Bacteroidetocins Target the Essential Outer Membrane Protein BamA of Bacteroidales Symbionts and Pathogens

ABSTRACT Bacteroidetocins are a family of antibacterial peptide toxins that are produced by and target members of the phylum Bacteroidetes. To date, 19 bacteroidetocins have been identified, and four have been tested and shown to kill diverse Bacteroidales species (M. J. Coyne, N. Béchon, L. M. Matano, V. L. McEneany, et al., Nat Commun 10:3460, 2019, https://doi.org/10.1038/s41467-019-11494-1). Here, we identify the target and likely mechanism of action of the bacteroidetocins. We selected seven spontaneous mutants of four different genera, all resistant to bacteroidetocin A (Bd-A) and found that all contained mutations in a single gene, bamA. Construction of three of these bamA mutants in the wild-type (WT) strains confirmed they confer resistance to Bd-A as well as to other bacteroidetocins. We identified an aspartate residue of BamA at the beginning of exterior loop 3 (eL3) that, when altered, renders strains resistant to Bd-A. Analysis of a panel of diverse Bacteroidales strains showed a correlation between the presence of this aspartate residue and Bd-A sensitivity. Fluorescence microscopy and transmission electron microscopy (TEM) analysis of Bd-A-treated cells showed cellular morphological changes consistent with a BamA defect. Transcriptomic analysis of Bd-A-treated cells revealed gene expression changes indicative of cell envelope stress. Studies in mice revealed that bacteroidetocin-resistant mutants are outcompeted by their WT strain in vivo. Analyses of longitudinal human gut isolates showed that bamA mutations leading to bacteroidetocin resistance do not become fixed in the human gut, even in bacteroidetocin-producing strains and nonproducing coresident strains. Together, these data lend further support to the applicability of the bacteroidetocins as therapeutic peptides in the treatment of maladies involving Bacteroidales species.

antimicrobial proteins [BSAPs]) (2)(3)(4)(5), the ubiquitin-like toxin Bacteroides fragilis Ubb (BfUbb) (6), and the bacteroidetocins (7). MACPF domain toxins are produced by diverse members of the phylum Bacteroidetes and target the glycan of lipopolysaccharides (LPS) or outer membrane proteins (OMPs) and likely form large pores in the outer membrane (OM). MACPF toxins described to date are specific for intraspecies killing. The BfUbb toxin is produced by a subset of Bacteroides fragilis strains, is 72 amino acids in its mature form, and is 82% similar to human ubiquitin. BfUbb specifically targets a subset of B. fragilis strains.
The bacteroidetocins are ribosomally produced peptide toxins with similarities to class IIa bacteriocins produced by Gram-positive bacteria (7). These molecules are synthesized with a leader sequence that is cleaved at a GG site, yielding mature peptides of 42 to 53 residues. To date, we have identified 19 distinct bacteroidetocins that are produced by diverse members of the Bacteroidetes phylum, four of which have been tested and shown to have toxin activity. Unlike the MACPF toxins and BfUbb, the bacteroidetocins have a broader targeting range. Bacteroidetocins A, B, and D (Bd-A, Bd-B, and Bd-D, respectively) kill Bacteroides, Parabacteroides, and Prevotella species (7), which are members of three distinct Bacteroidales families. Although these bacteroidetocins target similar strains, their killing profiles are not identical. In addition, not every Bacteroidales species is targeted by these bacteroidetocins, and phylogeny alone does not predict which will be sensitive or resistant. Bd-C is less similar to Bd-A, -B, and -D and targets a different set of Bacteroidales strains/species (7).
Class IIa bacteriocins target the cytoplasmic membrane components of the mannose phosphotransferase system (PTS) of Gram-positive bacteria, leading to permeabilization of the membrane and loss of cytoplasmic ATP (8)(9)(10)(11). Bacteroidales lack this PTS, and, unlike the class IIa bacteriocins, bacteroidetocins target Gram-negative Bacteroidetes. This study was designed to identify the target of the bacteroidetocins to better understand the variable susceptibility of Bacteroidales species to these toxins. In addition, as the bacteroidetocins target not only symbiotic gut species but also pathogenic oral and vaginal Prevotella species, a better understanding of how these toxic peptides kill cells is important to evaluate their utility in the treatment of maladies such as periodontal disease and bacterial vaginosis involving members of this order.

