Ancestral Genes Can Control the Ability of Horizontally Acquired Loci to Confer New Traits

Horizontally acquired genes typically function as autonomous units conferring new abilities when introduced into different species. However, we reasoned that proteins preexisting in an organism might constrain the functionality of a horizontally acquired gene product if it operates on an ancestral pathway. Here, we determine how the horizontally acquired pmrD gene product activates the ancestral PmrA/PmrB two-component system in Salmonella enterica but not in the closely related bacterium Escherichia coli. The Salmonella PmrD protein binds to the phosphorylated PmrA protein (PmrA-P), protecting it from dephosphorylation by the PmrB protein. This results in transcription of PmrA-dependent genes, including those conferring polymyxin B resistance. We now report that the E. coli PmrD protein can activate the PmrA/PmrB system in Salmonella even though it cannot do it in E. coli, suggesting that these two species differ in an additional component controlling PmrA-P levels. We establish that the E. coli PmrB displays higher phosphatase activity towards PmrA-P than the Salmonella PmrB, and we identified a PmrB subdomain responsible for this property. Replacement of the E. coli pmrB gene with the Salmonella homolog was sufficient to render E. coli resistant to polymyxin B under PmrD-inducing conditions. Our findings provide a singular example whereby quantitative differences in the biochemical activities of orthologous ancestral proteins dictate the ability of a horizontally acquired gene product to confer species-specific traits. And they suggest that horizontally acquired genes can potentiate selection at ancestral loci.


Introduction
Closely related bacterial species often exhibit significant differences in gene content [1]. Whereas these differences can arise from the duplication and divergence or the loss of ancestral genes [2][3][4], the vast majority of species-specific DNA sequences in bacteria appear to have arisen as a result of horizontal gene transfer events [1,[5][6][7]. Species-specific sequences often endow bacteria with new abilities, which enable them to access unexplored niches such as animal and plant tissues [1,8]. For example, some genomic islands from pathogenic bacteria function as freestanding virulence ''cassettes'', and thus, the pathogenic properties they specify can be ''played'' effectively when the genomic island is introduced into a different bacterial species, even a non-pathogenic one [9]. Still, other horizontally acquired genes operate on ancestral cellular pathways [10,11], raising questions as to the changes experienced both by the horizontally acquired DNA and the host bacterial genome that enable an organism to realize the fitness gains mediated by the acquired sequences.
This question can be addressed by first identifying a horizontally acquired gene product that behaves differently in present-day bacterial species, and by subsequently determining the changes in the horizontally acquired locus and/or associated ancestral genes that are responsible for a change in function. Such a strategy can provide insight into the evolutionary events that yielded extant diversity. Here, we provide a singular example of how an ancestral locus controls the ability of a horizontally acquired gene to confer a new ability.
The pmrD gene product enables Salmonella enterica serovar Typhimurium to express loci governing modifications of the lipopolysaccharide (LPS) [12], rendering the organism resistant to killing by the antibiotic polymyxin B ( Figure 1A) [13][14][15][16]. Transcription of these loci is controlled by the PmrA/PmrB two-component regulatory system [13,[17][18][19]. The PhoP-activated pmrD gene is expressed when Salmonella experience low levels of extracytoplasmic Mg 2+ , which is an inducing signal for the PhoP/ PhoQ two-component system [20]. The PmrD protein binds to the phosphorylated form of the DNA binding protein PmrA (PmrA-P), protecting it from dephosphorylation by the sensor PmrB; this enhances PmrA-P levels and promotes transcription of PmrA-dependent genes [21]. Thus, Salmonella can resist killing by polymyxin B not only in response to the Fe 3+ signal sensed by PmrB [22] but also in low Mg 2+ environments [12].
The closely related species Escherichia coli induces expression of PmrA-regulated genes and resistance to polymyxin B following growth in the presence of Fe 3+ ( Figure 1B) [14,16,23]. However, E.
coli does not activate the PmrA/PmrB system in low Mg 2+ conditions even though its pmrD gene is expressed in a PhoP/ PhoQ-dependent manner [23]. The inability of E. coli to activate the PmrA/PmrB system, and thus, to resist killing by polymyxin B in low Mg 2+ , has been attributed to its PmrD protein being highly divergent from the Salmonella PmrD [23]. Indeed, replacement of the E. coli pmrD gene by the Salmonella homolog enabled activation of the E. coli PmrA/PmrB system in low Mg 2+ [23].
We now show how an ancestral protein can control the ability of a horizontally acquired gene product to confer new traits. We present evidence suggesting that pmrD was acquired by lateral transfer. We demonstrate that the highly divergent E. coli pmrD gene can rescue a Salmonella pmrD mutant to activate the PmrA/ PmrB system during growth in low Mg 2+ . Beyond reinforcing the notion that the E. coli pmrD gene specifies a functional protein [24], our findings indicate that Salmonella and E. coli must differ in an additional component controlling the levels of PmrA-P. We establish that this component is PmrB as the E. coli PmrB exhibits significantly higher phosphatase activity than the Salmonella homolog, and that replacement of the E. coli pmrB gene with the Salmonella homolog is sufficient to restore transcription of PmrAactivated genes and polymyxin B resistance to E. coli grown in low Mg2+. Our findings provide a singular example whereby the ability of a horizontally acquired gene to confer a selective advantage upon a recipient organism is dictated by quantitative differences in the biochemical activities of ancestral proteins.

