Phenotypic differences between Salmonella and Escherichia coli resulting from the disparate regulation of homologous genes

Abstract

Phenotypic differences among closely related bacteria have been largely ascribed to species-specific genes, such as those residing in pathogenicity islands. However, we now report that the differential regulation of homologous genes is the mechanism responsible for the divergence of the enteric bacteria Salmonella enterica and Escherichia coli in their ability to make LPS modifications mediating resistance to the antibiotic polymyxin B. In S. enterica serovar Typhimurium, the PmrA/PmrB two-component system governing polymyxin B resistance is induced in low Mg²⁺ in a process that requires the PmrD protein and by Fe³⁺ in a PmrD-independent fashion. We establish that E. coli K-12 induces PmrA-activated gene transcription and polymyxin B resistance in response to Fe³⁺, but that it is blind to the low Mg²⁺ signal. The highly divergent PmrD protein is responsible for this phenotype as replacement of the E. coli pmrD gene by its Salmonella counterpart resulted in an E. coli strain that transcribed PmrA-activated genes and displayed polymyxin B resistance under the same conditions as Salmonella. Molecular analysis of natural isolates of E. coli and Salmonella revealed that the PmrD proteins are conserved within each genus and that selection might have driven the divergence between the Salmonella and E. coli PmrD proteins. Investigation of PmrD function demonstrated statistically different distributions for the Salmonella and E. coli isolates in PmrD-dependent transcription occurring in low Mg²⁺. Our results suggest that the differential regulation of conserved genes may have ecological consequences, determining the range of niches a microorganism can occupy.

The presence or absence of DNA sequences has become the paradigm to explain the genetic basis for the intrinsic characteristics of a bacterial species. Indeed, analysis of the enteric bacteria Escherichia coli K-12 and Salmonella enterica serovar Typhimurium reveals that E. coli has >800 genes absent from the Salmonella genome, and that >1,100 Salmonella genes lack counterparts in E. coli (1). These include the lac operon, which allows E. coli to grow on lactose as a sole carbon source, and several pathogenicity islands that endow Salmonella with the ability to invade epithelial cells and to survive within phagocytic cells (2–5). However, other genetic mechanisms, such as the differential regulation of orthologous genes or allelic differences in orthologous sequences, could be responsible for phenotypic traits that differentiate closely related bacterial species but have remained largely unexplored.

The PmrA/PmrB two-component system of S. enterica is required for virulence in mice (6), colonization of chicken macrophages (7), survival in soil (8), and resistance to Fe³⁺ (9) and to the antibiotic polymyxin B (10). The best-characterized PmrA-activated targets are those mediating modifications of the LPS necessary for resistance to polymyxin B. These include the pbgP [also called pmrHFIJKLM (11) and arn (12–14)] operon and ugd gene (15), which are responsible for the synthesis and incorporation of 4-aminoarabinose into the LPS (11, 14, 16), and the pmrC gene, which encodes a putative aminotransferase necessary for decoration of the LPS with phosphoethanolamine (17). Although additional PmrA-regulated genes have been identified (18, 19), their biochemical activities and the role(s) they play in Salmonella's lifestyle remain unknown.

The PmrA/PmrB system is unique in that transcription of PmrA-activated genes is induced not only by Fe³⁺, which is the specific signal recognized by the sensor PmrB (9), but also by low Mg²⁺, which is detected by the noncognate sensor PhoQ (20). The low Mg²⁺ activation also requires PhoQ's cognate regulator PhoP, and the PmrA and PmrB proteins (21), as well as the PhoP-activated PmrD protein (22) (Fig. 1A). Thus, Salmonella can express PmrA-activated genes and display resistance to polymyxin B in response to Fe³⁺ and/or low Mg²⁺ signals (22).

Fig. 1.

