Cell Separation in Vibrio cholerae Is Mediated by a Single Amidase Whose Action Is Modulated by Two Nonredundant Activators

Abstract

Synthesis and hydrolysis of septal peptidoglycan (PG) are critical processes at the conclusion of cell division that enable separation of daughter cells. Cleavage of septal PG is mediated by PG amidases, hydrolytic enzymes that release peptide side chains from the glycan strand. Most gammaproteobacteria, including Escherichia coli, encode several functionally redundant periplasmic amidases. However, members of the Vibrio genus, including the enteric pathogen Vibrio cholerae, encode only a single PG amidase, AmiB. Here, we show that V. cholerae AmiB is crucial for cell division and growth. Genetic and biochemical analyses indicated that AmiB is regulated by two activators, EnvC and NlpD, at least one of which is required for AmiB's localization to the cell division site. Localization of the activators (and thus of AmiB) is dependent upon the cell division protein FtsN. These factors mediate septal PG cleavage in E. coli as well; however, their precise roles vary between the two organisms in a number of ways. Notably, even though V. cholerae EnvC and NlpD appear to be functionally redundant under most growth conditions tested, NlpD is specifically required for intestinal colonization in the infant mouse model of cholera and for V. cholerae resistance against bile salts, perhaps due to environmental regulation of AmiB or its activators. Collectively, our findings reveal that although the cellular components that enable cleavage of septal PG appear to be generally conserved between E. coli and V. cholerae, they can be combined into diverse functional regulatory networks.

INTRODUCTION

Peptidoglycan (PG) is a critical cell wall component for most bacteria (1). This polymer preserves cell integrity against osmotic pressure and is essential for maintenance of cell shape. PG is composed of glycan chains with alternating β-1,4-linked N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc) peptide residues, which are connected through cross-links of the peptide side chains. The PG mesh is a dynamic structure that is constantly expanded and modified as cells elongate, due to the activities of cell wall synthetic and lytic enzymes. PG synthesis and modification are also critical components of cell division, as a PG-containing septum must simultaneously be synthesized at the bacterial division plane and cleaved to allow the subsequent dissociation of daughter cells. The PG hydrolases that mediate septal remodeling must be highly regulated; they should be active enough to avoid stalling cell division yet not so much as to trigger cell lysis (1). In Escherichia coli, both the localization and the enzymatic activity of PG hydrolases are regulated by their interaction with other cellular components, such that cleavage of septal PG is tightly coordinated with other steps in the cell division process (2, 3).

Periplasmic amidases have been shown to play a critical role in the cleavage of septal PG (4–7). These enzymes hydrolyze the amide bond between the N-terminal l-alanine residue of the peptide side chain and MurNAc. Most studies of septal PG cleavage have been carried out with E. coli, an organism that produces three functionally redundant periplasmic PG amidases: AmiA, AmiB, and AmiC (7, 8). The absence of any single amidase causes only minor changes in E. coli growth and cell morphology, which are mainly apparent in stationary phase, and does not markedly alter PG composition. However, E. coli lacking all three amidases forms long chains of cells with distinct cytoplasmic compartments but a shared PG layer between cells, due to an inability to divide septal PG between daughter cells (7). Other hydrolases (e.g., endopeptidases and lytic transglycosylases) also likely have some activity on septal PG, but the effect of their absence is far less dramatic, suggesting that they play a less significant role in septal remodeling (9).

All three E. coli amidases contain a C-terminal catalytic amidase_3 domain, and AmiB and AmiC each contain an N-terminal targeting domain that mediates their accumulation at the division site (2, 10). It was recently shown that AmiC's targeting domain is a novel PG binding AMIN domain (11). The septal localization of AmiB and AmiC is also dependent on the essential cell division protein FtsN (2, 10) and thus on maturation of the divisome, which is typically followed by synthesis of septal PG and cell constriction. Finally, recruitment of AmiB and AmiC to the divisome can be blocked by the PBP3 inhibitor cephalexin, suggesting that septal PG synthesis is also required for recruitment of AmiB and AmiC to the midcell (2). It is hypothesized that such regulated recruitment of the amidases to the midcell prevents excessive and potentially catastrophic PG cleavage that could lead to cell lysis.

Biochemical and genetic studies suggest that the E. coli amidases require activation through direct contact with the M23 peptidase-like proteins EnvC and NlpD (3). In particular, AmiA and AmiB are activated by EnvC, whereas AmiC is activated by NlpD. Additional studies suggest that EnvC and NlpD activate their cognate amidases by inducing conformational changes that remove autoinhibitory α-helices from the amidase active sites (11, 12). The phenotype of a strain lacking both activators is similar to that of the amidase triple mutant (13). Notably, both EnvC and NlpD lack critical residues in their respective peptidase active sites, and neither has peptidase activity (3). However, EnvC's ability to activate amidases is abrogated when point mutations are introduced into its degenerate active site, suggesting that the M23 peptidase domain nonetheless mediates amidase activation (14). Similar studies have not been carried out with NlpD, but it is assumed that its M23 domain is likewise required for the activation of AmiC.

