Amino acid addition to Vibrio cholerae LPS establishes a link between surface remodeling in Gram-positive and Gram-negative bacteria

Abstract

Historically, the O1 El Tor and classical biotypes of Vibrio cholerae have been differentiated by their resistance to the antimicrobial peptide polymyxin B. However, the molecular mechanisms associated with this phenotypic distinction have remained a mystery for 50 y. Both Gram-negative and Gram-positive bacteria modify their cell wall components with amine-containing substituents to reduce the net negative charge of the bacterial surface, thereby promoting cationic antimicrobial peptide resistance. In the present study, we demonstrate that V. cholerae modify the lipid A anchor of LPS with glycine and diglycine residues. This previously uncharacterized lipid A modification confers polymyxin resistance in V. cholerae El Tor, requiring three V. cholerae proteins: Vc1577 (AlmG), Vc1578 (AlmF), and Vc1579 (AlmE). Interestingly, the protein machinery required for glycine addition is reminiscent of the Gram-positive system responsible for d-alanylation of teichoic acids. Such machinery was not thought to be used by Gram-negative organisms. V. cholerae O1 El Tor mutants lacking genes involved in transferring glycine to LPS showed a 100-fold increase in sensitivity to polymyxin B. This work reveals a unique lipid A modification and demonstrates a charge-based remodeling strategy shared between Gram-positive and Gram-negative organisms.

The Gram-negative pathogen Vibrio cholerae is the causative agent responsible for ∼300,000 reported cases annually of the severe diarrheal disease cholera. Since 1961, resistance to polymyxin B, a cationic antimicrobial peptide (CAMP), has been used to clinically differentiate between the two V. cholerae O1 biotypes, El Tor and classical (1). Interestingly the O1 classical biotype, which is polymyxin-sensitive, caused the first six cholera pandemics; however, polymyxin-resistant O1 El Tor strains are responsible for the current, seventh pandemic. Despite its importance, the molecular mechanisms accounting for this phenotypic difference have remained unknown for nearly 50 y.

During infection, the cells of the innate immune system secrete CAMPs, which are small, positively charged proteins. Much like polymyxin, these peptides bind to and disrupt the bacterial cell membrane, ultimately resulting in death of the invading bacterial cell. Although Gram-negative and Gram-positive bacteria possess different cell wall structures, both have evolved mechanisms to remodel their cell envelope in response to CAMPs. These mechanisms often involve a common theme: neutralizing the negative charge of major cell wall components.

The major surface component of Gram-negative bacteria is LPS, which is composed of three distinct regions: lipid A, core oligosaccharide, and O-antigen polysaccharide (2). Lipid A is the bioactive portion of LPS, which activates the human innate immune system through the Toll-like receptor 4 (TLR4)/myeloid differentiation factor 2 complex (3). The negatively charged lipid A domain is synthesized through a well-conserved pathway, The Raetz pathway, that results in a β-1′,6-linked disaccharide of glucosamine that is hexa-acylated and bis-phosphorylated (2, 4, 5). Some Gram-negative bacteria (e.g., Salmonella enterica) modify their lipid A domain in response to CAMPs by masking the 1- or 4′-phosphate groups with either phosphoethanolamine or aminoarabinose (Fig. 1) (5–8). The addition of an amine-containing substituent to the lipid A anchor of LPS disrupts the binding of CAMPs to the bacterial surface.

Fig. 1.

Gram-negative and Gram-positive bacteria modify their cell wall components with amine-containing substituents (shown in red). Some Gram-negative bacteria (e.g., Salmonella enterica) modify their lipid A phosphate groups with phosphoethanolamine and aminoarabinose. Additionally, glycerophospholipids modified with amino acids, such as lysine, represent major membrane lipids in Gram-positive bacteria (e.g., S. aureus). Gram-positives can further modify lipoteichoic acids and wall teichoic acids with d-alanine. The addition of amine-containing substituents to these cell wall components helps decrease the overall negative charge of the bacterial cell wall and protects the bacterium against attack by CAMPs.

