Burkholderia BcpA mediates biofilm formation independently of interbacterial contact-dependent growth inhibition


For correspondence. E-mail peggy_cotter@med.unc.edu; Tel. (+1) 919 966 2627; Fax (+1) 805 893 4724.


Contact-dependent growth inhibition (CDI) is a phenomenon in which Gram-negative bacteria use the toxic C-terminus of a large surface-exposed exoprotein to inhibit the growth of susceptible bacteria upon cell–cell contact. Little is known about when and where bacteria express the genes encoding CDI system proteins and how these systems contribute to the survival of bacteria in their natural niche. Here we establish that, in addition to mediating interbacterial competition, the Burkholderia thailandensis CDI system exoprotein BcpA is required for biofilm development. We also provide evidence that the catalytic activity of BcpA and extracellular DNA are required for the characteristic biofilm pillars to form. We show using a bcpAgfp fusion that within the biofilm, expression of the CDI system-encoding genes is below the limit of detection for the majority of bacteria and only a subset of cells express the genes strongly at any given time. Analysis of a strain constitutively expressing the genes indicates that native expression is critical for biofilm architecture. Although CDI systems have so far only been demonstrated to be involved in interbacterial competition, constitutive production of the system's immunity protein in the entire bacterial population did not alter biofilm formation, indicating a CDI-independent role for BcpA in this process. We propose, therefore, that bacteria may use CDI proteins in cooperative behaviours, like building biofilm communities, and in competitive behaviours that prevent non-self bacteria from entering the community.


Within both eukaryotic hosts and environmental niches, bacteria communicate, cooperate, and compete with other microorganisms. Contact-dependent growth inhibition (CDI) has been characterized primarily in Escherichia coli as an interbacterial competition phenomenon. Mediated by Two Partner Secretion (TPS) system proteins, CDI is thought to occur when the toxic C-terminal ∼300 residues of the large (> 3000 aa) exoprotein (called CdiA in E. coli, CdiA-CT for the C-terminal ∼300 aa) is delivered to the cytoplasm of a susceptible target bacterium (Aoki et al., 2005; 2010; Webb et al., 2013). Bacteria are protected from CDI if they produce a small immunity protein (called CdiI in E. coli).

Characterization of TPS family proteins in Burkholderia pseudomallei led to the discovery that genes encoding CDI systems are present in a broad range of proteobacteria and that the C-terminal regions of the large exoproteins and the immunity proteins are highly variable (Aoki et al., 2010). CDI system-encoding genes fall into two main categories: E. coli-type, which have the gene order cdiBAI, and Burkholderia-type, which have the locus structure bcpAIOB. The first CdiA protein examined (in E. coli EC93) affects the target cell membrane (Aoki et al., 2009), resulting in growth inhibition rather than cell death (Aoki et al., 2005). All other CdiA-CTs examined so far display nuclease activity in vitro, cleaving tRNA or DNA substrates (Aoki et al., 2010; Morse et al., 2012; Nikolakakis et al., 2012), likely killing the target cells. Consistent with this hypothesis, cells targeted by a CdiA-CT DNase have been shown to lose DAPI staining (Webb et al., 2013). Immunity proteins bind specifically to their cognate CdiA-CTs or BcpA-CTs (Aoki et al., 2010; Morse et al., 2012; Nikolakakis et al., 2012) and, when produced intracellularly, protect a bacterium from CdiA-CT- or BcpA-CT-mediated toxicity (Aoki et al., 2010; Anderson et al., 2012; Morse et al., 2012; Nikolakakis et al., 2012).

Although CDI has been clearly demonstrated in vitro, whether interbacterial competition is the main or only function performed by CDI systems in nature is unknown. In the E. coli strain in which CDI was first described, the cdiBAI genes are expressed constitutively (Aoki et al., 2005). In all other strains examined, within and outside the Escherichia genus, cdiBAI gene expression appears to be tightly regulated and, in many cases, genetic manipulation to constitutively express cdiBAI genes is required for CDI to occur (Aoki et al., 2010). Furthermore, CDI has only been observed between wild-type bacteria (or bacteria constitutively expressing the cdiBAI or bcpAIOB genes from a multicopy plasmid) and mutants lacking the cdiBAI or bcpAIOB genes (Aoki et al., 2010; Anderson et al., 2012; Nikolakakis et al., 2012), a situation that is unlikely to occur in nature. Moreover, while mutation of the cdi or bcp locus in Dickeya (Aoki et al., 2010) or Burkholderia (Anderson et al., 2012; Nikolakakis et al., 2012), respectively, is sufficient to render these bacteria susceptible to CDI in vitro, the lack of the conserved surface structures capsule (Aoki et al., 2008) or P pili (Aoki et al., 2005) on target cells is also required for efficient CDI in E. coli. Thus, the biological relevance of CDI remains unclear.

Bacteria often grow as biofilms, communities of microbes enclosed within an extracellular polymeric matrix (Davey and O'Toole, 2000; Vlamakis et al., 2013). Biofilms can form on a wide variety of surfaces in diverse environments, including plant rhizospheres, indwelling medical devices and mammalian mucosal surfaces and can exacerbate human disease. Especially within environmental niches, biofilms are frequently polymicrobial, although the mechanisms utilized by bacteria to cooperate and/or compete in these complex sociomicrobiological communities are not well understood.

Here, using Burkholderia thailandensis as a model, we demonstrate that the CDI system-encoding genes bcpAIOB are essential for the development of multicellular biofilm communities. We additionally show that although the activity of BcpA-CT is required for this process, the role of BcpA in biofilm formation is independent of interbacterial growth inhibition. These data suggest a model in which the Burkholderia CDI proteins function to mediate both competitive and cooperative bacterial behaviours.