Model illustrating the stimuli that activate the PhoP/PhoQ and PmrA/PmrB two-component systems in Salmonella and E. coli. (A) In Salmonella, transcription of PmrA-activated genes is promoted during growth in low Mg²⁺ via the PhoP/PhoQ system, the PmrD protein, and the PmrA/PmrB system; and in the presence of Fe³⁺ via the PmrA/PmrB system, independently of PhoP/PhoQ and PmrD. The PmrA protein represses transcription of the pmrD gene. (B) In E. coli, transcription of PmrA-activated genes is promoted in the presence of Fe³⁺ via the PmrA/PmrB system. The PmrD protein is produced in low Mg²⁺ in a PhoP-dependent manner, but it fails to activate the PmrA/PmrB system.

E. coli K-12 is unable to make the LPS modifications required for polymyxin B resistance when grown in LB broth (23), which is a medium of relatively low levels of Mg²⁺. This result is surprising because: (i) E. coli encodes homologues of the PhoP/PhoQ and PmrA/PmrB systems, the PmrD protein, as well as the PmrA-regulated genes mediating the LPS modifications required for polymyxin B resistance (24); (ii) its PhoP/PhoQ system also responds to Mg²⁺ (25); and (iii) mutations in its pmrA gene produce strains with the same phenotypes as Salmonella pmrA mutants (8, 12). We now demonstrate that E. coli differs from Salmonella in the ability of the PmrA/PmrB system to respond to the low Mg²⁺ signal and establish that this is due to allelic differences in the regulatory protein PmrD. Our results suggest that the disparate control of a regulatory system can mediate critical differences in phenotypic traits that have ecological consequences for a bacterial species.

Materials and Methods

Bacterial Strains, Plasmids, and Growth Conditions. Bacterial strains and plasmids used in this study are listed in Table 2, which is published as supporting information on the PNAS web site. Except where indicated, strains are derived from S. enterica serovar Typhimurium strain 14028s or E. coli K-12 strain MG1655. Bacteria were grown at 37°C in N-minimal media, pH 7.7 or 5.8 (26), supplemented with 0.1% casamino acids/38 mM glycerol or 0.2% glucose/2% LB (when necessary for Salmonella natural isolate auxotrophs)/10 μM or 10 mM MgCl₂/100 μM FeS0₄, as indicated. P1 transduction of E. coli strains was performed as described (27). E. coli DH5α was used as the host for preparation of plasmid DNA. Ampicillin and kanamycin were used at 50 μg/ml, chloramphenicol at 20 μg/ml and tetracycline at 10 μg/ml. The DNA sequences of primers used in this study are listed in Table 3, which is published as supporting information on the PNAS web site.

Introduction of Gene Fusions and Allelic Replacements in the Chromosomal pmrD Loci. The pmrD gene was deleted from E. coli and Salmonella natural isolates by using primer pairs 1739, 1705 and 1485, 1486, respectively, as described (28). Primers 1739 and 2539 were used to construct the E. coli strain harboring a lacZY transcriptional fusion immediately downstream of the pmrD-coding region, as described (29).

To construct isogenic E. coli derivatives of strain MG1655 harboring either the E. coli pmrD or Salmonella pmrD ORF, we used a modification of the one-step inactivation protocol (28). We generated DNA fragments encompassing the C-terminally FLAG-tagged E. coli or Salmonella pmrD coding regions and a Cm^R cassette by performing two sequential PCR reactions. The C-terminally FLAG-tagged E. coli or Salmonella pmrD coding regions were amplified from the E. coli or Salmonella chromosomes by using primers pairs 2128, 2315 and 2282, 2316, respectively. These fragments also harbored 40 bp of E. coli pmrD upstream sequence on the 5′ end and 19 bp of 5′ sequence from the Cm^R cassette on the 3′ end. In addition, we amplified the Cm^R cassette from the plasmid pKD3 by using primers 2126 and 2127. The generated Cm^R cassette had 40 bp of E. coli pmrD downstream sequence on the 3′ end. A second PCR was performed by using the first two amplicons as the template to combine the sequences encoding the C-terminally FLAG-tagged E. coli or Salmonella pmrD ORF and the Cm^R cassette, both flanked by E. coli pmrD upstream and downstream sequences by using primers 2130 and 2131. The PCR products were purified, digested, and specifically recombined into the wild-type E. coli chromosome in place of the endogenous E. coli pmrD ORF, as described (28). Gene replacements were confirmed by sequence analysis, and protein expression was confirmed by Western blot analysis, as described (30).