Bacteria also modify other essential cell wall components in response to CAMPs. For example, many Gram-positive bacteria undergo d-alanylation of both wall teichoic acids and lipoteichoic acids (Fig. 1). Wall teichoic acids, which are linked to peptidoglycan, typically consist of a disaccharide linkage unit attached to a polyribitol phosphate or polyglycerol phosphate chain (9). Lipoteichoic acids are anchored to the membrane glycolipids of Gram-positive bacteria and consist of a (poly)glycerolphosphate backbone (9). Transfer of d-alanine to the ribitol phosphate or glycerolphosphate units of these polymers reduces the net negative charge of the Gram-positive cell wall. Modified forms of glycerophospholipids have also been described in bacteria. Aminoacyl esters of phosphatidylglycerol involving l-alanine and l-lysine (Fig. 1) represent major membrane lipids in several Gram-positives (10, 11) and have been implicated in conferring resistance to CAMPs, as well as enhancing survival in acidic conditions (10, 11).

The mechanism of V. cholerae polymyxin B resistance has been poorly understood. In these findings, we identify the addition of a glycine and a diglycine residue to the V. cholerae O1 El Tor lipid A species, which is required for polymyxin B resistance. Three V. cholerae proteins, Vc1577 (AlmG), Vc1578 (AlmF), and Vc1579 (AlmE), were found to be essential for this unique lipid A modification. Furthermore, these proteins share sequence homology with machinery involved in d-alanylation of lipoteichoic acid in Gram-positive bacteria. Discovery of an amino acid modification of the lipid A of V. cholerae illuminates an interesting link between Gram-negative and Gram-positive cell wall modification systems. Additionally, our findings provide a well-defined mechanism for the different polymyxin-resistant phenotypes observed in V. cholerae classical and El Tor biotypes for 50 y.

Results

V. cholerae O1 El Tor Synthesize Glycine-Modified Lipid A Species.

V. cholerae O1 classical and El Tor biotypes synthesize a common lipid A species (Fig. 2A) that can be detected at m/z 1,756 using MALDI-TOF MS in the negative-ion mode (Fig. 2 B and C). In a previous report from our laboratory (12), the O1 classical lipid A structure was elucidated by interrogating the singly deprotonated species (m/z 1,756) by tandem mass spectrometry (MS/MS). MALDI-TOF MS results also showed that V. cholerae O1 El Tor biotypes synthesize additional lipid A species (m/z 1,813.9 and 1,870.9) (Fig. 2B), which are not produced by the classical biotype (Fig. 2C).

Fig. 2.

MALDI-TOF MS of V. cholerae O1 El Tor and O1 classical lipid A species. Previously, V. cholerae O1 El Tor and O1 classical strains were shown to synthesize a similar hexa-acylated lipid A species (12) (A). This common lipid A species is shown at m/z 1,755.93 and m/z 1,756.1 for the El Tor (B) and classical (C) biotypes, respectively. The El Tor biotype produces additional lipid A species with peaks at m/z 1,813.9 and m/z 1,870.9.

First, to verify that both classical and El Tor strains have a common lipid A species on their surface, MS/MS was performed on the ion of m/z 1,756 peak produced by the El Tor biotype. The 193-nm UV photodissociation (UVPD) method was used because it has been shown to increase the depth of structural information for lipid A species over standard MS/MS techniques (13). Fig. 3A shows the UVPD spectrum of the [M–H]⁻ ion (m/z 1,756) of the V. cholerae El Tor lipid A, as well as its associated fragmentation map, which matches m/z values of observed fragment ions to distinctive cleavage sites along the lipid A structure (Fig. 3A and Fig. S1). The inter-ring glucosamine cleavage that generated the product ion of m/z 682.42 confirms the acyl chain configuration: the distal side of the molecule contains four acyl chains, whereas the proximal side contains only two. The combination of C-O bond cleavages at the phosphate groups (cleavage 1) and at the 3′- and 3-linked primary acyl chains (cleavages 4, 5, 6, and 7), and 2′- and 3′-linked secondary acyl chains (cleavages 8 and 10) allow identification of the basic structure of V. cholerae O1 El Tor lipid A. More specifically, cleavage at the 3′-linked secondary acyl chain, denoted as cleavage 8, confirms the general position of hydroxylation. Importantly, the abundant ion of m/z 1,627.83 (cleavage 23) identifies the specific position of the hydroxyl group on the 3′-linked secondary acyl chain. This preferential cleavage by UVPD has been used previously to locate the same hydroxyl group in the classical biotype (12). For a comparison with UVPD, collision-induced dissociation (CID) was also used to interrogate the same singly deprotonated ion of m/z 1,756 (Fig. S2) and yielded essentially the same C-O bond cleavages seen for UVPD. The combination of the more rich UVPD patterns and the simpler CID fragmentation patterns work well as complementary methods and confirm the structural assignment of lipid A from the El Tor biotype.

