A conserved hub protein required for peptidoglycan remodeling and cell division in Acinetobacter baumannii

Significance

Bacteria rely on a rigid cell wall, termed peptidoglycan, that must be continuously remodeled to allow growth and division. This process, known as peptidoglycan turnover, is essential for cell integrity but must be carefully controlled to prevent cell death. The critical pathogen Acinetobacter baumannii was missing known peptidoglycan amidases, a class of turnover enzymes, and the key activator that controls their activity during cell division. We have identified WthA as having a role in cell division most likely as an amidase activator. WthA homologs were widely distributed in bacteria and impacts activity of two other types of turnover enzymes. We explore the possible functions of this family of proteins that serves as a hub for impacting peptidoglycan turnover.

Abstract

Gram-negative bacteria build a multilayered cell envelope in which the peptidoglycan layer is sandwiched between an inner membrane of glycerophospholipids and an asymmetric outer membrane composed of glycerophospholipids and lipopolysaccharide (LPS). Acinetobacter baumannii, however, synthesize lipooligosaccharide (LOS), an LPS variant lacking O-antigen. Although LPS/LOS is typically essential, A. baumannii can survive without LOS, offering the opportunity to examine how the Gram-negative envelope remains stable in the absence of this major glycolipid. We previously found that the peptidoglycan biogenesis protein NlpD, an activator of peptidoglycan degradation by cell division amidases, is critical for fitness during LOS deficiency. Here, we show that NlpD is required under these conditions because a second putative amidase activator, WthA (cell wall turnover hub protein A), no longer functions in LOS-deficient cells. Mutants lacking WthA exhibited severe cell-division defects and were synthetically sick with loss of NlpD. Acinetobacter lack canonical periplasmic amidases, raising the question of which enzymes partner with NlpD and WthA. Previous work showed that overexpression of an Acinetobacter β-lactamase increased denuded peptidoglycan, a product of amidase activity. Guided by this finding, we examined the chromosomally encoded β-lactamase Oxa51 and found that its coexpression with WthA or NlpD enhanced release of amidase products, suggesting that Oxa51 participates in peptidoglycan degradation and that WthA is an amidase activator. Further, WthA influenced peptidoglycan endopeptidases and lytic transglycosylases through a network of protein interactions. Altogether, these findings identify WthA as a missing regulator in Acinetobacter peptidoglycan biogenesis and a hub that coordinates peptidoglycan turnover and cell division.

Acinetobacter baumannii is a Gram-negative, opportunistic pathogen with a high incidence of multi- and extreme-drug resistance (1–3). Like other Gram-negative bacteria, its resistance to many antibiotics is partly due to the structure of its dual-membrane cell envelope. The inner membrane (IM) consists of glycerophospholipids, while the outer membrane (OM) is asymmetrical with glycerophospholipids in the inner leaflet and glycolipids, either lipooligosaccharide (LOS) or lipopolysaccharide (LPS), in the outer leaflet (4, 5). Both LOS and LPS contain a core oligosaccharide attached to a negatively charged lipid A anchor, but LPS also contains a repeating O-antigen polysaccharide (6). The tight packing of these amphipathic glycolipids on the surface of the OM creates a highly impermeable barrier that contributes both to antimicrobial resistance and to the mechanical stability of the OM (7–11).

Cationic antimicrobial peptides, including the polymyxin antibiotics, have taken advantage of the conserved negative charge of LPS/LOS to bind and disrupt the OM (12–14). A. baumannii naturally produces LOS in its OM, but in response to high levels of polymyxins can obtain mutations that completely abolish LOS synthesis (15, 16). Although this adaptation confers polymyxin resistance, it has many trade-offs including slow growth, increased sensitivity to other antibiotics, and stress in maintaining envelope integrity (11, 15–19). Few bacteria can tolerate the complete loss of LPS or LOS (15, 20–22), which makes A. baumannii a valuable model to explore novel mechanisms of cell envelope biology.

To identify key factors that support survival in the absence of LOS, we performed genome-wide transposon sequencing (Tn-seq) in cells with and without LOS (11). Genes impacting fitness in LOS-deficient cells were enriched for functions involved in the growth, maintenance, and division of the peptidoglycan (PG) cell wall. PG is a rigid mesh with chains of alternating GlcNAc and MurNAc sugars (N-acetylglucosamine and N-acetylmuramic acid, respectively) that are crosslinked by short peptides attached to MurNAc residues (23). The PG precursor, lipid II, is synthesized in the cytoplasm and is composed of the GlcNAc-MurNAc-pentapeptide subunit attached to an undecaprenyl-phosphate lipid carrier (23). Once Lipid II is flipped across the IM (24), PG synthases polymerize and crosslink new subunits into the PG layer through glycosyltransferase and transpeptidase activities, respectively (23). Because the cell wall is rigid and continuous, its expansion requires carefully regulated hydrolases (Fig. 1A) that cleave existing bonds to insert new subunits and to separate daughter cells after division. This remodeling process, known as PG turnover, must be precisely coordinated with synthesis to prevent uncontrolled degradation and cell lysis.

Fig. 1.

Acinetobacter baumannii is missing homologs of key components involved in cell division. (A) Schematic showing sites of cleavage for major classes of PG turnover enzymes. (B) In Escherichia coli, amidases (AmiA, AmiB, or AmiC) that remove PG stem peptides are highly regulated. During cell division, ATP hydrolysis by FtsEX induces a conformational change in EnvC releasing a self-inhibiting helix, allowing EnvC to activate AmiA and AmiB. These amidases generate “denuded” glycan strands lacking peptide crosslinks, which are recognized by the SPOR domain of FtsN. FtsN then stimulates divisome synthases by signaling through both FtsQLB and FtsA to activate PBP3 (transpeptidase) and FtsW (glycosyltransferase). This positive feedback loop of PG synthesis and remodeling drives completion of cytokinesis. See main text for key references. (C) A. baumannii lacks homologs of EnvC, FtsEX, and the canonical amidases, yet retains an essential FtsN homolog, suggesting that it utilizes divergent amidases and activators to coordinate PG turnover during cell division.