Evidence That pmrD Is a Horizontally Acquired Gene
Several lines of evidence support the previously suggested hypothesis that pmrD is a horizontally acquired gene [25]. First, a search of the completed genomes in the NCBI microbial genomes database for the presence of sequences homologous to the S. enterica 14028s or E. coli K12 MG1655 PmrD protein (conducted on May 2011) showed that pmrD is limited to enteric bacteria of the Klebsiella, Citrobacter, Salmonella and Escherichia lineages ( Figure S1A). Second, a BLAST search identified a pmrD-like gene encoded on plasmid pKP187 from the nitrogen-fixing endophyte K. pneumoniae 342 [26], which specifies a product displaying ,30% amino acid identity to the Salmonella and E. coli PmrD proteins ( Figure S1). The presence of pmrD on a mobile genetic element could provide a means for its transfer among bacteria. Third, phylogenetic analysis of pmrD homologs present in enteric bacteria revealed that the pmrD neighbor-joining tree was similar to the species phylogeny constructed using four housekeeping genes (gapA, groEL, gyrA and pgi). The only exception was the Citrobacter pmrD gene, which clustered with the homolog identified in K. pneumoniae 342 pKP187 ( Figure S1A) and this incongruence with the accepted species phylogeny is likely a consequence of horizontal gene transfer. Fourth, the pmrD gene is located in the same chromosomal context in the Citrobacter, Salmonella and Escherichia genomes, which is different from that of the Klebsiella genome ( Figure S1B). Collectively, these analyses suggest that the pmrD gene was horizontally acquired by one or several transfer events.
The E. coli PmrD Protein Connects the PhoP/PhoQ and PmrA/PmrB Systems in Salmonella But Not in E. coli To explore the possibility that the E. coli PmrD protein might activate the PmrA/PmrB system in Salmonella experiencing low In wild-type Salmonella, transcription of PmrA-activated genes is promoted during growth in low Mg 2+ via the PhoP/PhoQ system, the PmrD protein and the PmrA/PmrB system, and in the presence of Fe 3+ via the PmrA/PmrB system and independently of PhoP/PhoQ and PmrD. The PmrA protein represses transcription of the pmrD gene. (B) In wild-type E. coli, transcription of PmrA-activated genes is promoted in the presence of Fe 3+ via the PmrA/PmrB system. The PmrD protein is produced in low Mg 2+ in a PhoP-dependent manner but fails to activate the PmrA/PmrB system. doi:10.1371/journal.pgen.1002184.g001

