Differential Regulation by Magnesium of the Two MsbB Paralogs of Shigella flexneri

Abstract

Shigella flexneri, a gram-negative enteric pathogen, is unusual in that it contains two nonredundant paralogous genes that encode the myristoyl transferase MsbB (LpxM) that catalyzes the final step in the synthesis of the lipid A moiety of lipopolysaccharide. MsbB1 is encoded on the chromosome, and MsbB2 is encoded on the large virulence plasmid present in all pathogenic shigellae. We demonstrate that myristoyl transferase activity due to MsbB2 is detected in limited magnesium medium, but not in replete magnesium medium, whereas that due to MsbB1 is detected under both conditions. MsbB2 increases overall hexa-acylation of lipid A under limited magnesium conditions. Regulation of MsbB2 by magnesium occurs at the level of transcription and is dependent on the conserved magnesium-inducible PhoPQ two-component regulatory pathway. Direct hexanucleotide repeats within the promoter upstream of msbB2 were identified as a putative PhoP binding site, and mutations within the repeats led to diminished PhoP-dependent expression of a transcriptional fusion of lacZ to this promoter. Thus, the virulence plasmid-encoded paralog of msbB is induced under limited magnesium in a PhoPQ-dependent manner. PhoPQ regulates the response of many Enterobacteriaceae to environmental signals, which include modifications of lipid A that confer increased resistance of the organism to stressful environments and antimicrobial peptides. The findings reported here are the first example of gene duplication in which one paralog has selectively acquired the mechanism for differential regulation by PhoPQ. Our findings provide molecular insight into the mechanisms by which each of the two MsbB proteins of S. flexneri likely contributes to pathogenesis.

Shigella flexneri is a gram-negative facultative intracellular pathogen that causes bacillary dysentery and diarrhea by infection of colonic epithelial cells. During human disease, the organism traverses multiple environmentally varied niches within the host. Following ingestion, transit through the intestine to the colon exposes S. flexneri to a broad range of pH and salt conditions. Protection of gram-negative bacteria from the environment is mediated in part by the presence of the outer membrane, the composition of which may be altered in response to environmental signals encountered during infection. The outer membrane is an asymmetric lipid bilayer in which the inner leaflet is composed of phospholipids and the outer leaflet of lipopolysaccharide (LPS), a complex molecule that comprises three distinct moieties: lipid A (endotoxin), core oligosaccharide, and O-antigen polysaccharide. Lipid A, which is the major lipid constituent of the outer leaflet of the outer membrane, anchors the saccharide moieties to the outer membrane. In Escherichia coli and S. flexneri, lipid A is a glucosamine disaccharide modified by six acyl chains: primary acyl chains at the 2, 3, 2′, and 3′ positions and secondary acyl chains attached to the 2′ and 3′ primary chains. Attachment of the secondary acyl chains to the 2′ and 3′ primary chains requires the activity of the late acyltransferases LpxL (HtrB) and MsbB (LpxM), respectively (Fig. 1). The reaction catalyzed by MsbB, transfer of a myristate residue from myristoyl-ACP to penta-acyl lipid A, is the final step in the synthesis of lipid A (6, 14).

FIG. 1.

Enzymatic activities of the late acyltransferases MsbB and LpxL in lipid A biosynthesis. LpxL transfers a secondary acyl chain to the 2′ primary chain, and MsbB transfers a secondary acyl chain to the 3′ primary chain.

This form of hexa-acylated LPS is found in many gram-negative bacteria (46). Lipid A molecules interact extensively with one another, conferring on the outer membrane a quasicrystalline arrangement and low permeability (28, 29, 44). Consequently, many strains carrying mutations in lipid A biosynthetic enzymes display increased sensitivity to drugs, temperature, or both (55, 57). msbB mutants of S. flexneri, pathogenic E. coli, Salmonella enterica serovar Typhimurium, and Klebsiella pneumoniae are each attenuated for virulence (14, 32, 49). Introduction of a constitutive mutation in pmrA into a serovar Typhimurium msbB strain leads to decoration of the lipid A with palmitate and phosphoethanolamine (40) and a concomitant increase in resistance to the chelator EGTA and polymyxin, a cationic antibiotic that binds to lipid A, suggesting that the lipid A defects of the serovar Typhimurium msbB mutant are responsible for the observed sensitivity of this strain to these agents. In E. coli, msbB mutants are neither temperature sensitive for growth nor significantly more susceptible to hydrophobic antibiotics than the wild type (25, 55).

Shigella spp. are unusual in that they possesses two paralogous msbB genes, msbB1, encoded on the chromosome, and msbB2, encoded on the large virulence plasmid that is present in all pathogenic shigellae (56, 59). MsbB1 and MsbB2 are 69% identical, and each catalyzes the transfer of myristate to lipid A (14). In a rabbit ileal loop model of S. flexneri infection, a mutant carrying disruptions of both msbB1 and msbB2 is significantly defective compared to the wild-type strain in the elicitation of tumor necrosis factor alpha production, fluid accumulation, and destruction of intestinal villi (14). Mutants carrying a single mutation in msbB1 or msbB2 are each more defective in these phenotypes than the wild-type strain but less defective than the double mutant (14). Thus, both MsbB proteins are required for full virulence, indicating that neither protein is fully redundant with the other in vivo.

We hypothesized that regulation of the activity of MsbB2 in response to environmental conditions encountered in the host might be different from that of MsbB1. We report that myristoyl transferase activity due to MsbB2 is detected during growth in limited magnesium medium, but not in replete magnesium, whereas that due to MsbB1 is detected under both conditions. Activity of MsbB2 in limiting magnesium is associated with an increase in the transcription of msbB2, but not msbB1, under these conditions. We also demonstrate that the magnesium dependence of MsbB2 activity is controlled by the conserved PhoPQ two-component signal transduction pathway. Thus, the acquisition of a second msbB gene has enabled S. flexneri to enhance its protective outer membrane layer in response to specific environmental signals.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

Bacterial strains and plasmids used in this study are listed in Table 1. Bacterial strains were grown in Luria-Bertani (LB) broth, tryptic soy broth (TCS), or N-minimal medium (42). E. coli K-12 strain JM83 and its msbB derivative, BMS67C12, were gifts of R. Darveau. N-minimal medium was supplemented with 0.2% Casamino Acids, 0.2% glucose, 0.5 mg ml⁻¹ thiamine, 0.5 mg ml⁻¹ tryptophan, 0.5 mg ml⁻¹ nicotinic acid, and MgSO₄ as indicated below in the text, buffered to pH 7.4 with 100 mM HEPES, pH 7.4. To assure maintenance of the large virulence plasmid, cultures of S. flexneri strains were set up from colonies that were red on tryptic soy agar containing 0.01% (wt/vol) Congo red. Where appropriate, antibiotics were used at the following concentrations: ampicillin, 100 μg/ml; kanamycin, 50 μg/ml; chloramphenicol, 25 μg/ml; tetracycline, 12.5 μg/ml; spectinomycin, 100 μg/ml.

TABLE 1.

