PmrA(Con) Confers pmrHFIJKL-Dependent EGTA and Polymyxin Resistance on msbB Salmonella by Decorating Lipid A with Phosphoethanolamine

The bacterial strains used in this study are listed in Table 1. The P22 mutant HT105/1int201 (obtained from the Salmonella Genetic Stock Center, Calgary, Canada) was used for Salmonella transductions. Salmonella enterica serovar Typhimurium strains were grown on LB-0 or MSB agar or in MSB broth. MSB medium consists of LB (21) with no NaCl and supplemented with 2 mM MgSO₄ and 2 mM CaCl₂. LB-0 is LB medium with no NaCl. These media were used for the growth of strains under nonselective conditions. LB-0 agar was used when genetic selections with antibiotics were performed. Plates were solidified with 1.5% agar. LB-0 agar and MSB broth were supplemented with chloramphenicol (15 μg/ml in broth; 25 μg/ml in agar), EGTA (Sigma, St. Louis, MO; 6 mM or 6.5 mM), streptomycin (200 μg/ml), or tetracycline (3, 5, or 20 μg/ml) as needed. A 350 mM stock of EGTA at pH 8.0 (adjusted with NaOH) was dissolved and then autoclaved. Antibiotics were added to LB-0 agar after cooling it to 45°C. MacConkey Agar Base (Difco) was used to prepare galactose MacConkey agar.

Phenotypes of strains were confirmed by replica plating. Replica plating was performed using the double-velvet technique (17).

A standard protocol for making electrocompetent cells (25) was modified as described previously (22).

Salmonella sp. strain P22 transductions were carried out as previously described (7), except that EGTA was not added to the medium. A Bio-Rad Gene Pulser was used for transformation, following the unique electroporation protocol described previously (22).

Ten milliliters of MSB broth was inoculated from a patch on a master plate, with phenotypes confirmed by replica plating. Cultures were grown in 2.5-cm-diameter glass tubes to an optical density at 600 nm (OD₆₀₀) of 0.4 and held on ice. Once all cultures reached the appropriate OD and sat on ice for at least 15 min, dilutions were made in ice-cold plastic Eppendorf tubes containing MSB broth. The cells were diluted so that approximately 300 cells would be present in a 500-μl aliquot. Polymyxin B sulfate was added to media to achieve a final concentration of 0.1 μg/ml (a 500-μl volume of 2× polymyxin B sulfate in MSB broth was added to a 500-μl aliquot of cells, giving a total of 1 ml in each tube) in a plastic Eppendorf tube for each strain, and they were incubated for 1.5 h in a 37°C incubator without shaking; 500 μl from each tube was spread onto MSB agar. The plates were incubated overnight at 37°C, and CFU were determined.

Lipid A samples were purified from at least two independent cultures, and all independent cultures from a given strain yielded nearly identical lipid A mass spectra. The strains were grown in 500-ml cultures of MSB broth to an OD₆₀₀ of 0.10 or 100 ml cultures of MSB broth without MgSO₄ (pH 8.0) were grown to an OD₆₀₀ of 1.0. (Cultures were stopped when they had grown to an OD₆₀₀ of 0.1 in order to decrease the chance of jackpots with derivatives. Jackpots arise when a suppressor mutation that confers a growth advantage spontaneously occurs early in the growth of a culture and cells with this secondary mutation overgrow the original clone, which lacks the second mutation for faster division. Later, we found that this was unnecessary because the frequency of suppressors is relatively constant in MSB broth between OD₆₀₀s of 0.1 and 1.0. Furthermore, we detected no changes in lipid A structure in our strains grown under both conditions.) Cultures in MSB broth lacking MgSO₄ (pH 8.0) were inoculated with 1:500 dilutions of overnight cultures grown in the same broth, and 1 mM EGTA was added after 40 min of incubation at 37°C. MSB broth cultures were grown with 100 rpm of translational movement, and MSB broth cultures lacking MgSO₄ (pH 8.0) were grown at 125 rpm of translational movement. LPS was purified by the Mg²⁺-ethanol precipitation method as previously described (5). Lipid A was purified by hydrolysis in 1% sodium dodecyl sulfate at pH 4.5 (4). Before being applied on a sample plate, the lyophilized lipid A was dissolved in 20 μl of 5-chloro-2-mercaptobenzothiazole matrix-assisted laser desorption ionization (MALDI) matrix in chloroform-methanol (1:1). Negative-ion MALDI-time of flight (TOF) was performed as described previously (8).