S1 Nuclease Assay. S1 nuclease assays were performed as described (22) by using RNA isolated from early-logarithmic phase cultures (A₆₀₀, 0.250) grown in 50 ml of N-minimal medium, pH 7.7, containing 38 mM glycerol with either 10 μM MgCl₂,10mM MgCl₂, or 10 μM MgCl₂ and 100 μM FeSO₄. To generate dsDNA probes, we used primers 995 and 767 for the Salmonella pbgP promoter and primers 1618 and 1619 for the E. coli pbgP promoter. Hybridization reactions were performed at 30°C and 45°C for Salmonella and E. coli RNA, respectively. Assays were performed in triplicate.

β-Galactosidase Assay. E. coli K-12 strains were grown in N-minimal media, pH 7.7, supplemented with 0.2% glucose (to suppress β-galactosidase activity from the plac promoter), and either 10 μM MgCl₂ or 10 mM MgCl₂. Activity was determined as described (27) after 4 h of growth at 37°C. Assays were performed in triplicate.

GFP Expression Assay. Strains harboring the pAG, pAG-rpsM, pAG-pbgP_Salmonella, or pAG-pbgP_{E. coli} 1–7 or 9 plasmids were grown in N-minimal media, pH 7.7, containing 38 mM glycerol with either 10 μM MgCl₂, 10 mM MgCl₂, or 10 μM MgCl₂ and 100 μM FeSO₄ and supplemented with 10 μg/ml tetracycline and 2% LB when necessary. GFP expression was analyzed after 4 h of growth at 37°C by using a Becton Dickinson fluorescent-activated cell sorter. Assays were performed in triplicate.

Polymyxin B Susceptibility Assay. Cells were grown to logarithmic phase in N-minimal media, pH 5.8, containing 38 mM glycerol with either 10 μM MgCl₂, 10 mM MgCl₂, or 10 μM MgCl₂ and 100 μM FeSO₄, washed, and incubated in the presence of 2.5 μg/ml polymyxin B at 37°C for 1 h. Samples were serially diluted in PBS and plated onto LB agar plates and incubated overnight at 37°C for viability counts. Survival values were calculated by dividing the number of bacteria after treatment with polymyxin B relative to those incubated in the presence of PBS and then multiplied by 100. Assays were performed in triplicate.

Sequence Data. Genes of interest were amplified with AccuPrime TaqDNA polymerase high fidelity (Invitrogen) by using primers upstream and downstream of the ORFs, by using the following primer pairs: Salmonella pmrA, 2876 and 2877; Salmonella pmrD, 641 and 1897; E. coli pmrA, 2591 and 2592; E. coli pmrD, 1782 and 2644; and E. coli pbgP, 4096 and 2416. PCR products were purified with the QIAquick PCR purification kit (Qiagen, Chatsworth, CA). Sequencing reactions were initiated by using the following primers: Salmonella pmrA, 2878, 2879, and 2880; Salmonella pmrD, 2881; E. coli pmrA, 3080 and 2595; E. coli pmrD, 2131; and E. coli pbgP, 4097, performed by using big dye 3.1 (Applied Biosystems) and analyzed on a 310 Genetic Analyzer (Perkin–Elmer). DNA sequences were translated by using editseq 3.92 (DNASTAR, Madison, WI). Percent amino acid identities was determined by megalign (DNASTAR), which uses the clustalv algorithm (31). Multiple sequence alignments were performed with clustalx (32).