Strains and plasmids used in this study

Strain or plasmid	Genotype or description	Reference or source
E. coli strains
β2155	thrB1004 pro thi strA hsdS lacZΔM15 (F′ lacZΔM15 lacIqtraD36 proA+proB+) ΔdapA::erm pir::RP4 Ermr Kmr	10
BMS67C12	JM83 msbB::Tn5, Kmr	49
JM83	F−ara Δ(lac-proAB) rpsL [φ80 dlacΔ(lacZ)M15], Strr	62
SM10λpir	C600 thi-1 thr-1 leuB6 supE44 tonA21 lacY1 recA::RP4-2-Tet::Mu λpir Kmr	36
S. flexneri strains
BS109	2457T galU::Tn10 Tetr	34
MBG283	2457T icsA::Ω Specr	51
SC577	M90T msbB1::Km msbB2::Amp Kmr Ampr	14
SRG10	YSH6000T icsA::Ω Specr	This work
SRG15	YSH6000T msbB2::Amp Ampr	This work
SRG19	YSH6000T msbB1::Km Kmr	This work
SRG20	SRG15 msbB1::Km Ampr Kmr	This work
SRG30	YSH6000T Pshf-lacZ Tetr	This work
SRG31	SRG10 phoP::pSRG63 Cmr Specr	This work
SRG32	YSH6000T phoP::pSRG63 Cmr	This work
SRG33	SRG15 phoP::pSRG63 Ampr Cmr	This work
SRG34	SRG19 phoP::pSRG63 Kmr Cmr	This work
SRG41	SRG30 phoP::pSRG63 Tetr Cmr	This work
YSH6000T	Wild-type serotype 2a	45
YSH6000T galU::Tn10	YSH6000T galU::Tn10 Tetr	This work
YSH6000T Pshf-lacZ	YSH6000T Pshf-lacZ Ampr	This work
YSH6000T Pshf*-lacZ	YSH6000T Pshf*-lacZ Ampr	This work
YSH6000T Pshf-lacZ phoP	YSH6000T Pshf-lacZ phoP::pSRG63 Ampr Cmr	This work
YSH6000T Pshf*-lacZ phoP	YSH6000T Pshf*-lacZ phoP::pSRG63 Ampr Cmr	This work
YSH6000T Pshf-lacZ phoP p-SRG41	YSH6000T Pshf-lacZ phoP::pSRG63 pACYC184-PphoP-phoPQ Ampr Cmr Tetr	This work
Plasmids
pACYC184	Cloning vector, Tetr Cmr	5
pAWY6	pFSV-1-lacZ Tetr	This work
pBlueScriptII KS+	Cloning vector, Ampr	Stratagene
pDH95	pBR322-aad9 Specr	Gift of D. Hava and A. Camilli
pFSV-1	Suicide vector, Tetr Cmr	4
pGP704	Suicide vector, Ampr	36
pGP704-Pshf-lacZ	pGP704-Pshf-lacZ Ampr	This work
pGP704-Pshf-lacZ	pGP704-Pshf-lacZ Ampr	This work
pRW50	Broad-host-range, low-copy-number lacZ expression vector, pRK2501 ori Tetr	31
pSRG20	pBlueScript II KS(+) Pshf	This work
pSRG20*	pBlueScript II KS(+) Pshf(T−138A, A−144T)	This work
pSRG23	pAWY6-Pshf-lacZ Tetr	This work
pSRG23*	pAWY6-Pshf(T−138A, A−144T)-lacZ, Tetr	This work
pSRG26	pACYC184-Pshf-shf-wabB-virK-msbB2 Cmr	This work
pSRG33	pGP704-msbB2 Cmr	This work
pSRG37	pACYC184-Pshf-msbB2 native promoter, Cmr	This work
pSRG40	pRW50-Pshf-lacZ Tetr	This work
pSRG41	pACYC184-PphoP-phoPQ Tetr	This work
pSRG42	pBluescript II KS(+)-Pshf Ampr	This work
pSRG42*	pBluescript II KS(+)-Pshf(T−138A, A−144T) Ampr	This work
pSRG50	pBluescript II KS(+)-PmsbB1, Ampr	This work
pSRG51	pBluescript II KS(+)-PmgtA, Ampr	This work
pSRG52	pBluescript II KS(+)-shf108-512, Ampr	This work
pSRG54	pRW50-PmsbB1-lacZ Tetr	This work
pSRG55	pRW50-PmgtA-lacZ Tetr	This work
pSRG56	pRW50-shf108-512-lacZ Tetr	This work
pSRG58	pSRG41-aad9 Specr	This work
pSRG63	pFSV-1-phoP65-624 Cmr	This work
pSRG70	pRW50-Pshf*-lacZ Tetr	This work
pSRG74	pACYC184-aad9 Specr	This work
pSRG76	pACYC184-PphoP-phoP Tetr	This work

Construction of S. flexneri strains.

An msbB1::kan mutation was introduced into wild-type strain YSH6000T by transduction of the kanamycin resistance locus from SC577, which is 2457T msbB1::kan msbB2::amp (gift of P. Sansonetti) (14), using phage P1L4, thereby generating strain SRG19. To obtain high-titer lysates on SC577, it was necessary to complement the msbB2::amp mutation with msbB2 in trans on pSRG26.

A strain in which msbB2 was disrupted was generated by single-crossover integration of a suicide plasmid carrying an internal fragment of msbB2. A 739-bp internal fragment of msbB2 (nucleotides +78 to +810 relative to the translation initiation codon) was amplified by PCR and cloned into pGP704 (36), thereby generating pSRG33. pSRG33 was moved from SM10λpir into YSH6000T galU::tet by conjugation, and transconjugants were selected on medium containing ampicillin and tetracycline and screened for Congo red binding. YSH6000T galU::tet was generated by introduction of the tetracycline resistance locus from BS109 (34) into YSH6000T by phage transduction. The msbB2::pSRG33-amp locus from YSH6000T galU::tet was then introduced into wild-type strain YSH6000T by phage transduction, thereby generating strain SRG15. An msbB1 msbB2 YSH6000T double mutant (SRG20) was generated by introduction of the msbB1::kan locus from SC577 into SRG15 by phage transduction.

A strain in which phoP was disrupted was generated by single-crossover integration of a suicide vector carrying an internal fragment of phoP. A 566-bp internal fragment of phoP (nucleotides +65 to +624 relative to the translation initiation codon) was amplified by PCR and cloned into pFSV-1 to generate pSRG63. This plasmid was moved from SM10λpir into SRG10 by conjugation. Transconjugants were selected on medium containing chloramphenicol and spectinomycin and screened for Congo red binding. The phoP::pSRG63-cm locus from this strain (SRG31) was then introduced into wild-type (WT; YSH6000T), msbB2 (SRG15), and msbB1 (SRG19) S. flexneri strains via phage transduction, thereby generating a phoP mutant in a clean background (SRG32) and msbB2 phoP (SRG33) and msbB1 phoP (SRG34) strains, respectively.

Verification that the correct mutation was present in each strain was performed by PCR. For experiments in which the expression from β-galactosidase transcriptional reporter plasmids was determined in these strain backgrounds, fresh transformants were used.

Construction of reporter plasmids.

