The PmrAB System-inducing Conditions Control Both Lipid A Remodeling and O-antigen Length Distribution, Influencing the Salmonella Typhimurium-Host Interactions*

Background: Salmonella Typhimurium LPS structure is regulated by the PmrAB, PhoPQ, and RcsCDB systems. Results: Wzzst is required for lipid A modifications. PbgE2 and PbgE3 control formation of short O-antigen region. Conclusion: PmrAB system is the master regulator of LPS remodeling, modulating genes that modify lipid A, core, and O-antigen. Significance: Salmonella exhibits complex mechanisms to modulate its LPS, which influences host interaction. The Salmonella enterica serovar Typhimurium lipopolysaccharide consisting of covalently linked lipid A, non-repeating core oligosaccharide, and the O-antigen polysaccharide is the most exposed component of the cell envelope. Previous studies demonstrated that all of these regions act against the host immunity barrier. The aim of this study was to define the role and interaction of PmrAB-dependent gene products required for the lipopolysaccharide component synthesis or modification mainly during the Salmonella infection. The PmrAB two-component system activation promotes a remodeling of lipid A and the core region by addition of 4-aminoarabinose and/or phosphoethanolamine. These PmrA-dependent activities are produced by activation of ugd, pbgPE, pmrC, cpta, and pmrG transcription. In addition, under PmrA regulator activation, the expression of wzzfepE and wzzst genes is induced, and their products are required to determine the O-antigen chain length. Here we report for the first time that Wzzst protein is necessary to maintain the balance of 4-aminoarabinose and phosphoethanolamine lipid A modifications. Moreover, we demonstrate that the interaction of the PmrA-dependent pbgE2 and pbgE3 gene products is important for the formation of the short O-antigen region. Our results establish that PmrAB is the global regulatory system that controls lipopolysaccharide modification, leading to a coordinate regulation of 4-aminoarabinose incorporation and O-antigen chain length to respond against the host defense mechanisms.

The Salmonella enterica serovar Typhimurium lipopolysaccharide consisting of covalently linked lipid A, non-repeating core oligosaccharide, and the O-antigen polysaccharide is the most exposed component of the cell envelope. Previous studies demonstrated that all of these regions act against the host immunity barrier. The aim of this study was to define the role and interaction of PmrAB-dependent gene products required for the lipopolysaccharide component synthesis or modification mainly during the Salmonella infection. The PmrAB two-component system activation promotes a remodeling of lipid A and the core region by addition of 4-aminoarabinose and/or phosphoethanolamine. These PmrAdependent activities are produced by activation of ugd, pbgPE, pmrC, cpta, and pmrG transcription. In addition, under PmrA regulator activation, the expression of wzz fepE and wzz st genes is induced, and their products are required to determine the O-antigen chain length. Here we report for the first time that Wzz st protein is necessary to maintain the balance of 4-aminoarabinose and phosphoethanolamine lipid A modifications. Moreover, we demonstrate that the interaction of the PmrA-dependent pbgE 2

and pbgE 3 gene products is important for the formation of the short O-antigen region. Our results establish that PmrAB is the global regulatory system that controls lipopolysaccharide modification, leading to a coordinate regulation of 4-aminoarabinose incorporation and O-antigen chain length to respond against the host defense mechanisms.
Like other Gram-negative bacteria, the lipopolysaccharide (LPS) in Salmonella enterica serovar Typhimurium (S. Typhimurium) 4 is the major surface constituent located in the outer leaflet of the outer membrane (1,2). This LPS is composed of (i) the hydrophobic lipid A, which anchors LPS to the outer membrane (3); (ii) an oligosaccharide core; and (iii) an O-antigen polysaccharide. In S. Typhimurium, the lipid A is composed of sugars and fatty acids, whereas the O-antigen is a polymer of three to six sugar repeat units extended out from the cell surface (3,4).
To survive microenvironmental changes, pathogenic bacteria are able to remodel their outer membrane, mainly at the level of lipid A and O-antigen. The remodeling of lipid A is produced by palmitoylation and/or deacylation in a PhoPQ-dependent manner and by addition of 4-aminoarabinose (L-Ara4N) and phosphoethanolamine (pEtN) depending on PmrAB system activation (5)(6)(7)(8)(9)(10)(11)(12). These modifications allow bacteria to resist the host immunity barriers such as iron and cationic peptides (13)(14)(15)(16)(17). The PmrAB two-component system consists of the PmrA response regulator and the PmrB sensor, which is able to sense Fe 3ϩ , activating the system (16). This regulatory system can also be activated by low Mg 2ϩ in a PhoPQ-dependent pathway in which the PhoP-activated PmrD protein is required (18,19). These two systems control the above lipid A modification through the regulation of pagP, ugd, pbgPE, and pmrC genes (5,6,8,11,12). It has been demonstrated that expression of the pbgPE operon (also called arn or pmrHFIJKLM operon) and ugd gene, involved in synthesis and incorporation of L-Ara4N into lipid A, are induced by PmrAB activation (6,20,21). In addition, this system controls the expression of pmrC gene, which is responsible for addition of pEtN to lipid A (5,6,8), and pmrG and cptA genes, which modify the core LPS region (22,23). These results suggest that activation of the PmrAB system is strongly necessary to respond against the host immunity barriers.
On the other hand, the normal assembly of S. Typhimurium O-antigen heteropolymer requires functional Wzz fepE , Wzx, Wzy, and Wzz st proteins (3,4). This LPS component plays a direct role in the resistance to phagocytosis, antimicrobial peptides, and serum complement, which are dependent on O-antigen chain length (5, 6, 8, 11, 12, 24 -28). It has been demonstrated that the O-antigen follows a bimodal distribution of its subunits attached to the lipid A-core (4,29,30). This distribution is controlled by the Wzz st protein (also known as Cld, WzzB, or Rol), which mediates the production of long O-antigen containing 16 -35 subunits (L-type) (4,29), and the Wzz fepE protein, which is responsible for the very long O-antigen containing Ͼ100 subunits (VL-type) (30).
We reported previously that the wzz st gene is under control of the PmrA and RcsB regulators (24,25). This result raises the possibility that the Wzz st protein is required in more than one bacterial membrane modification process. In this connection, we demonstrated that when RcsB is activated the Wzz st protein is involved in the negative control of flagella in hyperflagellated bacteria (24). However, no new function was determined for this protein when it is expressed under PmrA activation. On the other hand, in previous studies, we observed that S. Typhimurium displays a third O-antigen region of low molecular weight. This observation and the fact that PmrA also regulates the expression of wzz fepE and wzz st prompted us to investigate the possibility that other PmrA-dependent genes might control the formation of this third O-antigen region.
Taken together, the above observations suggest that there is a very specialized PmrA-dependent regulation mechanism acting on LPS components. This hypothesis led us to study the role of the PmrAB-dependent Wzz st protein in lipid A modification, the participation of other PmrAB-dependent gene products in the synthesis of the third O-antigen region, and their relevance in the Salmonella infection. Our results demonstrate for the first time that Wzz st protein is required to maintain the balance of L-Ara4N and pEtN modifications at the level of lipid A. We also found that the absence of the last two genes of pbgPE operon, pbgE 2 and pbgE 3 , results in an O-antigen without the low molecular weight region (1-15 subunits) and that the interaction of both gene products is necessary to control its formation. The latter allows us to propose that in S. Typhimurium the O-antigen subunits are distributed in the following regions: (i) low molecular weight or short (S), controlled by pbgE 2 and pbgE 3 genes; (ii) long (L), controlled by wzz st ; and (iii) very long (VL), controlled by wzz fepE . Here we conclude that the PmrAB system is the master regulator of the LPS remodeling and that both O-antigen formation and lipid A modifications occur simultaneously during Salmonella infection.
Introduction of Gene Mutations in the Salmonella Chromosome and Plasmid Construction-The one-step gene inactivation method (34) was used to generate deletion of the pbgE2 or pbgE3 complete coding sequence. The Cm cassette was amplified by using pKD3 plasmid as template and primers 7008, 7009, 7010, and 7011 ( Table 2). The correct insertion of the cassette in the mutant was confirmed by direct nucleotide sequencing.
The cloning of the pbgE2 and pbgE3 genes was carried out by PCR amplification using genomic DNA extracted from wildtype S. Typhimurium 14028s strain and primers 1001, 1002, 1003, and 1004 ( Table 2). The PCR products were cloned into pUHE2-2lacI q vector using XbaI restriction enzyme, resulting in derivative plasmids listed in Table 1. The correct orientation of the insert was confirmed by both PCR and direct nucleotide sequencing.
Serum Complement Sensitivity Assay-The analysis of serum sensitivity of the strains grown in N-minimal medium with low Mg 2ϩ was carried out as described (24). Briefly, 10 4 cells/ml were incubated for 1 h at 37°C with PBS buffer as control or with PBS containing 20% human serum (Sigma). The colonyforming units (cfu) produced by each treatment were determined by serial dilutions plated on LB agar medium and incubated at 37°C. The results were expressed as a percentage of the control (strains incubated in PBS buffer) as described previously (24).
Bacterial Infection of Eukaryotic Cells-The strains grown overnight in N-minimal medium with low Mg 2ϩ were used to infect the Raw 264.7 mouse macrophages as described (36). To test the macrophage phagocytic ability, the cells were lysed after 30 min of infection using 1% Triton X-100, and the number of viable bacteria that survived to the gentamicin treatment was determined by subsequent plating onto LB agar medium. The same procedure was used to evaluate replication ability, but the number of viable bacteria was determined after 6 and 18 h of infection. Results were expressed as a percentage of survival to gentamicin calculated as 100 ϫ (cfu ml Ϫ1 mutant bacteria)/(cfu ml Ϫ1 wild-type bacteria) at each time point.
Peptide Killing Assays-The polymyxin survival assay was carried out following the protocol described by Lee et al. (8).
Overnight cultures of bacteria grown in N-minimal medium with high Mg 2ϩ were diluted 1:100 in N-minimal medium with low Mg 2ϩ and grown at 37°C to reach an A 600 of 0.3-0.4. Then 50 l of a 1:100 dilution of these cultures were mixed with 50 l of polymyxin B or E dissolved in PBS at 20 g/ml (final concentration, 10 g/ml) in a 96-well plate. After 1 h of incubation at 37°C, cfu were determined by serial dilution on LB agar medium. The number of viable bacteria was represented as 100 ϫ (cfu ml Ϫ1 mutant bacteria polymyxin-treated)/(cfu ml Ϫ1 wild-type bacteria polymyxin-treated).
Protein-Protein Interaction Assay-This assay was performed as described by Karimova et al. (37). The pbgE 2 , pbgE 3 , and wzz st genes were cloned into the pUT18, pUT18C, pKT25, and pKTN25 plasmids (Table 1). To this end, the gene sequences were amplified by PCR from wild-type S. Typhimurium 14028s genomic DNA using the primers listed in Table 2. The PCR products of pbgE 2 and pbgE 3 were digested and cloned with KpnI/XbaI restriction enzymes, whereas the wzz st PCR product was cloned using KpnI/BamHI. The plasmid derivatives were controlled by DNA sequencing and co-transformed in all compatible combinations into Escherichia coli DHM1 strain. To determine the potential protein interactions, ␤-galactosidase activity from the co-transformant strains grown to stationary phase in LB medium containing 0.5 mM IPTG was measured as described (38). The positive control was generated by transformation of E. coli DHM1 with pKT25-zip and pUT18C-zip plasmids containing the leucine zipper GCN4 domain, whereas the strain harboring empty vectors served as the negative control. ␤-Galactosidase activities were expressed as the mean values of three independent experiments done in duplicate.

Mass Spectrometry Analysis of Lipid A-
The lipid A samples were purified from the indicated bacterial strains grown in N-minimal medium with low Mg 2ϩ as described previously (17). The MALDI-TOF mass spectrometry assay, performed in the negative ion mode on a Voyager DE STR mass spectrometer (PerSeptive Biosystems) equipped with a 337 nm nitrogen laser and delayed extraction, was used for lipid A sample analysis as described previously (17).

The wzz st Gene Product Is Involved in Lipid A Modification in
a PmrAB-dependent Manner-In a previous work, we reported that expression of wzz st is controlled by two regulatory systems, PmrAB and RcsCDB. We also observed that following RcsCDB activation the Wzz st protein participates in the formation of the L O-antigen region and in the bacterial swarming behavior (24). The working hypothesis of this section was that, under PmrAB system activation, the Wzz st protein could be required in other physiological processes. As it is well known that the PmrAB system is required for lipid A modification and that the wzz st gene is located close to ugd, which takes part in the synthesis and attachment of L-Ara4N to lipid A (6, 14, 21), we investigated whether the lipid A is modified by Wzz st . To this end, we used negative ion mode MALDI-TOF mass spectrometry to analyze lipid A species as the [M Ϫ H] Ϫ ions from the wild-type 14028s strain and its isogenic pmrA (EG13307) and wzz st (EG14929) mutants (Table 1). These strains were grown in N-minimal medium at low Mg 2ϩ conditions to promote the transcription of PmrA-activated genes in a PhoPQ-dependent pathway (18,39). Consistent with previous reports (8, 17),

Role of wzz st , pbgE 2 , and pbgE 3 in LPS Modifications
under this assay condition, the wild-type strain (Fig. 1A)   the Wzz st -dependent modifications. These results indicate that expression of wzz st leads to lipid A modifications in a PmrA-dependent manner. Interestingly, a new lipid A peak at m/z 2281 appeared in the wzz st mutant, whereas the species at m/z 2158 was absent (Fig.  1C). Based on previous reports, we suggest that the ion at m/z 2281 could be the result of an extra pEtN residue (Х123 average mass units) added to the 1-or 4Ј-phosphate of lipid A that peaks at m/z 2158. It is important to note that the m/z 2158 species arises from the m/z 1796 ion modified by one pEtN and one palmitic acid group when the pmrC and pagP genes are expressed under PmrAB and PhoPQ activation, respectively (Fig. 1D) (8,9,40). As the chemical structures for most of the lipid A species in S. enterica have been determined previously (41)(42)(43), we confirmed the extra pEtN residue hypothesis by MALDI-TOF analysis of the lipid A purified from pmrC (EG13633), pagP (EG13678), wzz st pmrC (MDs1015), and wzz st pagP (MDs1016) mutants growing in low Mg 2ϩ . As shown in Fig. 2, the m/z 2281 peak is absent in the spectra arising from pmrC and pagP mutants and from pmrC wzz st and pagP wzz st double mutants. As expected, the peaks containing L-Ara4N and/or palmitic acid (m/z 1928, 1944, 1955, 2035, 2051, 2166, and 2182) or modified by L-Ara4N or pEtN (m/z 1928, 1944, 1919, and 1935) were maintained in the pmrC or in the pagP mutants, respectively (Fig. 2, A and C). These results confirm that the m/z 2281 ion is formed by addition of a second pEtN residue to the m/z 2158 species in the absence of Wzz st when the pmrC and pagP genes are expressed. Moreover, we observed that in the wzz st pmrC and wzz st pagP double mutants, as in wzz st , species harboring L-Ara4N (m/z 1928 and 2166) were also absent (Fig. 2, B and D). Taken together, our results suggest that the wzz st gene product is required to maintain the balance between modification in lipid A by L-Ara4N and pEtN, resembling the effect that Wzz st exerts on the Wzx and Wzy balance required for O-antigen long chain determination.
The pbgE 2 and pbgE 3 Gene Products Are Necessary for O-antigen Short Chain Length Determination-Because the above results demonstrated that wzz st is involved in lipid A modification and we previously observed that an O-antigen of low molecular weight was retained in the wzz st and wzz fepE mutants (24,25), we studied the role of other PmrA-controlled genes in the formation of this O-antigen region. To this end, we first investigated the participation of the not well characterized PbgE 2 and PbgE 3 proteins encoded by the last two genes of the pbgPE operon (Fig. 3A) (14,22,44). To test whether pbgE 2 or pbgE 3 deletion affects the O-antigen chain length, we analyzed the distribution of LPS in the strains harboring pbgE 2 or pbgE 3 nonpolar gene deletions (Table 1) when the PmrAB system was activated. We noticed that the LPS obtained from pbgE 2 and pbgE 3 mutants lacked a silver-staining material of low molecular weight as compared with that from wild-type, wzz st , or wzz fepE strains (Fig. 3B). Furthermore, the wild-type O-antigen chain length distribution was restored when pbgE 2 and pbgE 3 mutants were complemented by ppbgE 2 and ppbgE 3 plasmids, respectively (Fig. 3B). These results showed that deletion of pbgE 2 exhibited no polarity effect on pbgE 3 expression. Furthermore, our findings demonstrated that, in addition to Wzz st and Wzz fepE , the pbgE 2 and pbgE 3 gene products are involved in the control of O-antigen chain length distribution. In accordance with a previous report of Hölzer et al. (45), our data allow us to establish that the O-antigen of S. Typhimurium follows a trimodal length distribution and that formation of the poorly studied low molecular weight region named the S region (1-15 sugar subunits attached to lipid A-core) is controlled by the PbgE 2 and PbgE 3 proteins.
Murray et al. (30) found that the domain "PX 2 PX 4 SPK X 1 X 10 GGMXGAG" is strongly conserved in both Wzz fepE and Wzz st proteins. This sequence, located in the C-terminal region and overlapping the second transmembrane domain of both proteins, was found to be essential for their function (46,47). It was of interest to investigate whether this domain is also present in PbgE 2 and PbgE 3 . The bioinformatics analysis of PbgE 2 and PbgE 3 sequences, carried out using the Transmembrane Prediction Server (Stockholm University, Sweden), suggested that both are inner membrane proteins bearing four transmembrane domains comprising the amino acid residues 6 -13, 38 -61, 74 -77, and 97-106 in PbgE 2 and 10 -12, 51-57, 81-91, and 107-119 in PbgE 3 . In addition, the multiple alignment of these sequences with the PX 2 PX 4 SPKX 1 X 10 GGMXGAG domain from Wzz fepE and Wzz st showed a similarity of 48 and 43%, respectively. In both cases, this domain overlapped the third and fourth transmembrane domains comprising residues 49 -92 from PbgE 2 and residues 70 -107 from PbgE 3 . Interestingly, when the alignment was performed only with the Wzz st protein domain, we noticed that PbgE 2 conserved the first part of the above domain, the "PX 2 PX 4 SPK" motif ( Fig. 3C, filled line box), whereas the second portion, the "GGMXGAG" motif, was conserved in PbgE 3 (Fig. 3C, dotted line box). Because the PbgE 2 and PbgE 3 proteins are smaller than Wzz st , we suggest that interaction of both proteins is necessary to control the S O-antigen modal distribution.
The PbgE 2 and PbgE 3 Proteins Are Able to Interact-To test the above notion, we investigated whether PbgE 2 and PbgE 3 act together or individually in the control of the S O-antigen region. We carried out an in vivo protein-protein interaction assay following the bacterial adenylate cyclase two-hybrid (BACTH) protocol (37). We constructed the pUT18, pUT18C, pKT25, and pKTN25 derivative vectors harboring the sequences of the T18 or T25 fragments in-frame with the coding sequences of pbgE 2 and pbgE 3 . To this end, these genes were amplified using wild-type S. Typhimurium 14028s DNA and primers containing XbaI and KpnI restriction sites (Table 2). To investigate interactions with PbgE 2 and PbgE 3 , we also cloned the wzz st gene with primers containing BamHI and KpnI restriction sites to ensure the correct cloning orientation ( Table  2). The resulting plasmids were introduced into E. coli DHM1 in all possible combinations. Functional interaction was determined by measuring the ␤-galactosidase activity produced by the transformed bacteria. The E. coli DHM1 strain co-transformed with empty vectors, which was used as a negative control, showed basal levels of ␤-galactosidase activity (ϳ100 Miller units). A positive control generated by using pKT25-zip and pUT18C-zip plasmids harboring the leucine zipper GCN4 domain displayed high levels of ␤-galactosidase activity (Х6000 Miller units). We observed an increase in ␤-galactosidase activity compared with the negative control when the pbgE 2 and pbgE 3 genes were fused to the C-terminal domain of either T18 or T25 fragments ( Fig. 4; 2252 and 1190 Miller units). Similar results were obtained when both genes were cloned in the N-terminal region of these fragments (data not shown). These increased levels demonstrated that a strong protein-protein interaction occurs when pbgE 2 and pbgE 3 genes are expressed together. In these assays, we observed that PbgE 2 and PbgE 3 are not self-interacting proteins because in the E. coli DHM1 strain harboring the pT25-pbgE 2 /pT18C-pbgE 2 or pT25-pbgE 3 / pT18C-pbgE 3 combinations the ␤-galactosidase activity levels were hardly increased as compared with the negative control (Fig. 4).
On the other hand, our results indicated that Wzz st is unable to interact with itself or with the PbgE 2 or PbgE 3 protein. Here we observed only small differences in ␤-galactosidase activity levels when the corresponding fusion combinations were used (Fig. 4).
Physiological Role of the pbgE 2 and pbgE 3 Gene Products-Previously, it has been reported that bacterial resistance to complement-mediated killing is directly related to O-antigen length, which acts against the formation of a membrane attack complex (2, 26 -28). According to this, the resistance to serum complement of pbgE 2 and pbgE 3 mutants was used to examine the physiological importance of the S O-antigen region in the Salmonella LPS. The susceptibility of isogenic strains growing under PmrA-inducing conditions to serum complement was determined by exposing them to 20% human serum for 1 h. As shown in Fig. 5A, the pbgE 2 and pbgE 3 mutants as well as wzz st were more sensitive to complement-mediated killing than the wild-type strain (20% survival). In addition, when pbgE 2 , pbgE 3 , and wzz st mutants were complemented with plasmids harboring the corresponding genes, the resistance phenotype was restored (Fig. 5A, gray bars). We observed that the resistance of the complemented mutants was higher than in the wild-type strain probably due to an increase in the expression of the genes when they are controlled by the IPTG-inducible promoter of the vector. These results confirmed that the pbgE 2 and pbgE 3 gene products are involved in the formation of the S region, which is required for serum resistance. We reported previously that deletion of pmrA did not abolish resistance to serum complement and that the wzz st and wzz fepE mutants are more sensitive than the wild-type strain but not as sensitive as the pmrA mutant (24,25). Here we noticed that the pbgE 2 and pbgE 3 mutants displayed the same sensitivity patterns as wzz st and wzz fepE ; thus, we concluded that the wzz fepE , wzz st , pbgE 2 , and pbgE 3 gene products have to act in concert to reach full serum resistance of the wild-type strain when the PmrAB system is activated.
As the PmrA regulator is also required during bacterial replication within macrophages (14,15) where both O-antigen and lipid A are involved, we investigated the importance of pbgE 2 and pbgE 3 gene products for Salmonella virulence. To this end, we studied the ability of the wild-type strain (14028s) and pmrA (EG13307), pbgE 2 (MDs1102), and pbgE 3 (MDs1103) mutants growing in low Mg 2ϩ to infect and replicate within Raw 264.7 mouse macrophages. We observed that pbgE 2 and pbgE 3 mutants were less phagocytized (45 and 35%, respectively) by macrophages than wild-type and pmrA strains (Fig. 5B, 0 h). Moreover, the pbgE 2 and pbgE 3 mutants showed a decreased replication ability (80 and 90%) with respect to the wild-type strain (Fig. 5B, 6 and 18 h). It must be noted that the replication ability of the pmrA mutant decreased 18 h postinfection (Fig.  5B). To validate these results, we carried out complementation assays of pbgE 2 , pbgE 3 , and wzz st mutants with ppbgE 2 , ppbgE 3 , and pwzz st plasmids, respectively. We observed that the wildtype phagocytic and replicative capacities were restored in the three mutants (data not shown). These results suggest that pbgE 2 and pbgE 3 gene products play an important role not only in bacterial intracellular replication as happened with Wzz st and Wzz fepE but also in the ability to enter eukaryotic cells (25). These functions could be attributed to their participation in the O-antigen trimodal distribution and in the lipid A modifications.
Because pbgE 2 and pbgE 3 mutants were unable to replicate within macrophages, we investigated whether this phenotype results from defects in the resistance to cationic peptides. To test this possibility, we determined the ability of wild-type strain and pmrA, pbgE 2 , and pbgE 3 mutants to survive the antimicrobial effects of polymyxins B and E. The bacteria were grown in low Mg 2ϩ conditions and then treated with the antibiotics as described under "Experimental Procedures." Strains lacking pbgE 2 or pbgE 3 exhibited less resistance than the wildtype strain to both cationic peptides (Fig. 5C). However, they were more resistant than pmrA (Fig. 5C). These results demonstrated that the pbgE 2 and pbgE 3 mutant replication defects are in part due to their partial susceptibility to cationic peptides. In addition, our findings indicate that these gene products contribute to the wild-type polymyxin resistance under PmrA activation.

DISCUSSION
The O-antigen, the distal region of the Gram-negative bacteria LPS, protects against the bactericidal action of serum complement and cationic peptides (26 -28). Early work  directed to the study of synthesis, composition, and distribution of this LPS portion demonstrated that the O-antigen length is important for the above mentioned protective effect and that the S. Typhimurium O-antigen follows a bimodal distribution pattern (4, 29, 48 -51). In the past few years, our interest has been focused on the regulatory mechanisms that control the expression of genes involved in O-antigen length (24,25). We previously established that wzz fepE (VL O-antigen determinant) is positively regulated by the PmrAB system, whereas expression of wzz st (L O-antigen determinant) is controlled by the PmrAB and RcsCDB systems (24,25). Our findings led us to postulate that the Wzz st protein may play some additional role as occurs with the ugd gene product, which under RcsB activation participates in colanic acid synthesis, but when induced by PmrA it is involved in the incorporation of L-Ara4N into lipid A (6,(52)(53)(54). Previously, we investigated the above assumption for the Wzz st protein when only the RcsCDB system was activated and demonstrated that the wzz st mutation restored the precocious cell swarming behavior of an rcsB mutant (24,55,56). One of the aims of the present study was to examine the functions directed by Wzz st when its gene is controlled by the PmrAB system. We found that upon PmrA activation the L-Ara4N incorporation into specific lipid A species is affected by the absence of wzz st , resulting in the increase of pEtN-lipid A species. We also found that the wzz st mutant displays decreased levels of polymyxin E resistance as compared with the wild-type strain (data not shown). This phenotype may result from loss of the lipid A species modified by L-Ara4N more than from the appearance of a new lipid A containing an extra pEtN group (peak at m/z 2281) (6,23). These observations were consistent with a previous report of Zhou et al. (57). It is important to highlight that only when PmrA is activated is the regulation of wzz st gene required for this new function because no changes were observed in the lipid A profile of the rcsB mutant as compared with that of the wild-type strain (data not shown). Our results confirm that the Wzz st protein is involved in more than one function. We propose that the wzz st gene product is required to maintain the balance between L-Ara4N and pEtN incorporation into lipid A. This probably implies an interaction with one or more proteins to form a complex responsible for L-Ara4N and pEtN synthesis and/or incorporation into lipid A. This would be similar to what is observed with Wzz st , Wzx, and Wzy in the control of L O-antigen region (4,51). Further studies to clarify this issue are currently in progress in our laboratory. Although a bimodal distribution of S. Typhimurium O-antigen controlled by Wzz st and Wzz fepE proteins has been proposed, we observed that in the absence of wzz st and wzz fepE genes an O-antigen portion of low molecular weight was maintained in these mutants (24,25). This observation is in accordance with the reported by Hölzer et al. (45), who demonstrated that the S. Typhimurium O-antigen displays a trimodal distribution in the outer membrane. Taken together, our results not only confirm this O-antigen distribution but also clearly establish that the S region is under control of the pbgE 2 and pbgE 3 gene products. In agreement with our findings, Bennett and Clarke (44) reported that pbgE 2 and pbgE 3 participate in O-antigen synthesis in Photorhabdus luminescens, a Gram-negative bacterium pathogenic to insect larvae.
We demonstrated that there are regions in PbgE 2 and PbgE 3 that display similarity to the PX 2 PX 4 SPKX 1 X 10 GGMXGAG domain present in Wzz st and Wzz fepE proteins required for functional control of VL and L O-antigen (30). We observed that in PbgE 2 the first part of this domain is highly conserved, whereas the second part is conserved in the PbgE 3 protein.
These results and the observation that both proteins are shorter than Wzz st and Wzz fepE suggest that the complete functional PX 2 PX 4 SPKX 1 X 10 GGMXGAG domain might be formed only by interaction of PbgE 2 and PbgE 3 to control the S O-antigen formation. This hypothesis was confirmed by the in vivo protein-protein interaction assay (Fig. 4).
On the other hand, the physiological importance of the pbgE 2 and pbgE 3 genes in O-antigen length distribution was demonstrated by the fact that they confer resistance to complementmediated killing. Here we established that deletion of pbgE 2 and pbgE 3 genes decreased the serum-complement resistance levels as occurs with the wzz st mutant, suggesting that the L and S O-antigen regions act more effectively than the VL region (Fig.  5A). Furthermore, we demonstrated that in pbgE 2 and pbgE 3 mutants both the susceptibility to phagocytosis and the ability to replicate within the host were reduced by several orders of magnitude relative to the wild-type strain (Fig. 5B). Interestingly, this reduction was even more marked than in the pmrA mutant (Fig. 5B). In addition, the pbgE 2 and pbgE 3 mutants were less resistant to polymyxins B and E than the wild-type strain but more resistant than the pmrA mutant. However, these results contradict the previous findings by Gunn et al. (14), who observed that the pbgE 3 strain was as sensitive to polymyxin B as the pmrA mutant. Also, these authors found that loss of pbgE 2 left unchanged the wild-type resistance to this antibiotic. These discrepancies may be due to the different experimental conditions (mainly in the culture media) used. Collectively, our results suggest that the pbgE 2 and pbgE 3 mutations could lead to an attenuated virulence of Salmonella. Similar observations have been made by Bennett and Clarke (44) in P. luminescens.
In summary, in this work, we have identified a novel function for Wzz st in the lipid A remodeling through L-Ara4N and pEtN incorporation. Remarkably, the pbgPE operon products are also involved in this process. Furthermore, here we established that the pbgE 2 and pbgE 3 gene products as well as Wzz st participate in the synthesis of the O-antigen when they are under conditions in which PmrA is activated. These findings support the working hypothesis that guided this study and underscore the importance of the PmrAB system in the LPS modifications that contribute to bacterial adaptation within the host. A challenge for future investigations will be the elucidation of the molecular mechanism involved in the balanced incorporation of the L-Ara4N and pEtN into lipid A, which allows bacteria to survive adverse conditions.