Identification of the lipopolysaccharide modifications controlled by the Salmonella PmrA/PmrB system mediating resistance to Fe(III) and Al(III)

The pbgE2 and pbgE3 genes are not directly involved in Fe(III) resistance

We determined that the minimal inhibitory concentration (MIC) for Fe(III) of strains deleted for the pbgE2 or pbgE3 genes is 50 μM, which is similar to the MIC of the ΔpmrAB mutant (i.e. 100 μM) and much lower than the MIC of the wild-type strain or mutants defective in either the pbgP or ugd genes (i.e. 3.2 mM) (Table 1). Plasmids expressing the pbgE2 or pbgE3 genes from a derivative of the tet promoter restored wild-type levels of resistance to the pbgE2 and pbgE3 mutants respectively (Table 1). Interestingly, wild-type levels of Fe(III) resistance could also be restored to the pbgE2 or pbgE3 mutants upon inactivation of the pbgP or ugd genes (Table 1), suggesting that the pbgE2 and pbgE3 gene products are not directly involved in Fe(III) resistance. Consistent with this notion, the plasmids expressing the pbgE2 or pbgE3 genes or one that expressed both pbgE2 and pbgE3 from the tet promoter derivative failed to confer Fe(III) resistance upon the ΔpmrAB mutant (Table 1). Thus, the connection between the pbgE2 or pbgE3 genes and Fe(III) resistance was not investigated further.

Identification of Fe(III) resistance genes

To uncover the PmrA-regulated genes mediating Fe(III) resistance, we prepared a genomic library from the Fe(III)-susceptible ΔpmrAB mutant in the multicopy number plasmid pBR322 (Bolivar et al., 1977). Plasmid DNA was isolated from a pool of 26 000 transformants, introduced into the same ΔpmrAB mutant and transformants that could grow on agar plates containing 200 μM Fe(III) and 50 μg ml⁻¹ ampicillin were recovered. The rationale behind this strategy was that a PmrA-regulated Fe(III) resistance gene(s) might be expressed from the tet promoter in pBR322 and phenotypically rescue the ΔpmrAB mutant. Then, we isolated plasmid DNA from each of 52 purified transformants and used it to retransform the ΔpmrAB mutant. All 52 plasmids conferred Fe(III) resistance, indicative that the plasmids harboured Fe(III) resistance genes.

Sequence analysis of the inserts in the 52 plasmid clones revealed that DNA originating from seven different regions of the Salmonella chromosome could confer Fe(III) resistance upon the ΔpmrAB mutant (Fig. S1). These inserts encompassed a total of 16 open reading frames (ORFs), which were individually subcloned into pBR322. Eight of the resulting subclones could still confer Fe(III) resistance upon the ΔpmrAB mutant but to different degrees (Fig. S2). We focused our attention on the pmrC and pmrG genes because they bestowed the highest levels of Fe(III) resistance upon the ΔpmrAB mutant (Fig. S2), and because they were known to be directly regulated by the PmrA protein (Gunn et al., 1998; Wosten and Groisman, 1999); this is in contrast to the six other genes, whose expression was not PmrA-dependent (data not shown).

The PmrA-activated ugd, pbgP, pmrC and pmrG genes are required for Fe(III) resistance

To examine whether the pmrG and pmrC genes are necessary for Fe(III) resistance, we constructed strains deleted for the chromosomal copies of these genes (see Experimental procedures). Both ΔpmrG and ΔpmrC single mutants and a ΔpmrG pmrC double mutant behaved like the wild-type parent (Table 1). We also made double mutants deleted in additional PmrA-regulated genes or operons (i.e. ugd, yibD, and the pbgPE operon). These mutants also resembled the wild-type parent in terms of MIC for Fe(III) (Table 1). Thus, we constructed strains deleted for multiple PmrA-regulated genes with the hope of creating a strain that recapitulated the Fe(III) hypersensitivity of the ΔpmrAB mutant.

A strain deleted for the pbgPE operon and the pmrC, ugd, pmrG and yibD genes exhibited the same hypersensitivity to Fe(III) as the ΔpmrAB mutant (Table 1). This was also true for a strain deleted for the pmrC, ugd, pmrG and yibD genes (Table 1), which is consistent with the fact that the proteins encoded by the ugd gene and pbgP operon participate in the same pathway of synthesis and incorporation of 4-aminoarabinose into lipid A (Gunn et al., 1998; Zhou et al., 2001; Breazeale et al., 2003). We investigated all possible combinations of double and triple mutants with deletions in the pmrC, ugd, pmrG and yibD genes and determined that only the ΔpmrC ugd pmrG triple mutant was as hypersensitive to Fe(III) as the ΔpmrAB mutant (Table 1). Furthermore, a strain deleted for of all three pmrC, ugd and pmrG genes and harbouring the pmrA505 allele, which encodes a PmrA protein that promotes transcription of PmrA-activated genes even under non-inducing conditions (Kox et al., 2000), was as susceptible to Fe(III) as the ΔpmrAB mutant (Table 1). Collectively, these results strongly suggest that the pmrC, ugd and pmrG genes and the pbgPE operon are the only PmrA-regulated determinants required for Fe(III) resistance.