Like AmiB and AmiC, NlpD and EnvC localize to the division septum, although the localization of these factors is not interdependent, and they do not all reach their target site simultaneously (2). Septal localization of NlpD, like that of AmiB and AmiC, is FtsN dependent, but EnvC localization is FtsN independent and occurs earlier than localization of its cognate amidase. Localization of EnvC is instead dependent on the ABC transporter-like complex FtsEX, which is essential for cell division in media of low osmotic strength (15). EnvC interacts directly with FtsX but not FtsE; however, the ATPase activity of FtsE seems to be required for EnvC activity (16). For Gram-negative bacteria other than E. coli, there is relatively little knowledge of the amidase-activator complex(es) that cleaves septal PG. In this study, we investigated septal PG cleavage in Vibrio cholerae, a Gram-negative, rod-shaped bacterium native to marine and estuarine environments, which can colonize the human small intestine and cause the diarrheal disease cholera. Notably, this facultative pathogen encodes only a single putative PG amidase, now called AmiB, which we show is crucial for V. cholerae growth and cell division. Genetic analyses suggest that AmiB is dependent upon two partially redundant proteins, EnvC and NlpD, which based on their similarities to the corresponding E. coli proteins are presumed to function as activators. In bacterial two-hybrid analyses, both activators interact with AmiB. The presence of either EnvC or NlpD is sufficient to mediate normal V. cholerae growth and cell division under most conditions tested. However, we find that nlpD is necessary for optimal colonization of the intestine in the infant mouse model of cholera, whereas envC is not, suggesting that the regulation and roles of its two activators are not equivalent.

MATERIALS AND METHODS

Strains, media, and growth conditions.

All V. cholerae strains described in this article are derivatives of V. cholerae wild-type (WT) El Tor strain C6706.

Cells were grown at 37°C or 30°C in Luria-Bertani (LB) medium, salt-free LB medium, or M9 medium supplemented with 0.2% glucose (M9) or glucose and 0.2% Casamino Acids (M9CA). Media were supplemented with 200 μg/ml of streptomycin, 50 μg/ml of kanamycin, 50 μg/ml of carbenicillin, or 5 μg/ml (V. cholerae) and 20 μg/ml (E. coli) of chloramphenicol. For induction of genes encoding fluorescent protein fusions under the control of arabinose- or lactose-inducible promoters, strains were grown in media supplemented with 0.2% l-arabinose or 0.2 mM isopropyl-β-d-1-thiogalactopyranoside (IPTG) for 2 to 5 h. To test for sensitivity against bile, LB medium was supplemented with bile salts (Sigma) as indicated below. For blue-white screens, media were supplemented with 40 μg/ml of 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal).

E. coli DH5α λpir, XL10-Gold (Stratagene), and XL1-Blue (Stratagene) were used for general cloning purposes. E. coli SM10 λpir and MFDpir (17) were used for conjugation. E. coli BTH101 (Euromedex) was used for bacterial adenylate cyclase two-hybrid (BACTH) assays and grown at 30°C on LB plates supplemented with X-Gal, 0.5 mM IPTG, carbenicillin, and kanamycin.

For growth curves, at least 4 replicates per strain and condition were grown in 200 μl of medium in a 100-well honeycomb plate inoculated 1:100 from an exponentially growing preculture (optical density at 600 nm [OD₆₀₀], ∼0.02) and analyzed in a Biotek growth plate reader at 10-min intervals. Data were analyzed using GraphPad PRISM for Windows (GraphPad, San Diego, CA).

Construction of plasmids and strains.

Plasmids and strains are described in detail in Tables S1 to S4 in the supplemental material. Plasmids were generated with Gibson assembly (18). To generate plasmids for in-frame deletions, 400- to 900-bp regions up- and downstream of the gene of interest were amplified by PCR and cloned into either of the sacB-containing suicide vectors pCVD442 (19) and pDS132 (20). Allele changes were introduced by transformation of V. cholerae C6706 with pAM231 (ΔamiB), pAM232 (ΔnlpD), pAM235 (ΔftsN), or pTDΔ0335 (ΔenvC), followed by sucrose-based counterselection, and mutant strains were verified by PCR. For inducible constructs, genes of interest were cloned into vector pBAD18-kan or pBAD33 (21) for expression from the araC (P_BAD) promoter or into pHL100 (22) or pAM227 for expression from the lacZ (P_lac) promoter. For BACTH studies, genes encoding cytoplasmic and transmembrane proteins were amplified by PCR and introduced into plasmids pKT25 and pUT18C (23), and genes encoding periplasmic proteins were amplified omitting the region encoding the signal sequence and cloned into pAM_T25pp and pAM_T18pp (24). All plasmid sequences were verified by sequencing.

TnSeq.

Transposon insertion sequencing (TnSeq) was performed as described previously (25). In brief, 200,000 to 300,000 transposon mutants were generated for each strain by transformation with transposon delivery vector pSC189, and pooled genomic DNA fragments were analyzed on an Illumina MiSeq benchtop sequencer (Illumina, San Diego, CA). Insertion sites were identified as described previously (25). Results were visualized using Artemis (26). Each strain was analyzed in two separate experiments.

Mouse experiments.

Intestinal colonization in infant mice was tested as described previously (27). In brief, cells were grown overnight at 30°C in LB medium. For exponential-phase inoculum, cultures were diluted 1:200 and grown to an OD of ∼0.1. Five-day-old suckling CD-1 mice (Charles River) were orogastrically inoculated with approximately 2 × 10⁵ CFU in 50 μl. For competition assays, mice were infected with equivalent amounts of a mutant and a wild-type strain, of which one was lacZ negative. Mice were euthanized after 24 h, and the intestine was homogenized in 7 ml of LB medium. Dilutions of the inoculum and the homogenate were plated on LB agar supplemented with streptomycin and X-Gal for determination of viable counts. Competitive indices (CI) were calculated as the ratio of mutant to wild-type bacteria isolated from intestines normalized to the input ratio. Data were analyzed using PRISM. Statistical significance was determined using a two-tailed Mann-Whitney t test (P < 0.01). This study was approved by the Institutional Animal Care and Use Committee (IACUC).

Microscopy.

