Relaxed specificity of BcpB transporters mediates interactions between Burkholderia cepacia complex contact-dependent growth inhibition systems

ABSTRACT Belonging to the two-partner secretion family of proteins, contact-dependent growth inhibition (CDI) systems mediate interbacterial antagonism among closely related Gram-negative bacteria. The toxic portion of a large surface protein, BcpA/CdiA, is delivered to the cytoplasm of neighboring cells where it inhibits growth. Translocation of the antibacterial polypeptide out of the producing cell requires an associated outer membrane transporter, BcpB/CdiB. Some bacteria, including many Burkholderia species, encode multiple distinct CDI systems, but whether there is interaction between these systems is largely unknown. Using Burkholderia cepacia complex species as a model, here we show that related BcpB transporters exhibit considerable secretion flexibility and can secrete both cognate and non-cognate BcpA substrates. We also identified an additional unique Burkholderia dolosa CDI system capable of mediating interbacterial competition and demonstrated that its BcpB transporter has similar relaxed substrate specificity. Our results showed that two BcpB transporters (BcpB-2 and BcpB-3) were able to secrete all four of the B. dolosa BcpA toxins, while one transporter (BcpB-1) appeared unable to secrete even its cognate BcpA substrate under the tested conditions. This flexibility provided a competitive advantage, as strains lacking the full repertoire of BcpB proteins had decreased CDI activity. Similar results were obtained in Burkholderia multivorans, suggesting that secretion flexibility may be a conserved feature of Burkholderia CDI systems. Together these findings suggest that the interaction between distinct CDI systems enhances the efficiency of bacterial antagonism. IMPORTANCE The Burkholderia cepacia complex (Bcc) is a group of related opportunistic bacterial pathogens that occupy a diverse range of ecological niches and exacerbate disease in patients with underlying conditions. Contact-dependent growth inhibition (CDI) system proteins, produced by Gram-negative bacteria, contain antagonistic properties that allow for intoxication of closely related neighboring bacteria via a secreted protein, BcpA. Multiple unique CDI systems can be found in the same bacterial strain, and here we show that these distinct systems interact in several Bcc species. Our findings suggest that the interaction between CDI system proteins is important for interbacterial toxicity. Understanding the mechanism of interplay between CDI systems provides further insight into the complexity of bacterial antagonism. Moreover, since many bacterial species are predicted to encode multiple CDI systems, this study suggests that interactions between these distinct systems likely contribute to the overall competitive fitness of these species.

I n both environmental niches and infection sites, bacteria often reside in diverse polymicrobial communities, where they engage in competitive and cooperative interactions with other microorganisms. Contact-dependent growth inhibition (CDI) system proteins mediate competition among closely related Gram-negative proteobac teria. In CDI systems, the toxic C-terminus of a large surface-localized exoprotein is delivered to the cytoplasm of a neighboring bacterium upon direct cell-to-cell contact (1). Delivery of the toxic domain, which frequently has nuclease activity, results in cell death or growth arrest of the recipient cells. Auto-inhibition is prevented by the production of a cognate immunity protein that binds to the toxin and blocks its activity. Because protection by immunity proteins is allele specific, exchange of CDI system toxins allows for self vs non-self discrimination (1,2). Some organisms encode multiple CDI systems that each containa distinct toxinimmunity pair. In Escherichia coli, Acinetobacter baumannii, Pseudomonas aeruginosa, and several Burkholderia cepacia complex (Bcc) species, these systems have been shown to independently mediate interbacterial competition, often displaying differences in gene expression or toxin potency (3)(4)(5)(6)(7). Whether cross talk may occur among CDI systems produced by the same strain has not been examined.
CDI systems are a subset of two-partner secretion (TPS) pathway (type Vb secre tion) proteins, consisting of an outer membrane transporter (the "TpsB" partner) that facilitates the secretion of its cognate exoprotein ("TpsA") onto the cell surface (8). In CDI systems in Burkholderia and related genera, the bcpA and bcpB genes (termed cdiA and cdiB in other species) encode the exoprotein and outer membrane trans porter, respectively (9,10). Additional components of Burkholderia CDI systems are the immunity protein BcpI and, sometimes, a predicted lipoprotein of unknown function, BcpO.
Research on representative TPS systems has defined a model for the secretion pathway of these proteins. After transport into the periplasm via the Sec machinery, TpsA remains in an unfolded state until the TpsB transporter incorporates TpsA into the outer membrane, and it is progressively folded at the cell surface (11)(12)(13). The TpsA proteins are large filamentous exoproteins with typical hemagglutinin repeats and a conserved N-terminal TPS domain required for recognition by a TpsB transporter (11,14). The TpsB transporters are Omp85 superfamily members that consist of an outer membrane-embedded β-barrel channel, an N-terminal α-helix (H1) plug that inserts into the barrel pore, a short periplasmic polypeptide linker, and two periplasmic polypep tide transport-associated (POTRA) domains (15)(16)(17). The POTRA domains interact with the TPS domain on the TpsA protein and are necessary for substrate recognition and secretion (18,19).
The specificity of a TpsB transporter for its cognate TpsA partner varies between systems. Many TpsB transporters can secrete only their cognate partner, while other transporters can secrete more than one TpsA effector (20). The CdiB transporters from A. baumannii ACICU and E. coli EC93, which share ~23% amino acid sequence identity, are not interchangeable and specifically secrete their cognate CdiA proteins (21). However, little is known about the specificity of more closely related BcpB or CdiB transporters, such as those that would be produced by an organism with multiple CDI systems.
Here, we use B. cepacia complex species that each produced multiple distinct CDI systems to examine the specificity of BcpB transporters for BcpA toxin secretion. The results show that even though each complete CDI system includes an associated BcpB, the transporters display a high degree of promiscuity and generally secrete both cognate and non-cognate BcpA proteins efficiently. While three BcpB proteins in Burkholderia dolosa each secreted multiple BcpA toxins, differences in gene expression appeared to limit which transporters were available. We also report that the relaxed specificity of BcpB proteins extends to Burkholderia multivorans, suggesting that interaction of non-cognate BcpB-BcpA pairs may be a common characteristic of bacterial species that produce multiple CDI systems.