Author Summary
The traits that distinguish closely related bacterial species are often ascribed to differences in gene content, which arise primarily through horizontal gene transfer. In some instances, the genes mediating a new trait act as independent entities that function in a variety of organisms. However, the ability of a horizontally acquired gene product(s) to operate on an ancestral pathway might be constrained by subtle differences between orthologous ancestral proteins. Here, we examine why the horizontally acquired pmrD gene product post-translationally activates the ancestral PmrA/PmrB two-component system in Salmonella enterica but not in the closely related species Escherichia coli. This allows Salmonella, but not E. coli, to transcribe PmrA-activated genes including those conferring antibiotic resistance when grown in low Mg 2+ , which is a condition that promotes PmrD expression. We now demonstrate that, paradoxically, the E. coli PmrD protein activates the PmrA/PmrB system in Salmonella even though it fails to do so in E. coli. We establish that quantitative differences in the biochemical activities of the PmrB proteins from Salmonella and E. coli dictate the functionality of PmrD, which protects phosphorylated PmrA from PmrB's phosphatase activity. Our findings demonstrate that ancestral genes can control the ability of horizontally acquired genes to confer species-specific traits upon different organisms.
Mg 2+ (even though it does not function in this capacity in E. coli [23]), we engineered a Salmonella strain that expresses a FLAGtagged E. coli pmrD gene under the control of the Salmonella pmrD promoter in the normal chromosomal location. The resulting strain transcribed the PmrA-activated pbgP gene in low Mg 2+ , like an isogenic strain expressing Salmonella's own pmrD gene or a FLAG-tagged version of it ( Figure 2). As expected, all three strains expressed pbgP to high levels when grown in the presence of Fe 3+ , which activates the PmrA/PmrB system in a PmrD-independent manner [22], but not under non-inducing (high Mg 2+ and no Fe 3+ ) conditions ( Figure 2). Moreover, a Salmonella pmrD mutant harboring a plasmid encoding the E. coli pmrD gene or a FLAGtagged version of it transcribed pbgP in low Mg 2+ , unlike cells harboring the plasmid vector ( Figure S2). These results indicate that the ability of the E. coli PmrD protein to connect the PhoP/ PhoQ and PmrA/PmrB systems is context-dependent: it functions in Salmonella, but surprisingly, not in E. coli.
The E. coli PmrB Protein Exhibits Higher Phosphatase Activity towards PmrA-P Than the Salmonella PmrB Protein Why does expression of the E. coli pmrD gene result in pbgP transcription in Salmonella but not in E. coli during growth in low Mg 2+ ? When bacteria experience inducing conditions for the ancestral PmrA/PmrB system, the sensor PmrB promotes the phosphorylated state of the DNA binding protein PmrA [22,27], resulting in transcription of PmrA-activated genes [28]. In addition to being an autokinase and phosphotransferase, PmrB is a PmrA-P phosphatase, an activity present primarily under noninducing conditions [29]. The PmrD protein, which is produced in low Mg 2+ , promotes the phosphorylated state of PmrA by antagonizing PmrB-mediated dephosphorylation of PmrA-P [21]. Thus, we hypothesized that the orthologous PmrA and/or PmrB proteins of these two species might differ in one or more of their biochemical properties, resulting in lower levels of PmrA-Pthe target of PmrD -in E. coli than in Salmonella. To evaluate this possibility, we compared the behaviors of the purified cytoplasmic domain of the PmrB protein (PmrB c ) from the two species because it retains all the known enzymatic activities of the full-length PmrB protein, and because PmrB c expression recapitulates the PmrDdependent activation of PmrA-regulated genes displayed by the full-length PmrB protein in vivo ( Figure S3) [21].
First, we examined the autophosphorylation rate of the two proteins and found that the ATP concentration at which it was half maximal was two-fold higher for the E. coli PmrB c protein than for the Salmonella PmrB c protein ( Figure S4A; see Table 1 for K M values). In addition, the E. coli PmrB c protein autophosphorylated at a rate that was two-fold higher than the Salmonella PmrB c protein ( Figure S4B; see Table 1 for k observed values). Given that the intracellular ATP concentration is ,3 mM in E. coli [30], which is .30 times the experimentally determined K M for ATP of the Salmonella and E. coli PmrB c proteins (Table 1), we anticipate that these PmrB orthologs autophosphorylate at similar rates in vivo.
We next analyzed phosphotransfer from the Salmonella PmrB c protein to the Salmonella PmrA protein. The amount of PmrA-P increased steadily, reaching a maximum at 30 min ( Figure 3A and 3C). The increase in the levels of PmrA-P was accompanied by a decrease in the levels of PmrB c -P ( Figure 3A). By contrast, when the phosphotransfer reaction was performed with the PmrB c -P and PmrA proteins from E. coli, there was a rapid increase in the levels of PmrA-P during the first 5 min, followed by a steady decrease ( Figure 3B and 3C). The decrease in PmrA-P levels was not due to a contaminating protease because the total PmrA amounts remained constant throughout the course of the reaction ( Figure  S4C). The different rates of dephosphorylation of the orthologous PmrA-P proteins after reaching their maximal values reflect the source of the PmrB c proteins (as opposed to resulting from differences in the rate of spontaneous hydrolysis of the phosphoryl group from the PmrA-P proteins) because the stabilities of the Salmonella and E. coli PmrA-P proteins were similar when the PmrB c proteins were omitted from the reactions ( Figure S4D). Cumulatively, these experiments suggested that the Salmonella and E. coli PmrB c proteins differ in their phosphatase activity towards PmrA-P.
We established that the rate with which PmrB c dephosphorylates PmrA-P is 10-fold higher for the reaction with the E. coli PmrB c and PmrA-P proteins than that displayed by the Salmonella PmrB c and PmrA-P proteins ( Figure S4E and Table 2). Furthermore, the different rates of PmrB c -promoted dephosphorylation of PmrA-P are largely dependent on the source of PmrB c (rather than on PmrA): the E. coli PmrB c dephosphorylated the PmrA-P proteins from Salmonella and E. coli at rates that were 3-10-fold higher than those with which the Salmonella PmrB c protein dephosphorylated PmrA-P from Salmonella or from E. coli ( Figure  S4E and Table 2). In sum, the enhanced phosphatase activity of the E. coli PmrB c towards PmrA-P ( Figure 3 and Table 2) is predicted to generate lower levels of PmrA-P in E. coli than in Salmonella, which would result in reduced transcription of PmrAactivated genes [31].  To explore the molecular basis for the differences in phosphatase activity exhibited by the Salmonella and E. coli PmrB proteins, we carried out homology modeling of their cytoplasmic domains using the structure of the sensor kinase HK853 from Thermotoga maritima [32]. This allowed us to predict the secondary structural features of the PmrB homologs and to identify subdomains that differed between the two proteins ( Figure S5). Next, we created chimeric PmrB c proteins whereby particular subdomains of the Salmonella PmrB c protein were replaced by amino acid sequences corresponding to those from the E. coli PmrB c protein, aiming to preserve the secondary structure of PmrB c . We then examined the ability of the chimeric proteins to dephosphorylate Salmonella PmrA-P.
The PmrB c -a2 chimera, which differs from the wild-type Salmonella PmrB c sequence in five residues located in alpha helix 2 ( Figure S5), dephosphorylated Salmonella PmrA-P at a higher rate than the Salmonella PmrB c protein, though not as high as the E. coli PmrB c protein ( Figure 4). The PmrB c -a2 chimera displayed a phosphotransfer profile that was similar to that of the E. coli PmrB protein: when phosphorylated PmrB c -a2 protein was incubated with Salmonella PmrA protein, the levels of PmrA-P increased between 0-5 min and then decreased ( Figure S6). By contrast, five other chimeric PmrB c proteins exhibited similar or lower levels of phosphatase activity towards Salmonella PmrA-P compared to the Salmonella PmrB c protein ( Figure S7). As expected, an E. coli PmrB c -a2 chimeric protein harboring the Salmonella PmrB a2 subdomain dephosphorylated E. coli PmrA-P at a lower rate than the E. coli PmrB c ( Figure S8). Our results indicate that residues in alpha helix 2 are critical modulators of PmrB's phosphatase activity.   We hypothesized that the heightened phosphatase activity of the E. coli PmrB protein may hinder accumulation of high enough levels of PmrA-P, which constitutes the target of the PmrD protein [21,23], thereby preventing transcription of PmrA-activated genes and resistance to polymyxin B during growth in low Mg 2+ . To test this hypothesis, we engineered isogenic E. coli strains expressing either the Salmonella or E. coli pmrB genes from the normal E. coli chromosomal location. We then investigated the PmrD-mediated connectivity between the PhoP/PhoQ and PmrA/PmrB systems by determining the ratio of the mRNA levels corresponding to the PmrA-activated genes pbgP, pmrC and pmrG produced during growth in low Mg 2+ and those produced following growth in the presence of Fe 3+ [23], which are PmrD-dependent andindependent processes in Salmonella.
PmrD connectivity was higher in the E. coli strain expressing the Salmonella pmrB gene than in the isogenic strain expressing E. coli's own pmrB gene (Figure 5A, 5B; Figure S9A and S9B). An E. coli strain harboring the Salmonella pmrB-a2 gene also exhibited higher PmrD connectivity than the strain expressing the E. coli pmrB gene, though not to the level displayed by the strain with the Salmonella homolog ( Figure S9B). As expected, these strains displayed a similar ratio (,1) for the PmrA-independent rpoD gene ( Figure 5C and 5D), which was used as a control. The low PmrD connectivity in the E. coli strain harboring its own pmrB gene is ascribed to its higher phosphatase activity because an E. coli strain expressing the PmrB T156R protein, which dephosphorylates PmrA-P at a lower rate than the E. coli PmrB protein in vitro ( Figure S4F), transcribes PmrA-activated genes in low Mg 2+ ( Figure S9C). The disparate PmrD connectivities were not due to differences in the responses of the orthologous PmrB proteins to Fe 3+ because these strains displayed similar levels of iron-promoted pbgP and pmrC transcription ( Figure S9D). The heightened expression displayed by the engineered E. coli strain harboring the Salmonella pmrB gene required the PmrD protein because there was no significant transcription of PmrA-activated genes in a pmrD mutant following growth in low Mg 2+ ( Figure S9E).
E. coli displays resistance to polymyxin B when grown in the presence of Fe 3+ (but not following growth in low Mg 2+ ) in a process that requires the pmrA and pbgP genes [23]. We reasoned that the E. coli strain expressing the Salmonella pmrB gene should display resistance towards polymyxin B when experiencing low Mg 2+ because PmrA-activated genes are transcribed ( Figure 5A). Indeed, the E. coli strain expressing the Salmonella pmrB gene was ,4-fold more resistant to polymyxin B than the strain expressing the E. coli pmrB gene when bacteria were grown in low Mg 2+ ( Figure 5E and 5F). As expected, both strains displayed similar levels of polymyxin B resistance following growth in the presence of Fe 3+ (Figure 5E and 5F). Collectively, these findings demonstrate that the source of the pmrB gene (i.e., Salmonella or E. coli) determines the ability of E. coli to transcribe PmrA-activated genes and to exhibit polymyxin B resistance in response to low Mg 2+ .