Fig. 3.

V. cholerae O1 El Tor synthesize glycine-modified lipid A species. UVPD MS was used to analyze the V. cholerae ions ([M – H]⁻, m/z 1,756; [M – H]⁻, m/z 1,813; and [M – H]⁻, m/z 1,870 [A–C]). Cleavages are indicated by dashed lines and the m/z values of the resulting fragment ions are shown in Fig. S1. The m/z values and cleavage sites highlighted in red font represent the unique fingerprints associated with the glycine or diglycine modifications. The precursor ion is denoted by an asterisk. The fragmentation patterns support that V. cholerae synthesizes a glycine- and diglycine-modified lipid A structure at the 3′-position of the glucosamine disaccharide. “x55,” “x65,” and “x20” denote a section of the spectrum that has been magnified 55, 65, or 20 times, respectively, to more easily visualize product ions.

Fig. 3B shows the UVPD spectrum of the singly deprotonated species of m/z 1,813, which is presumably lipid A with a 57-Da modification. Several of the product ions observed for the unmodified lipid A in Fig. 3A are also seen in Fig. 3B, confirming the general V. cholerae lipid A structure. Cleavages 3, 4, 6, and 8 localize the modification to the 3′-linked primary acyl chain; more specifically, cleavages 3 (neutral loss of the 57-Da modification) and 8 (loss of 3′-linked secondary acyl chain) confirm the general location of the 57-Da modification on the 3′-linked secondary acyl chain. The product ion of m/z 1,685.83 (cleavage 23) also provides further evidence concerning the position of the 57-Da modification; however, this product is isobaric with the potential ion produced from C-C cleavage adjacent to the hydroxyl group on the 3-linked acyl chain, thus making it somewhat ambiguous. The abundant ions of m/z 1,738.00, 1,755.75, 1,785.00, and 1,796.00 that correspond to cleavages 3, 11, 17, and 18 (labeled in red in Fig. 3B; the latter three all uniquely observed upon upon UVPD yet not by CID), respectively, confirm that the 57 Da modification is consistent with a glycine moiety. The rich level of detail in the region of m/z 1,650–1,800 of the UVPD spectrum is essential for mapping the glycine modification. The simpler CID spectrum seen in Fig. S2, which yields the general construction of the lipid A and complements the more detailed structural information obtained by UVPD, also aids in the confident structural assignment of the V. cholerae m/z 1,813 peak.

Fig. 3C displays the UVPD spectrum of the singly deprotonated ion of m/z 1,870 of the V. cholerae lipid A that contains a 114-Da diglycine modification. UVPD again yielded the same types of complementary product ions (Fig. 3C and Fig. S1) for the diglycine-modified lipid A as were observed for the glycine-modified species in Fig. 3B, with considerable fragmentation detail related to the diglycine modification. The primary structure and site-specificity of the diglycine modification at the same 3′-linked secondary acyl chain were confirmed based on the abundant ions labeled in red of m/z 1,741.92, 1,738.17, 1,756.17, 1,812.17, 1,842.17, 1,853.17, and 1,825.17 that correspond to cleavages 23, 3, 11, 20, 21, and 22. Interestingly, the ion of m/z 1,812.17 is the most abundant one in the spectrum and is a product from the cleavage of the diglycine backbone between the amine and carbonyl group, which is the same type of bond (i.e., amide bond) most preferentially cleaved upon UVPD of singly deprotonated lipid A ions. This product ion corresponds predictably with established UVPD fragmentation behavior (12, 13) and supports the assignment of the 114-Da modification as a diglycine moiety. The corresponding CID spectrum is shown in Fig. S2.

Vc1577 Is Required for Glycine Modification of V. cholerae O1 El Tor Lipid A.