B. dolosa AU0158 contains an additional putative CDI system
B. dolosa strain AU0158 (BdAU0158) was shown to produce three unique CDI systems capable of mediating interbacterial competition, but only system-1 and system-2 were expressed in laboratory conditions (6). Each of the three CDI systems encodes a distinct BcpB transporter, sharing ~80% amino acid identity. Additionally, we identified a fourth bcpB gene downstream of a cryptic bcp locus (referred to as bcp-4) located on BdAU0158 chromosome 3 (Fig. 1A). The bcp-4 region resembles other loci that encode Burkholde ria-type CDI systems, with the gene order bcpAI(O)B. However, the distance between the bcpI and bcpB genes is ~8,000 bp, a gap larger than what is typically found for Burkholderia-type CDI loci. The bcp-4 locus has multiple open reading frames (ORFs) between the bcpI and bcpB genes, although none of these ORFs are predicted to encode a BcpO lipoprotein. Instead, many of the ORFs are predicted to encode transposases, integrases, or genes that produce uncharacterized hypothetical proteins. Interestingly, immediately downstream of bcpAI-4 are additional bcp-like genes: an ORF annotated to encode an immunity 45 family protein and truncated bcpB and bcpA genes. Despite the chromosomal distance between the bcpAI-4 and bcpB-4 genes, the bcp-4 region contains the genetic components necessary to produce a CDI system.

Putative BdAU0158 bcpA-4 promoter is active under in vitro competition conditions
To examine the expression of the bcp-4 genes, ~500 bp 5′ to the bcpA-4 and bcpB-4 translational start sites were fused to promoterless lacZ genes and delivered to attTn7 sites of BdAU0158. The resulting reporter strains were compared to similar reporter strains generated for the other three BdAU0158 bcpA genes (6). When grown in monocul ture under the same conditions as those used for competition experiments, the bcpA-1 and bcpA-2 reporters showed low levels of β-galactosidase activity (Fig. 1B). By contrast, P bcpA-3 -lacZ showed no detectable activity (Fig. 1B), as previously demonstrated (6). The P bcpA-4 -lacZ reporter also showed low levels of β-galactosidase activity, while the P bcpB-4 -lacZ activity levels did not significantly differ from the promoterless control. These data suggest that bcpA-4 is expressed and, therefore, may produce a functional CDI system protein, while bcpB-4 is likely not expressed under the conditions tested.

BdAU0158 bcp-4 encodes a functional CDI system
To determine the functionality of the BdAU0158 bcp-4 CDI system, a mutant strain containing an unmarked, in-frame deletion of bcpA-4 through bcpB-4 was generated (∆bcp-4). When wild-type donor cells were competed against the ∆bcp-4 recipient cells at a 1:1 or 10:1 ratio, there was no difference in the competitive index (CI) between nonimmune recipient cells or those producing cognate immunity protein BcpI-4 (Fig. 1C). These data suggest that the BdAU0158 BcpAIB-4 CDI system does not mediate interbac terial competition under conditions of native expression in vitro.
We hypothesized that native expression of bcpA-4 is not sufficient for CDI-mediated competition and constructed a strain in which the putative bcpA-4 promoter was replaced with the strong, constitutively active Burkholderia thailandensis rpsL (ribosomal S12 subunit) promoter, resulting in strain bcp-4 C . Following co-culture, the BdAU0158 bcp-4 C donor bacteria outcompeted the ∆bcp-4 mutant by ~10-fold when inoculated at a 1:1 ratio and by ~100-fold when inoculated at a 10:1 ratio (Fig. 1C). Introduction of the cognate bcpI-4 immunity gene protected recipient cells from bcpA-4-mediated killing. As expected, complementation with a gene encoding a heterologous immunity protein from B. multivorans (BmCGD2M bcpI-2) did not provide protection against CDI, indicating that bcpI-4 protection was allele specific. These results indicate that the bcp-4 locus encodes a functional CDI system that can mediate interbacterial competition when expressed at a high level.