RESULTS
Selection of Bd-A resistant mutants. The receptors of the MACPF toxins are important fitness determinants for gut colonization but are not essential molecules in vitro (3). Therefore, it was possible to identify these receptors using transposon mutagenesis. Attempts to identify Bacteroides mutants resistant to Bd-A by transposon mutagenesis were not successful despite repeated attempts, suggesting that the target of Bd-A may be essential. We tested if we could isolate spontaneous mutants resistant to Bd-A by selecting for survivors that continue to grow after repeated passage in the presence of the peptide. For these assays, we used five different wild-type (WT) Bacteroidales strains from three different families: Parabacteroides johnsonii CL02T12C29, Bacteroides fragilis 638R, Bacteroides vulgatus ATCC 8482 (also known as Phocaeicola vulgatus), Prevotella bivia DNF00188, and P. bivia ATCC 29303. These strains were passaged sequentially in increasing concentrations of Bd-A (Fig. 1A), and those that grew in the presence of the toxin were single-cell cloned and tested individually in overlay assays. With the exception of P. johnsonii CL02T12C29, for which the WT strain is not targeted by Bd-D, most strains that grew in the presence of 10 ng/ml Bd-A were resistant to both Bd-A and Bd-D (Fig. 1B). P. bivia ATCC 29303 mutant (mut 1) was somewhat resistant to Bd-A ( Fig. 1B and D), enough to identify the mutant during selection, but was completely resistant to killing by Bd-D (Fig. 1B). Synthetic Bd-B peptide is insoluble in water at neutral pH; therefore, we tested the resistance of four of these mutant strains to Bd-A and Bd-B produced from native Bd-B-producing Bacteroides strains. We found that these mutants were also resistant to these natively produced toxins (Fig. 1C), showing that most Bd-A-resistant mutants are also resistant to Bd-B and Bd-D. Bd-A-resistant mutants were not growth defective in vitro and grew well in the presence of 10 ng/ml of Bd-A, a concentration that completely inhibited the growth of the WT strains (Fig. 1D).
Identification of mutations leading to Bd-A resistance. To determine what mutations occurred to confer resistance to Bd-A, whole-genome sequencing (WGS) was performed on the five WT strains and the seven resistant mutants. Comparison of WT and mutant genomes revealed that each mutant has an alteration in bamA (see Table S1 in the supplemental material), encoding an essential OMP protein of Gram-negative bacteria. BamA is part of the BAM complex responsible for shuttling, assembling, and inserting b-barrel proteins into the OM (reviewed in reference 12). The BAM complex consists of BamA and four other lipoproteins, BamBCDE, of which only BamA and BamD are essential. Conserved cysteine residues are highlighted in green, and identical residues are highlighted in yellow. (B) Agar spot overlay assays showing seven bacteroidetocin-resistant mutants selected from five different Bacteroidales strains. Mutants selected for resistance to Bd-A were cross resistant to Bd-D (with the exception of P. johnsonii CL02T12C29, which is not sensitive to Bd-D). (C) Overlays of Parabacteroides johnsonii CL02T12C29, Bacteroides fragilis 638R, Bacteroides vulgatus ATCC 8482, and bamA mutants. Spots are Bd-A-producing Bacteroides vulgatus CL01T12C17 (BvCL01) and Bd-B-producing Bacteroides thetaiotaomicron CL15T12C11 (BtCL15). BtCL15 DBd-B lacks the Bd-B-encoding gene and is shown as a control. (D) Growth curves of WT strains and resistant mutants with or without the addition of 10 ng/ml Bd-A. Data were recorded every 10 min and plotted as the averages from at least three biological replicates 6 SDs.
