The Burkholderia bcpAIOB Genes Define Unique Classes of Two-Partner Secretion and Contact Dependent Growth Inhibition Systems

Microbes have evolved many strategies to adapt to changes in environmental conditions and population structures, including cooperation and competition. One apparently competitive mechanism is contact dependent growth inhibition (CDI). Identified in Escherichia coli, CDI is mediated by Two–Partner Secretion (TPS) pathway proteins, CdiA and CdiB. Upon cell contact, the toxic C-terminus of the TpsA family member CdiA, called the CdiA-CT, inhibits the growth of CDI− bacteria. CDI+ bacteria are protected from autoinhibition by an immunity protein, CdiI. Bioinformatic analyses indicate that CDI systems are widespread amongst α, β, and γ proteobacteria and that the CdiA-CTs and CdiI proteins are highly variable. CdiI proteins protect against CDI in an allele-specific manner. Here we identify predicted CDI system-encoding loci in species of Burkholderia, Ralstonia and Cupriavidus, named bcpAIOB, that are distinguished from previously-described CDI systems by gene order and the presence of a small ORF, bcpO, located 5′ to the gene encoding the TpsB family member. A requirement for bcpO in function of BcpA (the TpsA family member) was demonstrated, indicating that bcpAIOB define a novel class of TPS system. Using fluorescence microscopy and flow cytometry, we show that these genes are expressed in a probabilistic manner during culture of Burkholderia thailandensis in liquid medium. The bcpAIOB genes and extracellular DNA were required for autoaggregation and adherence to an abiotic surface, suggesting that CDI is required for biofilm formation, an activity not previously attributed to CDI. By contrast to what has been observed in E. coli, the B. thailandensis bcpAIOB genes only mediated interbacterial competition on a solid surface. Competition occurred in a defined spatiotemporal manner and was abrogated by allele-specific immunity. Our data indicate that the bcpAIOB genes encode distinct classes of CDI and TPS systems that appear to function in sociomicrobiological community development.


Introduction
Microbes rarely live alone.Whether free in the environment or in close association with eukaryotic hosts, microbes typically share their living space with other viral, prokaryotic, and/or eukaryotic microorganisms.Survival under these conditions requires mechanisms for sensing, responding to, and cooperating or competing with neighboring organisms.Contact dependent growth inhibition (CDI) systems are protein toxin delivery mechanisms that appear to be involved in interbacterial competition [1].CDI was discovered in Escherichia coli strain EC93 due to its ability to inhibit the growth of specific CDI 2 E. coli strains upon cell-to-cell contact.CDI is mediated by Two-Partner Secretion (TPS) system proteins CdiA and CdiB [1].TPS systems are widespread amongst Gram-negative bacteria.They export large exoproteins (TpsA family members such as CdiA) across the outer membrane using pore-forming b-barrel proteins (TpsB family members such as CdiB) [2,3].Functions attributed to TpsA proteins before the discovery of CDI included adherence to eukaryotic cells, induction of cytolysis in host cells, iron uptake, and autoaggregation [2,3].Characterization of CDI in E. coli revealed an additional TpsA-mediated function: inhibition of 'target' bacterial cell growth upon contact.CDI + bacteria are protected from autoinhibition because they produce CdiI, a 79 amino acid 'immunity' protein encoded immediately 39 to cdiA [1].
Research in our lab on Burkholderia pseudomallei led to the discoveries that genes predicted to encode CDI systems are present in a large number of a-, b-, and c-proteobacteria, that the Cterminal ,300 aa of CdiA proteins (CdiA-CTs) and CdiI proteins are highly variable, and that CdiA-CTs are sufficient to confer toxicity when produced intracellularly in E. coli [4].Some CdiA-CTs have been demonstrated to possess nuclease activity, functioning as DNases or tRNases [4,5].CdiI proteins bind to cognate CdiA-CT proteins (those encoded by the same cdi locus), blocking their nuclease activity, but not heterologous CdiA-CT proteins (those encoded by different cdi loci) [4,5].Experiments with E. coli strains producing chimeric CdiA proteins showed that CdiI proteins provide immunity against interbacterial growth inhibition in an allele-specific manner, conferring protection only towards cognate CdiA-CTs but not heterologous CdiA-CTs [4].Although chimeric CdiA proteins containing CdiA-CTs encoded by different species of bacteria were effective at mediating interbacterial competition in E. coli, CDI has so far only been shown to function between members of the same species [1,4].The current model for CDI states that, upon cell-to-cell contact with a closely related bacterium (possibly by interacting directly with the outer membrane protein BamA [6]), the CdiA-CT from a CDI + cell is delivered to the cytoplasm of the target cell where it inhibits cell growth by degrading DNA or specific tRNAs.If present in the target cell, the cognate CdiI immunity protein binds to the CdiA-CT, blocking its nuclease activity [4].While considerable insight has been gained regarding how CDI systems function mechanistically in E. coli, almost nothing is known about when or why these systems are deployed by bacteria in nature.
Initial bioinformatic analysis of available bacterial genomes revealed that putative CDI systems fall into two distinct classes, ''E.coli-type,'' which include systems found in bacterial genera other than Burkholderia, and ''Burkholderia-type,'' found only in Burkholderia spp [4].E. coli-type CDI systems are encoded by genes with the order cdiBAI and predicted CdiA proteins contain a highly conserved VENN motif that separates the conserved (,2700 aa) N-terminus from the variable (,300 aa) C-terminus (the CdiA-CT).Burkholderia-type CDI systems are encoded by genes with the order cdiAIB and putative CdiA proteins contain an NxxLYN motif instead of VENN [4].Whether the Burkholderia proteins actually function as CDI systems has not yet been demonstrated.
Burkholderia spp are Gram-negative soil saprophytes and many are opportunistic pathogens [7,8,9].Burkholderia cepacia complex (Bcc) strains, for example, cause life-threatening respiratory infections in cystic fibrosis patients [10,11], and B. pseudomallei strains cause melioidosis, a disease that can range from localized wound infections and abscesses to fulminant pneumonia and septicemia [9,12].Because it is highly virulent by the aerosol route, resistant to most commonly used antibiotics, and extremely closely related to Burkholderia mallei, which has been used as a bioterrorism agent in the past, B. pseudomallei is an NIAID Category B Biothreat pathogen and select agent [13].Working with B. pseudomallei in the laboratory requires BSL-3 practices and rigorous security measures.Burkholderia thailandensis is closely related to B. pseudomallei and occupies the same environmental niche (both are endemic to southeast Asia and northern Australia) [14,15,16], but is not a human pathogen, is not a select agent, and requires only BSL-1 practices.
Here, we characterize the unique class of CDI systems produced by Burkholderia spp.We show that these systems compose a novel class of TPS system that requires a third protein for the large exoprotein to function, that expression of Burkholderia CDI proteinencoding genes is regulated in a probabilistic manner, that the gene products contribute to biofilm formation, and that CDImediated interbacterial competition in Burkholderia occurs on solid surfaces in a unique temporal and spatial pattern.

Results
bcpAIOB genes of Burkholderia spp compose a unique class of CDI system-encoding loci Identification and general features of Burkholderia-type CDI system-encoding loci.Initial comparisons of putative CDI system-encoding loci included six from Burkholderia spp and indicated that the Burkholderia loci had a different gene order and a different motif separating the conserved and variable regions of the predicted CdiA orthologs compared with E. coli-type CDI systemencoding loci [4].Upon closer inspection of putative cdi loci in B. pseudomallei strains, we noticed an additional small open reading frame (ORF) located between the predicted cdiI and cdiB homologs (Figure 1A).Because further investigation indicated that Burkholderia-type CDI systems form a distinct class of CDI system and also a novel class of TPS system, we chose to give them a distinct nomenclature.We therefore named the genes encoding Burkholderia-type CDI systems bcpAIOB, for Burkholderia CDI proteins A, I, O and B.
To identify all Burkholderia-type CDI-encoding loci in organisms with available genome sequence information, we performed a BLAST search using the predicted BcpB protein from B. pseudomallei K96243, downloaded approximately 20 kbp of DNA sequence flanking each identified bcpB homolog, and used Vector NTI to identify ORFs and to perform sequence comparisons.We found 58 loci in which there was a large ORF (,10 kb) predicted to encode a TpsA protein (the bcpA homolog) 59 to the bcpB homolog and a small ORF (,300 bp) immediately 39 to the bcpA homolog (Table S1, Figure S1).(Note that finished genome sequences were available for four strains of B. pseudomallei and about half of the other organisms in which we identified Burkholderia-type CDI-encoding loci (as indicated in Table S1).The other loci were identified in publically available contigs from unfinished genome sequencing projects.)Thirty of the identified loci were present in 20 B. pseudomallei strains (four strains contain two loci and three strains contain three loci) and two were identified in B. thailandensis strains.The other 26 loci were present in Bcc strains (Burkholderia ambifaria, Burkholderia cenocepacia, Burkholderia dolosa, Burkholderia multivorans, and Burkholderia vietnamiensis), Burkholderia gladioli, Burkholderia glumae, Burkholderia phymatum, Burkholderia rhizoxinica, Burkholderia ubonensis, Burkholderia xenovorans, Ralstonia solanacearum, Ralstonia syzygii, and Cupriavidus metallidurans.Ralstonia and Cupriavidus are the two genera most closely related to Burkholderia.No Burkholderia-type CDI-encoding loci were identified in any other bacteria.However, E. coli-type CDI-encoding loci were identified in some closely related bacteria (e.g., Cupriavidus taiwanensis) [4].In no case were both E. coli-type and Burkholderiatype CDI-encoding loci present in a single strain or even different strains of the same species.Amongst all identified Burkholderia-type CDI-encoding loci, predicted BcpA proteins ranged in size from 2814 aa to 3651 aa, with most about 3100 aa, and nearly all contained a Nx(E/Q)LYN motif located approximately 350 aa from the Cterminus.Predicted BcpI proteins ranged in size from 44 aa to 301 aa, with most about 100 aa.In all strains except one, the first