Our Tn-seq screen revealed that several genes involved in cell division (nlpD, tolA, tolB, tolQ, tolR, pal, and yraP) became critical for fitness in the absence of LOS (11). Many of these genes are known to coordinate PG remodeling during cytokinesis. The Tol-Pal system, which includes TolABQR and Pal, helps couple OM constriction (25) with PG remodeling and has been implicated in coordinating cell division PG turnover through NlpD (26). YraP, an outer membrane lipoprotein of unknown function, has also been implicated in this pathway (26). NlpD itself is a degenerate member of the LytM family of PG hydrolyases, but lacks the catalytic residues for PG degradation (27). Instead, NlpD localizes to septa and activates PG amidases that cleave the amide bond between MurNAc and crosslinked peptides (27, 28). These amidases completely release crosslinking peptides and produce so-called “denuded” glycan strands that signal progression of septal PG synthesis (Fig. 1B). In Escherichia coli, NlpD functions redundantly with other amidase activators, including EnvC and ActS (28–30). EnvC is the primary amidase activator in most proteobacteria (Fig. 1B), but defects in envC mutants are exacerbated by loss of NlpD, Tol-Pal, and YraP (26). Because loss of these genes decreased fitness in LOS-deficient A. baumannii, we suspected that a cell division amidase activator was no longer functioning when the OM lacks LOS. However, A. baumannii was missing homologs of envC and missing any known amidases (Fig. 1C).

From work performed largely in E. coli, cell division amidases have been proposed to participate in a positive feedback loop that drives septal PG synthesis and promotes cell division (Fig. 1B). Recruitment of divisome components is highly ordered and reviewed well by others (31, 32). Important here, the FtsEX complex acts upstream to regulate EnvC (23). After the Z-ring and divisome complexes have assembled and septal PG synthesis has begun, ATP hydrolysis by FtsEX triggers a conformational change in EnvC (33) and releases an inhibitory domain from its degenerate LytM region (34–36). Activated EnvC then stimulates the amidases AmiA or AmiB to cleave peptides from the septal PG. The resulting denuded PG is recognized by the SPOR domain of FtsN (37). Binding of FtsN to these sites triggers conformational signals that travel through two independent routes: one through FtsQLB to activate PBP3/FtsW and the other through FtsA to activate the glycosyltransferase FtsW (38–44). Increased septal PG synthesis by PBP3/FtsW accelerates recruitment of more FtsN leading to a self-enhancing cycle of septal PG synthesis that drives completion of cell division (38). A. baumannii contains a single, essential homolog of FtsN (45, 46), suggesting that it retains this cell division reinforcement loop (Fig. 1C). However, it lacks homologs of FtsEX, EnvC, and the canonical amidases AmiABC, leaving key steps in this regulatory loop unresolved.

Using the LOS-deficient background as a genetic tool, we set out to identify factors important for cell division including the missing amidase activation machinery. We found that HMPREF0010_01334 from strain ATCC 19606, which we designate as WthA, plays a critical role in cell division and growth. WthA is predicted to be localized to the periplasm and contains a high number of TPR (tetratricopeptide repeat) domains that facilitate protein–protein interactions. Cells lacking wthA exhibited severe chaining, lysis, and osmotic sensitivity. These defects were exacerbated with loss of nlpD suggesting that WthA functions as an amidase activator. Indeed, there is a precedent for involvement of TPR domains in amidase activation, as Staphylococcus aureus ActH also uses TPR domains for this purpose (47, 48). Additionally, we demonstrate that both NlpD and WthA induce formation of PG amidase products when coexpressed with the β-lactamase Oxa51, consistent with an amidase activator role. These findings also reinforce the recent proposal that A. baumannii β-lactamases can both degrade β-lactams and cleave PG (49, 50).

Beyond amidase activation, we find that WthA interacts with multiple PG turnover enzymes and have named it cell wall turnover hub protein A. WthA homologs are widely distributed among Proteobacteria, Bacteroidota, and Chlorobiota. LbcA, an ortholog of WthA in Pseudomonas aeruginosa, binds to lytic transglycosylases that cleave the glycan chains of PG and is an adaptor protein for proteolysis of endopeptidases that cleave PG crosslinks (47, 48, Fig. 1A). We show that LbcA’s role in regulating multiple PG turnover enzymes is conserved in A. baumannii WthA. However, the existing literature and data presented here indicate that the Pseudomonas homolog likely does not serve as an amidase activator. Overall, our findings suggest that WthA/LbcA homologs are a family of hub proteins for regulation of PG turnover enzymes with distinct mechanisms.

Results

HMPREF0010_01334 Is Involved in Cell Division and Does Not Function in LOS-Deficient Cells.

We suspected that an unknown cell division amidase activator was not functioning in LOS-deficient A. baumannii based on the following logic. In E. coli, loss of NlpD, Tol-Pal, or YraP results in a synthetic sick phenotype with loss of the primary amidase activator EnvC (26, 28). In LOS-deficient A. baumannii, NlpD, Tol-Pal, and YraP were found to be critical for fitness (11) (SI Appendix, Fig. S1A). A. baumannii lacks a homolog of EnvC, but retains an essential FtsN homolog (45, 46) (HMPREF0010_03200) that contains a SPOR domain for sensing denuded PG, the product of amidase activity (Fig. 1). This suggested that an amidase activator must exist, even though its identity was unknown. NlpD was an unlikely candidate for the primary activator, since it is not essential in LOS-containing cells and its disruption causes only rare chaining (SI Appendix, Fig. S1). In agreement with this logic, LOS-deficient strains exhibit chaining defects (19) (Fig. 2A) and naturally increase expression of NlpD (16), possibly to compensate for reduced function of another amidase activator. If the missing activator were nonfunctional in LOS-deficient cells, the gene encoding it may have a positive fold-change in our LOS-deficient Tn-seq data, similar to all genes involved in synthesis and transport of LOS. For example, lpx and lpt genes are easily disrupted in LOS-deficient cells because they are dispensable in this genetic background (11). Among candidate genes, HMPREF0010_01334 (renamed here wthA for cell wall turnover hub protein A) fit these criteria. Tn-seq analysis showed low read counts (RPKMs) for insertions in wthA in cells with LOS and a positive-fold change in LOS-deficient cells (11) (SI Appendix, Fig. S1A).