The mutations identified in the resistant mutants mapped to four residues predicted to be part of extracellular loops 2, 3, or 4 (eL2, eL3, or eL4, respectively) ( Fig. 2A and B; see also Fig. S1a and b). These three residues of BamA are each conserved in all four species used for mutation (Fig. S1b). In four of the seven mutants, the aspartate at the beginning of eL3 was altered. Aspartate is negatively charged at physiological pH, and these substitutions removed that charge by several different mutations: a threebase-pair codon deletion (D546D) resulting in loss of the aspartate, a mutation resulting in its replacement with the polar but uncharged residue asparagine (D547N), or a mutation changing it to the nonpolar residue tyrosine (D545Y and D547Y) ( Fig. 2A and B). The structure of B. vulgatus BamA was modeled with Phyre2 (13) using the crystal structure of the BamA of Neisseria gonorrhoeae (14) (PDB accession 4K3B) (see Fig. S2). In this structural prediction, the residues identified as conferring resistance to Bd-A appear spatially aligned (Fig. S2), potentially allowing for interaction of Bd-A with all three residues. To confirm that these mutations confer resistance, three of these mutations were created in the WT strains B. fragilis 638R (D545Y) and B. vulgatus ATCC 8482 (D546D and W624R). As expected, these constructed mutants conferred resistance to the synthesized Bd-A and Bd-D peptides as well as to the native Bd-A and Bd-B toxins produced by BvCL01T12C17 and BtCL15T12C11 (Fig. 2C). These mutants also grew well in vitro and were resistant to the bacteroidetocins in liquid culture, similarly to the spontaneous mutants (Fig. 2D).
Importance of the aspartate of eL3 for Bd-A sensitivity. The mutations affecting the aspartate residue at the beginning of eL3 suggested a charge interaction that may be important for the activity of the bacteroidetocins. To determine the importance of the aspartate residue, we aligned the BamA sequences of all Bacteroidales species previously tested for sensitivity to Bd-A (7) and found that all Bd-A-sensitive strains have a BamA with the aspartate residue at the N-terminal end of eL3, whereas the two resistant strains, Prevotella nigrescens F0103 and Porphyromonas gingivalis W83, do not (Fig. 3A). To further correlate this aspartate residue with sensitivity or resistance to Bd-A, we analyzed the BamA sequences of 17 additional Bacteroidales strains in our collection or available from public repositories. The new panel of strains included members of the genera Prevotella, Bacteroides (several species also known as Phocaeicola), Paraprevotella, Dysgonomonas, and Alistipes (Fig. 3B). Twelve of these 17 strains contain the aspartate residue at the beginning of eL3 and five do not. We tested each of these 17 strains for Bd-A sensitivity/resistance and found that all five of the strains that lack the aspartate residue at this site are resistant to Bd-A (Fig. 3C). Eleven of the 12 strains that contained eL3D are sensitive to Bd-A, with Prevotella buccae D17 being the only exception. BamA sequences vary between members of the same genera and even members of the same species. Of note are two Prevotella oris strains whose BamA sequences are extremely similar (99.2%), with only seven different residues along the entire 877-residue proteins (see Fig. S3). Four of these differences occur in the region of eL3D, where BamA of P. oris F0302 contains a glycine instead of an aspartate ( Fig. 3B; Fig. S3). We tested these two strains and found that P. oris F0302 is resistant to Bd-A, while P. oris DSM 18711 is sensitive (Fig. 3C). These data confirm the importance of eL3D to the activity of Bd-A.