Author Summary
Contact dependent growth inhibition (CDI) is a phenomenon discovered in Escherichia coli in which CDI + bacteria inhibit the growth of CDI 2 bacteria upon cell-to-cell contact.CDI is mediated by large toxic ''exoproteins'' present on the bacterial cell surface.An 'immunity' protein protects CDI + cells from killing themselves.While predicted CDI systems are widespread throughout bacterial genera, the role of these systems in nature has remained elusive.Here we identify a distinct class of CDI system in Burkholderia species.The genes encoding these systems are expressed in a stochastic manner such that only a few cells in the population produce the proteins at any given time when grown in broth.We also show that these systems are required for aggregation on an abiotic surface, suggesting an important role for CDI in biofilm formation.Finally, we show that CDI mediates competition under specific conditions in a precise spatiotemporal pattern when bacteria are grown on a solid surface.Our data suggest that in nature, CDI systems may be used by bacteria to establish complex sociomicrobial communities.codons of the bcpI genes were within 11 bp of the stop codons of the bcpA genes, and in some cases they were 59 to the bcpA stop codons, suggesting that the bcpA and bcpI genes may be both transcriptionally and translationally coupled.The one strain that differed in this respect was B. thailandensis TXDOH in which the predicted bcpI gene was located 313 bp 39 to the end of the bcpA gene.The predicted BcpA-CT of this strain (i.e., the region of the BcpA protein from the Nx(E/Q)LYN motif to the C-terminus) is about 90 aa shorter than most of the other BcpA-CTs, suggesting the possibility that the identified bcpA stop codon is the result of a sequencing error.However, translating the DNA sequence between the identified stop codon and the 59 end of the identified bcpI gene in all three reading frames did not result in one continuous ORF without stop codons, so the discrepancy with this strain cannot be explained by a single nucleotide sequencing error.In all but eight loci, a small ORF was present between the bcpI and bcpB genes.We named this ORF bcpO.
Comparison of B. pseudomallei and B. thailandensis bcpAIOB genes.The 32 loci present in B. pseudomallei and B. thailandensis strains comprised 16 different alleles in which the nucleotide sequences from 59 of the bcpA gene to 39 of the bcpB gene (i.e., the entire bcpAIOB locus) were nearly identical (Table 1, Table S1).Some alleles were present in multiple strains (e.g., each of the three different alleles present in B. pseudomallei 1106A is present in other strains), while others are present in only a single strain.If a strain contains more than one allele, the alleles are different (duplicated alleles in the same strain were not found).Moreover, Tuanyok et al. identified genomic islands in five B. pseudomallei strains and found that these loci (which they referred to as 'fha gene clusters') were present on genomic islands [17].Based on the limited number of strains used in that study, it appears that specific alleles are associated with specific genomic islands, i.e., if two strains contain the same allele, the alleles are present in the same location (same genomic island) in the chromosome.
The predicted aa sequences of the BcpB proteins encoded by all alleles were highly similar (89% similarity with the consensus sequence), but separated into three groups phylogenetically (Figure 1B).Similarly, the predicted aa sequences of the Nterminal ,2750 aa of the BcpA proteins encoded by all alleles (Nterminal to the Nx(E/Q)LYN motif) were highly similar, but also separated into the same three phylogenetic groups.In contrast, the aa sequences of the C-terminal ,350 aa of the BcpA proteins encoded by each of the different alleles were typically less than 10% similar.For a few of the alleles, however, the BcpA proteins appeared to be mosaics, with only the last ,100-200 aa being different (e.g., alleles 3-7, and alleles 8 and 9 in Figure 1B).In all cases, the predicted aa sequences of the BcpI proteins differed dramatically with no more than 10% similarity between any two alleles.In a few of the E. coli-type CDI systems, the immunity proteins have been shown to bind to cognate CdiA-CTs in an allele-specific manner [4].If the Burkholderia-type CDI systems function analogously and the immunity proteins for alleles 3-9 function in an allele-specific manner, it will suggest that BcpI binds to the C-terminal ,100-200 aa of BcpA.
Bioinformatic analysis of the small ORFs (bcpO genes) unique to Burkholderia-type CDI-encoding loci.The predicted BcpO proteins encoded by alleles 1-5, 7, and 8 are 73 or 74 aa long and contain N-terminal signal sequences with lipoboxes.Although the signal sequences of the preproteins encoded by the different alleles vary, the predicted mature proteins are nearly identical, i.e., the variation is located within the signal sequence.None of the predicted BcpO proteins contains a Lol avoidance signal (an Asp residue following the N-terminal Cys of the mature lipoprotein), therefore these BcpO lipoproteins are predicted to localize to the inner leaflet of the outer membrane.The fact that the mature BcpO proteins are identical within this family suggests that they play a conserved, non-allelespecific function, however they share no similarity to other characterized proteins.
The predicted BcpO proteins for alleles 10-16 are more heterogeneous, and their sequences vary in an allele-specific manner with their cognate BcpA-CT and BcpI proteins for the most part (alleles 14-16 being exceptions).They vary in size from 57 aa to 172 aa.None contains a predicted N-terminal signal sequence, and they have no similarity to any characterized proteins or protein domains.

The bcpAIOB genes are transcribed as an operon in B. thailandensis
To investigate the operon structure of the bcp locus in B. thailandensis strain E264, reverse transcriptase (RT) PCR was performed on RNA extracted from bacteria cultured in low salt LB broth (LSLB), the standard medium used for culturing B. thailandensis.Primer sets (Table S3) flanking the junctions between bcpA and bcpI, bcpI and bcpO, and bcpO and bcpB (1, 2, and 3, respectively, in Figure 2C) yielded products of the expected sizes (Figure 2A, top panel), indicating the bcpAIOB genes form an To measure expression of bcpAIOB, we constructed a strain containing a bcpA promoter-lacZ fusion (P bcpA -lacZ) inserted at the attTn7 site on the chromosome in B. thailandensis E264 and measured b-galactosidase activity in cells cultured under various conditions.For comparison, we also constructed two additional strains, one in which the promoter of the gene encoding the ribosomal S12 subunit (P S12 ) was fused to lacZ and one with no promoter (P neg ) 59 to lacZ.Approximately 2,000 Miller units of bgalactosidase activity were produced in the P bcpA -lacZ fusion strain cultured under all conditions tested (Figure 2B).The P S12 -lacZ fusion produced ,11,000 Miller units of b-galactosidase activity and the P neg -lacZ fusion produced ,500 Miller units of bgalactosidase activity in cells cultured in LSLB broth.These data suggest bcpA is transcribed at relatively low levels under each of the culture conditions tested.
We next constructed strains with in-frame deletion mutations in genes in the bcpAIOB operon.While it was possible to construct a DbcpO strain and a DbcpB strain by allelic exchange, and a strain in which the entire bcpAIOB operon was replaced with a gene encoding kanamycin resistance by natural transformation (Figure 2C), it was not possible to construct a DbcpI strain, suggesting bcpI is essential or, analogous to the E. coli CDI system, BcpI is required to protect against BcpA-mediated toxicity.For complementation experiments, we constructed plasmids to deliver bcpO or bcpB, driven by the constitutively active P S12 promoter or the native promoter (P bcpA ) to the attTn7 site.All mutant strains grew equally compared to wild type E264 in LSLB medium (data not shown).
BcpA protein production in wild type, DbcpO, and DbcpB B. thailandensis Initial attempts to investigate the contribution of bcpO and bcpB to BcpA production were carried out by performing Western blots from cell lysates of bacteria expressing the bcpAIOB genes from their native promoter, P bcpA , and producing BcpA containing a hemagglutinin (HA) epitope (BcpA-HA) N-terminal to the predicted Nx(E/Q)LYN sequence of the mature protein (i.e., immediately C-terminal to F2633, 141 amino acids from the N- Top panel, analysis of the operon structure of bcpAIOB.Primer sets 1, 2, and 3 flanking the intergenic region of bcpA and bcpI, bcpI and bcpO, and bcpO and bcpB, respectively, (as indicated in 2C) were used for RT-PCR.Middle and bottom panels, analysis of the approximate start site of transcription of bcpA.RT-PCR was performed using primers annealing 50, 70, 120, 150, 200, 250, and 300 nt 59 to bcpA with a reverse primer internal to bcpA.All products were visualized by ethidium bromide.B) Expression of bcpA.P S12 -lacZ and P neg -lacZ (dark gray bars) and P bcpA -lacZ (light gray bars) reporter strains were cultured under various conditions and assayed for b-galactosidase activity.Error bars represent the mean 6 1 SEM.C) Schematic of B. thailandensis E264 strain constructs, including WT, DbcpO, DbcpB, and DbcpAIOB, used throughout the study.doi:10.1371/journal.pgen.1002877.g002 terminal side of the Nx(E/Q)LYN sequence).We were unable to detect BcpA in this strain, possibly because bcpAIOB expression was insufficient under the growth conditions used.We therefore constructed B. thailandensis strains in which the bcpAIOB operon was controlled by the constitutively active ribosomal S12 subunit promoter, P S12 .In immunoblots of whole cell lysates of otherwise wild type B. thailandensis (E264BcpA-HA::pECG22), anti-HA antibodies recognized a polypeptide with considerably slower mobility than the 250 kDa molecular weight marker (possibly corresponding to 306 kDa, the predicted molecular mass of BcpA), plus three slightly smaller and much less abundant polypeptides (Figure 3).The polypeptide profile detected in the DbcpO strain was identical to that of E264BcpA-HA::pECG22.By contrast, only a very low level of the largest polypeptide was detected in whole cell lysates of the DbcpB strain (E264BcpA-HADbcpB::-pECG22) (Figure 3).In other TPS systems, the TpsA protein is undetectable when the strain contains a loss-of-function mutation in the TpsB-encoding gene, presumably because the TpsA protein, which cannot be translocated across the outer membrane, is degraded in the periplasm [18,19].Our data suggest that the same is true for bcpA and bcpB.Complementation of the DbcpO and DbcpB strains with bcpO and bcpB, respectively, expressed from the P S12 promoter did not alter the polypeptide profiles of these strains (Figure 3).However, the fact that BcpA-HA protein was detectable in the DbcpO strain and not in the DbcpB strain indicates that the DbcpO mutation does not have polar effects that abrogate expression of bcpB.Similarly, RT-PCR indicated that bcpB transcription was not abrogated by the in-frame deletion in the DbcpO strain (Figure S2).