The Fe(III)-susceptible mutants display increased Fe(III) binding

The LPS is one of the major components and the most accessible surface molecule in the outer leaflet of the bacterial outer membrane. It consists of three structurally distinct regions: the outermost O-antigen, a central core, and the innermost lipid A (Raetz and Whitfield, 2002). The pmrC gene product and the proteins encoded in the ugd gene and pbgPE operon mediate the modification of the two lipid A phosphates with phosphoethanolamine (Lee et al., 2004) and 4-aminoarabinose (Gunn et al., 1998) respectively. This suggested that Fe(III) resistance may result from hindering electrostatic interactions between the positively charged Fe(III) and the negatively charged phosphates in the lipid A; and it predicted that the Fe(III)-susceptible mutants would bind more iron than the wild-type strain. Consistent with this notion, there were 3–3.5 times more ⁵⁹Fe associated with the ΔpmrAB mutant and the ΔpmrC ugd pmrG triple mutant than with the wild-type strain (Fig. 1A). On the other hand, ⁵⁹Fe association with the pmrC, ugd and pmrG single mutants was similar to that exhibited by the wild-type parent (Fig. 1A). These results demonstrated that those mutants that bind more Fe(III) are more readily killed by this metal.

Polymyxin B nonapeptide rescues the ΔpmrC ugd pmrG triple mutant from Fe(III)-mediated killing

Modification of the lipid A phosphates with phosphoethanolamine and 4-aminoarabinose is required for resistance to both polymxyin B (Gunn et al., 1998; Lee et al., 2004) and Fe(III) (Table 1), raising the possibility that these bactericidal agents bind to similar or adjacent sites in lipid A. Consistent with this notion, the non-toxic polymyxin B nonapeptide, which binds to lipid A like the bactericidal polymyxin B (Yin et al., 2003), could partially rescue the ΔpmrC ugd pmrG triple mutant from Fe(III)-mediated killing (Fig. 1B). These results reinforce the notion that Fe(III) resistance is associated with modification or occlusion of the lipid A phosphates in the LPS.

PmrG is a phosphatase that targets the Hep(II) phosphate in the core region of the LPS

We then turned our attention to the pmrG mutant and determined that it exhibited a wild-type lipid A profile and resistance to polymyxin B (Fig. S3 and data not shown), which was in contrast to the phenotype of strains defective in the pmrC, pbgP or ugd genes (Gunn et al., 1998; Lee et al., 2004). As overexpression of the pmrG gene did not alter the lipid A profile of the ΔpmrC ugd pmrG triple mutant (Fig. S3), our findings implied that the PmrG protein was not likely to target the lipid A.

A hint to the potential function of the PmrG protein came from an independent genetic screening in which we mutagenized the ΔpmrAB strain and isolated pseudorevertants that could grow on agar plates containing 400 μM Fe(III). One of the pseudorevertants harboured a EZ-Tn5 (http://www.epibio.com/category.asp?id=284) transposon insertion in the rfaY gene (also known as waaY), which is necessary for phosphorylating the Hep(II) heptose in the core region of the LPS (Yethon et al., 1998) (Fig. 2A). Reconstruction experiments demonstrated that deletion of the rfaY gene suppressed Fe(III)-mediated killing not only in the ΔpmrAB strain but also in the ΔpmrC pmrG ugd triple mutant (Table 2), restoring Fe(III) resistance to the levels exhibited by the ΔpmrAB strain harbouring the plasmid expressing the wild-type pmrG gene (Table 2).

We hypothesized that the PmrG protein might counteract the action of the RfaY protein, removing the phosphate from the Hep(II) heptose. Thus, we purified a C-terminally His-tagged PmrG protein (Fig. 2B) and investigated its ability to dephosphorylate the core region of the LPS. The PmrG-His protein exhibited phosphatase activity towards core oligosaccharide (OS) prepared from the wild-type strain (Fig. 2C), but not against OS originating from the ΔrfaY mutant (Fig. 2C), which lacks the phosphate on Hep(II) (Yethon et al., 1998). The data presented above indicate that the PmrG protein encodes a phosphatase that targets the Hep(II) phosphate in the core region of the LPS.