Lipid A was labeled with [³³P]orthophosphate, purified, and analyzed by thin-layer chromatography as previously described (33), except that cells were grown in MSB broth to an OD₆₀₀ of 0.60 instead of 1.0.

Strains were grown as described above. The rapid LPS purification technique for gas chromatography was performed as previously described (27). Fatty acid components of LPS were converted to methyl esters by methanolysis (27).

To determine if mutations in pmrA could suppress msbB growth defects, we transduced loss-of-function and constitutive alleles of pmrA into Salmonella strain ATCC 14028 msbB. As we hypothesized, a constitutive mutation [pmrA505, referred to as pmrA(Con)], but not a loss-of-function mutation, in pmrA partially suppressed EGTA sensitivity in an msbB background (Fig. 2). In contrast, it did not rescue growth on galactose-MacConkey medium.

Polymyxin resistance has previously been correlated with EDTA resistance in pmrA mutants (34). Constitutive activation of pmrA has been shown to result in the modification of the phosphate groups of lipid A with aminoarabinose and/or phosphoethanolamine (9, 13, 30, 39). We propose that these lipid A structural modifications suppress msbB growth defects by decreasing the electrostatic repulsion between neighboring phosphate groups, which would make the cells less dependent on divalent cations, like Mg²⁺ and Ca²⁺.

The Salmonella aminoarabinose operon has been described previously (9, 11). To determine if the aminoarabinose biosynthetic genes are necessary for the suppression of msbB growth defects in the pmrA(Con) strain, we moved our msbB::Ωtet marker into pmrA(Con) strains with nonpolar deletions in pmrH, pmrI, pmrJ, pmrK, and pmrL. As shown in Fig. 3, these nonpolar deletions in pmrH, pmrI, pmrJ, pmrK, and pmrL have an unsuppressed msbB phenotype despite the presence of the pmrA(Con) mutation. In contrast, loss of pmrM, which is not believed to play a role in the aminoarabinose biosynthetic pathway (11), did not alter the EGTA-resistant phenotype of the msbB pmrA(Con) strain (Fig. 3). The rpsL allele present in some of these triple mutants, which yields resistance to streptomycin, does not alter the EGTA or galactose-MacConkey phenotypes (data not shown).

Our results (see below) demonstrate that the msbB mutation confers polymyxin sensitivity on Salmonella strain ATCC 14028 grown in MSB broth, which is essential to avoid enrichment for suppressor mutations. In comparison, another study, using LB broth, also reported that msbB conferred polymyxin sensitivity on Salmonella strain C5 (31). The pmrA(Con) mutation was isolated from a screen for polymyxin-resistant Salmonella in LB medium (26). To confirm that the pmrA(Con) mutation also increases polymyxin resistance in Salmonella strain ATCC 14028 msbB grown in MSB medium, polymyxin survival rates for the various msbB strains tested are shown in Fig. 4. Our results show that the pmrA(Con) mutation confers polymyxin resistance on Salmonella strain ATCC 14028 msbB grown in MSB broth. In a wild-type background, the pmrA(Con) mutation allows ∼73% of the cells to survive 1.5 h of exposure to 3.0 μg/ml polymyxin B sulfate in comparison to ∼1.3% survival in ATCC 14028 (not shown). In an msbB background, the pmrA(Con) mutation allows ∼53% of cells to survive a 1.5-h exposure to 0.1 μg/ml polymyxin B sulfate, whereas ∼0% of Salmonella strain ATCC 14028 msbB bacteria survive this treatment. Thus, although the msbB mutation confers polymyxin sensitivity (31), the pmrA(Con) allele significantly reduces this sensitivity.

Nonpolar deletions in pmrHFIJKL restored both EGTA (Fig. 3) and polymyxin (Fig. 4) sensitivity to Salmonella strain ATCC 14028 msbB pmrA(Con). Loss of pmrM, which has not been shown to play a role in lipid A modification, does not affect either EGTA or polymyxin resistance.