B. thailandensis biofilm formation requires bcpAIOB

Previous work from our laboratory demonstrated that the CDI system-encoding genes in B. thailandensis, bcpAIOB, are required for autoaggregation during culture in minimal medium, suggesting a role for these genes in biofilm formation (Anderson et al., 2012). To investigate this possibility, we used confocal laser scanning microscopy (CLSM) to measure B. thailandensis E264 static, submerged biofilm formation on glass coverslips (Fig. 1A). Six hours after inoculation, wild-type bacteria, a mutant lacking the entire bcpAIOB locus (ΔbcpAIOB), and a mutant lacking the bcpB gene that encodes the outer membrane channel for surface localization of BcpA (ΔbcpB) adhered similarly to the coverslip in small aggregates. The mutants were defective, however, in several subsequent steps of biofilm development, resulting in significantly less biomass than biofilms formed by wild-type bacteria (Fig. 1A and B). The mutants formed slightly sparser monolayers at 16 h and their biofilms at 24 h contained fewer microcolonies than those formed by the wild-type strain. The microcolonies that the mutants did develop were an average of 10 μm shorter than those formed by wild-type bacteria. Furthermore, while wild-type microcolonies developed into large pillar structures extending over 35 ± 2 μm above the coverslip by 72 h, the sparse ΔbcpAIOB microcolonies attained a maximum height of only 23 ± 2 μm.

Figure 1.

Biofilm development by B. thailandensisbcpAIOB mutants.

A. CLSM of biofilms formed by wild-type B. thailandensis (top row), ΔbcpAIOB mutant bacteria (middle row) and ΔbcpB mutant bacteria (bottom row) at 6, 16, 24 and 72 h post inoculation. All strains carry a gfp gene expressed constitutively from PS12 located at one of the two attTn7 sites in the B. thailandensis genome. In Z stack renderings (24 h and 72 h), cross-sections through the plane parallel to the coverslip (shown by large centre image) were set at 6 μm above the substratum for 24 h and 10 μm for 72 h. Scale bars represent 10 μm.

B. For each of the strains in (A), COMSTAT analysis of Z stacks collected after 24 (white bars) or 72 h (grey bars) of biofilm development was used to determine biofilm total biomass (left panel), average thickness (middle panel) and maximum thickness (right panel). Bars represent the mean of at least three independent experiments and error bars indicate the SEM. Significance was determined by two-tailed t-tests, comparing each mutant to the wild-type strain at the indicated time point. *P < 0.05; **P < 0.001.

Extracellular DNA contributes to B. thailandensis biofilms

For many bacterial species, extracellular DNA (eDNA) is an essential component of the biofilm matrix (Whitchurch et al., 2002; Lappann et al., 2010; Seper et al., 2011). To test the role of eDNA in B. thailandensis biofilms, DNase I was added during biofilm inoculation. After 6 h of growth, no difference was observed between DNase-treated and control biofilms (Fig. S1), indicating that DNA is not required for the initial attachment step. However, after 16 h of growth, monolayers of both wild-type and ΔbcpAIOB mutant bacteria in DNase-treated biofilms contained fewer substrate-associated cells than in samples treated with buffer alone (Fig. 2A), demonstrating that eDNA is required for this process. While necessary for biofilm formation, however, addition of excess eDNA was not sufficient to rescue the biofilm defect of ΔbcpAIOB mutant bacteria (Fig. 2B). Interestingly, however, while the addition of chromosomal DNA generally had no effect on wild-type biofilms (Fig. 2B), it occasionally appeared to enhance biofilm formation (Fig. S1).

Figure 2.

Contribution of extracellular DNA to B. thailandensis biofilms.

A. CLSM of 16 h biofilms formed by wild-type (top) and ΔbcpAIOB mutant (bottom) bacteria that were co-inoculated with 20 U DNase I (right) or buffer alone (left).

B. CLSM of 24 h biofilms formed by wild-type (top) and ΔbcpAIOB mutant (bottom) bacteria that were co-inoculated with 3 μg purified B. thailandensis chromosomal DNA (right) or buffer alone (left). Cross-sections through the plane parallel to the coverslip (shown by large centre image) were set at 6 μm above the substratum. The wild-type strain carries a gfp gene expressed constitutively from PS12 at the attTn7 site of B. thailandensis chromosome 2 and the ΔbcpAIOB mutant carries the same construct at the attTn7 site on chromosome 1.

Scale bars represent 10 μm.

Native expression of bcpAIOB is necessary for biofilm architecture

Our previous studies demonstrated that when B. thailandensis was cultured in liquid medium, bcpAIOB gene expression was stochastic, with only approximately one in 1000 bacteria expressing a bcpA–gfp fusion at a high level at a given time (Anderson et al., 2012). When grown on a solid surface, such as in the colony biofilm competition assay, bcpA–gfp was not expressed highly enough to see GFP+ bacteria by microscopy, but our data indicated that most, if not all, bacteria had expressed the bcpAIOB genes highly enough to mediate interbacterial competition (Anderson et al., 2012). To investigate bcpAIOB expression in a biofilm, we constructed a strain containing a PbcpA–gfp reporter and a constitutive rfp gene, each in single copy at one of the two attTn7 sites in the B. thailandensis genome. Examination of this strain indicated that, similar to what was observed during liquid culture (Anderson et al., 2012), only a small subset of bacteria expressed the bcpAIOB genes to high levels at a given time within the biofilm (Fig. 3A). In addition to individual GFP+ bacteria within the monolayer of cells adhering to the coverslip, 72 h biofilms also rarely contained entire pillar structures composed almost exclusively of GFP+ bacteria. However, the majority of pillars contained bacteria that were not highly expressing bcpAIOB (i.e. GFP), and the GFP+ pillars were never observed at earlier time points. Strong constitutive expression of the bcpAIOB genes from the rpsL promoter, PS12, resulted in an abnormally dense, flat biofilm that lacked discrete pillar structures (Fig. 3A). This was particularly evident at 72 h, when biofilms formed by wild-type bacteria were significantly thicker than those made by the bcpC strain (Fig. 3B). Furthermore, when biofilm heterogeneity was quantified, 72 h biofilms produced by the bcpC mutant were significantly (P < 0.0001) more uniform in thickness (roughness coefficient of 0.14 ± 0.01) than the highly textured biofilms formed by wild-type B. thailandensis (roughness coefficient of 0.36 ± 0.03). These data indicate that native expression of bcpAIOB, which appears to include strong expression by a small subpopulation of cells, is critical for biofilm formation.