The Clusters of Orthologous Groups database (14) identifies three putative V. cholerae lipid A late acyltransferase homologs: Vc0212, Vc0213, and Vc1577. Previous work from our laboratory (12, 15) identified Vc0212 (LpxN) and Vc0213 (LpxL) as functional V. cholerae lipid A late acyltransferases, which are responsible for the addition of myristate (C14:0) and 3-hydroxylaurate (3-OH C12:0) to the 2′- and 3′-positions of the glucosamine disaccharide, respectively (Fig. 2A and Fig. S3). Although BLASTp results indicate that Vc1577 is an unlikely lipid A late acyltransferase homolog (E-value 3.9, 22% identity), the protein contains a lysophospholipid acyltransferase (LPLAT) domain. Because proteins of the LPLAT superfamily are typically involved in the synthesis of membrane lipids (e.g., glycerophospholipids) catalyzing the transfer of an acyl chain from an acyl donor (e.g., acyl-acyl carrier protein, acyl-ACP) to lipid precursors, we hypothesized that Vc1577 may play a role in the synthesis of lipid A.

To determine if Vc1577 was involved in transfer of a glycine residue, a vc1577 mutant was generated in a V. cholerae O1 El Tor strain (Table S1). Purified lipid A from the vc1577 mutant was analyzed by MALDI-TOF MS and produced a predominant peak at m/z 1,757.1, consistent with the exact mass of the hexa-acylated V. cholerae lipid A species lacking the glycine modification (Fig. 4A, Left). Complementation of the vc1577 mutant restored the addition of glycine and diglycine, producing predominant peaks of m/z 1,812.9 and 1,869.9 (Fig. 4A, Right).

Fig. 4.

vc1577 is involved in glycine modification of V. cholerae lipid A and is cotranscribed with vc1578 and vc1579. (A) Lipid A was isolated from O1 El Tor vc1577::kan and analyzed by MALDI-TOF MS. A predominant peak at m/z 1,757.1 was consistent with the loss of a glycine. MALDI-TOF analysis of lipid A isolated from vc1577::kan, vc1577 resulted in major peaks at m/z 1,812.9 and 1,869.9, which is consistent with the addition of glycine and diglycine residues, respectively. Unmodified hexa-acylated lipid A is also present at m/z 1,755.9. (B) A schematic of the genetic organization of vc1577 (almG), vc1578 (almF), and vc1579 (almE) is shown. RT-PCR was done to determine if vc1577, vc1578, and vc1579 are cotranscribed and indicated a read-through transcript (product 1) containing vc1579-77. Additional RT-PCRs confirmed read-through transcripts for vc1579-78 (product 2) and vc1578-77 (product 3). V. cholerae genomic DNA template was used as a positive control for primers and amplified product sizes; however, for the negative control (−RT), cDNA without reverse transcriptase added was used as template, verifying that no DNA contamination had occurred.

Vc1578 and Vc1579 Also Contribute to Glycine-Modification of V. cholerae Lipid A.

Examination of the genomic context vc1577 suggests that it is cotranscribed with vc1578 and vc1579. To determine if vc1577, vc1578, and vc1579 are in an operon, RT-PCR was performed using RNA isolated from the O1 El Tor biotype. As shown in Fig. 4B, V. cholerae O1 El Tor synthesize a contiguous mRNA containing vc1577, vc1578, and vc1579 (Fig. 4B).

Vc1579 is annotated as a putative component (EntF) of enterobactin synthetase and contains an amino acid adenylation domain found in nonribosomal peptide synthetases involved in siderophore biosynthesis (16). However, amino acid adenylation domains are also found in proteins involved in modification of cell wall components of Gram-positives. Vc1579 shares homology (E-value 10⁻³³, 28% identity) with the Staphylococcus aureus d-alanine-d-alanyl carrier protein ligase (Dcl) (9, 17–19). Dcl activates d-alanine as an adenylate followed by condensation of the activated amino acid onto a small carrier protein, termed d-alanine carrier protein (Dcp). d-alanyl-Dcp serves as the donor substrate for modification of teichoic acids (9, 17–19).

Although no closely related homologs of Vc1578 were identified, the protein homology/analogY recognition engine (Phyre) (20) revealed that this small acidic protein has structural similarity to acyl-ACP (Fig. S4). ACP and its homologs serve as carrier proteins for the synthesis of a number of biologically important molecules, including fatty acids, polyketides, and nonribosomal peptides (21). These small acidic proteins share a conserved serine residue that is subject to posttranslational modification by a 4′-phosphopantetheine moiety that “carries” the substrate (e.g., d-alanine) as a thioester. Based on amino acid sequence alignments (Fig. S4A) and 3D structural predictions (Fig. S4B), Vc1578 shares the conserved serine (residue 34) as well as the overall helix bundle structure found in other characterized carrier proteins (21).