Microscopy was performed using exponentially growing cells (OD₆₀₀, ∼0.2 to 0.6). For visualization of fluorescent fusion proteins, cells were cultured in M9. Bacteria were immobilized on pads containing 1% agarose, 10% LB medium, and 1× phosphate-buffered saline (PBS) and were visualized using a Zeiss AxioImager.2 microscope equipped with a Plan Neofluor 100×/1.3 oil Ph3 objective and a Hamamatsu Orca ER 1394 camera. Images were processed with ImageJ (28). Statistical analyses of fluorescence distribution were performed using MicrobeTracker (29) and Matlab (Mathworks). To visualize cell cycle-dependent localization via “demographs,” the cell length and signal information for each cell was extracted using a custom-written script (see the supplemental material). Integrated fluorescence signal vectors were sorted by cell length, and fluorescence values were normalized to maximum fluorescence. Subsequently, heat maps were generated from the resulting matrices.

Staining with 7-hydroxycoumarin-3-carboxylic acid-coupled d-alanine (HADA) was performed essentially as described previously (30). In short, cells were grown to exponential phase in LB or M9 medium and cultured in the presence of 100 μM HADA for 45 min at 37°C. Subsequently, cells were incubated for 15 min at room temperature (RT) with 5 μM Syto 9 (Molecular Probes), washed twice in LB or M9 medium without HADA, and imaged as described above.

Isolation of peptidoglycan and UPLC or high-performance liquid chromatography (HPLC) analysis.

Peptidoglycan of V. cholerae strains was isolated as described previously (22), starting with cells from 1 liter (WT and ΔenvC ΔnlpD) or 3 liters (ΔamiB ΔenvC ΔnlpD) of an exponentially growing culture. Culture pellets were resuspended in PBS and slowly added to boiling 10% SDS with stirring. Samples were boiled for 4 h and then stirred overnight at 37°C. Cell wall material was then pelleted by ultracentrifugation (110,000 rpm, 1 h) and washed three times in Milli-Q water. Samples were digested with pronase E (100 μg/ml) in 10 mM Tris-HCl, pH 7.5, for 1 h at 60°C to remove Braun's lipoprotein. After addition of SDS to a final concentration of 1% (wt/vol), reaction mixtures were heat inactivated and detergent was removed by washing in Milli-Q water. Purified peptidoglycan was resuspended in 50 mM NaPO₄ buffer, pH 4.9, and treated with 100 μg/ml of muramidase for 16 h at 37°C. Muramidase digestion was stopped by incubation in a boiling water bath, and coagulated proteins were removed by 10 min of centrifugation at 14,000 rpm. The supernatants were reduced by adding 0.5 M sodium borate, pH 9.5, and sodium borohydride to a final concentration of 10 mg/ml and incubating the mixtures at RT for 30 min. Finally, samples were adjusted to pH 3.5 with phosphoric acid. Muropeptides were separated in a 20-min linear gradient of 50 mM NaPO₄, pH 4.35, to 50 mM NaPO₄, pH 4.95, and 15% methanol (vol/vol) on an AQUITY ultraperformance liquid chromatography (UPLC) BEHC18 column (130 Å, 1.7 μm, 2.1 mm by 150 mm; Waters, USA), and peptides were detected at A₂₀₄. The relative amounts of the muropeptides were calculated as described by Glauner (31). Values are the means of two independent experiments. Data were analyzed using PRISM. Variations of less than 1% were considered chance variations. Statistical significance was determined using an unpaired two-tailed t test (P < 0.01).

BACTH analysis.

BACTH analysis was performed as described previously (23). BTH101 was cotransformed with plasmids that encoded fusions of the protein of interest to either the T25 or the T18 subunit of Cya. Periplasmic proteins were linked to cytoplasmic T25 and T18 Cya subunits with a transmembrane anchor derived from E. coli MalG, as described previously (24). As a negative control, all fusion vectors were also assayed in conjunction with control (empty) vectors either including or lacking the transmembrane anchor.

Protein sequence analyses.

Protein sequences were retrieved from the NCBI database (http://www.ncbi.nlm.nih.gov/). Protein primary sequence alignments were generated with Clustal Omega (32) and edited with BioEdit 7.0.5 (Ibis Therapeutics, USA). Proteins containing the amidase_3 (PF01520) or M23 peptidase (PF01551) domain were retrieved from the UniProt Knowledgebase (33), and phylogenetic trees based on the NCBI taxonomy tree were generated and edited using iTOL (34). Protein architectures were determined using SMART analyses (35). SPII cleavage sites were predicted using LipoP 1.0 (36).

RESULTS

AmiB is required for V. cholerae cell division and localizes to the cell division site.

In contrast to E. coli, V. cholerae has only one predicted periplasmic PG amidase, which is encoded by VC0344. Like E. coli AmiB, VC0344 contains an N-terminal signal sequence for export via the Sec system and an amidase_3 domain, and like E. coli AmiC, it contains an N-terminal AMIN domain (Fig. 1A). Based on VC0344's slightly higher similarity to AmiB than to the other E. coli PG amidases (38% identity [ID] and total score [TS] of 276 for E. coli AmiB, 35% ID and TS of 234 for AmiC, and 38% ID and TS of 155 for AmiA), we refer to it as AmiB from here on. V. cholerae AmiB also contains three putative LysM PG-binding domains at its C terminus that are not present in the E. coli PG amidases. Analysis of previously reported transposon insertions within VC0344 indicates that disruption of this C-terminal region is less detrimental than disruption of the remainder of the gene (25), suggesting that the LysM domains are not critical to the function of AmiB.

FIG 1.