Microscopic analyses of Bd-A-treated strains. To determine if Bd-A binds the surfaces of sensitive cells but not those of resistant mutants, Bd-A was synthesized with a lysine-carboxyfluorescein (FAM) modification at the C terminus (Bd-A-FAM) (Fig. 4A). We first confirmed that the fluorescently labeled peptide is still toxic to WT but not to mutant cells. We chose to analyze WT B. vulgatus ATCC 8482 and both of the BamA mutants of this strain: D546D and W624R. We found that the CFU count of the WT strain treated with 10 ng/ml Bd-A-FAM decreased within 3 h, with an approximately 3log decrease by 5 h (Fig. 4B). In contrast, the CFU count of the two bamA mutants treated with Bd-A-FAM increased over time equivalent to the untreated control  Fig. S4 and S5). By 5 h, the WT cells displayed significant cell rounding and large membrane protrusions, whereas the mutant strains remained unchanged. WT B. vulgatus ATCC 8482 cells treated with Bd-A for 3 h were fixed, embedded, and sectioned for transmission electron microscopy (TEM) imaging. Cross-sectional analyses showed severe OM defects, including ruptures and ballooning of the OM, which were not evident in the untreated control ( Fig. 4G; Fig. S5). These phenotypes have been described in several bacteria when BamA activity is inhibited either with a toxin (15), with BamA inhibitors (16), or when knocked down with CRISPR interference (CRISPRi) (17). These data further support that Bd-A interacts with BamA and likely inhibits its function.
Exposure of sensitive cells to Bd-A results in transcriptional changes indicative of cell envelope stress. To provide further evidence that Bd-A interferes with the function of BamA, we performed transcriptome sequencing (RNA-Seq) analysis to determine if Bd-A treatment induces cell envelope stress. B. vulgatus ATCC 8482 was grown in liquid medium with or without a sublethal concentration of Bd-A (2 mg/ml) that still permitted growth albeit at a reduced rate. Triplicate cultures were harvested at an optical density at 600 nm (OD 600 ) of 0.8 and processed for RNA-Seq analysis. A gene was considered differentially expressed (DEG) if the absolute value of the fold change of the expression level in the Bd-A-treated bacteria versus that in untreated bacteria was greater than or equal to 2 and the adjusted P value (P adj ) for DESeq2 and the false-discovery rate (FDR) for edgeR were less than or equal to 0.05, as calculated by both statistical packages. Of the approximately 4,000 protein-encoding genes in the B. vulgatus ATCC 8482 genome, exposure to Bd-A resulted in the upregulation of  Table S1 in the supplemental material. 421 genes and the downregulation of 433 genes (see Table S2). Half of the 50 most downregulated genes are either susCD genes encoding a TonB-dependent OMP and an outer surface nutrient-binding protein involved in nutrient transport across the OM or other TonB-dependent OM receptors/transporters (Table S2; see also Fig. S6). These data suggest that the effect of intoxication with Bd-A is the downregulation of the energy-consuming processes of nutrient import. Although these TonB-dependent OM transporters were downregulated, the two TonB orthologs themselves were upregulated, one of which was the most significantly upregulated gene (Table S2; Fig. S6). Therefore, although the genes encoding numerous TonB-dependent transporters were downregulated, the organism may be redirecting TonB energetics to other OMPs, with two genes encoding porin proteins of unknown function among the top eight most highly upregulated genes. The list of upregulated genes reveals an organism Bacteroidetocins Target BamA ® that is dealing with substantial stress. Many of the genes that fall into Clusters of Orthologous Groups (COG) V "defense mechanisms" were upregulated when cells were treated with Bd-A (Table S2; Fig. S6). Four RND efflux protein systems, including all three genes of these tripartite complexes (permease, periplasmic adapter, and OM efflux protein) were upregulated. Two heat shock protein 20 (Hsp20) genes were highly upregulated as well as genes encoding the chaperones DnaK and GroEL. In addition, two periplasmic proteases that have distinct roles in maintaining OMP integrity (18) were highly upregulated. The first, BVU_0371, is a zinc metalloprotease with predicted structural similarity to BepA (PDB accession 6SAR) (19). BepA was shown to enhance the biogenesis of the essential b-barrel OMP LptD (20), which is involved in the assembly of LPS, and to also degrade misfolded LptD (20). The second is BVU_2737, which has predicted structural similarity to DegQ (PDB accession 6Z05) and also DegP. DegP and DegQ are involved in the removal of misfolded proteins in the periplasm. Therefore, two proteases involved in the removal of misfolded b-barrel OMPs are highly upregulated. Many of the DEGs in Bd-A-treated cells are similar to those identified in other organisms when BamA function is disrupted (see Discussion).
