagen Inc., Valencia, CA). RNA yield was determined by absorbance at 260 nm. Before hybridization, cRNA was fragmented to an average size of 50 to 200 bp by incubation in 100 mM potassium acetate, 30 mM magnesium acetate, and 40 mM tris-acetate for 35 min at 94°C. The biotinylated cRNA fragments were checked by gel electrophoresis in 1% agarose, and then purified using a Microcon clean-up column.

Microarray Experiments

D. vulgaris microarrays were designed by NimbleGen System Inc. (Madison, WI) using its Maskless Array Synthesizer technology (Nuwaysir et al. 2002). The D. vulgaris genome sequence and annotation were extracted from the Comprehensive Microbial Resource of The Institute of Genomics Research (TIGR, http://www.tigr.org ) (Heidelberg et al. 2004). The arrays containing all 3548 ORFs in D. vulgaris genome were manufactured as described before (Nuwaysir et al. 2002). For each ORF, 13 unique 24-mer oligonucleotides distributed throughout the ORF were printed onto glass microscope slides. Microarrays were hybridized with 10 μg cRNA in 300 μl, in the presence of 50 mM MES, 0.5 M NaCl, 10 mM EDTA, and 0.005% (v/v) Tween-20 for 16 h at 45°C. Before application to the array, samples were heated to 95°C for 5 min, allowed to cool, and then held at 45°C for 5 min, and spun at 14,000g for 5 min. Hybridization was performed in disposable adhesive hybridization chambers from Grace BioLabs in a hybridization oven with agitation. After hybridization, arrays were washed in nonstringent (NS) buffer (6 × SSPE, 0.01% [v/v] Tween-20) for 5 min at room temperature, followed by washing in stringent buffer (100 mM MES, 0.1 M NaCl, 0.01% Tween-20) for 30 min at 45°C. After washing, arrays were stained with streptavidin-cy3 conjugate from Amersham Pharmacia for 25 min at room temperature, followed by a 5-min wash in NS buffer, a 30-sec rinse in 1 × NimbleGen final rinse buffer, and a blow-dry step using high-pressure grade-5 Argon.

Data Normalization and Gene Expression Analysis

Prior to data extraction, images were rotated and doubled in size (without interpolation) using ImageJ software (http://www.rsb.info.nih. gov/ij/). Features were extracted using GenePix 3.0 software (Axon Instruments, Inc.), using a fixed feature size. The local background correction from the GenePix software was not applied to raw signal intensities. The data was normalized using tools available through the Bioconductor project (http://www.bioconductor.org). The gene calls were based on the Bioconductor implementation of the MAS 5 algorithms. For each experimental condition, there were four measurements for each gene: two replicates (each contains a pool of three biological replicates) that were each hybridized to duplicate microarrays. The genes with an anti-correlation expression pattern within replicates were manually excluded from the analysis. The Wilcoxon signed rank test statistical analysis was performed to check similarity of gene expression patterns, with high p value (>0.01) indicating high similarity as previously described (Ali et al. 1994).

Results

Survey and Structural Analysis of Putative HKs and RRs in D. vulgaris

To perform a phylogenetic analysis of a complete set of D. vulgaris HKs and RRs, we started by identifying all of the putative HK and RR encoding genes contained in D. vulgaris genome, as described in Materials and Methods. The approach identified a total of 59 HKs and 55 RRs from D. vulgaris genome (http://www.tigr.org), which is almost twice the number identified from E. coli and B. subtilis, although the size of the D. vulgaris genome is smaller (Koretke et al. 2000) (Table 2). DVU2580, which was originally annotated as a RR protein (Heidelberg et al. 2004), was re-named as a hybrid-type HK because it contains putative HisKA and HATPase_C at its N-terminus.