β-Galactosidase reporter plasmids were constructed by cloning DNA sequences upstream of the β-galactosidase gene in the low-copy-number plasmid pRW50, as described previously (60). A transcriptional fusion of the promoter region of the shf-wabB-virK-msbB2 operon to lacZ was generated by amplifying by PCR the 359-bp fragment containing nucleotides −335 to +24 relative to the translation start codon of shf. This DNA fragment was first cloned into the EcoRI and HindIII sites of pBluescript II KS(+) (Stratagene) to yield pSRG42 and then subcloned into pRW50 to yield pSRG40. Disruption of the predicted PhoP binding site within the shf promoter was performed by site-directed mutagenesis of the appropriate nucleotides on pSRG42 using QuikChange (Stratagene); the resultant promoter was designated shf* and the resultant plasmid pSRG42*. The DNA fragment containing shf* was then subcloned into the EcoRI and HindIII sites upstream of lacZ in pRW50 to yield pSRG70. Verification that this and all other cloned DNA fragments were correct was performed by sequence analysis.

A transcriptional fusion of the promoter region of msbB1 to lacZ was generated by amplifying by PCR the 478-bp fragment containing nucleotides −467 to +11 relative to the translation start codon of msbB1. This DNA fragment was first cloned into the BamHI and HindIII sites of pBluescript II KS(+) (Stratagene) to yield pSRG50 and then subcloned upstream of lacZ in pRW50 to yield pSRG54.

A transcriptional fusion of the promoter region of mgtA to lacZ was generated by amplifying by PCR the 477-bp fragment containing nucleotides −453 to +24 relative to the translation start codon of mgtA. This DNA fragment was first cloned into the EcoRI and HindIII sites of pBluescript II KS(+) (Stratagene) to yield pSRG51 and then subcloned upstream of lacZ in pRW50 to yield pSRG55.

As a control, a transcriptional fusion of an internal region of shf to lacZ was generated by amplifying by PCR a 404-bp fragment containing nucleotides +108 to +512 relative to the translation start codon of shf. This DNA fragment was first cloned into the EcoRI and HindIII sites of pBluescript II KS(+) (Stratagene) to yield pSRG52 and then subcloned upstream of lacZ in pRW50 to yield pSRG56.

Construction of shf::lacZ S. flexneri reporter strains.

Fusions of shf::lacZ and shf*::lacZ were generated at the shf-wabB-virK-msbB2 locus on the S. flexneri large virulence plasmid. For the generation of the shf::lacZ fusion, the coding sequence for lacZ was amplified by PCR, using as a template pRW50, and was cloned into pFSV-1 (4), thereby generating pAWY6. Then, a DNA fragment that contains the shf promoter (nucleotides −1442 to +48 relative to the shf translation start codon) was amplified by PCR and cloned into pBluescript II, creating pSRG20. For the generation of the shf*::lacZ fusion, the corresponding DNA fragment that contains the shf* promoter (nucleotides −1442 to +48 relative to the shf translation start codon) was amplified by PCR, using pSRG42* as template, and cloned into pSRG20, thereby generating pSRG20*. The DNA fragments encoding the shf and shf* promoters were each cloned into pAWY6, generating pSRG23 and pSRG23*, and the DNA fragments that encompass the shf promoter regions and the lacZ gene were then subcloned into pGP704, thereby generating pGP704-P_shf-lacZ and pGP704-P_shf*-lacZ. Each of these plasmids was then introduced into wild-type S. flexneri YSH6000T by conjugation, and single-crossover recombination events were selected by plating on ampicillin. The presence of the mutant promoter upstream of lacZ was verified by PCR and sequencing the appropriate sequences around the integrated fusion site. The phoP mutation was introduced into YSH6000T P_shf-lacZ and YSH6000T P_shf*-lacZ by phage transduction.

Construction of complementing vectors.

The msbB2 gene was cloned under the control of its native promoter. First, the entire shf-wabB-virK-msbB2 locus, from bp −310 relative to the shf translation start codon to bp 156 downstream of the msbB2 stop codon, was amplified via PCR from YSH6000T and cloned into pACYC184 to create pSRG26. Using oligonucleotide primers that were directed outwardly from the 5′ end of shf and the 3′ end of virK, the shf promoter region, the plasmid backbone, and the coding sequence for msbB2 were amplified as a single DNA fragment by PCR, the ends of which were ligated, thereby generating a plasmid (pSRG37) in which the coding sequences of shf-wabB-virK were deleted and placing msbB2 immediately downstream of the promoter.

The phoPQ locus was cloned under the control of its native promoter. The locus, from bp −180 relative to the phoP translation start codon to bp 110 downstream of the phoQ stop codon, was amplified by PCR and cloned into pACYC184 to create pSRG41. Then, the coding sequence for aad9, which encodes spectinomycin resistance, was cloned from pDH95 (gift of D. Hava and A. Camilli) into the BamHI site of pSRG41, thereby disrupting the tetracycline resistance gene and generating pSRG58. pSRG76, a complementing vector that encodes phoP alone under the control of the native promoter, was constructed by deleting from pSRG58 an internal fragment of phoQ that extends from bp 73 to bp 1182 relative to the phoQ start codon, thereby generating pSRG76. A control vector, pSRG74, was generated by first cloning the coding sequence for aad9 from pDH95 into the BamHI site of pACYC184 and then deleting the coding sequence for cat from the resultant plasmid.

The sequences of primers used in the construction and verification of all plasmids and strains are available upon request.

Transcriptional organization of the shf-wabB-virK-msbB2 locus.

The transcriptional organization of the shf-wabB-virK-msbB2 locus was determined by semiquantitative reverse transcriptase PCR (RT-PCR). RNA was prepared as described previously (11), and contaminating DNA was digested using DNA-free (Ambion, Austin, TX). An RNase inhibitor, RNaseOUT (Invitrogen, Carlsbad, CA), was routinely added to solutions containing aqueous RNA as per the manufacturer's instructions. RT-PCR was performed using the OneStep RT-PCR (Qiagen, Valencia, CA) according to the manufacturer's protocol. To control for DNA contamination of the RNA preparation, the reaction was performed in an identical manner, except that the reverse transcriptase was inactivated by preheating the reaction mix to 95°C for 15 min. Heating the reaction mix does not inactivate the polymerase. The sequences of primers used in this analysis are available upon request.

Lipid A analyses.

³²P-labeled lipid A was prepared in microscale quantities from overnight cultures of bacterial strains grown in the presence of 5 μCi ml^{−1 32}P, as orthophosphoric acid (Perkin-Elmer, Boston, MA), on a roller drum at 37°C, via Bligh/Dyer extraction as described previously (65). For analysis of the role of the magnesium concentration in lipid A synthesis, strains were grown in supplemented N-minimal medium containing either 10 mM or 8 μM MgCl₂. Since strains SRG33 and SRG34 grow slowly in medium containing 8 μM MgCl₂, in experiments involving these strains bacterial growth was continued 6 to 7 additional h, cells were harvested, washed, and stored frozen overnight, and lipid extraction was performed the following day. Antibiotics were typically not added to growth media except to select for maintenance of plasmids or the phoP::cm locus, which is unstable in medium containing 8 μM MgCl₂. For analysis of the role of pH in lipid A synthesis, lipid A was prepared from overnight cultures of bacterial strains grown in LB medium buffered to pH 5.4 with 100 mM morpholineethanesulfonic acid or to pH 7.4 with 100 mM morpholinepropanesulfonic acid. The buffering capacity of each medium was determined to be sufficient to maintain the medium at the desired pH throughout the course of the experiment.