The bcpAIOB operon is expressed in a probabilistic manner when B. thailandensis is cultured in liquid medium
Upon plating the P bcpA -lacZ fusion strain on solid medium containing X-gal, we found that the colonies were not all the same intensity blue; approximately 5-10% of the colonies were dark blue and 90-95% of the colonies were light blue (Figure 4A).To explore the possibility that bcpA was highly expressed in just a small proportion of cells in the population, we constructed fluorescent reporter fusion strains by delivering a gfp gene fused to P bcpA or P S12 to the attTn7 chromosomal insertion site in B. thailandensis  E264.Bacteria were cultured in liquid broth and visualized by confocal microscopy.All bacteria containing the P S12 -gfp fusion produced high levels of GFP (Figure 4B).By contrast, only a few fluorescent bacteria were present in cultures containing the P bcpAgfp fusion.These data suggest bcpA is differentially expressed within the bacterial population.
We next performed flow cytometry to measure bcpA-gfp expression in a large number of bacterial cells when cultured in liquid medium.Events distinct from the PBS control (Figure S3) were identified as bacteria and gated on for subsequent analysis.Approximately 99% of the bacteria in cultures of the P S12 -gfp fusion strain were GFP + (Figure 4C, Table S2), whereas only ,0.2% of the P bcpA -gfp fusion containing bacteria were GFP + , indicating that bcpA is differentially expressed within a population of bacteria when cultured in liquid medium.The mean relative fluorescence intensity of the P bcpA -gfp GFP + cells was similar to that of P S12 -gfp GFP + cells (Table S2), indicating expression was very high in the GFP + P bcpA -gfp bacteria.Taken together, these data show that only a small percentage of bacteria express bcpA when cultured in liquid medium, but those that do express bcpA do so at a high level.

The bcpAIOB genes are required for autoaggregation in M63 minimal medium
Culturing wild type B. thailandensis in M63 minimal medium resulted in a dramatic autoaggregation phenotype in which bacteria aggregated and adhered to the walls of glass test tubes (Figure 5A).By contrast, the DbcpAIOB strain grew as a homogenous suspension of planktonic cells (Figure 5A), indicating that the bcpAIOB genes are required for autoaggregation.The DbcpB and DbcpO strains also grew planktonically (Figure 5A).Because TpsB family members are required for secretion of TpsA proteins to the bacterial surface and Western blot data confirmed that no BcpA protein could be detected in the DbcpB strain (Figure 3), the DbcpB strain was expected to have the same phenotype as the DbcpAIOB strain.The fact that the DbcpO strain failed to autoaggregate, however, indicates that the BcpO protein is required for BcpA function.Because the BcpA protein profile in the DbcpO mutant was identical to that of wild type bacteria, BcpO likely contributes to maturation events that occur during or after translocation across the outer membrane.Because CDI systems have so far been demonstrated to function only in interbacterial growth inhibition between CDI + and CDI 2 bacteria, these data represent the first demonstration of a phenotype for a strain defective for expression of genes encoding a (putative) CDI system in a homogeneous culture.
Efforts to complement the bcpO and bcpB genes were not successful in restoring the wild type autoaggregation phenotype.Specifically, expression of bcpO and bcpB from the P S12 promoter or the native promoter, P bcpA , at the attTn7 site failed to restore autoaggregation in the DbcpO and DbcpB mutants, respectively (data not shown).To address the possibility that lack of autoaggregation in the DbcpO mutant was due to an unintended mutation other than the DbcpO mutation, we tested several independently constructed DbcpO mutant strains.All of these strains grew planktonically and failed to autoaggregate.Together, these data indicate that the bcpAIOB genes are required for autoaggregation, and they suggest that expression of the bcpO and bcpB genes from their native locus is critical for proper function of their gene products.
A strain in which expression of the bcpAIOB genes was controlled by the constitutively active P S12 promoter (P S12 -bcpAIOB) also autoaggregated when cultured in M63 minimal medium, but with different kinetics and aggregation characteristics compared with wild type B. thailandensis.The P S12 -bcpAIOB strain aggregated but was not adherent to the walls of the test tube at 24 hours, and by 48 hours, the adherent/aggregated bacteria had a smoother, more mucoid appearance (Figure 5B).These data indicate that controlled expression of the bcpAIOB genes (high in approximately 0.2% of the population and undetectable in the rest) is important for the wild type autoaggregation phenotype.
Autoaggregation and adherence to the walls of test tubes by B. thailandensis may be a form of biofilm.Because DNA has been shown to be an important component of many bacterial biofilms [20], we sought to determine if DNA contributed to the B. thailandensis autoaggregation phenotype.Addition of 4 U DNaseI decreased the amount of autoaggregation and addition of 10 U DNaseI completely abrogated autoaggregation (Figure 5A), causing the bacteria to grow planktonically.Extracellular DNA is therefore required for the autoaggregation phenotype.To quantify the amount of extracellular DNA in wild type cultures compared to DbcpAIOB cultures, supernatants were filter sterilized and analyzed by spectrophotometry.No difference in the quantity of DNA could be detected (data not shown), suggesting there may be variations in the quality of DNA present in the two cultures that mediate autoaggregation and/or that DNA-BcpA interactions are required for autoaggregation.
Burkholderia BcpA-CT and BcpI proteins are toxinimmunity pairs that function in an allele-specific manner Our inability to construct an E. coli strain producing the Cterminal ,350 aa of BcpA from B. pseudomallei 1026b unless the cognate BcpI protein was also produced provided the first evidence that the C-terminal domains of CdiA/BcpA proteins are sufficient to cause toxicity when produced intracellularly [4].Here, we constructed plasmids to encode the last ,350 aa (including the Nx(E/Q)LYN motif) of BcpA from B. pseudomallei K96243 (BcpA-CT BpK96243 ) and B. pseudomallei 1106A-2 (BcpA-CT Bp1106A-2 ) (with an added ATG at the 59 end) 39 to the rhamnose inducible promoter P rhaB .Overnight cultures of B. thailandensis containing these plasmids were supplemented with 0.2% glucose (to suppress P rhaB ) and diluted into LSLB containing either 0.2% glucose or 0.2% rhamnose.Bacterial viability was monitored after four hours of culture at 37uC by counting colony forming units (cfu).The number of cfu/ml of B. thailandensis strains containing these plasmids cultured in medium containing 0.2% glucose was not altered after four hours (Figure 6A).By contrast, when cultured with 0.2% rhamnose to induce P rhaB , the number of cfu/ml of B. thailandensis strains containing the plasmids encoding BcpA-CT BpK96243 and BcpA-CT Bp1106A-2 decreased by 3 and 2 logs, respectively (Figure 6A, left and right panels, respectively).These results indicate BcpA-CTs are sufficient to cause toxicity when produced intracellularly in B. thailandensis.
We next investigated the ability of BcpI to provide immunity to BcpA-CT-mediated intracellular toxicity.We constructed another set of plasmids encoding BcpI proteins from B. pseudomallei K96243 (BcpI BpK96243 ) and B. pseudomallei 1106A-2 (BcpI Bp1106A-2 ) also under control of P rhaB .B. thailandensis containing these plasmids in combination with the BcpA-CT-encoding plasmids were cultured as described above.Again, bacterial viability was monitored after four hours.The number of cfu/ml of B. thailandensis harboring the plasmids encoding cognate BcpA-CT BpK96243 and BcpI BpK96243 increased 0.5 log when cultured with 0.2% glucose and decreased only slightly when cultured with 0.2% rhamnose (Figure 6B, left panel).Similarly, the number of cfu/ml of B. thailandensis harboring the plasmids encoding cognate BcpA-CT Bp1106A-2 and BcpI Bp1106A-2 decreased slightly when cultured with 0.2% glucose and decreased less than 1 log when cultured with 0.2% rhamnose (Figure 6B, right panel).While the decrease in cfu/ml of B. thailandensis containing the plasmids encoding BcpA-CT Bp1106A-2 and BcpI Bp1106A-2 observed when cultured with 0.2% rhamnose was statistically significant (p,0.006), the log fold change of this strain compared to the log fold change of B. thailandensis harboring the BcpA-CT Bp1106A-2 encoding plasmid alone was also statistically significant (p,0.0001)when cultured with 0.2% rhamnose.These data indicate cognate BcpI proteins are able to rescue the toxic phenotypes observed when BcpA-CTs are produced intracellularly.By contrast, the number of cfu/ml of B. thailandensis harboring plasmids encoding the non-cognate pair BcpA-CT BpK96243 and BcpI Bp1106A-2 decreased by 3 logs when cultured with 0.2% rhamnose, but did not change when cultured with 0.2% glucose (Figure 6C, left panel); and B. thailandensis containing plasmids encoding the non-cognate pair BcpA-CT Bp1106A-2 and BcpI BpK96243 decreased by 2 logs when cultured with 0.2% rhamnose, but did not change when cultured with 0.2% glucose (Figure 6C, right panel).These data indicate BcpI proteins mediate immunity to BcpA-CT intracellular toxicity in an allelespecific manner.