Previous reports of lipid A structure in the wild type and Salmonella pmrA(Con) were from cultures grown in LB broth (9, 11, 18, 31, 32, 39). However, Salmonella msbB must be grown in MSB broth to prevent rapid overgrowth by suppressor mutants (22). To determine if there were any differences in lipid A structure based on growth conditions, we analyzed lipid A from strains grown in LB, LB-no-salt, and MSB broth using negative-ion MALDI-TOF mass spectrometry. The lipid A portion of LPS has been well characterized by mass spectrometry, with published mass-to-charge ratios correlated with distinct chemical structures (12, 38, 39). As shown in Fig. 5A, ATCC 14028 has both hexa-acylated (m/z 1797) and hepta-acylated (m/z 2036) lipid A when grown in LB broth. However, when grown in LB-no-salt broth (Fig. 5B) or MSB broth (not shown), lipid A lacks hepta-acylated lipid A (m/z 2036), showing that palmitoylation is upregulated by the presence of 1% NaCl.

Gas chromatography (Fig. 5E, C ₁₆ column) confirmed that there was an increase (∼2 to 3-fold) in palmitate addition in both 14028 and 14028 pmrA(Con) grown in LB broth compared to LB-no-salt broth. Thus, we see that palmitoylation, in both 14028 and 14028 pmrA(Con), is stimulated by the addition of 1% NaCl to LB-no-salt broth.

Similar investigations were carried out with ATCC 14028 pmrA(Con). As shown in Fig. 5C, the pmrA(Con) mutation, in an msbB⁺ genetic background, conferred aminoarabinose and phosphoethanolamine addition on lipid A. Peaks corresponding to hexa-acylated lipid A (m/z 1797) with phosphoethanolamine (m/z 1920) or aminoarabinose (m/z 1928) and hepta-acylated lipid A (m/z 2036) with phosphoethanolamine (m/z 2159) or aminoarabinose (m/z 2167) were apparent. Phosphoethanolamine addition shifted peaks 123 m/z units, while aminoarabinose addition shifted peaks 131 m/z units. Figure 5D shows the effect of the pmrA(Con) mutation on lipid A structure in rich medium lacking salt (LB-no-salt broth). Under these growth conditions, the pmrA(Con) mutation conferred phosphoethanolamine (m/z 1920) and aminoarabinose (m/z 1928) addition on hexa-acylated lipid A, and very small amounts of heptaacylated lipid A were present, either in LB-no-salt broth (Fig. 5D) or MSB broth (not shown).

To see if the pmrA(Con) mutation has the same effect on both wild-type and MsbB lipid A, we isolated lipid A from Salmonella strain ATCC 14028 msbB and msbB pmrA(Con). As seen in Fig. 6A, msbB lipid A contained a penta-acylated peak (m/z 1588) as its major lipid A species (as expected, since its lipid A lacks myristate, which yields an m/z shift of 210) and contained a minor amount of hexa-acylated lipid A (m/z 1826 due to palmitate addition). In contrast to the effects of the pmrA(Con) mutation on the wild type (Fig. 5D), the pmrA(Con) mutation conferred only phosphoethanolamine but not aminoarabinose addition on msbB lipid A (Fig. 6B). In addition to penta-acylated lipid A (m/z 1588) decorated with phosphoethanolamine (m/z 1711), there was a large increase in palmitate addition (m/z 1826) and some hexa-acylated lipid A with phosphoethanolamine was observed (m/z 1949). Since the pmrA(Con) mutation resulted in lipid A aminoarabinose addition in a wild-type (Fig. 5C and D) but not in an msbB (Fig. 6B) genetic background, we conclude that aminoarabinose is not added to lipid A lacking the myristate residue added by MsbB in MSB medium. A related finding was reported (31) from experiments using Salmonella C5 msbB, showing that aminoarabinose is not added to MsbB lipid A in LB broth. Perhaps the myristoyl group added by MsbB is required for the localization or proper positioning of one of the aminoarabinose biosynthetic enzymes, thereby conferring substrate specificity.