The ³²P-labeled lipid samples were dried under reduced pressure at 45°C in a desiccator (SpeedVac; Savant). The lipid A species were dissolved in CHCl₃-MeOH (4:1 [vol/vol]) and resolved by thin-layer chromatography (TLC). For each strain under each experimental condition, an amount of lipid A equivalent to 1,000 cpm was loaded onto silica gel 60 TLC plates (EMD Chemicals, Gibbstown, NJ). TLC plates were developed in CHCl₃-pyridine-88% formic acid-water (50:50:16:5 [vol/vol]). TLC plates were imaged with a Storm 820 PhosphorImager (Amersham) and analyzed with ImageQuant software (Amersham). The radioactive signal was measured for each spot and was normalized to the loading in the lane by dividing by the total radioactive signal for the lane. The relative signal from each lipid species was then expressed as the signal intensity relative to that of the corresponding species of the wild-type strain following growth in replete magnesium. The total bis-phosphate-specific signal and the total monophosphate-specific signal each varied minimally among the lanes on the TLC, indicating that the distribution of lipid A molecules into bis-phosphate and monophosphate species did not vary significantly among the different strains and different growth conditions.

Quantification of transcription from β-galactosidase reporter constructs.

Levels of transcription from the shf-wabB-virK-msbB2 promoter, derivatives of this promoter, and controls were determined by measuring β-galactosidase activity of transcriptional fusions of the appropriate DNA sequences to a β-galactosidase reporter. β-Galactosidase activity was measured in overnight cultures as the enzymatic hydrolysis of ortho-nitrophenyl galactoside, using the Miller protocol, as described previously (23). For each construct, transcriptional activity is reported as the mean ± standard deviation of β-galactosidase activity in each of three or more independent experiments. Statistical differences were determined using Student's t test.

RESULTS

MsbB2, but not MsbB1, modifies lipid A in a magnesium-dependent manner.

Two observations about the intracellular enteric pathogen S. enterica serovar Typhimurium suggested to us that the expression and/or activity of one or both of the two S. flexneri MsbB enzymes might be regulated by magnesium concentration, pH, or both. First, modification of serovar Typhimurium lipid A is regulated by magnesium concentration and pH (16). Second, immediately upstream of msbB2 in S. flexneri is virK, a gene of unknown function. In serovar Typhimurium, transcriptional activity of virK is increased following growth in acidic medium containing low concentrations of magnesium (13). Although virK is not linked with msbB2 in serovar Typhimurium, we postulated that similar mechanisms of regulation might exist for S. flexneri virK and that these might apply to the entire shf-wabB-virK-msbB locus.

We generated strains carrying a disruption of msbB1, msbB2, or both msbB1 and msbB2, as described in Materials and Methods. msbB1 is at the position of msbB in the E. coli backbone (59), and by analogy to E. coli, is likely organized as a single-gene operon (Fig. 2A, top). msbB2 is the distal gene in the four-gene locus shf-wabB-virK-msbB2 (Fig. 2A, bottom). Because of the organization of the two loci, neither of these mutations is likely to have polar effects on downstream genes.

FIG. 2.

Organization of the S. flexneri msbB1 and msbB2 loci. (A) Organization of the chromosomal locus containing msbB1 and the virulence plasmid locus containing msbB2. The promoter indicated upstream of msbB1 is by analogy to E. coli msbB. The promoter indicated upstream of shf is based on the presence of monocistronic mRNA, as determined in panel B. Primers used for RT-PCR are indicated by arrows numbered 1 to 6. (B) Analysis of the operon organization of the shf-wabB-virK-msbB2 locus by RT-PCR. Agarose gel electrophoresis of products from RT-PCRs was performed using primer pairs as indicated above the lanes. rrnA controls are indicated above the lanes. Reactions were carried out with polymerase in the presence of either active RT or heat-inactivated RT, as discussed in the text and indicated below the lanes. MW markers are indicated in kilobases to the left of the gel.

We tested whether differential regulation of MsbB1 and MsbB2 occurs in response to magnesium concentration (Fig. 3). Complete lipid A of S. flexneri has the same structure as lipid A of E. coli (14). Therefore, to permit the accurate assignment of lipid A species of S. flexneri, lipid A from wild-type and msbB1 msbB2 S. flexneri was compared with lipid A from wild-type and msbB E. coli (Fig. 3A), for which definitive species assignments have been published (57, 65). In the solvent used to separate the species on TLC, hexa-acyl lipid A migrates more rapidly than penta-acyl lipid A, and dephospho species migrate more rapidly than phosphorylated species. As expected, lipid A from the wild-type strains of the two organisms was identical and lipid A from the msbB mutants of the two organisms was identical. In lipid A from the wild-type strains, hexa-acyl species overwhelmingly predominated, whereas in lipid A from strains lacking functional msbB genes, penta-acyl species overwhelmingly predominated (Fig. 3A). Whereas the relative abundance of species having different numbers of acyl chains can be accurately analyzed using this technique, the relative abundance of the different phosphoryl species is not significant, since loss of phosphoryl moieties can occur during the preparation of lipid A.

FIG. 3.

MsbB2 modifies lipid A in a magnesium-dependent manner. Results shown are for TLC analysis of ³²P-labeled lipid A, following growth in modified N-minimal medium (42) buffered with 100 mM HEPES, pH 7.4, and supplemented with 0.2% Casamino Acids, 0.2% glucose, 0.5 mg ml⁻¹ thiamine, 0.5 mg ml⁻¹ tryptophan, 0.5 mg ml⁻¹ nicotinic acid, and MgSO₄ as indicated. (A) Lipid A profiles of WT and msbB1 msbB2 S. flexneri (YSH6000T and SRG20, respectively) used in this study compared to those of WT and msbB E. coli K-12 (JM83 and BMS67C12, respectively). Assignment of lipid A species, indicated to the left of the TLC, was based on published assignments for E. coli K-12 strains (57, 65). (B to D) Regulation of MsbB2 activity by magnesium. Data shown are representative of results from three individual experiments. (B) TLC of lipid A prepared from S. flexneri WT (YSH6000T), msbB1 (SRG19), msbB2 (SRG15), and msbB1 msbB2 (SRG20) grown in N-minimal medium supplemented with either 10 mM or 8 μM Mg²⁺. (C) Relative signal intensity from each lipid species from the wild-type strain following growth in 10 mM (replete) magnesium, as a percentage of the total signal from that lane. (D) Signal intensity of each lipid species examined relative to that of the corresponding species from the wild-type strain grown in 10 mM magnesium; the signal from each species of the wild-type strain grown in 10 mM magnesium was arbitrarily designated 1.0. (E) Rescue of the lipid A profile of msbB1 msbB2 S. flexneri in 8 μM magnesium, but not 10 mM magnesium, by expression of msbB2 in trans. The relatively low abundance of pyrophosphoryl lipid A species in panels B to D compared with panel A likely reflects the different growth media used for the two experiments (minimal medium and rich medium, respectively).