PmrB Determines the Expression Kinetics of PmrA-Activated Genes
Fe 3+ activates the PmrA/PmrB system independently of the PmrD protein both in Salmonella and E. coli [22,23]. In Salmonella, it has been demonstrated that such activation gives rise to a surge in the mRNA levels of PmrA-activated genes whereby the mRNAs increase, peak and then decrease to reach new steady-state levels [27], and that this reflects changes in the amount of PmrA-P protein [27]. Because the E. coli PmrB protein displays higher phosphatase activity than the Salmonella homolog (Table 2), we wondered whether the levels of PmrA-P, and thus the kinetics of PmrA-dependent gene expression, might differ temporally in E. coli when it expresses the Salmonella (instead of the E. coli) pmrB gene.
When bacteria were shifted to media containing Fe 3+ , transcription of PmrA-activated genes increased during the first 30 min and then decreased to reach steady-state levels in the E. coli strain harboring its own pmrB gene ( Figure 6). By contrast, in the E. coli strain expressing the Salmonella pmrB gene, the mRNA levels of PmrA-activated genes reached a peak at 45 min followed by a decrease to attain steady-state levels ( Figure 6). To determine whether the distinct temporal transciption profiles of PmrAactivated genes in these strains were due to differences in the rates of PmrB-mediated PmrA-P dephosphorylation, we examined mRNA levels in an E. coli strain expressing the E. coli PmrB T156R protein, which is compromised in its phosphatase activity ( Figure S4F). The mRNA levels of PmrA-activated genes were high even when bacteria were grown high Mg 2+ and no Fe 3+ , consistent with the notion that PmrB functions primarily as a PmrA-P phosphatase under these conditions ( Figure 6) [21]. When bacteria were shifted to media containing Fe 3+ , transcription of PmrA-activated genes increased and peaked at 45 min before decreasing to reach steady-state levels ( Figure 6). Altogether, our findings suggest that the higher phosphatase activity displayed by the E. coli PmrB protein results in PmrA-dependent transcription being turned off at an earlier timepoint in the E. coli strain expressing its own pmrB gene than in the isogenic strains harboring the Salmonella homolog or the E. coli pmrB T156R gene.