MALDI-TOF mass spectrometry was used to analyze overall lipid A structure in the EGTA-sensitive and EGTA-resistant strains. As seen in Fig. 6A, Salmonella strain ATCC 14028 msbB (EGTA sensitive) had a higher penta-acylated lipid A peak (at m/z 1588), whereas Salmonella strain ATCC 14028 msbB pmrA(Con) (EGTA resistant) (Fig. 6B) had a higher hexa-acylated lipid A peak (at m/z 1826), resulting from the addition of palmitate, as well as phosphoethanolamine peaks (penta-acylated lipid A with phosphoethanolamine at m/z 1711 and hexa-acylated lipid A with phosphoethanolamine at m/z 1949). Our mass spectra suggest that there are differences in terms of relative amounts of lipid A species between the strains. Since mass spectrometry is qualitative, we used thin-layer chromatography and gas chromatography to demonstrate quantitative differences between the amounts of lipid A species. Thin-layer chromatography of lipid A (Fig. 6E) suggested Salmonella strain ATCC 14028 msbB pmrA(Con) has more hexa-acylated (palmitoylated) lipid A (assignment was based on previously published data) (38, 39) than Salmonella strain ATCC 14028 msbB somA (Fig. 6E) or msbB (not shown) (Salmonella strain ATCC 14028 msbB somA and msbB strains have indistinguishable lipid A thin-layer chromatography profiles). Gas chromatography (Fig. 6F, compare C₁₆ bars 1 and 3) showed that the pmrA(Con) mutation results in an ∼7-fold increase in lipid A palmitate incorporation in msbB Salmonella strain ATCC 14028. Thus, three independent techniques indicated that the pmrA(Con) mutation, in an msbB background, increases lipid A palmitoylation. We also challenged Salmonella strain ATCC 14028 msbB with 1 mM EGTA to see if its lipid A structure became modified. As shown in Fig. 6F (C₁₆ column, second bar), EGTA challenge resulted in an ∼2-fold increase in lipid A palmitoylation in EGTA-sensitive Salmonella strain ATCC 14028 msbB. Mass spectrometry of lipid A from EGTA-challenged Salmonella strain ATCC 14028 msbB did not suggest any additional changes (not shown). Escherichia coli responds to EDTA similarly, by palmitoylating its lipid A molecules (14).

Since loss of the aminoarabinose genes conferred EGTA sensitivity on Salmonella strain ATCC 14028 msbB pmrA(Con) (Fig. 3) and we concluded that aminoarabinose cannot be added to msbB lipid A (compare Fig. 5C and D and 6B), we investigated lipid A structures in msbB pmrA(Con) strains with loss-of-function mutations in pmrH, pmrI, pmrJ, pmrK, pmrL, and pmrM to learn what affect these mutations have on lipid A structure in the msbB pmrA(Con) genetic background. The mutations in pmrH, pmrF, pmrI, pmrJ, pmrK, and pmrL are nonpolar deletions (11).

Figure 6C shows that a nonpolar-deletion mutation in pmrF, which restores EGTA sensitivity to Salmonella strain ATCC 14028 msbB pmrA(Con) (Fig. 3), also restores penta-acylated lipid A (peak at m/z 1588) as the dominant form and loses lipid A phosphoethanolamine incorporation (loss of peaks at m/z 1711 and m/z 1949). This decrease in palmitoylation and a loss of phosphoethanolamine incorporation were found consistently in the triple mutants. Lipid A from msbB pmrA(Con) strains with nonpolar deletions in pmrH, pmrI, pmrJ, pmrK, and pmrL reproducibly produce similar mass spectra, and the rpsL mutation in these triple mutants has no obvious effects on lipid A mass spectra (not shown). pmrM, although part of the transcriptional regulon pmrHFIJKLM, is not believed to be part of the aminoarabinose biosynthetic pathway and has a lipid A structural profile similar to that of msbB pmrA(Con) (not shown). Likewise, a loss-of-function mutation in pmrM does not make msbB pmrA(Con) EGTA sensitive (Fig. 3).