PmrG protein is a periplasmic protein

To examine the subcellular location of the PmrG protein, we used a Salmonella strain that expressed a PmrG-FLAG protein from the normal chromosomal pmrG promoter, and also the cytoplasmic enzyme β-galactosidase, which served as a marker for the purity of the extracytoplasmic fractions. Western blot analysis with anti-FLAG antibodies demonstrated that the PmrG-FLAG protein localizes to the periplasmic region (Fig. 3D), as predicted by the psort program (http://www.psort.org/). We also noticed a slightly larger band in the cytoplasmic fraction (Fig. 3D), which could correspond to the full-length (i.e. unprocessed) form of the PmrG protein.

Modification of the Hep(I) phosphate with phosphoethanolamine is not required for Fe(III) resistance

The core region of the LPS has two phosphates: one at Hep(II) that is targeted by PmrG (Fig. 2C), and one at Hep(I) that can be modified with phosphoethanolamine by the PmrA-activated cptA gene product (Tamayo et al., 2005). We determined that the latter modification is not required for Fe(III) resistance because deletion of the cptA gene did not alter the susceptibility of the wild-type strain, the ΔpmrAB mutant, the ΔpmrC ugd, ΔpmrC pmrG and Δugd pmrG double mutants, or the ΔpmrC ugd pmrG triple mutant (Table S3).

Metals that induce the PmrA/PmrB system promote killing of the ΔpmrC ugd pmrG mutant

Al(III), which had been previously shown to be an inducer of the PmrA/PmrB system (Wosten et al., 2000), also displayed bactericidal activity against the ΔpmrC ugd pmrG triple mutant (Table 3). On the other hand, neither Ca²⁺ nor Co²⁺, K⁺, Ni²⁺, Zn²⁺, Mn²⁺, Cu²⁺, Ga³⁺ or Ru³⁺ (at concentrations of up to 1 mM) could induce PmrA-activated genes (Wosten et al., 2000) or kill the ΔpmrC ugd pmrG triple mutant (data not shown). This suggests that there is correlation between the ability of a cation to activate the PmrA/PmrB system and its ability to kill the ΔpmrC ugd pmrG triple mutant.

The LPS modifications mediating Fe(III) resistance enhance survival in soil

Salmonella can survive in non-host environments such as soil and water for extended periods of time (Winfield and Groisman, 2003). Because Fe(III) and Al(III) are the most abundant metals in soil (Pina and Cervantes, 1996; Giesler et al., 2005) and soil is an environment that activates the PmrA/PmrB system (Chamnongpol et al., 2002), we hypothesized that the PmrA-regulated genes implicated in resistance to Fe(III) and Al(III) would be required for survival in soil. Indeed, the ΔpmrC ugd pmrG triple mutant was recovered 100-fold less than the wild-type strain following incubation in soil (Fig. 3).

Bacterial strains and plasmids

Bacterial strains and plasmids are listed in Table S1. For plasmid constructions, genes were amplified from the wild-type strain 14028s genomic DNA by PCR with primers listed in Table S2. The PCR fragments were cloned between the restriction enzyme sites of vector pBR322 as indicated in Table S1, which resulted in these genes being present in the same orientation and downstream of the tet promoter of the pBR322 vector (Bolivar et al., 1977). The construction of strains with deletions in particular genes was performed as described (Datsenko and Wanner, 2000) with the primers listed in Table S2. The drug-resistance genes were eliminated by using plasmid pCP20 as described (Datsenko and Wanner, 2000).

A strain expressing a chromosomally encoded PmrG-FLAG protein was constructed as described (Datsenko and Wanner, 2000) with primers 5′-AAACGGAAAACTCTTTTTAGATGGTGAATTTGTTCCCGGCGACTACAAGGACGACGATGACAAGTAGGTGTAGGCTGGAGCTGCTTC-3′ encoding the FLAG sequence immediately upstream of the stop codon of the pmrG gene following the priming site 1 sequence (Datsenko and Wanner, 2000) (sequence in bold face corresponds to the FLAG coding sequence) and 5′-GACAGCCGCTTCAGGCTGTCTTTACGCTTTTAACATGCGGCATATGAATATCCTCCTTAG-3′ harbouring the sequence immediately downstream of the stop codon of pmrG attached to priming site 2 (Datsenko and Wanner, 2000).

Selecting for Fe(III) resistance genes

A genomic library from the ΔpmrAB strain EG13937 was made in pBR322 as follows: chromosomal DNA prepared from an overnight culture of the ΔpmrAB strain was digested with Sau3AI (1 U μl⁻¹) for 15, 20, 30 or 40 min. The digested DNA was separated on 0.8% agarose gel, and fragments of approximately 2–10 kb in size were purified and ligated into the BamHI site of vector pBR322. The ligation products were transformed into E. coli DH5α selecting for ampicillin-resistant transformants. Plasmid DNA was prepared from a pool of 26 000 transformants and used to transform the ΔpmrAB deletion strain EG13937. Cells were plated on N-minimal pH 5.8 10 μM MgCl₂ agarose plates containing 50 μg of ampicillin/ml and 200 μM FeCl₃.