Fig. 2.

WthA (HMPREF0010_01334) is required for cell division and is nonfunctional in LOS-deficient A. baumannii. (A) Phase contrast microscopy of LOS-deficient and wthA mutants, representative of biological triplicates. (B) wthA operon structure and description of wthA mutant alleles. (C and D) Biological triplicate growth curves of wthA mutants in LB (C) or LB with 10% glycerol (D). Error bars are not visible when smaller than the size of the symbol. (E) Growth recovery of wthA and LOS-deficient mutants at 37 °C after storage of colonies for 7 to 8 d at either room temperature (RT) or 4 °C, representative of biological triplicate.

First, we constructed a wthA deletion mutant and assessed its morphology by microscopy. wthA is in an operon with lolB and contains the lolB ribosome-binding site within its open reading frame (ORF). To prevent polar effects on lolB, we deleted nucleotides 1-1535 of the 1638-nucleotide wthA ORF (Fig. 2B). We also constructed an in cis complement strain that mimicked this alteration to lolB, introducing a new RBS for lolB, while leaving wthA intact (Fig. 2B). The wthA mutant exhibited a severe cell division defect characterized by long, kinked chains, evident lysis, and occasional branched cells (Fig. 2A and SI Appendix, Fig. S2). The in cis complement strain was indistinguishable from the wild-type parent, indicating the division defects were not due to polar effects on lolB (Fig. 2A). Transmission electron microscopy confirmed these defects showing cells with detached OMs, elongated chains with several division sites, and extension of some division sites that produced the characteristic kinked appearance (SI Appendix, Fig. S2B).

Next, we performed growth curves in LB and in a hyperosmotic condition, LB supplemented with 10% glycerol. In LB, the mutant showed a drastic reduction in growth rate compared to wild type or the complement (Fig. 2D). Under hyperosmotic conditions, ΔwthA cells failed to grow (Fig. 2E), suggesting compromised cell wall rigidity. Loss of WthA also caused cell lysis during growth, as indicated by a drop in optical density after ~8-10 h (Fig. 2D) and the accumulation of cellular debris in stationary-phase cultures (SI Appendix, Fig. S2A).

To further exclude polar effects on lolB, we attempted genetic complementation in trans. The ΔwthA mutant was initially nontransformable, even with repeated attempts, so we disrupted a type II restriction endonuclease operon (HMPREF0010_00560 and HMPREF0010_00561) (SI Appendix, Fig. S3A) to improve transformation efficiency. The resulting strain, named 19606e for electrocompetent, exhibited a ~130-fold increase in transformation efficiency (SI Appendix, Fig. S3). Complementation with the wthA gene alone or in combination with lolB in trans fully restored growth and morphological defects, whereas (Fig. 3 and SI Appendix, Fig. S3) a plasmid carrying only lolB did not (Fig. 3 and SI Appendix, Fig. S3). These results confirmed that the phenotypes of ΔwthA were not due to polar effects on lolB.

Fig. 3.

Defects from deletion of wthA are not due to polar effects on lolB. (A) Biological triplicate growth curves in LB of the 19606e ΔwthA mutant with different plasmids to assess for complementation. Error bars are not visible when smaller than the size of the symbol. (B) Representative phase contrast microscopy of biological triplicates of the indicated strains. The scale bar is the same for all microscopy images. (A and B) IPTG was used at a final concentration of 25 μM.

While maintaining the mutant, we noted that ΔwthA colonies showed poor growth after agar plates were stored at 4 °C. This was striking because it is a phenotype shared by LOS-deficient strains. To test this formally, we compared the survival of ΔwthA (LOS⁺) and wild-type LOS-deficient cells after storage at room temperature (RT) or 4 °C for 7 to 8 d. When colonies were restruck on LB agar and incubated at 37 °C, both ΔwthA (LOS⁺) and LOS-deficient mutants exhibited markedly reduced growth after cold storage (Fig. 2F). This shared phenotype supports the idea that WthA activity is impaired in LOS-deficient cells, consistent with its proposed loss of function under these conditions.

We next attempted to characterize the cell division defect in ΔwthA using fluorescent labeling of the cell wall. The mutant incorporated fluorescent D-alanine derivatives poorly under all conditions tested; however, membranes and late-stage division planes could be labeled with the fluorescent FM 4-64 probe. The wild-type strain tolerated FM 4-64 labeling at all growth stages (SI Appendix, Fig. S4), whereas ΔwthA lysed extensively when labeled during logarithmic growth (OD₆₀₀ ~0.5 in cuvettes). Still, cells could be labeled without impacting viability when growth began to slow (OD₆₀₀ ~0.8 in cuvettes, SI Appendix, Figs. S5 and S6). Even with lysis, microscopy during rapid growth revealed numerous chains with visible late-stage division planes (SI Appendix, Fig. S5). Many of these chains showed partial lysis between adjacent compartments, suggesting that septation had occurred and cytoplasmic compartments in a chain were distinct. (SI Appendix, Fig. S5). With extended growth, a mixture of altered morphologies was evident including curved filaments, chains, and branched cells (SI Appendix, Fig. S6), consistent with severe disruption of the division process. It is not clear why FM 4-64 dye was toxic to the ΔwthA mutant during rapid growth, but it correlated to when cells were chained with a high number of unresolved septa.

WthA Impacts Multiple PG Turnover Enzymes.

The phenotypes observed were all consistent with loss of WthA causing a defect in PG biogenesis. To define which stage of PG synthesis or turnover was affected, we examined sensitivity to PG-targeting antibiotics. Loss of wthA did not impact sensitivity to fosfomycin, which targets MurA, suggesting that synthesis of lipid II precursors remained normal (SI Appendix, Table S1). The wthA mutant showed increased resistance to vancomycin, which targets D-Ala-D-Ala termini, and increased sensitivity to β-lactams that inhibit penicillin binding proteins (PBPs). This pattern suggested that transpeptidation activity by PBPs was low, resulting in a build-up of uncrosslinked D-Ala-D-Ala (SI Appendix, Table S1). Since amidase activity normally signals FtsN-dependent activation of the FtsW/PBP3 synthase complex (Fig. 1B), this antibiotic profile supported the possibility that WthA functions as a cell division amidase activator.