AmiB localizes to the septum and is required for cell division. (A) Predicted domain structure of AmiB. Shown are the Sec signal sequence (amino acids [AA] 1 to 34 [red]), AMIN domain (AA 40 to 157; AMIN), amidase_3 domain (AA 237 to 396; Ami_3), and three LysM domains (AA 417 to 460, 483 to 526, and 536 to 579). (B) Phase-contrast images of wild-type and ΔamiB V. cholerae cells grown to exponential phase in LB, M9CA, and M9 media and of ΔamiB cells expressing plasmid-encoded AmiB-mCherry grown in LB medium. (C) Growth curves of wild-type and ΔamiB V. cholerae grown in LB, M9, and salt-free LB (LBSF) media. OD₆₀₀ was measured at 10-min intervals. Data points represent means ± SEM of at least four replicates. (D) Representative images of wild-type and ΔamiB cells grown in LB and M9 media and stained with Syto 9 (which marks the cytoplasm) and HADA (which stains PG). (E) Localization of ectopically expressed AmiB-mCherry in wild-type (C6706) V. cholerae. Cells were grown in M9 medium and expression was induced with 0.2% arabinose for 3 h. A representative fluorescence image is shown, along with demographs depicting AmiB localization (left) and constriction (right) relative to cell length. Fluorescence intensity and constriction (width) are indicated by a rainbow color scale from blue (0) to dark red (100%). Scale bars: 5 μm.

V. cholerae containing an in-frame deletion of amiB formed long filaments similar to those produced by E. coli lacking all three amidases (Fig. 1B), consistent with V. cholerae's apparent lack of additional periplasmic PG amidases. Analyses of culture density confirmed that the ΔamiB strain has a pronounced growth defect in both LB and M9 media and revealed that AmiB is essential for growth under low-salt conditions (Fig. 1C). Syto 9 staining showed that chaining ΔamiB cells contain numerous distinct cytoplasmic compartments, which are particularly elongated in M9 medium. Notably, the fluorescent d-alanine analogue HADA, which labels PG (30), stains bright bands along the chains between these cytoplasmic compartments, suggesting that septal PG is present as a barrier between daughter cells (Fig. 1D). In contrast, HADA staining was fairly evenly distributed around the cell periphery in wild-type cells. Together, these observations suggest that in the absence of AmiB, cleavage of septal PG is markedly impaired and that this impairment prevents completion of cell division.

AmiB's subcellular localization was monitored with a functional C-terminal AmiB-mCherry fusion that complemented the cell division defect of the AmiB mutant (Fig. 1B). In predivisional cells, ectopically expressed AmiB-mCherry was found predominantly at the cell poles and dispersed along the cell envelope, presumably in the periplasm (Fig. 1E). In dividing cells, the fusion protein is also detected at the midcell, consistent with its apparent role in cell division. Accumulation of AmiB-mCherry at the midcell coincided with or slightly followed constriction, as seen by plotting fluorescence intensity and cell width against cell length (Fig. 1E, demograph analysis; also see also Fig. S1 in the supplemental material), suggesting that like E. coli AmiB and AmiC (2), V. cholerae AmiB is recruited to the cell division site during the late stages of cell division.

EnvC and NlpD are functionally redundant in vitro, and a strain lacking both proteins phenocopies a ΔamiB strain.

V. cholerae encodes two proteins that are similar to the E. coli amidase activators EnvC and NlpD: VC0335, which is most similar to EnvC (33% ID and TS of 214; subsequently referred to as EnvC), and VC0533, which is similar to NlpD (47% ID and TS of 239; subsequently referred to as NlpD) (Fig. 2A). As in E. coli, the C-terminal domains of the V. cholerae orthologues lack the active-site residues typically required for enzymatic activity (Fig. 2B), suggesting that they may have a regulatory rather than enzymatic function. Given that in E. coli, there is a single dedicated activator for each amidase, we anticipated that either EnvC or NlpD would be the cognate activator of AmiB; however, single deletions of envC or nlpD did not reproduce the cell division defect observed in the ΔamiB mutant (Fig. 2C). In addition, the growth rates of the ΔenvC and the ΔnlpD strains were indistinguishable from those of the wild type in complex, minimal, and salt-free media (Fig. 2D). We used fluorescent fusion proteins to explore the subcellular localization of EnvC and NlpD. EnvC-mCherry was evident at the division site prior to cell constriction; thus, like for E. coli EnvC, its localization at the midcell is apparently not dependent upon septal PG synthesis. Prior to midcell accumulation, EnvC-mCherry was enriched at cell poles and present at low levels in the remainder of the cell (Fig. 2E; see also Fig. S1 in the supplemental material). In contrast, fluorescence in cells expressing mCherry-NlpD was readily detectable across the entire cell, perhaps because the fusion protein is partially degraded (Fig. 2E; see also Fig. S1 and S2). Enrichment of NlpD was evident at cell poles and/or at the midcell in most cell images, although demograph analyses were less compelling than for EnvC, making these results, particularly regarding the timing of NlpD accumulation at the midcell, not entirely conclusive. It should be noted that the mCherry-NlpD fusion protein does not appear (based on complementation analyses described below) to be equal in activity to the untagged protein, and it is possible that the fusion protein's subcellular distribution also does not fully correspond to that of native NlpD.

FIG 2.