Data Analysis and Statistical Tests. Percent connectivity was calculated as the ratio between the values of the low Mg²⁺- and Fe³⁺-induced expression multiplied by 100. The calculation measures the ability of a strain to connect the PhoP/PhoQ and PmrA/PmrB systems: a higher percentage indicates an increased ability to make this connection. PmrD dependence was calculated as the fold decrease in the low Mg²⁺-induced phenotype of the ΔpmrD mutant compared with its isogenic pmrD⁺ strain. The calculation measures the contribution of the pmrD gene to the low Mg²⁺-induced phenotype; a higher value indicates that the wild-type phenotype has an increased requirement for the function of the PmrD protein. Data were plotted as frequency distributions for the E. coli and Salmonella isolates, respectively. Distributions were subjected to the Mann–Whitney test, a nonparametric test for the significance of the difference between the distributions of two independent samples, and P values are reported. The McDonald–Kreitman test was performed for the pmrD and pmrA sequence data by using dnasp 4 (33, 34).

Results

The PmrA/PmrB System of E. coli K-12 Is Blind to the Low Mg²⁺ Signal. We determined that resistance to the antibiotic polymyxin B in E. coli depends on the pmrA and pbgP genes (data not shown), like in Salmonella (11, 15, 16). But surprisingly, E. coli was resistant to polymyxin B if grown in the presence of Fe³⁺ but not in response to low Mg²⁺ (Fig. 2A), whereas Salmonella was resistant to polymyxin B after growth in either condition (Fig. 2B) (15, 22). This phenotype was reflected also at the level of transcription; the PmrA-activated pbgP gene was expressed by E. coli cells experiencing Fe³⁺ but not in response to the low Mg²⁺ signal (Fig. 2C), whereas a pbgP transcript was made by Salmonella in both conditions (Fig. 2D). [The Fe³⁺-promoted pbgP transcription in E. coli required a functional pmrA gene (data not shown), as demonstrated in Salmonella (9)]. As expected, there was no pbgP transcription when either E. coli or Salmonella was grown in high Mg²⁺ (Fig. 2 C and D), a condition that represses expression of the PhoP/PhoQ-activated pmrD gene (22). Cumulatively, these results demonstrate that E. coli harbors a functional PmrA/PmrB system that responds to Fe³⁺ but is blind to low Mg²⁺ (Fig. 1B).

Fig. 2.

Low Mg²⁺ promotes transcription of the PmrA-activated pbgP gene and polymyxin B resistance in Salmonella but not in E. coli. (A) Survival of wild-type E. coli (MG1655) to polymyxin B (2.5 μg/ml) after growth in N-minimal medium, pH 5.8, containing 38 mM glycerol with 10 μM Mg²⁺ (–) or 10 μM Mg²⁺ and 100 μM Fe³⁺ (+). (B) Survival of wild-type Salmonella (14028s) to polymyxin B (2.5 μg/ml) after bacterial growth as described in A. Assays were performed as described in Materials and Methods. (C) S1 mapping of the wild-type E. coli (MG1655) pbgP promoter in bacteria grown in N-minimal medium, pH 7.7, containing 38 mM glycerol with 10 μM Mg²⁺ (L, –), 10 mM Mg²⁺ (H, –), or 10 μM Mg²⁺ and 100 μM Fe³⁺ (L, +) and harvested during logarithmic growth phase. (D) S1 mapping of the wild-type Salmonella (14028s) pbgP promoter in bacteria grown as described in C. Lanes G, A, T, and C correspond to dideoxy chain-termination sequence reactions corresponding to this region. The sequence spanning the transcription start site is shown, and the start site is marked with an arrow.

Transcription of the E. coli pmrD Gene Is Promoted in Low Mg²⁺ in a PhoP-Dependent Manner. E. coli's behavior mimicked that displayed by a Salmonella pmrD mutant (22), raising the possibility that the pmrD gene may not be expressed in E. coli. However, the E. coli pmrD gene was induced in low Mg²⁺ in a phoP-dependent manner (Fig. 5A, which is published as supporting information on the PNAS web site), like other members of the E. coli PhoP regulon (25) and the Salmonella pmrD gene (22). Furthermore, low Mg²⁺ stimulated the production of a chromosomally encoded FLAG epitope-tagged PmrD protein in wild-type E. coli but not in an isogenic phoP mutant (Fig. 5B). These results demonstrate that E. coli produces a PmrD protein in response to low Mg²⁺ in a PhoP-dependent manner.