Figure 3.

bcpAIOB gene expression during B. thailandensis biofilm formation and effect of constitutive expression on biofilm architecture.

A. CLSM of biofilms formed by a B. thailandensis fluorescent reporter strain carrying a gfp gene driven by PbcpA located at the attTn7 site on B. thailandensis chromosome 2 and an rfp gene expressed constitutively from PS12 at the attTn7 site on B. thailandensis chromosome 1 (top row), and mutant bacteria constitutively expressing bcpAIOB (bcpC) (bottom row). In Z stack renderings (24 h and 72 h), cross-sections through the plane parallel to the coverslip (shown by large centre image) were set at 6 μm above the substratum for 24 h and 10 μm for 72 h. Scale bars represent 10 μm.

B. For each of the strains in (A), COMSTAT analysis of Z stacks collected after 24 (white bars) or 72 h (grey bars) of biofilm development was used to determine biofilm total biomass (left panel), average thickness (middle panel) and maximum thickness (right panel). Wild-type data from Fig. 1 are re-plotted here for comparison. Bars represent the mean of at least three independent experiments and error bars indicate the SEM. Significance was determined by two-tailed t-tests, comparing each mutant to the wild-type strain at the indicated time point. *P < 0.05; **P < 0.001.

BcpA shows homology to Holliday junction resolvases and conserved residues are required for interbacterial competition

Other TPS system exoproteins, including FHA of Bordetella species, have been shown to mediate autoaggregation (Menozzi et al., 1994) and biofilm formation (Irie et al., 2004; Serra et al., 2011) via intermolecular interactions that lead to interbacterial adhesion. To assess a putative role for interbacterial growth inhibition during biofilm formation and/or a potential role for BcpA functioning as an interbacterial adhesin, we constructed a strain producing BcpA protein with single amino acid substitutions that rendered it defective for CDI. A 50-amino-acid region within the C-terminal 130 residues of BcpA is predicted by PHYRE2 protein fold recognition (Kelley and Sternberg, 2009) analysis to share structural similarity with archaeal Holliday junction resolvases and endonucleases. Protein sequence alignment of the putative structurally conserved region with the PHYRE2 hits and their homologues showed striking conservation of several amino acids. Interestingly, three of these residues (corresponding to D3051, E3064 and K3066 in BcpA) have been shown to be required for DNA substrate cleavage by the Pyrococcus furiousus Holliday junction resolvase, Hjc (Komori et al., 2000) (Fig. 4A).

Figure 4.

Identification of BcpAE264 putative catalytic residues E3064 and K3066 and their role in interbacterial competition.

A. Alignment of BcpA residues 3018–3068 (top line) with PHYRE2 secondary structure prediction hits (bolded) and their homologues. See Experimental procedures for accession numbers. Highlighted residues are conserved in at least 5/10 sequences and asterisks indicate amino acids (E3064 and K3066) mutated in strain BtEKA.

B. Anti-HA Western blot of whole-cell lysates prepared from wild-type and BtEKA bacteria constitutively producing HA-tagged BcpA (BcpAWT-HA and BcpAEKA-HA, corresponding to strains E264BcpA-HA::pECG22 and BtEKABcpA-HA::pECG22 respectively) or wild-type bacteria constitutively producing untagged BcpA [WT (no HA), corresponding to strain bcpC]. HA-encoding sequences were inserted after the F2633 codon of bcpA on the chromosomes of wild-type and BtEKA bacteria.

C. Dot blot of the strains in (B) that have been modified to also contain phoA driven by PS12 at one of the two attTn7 sites in the B. thailandensis genome. Boiled (+) and unboiled (−) cells were probed with anti-HA and anti-PhoA antibodies to assess surface exposure of BcpA-HA.

D. Competitive indices (CI) (inhibitor:target) after 24 h of interbacterial competition. Inhibitor bacteria [WT(Cm) (E264CmR) or BtEKA(Cm)] were mixed at a 1:1 ratio with the indicated target bacteria [ΔbcpAIOB or BtEKA(Km)] and spotted onto agar. After 24 h of incubation, samples from the centre of the colony spot were collected and plated with antibiotic selection to determine the cfu of each strain and calculate CI (shown as log CI). Data represent two independent experiments, each performed in triplicate.

To determine whether these conserved amino acids are important for BcpA function, we mutated two of the corresponding codons to alanine codons on the B. thailandensis chromosome, generating a strain (BtEKA) with E3064A and K3066A substitutions in BcpA. To measure the BcpA protein produced by this strain, nucleotides encoding an HA epitope were introduced into bcpA (immediately 3′ to the F2633 codon) in both the wild-type and BtEKA strains (resulting in E264BcpA-HA and BtEKABcpA-HA respectively). Because bcpAIOB expression in vitro is insufficient to detect BcpA-HA protein production (Anderson et al., 2012), the locus's native promoter was replaced with the strong constitutive promoter PS12 in the HA-tagged strains. Western blot analysis of cell lysates prepared from these strains showed that BtEKABcpA-HA produced total BcpA-HA similar to the wild-type (E264BcpA-HA) strain (Fig. 4B). To examine BcpA-HA localization, the HA-tagged strains were further modified to carry a copy of the E. coli phoA gene at a neutral chromosomal site. Whole-cell dot blots indicated that both E264BcpA-HA::phoA and BtEKABcpA-HA::phoA produced similar levels of surface-exposed BcpA-HA (Fig. 4C), indicating that the E3064A and K3066A substitutions do not affect BcpA protein production or localization.