To test the role of magnesium limitation on myristoyl transferase activity of MsbB1 and MsbB2, the lipid A species from ³²P-labeled S. flexneri wild-type, msbB1, msbB2, and msbB1 msbB2 strains grown in limiting (8 μM) or replete (10 mM) Mg²⁺ were isolated and resolved by TLC (Fig. 3B to D). These concentrations of magnesium were chosen because growth of S. enterica serovar Typhimurium in media containing these concentrations leads to differential modification of lipid A (16). After growth in replete magnesium, the msbB1 mutant reproducibly produced predominantly penta-acyl lipid A, and the relative amounts of hexa-acyl and penta-acyl lipid A that were produced by the msbB1 mutant were indistinguishable from those produced by the msbB1 msbB2 double mutant (Fig. 3B and D), indicating that hexa-acyl lipid A produced as a result of the activity of MsbB2 is undetectable in replete magnesium. In contrast, after growth under limited magnesium conditions, the msbB1 mutant reproducibly produced substantially more hexa-acyl lipid A than the msbB1 msbB2 double mutant (Fig. 3B and D), indicating that activity due to MsbB2 is detected in limited magnesium. The amount of hexa-acyl lipid A produced by the msbB1 mutant in limited magnesium was nevertheless substantially less than that produced by the wild type (Fig. 3B and D), indicating that activity due to MsbB2 does not fully compensate for the absence of MsbB1 under these conditions. Based on densitometry of a representative TLC, the distribution of signal into different lipid A species for the various strains under the two conditions is indicated (Fig. 3C and D). It should be noted that the densitometry signal depends not only on the relative amount of the species but also on the number of phosphates bound to the species. The signals from individual species are expressed as the intensity relative to that of the corresponding species of the wild-type strain following growth in replete magnesium (Fig. 3D), after correction for loading; although these numbers do not directly correlate with the absolute amounts of individual species, they permit an analysis of the relative amounts of the corresponding species from the various strains under the two conditions. The results generated in these studies indicate that magnesium concentration differentially regulates MsbB2. This regulation could be at the level of MsbB2 protein production and/or MsbB2 enzymatic activity.

In contrast, in both replete and limited magnesium, the msbB2 mutant reproducibly produced substantially more hexa-acyl lipid A than the msbB1 msbB2 double mutant (Fig. 3B and D), indicating that myristoyl transferase activity due to MsbB1 is detected under both replete and limited magnesium conditions. However, under limited magnesium conditions, the msbB2 mutant reproducibly produced substantially less hexa-acyl and substantially more penta-acyl lipid A than was produced by the wild-type grown in replete magnesium, indicating that under limited magnesium, MsbB1 alone is not sufficient to bring the level of hexa-acyl lipid A up to that which is achieved by the wild-type strain under replete magnesium. As described above, under these conditions (limited magnesium), MsbB2 contributes significantly to the hexa-acylation of the lipid A. Thus, activity due to MsbB2 is present under limited magnesium and absent under replete magnesium growth conditions, whereas activity due to MsbB1 is present under both conditions.

To confirm that the observed magnesium-dependent regulation of acyltransferase activity in the msbB1 mutant was due to the absence of MsbB2, we complemented the msbB1 msbB2 double mutant with msbB2 in trans. The presence of msbB2 in the msbB1 msbB2 background led to a rescue of hexa-acyl lipid A production following growth in limited concentrations of magnesium, but not with replete concentrations of magnesium (Fig. 3E), confirming that activity due to MsbB2 is detected in limited but not replete magnesium growth conditions. The slight increase in hexa-acyl lipid A that was observed in the presence of msbB2 in replete magnesium is likely the result of a partial loss of regulation due to overexpression of MsbB2, since the complementing msbB2 is encoded on a high-copy-number vector.

S. enterica serovar Typhimurium virK is regulated by both pH and magnesium concentration (13). Therefore, we also examined whether MsbB2 in S. flexneri was regulated by pH, independent of its regulation by magnesium. ³²P-labeled lipid A was prepared from S. flexneri wild-type, msbB1, msbB2, and msbB1 msbB2 strains grown at pH 7.4 or pH 5.4, in buffered medium, so as to maintain the desired pH throughout the course of the experiment. These are the same pH conditions shown to regulate S. enterica serovar Typhimurium virK (13). As shown in Fig. 4, the lipid A profiles of the msbB1 mutant and the msbB2 mutant were indistinguishable from that of the wild type, indicating that MsbB1 and MsbB2 are not regulated by pH.

FIG. 4.

pH independence of MsbB1 and MsbB2 modification of lipid A. Results shown are from TLC analysis of ³²P-labeled lipid A prepared from S. flexneri WT (YSH6000T), msbB1 (SRG19), msbB2 (SRG15), and msbB1 msbB2 (SRG20) grown in LB buffered to either pH 7.4 with 100 mM morpholinepropanesulfonic acid or pH 5.4 with 100 mM morpholineethanesulfonic acid. The LB contained approximately 0.45 mM magnesium. Assignment of lipid A species is indicated to the left of the TLC.

Organization of the msbB operons.

Given that MsbB2 function was regulated by magnesium concentration, we wished to test whether the regulation of MsbB2 by magnesium was occurring at the level of transcription. To determine whether the four genes within the locus containing msbB2, shf-wabB-virK-msbB2, were transcribed as an operon, we analyzed mRNA transcripts from this locus in wild-type S. flexneri using reverse transcriptase PCR. Primer pairs were designed to amplify across the junctions of each pair of adjacent genes within the locus, as outlined in Fig. 2A. The presence of a product for each of these primer pairs demonstrated that the four genes are transcribed monocistronically (Fig. 2B). To control for possible contamination of the RNA preparation with DNA that might serve as a template for the polymerase, the reaction was performed in parallel following inactivation of the reverse transcriptase, but not the polymerase, by heating the reaction mix; no product was obtained (Fig. 2B, right lanes), indicating that DNA was not serving as a template. Thus, the promoter upstream of the shf gene drives transcription of msbB2, although we cannot eliminate the possibility that transcription from promoters internal to the four-gene cluster is also occurring.

Transcriptional activity of the operon that contains msbB2 is regulated by magnesium concentration.

We tested whether regulation by magnesium of MsbB2 activity was occurring at the level of transcription of the operon that contains msbB2, by measuring transcriptional activity of the promoter region upstream of shf, following growth in replete or limiting concentrations of magnesium. Transcription from the shf promoter region, as measured by β-galactosidase activity from an episomal P_shf-lacZ transcriptional fusion carried in wild-type S. flexneri, was increased 4.4-fold following growth in limited versus replete magnesium (Table 2), indicating that transcription of the operon containing msbB2 is regulated by magnesium. In contrast, transcription from the msbB1 promoter region was not altered by magnesium concentration (Table 2; induction ratio, 1.0), consistent with our observation that MsbB1 activity is not regulated by magnesium. Under limited magnesium conditions, transcription from the shf promoter was greater than 10-fold higher than transcription from the msbB1 promoter, possibly reflecting relatively higher levels of expression of MsbB2 than MsbB1 under these conditions. As expected, transcription of mgtA, which encodes a high-affinity magnesium transporter ATPase that is induced in limited magnesium (26, 48), increased significantly upon growth in limited magnesium (Table 2; induction ratio, 30.8), and transcription of a DNA sequence internal to the shf coding sequence led to no difference in lacZ expression after growth in limited versus replete magnesium (Table 2; induction ratio, 1.0). Thus, msbB2, and not msbB1, is regulated by magnesium at the level of transcription.