Cognate BcpB transporters are not required for B. dolosa CDI-mediated competition
Many bacterial species encode multiple CDI systems within the same strain. Since B. dolosa AU0158 contains four unique CDI systems, this strain provides a useful model for investigating potential interplay between distinct CDI systems. To examine the specificity of BdAU0158 BcpB transporters, strains containing unmarked, in-frame deletion mutations of each of the bcpB genes were generated, resulting in ∆bcpB-1, ∆bcpB-2, ∆bcpB-3, and ∆bcpB-4 mutants. Interbacterial competition assays between these bcpB mutants and the corresponding immunity-deficient (∆bcpAIOB) recipient cells were used to determine if each BcpB protein is required for the secretion of its cognate BcpA toxin. Surprisingly, the ∆bcpB-1 mutant inhibited the ∆bcp-1 recipient strain similarly to wild-type donors, implying that wild-type levels of BcpA-1 were still secreted and delivered to recipient bacteria in the absence of BcpB-1 ( Fig. 2A). The ∆bcpB-2 mutant was also able to outcompete susceptible recipient cells but showed a ~10-fold defect in and are indicated by slashes. Non-bcp genes associated with bcp-4 (dark gray) encode hypothetical proteins and putative transposases. (B) Beta-galactosidase assay of lacZ reporters for putative promoters of the BdAU0158 bcpA-1, bcpA-2, bcpA-3, bcpA-4, or bcpB-4 genes and control reporters P S12 -lacZ (constitutive) and promoterless lacZ. Bars show the mean of miller units from three independent experiments, each with three replicates. (C) Interbacterial competition assays between BdAU0158 wild type (WT) or constitutively expressed bcpA-4 (bcp-4 C ) donor bacteria and ∆bcp-4 recipient bacteria that were complemented with the cognate bcpI-4, bcpI-2 from B. multivorans CGD2M (bcpI-2 Bm ), or no bcpI (none). Symbols represent log 10 competitive index values (ratio of donor to recipient) from three independent experiments, and bars show the means (n = 9). Competition assays were performed at a 1:1 or 10:1 (donor to recipient) ratio as indicated. Dashed line shows log 10 competitive index = 0 (no competition). ns, not significant; ***P < 0.001; and ****P < 0.0001; compared to WT donor cells competed against no immunity (none) recipient cells in each panel for competition assays or promoterless reporter for the beta-galactosidase assay.
Research Article mSphere competitive index as compared to wild-type donor bacteria (Fig. 2B). Interbacterial killing was restored to wild-type levels when donor cells were complemented with bcpB-2 at a neutral chromosomal site. Thus, BcpB-2 is also unnecessary for secretion of its cognate BcpA protein but does appear to participate in BcpA-2 secretion. Because bcpA-3 is not expressed under laboratory conditions, competitions investigating this toxin were conducted in strains that constitutively expressed bcpA-3 due to replacement of the native bcpA-3 promoter with P S12 (bcp-3 C ), as previously described (6). The bcp-3 C ∆bcpB-3 mutant outcompeted ∆bcp-3 recipient cells, although the CDI activity was ~100-fold less than for the bcp-3 C parent strain (Fig. 2C). These results show that BcpB-3 contributes to but is not required for cognate BcpA-3 secretion. Competitive Index (CI) for competition assays were calculated as (output donor CFU/recipient CFU)/(input donor CFU/recipient CFU). Symbols represent log 10 CI values from three independent experiments, and horizontal bar shows means (n = 9-18). Competition assays in panel D were performed at a 10:1 ratio. Dashed line shows log 10 competitive index = 0 (no competition). ns, not significant and ****P < 0.0001; compared to WT donor cells competed against no immunity recipient cells. (E) Amino acid alignments of BdAU0158 BcpB-1, BcpB-2, BcpB-3, and BcpB-4 polypeptide transport-associated domains. Similarity is denoted by grayscale; residues similar in all sequences are highlighted in black, and residues similar in 50% of sequences are highlighted in gray. Regions underlined in blue or orange represent POTRA-1 or POTRA-2 domains, respectively.

Research Article mSphere
The bcp-4 overexpression strain (bcp-4 C , Fig. 1C) was also used here to test the contribution of bcpB-4. Similar to bcpB-1, there was no defect in growth inhibition of ∆bcp-4 recipient bacteria by donor cells lacking bcpB-4 ( Fig. 2D), indicating that BcpB-4 is not required for BcpA-4 secretion. However, complementation of the ∆bcpB-4 mutant with overexpressed bcpB-4 resulted in a high level of BcpA-4-mediated CDI, representing a ~1,000-fold increase in competitive index as compared to the bcp-4 C parent or ∆bcpB-4 mutant strains. This high level of CDI activity was eliminated when the complemented donor strain (bcp-4 C ∆bcpB-4+bcpB-4) was competed against recipient cells supplemen ted with cognate bcpI-4 immunity. These data show that BcpB-4 is not necessary for BcpA-4 secretion but suggest that the transporter can secrete the toxin when it is overproduced.
Altogether, these data indicate that the four BcpA toxins still mediate CDI in the absence of their cognate BcpB transporters. Because the percent identity among the BdAU0158 BcpB polypeptide transport-associated domains, POTRA-1 and POTRA-2 are 74% and 93%, respectively ( Fig. 2E; Fig. S1), we hypothesized that BcpA proteins could be secreted by non-cognate BcpB transporters.