Deletions of envC and nlpD have only minor effects on cell division. (A) Comparison of predicted domain structures of V. cholerae and E. coli EnvC and NlpD. EnvC contains a Sec export sequence (red), two coiled-coil regions (AA 38 to 128 and AA 153 to 250; CC), and a C-terminal M23 peptidase domain (AA 282 to 376). NlpD contains a Sec export sequence followed by an SPII cleavage site (AA20; black arrowhead), a LysM domain (AA 47 to 91), and a M23 peptidase domain (AA 211 to 306). (B) Sequence comparison of active center regions of V. cholerae EnvC (VC_0335) and NlpD (VC_0533) and E. coli EnvC (ENVC_Ecoli) and NlpD (NLPD_Ecoli) with active center regions of active M23 peptidases Staphylococcus aureus LytM (LYTM_STAAC) and Bacillus subtilis SpoIIQ (SP2Q_BACSU). The first (Y/X) and last (H) residues of the zinc binding motif are degenerate in EnvC and NlpD. (C) Representative images of V. cholerae wild-type, ΔenvC, and ΔamiB cells in LB medium. (D) Growth curves of wild-type, ΔenvC, and ΔnlpD V. cholerae in LB, salt-free LB, and M9 media. Bars represent means ± SEM for at least 4 replicates. (E) Representative fluorescence images showing localization of ectopically expressed mCherry-nlpD and envC-mCherry in wild-type cells grown in M9 medium are shown along with demographs of mCherry-NlpD and EnvC-mCherry localization and constriction. Fluorescence intensity and constriction (width) are indicated by a rainbow color scale from blue (0) to dark red (100%). Scale bars: 5 μm.

Since deletion of envC and nlpD individually had no effect on cell division, we suspected that EnvC and NlpD might be functionally redundant in activation of AmiB. Consistent with this hypothesis, we found that a mutant lacking both envC and nlpD exhibited a marked cell division defect that phenocopies the cell shape and cell growth deficiencies of the ΔamiB strain in several growth media (Fig. 3A to C). Additionally, as in the ΔamiB strain, HADA-stained PG accumulated at the cell division sites of ΔenvC ΔnlpD cells, and Syto 9 staining showed that chaining cells were comprised of distinct cytoplasmic compartments (Fig. 3C). Thus, our data suggest that the presence of EnvC or NlpD, like that of AmiB, is required for partitioning of septal PG and completion of cell division by V. cholerae.

FIG 3.

An nlpD envC mutant phenocopies the amiB mutant. (A) The ΔenvC ΔnlpD double mutant phenocopies the ΔamiB mutant. Shown are growth curves of wild-type, ΔamiB, and ΔenvC ΔnlpD V. cholerae in LB, M9CA, and M9 media and representative phase-contrast images of ΔenvC ΔnlpD cells grown to exponential phase in these media. (B) Growth curves of wild-type, ΔamiB, and ΔenvC ΔnlpD V. cholerae in salt-free LB medium. (C) Representative fluorescence images of wild-type and ΔenvC ΔnlpD cells grown in LB or M9 and stained with HADA (which labels PG) or Syto 9 (which marks the cytoplasm) are shown. (D and E) Artemis plots of transposon insertions in envC (VC0335) and nlpD (VC0533) in the genome of wild-type V. cholerae (C6706) versus those of the ΔnlpD (D) and ΔenvC (E) mutants. Red bars represent the numbers of reads from insertions facing forward on the plus strand, and green bars reflect those from insertions on the minus strand. Horizontal black bars represent all potential transposon insertion sites (TA sites) in the indicated regions. (F) Representative phase-contrast images of ΔenvC ΔnlpD cells and ΔenvCΔnlpD cells ectopically expressing EnvC-mCherry or NlpD. (G) Artemis plots of transposon insertions in ftsEX in the genomes of wild-type, ΔenvC, and ΔnlpD V. cholerae. Scale bars: 5 μm.

To validate the hypothesis that NlpD and EnvC are functionally redundant, we performed transposon insertion sequencing (TnSeq) analysis on both ΔenvC and ΔnlpD single mutants. We observed a marked reduction in the frequency of transposon insertions in envC in the absence of nlpD (relative to that in the wild-type strain; P < 1 × 10⁻¹⁵ versus the value for the WT) (Fig. 3D) and a comparable deficiency of insertions nlpD in the absence of envC (P < 1 × 10⁻¹⁵ versus the value for the WT) (Fig. 3E). Additionally, we found that there were significantly fewer transposon insertions in the 5′ region, but not the 3′ region, of nlpD in the absence of envC than in wild-type cells (P < 0.001). These results, coupled with knowledge of these proteins from other organisms, suggest that EnvC and the N-terminal region of NlpD are functionally redundant. Additionally, they suggest that the M23 domain encoded in the C-terminal part of nlpD is dispensable even in the absence of EnvC and thus that the mechanism of action of V. cholerae NlpD and E. coli NlpD (assuming that it functions in a manner similar to that of EnvC [14]) may not be equivalent. Unlike expression of EnvC-mCherry and untagged NlpD, expression of mCherry-NlpD and of an untagged truncated NlpD lacking the M23 domain did not complement the cell division defect of the ΔenvC ΔnlpD mutant (Fig. 3F and data not shown). However, the latter result does not rule out the possibility that the C terminus of NlpD is dispensable, as the truncated protein we generated may be misfolded or unstable.

We examined our TnSeq data for additional loci that are essential in the nlpD or envC mutants but not in wild-type cells, as these are likely to also contribute to AmiB activity. We found that there were almost no transposon insertions in ftsE and ftsX in the absence of NlpD, while these two cell division genes were disrupted in the envC mutant as frequently as the in the wild-type background (P = 0.00003 [ftsE] and P < 1 × 10⁻¹⁵ [ftsX]), suggesting that FtsEX likely contribute to EnvC activity. No other genes were found to be conditionally essential (i.e., conferred a severe growth deficiency) only in the envC or nlpD mutants.

Localization of AmiB is dependent on FtsN, EnvC, and NlpD.