Ancestral Genes Can Dictate the Functionality of Horizontally Acquired Genes
Horizontal gene transfer can endow bacteria with new capabilities, such as the ability to resist killing by an antibiotic or to gain access to a eukaryotic cell, and thus, it can potentially expand their ecological niches [1,[5][6][7]. We have now established that variation in the amino acid sequences of the Salmonella and E. coli PmrB sensors results in quantitative differences in their biochemical activities ( Figure 3 and Table 2) and governs the ability of these bacteria to survive killing by polymyxin B in low Mg 2+ ( Figure 5). We proposed that, by generating different amounts of PmrA-P, the different phosphatase activities of the PmrB orthologs determine the ability of PmrD to activate the PmrA/PmrB system (Figure 2 and Figure 5) [23]. Therefore, unlike self-contained horizontally acquired gene clusters that readily confer a new trait upon introgression [1,8,9], the functionality of a horizontally acquired gene product operating on ancestral pathways depends largely upon the biochemical properties of existing ancestral proteins, which can act in a variety of ways [33][34][35].
The effect of genetic context on protein function is not limited to horizontally acquired genes in bacteria. A recent genome-wide survey of genes that were essential in one budding yeast strain but not in another strain demonstrated that this essential character was due to distinct genetic interactions, which differed in a strainspecific manner [36]. In higher eukaryotes, the development of complex traits involves multiple interacting genetic loci; thus, the effect of a certain gene may require other genes controlling the same trait [37]. Taken together, the contribution of a given gene to a particular phenotype is largely determined by its interactions with other genes functioning in the same genetic network.

Particular Combinations of Gene Interactions Underlie Qualitatively Similar Traits
Given that the heightened phosphatase activity of the E. coli PmrB protein prevents the E. coli PmrD protein from promoting the accumulation of sufficient levels of PmrA-P (Figure 3 and Figure 5, Table 2), why is it that expression of the Salmonella pmrD gene allows transcription of PmrA-activated genes in E. coli experiencing low Mg 2+ [23]? Sequence analysis of the pmrD genes suggested that selection and divergence between the Salmonella and E. coli pmrD loci [23] allowed the Salmonella PmrD, but not the E. coli ortholog, to connect the PhoP/PhoQ and PmrA/PmrB systems in E. coli. These results indicate that particular combinations of pmrD and pmrB alleles dictate whether bacteria experiencing low Mg 2+ accumulate sufficient levels of PmrA-P to promote expression of PmrA-dependent genes and to resist killing by polymyxin B (Figure 2 and Figure 5) [23]. Our work, together with studies of gene interactions in model organisms such as budding yeast [36] and analyses of the existing phenotypic variation within natural populations [38], demonstrate that different combinations of particular alleles within a genetic network can lead to qualitatively similar phenotypes in related organisms [37]. Therefore, understanding the phenotypic variation that exists within or between species demands going beyond the study of single-gene effects to investigating how genetic variation at different loci influence one another.

Quantitative Differences in the Phosphatase Activity of Sensor Proteins Determine Expression Kinetics
Organisms respond to a stress condition by altering their gene expression programs and/or behaviors with temporal dynamics specified by their signaling systems [27,39,40]. The distinct dynamics that distinguish the responses of closely related organisms are typically ascribed to differences in the genetic architecture of their signaling systems [41]. For example, the enteric species Yersinia pestis, K. pneumoniae and S. enterica utilize different genetic architectures to promote transcription of the pbgP gene when experiencing low Mg 2+ , resulting in distinct expression profiles [25,42,43]. We now provide a singular example where variation in the biochemical activity of orthologous signaling proteins can affect expression dynamics in genetic systems with identical architectures. The identified differences in phosphatase activities exhibited by the Salmonella and E. coli PmrB proteins lead to dissimilar kinetics of PmrA-dependent gene expression ( Figure 6, Table 2), which could affect resistance to Fe 3+ [44] and contribute to the distinct ability of these two species to survive in soil [45].
The Salmonella and E. coli PmrB proteins are the first example, to our knowledge, where amino acid changes between orthologous bifunctional sensor kinases bring about dramatic differences in phosphatase activity ( Table 2) without much effect on autokinase activity (Table 1). These two enzymatic activities are carried out by the cytoplasmic portion of these proteins, which consists of the dimerization and histidine phosphotransfer (DHp) domain joined by a flexible linker to the catalytic and ATPbinding (CA) domain [32,46,47]. Amino acid residues identified in the DHp [48][49][50][51] and CA domains [52] of other sensor kinases as being important for phosphatase activity are conserved between the Salmonella and E. coli PmrB proteins. This was expected because mutation of these residues usually abolishes phosphatase activity and both the Salmonella and E. coli PmrB proteins exhibit phosphatase activity towards PmrA-P, albeit to different levels ( Table 2). One or all the five amino acids residues differing between the orthologous PmrB proteins in the alpha-2 helix ( Figure S5) modulate phosphatase activity (Figure 4 and Figure S8). Some of these amino acids lie within a weakly conserved region (termed X-region) of the DHp domain and linker region that is proposed to stimulate sensor protein phosphatase activity by properly orienting the DHp and CA domains towards each other ( Figure S5) [49,51]. Thus, one or more of these five amino acids might also modulate contacts between the DHp and CA domains of the PmrB proteins.