Suppressor selections were performed during growth in hyperosmotic media to further understand the PG defect. The ΔwthA mutant was serially passaged in LB with increasing concentrations of glycerol. After transfer from media containing 5% to 10% glycerol, two populations regained growth and were subjected to whole-genome sequencing. Population 1 contained a frameshift mutation that would disrupt the MepM endopeptidase, mepM(G156fs), and population 2 contained a frameshift that would disrupt the MltD lytic transglycosylase, mltD(V307fs) (SI Appendix, Fig. S7A). The P. aeruginosa WthA ortholog, named LbcA, binds both MepM and MltD (51, 52) (SI Appendix, Fig. S7B). LbcA serves as an adaptor to direct three endopeptidases (MepM, PA1198, and PA1199) to the protease CtpA (51, 52). LbcA also binds the lytic transglycosylases, MltD and RlpA, although these proteins are not protease substrates (52). The isolation of mepM and mltD suppressors in Acinetobacter suggested that WthA influences these enzymes through a conserved mechanism similar to Pseudomonas LbcA (SI Appendix, Fig. S7 B and C). To test this directly, both mepM and mltD were deleted individually or in combination with wthA. The mepM and mltD single mutants grew normally in LB and LB with 10% glycerol (SI Appendix, Fig. S7 D and E). When combined with ΔwthA, both the ΔmepM ΔwthA and ΔmltD ΔwthA double mutants partially rescued growth in LB with 10% glycerol and fully rescued growth in LB (SI Appendix, Fig. 4 A and B).

Fig. 4.

WthA is a hub protein that interacts with multiple peptidoglycan turnover enzymes. (A and B) Growth curves of ΔwthA suppressors in LB + 10% glycerol (A) or LB alone (B). Data are from biological triplicate and error bars are not visible when smaller than the size of the symbol. (C) Immunoblot showing chromosomally expressed MepM-His levels in the presence and absence of WthA and the protease CtpA. Immunoblot of PBP1A was performed as a loading control. Biological triplicates are shown. (D) PG biogenesis proteins copurified with HA-tagged WthA identified by shotgun proteomics following immunoprecipitation. FC denotes fold-change in peptide abundance relative to control immunoprecipitations from strains containing empty vector. (E) Listing of phenotypes observed for ΔlbcA and ΔwthA mutants. (F) Phase contrast microscopy of ΔwthA suppressors, representative of biological triplicates.

These results suggested that WthA impacts MepM and MltD similarly to LbcA. To determine whether WthA facilitates CtpA-dependent proteolysis of MepM, we monitored protein levels of a chromosomal His-tagged allele of mepM in strains with or without WthA and CtpA. MepM-His was undetectable when both WthA and CtpA were present but increased dramatically when either protein was absent (Fig. 4C), indicating that WthA participates in proteolytic regulation of MepM. To test whether WthA associates with multiple PG biogenesis proteins similar to LbcA, HA-tagged WthA was overexpressed, crosslinked with EGS [ethylene glycolbis(succinimidyl succinate)], and immunoprecipitated from wild-type A. baumannii. The elutant was subjected to mass spectrometry to identify coimmunoprecipitated proteins. Significantly enriched hits were defined relative to two independent controls: an empty plasmid and a cytoplasmic HA-tagged control protein (ElsL-HA) (Dataset S1). HA-WthA copurified with CtpA, an RlpA-like protein, and MltD, interactions conserved with its LbcA ortholog (Fig. 4D). Additional PG biogenesis proteins were also enriched in the pull-down (Fig. 4D), confirming that WthA functions as a hub for multiple PG turnover enzymes in Acinetobacter.

We suspected that changes in MepM and MltD function alone were unlikely to explain the pronounced cell division defects of ΔwthA because it is well documented that Pseudomonas lbcA mutants show no growth or cell division defects (Fig. 4E) (51). Increased activity of MepM and MltD would be expected to be highly toxic if PG synthesis were already reduced, as suggested by the antibiotic sensitivity profile of a ΔwthA strain (SI Appendix, Table S1). A similar effect has been observed for β-lactam inhibition of PG synthesis, during which lysis is attributed to PG degradation still being high while synthesis is low (28, 53–56). The ability of ΔmepM and ΔmltD to rescue ΔwthA cell division defects was assessed and both ΔmepM ΔwthA and ΔmltD ΔwthA retained some chaining and morphological defects (Fig. 4F and SI Appendix, Fig. S7F). This suggested that our initial suppressor mutations rescue growth by reducing PG degradation to match decreased PG synthesis. Notably, P. aeruginosa utilize EnvC as the primary amidase activator for cell division (57), whereas A. baumannii are missing EnvC (SI Appendix, Fig. S7 B and C). Together, these observations led us to hypothesize that WthA also contributes to amidase activation during cell division, providing a functional link between its hub activity and septal PG turnover.

WthA and NlpD Alterations Have Synthetic Effects, Suggesting Partial Redundancy.

Since LOS-deficient cells increase expression of NlpD (16) and WthA activity is compromised in this background, we first tested whether overexpression was beneficial to wthA mutants. Overexpression of NlpD from an IPTG-inducible plasmid slightly rescued the growth defect of ΔwthA but did not correct cell chaining or morphological defects (Fig. 5 A and B and SI Appendix, Fig. S8A). These data are consistent with partial redundancy between NlpD and WthA as amidase activators.

Fig. 5.

WthA and NlpD are functionally redundant, but not equivalent. (A and B) Growth curves (A) and phase-contrast microscopy (B) of the 19606e ΔwthA strain with or without nlpD overexpression. LB was supplemented with 10 µM IPTG. (C) Biological triplicate growth curves in LB with or without 10 nM aTc inducer. A single backdilution (indicated by arrow) was used to facilitate depletion of the aTc-regulated wthA allele. (A and C) Error bars indicate SD and are not visible when smaller than the size of the symbol. (D) Phase-contrast microscopy of wthA depletion strains grown in LB without aTc. (B and D) Representative of three biological replicates. (E) Minimal inhibitory concentrations of select antibiotics for wthA depletion strains grown without aTc. Average and SD of biological triplicates are provided.