It is well established that BcpA-CT (and CdiA-CT) proteins bind to their cognate BcpI (or CdiI) proteins in vitro (Aoki et al., 2010; Nikolakakis et al., 2012). To determine whether BcpA-CTEKA retained the ability to bind BcpI, His-tagged BcpA-CT proteins from wild-type or BtEKA bacteria were co-produced in E. coli with untagged BcpI. In pull-down purification experiments, native, but not denatured BcpI bound to His-BcpA-CT proteins from both wild-type and BtEKA bacteria (Fig. S2), suggesting that the mutant BcpA protein folds properly.

To measure the competitive fitness of the BtEKA mutant, the untagged BtEKA strain was tested in interbacterial competition assays. While wild-type bacteria outcompeted a ΔbcpAIOB mutant more than 100-fold, the BtEKA mutant failed to outcompete the CDI-susceptible strain (Fig. 4D), demonstrating the requirement of E3064 and K3066 for CDI. Because bcpI is unaltered in this strain, however, BtEKA bacteria remain protected from interbacterial killing (Fig. 4D).

E3064 and K3066 are conserved in other BcpA/CdiA proteins

Interestingly, the residues required for BcpAE264 interbacterial killing (E3064 and K3066) appear to be conserved in other putative CDI system exoproteins (Fig. S3). This E/DxK motif is found in approximately the same location (> 100 amino acids from the C-terminus) in exoproteins of both Burkholderia- and E. coli-type CDI systems, including those found in Burkholderia multivorans, Acinetobacter baumannii and Neisseria meningitidis. While several of the proteins containing the E/DxK motif appear to be full-length BcpA/CdiA proteins (e.g. sequences from Burkholderia dolosa and N. meningitidis alpha14), others seem to be orphan BcpA-CTs/CdiA-CTs (e.g. sequences from N. meningitidis H44/76 and G2136). These truncated bcpA/cdiA genes are encoded downstream of an intact locus, have been shown in E. coli to mediate interbacterial killing when fused to the N-terminal portion of CdiA, and are thought to represent a pool of toxin diversity (Poole et al., 2011). Other E/DxK-containing proteins appear to be orphan BcpA-CTs/CdiA-CTs unassociated with an intact bcpAIOB/cdiBAI locus and one example (the protein from Solitalea canadensis) seems to be a putative Rhs protein. Analogous to CDI system proteins, Rhs proteins have recently been shown to mediate interbacterial competition (Poole et al., 2011; Koskiniemi et al., 2013). Although the sequences of the entire variable C-terminal domains of all these proteins (BcpA-CTs/CdiA-CTs), as well as their associated immunity proteins, do not share striking similarity, localized homology exists in the region surrounding the putative catalytic region, suggesting that this predicted activity of BcpA-CTE264 may be conserved in other systems.

Biofilm development requires BcpA activity

To define the role of BcpA-CT activity in B. thailandensis biofilm development, we tested the ability of the BtEKA mutant to form a biofilm. Like ΔbcpAIOB mutant bacteria, the BtEKA mutant adhered to the glass coverslip similarly to wild-type bacteria, but was defective for biofilm development, forming a thin biofilm with sparse microcolonies (Fig. 5A) that reached a maximum height of only 22 μm by 72 h (Fig. 5B). Thus, activity of BcpA, not simply surface-localized protein utilized for interbacterial adhesion, is required for biofilm development. BcpAIOB-dependent autoaggregation of B. thailandensis (Anderson et al., 2012) also appears to require BcpA activity, as BtEKA mutant bacteria did not aggregate when cultured in minimal medium (data not shown).

Figure 5.

Biofilm development by the CDI-inactive BtEKA mutant and effect of constitutive bcpI expression on biofilm formation.

A. CLSM of biofilms formed by BtEKA mutant bacteria (top row) and wild-type bacteria constitutively expressing bcpI (WT::bcpIC) (bottom row) at 6, 16, 24 and 72 h post inoculation. Strains carry a gfp gene expressed constitutively from PS12 located at the attTn7 site on B. thailandensis chromosome 1. In Z stack renderings (24 h and 72 h), cross-sections through the plane parallel to the coverslip (shown by large centre image) were set at 6 μm above the substratum for 24 h and 10 μm for 72 h. Scale bars represent 10 μm.

B. For each of the strains in (A), COMSTAT analysis of Z stacks collected after 24 (white bars) or 72 h (grey bars) of biofilm development was used to determine biofilm total biomass (left panel), average thickness (middle panel) and maximum thickness (right panel). Wild-type data from Fig. 1 are re-plotted here for comparison. Bars represent the mean of at least three independent experiments and error bars indicate the SEM. Significance was determined by two-tailed t-tests, comparing each mutant to the wild-type strain at the indicated time point. *P < 0.05; **P < 0.001.

Role for BcpA activity is CDI-independent

Although it has not been demonstrated that BcpA-mediated CDI results in death of target bacteria, our data suggested a model in which biofilm development requires that some bcpAIOB-expressing bacteria in the community kill other bacteria that express bcpAIOB at a much lower level (or not at all), causing them to release a key nutrient, signalling molecule, or matrix component (perhaps eDNA) that is required for the biofilm to form. This model predicts that constitutive expression of the immunity-encoding gene, bcpI, in the entire population will prevent BcpA-mediated toxicity and, therefore, biofilm formation. Surprisingly, although bcpI protects from interbacterial inhibition in a competition assay (Anderson et al., 2012), its constitutive expression in wild-type B. thailandensis only slightly decreased monolayer formation (16 h) and had no effect on biofilm pillar development (Fig. 5A), biomass or thickness (Fig. 5B). These results indicate that biofilm formation does not require interbacterial killing or growth inhibition, a finding that is inconsistent with the paradigm that CDI systems are dedicated to interbacterial competition.