To confirm that Vc1579 and Vc1578 are required for glycine modification of lipid A, vc1578 and vc1579 transposon insertions (Fig. S5) were obtained from the V. cholerae O1 El Tor nonredundant transposon insertion library (22). MALDI-TOF analysis of lipid A isolated from vc1578 and vc1579 mutants showed a lack of glycine and diglycine modified forms (Fig. S6). Introduction of pVc1579, an arabinose inducible covering plasmid (Table S1), into vc1579 restored lipid A modification. Complementation of the vc1578 mutant also resulted in glycine modified species, albeit at lower levels compared with wild-type (Fig. S6). Given their role in aminoacyl lipid modification, we propose that Vc1577, Vc1578, and Vc1579 be assigned the designations, AlmG, AlmF, and AlmE, respectively.

Glycine Modification of Lipid A Is Required for Polymyxin B Resistance of V. cholerae Strains.

V. cholerae O1 El Tor strains were found to be highly resistant to polymyxin B, exhibiting a minimum inhibitory concentration (MIC) of ∼100 μg/mL using ETest polymyxin B strips (Fig. 5A). As a comparison, polymyxin-resistant variants of E. coli or S. enterica are resistant at ∼10–12 μg/mL (23). Deletion of vc1577 (almG) resulted in nearly a 100-fold decrease in polymyxin resistance compared with wild-type levels; however, the polymyxin-resistant phenotype can be restored when vc1577 (almG) is expressed in trans (Fig. 5A). Additionally, vc1578 (almF) and vc1579 (almE) El Tor mutants were more than 100-fold more sensitive to polymyxin B compared with the El Tor wild-type strain (Table S2). Consistent with MS results (Fig. S6), complementation of vc1579 (almE) restored antimicrobial peptide resistance similar to wild-type levels (MIC, 64 μg/mL), whereas complementation of the vc1578 (almF) mutant was only partial (MIC, 12 μg/mL). Still, expression of vc1578 in trans yields a 24-fold increase in resistance in comparison with the vc1578 mutant (MIC, 0.5 μg/mL) (Table S2). This partial complementation could arise due to the stoichiometry and flux of the system.

Fig. 5.

AlmG (Vc1577), AlmF (Vc1578), and AlmE (Vc1579) confer polymyxin resistance to V. cholerae. Polymyxin MIC for various strains was determined on LB agar using ETest polymyxin B strips. (A) Wild-type V. cholerae O1 El Tor are polymyxin-resistant with a MIC of 96 μg/mL However, the vc1577 mutant, lacking the glycine modification of lipid A, is nearly 100-fold more sensitive to polymyxin B. When O1 El Tor vc1577::kan is complemented in trans, the strain gains polymyxin resistance. (B) Introduction of a plasmid expressing vc1579-77 confers polymyxin resistance to O1 classical strains. (C) MALDI-TOF MS of the lipid A of the classical biotype expressing pW77-79 revealed ions at m/z 1,813.5 and m/z 1,870.5, corresponding to lipid A modified by either one or two glycine residues, respectively. The minor peak at m/z 1,757.5 represents unmodified hexa-acylated V. cholerae lipid A.

Although the classical strain O395 synthesizes unmodified hexa-acylated lipid A, expression of vc1577-79 (almEFG) in trans resulted in an 85-fold increase in polymyxin resistance compared with the wild-type strain (Fig. 5B). MALDI-TOF MS analysis of lipid A isolated from O1 classical expressing AlmEFG showed glycine modified species with major ions at m/z 1,813.5 and 1,870.5 (Fig. 5C). Taken together, these results confirm that AlmEFG confers polymyxin resistance to V. cholerae by glycine addition to surface-exposed LPS.

Polymyxin Resistance Correlates with Glycine Modification in Clinical and Environmental V. cholerae Strains.

To determine if other polymyxin-resistant V. cholerae isolates synthesize glycine-modified LPS, lipid A from various environmental/clinical V. cholerae isolates (O1 El Tor, O1 classical, and non-O1) was analyzed by MALDI-TOF MS (Table S3). Our results indicated glycine-modified lipid A forms in all but one strain, whereas polymyxin-sensitive strains did not contain these lipids (Table S3). These results provide a strong correlation between polymyxin resistance and glycine modification of the V. cholerae surface.