To test for possible synthetic genetic interactions, nlpD was deleted in a wthA depletion strain in which expression is controlled by an anhydro-tetracycline (aTc)-inducible promoter. We also combined ΔtolA with the wthA depletion allele to assess another pathway that impacts NlpD function. Depletion of wthA alone (no aTc) resulted in a growth defect and mild chaining, although less severe than the complete deletion (Fig. 5 C and D). In contrast, deleting nlpD or tolA in combination with wthA depletion resulted in severe exacerbation of chaining (Fig. 5 C and D and SI Appendix, Fig. S8B) and only a modest further decrease in growth.

Assessment of antibiotic sensitivity supported these interactions. The wthA depletion strain did not have as severe of sensitivities to rifampicin, aztreonam, and ceftazidime-avibactam as compared to the deletion mutant (Fig. 5E). However, combining ΔnlpD or ΔtolA with the wthA depletion allele increased antibiotic sensitivities to levels comparable to ΔwthA (Fig. 5E). Together, these phenotypes demonstrate that loss of NlpD or Tol-Pal is synthetic sick with reduced WthA function and support a model in which WthA acts as an amidase activator that is partially redundant with NlpD.

WthA and NlpD Stimulate a Toxic Amidase Activity When Coexpressed With Β-Lactamase, Oxa51.

A. baumannii lacks homologs of known amidases but retains the amidase activator NlpD, suggesting this organism may employ a different family of amidases. Recently, it was observed that overexpression of β-lactamases, including plasmid-encoded Oxa23 and chromosomally encoded AmpC, caused physiological changes in A. baumannii (49, 50). Oxa23 overexpression increased amidase products, PG strands denuded of peptide crosslinks (49), while A. baumannii AmpC displayed weak D-Ala-D-Ala carboxypeptidase activity (50). Consistent with a possible connection, AmpC was identified among WthA interaction partners (Fig. 4D). Oxa23 is not present in our strain, but all A. baumannii strains carry a chromosomal oxa51 allele. Oxa51 has poor carbapenemase activity compared with Oxa23 due to a sterically hindered β-lactam cleavage site (58). This conservation despite weak β-lactamase activity suggests that Oxa51 may have an alternative physiological function. Serine β-lactamases are distantly related to PG hydrolases, including endopeptidases and D-Ala-D-Ala carboxypeptidases (59), suggesting that they could act on PG peptides. In our hands, Oxa51 conferred no resistance to cephalosporins and only weak resistance to first-, third-, and fourth-generation penicillins compared with the well-characterized β-lactamase, TEM-1 (SI Appendix, Fig. S9). Therefore, we examined whether Oxa51 has amidase activity that could be enhanced by NlpD or WthA.

During plasmid construction, it was immediately evident that coexpression of NlpD or WthA with Oxa51 was toxic to E. coli cloning strains and expression was therefore repressed during strain maintenance. When introduced into wild-type E. coli (W3110) or A. baumannii, coexpression of NlpD/Oxa51 or WthA/Oxa51 was highly toxic (Fig. 6 A and B and SI Appendix, Fig. S10 A–D), whereas expression of the putative activators or Oxa51 alone showed no toxicity. These results suggested that NlpD and WthA may activate Oxa51 to trigger uncontrolled PG degradation. Oxa51 is expressed at low levels in A. baumannii, which may explain why it was not detected by mass spectrometry in the HA-WthA pull-down experiments. However, direct protein–protein interactions were obvious by reciprocal coimmunoprecipitations experiments. HA-tagged NlpD and WthA both copurified with His-tagged Oxa51, and vice versa (SI Appendix, Fig. S11).

Fig. 6.

A. baumannii WthA and NlpD stimulate a toxic amidase activity when coexpressed with the β-lactamase Oxa51. (A and B) Biological triplicate growth curves of E. coli expressing only Oxa51 or in combination with putative amidase activators, NlpD (A) or WthA (B). Concentrations of IPTG are indicated. (C) Diagram illustrating E. coli import systems that recycle PG turnover products. AmpG imports sugar-linked muropeptides generated by endopeptidases (EP) and lytic transglycosylases (LT). MppA/Opp and CadB import freed PG peptides released by amidases. (D) Amidase Release Assay: The AmpG-only E. coli strain (W3110 Δopp ΔcadB) specifically releases amidase-freed PG peptides into the supernatant while recycling other PG fragments. Cells were radiolabeled with [³H]-Dap, washed to remove excess label, and release of [³H]-amidase products was quantified following subtoxic expression of Oxa51 alone or in combination with NlpD or WthA (5 µM IPTG for all). (E) Amidase Uptake Assay: Spent media from the Amidase Release Assay was collected, filter-sterilized, and refreshed for carbon and nitrogen sources. Uptake of the released tritiated PG peptides was monitored using an Opp-only E. coli strain (W3110 ΔampG ΔamiD ΔampD, p-OppBCDF) or an Opp- control strain (W3110 ΔampG ΔamiD ΔampD ΔoppABCDF, p-empty). Counts per minute (CPM) of cell pellets reflect imported radiolabel. Error bars indicate the SD from three biological replicates and are not visible when smaller than the plotted symbols. Statistical significance was determined by the two-tailed t test with Welch’s correction: *P < 0.05, **P < 0.01, NS = not significant. Subtoxic induction levels were established in SI Appendix, Fig. S10 E and F; controls confirming absence of lysis during the Amidase Release Assays are shown in SI Appendix, Fig. S10 G–I.