In vivo analyses of bamA mutant fitness and analysis of strains from the human gut. As bacteroidetocins have potential as therapeutics or in modulation of the gut microbiota, the evolution of resistance to these peptides would reduce their usefulness for such purposes. We did not detect in vitro growth defects in any of the bamA mutants, with all growing at least as well as the WT strains. However, in vitro growth does not indicate if the mutant strains will be fit in the mammalian gut. We tested the fitness of bamA mutants in vivo in competition with the WT strain using a Swiss-Webster gnotobiotic competitive colonization mouse model. A mixed inoculum of WT and mutant strains was introduced into germ-free mice by gavage. After 1 week, feces were collected and the proportions of each strain were quantified. This experiment was performed using B. fragilis 638R WT/BamA D545Y and the B. vulgatus ATCC 8482 WT/BamA W624R pairs ( Fig. 5A and  B). In both cases, the bamA mutant strains dropped to ,1% abundance by 1 week, suggesting a severe fitness attenuation in vivo. It is therefore likely that if these mutations did arise in vivo, they would not reach fixation.
Next, we analyzed several isolates from the human gut to further determine if naturally evolved bacteroidetocin-resistant bamA mutants are present in toxin-producing strains. The bamA sequences of the two Bd-A-producing strains (B. vulgatus CL01T12C17 and B. vulgatus CL14T03C19) and two Bd-B-producing strains (Bacteroides ovatus CL02T12C04 and Bacteroides thetaiotaomicron CL15T12C11) do not have mutations and are sensitive to their own bacteroidetocins (7). We did not know, however, how stable bacteroidetocin-producing strains are in the human gut, if they would be lost over time, or if mutations in bamA would eventually arise allowing for longer-term colonization. We were able to collect a fecal sample from the original CL15 donor approximately 9 years (month 119 [T119]) after collection of an original sample (month 12 of the study [T12]), from which the original Bd-B-producing strain B. thetaiotaomicron CL15T12C11 was isolated. This T119 fecal sample was diluted and selectively plated for growth of Bacteroidales strains, and 63 individual colonies were screened with primers specific to the Bd-B biosynthesis region (Table S1b). Two strains were selected that produced a PCR product, and the 16S rRNA genes of these two strains were sequenced to confirm that they are B. thetaiotaomicron strains, as was the initial isolate 9 years earlier. These new strains were designated BtCL15T119C47 and BtCL15T119C52 (Bt for B. thetaiotaomicron, CL15 indicates they are from human subject 15, T119 is the month of isolation, and the number following the C is the colony or isolate number). Western immunoblot analysis showed that these later strains produced Bd-B (Fig. 5C) and that the toxin was active (Fig. 5D). Lastly, the genomes from both the original BtCL15T12C11 strain and the BtCL15T119C52 strain isolated 9 years later from the same human gut ecosystem were sequenced using SMRT (Pacific Biosciences) sequencing, which confirmed that these strains are near isogenic, with no differences in the Bd-B biosynthesis region. Importantly, even though the T119 strain still produces Bd-B, this strain did not evolve mutations in bamA and still self-intoxicates (Fig. 5E). From the T119 fecal sample, we also isolated several Bacteroides strains that were species matched with strains collected 9 years earlier (Table S1a). We used arbitrarily primed PCR (AP-PCR) to assess if these early and late species-matched stains are likely near isogenic. We identified two such pairs of strains (B. ovatus and Bacteroides uniformis) that produced similar PCR amplicons in all three AP-PCRs, using both the early and late strains as the template, that were different to those of nonisogenic type strains (Fig. S7). Sequence analysis of the bamA of these four strains revealed no differences in this gene between the early and late strains, and analyses in agar overlays showed continued sensitivity to Bd-B (Fig. 5E), despite coexisting with the Bd-B-positive B. thetaiotaomicron strain for nearly 9 years. To further evaluate if any Bacteroidales strains harboring Bd-A, Bd-B, or Bd-D have an alteration in the BamA sites shown in this study to confer resistance, we searched our Bacteroidales isolate collection of 1,434 sequenced genomes (21). We found 19 strains encoding either a Bd-A, Bd-B, or Bd-D peptide, and all BamA proteins of these strains contain the conserved aspartate, tryptophan, and asparagine residues shown in various Bacteroidales species to confer resistance when altered. The glycine residue of P. bivia that conferred resistance to Bd-D when changed to aspartate was either a glycine or the similar residue alanine in these 19 strains, and either of these two residues is normally present in the BamA of nonbacteroidetocin-harboring WT strains. These experiments suggest that organisms that evolve bamA mutations leading to resistance to these bacteroidetocins are not fit in the mammalian gut, further supporting the utility of the bacteroidetocins as therapeutic antibacterial molecules.