Table 2 Number of HKs and hybrid-type HKs in selected bacterial genomesa
Full size table

The typical HK is a transmembrane receptor with an amino-terminal extracellular sensing domain and a carboxy-terminal cytosolic two-component signaling domain (West and Stock 2001). Domain analysis of D. vulgaris HKs showed that 40 HKs (Fig. 2) possess at least one sensory domain at their N-termini, along with the HATPase_C domain and HisKA at the C-termini. The latter two domains are necessary for the binding of ATP, autophosphorylation at the histidine residue located in the His-box, and phosphate transfer to the aspartate residue in the RR receiver domain (Fig. 1). Three domains involved in sensory function were detected in D. vulgaris HKs; PAS domain was found in 36 HKs where they may be involved in sensing energy-related environmental factors such as oxygen, redox potential, or light (Taylor et al. 1999); GAF domain in 4 HKs where they may be involved in binding cyclic nucleotides (Aravind and Ponting 1997); and HAMP domain in 8 HKs, which is a common structural element hypo- thesized to transmit signals between external input domains and cytoplasmic output modules in sensor HKs and methyl-accepting chemotaxis proteins (MCPs) (Aravind and Ponting 1999; Williams and Stewart 1999). However, 12 HK proteins were found containing only HisKA and HATPase_C domains, so it is possible that their sensory domains are located on a separate protein, or they may function as intermediate proteins within complex signal cascades (Wang et al. 2002). Seven groups (Groups 8–14), with 21 putative HK proteins, met the criteria of hybrid-type HKs because they contain both HK and receiver domains in a single molecule (Fig. 2).

Figure 2
figure 2

Schematic representation of the functional domains found in the histidine kinases of D. vulgaris. Group 1 includes 12 HKs with only histidine kinase domains. Groups 2–7 contain HKs with both histidine kinase and sensory domains. Group 8–14 contain HKs with both histidine kinase and receiver domains. Number of genes in each group is listed. HisKA, His kinase A (phosphoacceptor) domain; HATPase_C domain, with histidine kinase-like ATPases; HAMP, (histidine kinases, adenylyl cyclases, methyl binding proteins, phosphatases) domain; GAF, domain present in phytochromes and cGMP-specific phosphodiesterases; PAS, domain sensing energy-related environmental factors; REC, receiver domain of RR. The sizes of these domains are not drawn to scale. Multiple occurrence of same sensory domain in the same sequence is not shown. The name of genes belonging to various groups is provided in Supplementary Table 1.

Full size image

According to their modular structure, the 55 RRs in D. vulgaris were classified into 10 groups (Fig. 3). Twenty-four RRs were found possessing only receiver domains without any identifiable output domains, whereas all other RRs have output functional domains located in their C-terminus that may have either DNA-binding or catalytic activity. The most commonly found output domain was the AAA domain, identified in 18 RR proteins. The AAA domain has been suggested to have ATPase activity associated with a variety of cellular activities, including DNA replication, recombination, and restriction, prokaryotic NtrC-related transcription, and the sporulation of Bacillus (Neuwald et al. 1999). Other identified functional domains include the PAS domain commonly involved in environmental signaling; HDc domain that is found in a superfamily of enzymes with phosphohydrolase or 3′, 5′-cGMP phosphodiesterase activity (Aravind and Koonin 1998); HATPase_C encoding histidine kinase-like ATPases; LuxR domain which binds DNA through its “helix-turn-helix” motif and activates the bioluminescence operon in Vibrio fischeri (Sitnikov et al. 1995); PP2C encoding serine/threonine phosphatase activity (Bork et al. 1996); and HPT (the histidine-containing phosphotransfer) domain with an active histidine residue involved in multistep phosphorelay signaling systems (Kato et al. 1997) (Fig. 3).

Figure 3
figure 3

Schematic representation of the functional domains found in the response regulators of D. vulgaris. Domain descriptions: REC, receiver domain of RR; AAA domain possessing ATPase activity; PAS domain sensing energy-related environmental factors; HDc domain with phosphohydrolase or 3′,5′-cGMP phosphodiesterase activity; HATPase_C with histidine kinase-like ATPases; LuxR domain with DNA binding ability; PP2C domain with serine/threonine phosphatase activity; and HPT domain involved in multistep phosphorelay signaling systems. Number of genes in each group is listed. The sizes of these domains are not drawn to scale. The name of genes belonging to different groups is provided in Supplementary Table 2.