In addition to TnSeq and growth analyses, we assessed whether there was interdependent localization of the components of AmiB-related pathways. We found that AmiB-mCherry was distributed normally—predominantly at cell poles and at the midcell—in ΔenvC and ΔnlpD cells; however, it showed a largely diffuse distribution in cells lacking both activators (Fig. 4A). Occasional foci of AmiB-mCherry were observed; however, they generally did not colocalize with cell constriction sites or poles, suggesting that they might result from nonspecific aggregation. Thus, our data suggest that in contrast to that of E. coli amidases (2), reliable localization of V. cholerae AmiB at the septum is dependent on the presence of at least one of its activators.

FIG 4.

Localization determinants of AmiB, EnvC, and NlpD. (A) Representative phase-contrast and/or fluorescence images of ΔenvC, ΔnlpD, and ΔenvC ΔnlpD cells expressing ectopic amiB-mCherry. (B) Representative fluorescence images of cells expressing FtsN under the control of an IPTG-inducible promoter, grown in M9 in the presence of IPTG (+FtsN) or for 12 h without IPTG (depletion). Cells also produce AmiB-mCherry, EnvC-mCherry, or mCherry-NlpD, whose expression was induced with 0.2% or 0.4% (EnvC-mCherry) arabinose for 2 to 3 h. (C) Representative fluorescence images of ftsX-deficient strains expressing ectopic amiB-mCherry, envC-mCherry, or mCherry-nlpD. (D) Bacterial adenylate cyclase two-hybrid analysis of interactions between AmiB, EnvC, NlpD, FtsE, and FtsX. Shown are colonies of cya-negative strains producing T25 and T18 fusions of the respective proteins on LB medium supplemented with X-Gal and IPTG. Periplasmic proteins were fused to an exogenous transmembrane domain (tm), as described previously (24). Positive interactions are indicated by blue color. Scale bars: 5 μm.

We also assessed whether recruitment of AmiB and/or its activators was dependent on the presence of FtsN at the midcell, i.e., on complete assembly of a divisome. Since FtsN is an essential protein, these experiments were performed with a strain containing inducible ftsN, in which FtsN depletion is reflected by formation of elongated cell chains without evident constrictions (Fig. 4B). AmiB-mCherry, mCherry-NlpD, and EnvC-mCherry were not found in bands spanning the width of these cells (i.e., septum-like structures), suggesting that FtsN or other septal factors are needed for localization of both activators and thus for localization of AmiB as well. Notably, none of the fusion proteins appear to be mislocalized in the absence of FtsX or FtsE (Fig. 4C; see also Fig. S3 in the supplemental material; however, as noted above, discrete localization of NlpD was not always evident). Thus, V. cholerae's means of localizing amidase activators (specifically EnvC) as well as localization of its PG amidase differ from that observed in E. coli. Given that localization of amidases and activators is thought to be an important means of regulating amidase activity in E. coli (2), it seems likely that V. cholerae has evolved alternate means to control its PG amidase.

BACTH analysis suggests that multiple interactions occur among septal amidase-associated proteins.

To gain further insight into the means by which AmiB, EnvC, and NlpD are recruited to the division site, we assessed whether these proteins interact with each other and with FtsEX, which appear to be required for EnvC-mediated AmiB activity (although not localization). Somewhat unexpectedly, our bacterial adenylate cyclase two-hybrid (BACTH) assay revealed interactions between almost all constituents (Fig. 4D). Only FtsE (a cytoplasmic protein) showed a relatively limited interaction profile, interacting only with its known partner FtsX (a transmembrane protein) and not with the other proteins tested, consistent with their periplasmic localization. (For the BACTH assays, periplasmic proteins—i.e., AmiB, NlpD, and EnvC—were linked to the cytoplasmic reporter domains via an exogenous transmembrane domain that was also included in the control vectors, as described previously [24].) AmiB interacted with NlpD and EnvC, as has been presumed from in vitro assays (3) but not previously demonstrated in vivo, and also interacted with FtsX. However, AmiB interacted with FtsX and NlpD only when expressed from a high-copy-number plasmid (T18) and not when expressed from the lower-copy-number T25, suggesting that these AmiB interactions may be relatively weak or transient and therefore require higher protein concentrations for detection. Interestingly, FtsX interacted with both NlpD and EnvC, although genetic analyses suggested that it is required only for EnvC activity, and NlpD and EnvC also interacted with each other. Collectively, these results suggest that amidase activity may be associated with assembly of relatively large complexes of periplasmic proteins at the division septum and potentially with regulatory cross talk among these factors.

PG compositions are similar but not identical in the ΔamiB and ΔenvC ΔnlpD mutants.

To further analyze the function of AmiB, EnvC, and NlpD in V. cholerae, we analyzed the PG compositions of all three single mutants as well as the ΔenvC ΔnlpD double mutant (Fig. 5A). The most marked change in PG composition for the ΔamiB mutant was in the level of M2 peptides, which was elevated >10 times relative to PG from wild-type bacteria, suggesting that M2 peptides are the preferred substrate for this enzyme. The ΔnlpD and ΔenvC single mutants did not contain an increased amount of M2 peptides, but M2 peptides were elevated ∼4 times for the ΔenvC ΔnlpD mutant, consistent with the expected role of EnvC and NlpD as functionally redundant AmiB activators. Unexpectedly, PG from the ΔenvC ΔnlpD mutant also contained increased levels of M3 and M5 peptides, which were not evident in PG from the ΔamiB strain. This observation raises the possibility that EnvC and NlpD also modulate the activity of additional PG-modifying enzymes, such as endopeptidases that reduce the length of PG peptide side chains prior to their cleavage by AmiB, which might also account for the less dramatic increase in M2 peptides for the activator-deficient strain compared to the ΔamiB mutant.

FIG 5.