Effect of Glycine Modification on Activation of TLR4.

Previously, we demonstrated that reduction in the number of acyl chains drastically decreased the endotoxicity of V. cholerae LPS (12). To determine the impact of glycine modification on Toll activation, we performed reporter cell signaling assays with purified LPS from V. cholerae O1 El Tor and vc1577::kan (Fig. S7). Whereas neither strain stimulated TLR2, a TLR activated by bacterial lipoproteins and lipoteichoic acids (24), the absence of glycine on lipid A caused only a modest decrease (P < 0.01) in TLR4 activation with 10 ng/mL of LPS and higher (Fig. S7). Thus, it does not appear that this particular lipid A modification plays a major role in endotoxicity.

Discussion

To adapt to their surrounding environment, bacteria have evolved the ability to remodel their cellular envelope. The modification of bacterial cell wall components (phospholipids, LPS, or techoic acids) has been shown to play a vital role in antimicrobial peptide resistance in Gram-negative and Gram-positive bacteria. Both organisms use amine-containing substituents to alter the overall negative charge of their surface in response to CAMP attack (Fig. 1). Our findings herein identify a unique lipid A modification that occurs in the Gram-negative organism V. cholerae. The rich level of structural detail afforded by UVPD (13) (Fig. 3 and Fig. S1) proved critical for mapping the identity and location of the glycine/diglycine modification. The presence of a glycine residue was found to be essential for polymyxin resistance and presumably decreases the negative charge of the bacterial surface. However, it is also possible that glycine modification alters membrane fluidity, thereby influencing antimicrobial peptide resistance.

Although prevalent in Gram-positives, aminoacylation of glycerophospholipids has been observed only in select Gram-negatives (i.e., Pseudomonas and Rhizobium) and depend on aminoacyl-tRNAs for their synthesis (10, 11). However, V. cholerae has evolved an amino acid modification system similar to that found in Gram-positive bacteria for modification of teichoic acids. Sequence analysis revealed that Vc1579 (AlmE) shares sequence homology to the Gram-positive Dcl (E-value 10⁻⁴⁷, 33% identity) (9, 19). Based on bioinformatic and structural analysis, Vc1578 was hypothesized to serve as a glycine carrier protein similar to the d-alanine carrier protein Dcp (Fig. S4) (9, 17–19). Sequence alignments show Vc1578 (AlmF) possesses a conserved serine at residue 34 that likely acts as the site of attachment of a 4′-phosphopantetheine group. Although the d-alanine transferase is unknown in Gram-positive bacteria, we hypothesize that Vc1577 (AlmG) catalyzes the transfer of glycine to the unmodified hexa-acylated V. cholerae lipid A. AlmG contains a conserved LPLAT domain that may provide recognition of both the lipid A domain of LPS and the amino acid-charged carrier protein. A model for glycine modification of V. cholerae lipid A species is shown in Fig. 6.

Fig. 6.

Proposed model for the synthesis of glycine-modified lipid A species in V. cholerae. The model indicates glycine is ligated to the carrier protein, AlmF-SH (Vc1578), as a thioester. The reaction is catalyzed by the amino acid ligase, AlmE (Vc1579), in the cytoplasm. AlmG (Vc1577) then catalyzes the transfer of glycine from AllmF-S-glycine to the hexa-acylated V. cholerae lipid A species in the inner membrane. Presumably, diglycine would arise from a second addition to the lipid A molecule. Glycine-modified forms of lipid A are then transported to the bacterial surface providing resistance to polymyxin.

The proteins involved in glycine modification are predicted soluble proteins by TMHMM (25) and contain no signal peptides required for secretion across the inner membrane. This prediction is consistent with the fact that the amino acid ligase (AlmE) would require cytoplasmic substrates, such as ATP and glycine, for activity (Fig. 6). Given that AlmG uses hexa-acylated lipid A as a substrate, one would expect that amino acid addition occur at the membrane. Indeed, a previous report from our laboratory showed that AlmG (Vc1577) is membrane-associated (15). A periplasmic localiziation would require charged AlmF to be transported across the inner membrane to serve as a donor for AlmG. Generally, lipid A modification systems are located in the extracytoplasmic compartments of the bacterial cell separated from the conserved biosynthetic pathway known as the Raetz pathway (5). Although our data suggests that glycine addition occurs on the cytoplasmic face of the inner membrane, further biochemical characterization is necessary to confirm the proposed model (Fig. 6).