Lack of evidence for BcpAIOB-mediated competition in liquid medium
Previous work with E. coli demonstrated that E. coli-type CDI systems function in interbacterial competition in liquid medium when the cdiBAI genes are expressed from a constitutive or inducible promoter [1,4].Additionally, it was demonstrated that expression of the cognate immunity gene in target bacteria was protective against CDI [4].To determine if the bcpAIOB genes also mediate interbacterial competition and if bcpI provides protection, similar competition assays were performed.Wild type inhibitor (E264Cm R ) and DbcpAIOB mutant target bacteria were mixed at a 1:1 ratio and cultured in LSLB broth at 37uC for 24 hours without antibiotic selection.Wild type bacteria were also cultured at a 1:1 ratio with DbcpAIOB bacteria constitutively expressing the cognate immunity gene (E264DbcpAIOB::bcpI E264 ) under the same conditions.After 24 hours, the cultures were serially diluted and plated on selective media to determine the number of cfu of each strain.In the competition between wild type and DbcpAIOB mutant bacteria, both strains grew equally (Figure 7A), and therefore no competitive advantage was observed for the wild type strain in this assay.Both strains also grew equally in the competition between wild type and DbcpAIOB::bcpI E264 bacteria (Figure 7B).Since our expression data indicated that B. thailandensis bcpAIOB genes are expressed in only about one in one thousand wild type bacterial cells cultured in liquid medium (Figure 4), we hypothesized two non-mutually exclusive reasons for the lack of apparent competition between wild type bacteria and the DbcpAIOB mutant.1) In addition to inhibiting DbcpAIOB mutant bacteria, the wild type bacteria expressing their bcpAIOB genes might also inhibit the growth of wild type bacteria not expressing their bcpAIOB genes because these bacteria would not be producing BcpI, and hence both wild type and DbcpAIOB bacteria would be inhibited to the same extent.2) Regardless of the susceptibility of wild type bacteria to growth inhibition by other wild type bacteria, the number of DbcpAIOB mutant bacteria inhibited by wild type bacteria may be insignificant because of the low number of wild type bacteria expressing their bcpAIOB genes.
To test the first hypothesis, we constructed a strain constitutively expressing the cognate bcpI gene (from B. thailandensis E264) in a wild type E264 background (E264::bcpI E264 ).Constitutive expression of bcpI E264 in the wild type strain did not alter the results of the competition assay; both strains again grew equally (Figure 7C), suggesting that wild type bacteria expressing bcpAIOB were not inhibiting wild type bacteria that were not expressing bcpAIOB to an appreciable level.To test the second hypothesis, we performed a competition experiment with the strain expressing bcpAIOB from the P S12 promoter (P S12 -bcpAIOB) and DbcpAIOB bacteria.Again, both strains grew equally (Figure 7D), suggesting expression of bcpAIOB in every wild type bacterium does not lead to growth inhibition of mutant target bacteria in liquid medium.Collectively, these data indicate that expression of bcpAIOB from either the native promoter or a constitutively active promoter (in single copy on the chromosome) is not sufficient to cause interbacterial competition against DbcpAIOB target bacteria when mixed at a 1:1 ratio in liquid medium.
The bcpAIOB-encoded CDI system mediates interbacterial competition on solid medium Initial characterization of BcpAIOB-mediated CDI on solid medium.We set out to determine if the bcpAIOB gene products affected growth and interbacterial competition on solid medium.To begin, overnight cultures of wild type and DbcpAIOB strains were diluted to OD 600 = 0.2 and 20 ml of culture was spotted onto LSLB agar to initiate colony biofilm formation and observed for four days.Competition colony biofilms (strains mixed at a 1:1 ratio) were plated at the same dilution and also observed for four days (Figure 8A).The initial boundaries of the colony biofilms were apparent by day 1.Between two and four days, the bacteria began to migrate outward, giving rise to a leading edge that was distinct from the initial boundary, thereby increasing the diameter of the colony biofilm.The wild type and DbcpAIOB mutant colony biofilms exhibited distinct morphological properties after four days.However, both strains grew past the initial boundary and migrated equal distances after four days, from ,9 mm to 11 mm, indicating that B. thailandensis is capable of moving across a solid surface and that the bcpAIOB genes are not required for this phenotype.Each colony biofilm formed from a heterogeneous population (the competition colony biofilms) also had a distinct morphology, but again the bacteria were able to migrate past the initial boundary.
Bacteria were picked from the center of the competition colony biofilms on days 1, 2, and 4 post inoculation using a sterile pipette tip, suspended in PBS, diluted, and aliquots were plated on LSLB agar containing Cm to select for wild type bacteria and LSLB agar containing Km to select for DbcpAIOB mutant bacteria, in order to determine the competitive index of CDI + inhibitors compared to CDI 2 targets.Wild type (E264Cm R ) bacteria outcompeted the DbcpAIOB mutant by approximately 2.5 logs at one day and this increased to ,3 logs at four days (Figure 8B, left panel).Constitutive expression of the cognate bcpI gene in the DbcpAIOB mutant strain (E264DbcpAIOB::bcpI E264 ) provided protection, as this strain competed equally with the wild type strain at one day and was only slightly outcompeted at two and four days (Figure 8B, middle panel).Constitutive expression of a heterologous bcpI gene (from B. pseudomallei K92643) however, did not provide protection to the DbcpAIOB mutant strain (E264DbcpAIOB::bcpI K92643 ), which was outcompeted by approximately 2.5 logs at one day and ,3 logs at four days (Figure 8B, right panel).Together, these data show that the bcpAIOB chromosomally-encoded CDI system in B. thailandensis mediates interbacterial competition on a solid agar surface and that immunity to interbacterial CDI is allele-specific.wild type and DbcpAIOB bacteria (Figure 8C, left panel) or wild type and DbcpAIOB::bcpI K92643 bacteria (Figure 8C, right panel) at any time point (the competitive index is therefore greater than or equal to 4.4), suggesting CDI-mediated competition is very strong at the leading edge.Constitutive expression of the cognate bcpI gene in DbcpAIOB mutant bacteria (E264DbcpAIOB::bcpI E264 ) abrogated competition, as this strain was present in equal numbers to wild type bacteria on each day tested (Figure 8C, middle panel).Collectively, these data indicate that the majority of BcpAIOBmediated competition occurs prior to day 1, and suggest competition is more robust at the leading edge than in the center of the colony biofilm.
Analysis of early time points of BcpAIOB-mediated CDI.To gain a better understanding of BcpAIOB-mediated CDI, we conducted a 24-hour time course competition between wild type and DbcpAIOB bacteria.Competition was apparent at six hours at the edge of the colony biofilm, and by 12 hours, no DbcpAIOB bacteria were recovered in about half of the samples (the competitive index is therefore greater than or equal to the represented value, red data points) (Figure 9A, bottom panel).By contrast, in the center of the colony biofilm, competition was not apparent until 12 hours, and DbcpAIOB bacteria were still recovered at 24 hours, albeit as a very low proportion of the population (Figure 9A, top panel).These results indicate that the effects of CDI can be observed earlier at the edge of the colony biofilm than in the center.
To investigate the difference in competition at the center and edge of the colony biofilm in more detail, we visualized the colony biofilms by live-image microscopy.One hour after plating the bacteria, the initial edge contained densely packed bacteria, whereas the center of the colony biofilm contained sporadically distributed bacteria, most of which were not in contact with any other bacteria (Figure S4).By 12 hours, bacteria in the center of the colony biofilm had divided to the point that most were in contact with other bacteria (Figure S4).The difference in density of bacteria over time within the colony biofilm therefore, may explain the difference in competition observed between the center and the edge.
To distinguish wild type and mutant bacteria within the colony biofilms, we labeled wild type bacteria with RFP (E264::P S12 -rfp) and DbcpAIOB bacteria with GFP (E264DbcpAIOB::P S12 -gfp) and visualized the colony biofilms by confocal microscopy.(Note that after addition of a cover slip to the top of the colony biofilm, the original architecture was disrupted, and bacteria were moderately displaced (i.e.spread out) from their original location.For authentic architecture of the colony biofilms, refer to Figure S4.)After one hour of co-incubation on agar, the bacteria were coccishaped (Figure 9B, first row).(Again, note that bacteria from the edge of the colony biofilms particularly at one hour were spread out from their original location when visualized for this assay.Prior to manipulation, these bacteria were in close association with one another.)Strikingly, after six hours on a solid surface, the morphology of the bacteria changed from cocci to more rodshaped (Figure 9B, second row), and by 12 hours, the bacteria were long rods (Figure 9B, third row).These data suggest the bacteria undergo a change in gene expression and cell morphology in response to either their change in environment upon plating on solid medium or due to the change from stationary phase growth (in liquid medium prior to plating) to log phase on solid medium.Interestingly, some DbcpAIOB::P S12 -gfp bacteria at the edge at six hours still appear to be cocci-shaped, suggesting they were previously growth inhibited by wild type bacteria and therefore unable to transition to rod-shaped, whereas this was not observed in colony biofilms containing only wild type bacteria (Figure S5).
With respect to the proportion of wild type (RFP + ) and DbcpAIOB (GFP + ) bacteria, both were present in equal numbers in the center and edge of the colony biofilm at one hour (Figure 9B, first row).At six hours, RFP + and GFP + bacteria were again detected in equal numbers in the center of the colony biofilm, while a greater proportion of wild type (RFP + ) bacteria than DbcpAIOB (GFP + ) bacteria were present along the edge (Figure 9B, second row).By 12 hours, the proportion of RFP + bacteria had increased in both the center and edge of the colony biofilm (Figure 9B, third row), and after 24 hours of co-incubation, only RFP + bacteria were detected along the edge, while some GFP + bacteria could still be detected in some samples from the center of the colony biofilm, a representative of which is shown in Figure 9B.As a control, we mixed wild type strains expressing rfp or gfp (E264::P S12 -rfp and E264::P S12 -gfp) at a 1:1 ratio, plated them on agar, and looked microscopically at 1, 6, 12, and 24 hours.For all time points, RFP + and GFP + bacteria were present in equal numbers (Figure S5).These data support the competition data presented in Figure 9A indicating that competition can be observed at earlier time points and to a greater extent at the edge of the colony biofilm compared to the center.
Together, these data show that wild type B. thailandensis is able to outcompete otherwise isogenic DbcpAIOB mutant bacteria via CDI when mixed on an agar surface in the center of, and to an even greater extent, at the initial boundary of a colony biofilm within 24 hours.Immunity to interbacterial CDI is allele-specific, as only expression of the cognate immunity gene and not a heterologous immunity gene provided protection to target bacteria.Bacteria in the center of the colony biofilm were not subject to CDI initially because they were not in direct contact with other bacteria.Upon sufficient growth to allow direct interbacterial interactions, some mutant bacteria however, were still surrounded only by DbcpAIOB bacteria, and therefore competition was not ''complete'' in the center of the colony biofilm.By contrast, at the initial edge of the colony biofilm, wild type and mutant bacteria were in direct contact immediately upon plating, and nearly every mutant bacterium was in contact with a wild type bacterium.Competition in this context took four to six hours to become apparent, suggesting a change in gene expression must occur to initiate CDI, and was ''complete'' by 12-24 hours.
Investigation of expression of the bcpAIOB genes on solid medium.When cultured in liquid medium, bcpAIOB gene expression was detected in only approximately one in one thousand bacteria (Figure 4).We hypothesized that expression occurred in a much greater proportion of the population on solid medium, given the level of competition that was observed in the colony biofilms (Figures 8 & 9).To investigate the expression of the bcpAIOB genes in bacteria in a colony biofilm, we performed microscopy and flow cytometry using our P bcpA -gfp strain as a reporter.Bacteria were scraped from the center and edge of colony biofilms on solid agar at 2, 4, 6, 8, 12, 24, and 48 hours post-plating, suspended in PBS, and analyzed.At no time point were any GFP + bacteria detected by either microscopy or flow cytometry (data not shown).However, our data indicate that BcpAIOB-mediated CDI had occurred under these conditions (Figures 8 & 9) and therefore the bcpAIOB genes must have been expressed.These data suggest therefore, that expression from the P bcpA promoter in bacteria grown on a solid surface is either too weak or too transient to be detected by our assays.
Because we could not detect bcpAIOB expression in cells during growth in colony biofilms, we developed an alternative method to determine the proportion of BcpAIOB + cells required to outcompete BcpAIOB 2 cells on a solid surface.We performed competition experiments with our strain in which the bcpAIOB genes were expressed constitutively (P S12 -bcpAIOB) mixed at a 1:1,000 ratio with DbcpAIOB bacteria to simulate expression similar to that in liquid medium.The competitive index did not change in either the center or the edge of the colony biofilm after 24 hours (Figure 10), indicating no competition occurred.When we mixed P S12 -bcpAIOB and DbcpAIOB bacteria at a 1:1 ratio to simulate 100% of the ''wild type'' population activating the bcpAIOB genes, P S12 -bcpAIOB bacteria outcompeted DbcpAIOB mutant bacteria by nearly 2 logs in the center of the colony biofilm and completely outcompeted the mutant along the leading edge after 24 hours (Figure 10), similar to competition between wild type and DbcpAIOB bacteria (Table 2).Collectively, these results strongly suggest that expression of the bcpAIOB genes occurs in a greater proportion of cells than one in one thousand, if not 100% of the population, on solid medium -indicating a change in gene expression in response to the solid surface colony biofilm environment.2).To determine if the small competitive advantage displayed by the DbcpO strain was in fact due to CDI, we competed DbcpO bacteria with DbcpAIOB mutant bacteria constitutively expressing the cognate bcpI gene (E264DbcpAIOB::bcpI E264 ) or a heterologous bcpI gene (E264DbcpAIOB::bcpI K96243 ).The competitive index for competition between the DbcpO strain and E264DbcpAIOB::bc-pI E264 was zero in the center and along the leading edge of the colony biofilm at 24 hours (Figure 11, column II), indicating constitutive expression of bcpI E264 in DbcpAIOB mutant bacteria is protective against CDI by the DbcpO strain.However, constitutive expression of bcpI K96243 in DbcpAIOB mutant bacteria was not protective, as DbcpO bacteria outcompeted E264DbcpAIOB::bc-pI K96243 bacteria by ,1 log in the center and ,2 logs along the leading edge of the colony biofilm at 24 hours (Figure 11, column III).Complemention of the DbcpO strain with a copy of bcpO expressed constitutively (E264DbcpO::bcpO) only slightly increased CDI activity in the center of the colony biofilm but fully restored activity at the edge, i.e.DbcpAIOB bacteria were completely outcompeted in this location (Figure 11, column IV, Table 2).Lack of restoration of autoaggregation and partial restoration of CDI by the bcpO complementation strain underscores the complexity of the bcpAIOB system and the role of BcpO in interbacterial CDI.
Together, our data indicate that the DbcpO strain has a greater than ten-fold defect in CDI-mediated interbacterial competition in the center of the colony biofilm and at least a thousand-fold defect along the leading edge compared to wild type bacteria after 24 hours (Table 2).BcpO, therefore, plays a substantial and critical role in CDI-mediated interbacterial competition in B. thailandensis.