Peptidoglycan composition of amidase-activator complex mutants. (A) UPLC analysis of PG from wild-type (C6706), ΔamiB, ΔenvC, ΔnlpD, and ΔenvC ΔnlpD V. cholerae. Graphs were normalized by scaling the chromatograms relative to the maximum intensity measured in each run. Altered peaks are indicated by arrows. (B) Comparison of relative molar abundances of M2, M3, and M5 monopeptides. Asterisks indicate statistically significant differences based on an unpaired t test (*, P = 0.01 to 0.1; **, P = 0.001 to 0.01; ***, P < 0.001). Schematic representations of the respective muropeptide units are depicted on the right. Relative amounts of muropeptides were calculated as described by Glauner (31). The values are the means of results from two independent experiments. Error bars represent standard deviations of the means.

Differential requirement for NlpD versus EnvC for V. cholerae intestinal colonization.

To elucidate the role of AmiB and its activators during V. cholerae intestinal colonization, we studied the colonization efficiency of the mutant strains in the infant mouse model of cholera. Given the severe cell division defects in the ΔamiB and ΔenvC ΔnlpD strains, our observation that these mutants were completely outcompeted by the wild type in vitro and in vivo (Fig. 6A) was expected. Surprisingly however, the single nlpD and envC mutants exhibited significantly different capacities to colonize the intestine. While neither strain exhibited a competitive defect in competition assays versus the wild-type strain in vitro, the ΔnlpD strain was significantly outcompeted in vivo (at least 20-fold), regardless of the growth phase of the inoculum (Fig. 6A). Thus, although nlpD and envC appear to function in a redundant fashion in laboratory media (Fig. 2), during growth in the host intestine, NlpD's role cannot be filled by EnvC. The increased importance of NlpD for in vivo growth is corroborated by two recent transposon insertion sequencing studies that revealed that nlpD mutants (but not envC mutants) have reduced fitness in the infant rabbit model of cholera (25, 37).

FIG 6.

NlpD is required for effective colonization of the infant mouse intestine and resistance against detergents and bile salts. (A) In vitro and in vivo competition assays between wild-type V. cholerae and the indicated mutants. Asterisks indicate statistically significant differences based on the Mann-Whitney t test (**, P = 0.001 to 0.01; ***, P < 0.001). When no mutant bacteria were detected in vivo, the limit of detection (based on assumed isolation of a single colony) was calculated and plotted as an open circle. “lacZ” indicates the lacZ-negative strain. (B) Growth curves of wild-type, ΔenvC, ΔnlpD, ΔamiB, and ΔenvC ΔnlpD V. cholerae strains in LB medium supplemented with 0.04% SDS. (C) Growth curves of strains in panel B in LB medium supplemented with 5% bile salts. (D) Representative images of wild-type, ΔamiB, ΔenvC, ΔnlpD, and ΔenvC ΔnlpD cells grown to exponential phase in LB medium containing bile salts. Scale bar: 5 μm.

Previous analyses have revealed that amidase-deficient E. coli isolates are sensitive to detergents (e.g., deoxycholate, a bile acid) and high-molecular-weight antibiotics that do not penetrate or inhibit growth of wild-type bacteria, suggesting that these mutants have impaired outer membrane integrity (9). As V. cholerae likely encounters bile within the intestinal tract, we assessed the susceptibility of our mutants to bile salts (1.25 to 10%) and to the detergent SDS (0.04%). Growth of the ΔnlpD, ΔamiB, and ΔenvC ΔnlpD mutants but not the ΔenvC mutant was reduced, relative to that of the wild-type strain, under all conditions (Fig. 6B and C). In the presence of SDS and 10% bile salts, the ΔnlpD, ΔamiB, and ΔenvC ΔnlpD strains all failed to grow; however, bile salts had more severe consequences for the ΔamiB and ΔenvC ΔnlpD strains than the ΔnlpD strain at lower concentrations (Fig. 6B and data not shown), which perhaps is reflected in the more severe colonization deficiencies of the ΔamiB and ΔenvC ΔnlpD strains. Further analysis revealed that a fraction of the ΔnlpD cells and all of the ΔamiB and ΔenvC ΔnlpD cells lyse in the presence of 1.25% bile salts, while no lysis is apparent for wild-type and ΔenvC cells (Fig. 6D). Thus, although in many assays EnvC and NlpD appear to be functionally redundant in V. cholerae, these analyses demonstrate that their roles do not fully overlap. It is not clear why amidase activators or AmiB influences outer membrane integrity, or whether the absence of NlpD impairs membrane integrity under all conditions or only when cells are subjected to particular environmental stresses.

DISCUSSION

Our analyses of the factors required for septal PG cleavage and cell separation in V. cholerae revealed that a single amidase—AmiB—generally permits septal PG cleavage when at least one of its two activators, EnvC and NlpD, is present. In the absence of the amidase or of both activators, V. cholerae forms extended chains of cells that are linked by shared bands of PG at cell poles. At best, these cells have a reduced growth rate relative to that of wild-type cells, and they fail to survive exposure to low-salt media, SDS, and bile salts. The activators EnvC and NlpD appear to be functionally redundant under most growth conditions; however, NlpD is more key to V. cholerae growth in a subset of conditions, including exposure to detergents and growth within an animal model of infection, indicating that AmiB's activators also likely have distinct roles and/or regulators that modulate PG hydrolysis in response to environmental stimuli.