Earlier reports indicated that the outer membrane porin, OmpU, contributes toward polymyxin resistance in the O1 classical biotype (26). However, as demonstrated here and by others (1, 27) O1 classical strains are quite sensitive to polymyxin (MIC of 0.75 μg/mL) (Table S2), making it difficult to determine phenotypic differences in classical strains. V. cholerae O1 El Tor strains lacking either of the major porins, OmpU or OmpT, show wild-type levels of polymyxin-resistant (MIC 96–128 μg/mL) (Table S2), whereas alm mutants are sensitive (MIC 0.5–1.0 μg/mL).

Both V. cholerae O1 El Tor and classical biotypes synthesize a hexa-acylated lipid A species with a 3-hydroxylaurate (3-OH C12:0) acyl chain linked at 3′-position of the molecule (Figs. 2 and 3) (12). In a previous report it was shown that the presence of the 3-hydroxylauroyl group was necessary for antimicrobial peptide resistance (12, 28). Our laboratory identified an unusual acyl transferase, LpxN, which is selective for hydroxylated acyl chains (Fig. S3) (12), and demonstrated that the hydroxyl group itself promotes polymyxin resistance. Here, we establish that the 3-hydroxyl group is actually the site of glycine modification. Our previous findings also demonstrated that reduction in the number of acyl chains drastically decreased the endotoxicity of V. cholerae LPS (12). To determine the impact of glycine modification on Toll activation, we performed reporter cell-signaling assays with purified LPS from V. cholerae O1 El Tor and vc1577::kan. Whereas neither strain stimulated TLR2, a TLR activated by bacterial lipoproteins and lipoteichoic acids (24), the absence of glycine on lipid A caused only a modest decrease (P < 0.01) in TLR4 activation with 10 ng/mL of LPS and higher (Fig. S7). Thus, it does not appear that this particular lipid A modification plays a major role in endotoxicity.

One key question is why classical strains lack glycine-modified lipid A. Comparison of the genomes of sequenced classical and El Tor strains showed that both biotypes contain the almEFG locus. However, sequence alignments comparing the coding region revealed that classical strain O395 has a nonsense mutation, resulting in a truncated almF carrier protein lacking the conserved serine (Fig. S8). When the classical strain is complemented with the El Tor almEFG region in trans, the polymyxin phenotype is restored and glycine-modified lipid A species are synthesized (Fig. 5 B and C). The O139 serogroup (strain MO10) that was associated with outbreaks in the 1990s is also polymyxin-resistant, synthesizing glycine-modified lipid A (Table S3). O139 arose from a progenitor O1 El Tor strain and has an intact almEFG locus.

The life cycle of V. cholerae is complex, having both human and environmental stages. Given that the El Tor biotype displaced the classical biotype, it is tempting to speculate that amino acid modification of the cell surface enhances bacterial survival. Another consideration is whether glycine addition is regulated in different environments. This work also provides the opportunity to investigate the biochemical mechanisms of cell envelope modifications aiding in development of new antimicrobial compounds. On the whole, our findings identify a unique LPS modification that mechanistically links Gram-positive and Gram-negative cell wall modification systems.

Materials and Methods

Bacterial Strains and Growth Conditions.

The bacterial strains and plasmids used in this study are summarized in Tables S1 and S4. V. cholerae strains were grown routinely at 37 °C in Luria-Bertani (LB) broth or on LB agar or in a modified g56 minimal media, described previously (12). E. coli strains were grown in LB at 37 °C. Antibiotics were used at the following concentrations: 100 μg/mL ampicillin, 60 μg/mL kanamycin, 10 μg/mL streptomycin. For V. cholerae complementation, 0.5 mM isopropyl β-d-1-thiogalactopyranoside was added to growth media.

MS and UVPD.

All MS experiments were undertaken on a Thermo Fisher Scientific LTQ XL linear ion-trap mass spectrometer equipped with a 193-nm Coherent ExciStar XS excimer laser. The setup was similar to that previously described (29, 30). An online nanoelectrospray setup was used for direct infusion of lipid A samples as reported previously (13).

Other Methods.

Methods describing recombinant DNA and RNA techniques, lipid A isolation, polymyxin B MIC assays, complementation plasmids, and gene deletions are described in SI Materials and Methods.