The Salmonella pmrD Gene Enables E. coli to Express the pbgP Gene and to Display Resistance to Polymyxin B in Response to Low Mg²⁺. The E. coli PmrD protein is only 55.3% identical to the Salmonella PmrD protein, which is well below the median amino acid identity between homologous proteins of these two species (i.e., 90%) (1), including that exhibited by the PhoP, PhoQ, PmrA, and PmrB proteins (i.e., 84.0–93.3%). Because the PmrD protein is expressed in E. coli under the same conditions as in Salmonella, this result suggested that the highly divergent E. coli PmrD protein might not be functional. Moreover, it raised the possibility that expression of the Salmonella PmrD protein might be sufficient to activate the E. coli PmrA/PmrB system in response to the low Mg²⁺ signal. Thus, we constructed an E. coli strain in which the pmrD ORF was replaced by the Salmonella pmrD ORF, resulting in a strain that expressed the Salmonella PmrD protein from the E. coli pmrD promoter at its normal chromosomal location. When this strain was grown in low Mg²⁺, it transcribed the pbgP gene using the same transcription start site as wild-type E. coli exposed to Fe³⁺ (Fig. 3A), whereas the isogenic strain expressing E. coli's own pmrD gene was unable to promote pbgP transcription in response to low Mg²⁺ (Fig. 3A). As expected, there was no pbgP transcription when cells were grown in high Mg²⁺ (Fig. 3A). Moreover, the E. coli strain expressing the Salmonella pmrD gene displayed resistance to polymyxin B in low Mg²⁺ whereas the one expressing the E. coli pmrD gene did not (Fig. 3B).

Fig. 3.

The Salmonella pmrD gene enables E. coli to promote transcription of the PmrA-activated pbgP gene and to be resistant to polymyxin B after growth in low Mg²⁺. (A) S1 mapping of the pbgP promoter in wild-type E. coli (MG1655), E. coli expressing the Salmonella pmrD gene (EG13622), and an isogenic strain harboring the E. coli pmrD gene (EG15041). Bacteria were grown as described in the legend to Fig. 2C. Lanes G, A, T, and C correspond to dideoxy chain-termination sequence reactions corresponding to this region. The sequence spanning the transcription start site is shown, and the start site is marked with an arrow. (B) Survival of E. coli cells harboring the Salmonella pmrD (EG13622) or the isogenic strain with the E. coli pmrD gene (EG15041) to polymyxin B (2.5 μg/ml), following the assay conditions described in the legend to Fig. 2 A. (C) S1 mapping of the pbgP promoter in wild-type E. coli (MG1655) grown in N-minimal media containing 10 μM Mg²⁺ and 100 μM Fe³⁺ (1) or 10 mM Mg²⁺ (2), wild-type E. coli (MG1655) harboring plasmid pLK32 with E. coli pmrD (3), plasmid pLK24 with Salmonella pmrD (4), plasmid ppmrD C1 with a E. coli-Salmonella pmrD chimera (5), plasmid ppmrD C2 with a Salmonella–E. coli pmrD chimera (6), or the pUC19 plasmid vector (7). Strains in lanes 3–7 were grown in N-minimal media containing 10 mM Mg²⁺, 5 mM IPTG, and 50 μg/ml ampicillin. Strains were grown and RNA harvested as described in the legend to Fig. 2C.