Similar lipid A profiles were observed for these strains grown in MSB broth lacking MgSO₄ and in strains grown in this medium challenged with EGTA after entering exponential growth. As mentioned above, the only difference noticed in EGTA-challenged cells was a twofold increase in palmitoylation. Gas chromatography confirmed that loss-of-function mutations in pmrH and pmrF (Fig. 6F, C ₁₆ column) and pmrIJKL (not shown) reduce the levels of palmitate addition to those observed in unsuppressed msbB strains, despite the presence of the pmrA(Con) allele in the triple mutants. These data demonstrate that the aminoarabinose genes affect palmitate incorporation [i.e., as described above, loss of the aminoarabinose biosynthetic genes results in decreased palmitoylation of lipid A in Salmonella strain ATCC 14028 msbB pmrA(Con)].

Lipid A from Salmonella strain ATCC 14028 msbB pmrA(Con) has two distinct molecular differences from lipid A harvested from Salmonella strain ATTC 14028 msbB: increased palmitoylation and phosphoethanolamine addition. Either or both could be responsible for the EGTA and polymyxin resistance phenotypes. In order to address this, we transduced a pagP loss-of-function mutation into Salmonella strain ATTC 14028 msbB pmrA(Con). The msbB pmrA(Con) pagP transductants had EGTA-resistant (not shown) and polymyxin-resistant (Fig. 4) phenotypes that were indistinguishable from those of msbB pmrA(Con) strains, suggesting that palmitate addition is not necessary for EGTA (not shown) or polymyxin (Fig. 4) resistance. To confirm that lipid A phosphoethanolamine addition and no palmitoylation was occurring in these strains, we performed lipid A structural analyses.

Figure 6D shows the lipid A profile from an msbB pmrA(Con) pagP2::Tn10d strain. The spectra in this sample were calibrated differently and are shifted to the left by approximately 18 m/z units. As seen in Fig. 6D, the mutational block in pagP blocks hexa-acylation (no peak at m/z 1806) but not phosphoethanolamine (peak at m/z 1693) addition. The loss-of-function mutation in pagP greatly reduces lipid A palmitoylation (Fig. 6F, last bar). In the pagP mutant, ∼1.5% of total lipid A fatty acids were C₁₆. This suggests that another acyl transferase may be catalyzing the addition of palmitate at a low level or that our lipid A samples were contaminated with palmitate derived from phospholipids. In any case, our mass spectrometry data demonstrate that the dominant form of lipid A in msbB pmrA(Con) pagP strains is penta-acylated lipid A decorated with phosphoethanolamine. Since this strain's lipid A seems to be enriched for phosphoethanolamine, it raises the possibility that palmitoylation may inhibit phosphoethanolamine addition.

The msbB mutation is of clinical interest because it allows Salmonella to be safely administered to mammals, which is essential for live-vaccine or attenuated-bacterial-vector development. Suppressor mutations for msbB allow Salmonella to avoid the septic shock response in a stable genetic background and thus are clinically useful. By choosing a suppressor mutation that confers the desired characteristics of a given product, such as tumor-targeting Salmonella VNP2009 (20, 23) being nontoxic and retaining tumor targeting and tumor inhibition, attenuated bacterial delivery vectors or live-vaccine strains can be genetically optimized.

In this study, we have shown that the aminoarabinose biosynthetic genes are required for the pmrA(Con) mutation to suppress msbB growth defects. Lipid structural analysis of these mutants provided an unexpected result, suggesting that the addition of phosphoethanolamine and palmitate to lipid A, and not aminoarabinose, was responsible for the suppressed phenotype in Salmonella strain ATTC 14028 msbB pmrA(Con). To determine if the addition of phosphoethanolamine or palmitate is responsible for the EGTA- and polymyxin-resistant phenotypes, we created an msbB pmrA(Con) pagP strain that had EGTA- and polymyxin-resistant phenotypes indistinguishable from that of Salmonella strain ATTC 14028 msbB pmrA(Con). This suggests that lipid A phosphoethanolamine, and not palmitate, addition is sufficient and necessary for both EGTA and polymyxin B resistance.

Additional experiments will be needed to determine the mechanism by which the aminoarabinose biosynthetic proteins affect lipid A phosphoethanolamine addition. One possibility is that the aminoarabinose genes are multifunctional, affecting both the aminoarabinose and phosphoethanolamine branches of the pmrA pathway. An alternative hypothesis is that both aminoarabinose and phosphoethanolamine biosynthetic proteins form a complex, and when essential proteins of the complex are not present, the complex may dissociate, resulting in a loss of proper protein localization and loss of activity for both pathways.