Wild-type bacteria cannot rescue bcpAIOB mutants in mixed strain biofilms

As an alternative approach to determine the role of BcpA activity during biofilm formation, we inoculated biofilm chambers with 1:1 mixtures of RFP- and GFP-labelled bacteria. Mixtures of differentially labelled wild-type B. thailandensis demonstrated that biofilm pillar structures are predominantly clonal, as solely GFP- or RFP-positive cells composed most large microcolonies (Fig. 6A). Twenty-four per cent of pillars appeared sectored, suggesting that they formed from the merging of neighbouring GFP+ and RFP+ populations. Planktonic bacteria entering pillar structures was apparently an infrequent event, as only 24% of pillars appeared to contain a mixture of GFP- and RFP-labelled cells. While some ‘mixed’ structures were indeed nearly homogenous mixtures of GFP- and RFP-labelled bacteria, many were predominantly composed of bacteria with one fluorescent label, containing oppositely labelled cells attached to the pillar surface.

Figure 6.

Analysis of biofilms formed by mixtures of wild-type B. thailandensis and bcpAIOB mutants. RFP-producing wild-type B. thailandensis was mixed and inoculated at a 1:1 ratio with GFP-producing (A) wild-type bacteria, (B) ΔbcpAIOB mutant (Δbcp), (C) ΔbcpB mutant, (D) BtEKA mutant, (E) ΔbcpAIOB mutant constitutively expressing bcpIbcp::bcpIC) or (F) wild-type bacteria constitutively expressing bcpI (WT::bcpIC). Images represent 48 h biofilms of each mixture. Each graph depicts the per cent of microcolonies observed that were composed of only RFP+ bacteria (WT, red bars), only GFP+ bacteria (mutant, green bars), half RFP+ and half GFP+ bacteria (sectored, grey bars), or containing a mixture RFP+ and GFP+ bacteria (mixed, yellow bars). It should be noted that most ‘mixed’ microcolonies were predominantly RFP+ with fewer GFP+ cells adhered to their surfaces. Microcolony composition was scored at 24 h in at least 10 randomly selected microscope fields in each of three independent experiments. Bars represent the mean and error bars show the standard error of the mean. Significance was determined by two-tailed t-test, comparing each proportion of GFP+ microcolonies (mutant, green bars) to the proportion of RFP+ structures (WT, red bars). *P < 0.05; ns, not significant.

The presence of wild-type bacteria in a biofilm was not sufficient to rescue the biofilm defects of ΔbcpAIOB (Fig. 6B), ΔbcpB (Fig. 6C) or BtEKA (Fig. 6D) mutant bacteria, supporting the conclusions that the purpose of BcpAIOB is not simply to release a factor from a subset of cells that is required for biofilm formation and that to participate in the biofilm, each bacterium must produce an active BcpA protein. As with monoculture biofilms, constitutive expression of bcpI did not alter mixed strain biofilms (Fig. 6E and F).

CDI occurs in the biofilm monolayer, but does not impact pillar development by ΔbcpAIOB mutant bacteria

While it appears to have no effect on the ability of ΔbcpAIOB mutant cells to develop pillar structures, CDI does occur in the submerged biofilm environment at later time points. At 72 h, mixtures containing RFP-labelled wild-type bacteria co-inoculated with GFP-labelled wild-type cells (Fig. 7A) or wild-type bacteria expressing constitutive bcpI (Fig. 7C) appeared to contain approximately 50% of each strain, as observed after 24 h of growth (Fig. 6A). Unlike earlier time points (Fig. 6B and E), while mixtures of wild-type B. thailandensis and ΔbcpAIOB mutant bacteria again contained fewer ΔbcpAIOB mutant cells (Fig. 7B), constitutive bcpI expression in ΔbcpAIOB allowed more mutant bacteria to persist with wild-type cells on the substratum (Fig. 7D). This result indicates that ΔbcpAIOB mutants were eliminated from the monolayer by wild-type bacteria via CDI. However, while it could persist in the monolayer, BcpI production did not allow the ΔbcpAIOB mutant to grow up into pillar structures that were composed solely of ΔbcpAIOB bacteria nor did it substantially affect the ability of ΔbcpAIOB bacteria to coexist in wild-type pillars. Similar results were seen with 72 h mixed biofilms consisting of wild-type and BtEKA mutant bacteria (data not shown), further supporting the conclusion that each biofilm pillar participant requires active BcpA on its surface. Additionally, as these experiments demonstrate that BcpA-CT molecules must be delivered into target cells within mixed strain biofilms, they suggest that delivery of BcpA-CT to a target cell, either producing (Figs 6C–F and 7D) or not producing (Figs 6B and 7B) BcpI, does not function to signal the recipient cell to develop a biofilm pillar structure. Together, these data indicate a distinct role for BcpA catalytic activity that is independent of both interbacterial growth inhibition and the presence of intracellular BcpI, implying that this activity occurs extracellularly.

Figure 7.

Interbacterial competition within B. thailandensis biofilms. RFP-marked wild-type B. thailandensis was mixed and inoculated at a 1:1 ratio with GFP-marked (A) wild-type bacteria, (B) ΔbcpAIOB mutant, (C) wild-type bacteria constitutively expressing bcpI (WT::bcpIC), or (D) ΔbcpAIOB mutant constitutively expressing bcpIbcpAIOB::bcpIC) and imaged at 72 h. For each mixture, the left image depicts the biofilm monolayer and the right image shows the Z stack rendering. Scale bars represent 10 μm.