Evolution of an Ancestral Regulatory System Targeted by a Horizontally Acquired Gene Product
We have established that quantitative differences between the biochemical properties of the orthologous PmrB proteins control the functionality of the pmrD gene product in present-day Salmonella and E. coli. Yet, when did these genetic events (i.e. horizontal acquisition of pmrD and mutational changes in pmrB) occur in the evolutionary history of enteric species? And how might the integration of horizontally acquired genes into ancestral regulatory circuits drive their evolution and facilitate bacterial adaptation to particular environments?
The pmrD gene appears to have been acquired by the common ancestor to the enteric species K. pneumoniae, Citrobacter spp., S. enterica and E. coli and incorporated into the regulatory circuit of the ancestral PhoP/PhoQ system because it is present in these four species but absent from all other examined enteric bacteria ( Figure  S1) [25]. This ancestral organism likely transcribed PmrAactivated genes in low Mg 2+ via the PmrD-mediated pathway as this ability is retained in Klebsiella and Salmonella [12,25]. After the Salmonella and E. coli lineages diverged, the ability to make low Mg 2+ -dependent, PmrA-controlled modifications of the LPS was retained by Salmonella but lost in E. coli, perhaps contributing towards their survival and proliferation in distinct host and nonhost environments [45]. This difference can be attributed, in part, to selection at the pmrD locus and divergence of the Salmonella and E. coli PmrD orthologs because expression of the Salmonella pmrD gene allows transcription of PmrA-activated genes under low Mg 2+ -inducing conditions in E. coli [23]. We further propose that selection for a Salmonella PmrB that has low phosphatase activity may have served to maintain the PmrD-mediated connection between the PhoP/PhoQ and PmrA/PmrB systems (Table 2), thus enabling Salmonella to display the LPS modifications, including those resulting in polymyxin B resistance in environments denoted by low Mg 2+ [12,21]. By contrast, the accumulation of PmrA-P in low Mg 2+ may be detrimental to E. coli's lifestyle because a mutant that constitutively activates the PmrA/PmrB system exhibits increased susceptibility to the bile detergent deoxycholic acid [53], which E. coli may experience in the mammalian intestine [54]. This may have driven selection for a PmrB protein with a higher phosphatase activity in E. coli (Table 2). Indeed, molecular analysis of the pmrB gene from natural isolates corresponding to the Salmonella reference collection C (SARC) [55] and natural isolates from the E. coli standard reference collection (ECOR) [56] revealed that the pmrB gene is evolving in a non-neutral fashion (Table S1).
In Salmonella, further integration of the pmrD gene into the regulatory circuits of the ancestral PhoP/PhoQ and PmrA/PmrB systems has entailed the evolution of a negative feedback loop where PmrA-P represses PhoP-promoted pmrD transcription, thus preventing excessive expression of PmrA-activated genes ( Figure 1A) [29]. This feedback loop is absent in E. coli because the PmrD protein does not activate the PmrA/PmrB system in this species following growth in low Mg 2+ ( Figure 1B) [23]. Our work, and that of others [57], suggests that changes in ancestral regulatory and metabolic networks arising from horizontal gene transfer contribute to the survival of bacterial species in their particular niches.

Concluding Remarks
In sum, genetic changes are often contingent upon those previously fixed within an evolutionary lineage [58]. For example, evolution of the ability to metabolize citrate in an experimental population of E. coli was shown to require the occurrence of earlier mutations, which potentiated the appearance of the latter mutation giving rise to the novel trait [59]. Our findings now suggest that horizontal acquisition of the pmrD gene might have potentiated selection at the Salmonella and E. coli pmrB loci, further cementing the ability of Salmonella, but not E. coli, to express PmrAactivated genes and to resist killing by polymyxin B in the presence of low Mg 2+ .
Finally, the behavior of an organism, or its survival in particular ecological niches, is often deduced from the presence or absence of genes and/or the predicted functions of the encoded gene products, which are based on experiments typically carried out with model organisms. However, more complex mechanisms could link an organism's genotype to its phenotype [60]. We have now established that quantitative differences in the biochemical activities of an ancestral protein can determine whether a horizontally acquired gene product connects different regulatory systems, thereby affecting the ability of a bacterial species to survive killing by cationic antimicrobial peptides. Consequently, subtle quantitative differences in the biochemical activities of conserved orthologous proteins, which are not easily predicted by sequence comparison and identification of conserved catalytic residues, can give rise to phenotypic variation between and within closely related species [38,61]. Our work demonstrates that protein function can only be appreciated in the context of other proteins operating within the same genetic network, and underscores the need to experimentally verify predictions derived from comparative genomics.