TABLE 2.

Transcriptional activities of episomal lacZ fusions to msbB promoters

Strain genotype	Promoter fusion	β-Galactosidase activity (Miller units, mean ± SD) when grown in:		Induction ratioa
		10 mM Mg2+	8 μM Mg2+
Wild type	Empty vector	67 ± 9	50 ± 5	0.7
	shf	690 ± 150	3,010 ± 150	4.4d
	msbB1	230 ± 14	240 ± 14	1.0
	mgtA	60 ± 7	1,850 ± 220	30.8b
	shf108-512	420 ± 62	420 ± 40	1.0
	shf*	250 ± 23	740 ± 32	3.0d
phoP p-empty vector	Empty vector	50 ± 8	<1 ± <1	ND
	shf	290 ± 93	250 ± 150	0.9e
	msbB1	260 ± 42	260 ± 100	1.0
	mgtA	20 ± 19	160 ± 210	8.0b,c
	shf108-512	430 ± 46	470 ± 110	1.1
	shf*	260 ± 45	270 ± 80	1.0e
phoP p-phoPQ	Empty vector	40 ± 3	<1 ± <1	ND
	shf	580 ± 130	2,850 ± 390	4.9f
	msbB1	220 ± 9	210 ± 20	1.0
	mgtA	40 ± 6	1,410 ± 170	35.3c
	shf108-512	400 ± 21	380 ± 20	1.0
	shf*	300 ± 32	560 ± 40	1.9f

^aRatio of β-galactosidase activity in 8 μM Mg²⁺ to that in 10 mM Mg²⁺. ND, not determined.

^bP = 0.01 for the comparison of mgtA in the wild type to mgtA in the phoP p-empty vector strain.

^cP = 0.001 for the comparison of mgtA in the phoP p-empty vector to the phoP p-phoPQ strain.

^dP = 0.02 for the comparison of shf to shf* in the wild-type strain.

^eP > 0.05 for the comparison of shf to shf* in the phoP p-empty vector strain.

^fP = 0.01 for the comparison of shf to shf* in the phoP p-phoPQ strain.

To test whether the observed transcriptional regulation of the msbB2 locus by magnesium was occurring in the native genetic context, a transcriptional fusion of the shf promoter region to lacZ was integrated at the native locus on the S. flexneri large virulence plasmid (strain YSH6000T P_shf-lacZ). In this strain, β-galactosidase activity was 2.0-fold higher following growth in limited magnesium compared to growth in replete magnesium (Table 3), indicating that regulation of the msbB2 locus by magnesium also occurs in the native context. Whereas these results indicate that magnesium regulates msbB2 at the level of transcription, they do not eliminate the possibility that additional regulation of MsbB2 by magnesium could also be occurring posttranscriptionally.

TABLE 3.

Transcriptional activities of integrated genomic lacZ fusions to the shf promoter

Strain genotype	Promoter fusion	β-Galactosidase activity (Miller units, mean ± SD) when grown in:		Induction ratioa
		10 mM Mg2+	8 μM Mg2+
Wild-type p-empty vector	shf	320 ± 34b	650 ± 130b,c	2.0d
phoP p-empty vector	shf	220 ± 33	170 ± 29c,e	0.8d,f
phoP p-phoPQ	shf	330 ± 110	530 ± 170e	1.6f
Wild type	shf	420 ± 73g	860 ± 26g,h	2.0i,j
phoP	shf	230 ± 18	140 ± 11h	0.6i,k
Wild type	shf*	250 ± 30	270 ± 23	1.1j
phoP	shf*	230 ± 34	150 ± 4	0.7k

^aRatio of β-galactosidase activity in 8 μM Mg²⁺ to that in 10 mM Mg²⁺.

^bP = 0.026 for the comparison of shf in 10 mM Mg²⁺ to shf in 8 μM Mg²⁺ in the wild-type p-empty vector strain.

^cP = 0.013 for the comparison of shf in the wild-type p-empty vector to the phoP p-empty vector strain, each in 8 μM Mg²⁺.

^dP = 0.0009 for the comparison of induction ratios for shf in the wild-type p-empty vector to the phoP p-empty vector strain.

^eP = 0.037 for the comparison of shf in the phoP p-empty vector to the phoP p-phoPQ strain, each in 8 μM Mg²⁺.

^fP = 0.01 for the comparison of induction ratios for shf in the phoP p-empty vector to the phoP p-phoPQ strain.

^gP = 0.005 for the comparison of shf in 10 μM Mg²⁺ to shf in 8 μM Mg²⁺ in the wild-type strain.

^hP = 0.00001 for the comparison of shf in the wild-type strain to shf in the phoP strain, each in 8 μM Mg²⁺.

ⁱP = 0.009 for the comparison of induction ratios for shf in the wild-type strain to the phoP strain.

^jP = 0.02 for the comparison of induction ratios for shf and shf*, each in the wild-type strain.

^kP = 0.3 for the comparison of induction ratios for shf and shf*, each in the phoP strain.

MsbB2 modification of lipid A is dependent on the PhoPQ two-component regulators.

The transcriptional regulation of many bacterial genes by magnesium is mediated by the PhoPQ two-component regulatory system (12, 15, 24, 26, 30, 37, 43, 64). Moreover, in serovar Typhimurium, certain lipid A modifications are magnesium regulated (16), and virK is regulated by PhoP (41). We therefore postulated that the magnesium-dependent regulation of MsbB2 activity and msbB2 transcription that we observed in S. flexneri might be dependent on PhoPQ.

As in E. coli and S. enterica serovar Typhimurium, phoP and phoQ are organized in an operon in S. flexneri. We introduced a phoP mutation into the msbB1 and the msbB2 mutants. To test the role of PhoPQ in regulation of MsbB2, we compared the lipid A profiles of the wild type, the phoP mutant, the phoP msbB1 mutant, the phoP msbB2 mutant, and the msbB1 msbB2 mutant, each grown in limited or replete magnesium (Fig. 5). In limited magnesium, the msbB1 phoP mutant produced predominantly penta-acyl lipid A, indicating that, in the absence of PhoP, activity due to MsbB2 is undetectable in limited magnesium. In contrast, under these same conditions, both the msbB1 mutant and the msbB1 phoP mutant complemented with phoPQ in trans produced near-wild-type levels of hexa-acyl lipid A (Fig. 3B and 6, respectively), indicating that MsbB2 is dependent on PhoPQ for its activity in limited magnesium. In these experiments, we complemented the phoP mutation with phoPQ; however, in experiments described below, complementation with phoP alone rescued the phenotype of a phoP mutant as well as complementation with phoPQ. In both limited and replete magnesium, the msbB2 phoP mutant produced near-wild-type levels of hexa-acyl lipid A (Fig. 5), indicating that activity due to MsbB1 is independent of PhoPQ under both growth conditions. Thus, MsbB2, but not MsbB1, is dependent on PhoPQ under limited magnesium growth conditions.

FIG. 5.