To directly test whether Oxa51 cleaves PG peptides, we used E. coli PG recycling mutants (60, 61) previously generated by our group as reporters of amidase activity (Fig. 6 C–E). E. coli recycles nearly 100% of PG turnover products using specialized transporters. MppA/OppBCDF and CadB import freed peptides generated by amidases, whereas AmpG imports sugar-linked peptides produced by endopeptidases and lytic transglycosylases (60–64) (Fig. 6C). Therefore, a strain expressing only AmpG (Δopp ΔcadB) will recycle sugar-linked peptides but releases freed peptides into the growth media. These freed peptides serve as a direct proxy for PG amidase activity. To measure peptide release, cells were grown in the presence of tritiated meso-diaminopimelic acid [(³H)-Dap] to label PG peptides. After removal of excess radiolabel, the level of [³H]-Dap released to the growth supernatants was quantified as a measurement of amidase activity (Fig. 6D). Plasmids expressing Oxa51, NlpD, and WthA alone or in combination were induced at sublethal levels (SI Appendix, Fig. S10 E-I) to prevent any signal from lysed cells. Expression of Oxa51, NlpD, or WthA alone produced no significant increase in released radioactivity compared to the empty vector (Fig. 6D). However, coexpression of NlpD/Oxa51 or WthA/Oxa51 released ~20-30% of total cellular tritiated PG (Fig. 6D). Lysis was excluded by monitoring colony-forming units throughout the experiment (SI Appendix, Fig. S10G), and our Oxa51 coexpressions caused no detectable lysis. Although WthA/Oxa51 overexpression caused a slight growth defect, this was not due to lysis under the limited level of expression used during the assay as there was no increase in the release of DNA or ATP into growth supernatants (SI Appendix, Fig. S10 H and I).

Next, we verified that the released radiolabeled material consisted of freed PG peptides from amidase activity. To do this, we performed a peptide uptake assay leveraging the ability of E. coli Opp transport system to scavenge amidase products (60). Spent media from the amidase release assays was collected, filter-sterilized, and refreshed for carbon and nitrogen sources. These supernatants containing the released tritiated PG fragments were fed to an Opp-only strain (ΔampG ΔampD ΔamiD p-OppBCDF). This strain was blocked at several stages from being able to take in other PG fragments [see methods and ref. 60], assuring that only Opp uptake of amidase products could occur. The Opp-only strain efficiently imported the radioactive PG products released during NlpD/Oxa51 or WthA/Oxa51 expression (Fig. 6E) in an Opp-dependent manner, strongly supporting that the released tritiated products were amidase-derived PG peptides.

Together, these data demonstrate that NlpD and WthA function as amidase activators that stimulate an amidase-like activity associated with Oxa51. Whether Oxa51 itself acts as the catalytic amidase or as part of a larger complex remains to be determined. Deletion of oxa51 alone did not cause cell-division defects (SI Appendix, Fig. S9); however this is consistent with the redundancy typical of amidases in other bacteria making identification of such enzymes difficult. We are in the process of identifying other potential Acinetobacter amidases.

WthA/LbcA Homologs Are Conserved in Gram-Negative Bacteria.

All major cell division amidase activators described to date contain degenerate LytM domains, including EnvC (27, 65), DipM of Caulobacter crescentus (66–71), and NlpD/ActS (27, 29, 65) (Fig. 7A). WthA is a putative cell division amidase activator composed instead of TPR domains. Although S. aureus utilizes the TPR-containing protein ActH to activate an amidase (Fig. 7A), its cognate amidase cleaves uncrosslinked PG peptides to stimulate PG synthesis through a distinct mechanism that remains under investigation (47, 48). To explore whether the function of WthA might be conserved across Gram-negative bacteria, we examined the genomic distribution of wthA homologs. The STRING database (72) revealed that wthA had synteny with lolB that was widespread in γ- and β-Proteobacteria as well as Bacteroidota. We extended this analysis across related phylogenetic groups and identified likely wthA homologs in α-, β-, and γ-Proteobacteria, as well as Chlorobiota (Fig. 7B). Notably, several β-, and γ-Proteobacteria lacking EnvC also contained wthA homologs, suggesting that WthA may substitute for EnvC-type activators in these species (Fig. 7B).

Fig. 7.

WthA/LbcA homologs are widely conserved but differ in their mechanisms of regulating PG turnover enzymes. (A) Domain architectures of known and putative amidase activators. Degenerate LytM (dLytM) domains lacking catalytic residues are used by EnvC, NlpD, ActS, and DipM to activate cell-division amidases. In contrast, ActH and likely WthA employ tetratricopeptide repeat (TPR) domains for amidase activation. “SP” denotes periplasmic signal peptides; “Lipo” indicates lipoprotein signal peptides targeting outer-membrane (OM) anchoring. (B) Conservation of dLytM-domain activators and WthA/LbcA homologs across α-, β-, and γ-Proteobacteria and FCB (Fibrobacterota, Chlorobiota, and Bacteroidota) phyla. Predicted synteny with lolB-like genes is shown; these encode OM lipoproteins of unknown function that STRING analysis suggests are divergent lolB homologs. WthA/LbcA homologs predicted to be OM-anchored lipoproteins are marked as “OM Lipo,” whereas those predicted as soluble periplasmic proteins are labeled “Peri.” Species names and locus tags are listed in Dataset S1. (C and D) Growth curves of E. coli W3110 expressing the P. aeruginosa amidase, PaAmiB, or the β-lactamase, PaOxa10, alone or in combination with P. aeruginosa NlpD, PaNlpD (C) or PaLbcA (D). Growth curves are representative of biological triplicates and were performed in LB with 50 µM IPTG.

Bacteroides species are generally thought to lack lolB homologs. However, in Bacteroidota, we found wthA in synteny with OM lipoproteins genes of unknown function that STRING annotated as possible lolB homologs (Fig. 7B). These genes could encode divergent LolB proteins [20.6% identity to E. coli LolB by Clustal Omega alignment (73)]. Interestingly, in Bacteroidota, envC was located within the wthA operon (Fig. 7B), suggesting coordination of PG turnover regulators in this lineage.