The current paradigm is that CDI systems mediate competition between bacterial cells. Although CDI system-dependent interbacterial competition clearly occurs, it has thus far only been observed between wild-type inhibitor cells (or cells that overproduce the CDI system proteins) and mutant target cells (CDI and sometimes capsule pili) (Aoki et al., 2005; 2010; Anderson et al., 2012; Nikolakakis et al., 2012). While CdiA-CT and BcpA-CT domains are toxic to a range of bacterial species when produced intracellularly (Aoki et al., 2010; Anderson et al., 2012; Nikolakakis et al., 2012) or delivered on chimeric CdiA or BcpA proteins (Aoki et al., 2010; Nikolakakis et al., 2012), interspecies or even inter-strain competition between wild-type strains, as is expected to occur in nature, has not yet been observed, raising the question of the biological role of CDI. Here we demonstrate that the BcpAIOB proteins, and specifically BcpA activity, are required for biofilm formation, functioning within a genetically identical population to facilitate sociomicrobiological community development.

Although BcpA activity was indispensible for biofilm development, suggesting that CDI was also required, expression of the immunity-encoding bcpI gene had no effect on biofilm growth. As BcpI production protects target B. thailandensis bacteria from CDI-meditated toxicity (Anderson et al., 2012), we must conclude that the role for BcpA activity during biofilm formation is CDI-independent. This surprising result implies that BcpA facilitates CDI and biofilm construction through distinct mechanisms. In contrast, the fact that the BtEKA mutant was defective for both CDI and biofilm formation suggests that, while distinct in their requirement for BcpI-mediated immunity, these two mechanisms require similar enzymatic activities and/or substrates. Resolving the apparently intertwined roles for BcpAIOB in biofilm development and CDI is our next goal.

While there are examples of regulated cell death and eDNA release affecting biofilm development in some bacterial species (Rice et al., 2007; Lopez et al., 2009; Thomas et al., 2009), in others eDNA release during biofilm formation is not dependent on cell lysis (Barnes et al., 2012). Our data provide evidence that CDI system protein biofilm-promoting activity in B. thailandensis is independent of interbacterial killing. Combined with our finding that transcomplementation of ΔbcpAIOB mutant biofilms by wild-type B. thailandensis does not occur, indicating that each biofilm participant must produce active BcpA on its surface, this result implies that the activity of BcpA during biofilm formation occurs extracellularly. The similarity of BcpAE264-CT to Holliday junction resolvases and its possible DNA nickase activity (Nikolakakis et al., 2012) suggest the hypothesis that BcpAE264 and related proteins may perform an enzymatic function on eDNA within the biofilm matrix, perhaps cross-linking it or covalently anchoring bacteria to the eDNA network. In Staphylococcus aureus, there is evidence for a mechanism of this nature. Beta toxin, a secreted protein that contributes to both virulence and biofilm formation, was shown to form cross-links in the presence of DNA in vitro, forming a nucleoprotein complex independent of its well-characterized sphingomyelinase activity (Huseby et al., 2010). However, although CDI systems are widespread among Gram-negative bacteria (Aoki et al., 2010), whether this putative biofilm mechanism is common to all CDI systems may be unlikely, as it is difficult to imagine how the BcpA/CdiA proteins that have been shown to have tRNase activity (Aoki et al., 2010; Morse et al., 2012; Nikolakakis et al., 2012) could function in this way.

The failure of the strain constitutively expressing the bcpAIOB genes to form pillar structures demonstrates that native bcpAIOB gene expression, perhaps involving turning bcpAIOB both on and off at appropriate stages, is necessary for proper biofilm development. Within biofilms, however, the majority of cells did not express the bcpA–gfp fusion strongly enough to be detected by fluorescence microscopy. However, our data from biofilms co-inoculated with wild-type and mutant strains suggest that each bacterium must produce active BcpA to participate in the biofilm structure. Thus, it seems that, similar to our previous findings on bcpAIOB expression in colony biofilms (Anderson et al., 2012), bcpAIOB expression in the majority of biofilm-associated cells is too low or too transient for bcpA–gfp to be detected by microscopy, but is sufficient to mediate biofilm development. In addition to this low-level expression, in both liquid culture (Anderson et al., 2012) and biofilms, a small subpopulation of bacteria expresses bcpAIOB very strongly, suggestive of a bistable regulation phenomenon. The finding that occasionally all B. thailandensis cells within a pillar structure appear to express bcpAIOB simultaneously suggests that these bacteria employ a sensing/signalling mechanism to co-ordinate BcpAIOB production. Quorum sensing, which has been demonstrated to affect B. thailandensis autoaggregation (Chandler et al., 2009) and is well-established to contribute to biofilm development in other species (Davies et al., 1998; Hammer and Bassler, 2003), provides one possible mechanism. Interestingly, these bcpAIOB-expressing pillar structures were only observed at later time points (72 h), suggesting that this expression pattern may play a role at a later stage of biofilm development or maturation, such as dispersal.

We show here that the putative catalytic region of BcpAE264, including the residues required for CDI and biofilm formation (E3064 and K3066), is conserved in other BcpA and CdiA proteins, suggesting that the biofilm-promoting activity of BcpA may be conserved. Moreover, cdi loci in N. meningitidis (Neil and Apicella, 2009), Xanthomonas axonopodis (Gottig et al., 2009) and Dickeya dadantii (Rojas et al., 2002) have been found previously to function in biofilm formation and autoaggregation, although they were not identified as CDI system-encoding genes at the time. Similarly, a TPS exoprotein with characteristics of CdiA proteins was found to be the fifth most abundant protein in the extracellular matrix of Pseudomonas aeruginosa colony biofilms (Toyofuku et al., 2012). These observations support the hypothesis that contribution to biofilm development is a common function for at least a subset of both Burkholderia-type and E. coli-type CDI systems.