To explore the molecular basis for the distinct phenotypes mediated by the different pmrD genes upon E. coli, we investigated pbgP transcription in wild-type E. coli harboring plasmids expressing PmrD chimeras between the E. coli and Salmonella PmrD proteins from the plac promoter (see Supporting Text, which is published as supporting information on the PNAS web site). Cells harboring plasmid ppmrD C2, which encodes a protein sequence corresponding to amino acids 1–44 from Salmonella PmrD and amino acids 45–88 from E. coli PmrD, expressed the pbgP gene although not to the levels exhibited by cells harboring a plasmid that expressed the full length Salmonella PmrD protein (Fig. 3C). In contrast, no expression was detected in cells harboring plasmid ppmrD C1, which encodes a protein sequence corresponding to amino acids 1–44 from E. coli PmrD and amino acids 45–85 from Salmonella PmrD, a plasmid expressing full length E. coli PmrD, or the plasmid vector (Fig. 3C). As expected, wild-type E. coli could express pbgP in response to Fe³⁺ even in the absence of plasmids. These results suggest that the N-terminal region of PmrD, which differs at 19 of 44 positions between the E. coli and Salmonella proteins, is in part responsible for the distinct behaviors of the PmrD proteins.

Sequence Analysis Reveals Nonneutral Evolution of the pmrD Gene. To explore whether the disparity in PmrD function between E. coli and Salmonella reflected a regulatory difference distinguishing these two enteric genera (as opposed to a difference pertaining to the two investigated strains), we examined a collection of 72 natural isolates of E. coli [E. coli standard reference collection (ECOR)] (35) and 16 natural isolates of Salmonella [Salmonella reference collection C (SARC)] (36) believed to represent the genetic diversity of these two genera. DNA sequence analysis of the pmrD genes revealed that their deduced amino acid sequences were 92.1–100% identical within E. coli (n = 71) (Fig. 6, which is published as supporting information on the PNAS web site) and 88.0–98.8% identical within Salmonella (n = 15) (Fig. 7, which is published as supporting information on the PNAS web site). Likewise, the PmrA proteins were also highly conserved, being 97.8–100% identical within E. coli (n = 65) (Fig. 8, which is published as supporting information on the PNAS web site) and 95.1–99.1% identical within Salmonella (n = 15) (Fig. 9, which is published as supporting information on the PNAS web site). On the other hand, the PmrD proteins were only 53.5–56.6% identical between the E. coli and Salmonella isolates, which is in contrast to the 89.7–92.4% identity exhibited between the PmrA proteins of the two genera.

Given the dramatic sequence divergence between the E. coli and Salmonella pmrD alleles, we predicted that this gene might be evolving in a nonneutral fashion. The neutral theory (37) predicts that the ratio of replacement (nonsynonymous) to silent (synonymous) substitutions observed between species equals the ratio of replacement to silent substitutions observed within species (38). Thus, we applied the McDonald–Kreitman test of similarity in ratios of replacement to silent changes within and between species (38) to examine the evolution of the pmrD gene. We determined that there is a significant excess of fixed amino acid changes at the pmrD locus but not the pmrA locus, which was used as a control (Table 1). This analysis rejects the null hypothesis of neutral evolution for the pmrD gene and suggests that selection might have driven the divergence of the E. coli and Salmonella PmrD proteins.

Table 1. McDonald–Kreitman test for protein evolution.

	pmrD*		pmrA†
	Fixed between species	Polymorphic within species	Fixed between species	Polymorphic within species
Replacement	41	22	14	19
Silent	23	54	53	146

Fisher's exact test, P = 0.000040 (pmrD), P = 0.095 not significant (pmrA).

^*n = 71 E. coli, n = 15 Salmonella.

^†n = 64 E. coli, n = 15 Salmonella.