Discussion
CDI has so far been demonstrated only in E. coli, the species in which it was discovered, and Dickeya dadantii, a phytopathogenic bacterium that infects a variety of crop plants [1,4].Although genetic loci in six Burkholderia strains were predicted to encode CDI systems based on the presence of small (,300 bp) ORFs immediately 39 to genes predicted to encode TpsA proteins, the gene order in these loci was different than that of the E. coli cdiAIB genes (and all other putative CDI system-encoding loci) and the motif separating the conserved and variable regions of the predicted TpsA protein was NxxLYN rather than VENN [4].We showed in this study that the Burkholderia bcpAIOB genes do in fact encode proteins that function in many ways like the CDI system of E. coli; the BcpA-CTs are toxic when produced intracellularly, the bcpI genes confer immunity in an allele-specific manner, and wild type B. thailandensis can outcompete a DbcpAIOB mutant if the mutant does not express the cognate immunity gene.B. thailandensis is therefore the third species in which CDI has been demonstrated.Moreover, together with the previous observations, our results indicate that Burkholderia bcpAIOB genes define novel classes of both CDI and TPS systems.We also showed in this study  that the Burkholderia bcpAIOB genes are required for autoaggregation and adherence to an abiotic surface, and that they are expressed in a probabilistic manner when the bacteria are cultured in liquid medium, two phenotypes not previously ascribed to CDI or CDI system-encoding genes.
The TPS pathway is one of the simplest mechanisms for the secretion of proteins to the surface of Gram-negative bacteria.The paradigm, based primarily on studies of the FHA/FhaC proteins of Bordetella species and the HMW1/HMW1B proteins of Haemophilus influenzae, states that the large b-helical exoprotein (the TpsA family member) is translocated across the cytoplasmic membrane by the general Sec pathway and then requires only one protein, the TpsB family member, for translocation across the outer membrane [2,3].Experimental support for the sufficiency of the TpsB protein in outer membrane translocation of the TpsA protein was recently obtained in a study using purified FhaC, liposomes, and a polypeptide corresponding to the N-terminal 370 aa of FHA [21].Our bioinformatic analysis identified a small ORF located 59 to bcpB in most Burkholderia-type putative CDI protein-encoding loci, which we named bcpO.BcpO of B. thailandensis E264 is predicted to be a lipoprotein that localizes to the inner leaflet of the outer membrane and deletion of bcpO resulted in loss of autoaggregation (identical to deletion of the entire bcpAIOB operon) and significantly reduced CDI activity.BcpA appeared to be produced and exported across the outer membrane in the bcpO mutant, based on the fact that BcpA was not degraded, as assessed by immunoblot, and was capable of mediating a very low level of CDI.Unfortunately, our attempts to visualize BcpA on the surface of wild type and DbcpO mutant bacteria were unsuccessful (data not shown).Based on the phenotypes of the bcpO mutant and the predicted cellular location of BcpO, we hypothesize that BcpO is involved in efficient, export across the outer membrane, maturation of BcpA into a functional protein, release of BcpA from the cell surface, and/or sensing interbacterial interactions.Regardless of its role, our data indicate that the Burkholderia BcpAIOB proteins define a novel class of TPS system that requires an additional small protein, BcpO, to produce a fully functional TpsA protein.
TpsB proteins are members of the Omp85-TpsB superfamily, which includes BamA (the main component of the Bam complex that inserts b-barrel proteins into the outer membranes of Gramnegative bacteria), Tob55/Sam50 (which inserts proteins into the outer membranes of mitochondria), and Toc75 (which inserts proteins into the outer membranes of chloroplasts) [22].In addition to BamA, the Bam complex contains four lipoproteins, BamB, C, D, and E, that localize to the inner leaflet of the outer membrane and play important but mostly unknown roles in outer membrane protein assembly [22].The BcpAIOB system of B. thailandensis E264 may function in an analogous manner to the Bam complex, as it appears to require at least one predicted periplasmic lipoprotein for proper secretion or maturation of its substrate.Curiously, some classes of predicted CDI systems in Burkholderia spp contain BcpO proteins that do not have signal sequences and that vary in an allele-specific manner with their cognate BcpA and BcpI proteins.Whether these proteins function similarly to BcpO of B. thailandensis E264 or perform completely different functions is unknown.These systems may represent yet additional variation of the TPS and CDI paradigms.
Our P bcpA -gfp studies showed that when B. thailandensis is cultured in liquid medium (either LSLB or M63), expression of bcpAIOB is high in approximately 0.2% of the bacteria and undetectable in the rest.The only other strain for which expression of CDI systemencoding genes has been demonstrated is E. coli EC93, which, in contrast to other strains that have been investigated, appears to express the cdiBAI genes constitutively [4].Our result suggests that, in liquid medium under the conditions tested, bcpAIOB gene expression is controlled in a probabilistic manner.Several cellular differentiation processes in organisms ranging from bacteria to humans are controlled probabilistically and transiently [23,24,25], one of the best understood being the development of competence in Bacillus subtilis.When starved for nutrients, a minority of B. subtilis cells in a population express genes required for DNA uptake (competence), while the rest commit to sporulation [26].Regulation of competence genes in B. subtilis involves an excitable core module containing both positive and negative feedback loops [27].We expect that equally complex regulatory circuits control bcpAIOB expression in Burkholderia, and our future experiments will be aimed at identifying and characterizing the systems involved.
When cultured in M63 minimal medium, wild type B. thailandensis aggregated and adhered to the walls of glass test tubes.This phenotype required BcpAIOB as the DbcpAIOB, DbcpO, and DbcpB mutants grew planktonically.This is the first demonstration of a phenotype for any CDI 2 strain other than susceptibility to growth inhibition by CDI + counterpart bacteria.Although we did not formally test for biofilm formation, Schwarz et al. showed that B. thailandensis forms a biofilm in a flow chamber model [28].Our data suggest that the BcpAIOB proteins contribute to biofilm formation and therefore their true role in nature may not be (just) mediating interbacterial competition.Interestingly, the conditions that promote B. thailandensis autoaggregation are the same conditions in which expression of the bcpAIOB genes occurs in only about 0.2% of the bacterial cells.Treating cultures with DNaseI abrogated autoaggregation, suggesting that DNA may be an essential component of an extracellular matrix that holds aggregated cells together.This observation suggested the intriguing hypothesis that CDI contributes to biofilm formation by killing a small proportion of cells in the population such that their released DNA can be used for extracellular matrix formation.However, the strain expressing bcpAIOB from the strong, constitutive P S12 promoter also aggregated, albeit with different kinetics.In this population, all cells should produce BcpI and therefore should be immune to CDI, although we cannot rule out the possibility that strong, constitutive expression of the bcpAIOB locus alters the functionality of the individual gene products.Moreover, we detected extracellular DNA in cultures of both wild type and DbcpAIOB bacteria, indicating that extracellular DNA is required but not sufficient for autoaggregation.These observations underscore the complexity of the aggregation/biofilm phenotype and indicate that determining the role of BcpAIOB in this microbial lifestyle will require an extensive amount of additional investigation.
By contrast to what was observed for B. thailandensis cultured in liquid, BcpAIOB-dependent interbacterial competition was easily observed when the bacteria were grown on a solid surface.Within the colony biofilm, wild type bacteria outcompeted DbcpAIOB mutants by ,2.5 logs in the center and completely ($4.4logs) at the edge after 24 hours of co-incubation.Examination of earlier time points indicated that competition occurred as early as 12 hours in the center and six hours at the edges.Live-cell imaging provided an explanation for the difference in competition at the two sites; bacteria in the center of the colony biofilm were just beginning to contact each other between 6 and 12 hours, while bacteria at the edges were in contact immediately upon plating.The efficiency of the competition when bacteria were in contact however, suggested that the bcpAIOB genes must be expressed in more than ,0.2% of wild type bacteria under these conditions.The fact that bacteria transitioned from coccobacilli to long rods after being moved from stationary phase growth in broth to the solid surface provided evidence for a change in gene expression, but we were unsuccessful in our attempts to detect GFP + bacteria when the P bcpA -gfp strain was used to form colony biofilms.However, the strain expressing bcpAIOB from the P S12 promoter was able to outcompete the DbcpAIOB strain when the colony biofilms were initiated with the bacteria present at a 1:1 ratio but not when present at a 1:1,000 ratio.Together, these data suggest that expression of the bcpAIOB operon is induced in a majority of the bacteria within a few hours after plating on agar.The competitive index did not increase between 24 and 96 hours in the center of the colony biofilm where there were still DbcpAIOB mutants present, however, suggesting that the early activation of bcpAIOB gene expression was only transient.Moreover, the fact that streaking the strain containing a P bcpA -lacZ fusion resulted in a heterogeneous population of dark and light blue colonies indicates that the signal that induces bcpAIOB gene expression is not solely the solid surface environment.We are currently investigating the possibility that bcpAIOB gene expression is induced in response to both environmental cues and recognition of neighboring bacteria, including sensing whether those neighboring bacteria have BcpA proteins on their surface.
By contrast with our observations with B. thailandensis, CDI in E. coli occurs in liquid medium [1].In strain EC93, the rat fecal isolate in which CDI was discovered, the cdiBAI genes are expressed constitutively [1].In human uropathogenic E. coli strain 536, cdiBAI expression was not detected under standard laboratory growth conditions [4].To measure CDI activity in 536, or in laboratory K12 strains of E. coli, therefore, cdiBAI genes were expressed from an inducible promoter on a high copy number plasmid [1,4].Moreover, it was shown that production of capsule or P or S pili in the target bacteria blocked CDI [1].These caveats raised questions about the biological relevance of CDI as observed under these conditions.In light of our findings, the possibilities that cdiBAI genes in E. coli and other bacteria are expressed in a probabilistic and transient manner and that they contribute to aggregation and/or biofilm formation are compelling hypotheses to test.
Another apparent difference between Burkholderia-type CDI systems and E. coli-type CDI systems is the presence of one or more ''orphan cdiA-CT/cdiI modules'' located 39 to cdiBAI genes [5].Comparative genome analyses suggest that these modules may move within or between bacteria into the functional cdiBAI locus, changing the allele that is expressed, and hence they may contribute to the diversity of cdiBAI alleles within a population and even within a clonal population [5].We searched specifically for bcpA-CT/bcpI modules in bacteria containing Burkholderia-type CDI systems but found no evidence for their existence.Diversity of CDI systems amongst Burkholderia must, therefore, occur by a different mechanism.
Our experiments, aimed at providing an initial characterization of the function of the bcpAIOB gene products and determining their contribution to aggregation and interbacterial competition, used wild type and mutant derivatives of B. thailandensis strain E264.As environmental saprotrophs, however, Burkholderia species share their habitats with a plethora of other bacteria, as well as viruses and eukaryotes.How CDI is used in this environment and whether it occurs only in an intra-species manner or also in an inter-species manner is not known.Our bioinformatic analysis identified 58 bcpAI(O)B loci and 41 different alleles.Some strains have multiple alleles and several alleles are present in multiple strains.Notably, one allele that is present in two different B. pseudomallei strains (1106A-1 and BCC215-2) is also present in B. gladioli (strain BSR3-1), suggesting that BcpAIOB-mediated inter-species CDI does occur.It is also notable that many, if not all, bcpAI(O)B genes are located on genomic islands.Several Burkholderia species have been shown to be naturally competent [29] and there is tremendous genomic diversity amongst Burkholderia strains, mediated in part by horizontal gene transfer [30].These observations raise several questions.For example, is there a hierarchy of potency amongst the various BcpA proteins?What happens if bacteria with different bcpAIOB alleles come into contact?Or if a bacterium containing one allele encounters a bacterium containing two alleles?Is it advantageous for a bacterium to contain multiple alleles?If so, why is the greatest number of alleles identified in a single strain only three?Is there a cost associated with bcpAIOB alleles that limits the number that can be tolerated within a single cell?The question of how diversity amongst bcpAIOB alleles is generated is a perplexing one as nucleotide changes in the region encoding BcpA-CT can presumably be tolerated only if compensatory changes occur in bcpI, and vice versa.Finally, the most important question and one related to CDI in general but perhaps the most difficult to address experimentally: What is the true role of CDI in nature?Is it used for competitive exclusion or for the acquisition of DNA that can be used for nutrition, as a biofilm matrix, or a source of genetic diversity?Perhaps it serves all of these purposes -or none and plays a role that we cannot yet imagine.Because life in a community is the rule rather than the exception for most bacteria, understanding how and why CDI systems function will be broadly relevant.
While this manuscript was in revision, Nikolakakis et al. published a related characterization of CDI systems in Burkholderia [31].Although these authors used different experimental approaches, their results are consistent with ours and support the conclusion that bcpAIOB genes encode CDI systems that can mediate interbacterial competition and that function in an allelespecific manner.Moreover, Nikolakakis et al. showed that the Cterminal domains of some BcpA proteins possess tRNase activity, similar to what has been demonstrated for the C-terminal domains of CdiA proteins in E. coli type CDI systems [31].