Full size image

Phylogenetic Analyses of Putative HKs and RRs in the D. vulgaris Genome

To infer the evolutionary courses of D. vulgaris HKs and RRs, phylogenetic analysis was performed using the deduced protein sequences of the HisKA and HATPase_C domains for HKs and receiver domains for RRs, rather than intact protein sequences, in order to increase the confidence level of sequence alignment. For hybrid-type HKs, both HisKA and HATPase_C domains and receiver domains were extracted and used in phylogenetic analysis of HKs or RRs, respectively. Protein sequences of the HK domains were aligned using the ClustalW program. A portion of alignment of the deduced protein sequences of several HKs was shown in Figure 4, with all the features characteristic of the histidine protein kinases family, including the conserved histidine residue, presumably the site of phosphorylation, and four other highly conserved sequences termed H, N, G1, F, and G2 boxes indicated (Dutta et al. 1999). The trees were constructed by both Neighbor Jointing (NJ) and Maximum Parsimony (MP) methods and evaluated by 1000 bootstrap replicates (Fig. 5 and Supplementary Fig. 1). Despite the low ratio between the number of aligned residues and the number of OTUs in both HK and response regulator data sets, which restricted the resolution of the phylogenetic analyses (Koretke et al. 2000), phylogenetic analyses showed that most of the D. vulgaris HKs and RRs were classified into several previously known subfamilies according to their clustering with those of E. coli and B. subtilis. Overall, better bootstrap support was obtained for HK tree, in which most of the major phylogenetic clusters occurred in >50% of 1000 random bootstrap replicates in both NJ and MP analyses (Fig. 5). Although several individual exceptions were present, the NJ and MP analyses presented very similar clustering characteristics for most of the HKs and RRs (Fig. 5 and Supplementary Figure 1).

Figure 4
figure 4

Alignment of HisKA and HATPase_C domains of HKs from D. vulgaris. Sequences from D. vulgaris were indicated by gene ID starting with DVU. In addition, NtrB and PhoQ of E. coli and PhoR of B. subtilis are also included in the alignment. The alignment is constructed using the ClustalW program (Version 1.82). The conserved residues in the highly conserved regions are marked by shading. The highly conserved sequences termed H, N, G1, F, and G2 boxes, characteristic of histidine kinases, are boxed. The conserved Histidine residue (H) is indicated in bold.

Full size image
Figure 5
figure 5

Phylogenetic trees of histidine kinases (A) and response regulators (B) constructed by Neighbor-Jointing method. Selective HKs and RRs from E. coli, B. subtilis, and all HKs from D. vulgaris are included in the trees and are designated with the prefix Ec, Bs, and DVU, respectively. Deduced protein sequences of the HisKA-HATPase_C domains of the histidine kinase (with average length of 220–240 amino acids) and deduced protein sequences of the receiver domains of response regulator proteins (with average length of 110 amino acids) are used for tree construction as described in Materials and Methods. Definitions of subfamilies are indicated as described by Koretke et al. (2000). Three pyruvate dehydrogenase kinases (PDK) are used as out-group for the HK tree. For each phylogenetic tree, 1000 bootstrap replicates are analyzed and the numbers of bootstrapping are indicated as percentage.

Full size image

The HKs and RRs from D. vulgaris were found to be distributed in six phylogenetic subfamilies—Che, Ntr, Pho, Lyt, Nar, and Hybrid—although bootstrap supports for some of the clusters are weak (Koretke et al. 2000). The Ntr subfamily, which is known to occur in systems regulating nitrogen assimilation (NtrB/NtrC, McFarland et al. 1981), short-chain fatty acid degradation (AtoS/AtoC, Jenkins and Nunn 1987) and hydrogenase activity (HydH/HydG, Stoker et al. 1989; Koretke et al. 2000), was the most common TCSTS system found in D. vulgaris with 34 HKs and 39 RRs according to the NJ analysis. The Hybrid subfamily was the next most common subfamily with 12 HKs and 24 RRs according to the NJ analysis; among them, 14 RR domains were from hybrid-type HKs. However, the subfamilies dominant in E. coli and B. subtilis TCSTSs, such as Pho systems that contain TCSTSs involved in phosphate regulation (PhoR/PhoB; Tommassen et al. 1982) and anaerobic nitrite reduction (ResE/ResD; Nakano et al. 1996), and Nar systems that contain TCSTSs regulating anaerobic respiration (NarQ/NarP; Darwin et al. 1998) and sugar phosphate uptake (UhpB/UhpA; Island et al. 1992), contain only a small fraction of TCSTSs in D. vulgaris (Fig. 5). In addition, no obvious DVU-specific cluster with bootstrap support was observed in either HK or RR phylogenetic trees (Fig. 5).