MsbB2 modification of lipid A is dependent on PhoP. Results shown are from TLC analysis of ³²P-labeled lipid A prepared from S. flexneri WT (YSH6000T), phoP (SRG32), msbB1 phoP (SRG24), msbB1 phoP p-phoPQ, msbB2 phoP (SRG33), and msbB1 msbB2 (SRG20) grown in N minimal medium supplemented with either 10 mM or 8 μM Mg²⁺. Each strain except msbB1 phoP p-phoPQ carries pSRG74, which is the empty vector control for p-phoPQ. Assignment of lipid A species is indicated to the left of the TLC.

FIG. 6.

Putative PhoP binding site within the shf promoter. (A) Nucleotide sequence of the region of the shf-wabB-virK-msbB2 promoter that contains the putative PhoP binding site. Numbering adjacent to the sequence represents the position relative to the translation start codon of shf. Putative −35 and −10 boxes are indicated above the sequence. Putative PhoP binding site direct repeats are indicated by arrows below the sequence. (B) Alignment of putative PhoP binding site within the shf-wabB-virK-msbB2 promoter with a consensus for PhoP binding sites generated from PhoP binding sites in E. coli and S. enterica serovar Typhimurium (30, 37). Mutations of the putative shf-wabB-virK-msbB2 PhoP binding site used in this study are indicated below the alignment.

To determine whether PhoPQ regulation of MsbB2 was occurring at the level of msbB2 transcription, we examined the effect of the phoP mutation on transcriptional activity from the shf-wabB-virK-msbB2 promoter. In the phoP background, β-galactosidase activity from the episomal P_shf-lacZ transcriptional fusion was unaffected by magnesium concentration (Table 2, induction ratio 0.9). Moreover, the basal level of expression in replete magnesium was approximately half the basal level of the same fusion in the wild-type background (Table 2) (290 ± 93 versus 690 ± 150 Miller units, respectively), indicating that PhoP enhances basal expression in the bacterial population even in replete magnesium. Both the dependence on magnesium and the level of basal expression were rescued by expression of phoPQ in trans (Table 2, induction ratio 4.9) (basal expression in replete magnesium, 580 ± 130 Miller units). Rescue by expression of phoP alone in trans was equivalent to that by expression of phoPQ in trans (data not shown). Induction of β-galactosidase activity from the mgtA promoter was also significantly diminished in the absence of phoP (Table 2) (P = 0.01 for the wild type compared with the phoP mutant and P = 0.001 for the phoP mutant compared to the complemented strain), consistent with previously published data indicating that it is dependent on PhoP (26, 50), whereas β-galactosidase activity from the msbB1 promoter was unaltered in the absence of PhoP (Table 2, induction ratios of 1.0 in the wild-type, the phoP mutant, and the complemented strain), indicating its independence of PhoP under these conditions. The absence of an effect of PhoP on transcription of msbB1 indicates that the effects of the phoP mutation on msbB2 and mgtA transcription are not a result of nonspecific global effects of the phoP mutation on transcription. The observed residual magnesium regulation of mgtA is likely due to direct effects of magnesium on mgtA expression, perhaps through the 5′ untranslated region of mgtA mRNA, which has been shown to respond to magnesium concentration (9). These results demonstrate that the induction of msbB2 transcription in limited magnesium is dependent on PhoPQ, whereas msbB1 transcription is independent of PhoPQ under the conditions used.

To test whether the observed transcriptional regulation of the msbB2 locus by PhoPQ was occurring in the native genetic context, we examined the effect of the phoP mutation on transcription from the native shf promoter. In the absence of PhoP, induction of transcription from the shf promoter at the native locus by limited magnesium conditions was significantly reduced (Table 3) (P = 0.0009). Basal transcription under replete magnesium conditions from the shf promoter at the native locus was reduced 1.5- to 2-fold compared to the same fusion in the wild-type background (Table 3), indicating that, as for the episomal fusion, PhoP enhances basal expression in replete magnesium in the bacterial population at the native locus. Complementation with phoPQ in trans rescued both the level of induction under limited magnesium conditions (Table 3) (P = 0.01) and the basal level of transcription under replete magnesium conditions (Table 3). Thus, at the native locus, PhoP is required both for induction of transcription of msbB2 in response to limited magnesium and for basal levels of transcription of msbB2.

Identification of a putative PhoP binding site in the promoter of the shf-wabB-virK-msbB2 operon.

Activation of transcription by the PhoP response regulator occurs by binding of a dimer of phosphorylated PhoP to direct hexanucleotide repeats in the promoters of cognate genes. The consensus sequence for PhoP binding sites of magnesium-regulated genes of E. coli and S. enterica serovar Typhimurium is (T/G)GTTTAN₅(T/G)GTTTA (26, 30, 37), although several variations (submotifs) of the PhoP box are also present and functional in these organisms (67). We identified a sequence in the promoter region of the shf-wabB-virK-msbB2 operon, located at −133 to −150 relative to the translation start codon of shf, for which 11 of 12 nucleotides are identical to the consensus PhoP binding site and the spacing of the repeats is appropriate (Fig. 6A and B). As for many PhoP binding sites (37), a putative −35 box overlies the upstream repeat (Fig. 6A). To test whether this sequence is required for PhoP regulation of shf-wabB-virK-msbB2 transcription, we changed two of the nucleotides that are among those most highly conserved within the repeats but which are not within the overlying putative −35 box (Fig. 6B, T⁻¹³⁸A and A⁻¹⁴⁴T). Of note, the adenine at position 6 within the hexanucleotide, which was changed to thymine within the upstream repeat in the mutant, is critical for PhoP-dependent transcription from mgtA in E. coli (61). Induction of transcription in limited magnesium from the mutant shf-wabB-virK-msbB2 promoter (shf*) was significantly reduced compared to the wild-type promoter (P = 0.02) (Tables 2 and 3). This difference in transcription induction between the shf* and shf promoter fusions was lost in the phoP mutant background (P > 0.05) (Tables 2 and 3) and was restored by complementation with phoPQ in trans (P = 0.01) (Tables 2 and 3). These results indicate that disruption of the direct repeats interferes with regulation by PhoP, suggesting that the direct repeats are a PhoP binding site. The residual activation observed for the mutant promoter may represent residual binding of PhoP to the site, since certain PhoP-dependent promoters that contain only a single hexanucleotide sequence are able to bind PhoP (30).

DISCUSSION

In this work, we demonstrate that the two paralogous msbB genes of S. flexneri are differentially regulated by extracellular magnesium concentration. msbB2, the paralog encoded on the large virulence plasmid, is activated under limited magnesium by the PhoPQ two-component regulatory pathway. msbB1, the paralog encoded on the chromosome, is active under both limited and replete magnesium growth conditions and is independent of PhoPQ. To our knowledge, this is the first example of differential regulation in which one paralog has selectively acquired the mechanism for regulation by PhoPQ. Our findings are consistent with the previously published observation that the two S. flexneri MsbB proteins are only partially redundant with one another in vivo (14), and the results provide molecular insight into the mechanisms by which each likely contributes to S. flexneri pathogenesis. The presence of a second copy of msbB that is regulated differently from the ancestral chromosomal copy suggests that the acquisition of the second copy enhanced the ability of Shigellae to survive.