Functional Divergence of the E. coli and Salmonella PmrD Proteins. To determine whether the sequence divergence between E. coli and Salmonella PmrD proteins resulted in a functional difference, we examined pbgP transcription in 21 nonpathogenic and 4 pathogenic E. coli isolates (35) and 12 Salmonella isolates. We constructed plasmids harboring isolate-specific pbgP promoters in front of a promoterless gfp gene and measured GFP fluorescence in strains harboring their specific pbgP promoter construct (i.e., ECOR1 would harbor a plasmid with the ECOR1 pbgP promoter in front of promoterless gfp) (see Supporting Text). The fluorescence values for the E. coli isolates ranged from 5 to 50 units during growth in low Mg²⁺ and from 20 to 200 units during growth in Fe³⁺ (Fig. 10A, which is published as supporting information on the PNAS web site). In contrast, the fluorescence values for the Salmonella isolates ranged from 40 to 240 units during growth in low Mg²⁺ and from 100 to 350 units during growth in Fe³⁺ (Fig. 10B). These data indicate that there is variation in the pbgP expression levels both within and between natural populations of E. coli and Salmonella, including cases where an E. coli isolate displays a Salmonella phenotype and vice versa.

The above data were used to determine the connectivity between the PhoP/PhoQ and PmrA/PmrB systems, which was defined as the ratio between the pbgP expression values at low Mg²⁺ and those obtained in response to Fe³⁺. The mean connectivity of the E. coli isolates was 26% as compared with 60% for the Salmonella isolates, with statistically different distributions across isolates (Mann–Whitney, P < 0.0001) (Fig. 4A). Because the PmrD protein is responsible for connecting the PhoP/PhoQ and PmrA/PmrB systems in Salmonella, we deleted the pmrD gene in a subset of the investigated isolates and determined the PmrD dependence of the pbgP expression taking place in low Mg²⁺ by calculating the ratio of pbgP expression exhibited by isogenic pmrD⁺ and ΔpmrD strains (Fig. 7 A and B). Deletion of the pmrD gene resulted in an average 6-fold reduction in pbgP expression among the E. coli isolates as compared with an average 31-fold reduction displayed by Salmonella isolates with statistically significant different distributions across isolates (Mann–Whitney, P < 0.0001) (Fig. 4B). In sum, these analyses indicate that the species E. coli does not generally have the PmrD-mediated connection between the PhoP/PhoQ and PmrA/PmrB systems whereas the genus Salmonella does.

Fig. 4.

E. coli and Salmonella natural isolates differ in the PmrD-dependent connectivity of the PhoP/PhoQ and PmrA/PmrB systems in low Mg²⁺. (A) Frequency distribution of the percent connectivity, calculated as the ratio between the pbgP expression in low Mg²⁺ and in Fe³⁺, in wild-type E. coli (n = 25, gray) and Salmonella (n = 12, black) isolates. The two distributions are significantly different from each other (Mann–Whitney test, P < 0.0001). (B) Frequency distribution of the PmrD dependence, calculated as the ratio of pbgP expression exhibited by isogenic pmrD⁺ and Δ pmrD strains of E. coli (n = 13 pairs, gray) and Salmonella (n = 7 pairs, black). The two distributions are significantly different from each other (Mann–Whitney test, P < 0.0001).

The PmrA Protein Does Not Repress pmrD Transcription in E. coli. When the Salmonella PmrA/PmrB system is activated independently of PmrD, such as by exposure to Fe³⁺, the PmrA protein binds to the pmrD promoter and represses pmrD transcription (30) (Fig. 1 A). We reasoned that if the PmrA-mediated repression of the Salmonella pmrD gene is designed to prevent the potentially detrimental production of PmrD protein, this negative feedback loop would be absent from E. coli because its PmrD protein is unable to activate the PmrA protein. Consistent with this notion, the transcriptional activity originating from a chromosomal pmrD-lacZ fusion was the same in isogenic wild-type and pmrA strains of E. coli K-12 (data not shown). Furthermore, the E. coli pmrD promoter does not appear to have a PmrA-binding site.

Discussion

It is becoming increasingly clear that the phenotypic variation that exists among higher eukaryotic species is to a large extent due to differences in gene regulation (as opposed to gene content) (39–42). We demonstrate here that this genetic mechanism is not limited to eukaryotes by documenting a case in which phenotypic differences between two closely related bacterial species result from disparate regulation of homologous genes.