As previously discovered by Koretke et al. (2000), the cognate pairs of HKs and RRs of E. coli and B. subtilis that are known to interact were overwhelmingly found in corresponding clusters in phylogenetic trees. In addition, the connection between the functional coupling of HK and RR genes and their chromosomal vicinity has been suggested by a recent survey showing that 75 of 84 HKs adjacent on the chromosome with RRs were with RR modules in corresponding phylogenetic subfamilies (Koretke et al. 2000). A phylogenetic approach was recently applied successfully in the identification of cognate TCSTSs in Streptococcus pneumoniae (Throup et al. 2000). Twenty-six pairs of HK-RR adjacent on the chromosome were identified from the D. vulgaris genome. Among these adjacent pairs, 16 pairs possess overlapping stop and start codons, a device that is thought to lead to translational coupling, and 6 pairs with an intergenic region smaller than 20 bp, only 4 pairs with intergenic region larger than 20 bp (Table 3), which suggested that most of these HK-RR pairs adjacent on the chromosome could be in the same gene clusters, and they are very likely to be cognate pairs. Analysis of their distribution in both HK and RR phylogenetic trees showed that 19 of 26 pairs have their cognate partners located in corresponding subfamilies (Fig. 5, Supplementary Fig. 1, and Table 3). In addition, the majority of the receiver domains from hybrid-type HKs were clustered together in the RR phylogenetic tree, concordant with their corresponding histidine kinase domains from the same hybrid-type HKs, which are also clustered mainly into one clade in the HK phylogenetic tree. The results initially suggested that the concordance of HKs and RRs in corresponding phylogenetic groups could be indicative of possible cognate HK-RR pairs. However, the fact that large number of TCSTSs in D. vulgaris were narrowly distributed in the Ntr and Hybrid subfamilies, as well as the relatively low bootstrap supports for RR trees impeded the deduction of cognate HK-RR pairs just according to their distribution in phylogenetic trees (Fig. 5).

Table 3 Phylogenetic distribution of HK-RR pairs adjacent on chromosome
Full size table

Analysis of Putative HKs and RRs in Response to Growth Condition or Genetic Changes

Attempts have been previously made to systematically determine mRNA level of all two-component regulatory system mutants of E. coli (Oshima et al. 2002) and B. subtilis (Kobayashi et al. 2001), based on the assumption that a change in expression of either sensor HK or RR will affect the expression of its cognate partner and corresponding target genes involved in regulation of cellular functions. The analyses revealed various target gene candidates for these TCSTSs, and allowed deduction of the probable physiological functions of some TCSTSs with previously unknown functions (Kobayashi et al. 2001). Using whole genome microarrays, the expression level of all TCSTS components was analyzed in response to 12 environmental and genetic changes: different carbon sources (lactate or formate), different growth phases (exponential or stationary phase), different stress conditions (oxidative or heat shock), different lengths of exposure to high hydrogen concentration, and four different gene knockout mutants involved in energy metabolism (Table 1). For each gene, the gene expression data were an average of four measurements of two biological replicates. Using the lactate at exponential phase condition as reference, ratio between expression levels of each environmental and genetic change and the reference condition was calculated for all TCSTS genes.

As a test, the approach of transcriptomic analysis was verified first by analyzing the expression patterns of the 26 HK and RR gene pairs adjacent on the chromosome. Using the criterion of differential expression of at least 1.5-fold in 3 or more of the 12 conditions as a cutoff, only 4 HK and RR gene pairs adjacent on the chromosome were responsive to the test conditions and can be used for expression pattern clustering analysis. Expression pattern analysis showed that all four HK and RR gene pairs adjacent on the chromosome possessed HKs and RRs differentially expressed in similar patterns across all experimental conditions, which was further confirmed by the Wilcoxon signed rank test with p values greater than 0.1 (Fig. 6A). Interestingly, all four HK-RR pairs belonged to the same Ntr subfamily by phylogenetic analysis (Fig. 5).