Determination of the MIC for Fe(III) and Al(III)

The MIC was determined on N-minimal pH 5.8 10 μM MgCl₂ agarose plates containing twofold dilutions of FeCl₃ or Al₂(SO4)₃. To determine the MIC, bacteria were grown overnight in N-minimal media with 10 mM MgCl₂, pH 7.7 at 37°C, then diluted into N-minimal media with 10 μM MgCl₂, pH 5.8, and organisms were tested at a final inoculum size of 10⁴ colony-forming units (cfu) per μl with the use of a multipoint inoculator (Sakuma Seisakusyo, Tokyo, Japan), following a 20 h incubation at 37°C. The MIC was the lowest concentration of compound that inhibited cell growth. MIC determinations were repeated at least three times.

Iron and soil killing assays, and polymyxin B nonapeptide competition assays

Overnight cultures grown in N-minimal media 10 mM MgCl₂ pH 7.7 at 37°C were washed and used to inoculate N-minimal media 10 μM MgCl₂ pH 5.8. After a 4 h incubation at 37°C, cells were diluted 1:100 in the same media. Fifty microlitres of diluted cells was mixed with 50 μl of FeCl₃ dissolved in the same media to a final concentration of 100 μM FeCl₃. For the competition assay, polymyxin B nonapeptide (Sigma) was added to this mixture at a final concentration of 1.25 mM. For polymyxin B killing assay, 50 μl of diluted cells were mixed with 50 μl of polymyxin B (Sigma) solution dissolved in the same media at a final concentration of 2.5 μg ml⁻¹. The mixtures in U-bottom 96-well microtiter plate were incubated at 37°C with shaking for 120 min. At this time, cells were put on ice, serial dilutions were conducted, 50 μl were plated onto LB agar plates and the number of colonies was recorded following overnight incubation at 37°C.

To investigate killing in soil, bacterial cultures were mixed with an equal volume of soil solution (dissolved in N-minimal media 10 μM MgCl₂ pH 5.8). The mixture was incubated for 5 h at 37°C before bacterial viability was assessed by dilution and plating as described above for the iron killing assay.

⁵⁹Fe association assay

⁵⁹Fe associated with cells was determined using a Packard Cobra II γ counter as described (Chamnongpol et al., 2002). The percentage of ⁵⁹Fe association was calculated as the amount of ⁵⁹Fe per colony-forming unit (cpm cfu⁻¹).

Mass spectrometry analysis of lipid A

Lipid A samples were prepared as described previously (Yi and Hackett, 2000), with a slight modification. Bacteria were grown overnight in N-minimal media with 10 mM MgCl₂, pH 7.7 at 37°C. Then, the cells were washed and diluted into N-minimal media with 10 μM MgCl₂, pH 5.8 to be a final inoculum size of 10⁴ cfu μl⁻¹. Sixty microlitres (60 spots of 1 μl diluted cells) of each strain was spotted onto an N-minimal pH 5.8 10 μM MgCl₂ agarose plate containing 50 μM FeCl₃ with the use of a multipoint inoculator (Sakuma Seisakusyo, Tokyo, Japan), followed by incubation at 37°C for 20 h. The cells were then harvested, washed, and suspended in 0.8 ml PBS, pH 7.4. Preparation of lipid A samples and MALDI-TOF mass spectrometry analyses of lipid A were performed in the negative ion mode on a Voyager DE STR mass spectrometer (PerSptive Biosystems) as described (Lee et al., 2004).

Phosphatase activity assay

Core OS was prepared as described (Yethon et al., 1998). The PhoP-His and PmrG-His proteins were purified using the Ni-NTA Spin kit (Qiagen) as described by the manufacturer. Production of free phosphate was assayed as described (Harder et al., 1994). Purified protein (1 μg) was incubated with core OS substrate for 30 min at 37°C. Free Pi generated in the assay was measured by malachite green assay using BIOMOL GREEN™ reagent (BIOMOL Res Laboratories) and a microplate reader at 620 nm. Data were corrected for background activity by subtracting the absorbance of the samples lacking core OS from those incubated in the presence of core OS.

β-Galactosidase assays

β-Galactosidase activity was determined as described (Miller, 1972).

Localization of the PmrG protein

Periplasmic and cytoplasmic fractions were prepared by the osmotic shock procedure as described (Neu and Heppel, 1965).