Although we have shown that limited magnesium induces expression of msbB2 in vitro, the niche in which msbB2 is induced and the precise inducer in vivo remain to be demonstrated. The requirement for both paralogs of msbB for full virulence in rabbits (14) suggests that conditions for msbB2 induction are encountered during host infection and that the function of the enzyme is critical to the survival of the pathogen under those conditions. In serovar Typhimurium, PhoPQ is activated within macrophage phagosomes (1) and for serovar Typhimurium and Yersinia pseudotuberculosis is required for survival within macrophages (1, 17). In vitro, divalent cations, such as magnesium, are able to activate S. enterica serovar Typhimurium PhoPQ signal transduction by binding the highly acidic periplasmic sensor domain of PhoQ (15). In the absence of bound divalent cation, PhoQ undergoes a conformational change (15), triggering a phosphorelay that activates the cytoplasmic response regulator PhoP. Activated PhoP binds to cognate promoters (37), thereby inducing or repressing PhoP-regulated genes. Moreover, for the PhoPQ-regulated serovar Typhimurium gene mgtA, regulation by magnesium occurs both at the level of transcription initiation, via PhoPQ, and at the level of transcriptional elongation by interaction of magnesium with the mgtA 5′ untranslated region (9). Activation of PhoPQ in RAW264.7 macrophage phagosomes is triggered by pH and proton pump inhibition (33); whether magnesium might activate PhoP at some point during macrophage infection by serovar Typhimurium is uncertain.

In S. flexneri, we show that induction of msbB2 in vitro occurs in limited magnesium and is independent of pH, suggesting that at least some of the environmental triggers of PhoPQ activation may differ for serovar Typhimurium and S. flexneri. Such differences may have evolved in response to the distinct intracellular lifestyles of the two organisms. In distinction from serovar Typhimurium, shortly after its uptake by phagocytosis S. flexneri lyses the phagosomal membrane (7, 47), such that it is exposed to the phagosomal environment for only a few minutes. Another signal that has been shown to activate the serovar Typhimurium PhoPQ pathway is cationic antimicrobial peptides (2, 35), an important arm of the innate immune response. In addition to being present in the phagosome of macrophages, antimicrobial peptides are secreted into the intestinal lumen by Paneth cells (58). Binding of antimicrobial peptides to the periplasmic sensor domain of PhoQ is thought to displace bound divalent cations, thereby triggering the phosphorelay and PhoP activation of cognate promoters. The PhoPQ response is triggered by exposure to sublethal concentrations of these peptides, and the resultant alteration in expression of structural molecules confers resistance to these peptides (22). PhoPQ signaling is also required for resistance of S. flexneri to antimicrobial peptides (38). Thus, while additional studies will be necessary to directly determine the precise timing of induction of S. flexneri msbB2 during infection, a potential trigger of PhoPQ in the intestinal lumen is antimicrobial peptides.

Cellular Toll-like receptors (TLRs), which recognize a variety of microbial molecules, signal the presence of pathogens and trigger a set of host defenses. LPS is a signal that is recognized by TLR4. A recent report demonstrated that penta-acylated LPS from an E. coli msbB mutant antagonized signaling by hexa-acylated E. coli LPS via TLR4 (8), suggesting that the state of LPS acylation is critical to appropriate recognition of the pathogen by the host. This may in part explain the reduced inflammation seen following infection with S. flexneri msbB mutants (14). Precisely how such a defect in TLR signaling per se would alter the course of the infection is unclear.

In S. enterica serovar Typhimurium, remodeling of the outer membrane that occurs upon activation of PhoPQ makes the cell more impermeable to a variety of small molecules and large lipophilic substances (39). We speculate that MsbB2 is critical for S. flexneri survival during host infection. Specifically, we speculate that the absence of msbB2 and/or msbB1 leads to a decrease in hexa-acylation of lipid A that may be associated with increased bacterial susceptibility to antimicrobial peptides and perhaps other killing mechanisms present in macrophages and monocytes or other sites in the host. Decreased survival of the organism in monocytes and macrophages might contribute to the decreased production of proinflammatory cytokines and decreased inflammation that have been observed in vivo (14).

It is noteworthy that, in limited magnesium, neither MsbB1 nor MsbB2 alone is sufficient to produce levels of hexa-acylated lipid A that are equivalent to those produced by the wild-type in replete magnesium (Fig. 3B and D). This suggests that S. flexneri maintains two myristoyl transferases to ensure that lipid A is maximally hexa-acylated under conditions where the activity of the chromosomally encoded paralog is insufficient. Serovar Typhimurium has only one copy of msbB, which has not been found to be regulated by PhoPQ or magnesium, and increased hexa-acylation per se is not triggered by the PhoPQ pathway. However, several distinct modifications of lipid A that reduce cell permeability are triggered by the PhoPQ pathway in this organism (3, 18-21, 52-54, 66). Thus, PhoPQ-regulated modifications of lipid A that adapt the two organisms to intracellular survival occur through distinct pathways.

We demonstrate that regulation of the msbB2 locus by magnesium is dependent on PhoP. A phoP mutant was attenuated in a mouse model of S. flexneri infection, was more susceptible to killing by macrophages, and was more susceptible to antimicrobial peptides, yet the PhoP-regulated gene(s) that were required for full virulence were not identified (38). The present study is the first identification and characterization of an S. flexneri PhoP-regulated gene involved in virulence. Based on the role of lipid A modification in survival within macrophages and resistance to antimicrobial peptides, it seems likely that PhoP regulation of msbB2 is at least in part responsible for the observed phenotype of the S. flexneri phoP mutant. However, there are likely to be additional S. flexneri genes regulated by PhoPQ, some of which may also contribute. Interestingly, our results demonstrate that under limited magnesium conditions, several lipid A species that were absent in the phoP mutant were present in the wild-type strain, and several that were absent in the wild-type strain were present in the phoP mutant (Fig. 5). Of note, many of these species were seen only in strain backgrounds that contained intact msbB1, suggesting the possibility of modifications, such as addition of amino-4-deoxy-l-arabinose, that require hexa-acylated lipid A substrate. Determination of what PhoP-dependent modifications might be represented by these spots will require biochemical identification of the species.

Enterohemorrhagic E. coli O157:H7 is the only organism other than Shigella spp. known to have two paralogs of msbB. The genetic context of msbB2 in O157:H7 is similar to that in Shigella spp., in that it is the last gene within an operon on a large virulence plasmid that consists of four genes, all but the third of which are highly homologous to their orthologs in Shigella (27). Environmental signals that regulate transcription of the two operons appear to be distinct. Whereas we have shown that transcription of the S. flexneri operon is regulated by magnesium, transcription of the O157:H7 operon is regulated by temperature (63). Whether the O157:H7 operon is also regulated by magnesium or PhoPQ has not been directly tested.

Our data are consistent with regulation of msbB2 by PhoP occurring by PhoP binding to the direct hexanucleotide repeat we identified within the shf-wabB-virK-msbB2 promoter and that we demonstrated was required for full induction by PhoP. Eleven of the 12 nucleotides within the repeats, as well as the spacing of the repeats, were identical to the consensus PhoP binding site for serovar Typhimurium and E. coli (Fig. 6A and B). Given the similarity of the msbB2 putative PhoP binding site to those of serovar Typhimurium and E. coli (26, 30, 37), PhoP binding sites in the promoters of additional PhoP-regulated genes in S. flexneri are likely to also be conserved.