Figure 6
figure 6

Expression pattern of TCSTS pairs in response to various environmental and genetic changes. (A) Expression patterns of four pairs of HK-RR adjacent on chromosome; (B) Expression patterns of four groups of HK-RR not linked on the chromosome. The X-axis indicates various test conditions: 1) Heat shock treatment versus untreated, 2) Oxidative stress versus untreated, 3) Stationary versus Exponential phases in lactate-based medium, 4) Stationary versus Exponential phases in formate-based medium, 5) Formate versus Lactate in exponential phase, 6) Formate versus Lactate in stationary phase, 7) Addition of high hydrogen concentration for 30 min versus untreated, 8) Addition of high hydrogen concentration for 3 h versus untreated, 9) ADH100 mutant versus wild type, 10) H801 mutant versus wild type, 11) Hyd100 mutant versus wild type, and 12) RK2 mutant versus wild type. The Y-axis indicates the fold change of TCSTS genes across conditions. The p values of the Wilcoxon signed rank test for predicted cognate HK-RR pairs (A, B, C, and D) are indicated in the figures, whereas the p values of the Wilcoxon signed rank test for various HK-RR combinations are indicated in Table 4.

Full size image

The approach was then applied to D. vulgaris TCSTSs whose components were not physically linked on the chromosome. Using the same criterion described for HK-RR gene pairs adjacent on the chromosome, 26 gene-encoding TCSTS components were determined as responsive under the testing condition and were eligible for expression pattern analysis, which included 13 genes encoding putative HKs and 13 gene-encoding putative RRs. Analysis of the expression pattern identified 11 sets of TCSTS genes expressed in a similar pattern across various experimental conditions, which included 6 gene sets containing only RR genes (data not shown) and 5 sets with both HK and RR genes (Fig. 6B). The similarity of gene expression patterns within each group was also confirmed by the high p values of the Wilcoxon signed rank test. Among the five gene sets, four gene sets each contain only one HK-RR pair: DVU2395-DVU2577 was up-regulated in stress conditions (heat, oxygen, stationary phase) and down-regulated in all other conditions; DVU2281-DVU0722, which may be part of a multiple-step phosphorelay system since the DVU2281 is a hybrid-type HK, was down-regulated in all conditions except one and most greatly down-regulated with heat shock and in stationary phase with formate medium; and DVU0119-DVU3334 was up-regulated in all four energy metabolism–related mutants. The mean expression of the last gene set containing 5 HKs (DVU0013, DVU0258, DVU0680, DVU3022, and DVU3058) and 2 RRs (DVU110 and DVU3305) was significantly down-regulated at stationary phase in formate-based medium and slightly up-regulated in the four energy metabolism mutants (Fig. 6B, Table 4). More studies are needed to distinguish the possible cognate pairs within this group of genes. Phylogenetic analysis of these possible cognate HK-RR pairs suggested that HK and RR might not necessarily belong to the same phylogenetic subfamily; for example, HK DVU2395 and DVU 0331 were clustered in the putative Ntr subfamily, whereas RR DVU2577 and DVU2675 were clustered in the Nar subfamily. However, extra caution should be taken because the bootstrapping supports for RR trees are low.

Table 4 p-values of correlation of group E of HKs and RRs
Full size table

Discussion

SRB of the genus Desulfovibrio are universally distributed in various anoxic environmental niches such as soils, sediments, waters, and active microbial mats and flocs, where they can use organic acids as electron donors for sulfate reduction (Voordouw 1995; Heidelberg et al. 2004). Although much has been known about the biochemistry of metabolic pathways, little is known about their regulation. One well-established mechanism to conduct signal transduction in bacteria is TCSTSs (West and Stock 2001). Although TCSTSs have been previously suggested to be involved in facultatively anaerobic metabolism in E. coli (Stewart 1993) and in regulation of extracellular toxin production in anaerobic Clostridium perfringens (Lyristis et al. 1994), information is still scarce for their possible roles in anaerobic bacteria. In the recently completed D. vulgaris genome, more than 100 putative TCSTS components were revealed, which prompted us to further examine metabolic regulation in D. vulgaris. As the first step in this direction, we reported here the structural and phylogenetic analysis of all putative HKs and RRs in D. vulgaris and initial results of using transcriptomic analysis to identify possible cognate HK-RR gene pairs.