Specificity of BcpB transporters for BcpA TPS domains
It has been previously shown for other TPS systems that a truncated TpsA protein consisting of the signal peptide and the TPS domain is efficiently secreted into the culture supernatant in a TpsB-dependent manner (18,20,22,23). To directly examine the secretion of the BcpA proteins, similar TpsA constructs were created for the two proteins that mediate CDI under laboratory conditions, BcpA-1 and BcpA-2. These genetic constructs encoded C-terminally truncated BcpAs encompassing the signal peptide, predicted TPS domain, a portion of the FHA β helical repeat domains, and a FLAG epitope tag (Fig. 3A). To determine the role each BcpB transporter plays in BcpA secretion, a quadruple ∆bcpB mutant lacking all four transporters (∆bcpB1-4) and a series of triple ∆bcpB deletion mutants that each contained only one natively expressed transporter were constructed. The two constructs encoding truncated BcpA-1 and BcpA-2 proteins, termed tpsA-1 and tpsA-2, were each delivered in single copy to a neutral chromosome site in these mutant strains.
TpsA-FLAG proteins of the expected size (~50 kDa) were only detected in culture supernatants ( Fig. 3B and C; Fig. S2). As expected, TpsA-1 and TpsA-2 were not detected in the ∆bcpB1-4 mutant or in the wild-type strain lacking tpsA constructs. Both TpsA-1 ( Fig. 3B) and TpsA-2( Fig. 3C) were detected in supernatants when they were produced in wild-type bacteria or the triple bcpB mutants containing bcpB-2 or bcpB-3 alone. Low levels of TpsA-1 were sometimes observed above the limit of detection in the culture supernatant of the strains containing bcpB-1 or bcpB-4 ( Fig. 3B; Fig. S2D). The nonsecreted TpsA-1 and TpsA-2 did not accumulate in the cytoplasm or insoluble (mem brane) fractions but appeared to be degraded (Fig. S2A).
Overall, these results indicate that truncated BcpA polypeptides are produced and secreted into the culture medium in a BcpB-dependent manner, primarily by BcpB-2 and BcpB-3. This indicates that the domains contained on these proteins are sufficient for BcpA secretion, which is consistent with observations in other TPS and CDI systems (19)(20)(21). Furthermore, these results support our previous findings that secretion of BcpA-1 and BcpA-2 is not dependent upon the cognate BcpB transporter.

BcpB-2 and BcpB-3 transporters can secrete all four BcpA toxins
Our findings suggest that both cognate and non-cognate BcpB transporters participate in BcpA secretion in B. dolosa. To determine the role each BcpB transporter plays in BcpA toxin secretion and delivery, we used the triple ∆bcpB mutants to individually examine the activity of one natively expressed transporter at a time. These bcpB deletion mutants were competed against a series of recipient cells that each lacked one CDI system (thus lacking immunity to only one BcpA protein). As expected, the quadruple ∆bcpB1-4 mutant did not outcompete any recipient strain as it lacks all BcpB transporters (Fig. 4).
Only the donor strains containing bcpB-2 or bcpB-3 outcompeted a ∆bcp-1 recipient, indicating that BcpB-2 and BcpB-3 each secreted BcpA-1 toxin that was capable of mediating CDI (Fig. 4A). However, these competitive indices were significantly less than those of wild-type donors, suggesting that production of only one BcpB transporter is not sufficient for maximum BcpA-1-mediated killing. The donor strains containing only bcpB-1 or bcpB-4 did not outcompete ∆bcp-1 recipient cells, suggesting that BcpA-1 was not secreted by natively produced BcpB-1 or BcpB-4 or that BcpA-1 secreted by these transporters was unable to mediate CDI. While Fig. 1A indicated that BcpB-1 is not required for BcpA-1 secretion, these results further suggest that BcpB-1 does not participate in the secretion of its cognate BcpA protein under these conditions. Similar results supporting the importance of BcpB-2 and BcpB-3 were also found for the remaining BcpA proteins. CDI activity against ∆bcp-2 or ∆bcp-3 mutant recipients was only observed for donor strains producing BcpB-2 or BcpB-3 ( Fig. 4B and C). In each case, donor strains producing the cognate BcpB transporter outcompeted recipient bacteria at levels similar to wild-type donors. Donor bacteria producing the non-cognate trans porter (either BcpB-2 or BcpB-3) also outcompeted recipient cells but at levels less than wild type. These data suggest that BcpA-2 and BcpA-3 can be secreted by multiple BcpB transporters but may prefer their cognate transporters.
BcpA-4 also appeared to utilize BcpB-2 and BcpB-3, but the competitive indices for these mutant co-cultures were significantly less than for the parent strain (Fig. 4D). While FLAG-tagged BcpA-1 TPS (tpsA-1) or (C) FLAG-tagged BcpA-2 TPS (tpsA-2). Wild-type bacteria that lack a tpsA construct (none) were used as a negative FLAG control. Equal protein amounts for each fraction (supernatant and cell lysate) were resolved on SDS-PAGE gels. Panels have shown blotting with anti-E. coli RNA polymerase β subunit (RpoB, top) or anti-FLAG peptide (middle) antibodies and total protein visualization by SYPRO Ruby-stained gels (bottom). Expected masses for TpsA-1, TpsA-2, and RpoB are ~53, ~56, and 150 kDa, respectively. Arrows show TpsA-FLAG or RpoB bands, and asterisks indicate non-specific bands.
Research Article mSphere not definitive, this result suggests that production of either BcpB-2 or BcpB-3 may not be sufficient for maximum BcpA-4-mediated killing. BcpA-4 activity was not observed when secretion depended on BcpB-4, likely due to poor bcpB-4 expression under these conditions (Fig. 1B). Moreover, previous data showed that bcpB-4 overexpression led to increased CDI by BcpA-4 ( Fig. 2D), implying that BcpA-4 can be efficiently secreted by its cognate transporter. By contrast, overexpression of bcpB-1 did not affect BcpA-1mediated CDI ( Fig. 2A), suggesting that low bcpB-1 expression may not explain the lack of CDI activity by donor bacteria that only contain bcpB-1.
Together these findings indicate that B. dolosa BcpA toxins mediate CDI activity when secreted from both cognate and non-cognate BcpB, but the toxins vary in their specificity for the transporters. All four BcpA proteins were secreted from strains the bcpA-3 and bcpA-4 promoters were replaced with the P S12 constitutive promoter to generate bcp-3 C and bcp-4 C parent strains, respectively. Symbols represent log 10 competitive index values (ratio of donor to recipient) from three independent experiments, and bars show the mean (n = 9). Experiments in (D) were performed at a 10:1 (donor to recipient) ratio. Dashed line shows log 10 competitive index = 0 (no competition). ns, not significant; *P < 0.05; ***P < 0.001; and ****P < 0.0001 compared to ∆bcpB1-4 donor cells, unless indicated by line or brackets.
Research Article mSphere containing either bcpB-2 or bcpB-3, but none of the toxins mediated CDI when only bcpB-1 or bcpB-4 was present. An implication of this result is that BcpB-3 must be produced and active, even though the bcpA-3 promoter is inactive under these conditions (6). Thus, the activities of this cognate BcpA/BcpB pair are uncoupled in B. dolosa.
These results are also generally consistent with the TpsA-1 and TpsA-2 secretion assays, which showed secretion primarily by BcpB-2 and BcpB-3 (Fig. 3). While we cannot rule out differences in the secretion of truncated BcpA polypeptides ("TpsA-1" and "TpsA-2") as compared to full-length BcpAs, it is likely that the occasional low level of TpsA secretion detected for BcpB-1 and/or BcpB-4 ( Fig. 3; Fig. S2) was insufficient to cause measurable CDI.