DISCUSSION
BamA is a b-barrel OMP containing periplasmic polypeptide translocation-associated (POTRA) domains responsible for binding chaperoned proteins in the periplasm and a b-barrel domain consisting of 16 b-strands that are involved in the final folding and inserting of OMPs into the OM (reviewed reference 12). The BamA proteins of Escherichia coli and Neisseria gonorrhoeae each contain seven extracellular loops (eLs), of which loop 6 (eL6) is highly conserved and contains the (V/I)RG(F/Y) motif which interacts with other residues in the b-strands of the barrel wall (14). The surface-exposed eL6 of E. coli BamA is necessary for its activity as it undergoes a conformational change during the OMP assembly reaction (22). The interaction of residues of eL6 with the interior of the nascent b-barrel wall is an important step in the folding of OMPs (23). Recent studies have shown that binding of diverse molecules to the eLs of BamA inhibits its function, leading to cell death. One example is the lectin-like bacteriocins (LlpAs) produced by Pseudomonas that initially interact with the common polysaccharide antigen of the LPS of similar species (24). This interaction is predicted to stabilize or facilitate its interaction with eL6 of BamA, inhibiting its function (25). Another example is the ribosomally synthesized modified small peptide darobactin produced by Photorhabdus khanii, which inhibits BamA activity, resulting in cell death (15). In contrast to Bd-A and LlpA which bind the surface of BamA, darobactin mimics the recognition signal of native substrates and binds the lateral gate of BamA of E. coli (26) near the periplasmic interface. Another study showed that the small molecule compound MRL-494 functions at the outer surface of E. coli by binding to BamA and inhibiting OMP biogenesis (16). In addition, a monoclonal antibody that binds eL4 of the BamA of E. coli inhibits OMP folding activity and is bactericidal (27).
The present study strongly suggests that Bd-A not only binds BamA but also inhibits its function. The only mutations identified that render Bacteroidales strains resistant to Bd-A killing were in bamA. We also showed that fluorescently labeled Bd-A binds to the surfaces of WT cells but not to those of isogenic strains with bamA mutations. In addition, several Bd-A-mediated effects in treated cells were similar to those described in other Gram-negative bacteria treated with molecules that inhibit BamA function. For example, the kinetics of killing by Bd-A is similar to that by darobactin, which was shown to lead to a two-log decrease in CFU by 2 h (15). Darobactin-mediated cellular morphological changes included membrane blebbing, cell rounding, and subsequent bursting of cell membranes, as we show here in Bd-A-treated cells. Another study using CRISPRi knockdown of bamA of Pseudomonas aeruginosa also showed cellular rounding and eventual cell lysis (17). Lastly, the transcriptomics data of Bd-A-treated cells demonstrate a response indicative of outer membrane stress. In Proteobacteria, inhibition of BamA function led to transcriptional responses where the s E stress response regulon was upregulated, including upregulation of degP (15) encoding the periplasmic protease involved in degrading misfolded proteins. Similarly, quantitative proteomic analysis revealed increased levels of DegP in E. coli treated with MRL-494 (16). Here, treatment of B. vulgatus with Bd-A resulted in the increase of transcripts encoding the two predicted periplasmic proteases, DegQ/P and BepA, involved in removing misfolded OMPs.