One of the interesting findings from structural analysis of putative HKs in D. vulgaris is the unexpected large number of hybrid-type HKs that contained both HK and receiver domains in one molecule. According to early study, hybrid-type HK is a less dominant subfamily in most microbial genomes, with an average number of only one to five per genome (Table 2) (Koretke et al. 2000). In contrast, eukaryotic cells generally possess more hybrid-type HKs that mediate multiple-step signal transduction; for example, Arabidopsis thaliana has 11 putative HKs, 9 of which are hybrid-type HKs (Urao et al. 2000; West and Stock 2001). The only exceptions to this relationship found so far are Synechocystis sp. PCC6803 (3.5Mb), Anabaena sp. PCC7120 (7.2Mb), and Pseudomonas aeruginosa PAO1 (6.2Mb) that contained 12, 55, and 16 hybrid-type HKs, respectively. However, it should be noted that filamentous, Anabaena sp. PCC7120 is capable of performing morphological differentiation and cell–cell communication similar to multicellular eukaryotes (Koretke et al. 2000; Wang et al. 2002), and Pseudomonas aeruginosa is one of the most important human pathogens (Rodrigue et al. 2000). Although data are lacking, it has been speculated that hybrid-type HKs found in bacteria could be involved in signal transduction required for cell–cell communication or differentiation. In support of this hypothesis, recent studies show that a hybrid-type HK (RcsC) and its RR (RcsB) were involved in the modulation of a characteristic behavior of E. coli cells during colony formation on the surface of agar plates (Takeda et al. 2001), and a hybrid-type HK (rpfF) and its cognate RR (rpfB) was involved in coupling the cell–cell signaling to the synthesis of pathogenicity factors in Xanthomonas campestris (Slater et al. 2000). The number of hybrid-type HKs identified in D. vulgaris was even higher than Synechocystis sp. PCC6803 and comparable with Anabaena sp. PCC7120 genomes, implying that multiple-step phosphorelay may be a common signal transduction mechanism in D. vulgaris. In addition, three HKs (DVU0681, DVU2677, and DVU3062) were found to contain more than one receiver domain, a phenomenon commonly observed in proteins involved in multiple-step phosphorelay systems (Hoch 2000; Wang et al. 2002). The regulatory systems and physiological functions these systems are associated with in nonphotosynthetic bacteria are of great interest; thus, D. vulgaris could be a model bacteria for the study of multiple-step phosphorelay systems.

In addition to various biochemical approaches such as two-hybrid or direct detection of phosphate transfer by labeling experiments (Wang et al. 2002), a number of computational techniques exist for predicting cognate HK-RR pairs including the similarity of phylogenetic trees method. The method has shown that interacting protein pairs show a greater degree of similarity (symmetry) in phylogenetic trees than noninteracting proteins do (Fryxell 1996; Valencia and Pazos 2002); thus, the concordance of HKs and RRs in corresponding phylogenetic subfamilies would be an important indication of cognate HKs and RRs. In this work, although initial finding of substantial high-order similarity between phylogenetic trees of HKs and RRs, with 19 of 26 D. vulgaris HK-RR pairs that are adjacent on the chromosome present in the corresponding phylogenetic subfamilies, suggested that the concordance of HKs and RRs in corresponding phylogenetic groups could be indicative of possible cognate HK-RR pairs, it was found that phylogenetic analysis alone is not enough to identify the specific cognate HK-RR pairs because most D. vulgaris TCSTSs narrowly distribute to only a few phylogenetic subfamilies.

An attempt was then made to identify the possible cognate HK-RR pairs in D. vulgaris based on the co-expression across selective experimental conditions. As noted in recent studies (Martin-Galiano et al. 2004), one of the difficulties in analyzing the expression pattern of TCSTSs is that most TCSTS genes are modestly expressed, and in most cases finding the right experimental conditions in which the TCSTS genes are differentially regulated can be difficult. Nevertheless, with the limited experimental conditions we tested in this study, 4 pairs of HK-RRs adjacent on the chromosome and 5 sets of HK-RRs not linked on the chromosome were found differentially expressed in the similar expression pattern, and were suggested as possible cognate HK-RR gene pairs. Although biochemical experiments will be needed to verify these predictions, our identification of a preliminary list of possible cognate TCSTS pairs constitutes a basis for further exploration of the physiological functions of TCSTSs in D. vulgaris.