In Salmonella, the PmrD protein enables the Fe³⁺-responding PmrA/PmrB system to also be activated in response to the low Mg²⁺ signal that induces the PhoP/PhoQ system. This connectivity capacity expands the spectrum of environments in which the PmrA regulon can be expressed and is not limited to a single strain of Salmonella but conserved across the Salmonella genus (Fig. 4). In contrast, E. coli is unable to connect the PhoP/PhoQ and PmrA/PmrB systems (Figs. 2C and 4A), which prevents it from being resistant to polymyxin B after growth in low Mg²⁺ (Fig. 2 A). This behavior is due to the highly divergent PmrD protein (Figs. 6 and 7), because replacement of the E. coli pmrD gene by the Salmonella pmrD gene endowed E. coli with the ability to transcribe the PmrA-activated pbgP gene and to exhibit resistance to polymyxin B in response to the low Mg²⁺ signal (Fig. 3). Although there was variation in the degree to which the PhoP/PhoQ and PmrA/PmrB systems were connected in different E. coli and Salmonella isolates, our analysis demonstrates a statistically significant difference in the distribution of the E. coli and Salmonella isolates with respect to this property (Fig. 4).

The disparate ability of Salmonella and E. coli to perform the PmrA-controlled modification of the LPS (23) may contribute to the differential survival of these two enteric species in host and nonhost environments. For example, the Salmonella pmrB and pbgP loci are required for colonization of chicken macrophages (7), whereas E. coli does not normally proliferate within phagocytic cells. Likewise, Salmonella demonstrates a high survival rate in nutrient-poor nonhost environments, such as water and soil, in which E. coli has a net negative growth rate (43). Survival in these nonhost environments has been correlated with the ability to adhere to soil particles (44), a characteristic believed to result from modifications in the bacterial outer membrane (43).

The sequence divergence between the E. coli and Salmonella pmrD genes appears to result from selection at this locus (Table 1). In Salmonella, selection may serve to maintain the PmrD-mediated connectivity between the PhoP/PhoQ and PmrA/PmrB systems. In E. coli, selection resulted in a PmrD protein that does not connect the PhoP/PhoQ and PmrA/PmrB systems. This could have driven PmrD to lose its function, because expressing PmrA-activated genes in low Mg²⁺ might be deleterious for E. coli's lifestyle. Indeed, pmrD appears to be a pseudogene in the two sequenced strains of Shigella flexneri (45, 46). Because E. coli and Shigella are considered members of the same species (46, 47), these data suggest that the PmrD function established in Salmonella has been lost from the E. coli–Shigella lineage. Thus, the PmrD-mediated connection between the PhoP/PhoQ and PmrA/PmrB systems is a regulatory difference that distinguishes Salmonella and E. coli, rather than a difference between pathogenic and nonpathogenic enteric bacteria. It is also possible that the E. coli PmrD protein has a yet-unidentified role both in driving the evolution of the pmrD gene and in preventing its loss from the E. coli genome.

Conclusion

Our analysis demonstrates that the differential regulation of conserved genes can mediate phenotypic traits that distinguish closely related bacterial species. This can be due to allelic variation in orthologous regulatory proteins, such as PmrD, or the result of differences in cis regulatory sequences. For example, the presence of a PmrA-binding site in the Salmonella pmrD promoter enables the PmrA protein to repress pmrD transcription (30), but such regulatory site is absent from the E. coli pmrD promoter, whose transcription is not affected by the presence of the PmrA protein (this work). Mutations in cis regulatory sites have been shown to mediate morphological divergence in fish (48) and Drosophila (49, 50) species (51).

Finally, our work suggests that the activation of a regulatory system by a different spectrum of stimuli may have major ecological consequences for E. coli and Salmonella, possibly determining both the range of niches available to these organisms and the evolution of these two bacterial genera. Moreover, it indicates that a complete elucidation of bacterial lifestyle differences demands investigations beyond those involving species-specific genes.