Paradoxically, while CDI systems have so far been described in terms of their ability to mediate interbacterial competition, our data indicate that these proteins can facilitate a cooperative bacterial behaviour. In nature, it is possible that both of these seemingly conflicting activities occur, perhaps under different environmental conditions. Indeed, while not required for biofilm development, our findings indicate that CDI does occur within the biofilm monolayer. The hallmark diversity of CdiA-CT/BcpA-CT proteins (Aoki et al., 2010; Anderson et al., 2012; Nikolakakis et al., 2012) and the observation that B. pseudomallei bcpAIOB genes expressed in B. thailandensis can inhibit B. thailandensis (Nikolakakis et al., 2012) support the hypothesis that interspecies CDI occurs in the environment. An intriguing possibility is that CDI systems function both to build biofilms, and to exclude non-self competitors from the community.

Experimental procedures

Bacterial strains and culture conditions

Burkholderia thailandensis E264 is an environmental isolate (Brett et al., 1998). All plasmids were maintained in E. coli DH5α and DH5αλpir and mated into E264 using E. coli donor strain RHO3 (López et al., 2009). Unless otherwise stated, bacteria were cultured overnight with aeration at 37°C in low salt Luria broth (LSLB, 0.5% NaCl) supplemented with 250 μg ml−1 (for B. thailandensis) or 50 μg ml−1 (for E. coli) kanamycin, 100 μg ml−1 ampicillin, 20 μg ml−1 tetracycline or 200 μg ml−1 diaminopimelic acid as appropriate. M63 minimal medium [110 mM KH2PO4, 200 mM K2HPO4, 75 mM (NH4)2SO4, 16 nM FeSO4] supplemented with 1 mM MgSO4, 0.2% glucose and 0.4% glycerol (Thongdee et al., 2008) was used where indicated. All constructed plasmids and bacterial strains were confirmed by DNA sequence analysis.

Burkholderia thailandensis strains E264CmR, ΔbcpAIOB, ΔbcpB, ΔbcpAIOB::bcpI, WT::bcpI and bcpC (PS12-bcpAIOB) were constructed previously (Anderson et al., 2012). BtEKA was constructed by allelic exchange. A DNA fragment corresponding to ∼700 bp of internal bcpA sequence containing three nucleotide changes (A9191C, A9196G, A9197C) to generate the E3064A and K3066A substitutions, as well as a single nucleotide change (T9186A) to introduce a translationally silent SnaBI restriction site, was generated by overlap PCR and cloned into pEXKm5, which was used for allelic exchange as described (López et al., 2009). For competition assays, a gene encoding kanamycin or chloramphenicol resistance was delivered to BtEKA at one of the strain's two attTn7 sites using pUC18T-miniTn7T-Kan (Choi et al., 2008), generating the strains BtEKA(Km) and BtEKA(Cm) respectively.

Strains for biofilm assays were marked with green or red fluorescent protein using the miniTn7kan–gfp and miniTn7kan–rfp plasmids (Norris et al., 2010) to deliver a constitutive (expression driven by PS12) gfp or rfp gene to one of the two attTn7 sites in the B. thailandensis genome (one site on each of two chromosomes). The PbcpA–gfp reporter strain was constructed previously (Anderson et al., 2012) and contains ∼500 bp immediately upstream of the bcpA start codon cloned 5′ to a promoterless gfp and delivered to the attTn7 site on B. thailandensis chromosome 2. The kan cassette was removed from this strain by Flp-FRT recombination as described (Choi et al., 2008) using pFlpTet, a pFlpe4-derived plasmid modified to carry tetracycline, rather than kanamycin, resistance. The resulting kanamycin-sensitive PbcpA–gfp reporter was marked at the attTn7 site on B. thailandensis chromosome 1 with constitutive rfp, as described above. The kan cassettes were also similarly excised from the wild-type PS12–gfp and ΔbcpAIOB strains. The miniTn7kan–gfp construct was used to deliver a gfp gene to the kanamycin-sensitive ΔbcpAIOB strain at the attTn7 site on chromosome 1. Plasmid pECG22 (Anderson et al., 2012), which contains the first ∼500 bp of bcpA immediately downstream of the strong constitutive rpsL promoter (PS12), was co-integrated into the kanamycin-sensitive wild-type PS12–gfp strain to generate the gfp-marked constitutive strain bcpC.

For protein analyses, the wild-type strain that encodes an HA tag located after F2633 in BcpA (E264BcpA-HA) was constructed previously (Anderson et al., 2012) and the HA-tagged BtEKA strain (BtEKABcpA-HA) was constructed in the same manner, by allelic exchange using pEXKm5 (López et al., 2009) to introduce HA-encoding sequence. For dot blot analysis, the phoA gene was PCR-amplified from E. coli RHO3 and cloned into pUC18T-miniTn7T-Km (Choi et al., 2008) immediately downstream of PS12. This phoA construct was integrated into one of the two attTn7 sites in the genomes of the HA-tagged strains. The kan cassette was removed from the phoA-expressing strains by Flp-FRT recombination as described above and plasmid pECG22 (Anderson et al., 2012), was co-integrated into all HA-tagged strains (both those with and those without phoA) to generate strains that constitutively expressed bcpAIOB from PS12.

For purification of BcpA-CT, DNA fragments containing the last 1146 bp of bcpA, encoding G2766 to the C-terminus, and bcpI (including the stop codon) were PCR-amplified from wild-type and BtEKA bacteria. Each fragment was cloned into pET28a (Novagen) in-frame with the vector's N-terminal His6 tag and including a 5′ ATG, resulting in plasmids pECG20 (encoding BcpA-CTWT) and pECG21 (encoding BcpA-CTEKA).