The BamA proteins of E. coli and B. vulgatus are not highly similar in their primary sequences, especially in eL3. The eL3 of B. vulgatus is 44 residues long, compared to that E. coli, which is 12 residues. Of these 44 residues of the B. vulgatus BamA, 22 are tyrosines and asparagines, residues rarely seen in low-complexity regions of prokaryotic proteins (28). Four of the seven bamA mutations that lead to Bd-A resistance in different genera of Bacteroidales altered the same aspartate residue predicted to be at the interface of the transmembrane domain and eL3. The other three bamA mutations that conferred Bd-A resistance all altered residues at the predicted transmembrane interface with eL2 or eL4. Mutants resistant to the LlpA bacteriocin also mapped to three different eLs of BamA (24), suggesting a three-dimensional interaction motif. Of the 38 Bacteroidales strains that have been tested with Bd-A, no strains that lack the aspartate residue of eL3 are susceptible to Bd-A, and all but one strain that contains the aspartate residue at this site are sensitive to Bd-A. It is our prediction, based on the fact that different bacteroidetocins target different strains and species, that different residues at the transmembrane interfaces of these eLs will dictate specificity. Once the molecular and structural interactions between the bacteroidetocins and BamA are elucidated, it should be possible to rationally design peptides to target the BamA of specific Bacteroidales pathogens, such as Porphyromonas gingivalis, for which we have not yet identified a toxic bacteroidetocin.
Competitive colonization experiments demonstrate that Bd-A escape mutants selected in our in vitro assay are not fit to compete in the mammalian gut. Even more relevant are the findings that Bd-producing Bacteroidales strains in the human gut do not contain bamA mutations and, in the one ecosystem we analyzed, were stable members of the human gut microbiota. We previously showed that Bd-B peptide is present in the stool of mice monocolonized with BtCL15T12C11 (7), showing that this strain produces Bd-B in the mammalian gut. These data raise the question of how a strain that has the ability to self-intoxicate would be fit for gut colonization. It is possible that the organism may upregulate the expression of protective genes in vivo that are not expressed in vitro. It is also possible that the toxin is only produced by a small proportion of cells. In addition, the bacteroidetocins may be produced only at high cell density, which may explain how the BtCL15 strains can grow as a lawn when inoculated at low density, such as in top agar of overlay assays. Future analysis of the regulation of toxin production and subsequent secretion from the cell will likely reveal important ecological properties of bacteroidetocin production and whether the bacteroidetocins confer an as-yet-unrecognized benefit to the producing cell.
Unlike previously discovered natural and synthetic inhibitors of BamA, the bacteroidetocins uniquely target species of Bacteroidetes. The data support that the bacteroidetocins are bactericidal and that resistant mutants are not fit in vivo, highlighting their therapeutic potential against pathogenic Bacteroidales species, in particular, pathogenic Prevotella species. Here, we show that Bd-A targets all four species of pathogenic vaginal Prevotella implicated in bacterial vaginosis, P. bivia, P. amnii, P. timonensis, and P. disiens. The high degree of treatment failure and subsequent recurrence of B. vulgatus necessitate new treatment strategies. The data support that bacteroidetocins could be important as a next-generation targeted treatment.

MATERIALS AND METHODS
Bacterial strains and growth. All strains are listed in Table S1a in the supplemental material. Bacteroides and Parabacteroides spp. were grown on supplemented brain heart infusion (BHIS) plates or were cultured in basal liquid medium (7). All other strains were grown on Brucella blood agar plates containing hemin and vitamin K (Remel R01255) and in liquid tryptic soy broth (TSB) containing per liter 30 g TSB, 5 g yeast extract, and 0.5 g cysteine-HCl, with the addition of hemin and vitamin K 1 following autoclaving. When appropriate, antibiotics were added at the following concentrations: erythromycin, 10 mg/ml; tetracycline, 6 mg/ml. E. coli cells were grown in LB broth or on plates with 100 mg/ml carbenicillin where appropriate.
Agar spot overlay assays. Parabacteroides and most Bacteroides strains were grown on BHIS agar plates, and Prevotella, Porphyromonas and Dysgonomonas as well as Bacteroides abscessus, B. coprophilus, and B. plebius were grown on Brucella blood plates. Synthesized mature Bd peptides were purchased from LifeTein (Somerset, NJ). Strains to test for sensitivity in overlays were inoculated into 5 ml of prereduced basal or tryptic soy broth (TSB) and grown until they reached mid-log phase; 50 ml of bacteria was added to 5 ml of 0.75% top agar (either BHIS or TSB based) and poured over plates. After the top agar had solidified, 10 ml of a 100-ng/ml solution of synthesized mature Bd-A, Bd-A-FAM, or Bd-D dissolved in deionized (DI) water or 10 ml of DI water (for controls) was spotted onto the top and allowed to dry. Testing for sensitivity to Bd-A and Bd-B produced from Bacteroides strains was performed as previously described (7). Zones of growth inhibition were analyzed after 24 to 72 h, depending on the rate of growth of the strain.
Effects of Bd-A on growth in broth. Bacteria were grown anaerobically to mid-log phase in either basal or TSB and then diluted to an OD 600 of 0.1; 10 ml of this culture was added to 100 ml of fresh medium containing 1 mg of bacteroidetocin A in at least triplicate wells of a 96-well plate. Control wells were inoculated with sterile DI water, and OD 600 was recorded every 10 min for 18 to 48 h using an Eon high-performance microplate spectrophotometer (BioTek Instruments, Winooski, VT). Data were recorded every 10 min and plotted as the average of at least three biological replicates 6 the standard deviation (SD) as calculated by Prism version 8.0.1 for 64-bit Windows (GraphPad Software, San Diego, CA).
Creation of BamA site mutants. Site mutations were created in bamA using the pLGB13 counterselection vector (29). Primers were designed to amplify B. fragilis 638R BamA D545Y , B. vulgatus ATCC 8482 BamA D546D , and B. vulgatus ATCC 8482 BamA W624R (Table S1b). PstI-and BamHI-digested pLGB13 was joined with the mutant bamA products using NEBuilder (New England BioLabs). DNA was transformed into E. coli S17 lpir, and colonies arising on carbenicillin plates were tested by PCR; the plasmids were then sequenced. The plasmids were transferred by conjugation from E. coli to the WT Bacteroides strain, and cointegrates were selected on erythromycin plates. Cointegrates were passaged in nonselective medium and then plated on BHIS plates containing 100 ng/ml anhydrotetracycline, selecting for double-crossover recombinants. Mutants were confirmed by sequencing bamA.
Competitive colonization assay in gnotobiotic mice. Mouse experiments were approved by the Institutional Animal Care and Use Committee (IACUC), Brigham and Women's Hospital. WT and mutant isogenic strains made tetracycline resistant (Tet r ) or erythromycin resistant (Em r ), respectively, were mixed in a 1:1 ratio, and 200 ml was gavaged into each of three male and three female Swiss-Webster germ-free mice housed separately under gnotobiotic conditions in Opti-mouse cages. Seven days after inoculation, fecal samples were collected, diluted, an plated on nonselective BHIS plates, and, after 2 days, samples were replica plated to tetracycline and erythromycin plates. WT and mutant colonies were quantified after 2 days.
Western immunoblot analysis. Bacteria from overnight cultures were boiled in lithium dodecyl sulfate (LDS) sample buffer, and the equivalent of 5 ml of bacterial lysate was separated by electrophoresis using NuPAGE 12% 2-(N-morpholino)ethanesulfonic acid (MES) polyacrylamide gels (Invitrogen). The contents of the gels were transferred to polyvinylidene difluoride membranes, blocked with skim milk, and probed with anti-Bd-B antibodies as previously described (7).
Additional methods. Text S1 contains the methods for widefield fluorescence and transmission electron microscopy, RNA-Seq analyses, isolation and sequencing of BtCL15 strains, and arbitrarily primed PCR analyses.
Data availability. RNA-seq data are provided in Table S2. The BioProject accession number for the genome sequences of BtCL15T12C11 and BtCL15T119C52 is PRJNA728094.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. TEXT S1, DOCX file, 0.1 MB.