Since A. baumannii wthA and P. aeruginosa lbcA mutants have distinct phenotypes, homologs of WthA/LbcA may differ in what types of PG turnover enzymes they modulate (SI Appendix, Fig. S7 B and C). We wondered whether P. aeruginosa LbcA could also act as an amidase activator, as this could have been missed due to its potential redundancy with EnvC and NlpD. We cloned codon-optimized alleles of P. aeruginosa nlpD and lbcA on to plasmids along with the known amidase gene, amiB, and intrinsic β-lactamase, oxa10, referred to as PaNlpD, PaLbcA, PaAmiB, and PaOxa10, respectively. Expression of each combination of activator/amidase was evaluated for toxicity in E. coli serving as a proxy for amidase activation (Fig. 7 C and D). As expected, expression of PaAmiB alone caused mild toxicity given that E. coli NlpD and EnvC can activate amidases of this family. In agreement with existing literature showing that PaNlpD functions with PaAmiB (57), coexpression of these two proteins was highly toxic and allowed for no growth (Fig. 7C and SI Appendix, Fig. S12). In contrast, PaNlpD coexpression with PaOxa10 showed no evidence of toxicity (Fig. 7C). Coexpression of PaLbcA with either PaAmiB or PaOxa10 also failed to enhance toxicity (Fig. 7D and SI Appendix, Fig. S12), indicating that LbcA does not stimulate any PG hydrolytic activities under these conditions. It remains possible that we did not test the correct amidase partner and that LbcA could activate P. aeruginosa AmiA or another β-lactamase. However, LbcA protein–protein interactions have been highly explored and identify AmiB as its only amidase-binding partner (52).

Since PaOxa10 had no evidence of toxicity, we wondered whether A. baumannii Oxa51 had evolved this new activity. Exploring this, we found that Oxa51 was toxic to E. coli when coexpressed with PaNlpD (SI Appendix, Fig. S12E), suggesting that Oxa51 can interact functionally with NlpD from multiple species. However, PaLbcA again did not enhance toxicity when coexpressed with Oxa51 (SI Appendix, Fig. S12F). These results highlight functional divergence between WthA and LbcA homologs, with WthA possibly retaining amidase-activator properties that are absent in LbcA.

Discussion

We have provided evidence that WthA plays a critical role in cell division in A. baumannii. Mutants lacking WthA grew poorly and exhibited severe morphological abnormalities, including chaining during active growth (SI Appendix, Fig. S5) and filamentation and branching as growth slowed (SI Appendix, Fig. S6). This explains why transposon insertions in wthA were underrepresented in our LOS-containing transposon mutant libraries (11). The connection between WthA and cell division emerged only because its function was already impaired in LOS-deficient cells, allowing us to obtain transposon insertions and a distinct Tn-seq signature. In support of WthA dysfunction in the absence of LOS, both LOS-deficient and wthA mutants displayed similar cell division defects and were difficult to culture after storage at 4 °C. Why WthA does not function in the absence of LOS remains unclear, but WthA may rely on an OM partner that fails to fold properly when LOS is absent.

Reduced function of WthA in LOS-deficient cells also provides a clear explanation for why NlpD, Tol-Pal, and YraP become essential for fitness under these conditions. Overexpression of NlpD partially restored growth of the wthA mutant, whereas loss of nlpD greatly intensified its division defects, consistent with the two proteins acting in overlapping pathways. These genetic relationships strongly suggest that WthA serves as an amidase activator during cell division. Notably, A. baumannii retains an essential SPOR-domain protein homologous to FtsN even though it lacks FtsEX and EnvC, indicating that amidase activation remains a key step in the septal synthesis feedback loop. WthA likely fulfills the role of EnvC in this system, linking PG cleavage to activation of the divisome and sustaining the self-reinforcing cycle of PG synthesis during cytokinesis. Although A. baumannii lacks recognizable homologs of the canonical periplasmic amidases, the conservation of three SPOR-domain proteins—one of which, HMPREF0010_03200 (FtsN), is essential—supports that this bacterium still depends on amidase-mediated PG remodeling.

Recently, overexpression of an Acinetobacter class D β-lactamase, Oxa23, was shown to increase denuded PG, raising the intriguing possibility that certain β-lactamases have been co-opted to perform amidase-like functions (49). We explored this hypothesis further and examined whether the broadly conserved and chromosomally encoded class D β-lactamase Oxa51 could be one of the missing Acinetobacter amidases. Oxa51 possesses a sterically hindered β-lactamase active site that reduces its carbapenemase activity (58). Indeed, we found Oxa51 to be a weak penicillinase compared to other β-lactamases (SI Appendix, Fig. S9A) suggesting functional divergence away from β-lactamases. This observation fits with the evolutionary link between class D β-lactamases and high molecular weight penicillin binding proteins (PBPs), the PG synthases from which they likely evolved (59). Notably, a chromosomally encoded class C β-lactamase, Acinetobacter AmpC, was recently found to possess weak D-Ala-D-Ala carboxypeptidase activity (50), supporting the broader concept that repurposed or promiscuous β-lactamases may contribute to PG metabolism in Acinetobacter. Oxa51 is not essential, and its deletion did not affect growth or cellular morphology (SI Appendix, Fig. S9), indicating that if it is an amidase, its function is likely redundant.

Our first indication that Oxa51 could act as a PG amidase was that its expression was toxic, but only when coexpressed with NlpD or WthA. Uncontrolled activation of PG amidases is known to be highly lethal to bacteria (27), consistent with Oxa51 contributing to PG hydrolysis. To evaluate this activity directly, we leveraged E. coli PG recycling mutants that selectively release or import PG amidase products (60, 61). Using these strains, we designed complementary assays to measure both amidase product release and uptake. Coexpression of WthA/Oxa51 or NlpD/Oxa51 both resulted in release of PG amidase products, supporting the most likely hypothesis that WthA/Oxa51 and NlpD/Oxa51 constitute amidase and activator pairs. However, since our assays were performed in whole cells, we cannot exclude the possibility that the increased amidase activity was indirect. Further characterization will be needed to identify additional A. baumannii amidases and establish the mechanism of catalysis through biochemical characterization. Interestingly, the amidase activity observed when Oxa23 was overexpressed did not require the catalytic serine of the β-lactamase active site (49). This raises the possibility that class D β-lactamases, like Oxa51, employ distinct residues or alternative catalytic centers to mediate PG cleavage, representing a repurposing of their ancestral penicillin-binding protein fold.

WthA’s protein interaction profile suggested that it functions as a hub protein similar to its Pseudomonas homolog LbcA. Our data show that in A. baumannii, WthA i) directs CtpA-dependent proteolytic regulation of the endopeptidase MepM, ii) interacts with the lytic transglycosylases RlpA and MltD, iii) influences cell division, potentially as an amidase activator, and iv) associates with additional PG biogenesis proteins (Fig. 8). The first two functions are conserved between WthA and LbcA (51, 52). However, whether and how LbcA impacts lytic transglycosylases has remained elusive in Pseudomonas, since these enzymes are not proteolytically regulated.