Materials and Methods
Bacterial Strains, Plasmids, and Growth Conditions S. enterica serovar Typhimurium strains were derived from the wild-type strain 14028s and E. coli strains from the wild-type strain MG1655. Bacteria were grown at 37uC with aeration in Luria-Bertani (LB) broth or in N-minimal media (pH 7.7), supplemented with 0.1% casamino acids, 38 mM glycerol, 10 mM or 10 mM MgCl 2 and 100 mM FeSO 4 [62]. When necessary, antibiotics were added at the following final concentrations: ampicillin, 50 mg/ml; chloramphenicol, 20 mg/ml; kanamycin, 50 mg/ml, and tetracycline, 10 mg/ml. P1 transduction of E. coli strains was performed as described [63]. P22 transduction of Salmonella strains was performed as described [64]. E. coli DH5a was used as a host for the preparation of plasmid DNA. We describe the construction of bacterial strains and plasmids as well as other molecular biology procedures in the Text S1. Bacterial strains and plasmids used in this study are listed in Table S2 and S3, respectively. b-Galactosidase Assay b-galactosidase assays were carried out in triplicate, and the activity was determined as described [63]. Data correspond to mean values of three independent experiments performed in duplicate.

Autokinase, Phosphotransferase, and Phosphatase Assays
Biochemical assays were carried out as in [21] and a detailed description is provided in Text S1.

RNA Isolation and Real-Time PCR to Determine Transcript Levels
E. coli cells from overnight cultures grown in N-minimal medium at pH 7.7 with 10 mM MgCl 2 were washed twice with N-minimal medium containing no Mg 2+ and added into the appropriate fresh media with 1:50 dilution. The bacterial cultures were grown to OD 600 0.5 in a shaking water bath at 37uC. 0.5 ml aliquots of cells were removed, mixed with RNAprotect Bacteria Reagent (Qiagen) for stabilization of RNA, and total RNA was isolated using RNeasy Kit (Qiagen) with on-column DNase treatment. cDNA was synthesized using TaqMan (Applied Biosystems) and random hexamers. Quantification of transcripts was carried out by real-time PCR using SYBR Green PCR Master Mix (Applied Biosystems) in an ABI 7500 Sequence Detection System (Applied Biosystems). The relative amount of cDNA was determined using a standard curve obtained from PCR with serially diluted genomic DNA, and results were normalized to the levels of 16S ribosomal RNA. Data shown is an average from at least three independent experiments.

Polymyxin B Susceptibility Assay
Polymyxin B susceptibility assays were carried out as described in [23]. Data correspond to mean values of three independent experiments performed in duplicate.

Homology Searches and Phylogenetic Analyses of the pmrD Gene
To identify homologs of the pmrD gene, the deduced PmrD protein sequences from E. coli K12 MG1655 and S. enterica serovar Typhimurium 14028s were used as query to search the NCBI microbial genomes database as well as the NCBI non-redundant nucleotide database using TBLASTN. In parallel, the UniProt database was searched for homologs using BLASTP [65]. Similar results were obtained in these searches and the protein sequences of the homologs were used for phylogenetic analyses. The sequences of the pmrD gene were aligned using ClustalX2 [66] and phylogenetic trees were constructed using the neighborjoining algorithm with default parameters and 1000 bootstrap replicates using MEGA 4 [67]. For comparison, the sequences of 4 housekeeping genes (gapA, groEL, gyrA and pgi) [68] were concatenated, aligned using ClustalX2 [66] and used for construction of a species tree.

