The toxin/immunity network of Burkholderia pseudomallei contact-dependent growth inhibition (CDI) systems


  • Kiel Nikolakakis,

    1. Department of Chemistry and Biochemistry, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
    • These primary authors contributed equally to the work.

  • Saba Amber,

    1. Department of Molecular, Cellular, and Developmental Biology, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
    • These primary authors contributed equally to the work.

  • J. Scott Wilbur,

    1. Department of Molecular, Cellular, and Developmental Biology, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
    • Present addresses: Vet. Sciences Micro Dept., 1117 E. Lowell St., Bld. 90, Room 301, University of Arizona, Tucson, AZ 85721;

  • Elie J. Diner,

    1. Biomolecular Science and Engineering Program, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
    • 415 Life Sciences Addition (LSA), University of California, Berkeley, CA 94720;

  • Stephanie K. Aoki,

    1. Department of Molecular, Cellular, and Developmental Biology, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
    • ETH Zürich, D-BSSE, WRO-1058, Warenannahme WRO-1007.P.18,Mattenstrasse 22,CH-4058 Basel, Switzerland;

  • Stephen J. Poole,

    1. Department of Molecular, Cellular, and Developmental Biology, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
  • Apichai Tuanyok,

    1. Center for Microbial Genetics and Genomics, Northern Arizona University, Flagstaff, AZ, USA
    Search for more papers by this author
  • Paul S. Keim,

    1. Center for Microbial Genetics and Genomics, Northern Arizona University, Flagstaff, AZ, USA
    2. The Translational Genomics Research Institute, Flagstaff, AZ, USA
    Search for more papers by this author
  • Sharon Peacock,

    1. Mahidol-Oxford Tropical Medicine Research Unit, Faculty of Tropical Medicine, Bangkok, Thailand
    Search for more papers by this author
    • Departments of Medicine and Pathology, University of Cambridge, Cambridge, UK.

  • Christopher S. Hayes,

    1. Department of Molecular, Cellular, and Developmental Biology, University of California, Santa Barbara, CA, USA
    2. Biomolecular Science and Engineering Program, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
    • These senior authors contributed equally to this work.

  • David A. Low

    Corresponding author
    1. Department of Molecular, Cellular, and Developmental Biology, University of California, Santa Barbara, CA, USA
    2. Biomolecular Science and Engineering Program, University of California, Santa Barbara, CA, USA
    Search for more papers by this author
    • These senior authors contributed equally to this work.


Burkholderia pseudomallei is a category B pathogen and the causative agent of melioidosis – a serious infectious disease that is typically acquired directly from environmental reservoirs. Nearly all B. pseudomallei strains sequenced to date (> 85 isolates) contain gene clusters that are related to the contact-dependent growth inhibition (CDI) systems of γ-proteobacteria. CDI systems from Escherichia coli and Dickeya dadantii play significant roles in bacterial competition, suggesting these systems may also contribute to the competitive fitness of B. pseudomallei. Here, we identify 10 distinct CDI systems in B. pseudomallei based on polymorphisms within the cdiA-CT/cdiI coding regions, which are predicted to encode CdiA-CT/CdiI toxin/immunity protein pairs. Biochemical analysis of three B. pseudomallei CdiA-CTs revealed that each protein possesses a distinct tRNase activity capable of inhibiting cell growth. These toxin activities are blocked by cognate CdiI immunity proteins, which specifically bind the CdiA-CT and protect cells from growth inhibition. Using Burkholderia thailandensis E264 as a model, we show that a CDI system from B. pseudomallei 1026b mediates CDI and is capable of delivering CdiA-CT toxins derived from other B. pseudomallei strains. These results demonstrate that Burkholderia species contain functional CDI systems, which may confer a competitive advantage to these bacteria.


Burkholderia pseudomallei (Bp) is an NIAID category B priority pathogen and the causative agent of melioidosis, a serious human disease endemic to southeast Asia and northern Australia and prevalent throughout tropical and subtropical regions (Limmathurotsakul and Peacock, 2011). Melioidosis can manifest as either acute or chronic disease and is often difficult to treat due to the bacterium's intrinsic resistance to multiple antibiotics. Bp infections are usually acquired from the environment, with soil and water constituting the main pathogen reservoir (DeShazer, 2004; Lazar Adler et al., 2009). Indeed, the incidence of melioidosis increases after periods of extended rainfall (Lin et al., 2011), consistent with a relationship between Bp infection and exposure to surface water. Given this mode of transmission, an understanding of the factors that contribute to pathogen growth and survival in the environment may provide strategies for eradication. Previous studies have shown that the distribution of Bp in the environment is complex (Chantratita et al., 2008; Corkeron et al., 2010). Bp often exhibits a discontinuous distribution with distinct sequence types (ST) dominating adjacent sites (Chantratita et al., 2008). Although each niche tends to be dominated by a particular ST, there does not appear to be a fixed strain hierarchy. A given ST may dominate another in one sample, yet be outcompeted in an adjacent site (Chantratita et al., 2008). Moreover, there is evidence that other Burkholderia species may exclude Bp strains from some environments (Lin et al., 2011). These and other observations suggest that Bp strains compete with one another and other species for niche dominance, but the underlying mechanisms are not understood (Chantratita et al., 2008).

Bacterial contact-dependent growth inhibition (CDI) provides one possible explanation for the complex distribution of Bp strains. CDI systems allow Gram-negative bacteria to inhibit the growth of neighbouring bacteria upon direct cell-to-cell contact and are widely distributed throughout the α-, β- and γ-proteobacteria (Aoki et al., 2005; 2010; Hayes et al., 2010). CDI is mediated by the CdiB/CdiA family of two-partner secretion proteins (Aoki et al., 2005). CdiB is an outer membrane protein required for the secretion and assembly of CdiA onto the inhibitor cell surface (Buscher et al., 2006; Mazar and Cotter, 2006; 2007). CdiA proteins carry the growth inhibitory activity and are predicted to form elongated filaments that extend from inhibitor cells to interact with target bacteria. Each CDI system also encodes an immunity protein (CdiI), which protects CDI+ cells from autoinhibition (Aoki et al., 2005; 2010). We recently localized the CDI toxin activity to the C-terminal region of CdiA (CdiA-CT). Remarkably, the CdiA-CT region is highly variable between different bacteria, suggesting that several distinct toxins are encoded by CDI systems. Indeed, the CdiA-CTs from uropathogenic Escherichia coli 536 and Dickeya dadantii 3937 have toxic tRNase and DNase activities respectively (Aoki et al., 2010). CdiI immunity proteins are also highly variable and specifically bind their corresponding CdiA-CTs to block toxin activity. Thus, a given CdiI protein confers immunity to its cognate CdiA-CT toxin, but provides no protection from the toxins deployed by other CDI systems.

Nearly all Bp strains contain two-partner secretion proteins related to CdiB and CdiA, but in most instances the predicted CdiA proteins are annotated as haemolysins, adhesins or filamentous haemagglutinins. The putative CDI systems of Burkholderia species are distinct from those found in other proteobacteria. First, the Burkholderia CDI genes are arranged as cdiAIB clusters in contrast to the cdiBAI organization found in all other genera for which genomic sequences are available. Additionally, Burkholderia CdiA proteins lack Pfam domain DUF638 (PF04829) and the associated VENN peptide motif that demarcates the variable CdiA-CT region in most other CDI systems (Aoki et al., 2010). Here, we examine CdiA-CT/CdiI pairs from B. pseudomallei and Burkholderia thailandensis to determine whether these proteins function as toxin/immunity pairs in inter-bacterial competition.

Expression of Bp CdiA-CTs in E. coli cells revealed that these proteins have distinct tRNase activities capable of inhibiting cell growth. Each of these toxic activities is specifically blocked by cognate CdiI immunity protein, demonstrating the toxin/immunity functions of these proteins. Moreover, we show that the cdiAIB genes of Bp strains and B. thailandensis mediate CDI. This growth inhibition is specifically blocked when target cells express the cognate cdiI immunity gene. These results indicate that Burkholderia CDI systems have the potential to contribute to the competition and survival of Bp strains in the environment.


Analysis of Burkholderia cdiAIB gene clusters

Bioinformatic analysis of fully sequenced and assembled Burkholderia genomes shows that most species contain at least one gene cluster encoding proteins related to CdiB and CdiA from γ-proteobacteria. Burkholderia cenocepacia and Burkholderia mallei are notable exceptions, although B. mallei strains contain proteins with haemagglutinin repeats that are distantly related to CdiA proteins. In contrast to the typical cdiBAI gene organization found in other bacteria (Aoki et al., 2010), the putative CDI genes in Burkholderia species are generally arranged as cdiAIB clusters (Fig. 1). Examination of the publically available B. pseudomallei (Bp) genomes revealed that all strains contain one to three cdiAIB clusters, and in most instances the predicted Bp CdiA proteins have been annotated as either haemolysins or putative adhesins. Several Bp CDI systems are linked to rhs genes (Fig. 1; for example, the Bp K96243 cdiA gene is adjacent to rhs locus BPSS2054), which share several features with cdiA genes and may also be involved in intercellular growth competition (Poole et al., 2011). In general, Burkholderia CdiA proteins lack Pfam DUF638 domains and the associated VENN peptide motifs that demarcate the variable CdiA-CT region in most other bacteria. Instead, Bp CdiA-CT sequences diverge at varying positions after a conserved (Q/E)LYN peptide motif (Fig. S1). Bp CdiA proteins fall into two classes based on the sequence of the conserved N-terminal region. CdiA sequences N-terminal to the (Q/E)LYN motif are 78–98% identical within each class, but only share ