|Home | About | Journals | Submit | Contact Us | Français|
Gram-negative bacteria often modify their lipopolysaccharide (LPS) thereby increasing resistance to antimicrobial agents and avoidance of the host immune system. However, it is unclear how bacteria adjust the levels and activities of LPS-modifying enzymes in response to the modification status of their LPS. We now address this question by investigating the major regulator of LPS modifications in Salmonella enterica. We report that the PmrA/PmrB system controls expression of a membrane peptide that inhibits the activity of LpxT, an enzyme responsible for increasing the LPS negative charge. LpxT’s inhibition and the PmrA-dependent incorporation of positively charged L-4-aminoarabinose into the LPS decrease Fe3+ binding to the bacterial cell. Because Fe3+ is an activating ligand for the sensor PmrB, transcription of PmrA-dependent LPS-modifying genes is reduced. This mechanism enables bacteria to sense their cell surface by its effect on the availability of an inducing signal for the system regulating cell surface modifications.
The outer membrane constitutes a permeability barrier that protects Gram-negative bacteria from certain antibiotics and harmful chemicals (Nikaido, 2003). This protection is largely due to the asymmetric nature of the outer membrane, which, in addition to proteins and lipoproteins, consists of phospholipids in the inner leaflet and the glycolipid lipopolysaccharide (LPS) in the outer one. The LPS is a dynamic structure that can be chemically decorated at various positions in response to changes in the environments experienced by a bacterium (Raetz et al., 2007). The extent of LPS decorations must be finely tuned because, while they are critical for resistance against bactericidal agents (Shai, 1999; Vaara et al., 1979) and for evasion of the host immune system (Takeda et al., 2003), the constitutive expression of LPS-modifying genes renders bacteria hypersusceptible to the bile detergent deoxycholate (Froelich et al., 2006). Here we address when and how bacteria adjust the production and activity of LPS-modifying gene products in response to the LPS modification status of their outer membrane.
The PmrA/PmrB two-component system is the major regulator of genes mediating modification of the LPS in the Gram-negative pathogen Salmonella enterica serovar Typhimurium (Gunn, 2008). PmrA-dependent modifications can occur at each of the three structurally distinct regions: the innermost lipid A, the central core region, and the outermost O-antigen (Raetz et al., 2007). These modifications include the incorporation of phosphoethanolamine (pEtN) and L-4-aminoarabinose (L-Ara4N) at the 1 and 4′ positions of lipid A (Figure 1A), which neutralizes the lipid A phosphates and confers resistance to the cationic antibiotic polymyxin B (Groisman et al., 1997; Gunn et al., 1998; Lee et al., 2004; Trent et al., 2001) and to the toxic metals Fe3+ and Al3+ (Nishino et al., 2006; Wösten et al., 2000). In the inner core, decoration of the heptose (I) phosphate with pEtN furthers resistance to polymyxin B (Tamayo et al., 2005a) and removal of the heptose (II) phosphate amplifies resistance to Fe3+ and Al3+ (Nishino et al., 2006). And increase of the O-antigen length results in heightened resistance to serum (Delgado et al., 2006; Pescaretti et al., 2011). The PmrA/PmrB system directly controls transcription of seven LPS-modification loci (i.e., ugd, pbgP, pmrC, pmrG, cptA, wzzST, and wzzFepE; note that pmrC is often referred to as eptA (Herrera et al., 2010) and that pbgP is referred to as pmrHFIJKLM (Gunn et al., 1998) or arn (Breazeale et al., 2003)). In addition, the PmrA/PmrB system uses a yet to be identified gene product to inhibit the activity of LpxT, a constitutively-synthesized inner membrane enzyme that generates diphosphorylated lipid A at the 1-position (1-PP) (Herrera et al., 2010) (Figure 1A). The combination of many of these PmrA-regulated modifications decreases the overall negative charge of the LPS. These modifications appear important for Salmonella pathogenicity because inactivation of the pmrA and/or pmrB genes attenuates virulence in mice (Gunn et al., 2000) and infection of chicken macrophages (Zhao et al., 2002).
The sensor PmrB responds to extracytoplasmic Fe3+ and Al3+ (Wösten et al., 2000) and acidic pH (Perez and Groisman, 2007) by promoting phosphorylation of its cognate regulator PmrA (Shin et al., 2006) (Figure 1B). This results in transcription of PmrA-activated genes (Shin et al., 2006) and repression of PmrA-repressed genes (Kato et al., 2003) because phosphorylated PmrA (i.e., PmrA-P) is the PmrA form that binds to its target DNA sequences in vivo (Shin and Groisman, 2005). Growth in low Mg2+, which is an inducing signal for the PhoP/PhoQ two-component system (Garcia Vescovi et al., 1996), also promotes transcription of PmrA-dependent genes in a process that requires the PhoP-activated PmrD protein (Kato and Groisman, 2004; Kox et al., 2000) (Figure 1B). The PmrA-P protein controls its own levels by positively regulating transcription of the pmrCAB operon (Gunn and Miller, 1996; Soncini and Groisman, 1996), by repressing expression of the pmrD gene (Kato et al., 2003) and by an intrinsic feedback mechanism whereby PmrB dephosphorylates PmrA-P shortly after bacteria begin experiencing Fe3+ (Yeo et al., 2012) (Figure 1B).
We now report that Salmonella dynamically fine-tunes the activation level of the PmrA/PmrB system in response to the modification status of its LPS. We uncover a PmrA-activated membrane peptide that inhibits the LPS-modifying enzyme LpxT thereby hindering the generation of 1-PP lipid A. This inhibition and the PmrA-dependent decoration of the lipid A with L-Ara4N together decrease the negative charge of the bacterial cell surface. Consequently, the amount of Fe3+ bound to the bacterial cell, and thus, available to stimulate the sensor PmrB is lowered, reducing expression of PmrA-dependent genes. The identified feedback mechanism differs from those typically acting on transcriptional regulators both in the timing and the target of control. Our findings suggest that bacteria continuously adjust the levels and activities of the determinants directing cell surface properties even when experiencing constant inducing conditions for the system controlling such determinants.
We identified a PmrA-activated promoter immediately downstream of and transcribed towards the pmrB coding region that had not been reported in previous experimental and computational searches of PmrA-activated targets (Aguirre et al., 2000; Marchal et al., 2004; Wosten and Groisman, 1999) (Figure 2A). The promoter’s PmrA box (i.e., the motif recognized by PmrA) (Aguirre et al., 2000; Marchal et al., 2004; Wosten and Groisman, 1999), which is also found at an equivalent location in the Escherichia coli, Klebsiella pneumoniae and Citrobacter koseri genomes (Figure 2A), was footprinted by the Salmonella PmrA-His6 protein (Figure S1A). S1-mapping experiments revealed a PmrA-dependent transcription start site in wild-type Salmonella grown under conditions that activate the PmrA/PmrB system (Figure S1B). This transcript was not detected in wild2 type Salmonella grown under non-inducing conditions, in a pmrA null mutant or in strains with a deletion in the identified PmrA box or with nucleotide substitutions in the conserved TTAA sequence of the PmrA box, regardless of growth conditions (Figure S1B).
The identified transcript includes a 30-codon ORF preceded by a putative ribosome2 binding site, and these features are conserved in related enteric species (Figures 2A and and2B).2B). The ORF is translated in vivo because an engineered strain with the sequence coding for the FLAG epitope inserted between the last codon and the stop codon of the ORF at the normal chromosomal location produced a product with the expected molecular weight (Figure S1C). The ORF does not exhibit sequence similarity to peptides or proteins of known function. For reasons explained below, it was termed PmrR for Pmr Regulator. Subcellular localization experiments followed by Western blotting with anti-FLAG antibodies demonstrated that a FLAG-His6-PmrR peptide localizes to the inner membrane fraction (Figure 2C), as predicted from the presence of a long stretch of hydrophobic amino acids (Figure 2B).
Activation of the PmrA/PmrB system inhibits the LpxT-dependent generation of 1-PP lipid A (Herrera et al., 2010) (Figure 1A). However, the identity of the PmrA-regulated gene product responsible for this inhibition has remained unknown. Given that both PmrR (Figure 2C) and LpxT (Tatar et al., 2007; Touze et al., 2008) localize to the inner membrane, and that most PmrA-activated gene products of known function are involved in modifying the LPS (Gunn, 2008), we explored the possibility of PmrR being the elusive PmrA-activated LpxT inhibitor.
To examine whether PmrR interacts with LpxT, we used the bacterial two-hybrid (BACTH) system (Karimova et al., 1998; Karimova et al., 2000), which had been previously utilized to uncover interactions between integral membrane proteins and/or peptides (Gerken et al., 2009; Lippa and Goulian, 2009). CO-expression of adenylate cyclase protein fusions to the N-terminus of PmrR and to LpxT resulted in β2 galactosidase activity that was ~10 times higher than that produced by cells carrying the positive control plasmids (Figure 3A). By contrast, there was little activity when the PmrR fusion was cO-expressed with protein fusions to YbjG, which is the closest LpxT homolog in Salmonella; to PmrC, which is a PmrA-activated inner membrane protein (Lee et al., 2004) that competes with LpxT for modification of lipid A (Herrera et al., 2010); or in cells harboring the negative control plasmids (Figure 3A). PmrR’s interaction with LpxT is dependent on the single tryptophan residue at position 25 (Figure 3B). (We targeted W25 because it is conserved in PmrR homologs (Figure 2B), and also because tryptophan residues in transmembrane helices localize preferentially to the interfacial regions of phospholipid bilayers (Yau et al., 1998) and can influence interactions between transmembrane proteins (White and von Heijne, 2005).)
The PmrR-LpxT interaction detected in the BATCH assay was verified in Salmonella expressing the FLAG-His6-PmrR peptide from a heterologous promoter and an LpxT with LacZ fused to its C-terminus from its normal chromosomal location. We pulled down LpxT-LacZ (Figure 3C) by using Ni-NTA affinity chromatography to capture complexes associated with FLAG-His6-PmrR in detergent-solubilized membrane preparations. The pull down is specific for LpxT-LacZ because: First, the CorA protein was not detected even though it localizes to the inner membrane (Snavely et al., 1989) like LpxT (Tatar et al., 2007). And second, LpxT-LacZ was not recovered from extracts prepared from the lpxT-lacZ strain harboring the plasmid vector or from a strain expressing FLAG-His6-pmrR and the wild-type lpxT gene (Figure 3C).
To determine whether PmrR’s interaction with LpxT alters the ability of LpxT to generate 1-PP lipid A, we analyzed the LPS from sets of isogenic strains grown under different conditions. We began by comparing the lipid A profiles of wild-type Salmonella harboring a plasmid that expresses pmrR from a heterologous promoter (or the corresponding plasmid vector) following growth under non-inducing conditions for the PhoP/PhoQ and PmrA/PmrB systems (i.e., high Mg2+). This enabled us to evaluate PmrR’s role in the absence of other PhoP- and PmrA-dependent LPS modifications. The pmrR-expressing strain exhibited decreased amounts of 1-PP and increased levels of pEtN (Figure 3D). PmrR’s action appears to be LpxT-dependent because the profiles of the pmrR-expressing and vector-containing organisms were the same in an lpxT mutant background: 1-PP was absent but pEtN levels were enhanced (Figure 3D). In agreement with previous results (Herrera et al., 2010), the profiles of vector-carrying wild-type and lpxT mutant Salmonella demonstrated that the presence of 1-PP in lipid A requires a functional lpxT gene (Figure 3D).
Next, we examined the lipid A profiles from four isogenic strains with mutations in the pmrR promoter and/or deleted for the lpxT gene following growth under activating conditions for both the PhoP/PhoQ and PmrA/PmrB systems (i.e., low Mg2+ and high Fe3+). This allowed us to evaluate the contributions of PmrR and LpxT in the presence of other PhoP- and PmrA-dependent LPS modifications. wild-type Salmonella produced highly modified lipid A species that were not found in bacteria grown under non-inducing conditions (Figure 3E) (Raetz et al., 2007). By contrast, the pmrR promoter mutant displayed reduced amounts of lipid A doubly-substituted with L-Ara4N and pEtN (Figure 3E). Both the lpxT and lpxT pmrR mutants contained much lower amounts of single pEtN and L-Ara4N modifications but higher levels of lipid A doubly-modified relative to the wild-type strain (Figure 3E) (Herrera et al., 2010). Together these data indicate that PmrR modifies the LPS indirectly by inhibiting the LpxT-dependent formation of 1-PP lipid A. Moreover, these results support the notion that LpxT and PmrC compete for modification of the phosphate at the 1-position of lipid A (Herrera et al., 2010), and influence the incorporation of L-Ara4N into lipid A.
We reasoned that the PmrR-dependent inhibition of LpxT might hamper electrostatic interactions between Salmonella and the positively charged Fe3+. Indeed, the pmrR promoter mutant and a pmrR start codon mutant bound more Fe3+ than the isogenic wild-type strain (albeit less than a strain deleted in the pmrA gene) (Chamnongpol et al., 2002) (Figure 4A). By contrast, the lpxT mutant bound less Fe3+ than wild-type Salmonella (Figure 4A). As expected, a mutant with a constitutively active PmrA/PmrB system (i.e. harboring the pmrA505 allele; (Roland et al., 1993)) showed the least Fe3+ binding (Figure 4A), possibly due to enhanced lipid A substitutions that decrease the negative charge of the LPS.
We hypothesized that the PmrR-promoted decrease in Fe3+ binding might result in diminished transcription of PmrA-dependent genes by virtue of reducing the amount of Fe3+ available to activate the sensor PmrB. To test this hypothesis, we determined the β2 galactosidase activity produced by strains harboring chromosomal lac transcriptional fusions to the PmrA-activated pbgP operon and ugd gene. When bacteria were grown with Fe3+ to activate the PmrA/PmrB system, the pmrR mutants produced higher levels of β-galactosidase than the isogenic pmrR+ strain (Figures 4B and and4C).4C). This behavior reflects neither the assay (i.e., β-galactosidase) nor that the chromosomal ugd-lac strains are unable to produce L-Ara4N (due to a lac insertion in the ugd gene) because the pmrR mutants produced heightened levels of ugd and pmrC mRNAs than the wild-type strain (Figures 4D and and4E).4E). Furthermore, normal ugd-lac expression was restored to a pmrR mutant by a plasmid expressing the pmrR coding region or one encoding a PmrR derivative tagged with FLAG-His6 at the N terminus (Figures 4F and S2A). By contrast, there was no complementation by the plasmid vector or by a plasmid specifying a PmrR tagged with a FLAG-His6 tag at the C terminus (Figures 4F and S2A). PmrR requires a tryptophan at position 25 to downregulate the activity of the PmrA/PmrB system because a plasmid specifying the PmrR W25A protein, which is defective for LpxT interaction (Figure 3B), could not restore normal ugd-lac expression to the pmrR mutant (Figure 4G).
PmrR appears to act exclusively via LpxT because the pmrR mutants no longer stimulated ugd-lac transcription in an lpxT mutant background (Figure 4C). Likewise, the heightened mRNA levels for the ugd and pmrC genes exhibited by the pmrR mutants decreased to wild-type levels upon deletion of lpxT (Figures 4D and and4E).4E). In addition, the enhanced Fe3+ binding exhibited by the pmrR mutants was not observed in an lpxT mutant (Figure 4A). Moreover, we determined that a pmrR ugd degP cpxP quadruple mutant was defective for growth in the presence of Fe3+ (Figure S2B) and that this defect was overcome by inactivation of the lpxT gene (Figure S2B). By contrast, PmrR function does not involve the PmrD protein because expression of ugd-lac was still observed in a pmrD null mutant (Figure S2C). Collectively, these results reinforce the notion that PmrR decreases Fe3+ association to bacteria, confers Fe3+ resistance and downregulates PmrA-dependent gene transcription by targeting the LPS-modifying enzyme LpxT.
We investigated whether other PmrA-activated gene products also exert negative feedback on the PmrA/PmrB system. First, we determined that the ugd2dependent incorporation of the positively charged L-Ara4N into the lipid A is critical for negative feedback. This is because the mRNA levels of the pbgP (Figure 5A) and pmrC (Figure 5B) genes were ~4-fold higher in a ugd mutant than in the wild-type strain. These mRNA levels were even higher in a ugd pmrR double mutant than in the isogenic pmrR strain (Figures 5A and and5B)5B) indicative that the L-Ara4N-mediated feedback is not contingent on the presence of pmrR. The ugd and ugd pmrR mutants bound more Fe3+ than the isogenic wild-type and pmrR strains, respectively (Figure 5C). Therefore, the ugd-mediated L-Ara4N modification of lipid A downregulates the PmrA/PmrB system by decreasing Fe3+ association with the bacterial cell.
Next, we examined whether the PmrC-dependent decoration of lipid A with pEtN impacted Fe3+ binding and transcription of PmrA-activated genes. The pmrC mutant and pmrC pmrR double mutant bound slightly more Fe3+ (Figure 5C) and displayed increased pbgP (Figures 5A) and pmrC (Figures 5B) transcript levels when compared to the wild-type strain. Deletion of the ugd gene in these mutant strains further enhanced mRNA levels of PmrA-activated genes (Figures 5A and and5B).5B). Although the pmrC ugd and pmrC ugd pmrR mutants bound as much Fe3+ as the isogenic ugd and ugd pmrR strains (Figure 5C), expression was lower in the pmrC mutant than in the pmrC+ strains (Figures 5A and and5B),5B), likely because the pmrC mutant is missing the positive feedback that PmrA exerts on the pmrCAB promoter (Wosten and Groisman, 1999).
Negative feedback on the PmrA/PmrB system appears to be exercised by PmrA-dependent gene products that target the lipid A phosphates and not by those modifying other portions of the LPS. On the one hand, a strain defective in the pmrG-mediated dephosphorylation of the core Hep (II) phosphate, a modification required for resistance to Fe3+ (Nishino et al., 2006), retained wild-type levels of Fe3+-binding and of PmrA-activated transcripts (Figure S3). The same was true for a strain lacking the PmrA-activated wzzST gene (Figure S3), which controls the O-antigen length (Delgado et al., 2006; Pescaretti et al., 2011). On the other hand, the increased Fe3+ binding (Figure 5D) and enhanced mRNA levels of the pbgP (Figure 5E) and pmrC (Figure 5F) genes displayed by the ugd mutant was completely suppressed upon inactivation of lpxT. These results demonstrated that the LpxT-dependent formation of 1-PP is critical for Fe3+ binding to the LPS as well as for both PmrR- and L-Ara4N-mediated negative feedback on the PmrA/PmrB system.
PmrA-activated transcripts are produced within 20 min of bacteria experiencing PmrB2 inducing conditions (Shin et al., 2006). Surprisingly, wild-type and ugd Salmonella exhibited virtually identical levels of Fe3+ binding (Figure 6A) and of pbgP mRNA (Figure 6B) when examined 20 min after being switched to Fe3+-containing media. That the ugd mutant behaved like the wild-type strain at 20 min reflects that PmrA-dependent lipid A modifications were undetectable (Figure 6C), possibly because these modifications are added to newly synthesized LPS (Raetz et al., 2007). By 120 min, the PmrA-dependent lipid A modifications were visible (Figure 6C) and wild-type Salmonella bound less Fe3+ (Figure 6A) and expressed lower amounts of pbgP transcript than the ugd mutant (Figures 6B). The L-Ara4N-mediated negative feedback is not limited to PmrA-dependent genes specifying LPS-modifying enzymes, because mRNA levels of the PmrA-activated aroQ (Tamayo et al., 2005b), which encodes a chorismate mutase, are lower in the wild-type than in the ugd mutant at 120 min, even though these strains produce similar amounts of aroQ transcript at 20 min (Figure S4A). The amount of Fe3+ bound by the wild-type strain decreased between 20 and 120 min whereas its resistance to polymyxin B increased >500 fold (Figure 6D), both indications that the bacterial cell surface became less negatively charged. By contrast, the ugd mutant displayed similar susceptibility to polymyxin B whether grown for 20 or 120 min in the presence of Fe3+ (Figure 6D).
We report a dynamic feedback mechanism whereby the activity of a regulatory system is continuously adjusted even when bacteria experience constant inducing conditions. This adjustment results from cellular changes, brought about by the regulatory system, that limit access to an activating signal for the system. Specifically, we determined that Salmonella monitors the modification status of its LPS by the effect it has on the availability of Fe3+, an inducing ligand for PmrA/PmrB, the regulatory system controlling expression of LPS modification genes.
Classical negative feedback typically entails a regulator controlling the expression of a gene product that reduces the levels and/or activity of the regulator (Thomas and D’Ari, 1990). In some cases, negative feedback functions by actively lowering the levels of an inducing signal. For example, lactose is converted to allolactose, which binds to LacI thereby lowering binding to the lac operator and derepressing the lac operon in E. coli; as lactose is metabolized and eventually depleted from the growth media, the lac operon is downregulated (Wilson et al., 2007).
We have now established that negative feedback on a regulatory system can function by controlling its access to an inducing signal. PmrA is the major regulator of genes mediating the decoration of the LPS (Gunn, 2008). Three such decorations are critical for feedback control on its cognate sensor PmrB: preventing the LpxT-mediated generation of 1-PP lipid A (Herrera et al., 2010) (Figures 3D and and3E)3E) and the covalent modification of the lipid A phosphates with L-Ara4N and pEtN (Figure 6C). These decorations decrease the overall negative charge of the bacterial cell surface, which reduces Fe3+ association to the bacterial cell (Figures 4A and and5C).5C). Consequently, less Fe3+ is available for binding to and activating PmrB in the periplasm, which is anticipated to decrease the levels of phosphorylated PmrA (Shin et al., 2006). Such a decrease lowers expression of PmrA-dependent genes (Figures 4 and and5)5) even when Fe3+ levels in the bacterial surroundings do not change.
There is a direct relationship between the amount of Fe3+ bound by the bacterium and the transcriptional output of PmrA-activated genes (Figures 4A, 4D, 4E and and5).5). Different lipid A modifications display dissimilar abilities in neutralizing the negatively charged lipid A (Nikaido, 2003). This may explain why the L-Ara4N modification has a larger effect on Fe3+ binding and activation of the PmrA/PmrB system than substitution with pEtN both in wild-type and pmrR backgrounds (Figure 6). For instance, the pEtN2 modified lipid A species is less abundant than the L-Ara4N2modified species (Zhou et al., 2001) (Figure 3E). Moreover, pEtN decreases the net charge of lipid A from −1.5 to −1 while L-Ara4N brings it down to 0 (Nikaido, 2003).
Our data highlights the critical role that 1-PP lipid A plays in Fe3+ binding and covalent modification of lipid A. This is because inactivation of lpxT hindered the generation of lipid A singly substituted with pEtN or L-Ara4N (Figure 3E), and it eliminated the increased Fe3+ association to the bacterial cell and transcription of PmrA-activated genes displayed by pmrR and ugd mutants (Figures 4A, 4D-E and 5D-F).
Conditions that modify the negative charge of the LPS can alter signal detection in the periplasm not only by PmrB but also by other sensors. On the one hand, the dominant cation neutralizing the negative charges in the LPS is Mg2+ (Nikaido, 2003), which hinders Fe3+-promoted expression of PmrA-dependent genes (Wösten et al., 2000). On the other hand, the activity of the PhoP/PhoQ system is enhanced by mutations that increase expression of the eptB gene in E. coli (Moon and Gottesman, 2009). This is because EptB modifies the outer KDO residue of the LPS core with pEtN (Reynolds et al., 2005), likely decreasing binding of divalent cations, which function as a negative signal for the PhoP/PhoQ system (Garcia Vescovi et al., 1996).
PmrR provides a singular example of a hydrophobic peptide that controls the modification of the lipid A indirectly. It binds to the LPS-modifying LpxT protein, preventing the generation of 1-PP lipid A (Herrera et al., 2010) (Figure 3). This decreases Fe3+ binding and activation of the PmrA/PmrB system (Figure 4). PmrR’s mode of action differs from that of peptides (Eguchi et al., 2007; Gerken et al., 2009; Lippa and Goulian, 2009) and periplasmic proteins (Raivio et al., 1999; Zhou et al., 2011) that directly modulate the output of two-component systems. Our findings add to the diversity of actions promoted by bacterial hydrophobic peptides (Hobbs et al., 2011), which have been shown to further proteolysis of membrane proteins (Alix and Blanc2 Potard, 2008), to stabilize macromolecular protein complexes within the inner membrane (Gassel et al., 1999) and to control the activity of sensor kinases (Eguchi et al., 2007; Lippa and Goulian, 2009). The regulatory mechanism described here is likely to be evolutionarily conserved given that the PmrA/PmrB two-component system governs transcription of LPS-modifying loci in response to extracytoplasmic Fe3+ in a variety of species (Mitrophanov et al., 2008; Viau et al., 2011; Winfield and Groisman, 2004; Winfield et al., 2005) that also specify PmrR (Figure 2A), LpxT and Ugd.
We propose that the activity of the PmrA/PmrB system is controlled by two distinct feedback mechanisms operating in different time scales. On the one hand, there is intrinsic negative feedback whereby the sensor PmrB shifts from being a kinase that promotes the phosphorylated state of PmrA to becoming a PmrA-P phosphatase (Shin et al., 2006; Yeo et al., 2012). This mechanism takes place within one-generation time, before PmrA-dependent LPS modifications are detectable (Figures 6C and S4B). On the other hand, there is the negative feedback mechanism described in this paper whereby changes in the bacterium’s LPS modification status decrease the availability of Fe3+ to the sensor PmrB thereby lowering activation of the PmrA/PmrB system. This mechanism is slower, requiring more than one cell generation (Figure S4A), and results in the progressive adjustment of the PmrA gene expression program (Figure 6).
We propose the following scenario for the feedback mechanism here described. In the absence of Fe3+, PmrA-dependent LPS modification genes are not transcribed and the LPS is negatively charged (Figure 1C). Under these conditions, Salmonella promotes expression of iron acquisition systems that scavenge iron from various sources (FaraldO-Gomez and Sansom, 2003). When bacteria initially transition from iron-poor to iron-replete environments, Fe3+ will associate with the negatively charged LPS and be transported across the outer membrane presumably by iron-uptake systems. Indeed, the crystal structure of the outer membrane ferrichrome-iron receptor FhuA revealed the presence of non-covalently associated LPS molecules (Ferguson et al., 1998), suggesting that the Fe3+ bound to the LPS might be readily accessible to FhuA, and potentially to other siderophore receptors, for its translocation into the periplasm. This would increase periplasmic Fe3+ levels, which would be sensed by PmrB, leading to activation of the PmrA/PmrB system and transcription of PmrA-dependent loci (Figure 1D). Over time, the PmrA-regulated modifications that lower the negative charge of the bacterial surface are incorporated into the lipid A moiety while the expression of iron transport genes is repressed due to rising cytoplasmic Fe2+, which combines with the aporepressor Fur to dampen expression of iron acquisition determinants (FaraldO-Gomez and Sansom, 2003). Thus, less Fe3+ is bound to the LPS and transported into the periplasm, reducing the amount of Fe3+ available for activating the PmrA/PmrB system (Figure 1E).