Strain construction and culture conditions
Burkholderia thailandensis E264 is an environmental isolate [14].Plasmids were maintained in E. coli DH5a and DH5alpir and mated into B. thailandensis using the donor E. coli strain RHO3 [32].
The B. thailandensis DbcpO and DbcpB strains were created by allelic exchange.Deletion constructs were created by PCR amplifying (PFU Ultra polymerase, Agilent) ,500 nucleotides 59 of bcpO and bcpB (including the first three codons of the gene) and the ,500 nucleotides 39 of bcpO and bcpB (including the last three codons of the gene) from E264 genomic DNA.DNA fragments were restriction digested and cloned into allelic exchange vector pSM112, which carries a pheS counter-selectable marker.The resulting plasmids, pMA11 and pMA12, were used to create the DbcpO and DbcpB strains, respectively.
The B. thailandensis DbcpAIOB strain was created by natural transformation as described [29].Briefly, ,800 nucleotides 59 of bcpA (including the first three codons of the gene) and ,800 nucleotides 39 of bcpB (including the last three codons of the gene) were amplified from genomic E264 DNA and the gene encoding kanamycin resistance (plus ,500 nucleotides to include the promoter) was amplified (PFU Ultra polymerase, Agilent) from pUC18miniTn7(Km).Overlap PCR was performed to construct a single DNA product in which the kanamycin resistance-encoding gene was flanked by the regions 59 to bcpA and 39 to bcpB, which was subsequently transformed into B. thailandensis.
The constitutively active strain (P S12 -bcpAIOB) was constructed as follows.Approximately 200 nucleotides 59 to the ribosomal S12 protein-encoding gene (containing the promoter, P S12 ) and ,750 nucleotides 39 of the bcpA translation start site were PCR amplified (PFU Ultra polymerase, Agilent) from genomic E264 DNA, joined by overlap PCR, and cloned into pEXKm5 [32] to create pECG22, which was then used for cointegration into the chromosome of E264.
B. thailandensis E264Cm R and E264Km R (microscopy and flow cytometery vector control) were constructed using pUC18mi-niTn7(Cm) and pUC18miniTn7(Km) to insert a gene encoding chloramphenicol or kanamycin resistance, respectively, on the chromosome at the attTn7 site.
All lacZ and gfp reporter strains were created using the Tn7 transposase method as well.Briefly, ,500 nucleotides 59 to bcpA (P bcpA ), ,200 nucleotides corresponding to the predicted S12 promoter (P S12 ), or no DNA (P neg ) were cloned 59 to a promoterless lacZ gene and ,500 nucleotides 59 to bcpA (P bcpA ) was cloned 59 to a promoterless gfp gene and cloned into pUC18miniTn7(Km).Transposition into the chromosome of B. thailandensis at the attTn7 site generated the P bcpA -lacZ, P S12 -lacZ, P neg -lacZ, and P bcpA -gfp strains.The miniTn7-kan-gfp plasmid [34] was used to generate the B. thailandensis P S12 -gfp reporter strain in a similar manner.
All strains constructed were verified by PCR and sequencing analysis (Eton BioScience).