Accession Numbers
GenBank accession numbers are available in Table S4. Transcript levels of the PmrA-activated pbgP gene were determined in a Salmonella pmrD strain (EG11491) harboring plasmids pUH-pmrD E. coli -FLAG, pUH-pmrD E. coli or the vector pUHE21-2lacI q . (B) The ratio between the pbgP mRNA levels produced during growth in low Mg 2+ and those produced following growth in the presence of Fe 3+ was determined in a Salmonella pmrD strain (EG11491) harboring plasmids pUH-pmrD E. coli -FLAG, pUH-pmrD E. coli or the vector pUHE21-2lacI q . Bacteria were grown with 0.5 mM IPTG in N-minimal medium at pH 7.7 containing 10 mM Mg 2+ (HL) or 10 mM Mg 2+ and 100 mM Fe 3+ (HL+Fe 3+ ). Data correspond to the mean of three independent experiments and error bars show standard deviation. (TIF) Figure S3 The E. coli PmrB cytoplasmic domain (PmrB c ) is sufficient for PmrD-dependent transcription of PmrA-activated genes. b-galactosidase activity (Miller units) from a chromosomal pbgP-lac transcriptional fusion were determined in DpmrB (EG10065) and DpmrB DpmrD (EG12060) Salmonella strains harboring pUH-pmrB, pUH-pmrB c or pUHE21-2lacI q . Bacteria were grown in N-minimal medium at pH 7.7, 0.05 mM IPTG and 50 mg/ml ampicillin with 10 mM Mg 2+ (HL) or 10 mM Mg 2+ (HH). Data correspond to the mean of three independent experiments performed in duplicate and error bars show standard deviation. (TIF) Figure S4 The total levels of PmrB c and PmrA protein remain constant throughout the course of the phosphotransfer reaction and the PmrA-P proteins display similar stabilities in the absence of PmrB c . (A) Levels of Salmonella or E. coli PmrB c -P following incubation of PmrB c (2.5 mM) with ATP at the concentrations indicated at the top of the figure according to the protocols described in Text S1. (B) Levels of Salmonella or E. coli PmrB c -P following incubation of PmrB c (2.5 mM) with 0.75 mM ATP at the times indicated at the top of the figure according to the protocols described in Text S1. (C) Levels of the PmrB c and PmrA proteins from Salmonella and E. coli were determined by SDS-PAGE and Coomassie staining following phosphotransfer from PmrB c -P (5 mM) to PmrA (10 mM) as described in Text S1 using aliquots taken at the indicated times. (D) Levels of PmrA-P proteins from Salmonella and E. coli were determined as described in Text S1 with aliquots taken at the indicated time points. (E) Levels of PmrA-P following incubation of Salmonella or E. coli PmrB c (5 mM) with the Salmonella or E. coli PmrA-P (2.5 mM) proteins for the indicated times according to the protocols described in Text S1. (F) Levels of PmrA-P following incubation of E. coli PmrB c T156R (5 mM) with the E. coli PmrA-P (2.5 mM) protein for the indicated times according to the protocols described in Text S1. (TIF) Figure S5 Primary sequence alignment of the Salmonella and E. coli PmrB sensor proteins. Residues that differ between the two proteins are highlighted in green (conserved changes) or red (nonconserved changes). A blue vertical dotted line is present before the first amino acid of the PmrB c cytoplasmic domain. The structures of the cytoplasmic domains of the PmrB proteins were predicted using a homology-modeling program (Phyre) [69] and were based on HK853 sensor histidine kinase cytoplasmic domain from Thermotoga maritima [32]. The dimerization and histidine phosphotransfer (DHp) domain consists of amino acids 128-200, while the catalytic and ATP-binding (CA) domain consists of amino acids 215-356 and 215-363 from the Salmonella and E. coli PmrB proteins, respectively. b sheets and a helices in the Salmonella and E. coli PmrB c proteins that contain non-identical amino acid residues are highlighted by colored boxes, and those that are identical between these proteins are highlighted by grey boxes. The X-region (amino acids 179-202), which is underlined with a black dashed line, was predicted based on sequence alignment to the sensor kinase EnvZ [49,51]. (TIF) Figure S6 Phosphotransfer from the PmrB c -a2 chimera to the Salmonella PmrA protein is similar to that displayed by the E. coli  Figure S9 The Salmonella pmrB gene restores low Mg 2+ -promoted transcription of PmrA-activated genes in E. coli. (A-C) The ratio between the mRNA levels of the PmrA-activated genes pmrG, pmrC and pbgP produced during growth in low Mg 2+ and those produced following growth in the presence of Fe 3+ was determined in E. coli expressing the Salmonella pmrB gene (DC3) or its own pmrB gene (DC5) or the Salmonella pmrB-a2 gene (DC68) or the E. coli pmrB T156R gene (DC116) from the normal chromosomal location. RNA samples were prepared from bacteria grown in N-minimal media, pH 7.7 with 10 mM Mg 2+ (HL) or 10 mM Mg 2+ and 100 mM Fe 3+ (HL+Fe 3+ ) to OD 600 of 0.5, and the levels of pmrG, pmrC or pbgP mRNA were determined by reversetranscription-qPCR analysis. (D) Transcript levels of the PmrAactivated genes pbgP and pmrC were determined in E. coli strains expressing either the Salmonella pmrB gene (DC3) or the E. coli pmrB gene (DC5). Bacteria were grown in N-minimal media, pH 7.7 with 1 mM Mg 2+ (HH) or 1 mM Mg 2+ and 100 mM Fe 3+ (HH+Fe 3+ ) to an OD 600 0.5 before RNA samples were prepared for reverse-transcription-qPCR analysis. (E) Transcript levels of the PmrA-activated genes pbgP and pmrC were determined in E. coli strains deleted for the pmrD gene and expressing either the Salmonella pmrB gene (DC9) or the E. coli pmrB gene (DC11). Bacteria were grown in N-minimal media, pH 7.7 with 10 mM Mg 2+ (HL) or 10 mM Mg 2+ and 100 mM Fe 3+ (HL+Fe 3+ ) to an OD 600 0.5 before RNA samples were prepared for reversetranscription-qPCR analysis. Data correspond to the mean of at least three independent experiments and error bars represent standard deviation. (TIF)