Competition between BcpA-1 and BcpA-2 for secretion by BcpB-3
Since our data indicate that multiple BcpA toxins are secreted by BcpB-2 and BcpB-3, we next sought to determine whether competition occurs for secretion by the available BcpB transporters. To do this, we compared the CDI activities of donor strains that utilized only BcpB-3 but had varying levels of potentially competing BcpA-1 and BcpA-2 proteins. This allowed us to ask whether secretion through a single transporter, BcpB-3, was impacted by levels of substate BcpA proteins. When examining BcpA-1 activity, we asked whether interbacterial killing was impacted by BcpA-2 levels, and vice versa.
Consistent with this hypothesis, a donor strain lacking bcpA-2 (∆bcp-2 ∆bcpB-1) showed significantly higher levels of BcpA-1-mediated CDI (Fig. 5A, panel 5). However, the reciprocal was not true. Loss of bcpA-1 did not alter interbacterial killing by BcpA-2 (Fig. 5B, panel 5). These data suggest that BcpA-2 secretion is less sensitive to the presence of BcpA-1 when both BcpA proteins are utilizing a single BcpB transporter, implying that the BcpB-3 transporter may have a higher affinity for BcpA-2 toxin. Altogether, these data indicate that there can be a competition among BcpA proteins for secretion by limited BcpB transporters, and the transporters likely have a higher affinity for specific BcpA toxins.

Replacing the BcpB-1 POTRA domains changes the functionality of BcpB-1
Previous data showed that natively expressed bcpB-1 does not allow donor cells to inhibit recipient cell growth even with cognate BcpA-1. To determine whether insufficient bcpB-1 gene expression contributes to this defect, we overexpressed bcpB-1 in the quadruple ∆bcpB1-4 mutant that does not contain any native bcpB genes. While bacteria that overproduced BcpB-2 secreted BcpA-1 (Fig. 6B) and BcpA-2 (Fig. 6C), donor bacteria overexpressing wild-type bcpB-1 did not show any interbacterial toxicity. Thus, even overproduced BcpB-1 is defective for some step of BcpA secretion or delivery to recipient cells.
To further investigate the apparent lack of BcpB-1 functionality, we utilized chimeric BcpB proteins. For TPS systems in some Neisseria species, swapping the POTRA domains can change TpsB specificity (19). To test whether differences in the BcpB-1 and BcpB-2 POTRA domains account for the proteins' functional differences, we generated chimeric