Fig. 8.

WthA functions as a central hub coordinating PG turnover in A. baumannii. Protein interactions and experimental evidence presented here suggest that WthA impacts multiple classes of PG turnover enzymes. (i) Endopeptidase regulation: WthA promotes CtpA-dependent proteolysis of the endopeptidase MepM. (ii) Lytic transglycosylase control: Suppressor analysis of ΔwthA disrupted MltD to rescue growth, suggesting WthA negatively regulates its activity to maintain PG homeostasis. The same is true for MepM supporting endopeptidase regulation. (iii) Amidase activation: Loss of WthA results in cell division defects that are exacerbated with alterations to NlpD. Furthermore, like NlpD, coexpression of WthA with Oxa51 was toxic and released amidase products, consistent with the protein functioning as an amidase activator. Oxa51 may represent one of the missing amidases in A. baumannii. Immunoprecipitation studies showed that WthA also interacts with HMPREF0010_02362, an RlpA-like lytic transglycosylase, which may act downstream of amidases to remove denuded PG strands. (iv) Coordination with PG synthesis. WthA’s strongest detected protein interaction was with the undecaprenyl phosphate recycling protein, UppP. Suggesting that WthA interaction network is even larger coordinating PG synthesis, turnover, and lipid carrier recycling.

Suppressor analyses in A. baumannii provided further insight into the WthA interaction network. Mutations in mepM or mltD rescued growth defects of the wthA mutant, suggesting that both enzymes become hyperactive in the absence of WthA. Increased MepM protein levels explain its elevated activity, whereas the mechanism controlling MltD may be either through direct negative regulation by WthA or indirect via coupling to MepM activity. These observations support a model in which WthA restrains multiple PG hydrolases to balance synthesis and degradation. This coupling is reminiscent of the classic observations that autolysis occurs when hydrolysis proceeds faster than synthesis, a principle that underlies β-lactam-induced lysis (28, 53–56).

Importantly, WthA’s impact on cell division appears distinct from LbcA-like activities. P. aeruginosa mutants lacking lbcA show no morphological abnormalities (51, 52), and in A. baumannii, deletion of mepM or mltD only partially restore growth and did not eliminate chaining or branching in ΔwthA cells. This separation of phenotypes indicates that WthA carries out additional functions—specifically, a role in cell division that is beyond its role in controlling PG hydrolyases (MepM, MltD). Therefore, WthA may combine two levels of PG coordination that in other organisms are divided among multiple proteins reinforcing the idea that WthA is a multifunctional hub.

Homologs of WthA/LbcA are widely distributed across a α-, β-, and γ-Proteobacteria, as well as Bacteroidota and Chlorobiota (Fig. 7B), suggesting that these proteins represent a broadly conserved strategy for coordinating PG turnover. Notably, several of these homologs have been implicated in maintaining cell envelope integrity in Brucella melitensis (74), Brucella abortus (75), and Rhizobium leguminosarum (76), suggesting that WthA/LbcA-like proteins may perform related functions in other genera. However, it is unlikely that all homologs participate directly in cell division. In support of this notion, we found no evidence that Pseudomonas LbcA could influence the activity of its native amidase AmiB, the P. aeruginosa β-lactamase Oxa10, or the putative A. baumannii amidase, Oxa51 (Fig. 7D). These findings underscore that WthA and LbcA differ in the range of PG turnover enzymes they regulate, reflecting functional diversification within this protein family (see SI Appendix, Fig. S7 B and C). How conserved each of these functions are across Gram-negative WthA/LbcA homologs will be interesting to explore.

Hub proteins that coordinate multiple PG turnover enzymes have been described in other bacteria. Among these, the roles proposed for Acinetobacter WthA most closely resemble those of Caulobacter DipM. DipM has been shown to function as a hub protein (69, 71), activate a cell division amidase (69–71), and interact with lytic transglycosylases (71). Similarly, Pseudomonas LbcA shows striking parallels to Escherichia NlpI as both employ TPR domains to direct proteolysis of endopeptidases through tail-specific proteases, CtpA in Pseudomonas (52, 77, 78) and Prc in Escherichia (79, 80). Together, these comparisons highlight a recurring theme that different species encode for multienzyme complexes that coordinate PG synthesis and hydrolysis to maintain envelope integrity.

It is also curious that LbcA and WthA differ in cellular localization, with LbcA anchored to the OM and WthA residing in the periplasm. All characterized amidase activators known to play major roles in cell division (EnvC, DipM, WthA) are soluble periplasmic proteins, whereas OM anchored activators (NlpD, ActS) exert more limited effects. This distinction may reflect the timing of key steps during cytokinesis in that periplasmic activators access nascent septal PG earlier, while OM-anchored factors act later as the OM invaginates. Consistent with this idea, all organisms possessing WthA/LbcA homologs but lacking EnvC encode soluble periplasmic variants (Fig. 7B). Altogether, our findings define WthA as a periplasmic hub protein that coordinates multiple PG turnover enzymes to promote balanced envelope remodeling during growth and division in A. baumannii.

Materials and Methods

Strains and Growth Conditions.

All strains and plasmids used in this study are listed in Dataset S1. Strain construction, assay descriptions, and other general methods are further described in the Supplementary text. All cultures and agar plates were performed with Difco™ Luria-Bertani Broth or Agar, Miller (BD). Where appropriate media were supplemented with glycerol (10% volume/volume), kanamycin (30 μg/mL), tetracycline (5 to 10 μg/mL), hygromycin B (250 μg/mL for A. baumannii and 100 μg/mL for E. coli), carbenicillin (100 μg/mL), and anhydro-tetracycline (10 nM, aTc). Concentrations of isopropyl β-D-1-thiogalactopyranoside (IPTG) are indicated in the experiment.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Data, Materials, and Software Availability

All study data are included in this article and/or supporting information. Primary data were deposited at https://zenodo.org/records/17602936 (81).