Septal cleavage in V. cholerae differs from the more-studied pathways in E. coli in several respects. First, in possessing only a single periplasmic PG amidase for cleavage of septal PG, V. cholerae differs from the majority of gammaproteobacteria, which typically contain three members of this amidase family. However, reliance on a single amidase is likely common among alphaproteobacteria, which encode only a single predicted amidase_3-type amidase (see Fig. S4 in the supplemental material). Second, V. cholerae's AmiB is generally functional when either of two potential activators is present. It is not clear why V. cholerae's activator proteins have a broader partner range than seen in E. coli; however, since most Vibrionaceae contain a large number of putative M23 peptidases (which include NlpD and EnvC homologues as well as other factors) (see Fig. S5 in the supplemental material), it is likely that this regulatory paradigm is maintained throughout this genus. Third, although the activity of V. cholerae EnvC appears to be dependent upon the cell division proteins FtsEX as in E. coli, localization of V. cholerae EnvC is FtsEX independent and instead requires the presence of FtsN. Finally, AmiB localization requires the presence of at least one of its activators, whereas localization of E. coli amidases is activator independent. Thus, although the cellular components and processes that culminate in cleavage of septal PG appear to be generally conserved among these related gammaproteobacteria, they can be combined into diverse and functional regulatory networks.

We had expected, based on mutational studies of E. coli EnvC, that NlpD's homologous C terminus would contribute to amidase activation. However, analyses of transposon insertion sites suggest that the C-terminal region of V. cholerae NlpD is dispensable, even in the absence of EnvC. Additional studies of truncated or mutated NlpD and EnvC are needed to more precisely map sequences required for AmiB activation. It is possible that NlpD family and EnvC family activators act via distinct mechanisms, despite their homology and similar roles; alternatively, it is also possible that homologous E. coli and V. cholerae activators have evolved distinct features, perhaps because their amidase targets in the two organisms do not correspond.

Although V. cholerae NlpD and EnvC appear to be functionally redundant under the majority of growth conditions assayed (LB, minimal media, salt-free media, etc.), strains lacking NlpD have a significant growth impediment within the infant mouse intestine and when cultured in the presence of bile salts or other detergents, and they show a marked tendency to lyse in bile. Similar, but more severe, phenotypes are seen for the AmiB-deficient strain, while the absence of EnvC has no apparent effect on wild-type cells. It has previously been reported that amidase-deficient E. coli is also sensitive to components of bile and other agents, which is thought to reflect weakness of the outer membrane (9). Resistance against bile has been previously linked to cell envelope structures, such as lipopolysaccharide (38), and to reduced levels of PG-bound lipoprotein (39, 40). Since NlpD is predicted to be an outer membrane protein, it may contribute to the release of Lpp-bound peptides by preferentially activating AmiB near outer-membrane-proximal PG. In contrast, EnvC is thought to be tethered to the inner membrane, via its interaction with FtsX, and therefore might play a less prominent role with respect to the outer membrane. Given the proteins' presumably disparate membrane targeting, as well as their independent activities, it is rather surprising that two-hybrid analyses revealed an interaction between NlpD and EnvC. Interactions spanning the PG layer are not unprecedented in Gram-negative organisms; indeed, PBP1A and PBP1B—both inner membrane proteins—are both thought to be activated by proteins in the outer membrane (LpoA and LpoB, respectively), probably via contacts through gaps in the PG layer. Nonetheless, from a mechanistic standpoint, it is not clear what is gained from an interaction between NlpD and EnvC. Potentially, this question could be addressed via biochemical assays of their activities; however, to date we have not found it possible to overexpress the V. cholerae activators or V. cholerae AmiB, and hence in vitro analyses of purified proteins have not been performed. We also cannot exclude the possibility that the observed interaction between NlpD and EnvC is a consequence of NlpD mislocalization due to the presence of additional sequences (including an inner membrane-spanning domain) in the NlpD BACTH constructs.

In future studies, purification of V. cholerae AmiB and its activators will also provide opportunities for precisely defining the substrate of V. cholerae AmiB. Our analyses of PG from the amiB mutant revealed that it contains an elevated level of monomeric M2 peptides, suggesting that AmiB is needed to release dipeptides from PG. Interestingly, PG amidase-deficient E coli accumulates trimeric and tetrameric muropeptides (8), and E. coli AmiB was able (in the presence of EnvC) to release tetrapeptides from PG in vitro (3), raising the possibility that the V. cholerae and E. coli enzymes have different substrate specificities. However, there is currently no evidence that the E. coli amidase cannot also utilize a dipeptide substrate or that V. cholerae AmiB cannot release longer peptide chains (e.g., trimers and tetramers); the latter may not accumulate in the V. cholerae amiB mutant simply because they can also be processed (e.g., to dimers) by other enzymes. Interestingly, tri- and tetrapeptides do accumulate when AmiB's activators are absent, suggesting that the activators may facilitate cleavage of larger side peptides by enzymes other than AmiB, as proposed for the E. coli proteins (3).

It is not known whether V. cholerae's specific requirement for NlpD rather than EnvC in the presence of bile and its heightened requirement for AmiB and both activators in low-salt media indicate that the proteins' expression or activities are altered under these conditions. It is possible that such conditions simply enable more sensitive assessment of the proteins' roles and of membrane integrity. However, many studies have demonstrated that PG synthesis is highly responsive to environmental conditions. For example, changes in PG composition and altered PG synthase activity contribute to bile resistance in Salmonella enterica (39, 41) and PG composition changes during stationary phase (42). Pathogenic bacteria can also remodel PG to avoid a host response (5, 43–46). Previous studies have shown that the number and activity of PG synthetic enzymes and their interaction partners vary significantly in different organisms (47). Our findings introduce the possibility that the amidase-activator complex is similarly adaptable. It will be interesting to attempt to uncover the explanation(s) for the differences in the content and control of amidase-activator complexes in diverse microorganisms in order to better understand their role in PG plasticity.