Bioinformatics
Burkholderia-type CDI-encoding loci were identified by using the predicted BcpB protein sequence from B. pseudomallei K96243 as a pBLAST query against the entire NCBI genome database.DNA sequences encoding predicted TPS system-like proteins were analyzed in Vector NTI Advance 11.A locus was considered to encode a Burkholderia-type CDI system if it was composed of a large ORF (,3000 codons) followed immediately 39 by a small ORF (,100 codons), both of which were 59 to the tpsB homolog.Alignments were performed in Vector NTI and then analyzed in Jalview 2.7 with Taylor residue coloring.

Reverse transcriptase PCR
Total RNA was extracted from bacterial cells cultured overnight in LSLB broth using TRIzol Reagent (Invitrogen) according to the manufacturer's protocol.RNA was treated with 4 U of DNaseI (Ambion) for 30 min at 37uC.2-5 mg of RNA was subsequently used for cDNA synthesis using SuperScript III First-Strand (Invitrogen) according to the manufacturer's protocol.RT-PCR was performed using GoTaq polymerase (Promega); PCR was also performed on E264 genomic DNA with identical primer sets as a control.PCR products were analyzed by 0.8% agarose gel electrophoresis and stained with ethidium bromide for visualization.

b-galactosidase assay
Reporter strains were cultured overnight in LSLB at 37uC with or without aeration, at room temperature (,25uC) with aeration, or in M63 or M9 minimal medium at 37uC with aeration.P neg -lacZ and P S12 -lacZ were cultured in LSLB at 37uC with aeration.Heat shocked samples were obtained by incubating 500 ml aliquots of overnight LSLB (37uC, aerated) culture at 42uC for 5 min.Cultures were normalized to OD 600 ,0.2-0.5 and b-galactosidase activity was measured as described [35].Two independent assays were performed in triplicate.

Autoaggregation
Overnight cultures of strains in LSLB medium were diluted to OD 600 = 0.2 into 2 ml of M63 minimal medium in glass tubes.Bacteria were cultured with aeration for ,48 h at 37uC with or without 4 or 10 U DNaseI with 106 DNaseI buffer (Ambion).