Finally, our findings indicate that a bacterium’s LPS modification status amplifies the level of Fe3+ sensed by PmrB at early times but reduces them at later time points. This might enable cells to lower the activity of the PmrA/PmrB system upon the alleviation of Fe3+-mediated toxicity through lipid A modifications that decrease Fe3+ binding to the cell surface. Moreover, it links the strength of the cell surface modifications to the activity of the regulatory system. This enables bacteria to monitor changes in their cell envelope and to fine-tune the expression of LPS-modifying enzymes to levels appropriate for the Fe3+ concentration they experience in their environments. Such reciprocal control between a bacterium’s regulatory system and the physiological state of the cell surface may constitute a widespread feedback mechanism, given that a variety of bacterial signal transduction systems participate in regulating membrane composition (Nikaido, 2003).
Bacterial strains and plasmids used in this study are listed in Table S1. Primers used in this study are listed in Table S2. We describe the growth conditions, construction of bacterial strains and plasmids, and other molecular biology procedures in the Supplemental Experimental Procedures.
For data presented in Figures 3A, 4B, 4C and 4F, a kinetic β-galactosidase assay was performed as described by (Camp and Losick, 2009) and in the Supplemental Experimental Procedures. β-galactosidase activity (Arbitrary units) is the rate of ONPG conversion (i.e., Velocity, with units of mOD415 per minute) divided by the OD595 of the bacterial culture.
E. coli DHP1 Δcya harboring derivatives of plasmids pT18 or pT25 were grown overnight in LB containing 100 μg/μl ampicillin and 50 μg/μl chloramphenicol, added at 1:50 dilution to 200 Pl of the same fresh media and grown shaking at 37°C for 24 h before β-galactosidase activities (Arbitrary units) were measured.
E. coli BTH101 cya- harboring derivatives of plasmids pUT18 and pKT25 were grown overnight in LB containing 100 μg/μl ampicillin and 50 μg/μl kanamycin, added at 1:100 dilution to 0.6 ml of the same fresh media containing 0.5 mM IPTG and grown shaking at 37°C for 4 h before β-galactosidase activities (Miller units) were measured.
Isolation of 32P-labeled lipid A was performed as described (Herrera et al., 2010) with the following modifications: 32P-labeled lipid A species (~200 cpm per lane) were analyzed using TLC in a solvent system of chloroform, pyridine, 88% formic acid and water (50:50:16:5, v/v) for 3 h. The plate was exposed to a BAS-MS2040 imaging plate for several days and visualized using an FLA-7000 imaging system (Fujifilm).
Bacteria were grown and then lysed as described in the Supplemental Experimental Procedures. The Quantichrom Iron Assay Kit was used to determine the amount of Fe according to the manufacturer’s instructions.
RNA isolation, cDNA synthesis and reaL-time quantitative PCR analyses were performed as described in (Chen et al., 2011) and results were normalized to 16S ribosomal RNA levels.
We thank Michael Maguire (Case Western Reserve University) for anti-CorA antibodies, Jennifer Aronson for editorial assistance and Andrew Goodman, Gisela Storz and Eunjin Lee for comments on the manuscript. This work was supported, in part, by Grant-in-Aid for Young Scientists (Start-up) 19810025 and (A) 23688013 from the Japan Society for the Promotion of Science (JSPS), the Kato Memorial Bioscience Foundation, the Uehara Memorial Foundation, the Mochida Foundation, and the Inamori Fundation to A.K., and by grant 42336 from the NIH to E.A.G., who is an Investigator of the Howard Hughes Medical Institute.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.