BcpB E264 , a close B. dolosa BcpB-1 homolog, can secrete BcpA-1 and BcpA-2
Interestingly, it has been previously reported that the Burkholderia thailandensis E264 bcp and B. dolosa AU0158 bcp-1 toxin and immunity alleles are functionally interchange able. The BdAU0158 BcpI-1 and BtE264 BcpI immunity proteins provided cross-protec tion against both the BdAU0158 BcpA-1 and BtE264 BcpA toxins (6). Even though the BdAU0158 BcpB-1 and BtE264 BcpB proteins are ~95% identical at the amino acid level (Fig. S3), based on our data we hypothesize that they are not functionally identical. Unlike BdAU0158, BtE264 produces only one CDI system with one bcpB gene, so we Research Article mSphere expect the BcpB BtE264 protein to be functional. To examine the functionality of the BtE264 BcpB protein for secretion of BdAU0158 BcpA proteins, we expressed BtE264 bcpB gene in the BdAU0158 ∆bcpB1-4 mutant. Unlike BdAU0158 BcpB-1, BtE264 BcpB was able to mediate CDI against ∆bcp-1 (Fig. 7A) or ∆bcp-2 (Fig. 7B) recipient cells at levels not significantly different from wild-type donor cells. These data indicate that the few amino acid differences between the two closely related BcpB proteins are responsible for a large difference in functionality (Fig. S3). Sequence comparison between BdAU0158 BcpB-1, which appears defective for one or more steps in BcpA secretion or delivery, and the CDI-competent BdAU0158 BcpB-2 and BtE264 BcpB shows a limited number of amino acid differences. Only 20 residues differ in BcpB-1 BdAU0158 as compared to BcpB-2 BdAU0158 or BcpB BtE264 (Fig. S3). Nine of these residues map to the predicted β-barrel of BcpB-1, including three each in β-strands β7 and β16 (Fig. 7C). Six residues are predicted to be found in or immediately adjacent to extracellular loops. Three residues are predicted in the POTRA-1 domain, and one residue is located in each of the plug-linker region and the region N-terminal to H1.

Research Article mSphere
dolosa BcpB-2 also inhibited the growth of recipient cells but to a lesser extent than cells producing the native transporter. Consistent with our previous observations, BdAU0158 BcpB-1 did not allow CDI activity in B. multivorans. Together, these findings indicate that the relaxed specificity of the BcpB transporters occurs in several Bcc species that produce multiple CDI systems. By comparing donor cells that overproduced BcpB proteins (Fig. 8), the relative secretion/delivery efficiencies of each BcpA substrate by each transporter could be determined. BcpB-4 BdAU0158 and BcpB-2 BmCGD2M appeared the most promiscuous, mediating wild-type levels of CDI from all three distinct BcpA substrates (Fig. 8A, B, and D). BcpB-1 BmCGD2M also secreted all three substrates, but cells producing this transporter displayed diminished inter bacterial inhibition (Fig. 8A, B, and C). Interestingly, BcpB-2 BdAU0158 and BcpB-3 BdAU0158 showed variable transporter function. Both proteins mediated wild-type levels of growth inhibition by BcpA-2 BdAU0158 but differed in their abilities to cause CDI by BcpA-1 BdAU0158 and BcpA-2 BmCGD2M (Fig. 8). Overall, the six distinct BcpB proteins examined here showed variable specificity that depended on the particular BcpA substrate.