Intracellular toxicity
The last ,350 codons of bcpA (including the Nx(E/Q)LYN motif) from B. pseudomallei strains K96243 and 1106A-2 were PCR amplified (with an added ATG at the 59 end) from genomic DNA and cloned into pSCrhaB2 [36].Plasmids carrying bcpI were cloned in a similar manner except the trimethoprim resistance gene of pSCrhaB2 was replaced with a kanamycin resistance gene to allow for selection of these plasmids in combination with the plasmids encoding BcpA-CTs.B. thailandensis containing the BcpA-CT-encoding plasmids alone or in combination with the BcpIencoding plasmids were cultured overnight in LSLB supplemented with trimethoprim (strains with BcpA-CT-encoding plasmids) or trimethoprim and kanamycin (strains with BcpA-CT-and BcpIencoding plasmids) and 0.2% glucose.Cultures were washed and diluted in fresh LSLB medium to OD 600 0.2-0.6 supplemented with antibiotics and 0.2% glucose or 0.2% rhamnose.Bacteria were cultured with aeration for 4 h at 37uC.Aliquots were taken at 0 h and 4 h, diluted in PBS, and plated on LSLB with antibiotic selection and 0.2% glucose to determine the cfu/ml of each strain.Two independent experiments were performed in triplicate.

Microscopy
Confocal microscopy-Liquid cultured bacteria preparation; bacteria were cultured overnight in LSLB broth and M63 minimal medium and concentrated to OD 600 ,50 in PBS.Solid medium cultured bacteria preparation; bacteria in colony biofilms on agar were cut out from petri dishes (agar included) and placed on a glass slide.Cover slips were added to the top of the colony biofilms.Bacteria were imaged on a Ziess LSM5 Pascal Confocal Laser Scanning microscope, 636 objective with oil immersion.Macroscope-Colony biofilms were imaged with a Leica M420 macroscope, 0.56 objective, at 1, 2, and 4 days post inoculation.Live-imaging microscopy-500 ml of LSLB agar was added to cover glass bottom dishes, Delta T, (Biotechs) and allowed to solidify.Bacteria were plated onto the agar and imaged on an Olympus IX81 inverted microscope at 206 objective.

Flow cytometry
Bacteria were cultured overnight in LSLB broth and M63 minimal medium supplemented with kanamycin and diluted to OD 600 ,0.1 in PBS.Bacterial samples were analyzed on a Gallios Flow Cytometer (Beckman Coulter); bacteria were gated based on forward scatter and side scatter (,1 um diameter and distinct from particles in the PBS only control (Figure S3)) for subsequent analysis.Data were analyzed with Kaluza 1.1 software.

Competitions
Liquid-bacteria were cultured overnight, washed with PBS, and diluted to OD 600 0.2 in fresh LSLB medium.Strains were mixed at a 1:1 ratio (final volume of 2 ml) without antibiotic selection and cultured with aeration for 24 h at 37uC.Aliquots were taken at 0 h and 24 h, diluted in PBS, and plated on LSLB with antibiotic selection to determine the cfu/ml of each strain in the competition.Two independent experiments were performed in triplicate.
Solid-bacteria were cultured overnight, washed with PBS, and diluted to OD 600 0.2 in LSLB.Strains were mixed at a 1:1 ratio and 20 ml of culture was plated on solid LSLB (1.5% agar) without antibiotic selection.The culture inoculum was plated on LSLB with antibiotic selection to determine the ratio at 0 hours.Agar plates were stored at room temperature (,25uC) for the duration of the experiment.Bacteria were picked from the colony biofilms with a sterile pipette tip, diluted in PBS, and plated on LSLB with antibiotic selection to determine the cfu of each strain in the competition at a particular dilution.The competitive index for each day was calculated as the ratio of wild type bacteria to mutant bacteria at time X hours divided by the ratio of wild type to mutant at time 0 hours.Two to three independent experiments were performed in triplicate.

Figure 1 .
Figure 1.Organization and diversity of Burkholderia-type CDI-encoding loci.A) Schematic of Burkholderia-and E. coli-type CDI-encoding loci.Burkholderia-type CDI loci are encoded by genes bcpAIOB and differ from E. coli-type CDI loci in gene order and content; Burkholderia-type CDI loci contain an additional ORF, bcpO, not present in E. coli-type CDI loci.B) Schematic of the amino acid sequence of a single representative allele for each of the B. thailandensis and B. pseudomallei CDI-encoding loci.The alleles have strong sequence similarity (,89% similar) amongst the N-terminal encoded ,2750 aa of BcpA, BcpB, and some BcpO proteins, but fall into three phylogenetic groups (indicated by shades of gray).The diversity of the CDI-encoding proteins is located within the ,350 aa C-terminal to the Nx(E/Q)LYN motif in BcpA, BcpI, and some BcpO proteins, with no more than ,10% similarity between any two alleles (indicated by different colors).For some BcpO proteins, the diversity is restricted to the predicted signal sequence peptide (alleles 1-5, 7, and 8) whereas the functional domain of the protein is highly conserved.doi:10.1371/journal.pgen.1002877.g001

Figure 2 .
Figure2.Examination of the bcpAIOB locus in B. thailandensis.A) Top panel, analysis of the operon structure of bcpAIOB.Primer sets 1, 2, and 3 flanking the intergenic region of bcpA and bcpI, bcpI and bcpO, and bcpO and bcpB, respectively, (as indicated in 2C) were used for RT-PCR.Middle and bottom panels, analysis of the approximate start site of transcription of bcpA.RT-PCR was performed using primers annealing 50, 70, 120, 150, 200, 250, and 300 nt 59 to bcpA with a reverse primer internal to bcpA.All products were visualized by ethidium bromide.B) Expression of bcpA.P S12 -lacZ and P neg -lacZ (dark gray bars) and P bcpA -lacZ (light gray bars) reporter strains were cultured under various conditions and assayed for b-galactosidase activity.Error bars represent the mean 6 1 SEM.C) Schematic of B. thailandensis E264 strain constructs, including WT, DbcpO, DbcpB, and DbcpAIOB, used throughout the study.doi:10.1371/journal.pgen.1002877.g002

Figure 4 .
Figure 4. Per cell analysis of the probabilistic expression of bcpA.A) P bcpA -lacZ reporter strain plated on M63 minimal medium agar supplemented with X-gal.B) Confocal microscopy of gfp reporter strains, P bcpA -gfp, P S12 -gfp, and E264Km R (vector), cultured overnight in liquid medium.C) Representative flow cytometry analysis of gfp reporter strains, P bcpA -gfp, P S12 -gfp, and E264Km R (vector), cultured overnight in liquid medium.The dot plot (top left panel) of side scatter intensity (SS INT) vs. forward scatter intensity (FS INT) shows the events corresponding to B. thailandensis that were gated on (''Gate'') for subsequent analysis.Histograms show the number of GFP-positive events (Count) vs. relative fluorescence intensity (FL1 INT) for each reporter strain, and the percent of GFP-positive bacteria detected in each culture is indicated.doi:10.1371/journal.pgen.1002877.g004

Figure 5 .
Figure 5. bcpAIOB-dependent autoaggregation in M63 minimal medium.A) Indicated strains were cultured in M63 minimal medium for ,48 hours with aeration.Four or 10 U of DNaseI was added to wild type E264 cultures upon inoculation.B) Wild type and P S12 -bcpAIOB strains were cultured in M63 minimal medium; photos were taken over the course of 48 hours.doi:10.1371/journal.pgen.1002877.g005

Figure 8 .
Figure 8. bcpAIOB-mediated contact dependent growth inhibition on solid medium.A) Colony biofilms of indicated strains and competitions were observed for four days by microscopy.B) Samples from the center of each colony biofilm competition were taken on each day and plated with antibiotic selection to determine the competitive index (C.I.) of wild type E264Cm R (WT) bacteria compared to DbcpAIOB (left panel), DbcpAIOB::bcpI E264 (middle panel), and DbcpAIOB::bcpI K96243 (right panel).C) Samples from a single location along the leading edge of each colony biofilm (as indicated by the red asterisks(*)) were plated as in B. In cases where only wild type bacteria were recovered, the actual C.I. is greater than or equal to the value stated.doi:10.1371/journal.pgen.1002877.g008

Figure 9 .
Figure 9. Analysis of 24-hour time course of CDI-mediated competition.A) Wild type E264Cm R (WT) bacteria and DbcpAIOB mutant bacteria were co-incubated on solid medium.Samples were

Figure 11 .
Figure 11.DbcpO-mediated contact dependent growth inhibition.Cultures of bacteria were mixed at a 1:1 ratio and colony biofilms were plated on solid LSLB agar.Samples were taken from the center (top panels) and leading edge (bottom panels) of the colony biofilms at day 1 and plated with antibiotic selection to determine the C.I. of DbcpO compared to DbcpAIOB (column I), DbcpO::bcpO E264 compared to DbcpAIOB (column II), DbcpO compared to DbcpAIOB::bcpI E264 (column III), and DbcpO compared to DbcpAIOB::bcpI K96243 (column IV).Red data points indicate only wild type bacteria were recovered, and the actual C.I. is therefore greater than or equal to the represented value.doi:10.1371/journal.pgen.1002877.g011