Static biofilm assay

Overnight cultures were washed in M63 medium and inoculated to an OD600 of 0.02 in 400 μl M63 in chambered coverglass dishes (Thermo Scientific). For mixed strain biofilms, GFP- and RFP-producing strains were mixed at a 1:1 ratio and inoculated as above. For DNase treatment, M63 medium was supplemented with 2 mM MgCl2 and 1 mM CaCl2 and 20 U DNase I (Ambion) was added during inoculation. For exogenous DNA addition, 3 μg chromosomal DNA, prepared from an overnight culture of E264 using Gentra Puregene reagents (Qiagen, according to the manufacturer's instructions), was added during biofilm inoculation. Biofilms were incubated in humidified chambers at 37°C for 6–72 h, washed four to five times with 400 μl PBS, overlaid with 400 μl PBS, and imaged by confocal laser scanning microscopy with a Zeiss LSM 700 using a 63× objective lens. Z stacks were processed with Imaris x64 v7.5.2 (Bitplane Scientific Software) and analysed with COMSTAT (Heydorn et al., 2000).

Protein alignments

Approximately 100 residues near the BcpA C-terminus were submitted to the PHYRE2 Protein Fold Recognition Server (Kelley and Sternberg, 2009) and the resulting high confidence/identity hits (Rp, Rhodopseudomonas palustris NP_945676; Ss, Sufolobus solfataricus AAK41424; Af, Archaeoglobus fulgidus NP_07377) were aligned to the BcpA region of homology (K3018-L3068). Also included in the alignment were homologues (determined by pBLAST) of each PHYRE2 hit (Bj, Bradyrhizobium japonicum YP_005605519; Bm, Brucella melitensis AAL52950; Mc, Metallosphaera cuprina YP_004459243; Pf, Pyrococcus furiosus NP_579232; Cm, Cyclobacterium marinum YP_004776468; Hm, Hippea maritima YP_004340293). Protein sequences were aligned using Vector NTI Advance 11 and alignments analysed with Jalview 2.7 with clustalx residue colouring.

For identification of the ‘E/DxK’ motif in other BcpA-CTs/CdiA-CTs, the region of BcpAE264 showing homology to Holliday junction resolvases (K3018-L3068) and the entire amino acid sequence of the putative truncated BcpA in B. thailandensis TXDOH (ZP_02369362) were used as pBLAST queries against the NCBI genome database. The resulting protein sequences, truncated to include solely the C-terminal 300–400 residues, were aligned as described above.


Overnight cultures were resuspended to an OD600 of 5.0 in 2× SDS-PAGE loading buffer, boiled 5 min, separated on a 5% SDS-PAGE gel, and transferred to nitrocellulose. Membranes were blocked in 5% milk in PBS with 0.01% Tween and probed with mouse monoclonal anti-HA.11 antibody (Covance) at 1:1000, followed by goat anti-mouse IgG conjugated to IRDye680 (Odyssey) at 1:15 000. Blots were imaged on a LiCor (Odyssey) with Odyssey v3.0 software.

Dot blots

Overnight cultures were washed in PBS and duplicate samples were resuspended to an OD600 of 4.0. One set of samples was boiled 5 min (lysate). Five microlitres of each sample (whole cells and lysate) were pipetted onto nitrocellulose and air-dried. Membranes were processed as described above or probed with mouse monoclonal anti-bacterial alkaline phosphatase (Sigma) at 1:1000 followed by secondary antibody and detection as described.

BcpA-CT/BcpI pull-down

BcpA-CT/BcpI complexes were purified as described previously (Aoki et al., 2010; Nikolakakis et al., 2012). Cultures (200 ml) of E. coli BL21(DE3) cells carrying plasmids pECG20 and pECG21 (encoding His-BcpA-CT and BcpI from wild-type and BtEKA bacteria respectively) were induced with 200 μM isopropyl-β-d-1-thiogalactopyranoside, resuspended in 10 ml sodium phosphate extraction buffer (20 mM sodium phosphate buffer pH 7, 150 mM NaCl, 0.05% Triton X-100, 100 mM β-mercaptoethanol) containing 1 mM phenylmethylsulphonyl fluoride and lysed by French press at 20 000 psi. Lysate was centrifuged twice at 3000 g at 4°C for 10 min and cleared lysate incubated with Ni2+NTA agarose resin (Qiagen) for 3 h at 4°C. Resin was washed twice with 10 ml sodium phosphate extraction buffer containing 20 mM imidazole and resuspended in 1 ml denaturing wash buffer (20 mM sodium phosphate buffer pH 7, 6 M guanidine-HCl, 10 mM β-mercaptoethanol). After incubation at room temperature for 5 min, flow-through was collected and resin was washed 8–10 times. Column material was resuspended in 300 μl denaturing elution buffer (20 mM sodium phosphate buffer pH 7, 6 M guanidine-HCl, 10 mM β-mercaptoethanol, 240 mM imidazole), incubated 5 min at room temperature, flow-through collected and elutions repeated five times. Fractions were analysed on a 12% SDS-PAGE gel.

Competition assay

Interbacterial competition was determined as previously described (Anderson et al., 2012). Bacteria [E264CmR, ΔbcpAIOB, BtEKA(Km) and BtEKA(Cm)] were cultured overnight with appropriate antibiotics, washed and diluted to OD600 0.2 in fresh LSLB without antibiotics. Strains were mixed at a 1:1 ratio and 20 μl 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 h. Agar plates were incubated at room temperature (∼25°C) for 24 h. 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. The competitive index (CI) was calculated as the ratio of the inhibitor strain to the target strain at time 24 h divided by the ratio at time 0 h.


The authors thank C. Robert Bagnell, Steven Ray and Victoria Madden at the UNC Microscopy Services Laboratory for microscopy assistance, Herbert Schweizer (Colorado State University) and Tung Hoang (University of Hawaii) for bacterial strains and plasmids, and John Leong (Tufts University) for insightful discussion. Research reported in this publication was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under Award Numbers U54 AI065359 (P.A.C.), R21 AI093154 (P.A.C.) and F32 AI096728 (E.C.G.). The content is solely the responsibility of the authors and does not represent the official views of the National Institutes of Health.