DISCUSSION
In this study, we investigated the impact of interactions between distinct contactdependent growth inhibition systems on interbacterial antagonism. Surprisingly, we found that BcpB transporters were dispensable for the secretion of their cognate BcpA toxin. BcpB transporters in multiple Burkholderia species showed relaxed specificity and secreted both cognate and non-cognate full-length BcpA toxins. One toxin (BdAU0158 BcpA-1) was secreted exclusively by non-cognate transporters, as its cognate BcpB protein appeared non-functional under the conditions tested here. The promiscuity of the BcpB transporters led to the observation that competition between CDI systems for substrate secretion can occur when transporters are limited. These findings suggest a model in which distinct CDI systems produced by the same organism may not function independently but instead interact to secrete the available pool of toxins (Fig. 9A).
Activity of cognate BcpB/BcpA pairs in B. dolosa was sometimes uncoupled, likely due to differences in gene expression. The bcpA-3 promoter was inactive under laboratory conditions and detection of BcpA-3-mediated CDI required introduction of a strong, constitutive promoter. However, BcpB-3 was highly active under these same conditions and contributed to the secretion of both BcpA-1 and BcpA-2. Expression of bcpB-3 may be due, in part, to an active promoter upstream of bcpI-3 that was previously identified (6). Putative promoters driving bcpA-4 and bcpB-4 also showed differences in promoter activity. This differential gene expression may produce distinct BcpA/BcpB repertoires in different conditions and, combined with the secretion flexibility we observed, could tune optimal toxin secretion for different environmental niches.
Surprisingly, B. dolosa BcpB-1 was not able to secrete or deliver sufficient toxin, either cognate or non-cognate, to mediate interbacterial competition. Changes to the BcpB-1 POTRA-1 domain by replacement with BcpB-2 sequences increased its activity slightly, suggesting that one or more of the 13 amino acid differences in this region contribute to BcpA recognition and/or secretion. In addition, although BdAU0158 BcpB-1 and BtE264 BcpB are ~95% identical, only BcpB BtE264 appeared functional for BcpA secretion. Compar ison of all three transporters (BcpB-1 BdAU0158 , BcpB-2 BdAU0158 , and BcpB BtE264 ) identified 20 residues that are unique to BcpB-1 BdAU0158 (Fig. 7C; Fig. S3). While additional work will be needed to elucidate their potential contributions, three unique residues are located in βbarrel strand β16, which is part of an interface (β1-β16) implicated to undergo rear rangements during CdiB secretion (21). It is possible that these sequence differences allow for BcpB-1 BdAU0158 activity under particular environmental conditions. Alternatively, these differences may represent an accumulation of mutations that decreased BcpB-1 function. Unlike BtE264, which produces a single BcpB transporter, detrimental muta tions might be tolerated in B. dolosa because it produces compensatory BcpB proteins.
The secretion specificity of TpsA-TpsB pairs has been shown to be dependent on recognition of the exoprotein TPS domain (22,23). Although the BcpA proteins are highly variable, the N-terminus which includes the TPS domain is well conserved (Fig. S2C and  S4). B. dolosa BcpA-1 and BcpA-2 share only ~37% identity overall, but their TPS domains are 76% identical. By contrast, the TPS domains of E. coli and A. baumannii CdiA proteins, which are not secreted by each other's CdiB transporters, share only 46% sequence identity (21). Thus, similarity of the TPS domains among BcpA proteins likely accounts for much of the relaxed specificity observed for the BcpB transporters. However, among the relatively similar B. cepacia complex BcpA proteins tested here, there does not appear to be a strong correlation between TPS domain similarity and substrate secretion. For example, the B. dolosa BcpA-1 and B. multivorans BcpA-2 TPS domains are ~95% identical (Fig. S4), but these substrates utilize BcpB transporters with differing efficiencies (Fig. 8A and D). The amino acid variations between these two TPS domains do not appear to map to a particular region (Fig. S4). However, these results suggest that the closely related BcpA and BcpB proteins examined here may provide a useful framework for investigating additional mechanistic details of CDI system protein secretion and toxin release. Moreover, given the precisely controlled release of partially secreted CdiA/BcpA that has been proposed to occur upon recipient cell engagement (25), it is likely that additional interactions between the substrate protein and BcpB transporter are critical to achieve optimal toxin delivery.
While specificity of a "TpsB" transporter for its partner "TpsA" exoprotein is a hallmark of two-partner secretion systems, substrate flexibility has been observed for other systems. Some organisms encode "orphan" TpsA proteins that do not occur with a partner transporter. Bordetella bronchiseptica produces a single transporter, FhaC, which secretes three distinct substrates-FhaB, FhaL, and FhaS (26,27). Similarly, Neisseria Research Article mSphere meningitidis TpsB2 secretes five TpsA proteins, including cognate, non-cognate, and orphan TpsA proteins, while TpsB1 secretes only two of these (20).
Our results indicate that the relaxed specificity of BcpB transporters leads to interactions between distinct CDI systems produced within the same B. cepacia complex strain. Many bacterial species encode two or more complete CDI systems, raising the possibility that similar interactions also occur in these organisms. An examination of >450 clinical and environmental Burkholderia pseudomallei isolates showed that 57% harbored two or three distinct bcpA (termed "fhaB3") gene clusters (28). Acinetobacter baumannii, A. baylyi, and 81% of P. aeruginosa strains carry two cdi loci, several of which have been shown to mediate interbacterial competition (4,5,(29)(30)(31)(32). Comparisons of the CdiB/BcpB proteins that co-occur in these species indicate similarities (Fig. 9B), suggest ing that CDI system interactions may not only occur in other Burkholderia species, such as B. pseudomallei, but also in other Gram-negative bacteria that produce multiple CDI systems.
B. cepacia complex bacteria can occupy various environmental niches and cause opportunistic infections in immunocompromised individuals. The natural niches in which CDI systems are active or provide a fitness advantage are not known, but it may be advantageous for Burkholderia species to produce multiple CDI systems within the same strain. In addition to providing increased toxin diversity and broader immun ity, encoding multiple CDI systems may increase secretion efficiency or flexibility by providing additional BcpB transporters.

Genetic manipulations
Plasmids used in this study are listed in Table S2. All plasmid inserts were confirmed by DNA sequencing (Eurofins Genomics or ACGT, Inc.), and bacterial mutant strains were verified by PCR.
In-frame deletion mutations were constructed by allelic exchange using plasmid pEXKm5 (33). Plasmids for gene deletions were constructed by PCR amplification of two fragments: one fragment ~500 bp 5′ to the ORF (including the first three to seven codons) and another ~500 bp 3′ of the ORF (including the last 3-20 codons).
To complement BdAU0158 or BmCGD2M mutants, the genes of interest were PCR amplified and cloned into an attTn7 site delivery plasmid as described in supplemental material. Bacterial mutants were marked with antibiotic resistance cassettes by attTn7 delivery of pUC18Tmini-Tn-Kan (34), pUCTet (10), or for BmCGD2M only, pUCCm (pMA41) (10). Markers were delivered via triparental matings of E. coli RHO3 with helper plasmid pTNS3, as previously described (35,36).
For strains constitutively expressing the bcp-4 genes, approximately 500 nucleotides 3′ to the bcpA-4 translational start site were PCR amplified and cloned immediately 3′ to the P S12 promoter of plasmid pUCS12. The plasmids were inserted immediately 5′ to the chromosomal copy of each bcp locus, the resulting strains were routinely cultured with kanamycin to select for plasmid retention. Additional details of plasmid and strain construction are described in the supplemental material.

Interbacterial competition assay
Interbacterial competition assays were performed as previously described (36). B. dolosa or B. multivorans strains carrying antibiotic resistance cassettes at attTn7 sites were cultured overnight without antibiotics and resuspended in sterile PBS to an OD 600 = 2. Unless noted otherwise, bacteria were mixed at a 1:1 ratio, 20 µL of the mixture was plated on LSLB agar in triplicate, and plates were incubated at 37°C for 24-26 hours. The culture inoculum was plated on LSLB with antibiotic selection to determine the input ratio (donor:recipient) at 0 hours. All donor and recipient bacteria are marked with kanamycin or tetracycline resistance cassettes, respectively. Bacteria were collected from co-cultures with a sterile loop, diluted in sterile PBS, and plated on LSLB with antibiotics to quantify each strain. CI was calculated as a ratio of the donor strain to the recipient strain at 24 hours divided by the input (donor:recipient) ratio at 0 hours. At least three independent experiments were performed in triplicate.

Beta-galactosidase assay
Reporter strains cultured in LSLB were spotted (20 µL) onto agar plates, incubated overnight at 37°C, and collected in PBS. Cell suspensions were normalized to an OD 600 = 1.5. Cells were permeabilized with SDS and chloroform, and beta-galactosi dase activity was measured as described (6) using a SpectraMax 5M plate reader (Molecular Devices). Three independent experiments were performed, each with three biological replicates.

Subcellular fractionation, secretion assay, and western blotting
Subcellular fractionation of BdAU0158 was performed as previously described (37) with modifications. Bacterial strains were cultured overnight at 37°C with agitation in LSLB. Cells were harvested by centrifugation at 12,000 × g for 15 minutes at 4°C, and the pellet was resuspended to an OD 600 = 8 in Tris resuspension buffer (50 mM Tris, pH 8 supplemented with Roche Complete Mini EDTA-free Protease Inhibitor Cocktail and Pierce Universal Nuclease for Cell Lysis). Proteins were precipitated from the culture supernatants with 15% trichloroacetic acid, washed with acetone, and dissolved in 10 mM Tris, pH 8 with 2% sodium dodecyl sulfate (SDS) (Tris/SDS buffer). Cells were broken by three passages through a chilled French Pressure cell (40,000 lb/in 2 ), and unbroken cells and large debris were removed by two centrifugations at 12,000 × g at 4°C for 15 minutes. Total membranes were separated by ultracentrifugation for 15 minutes at 100,000 × g at 4°C, and supernatants collected to analyze the cytoplasmic fraction. Total membranes were washed with Tris resuspension buffer, the resulting pellet was resuspended in Tris/SDS buffer. The cytoplasmic fraction was concentrated by methanol-chloroform precipitation, and resulting pellets were suspended in Tris/SDS buffer. Protein concentration for all fractions was determined by Microplate BCA Assay (Pierce).
Equal protein amounts for each fraction were analyzed by SDS-PAGE on Novex 10%-20% Tricine gels (Invitrogen) and transferred to polyvinylidene difluoride 0.2 µM membranes (Invitrogen). Immunoblots were probed as previously described (37) with mouse monoclonal anti-FLAG M2 (Sigma) or anti-E. coli RNA Polymerase β (Biolegend) and secondary antibodies coupled to IRDye 800CW (Licor). SYPRO Ruby staining of the gels containing the supernatant and cell lysate was used for visualization of protein loading. Immunoblots and gels were imaged on a Gel Doc EZ Imager (Bio-Rad).

Bioinformatics and statistics
The putative bcp-4 locus was identified in the complete genome of B. dolosa AU0158 by BLASTp, using BdAU0158 BcpB proteins for queries. Protein alignments were performed using the Clustal W alignment feature of Geneious Prime (2022.2.1), and phylogenetic trees generated using the associated Geneious Tree Builder. Domain predictions were performed using NCBI Conserved Domain search. Predicted B. dolosa BcpB-1 structure was generated by AlphaFold (38)(39)(40), and structure visualized and shaded using UCSF ChimeraX v1.6.1 (41). Data were analyzed by one-way ANOVA with Tukey post hoc test or Student's t test using the statistical package in GraphPad Prism (v9).

ACKNOWLEDGMENTS
This work was supported by a grant from the National Institutes of Health (R01AI150767 to E.C.G.).
We thank the Peggy Cotter laboratory (University of North Carolina-Chapel Hill) for providing plasmid components and mutant strains.

DATA AVAILABILITY
The raw data that support the findings of this study are found in the supplemental material (Table S3) or available from the corresponding author upon request.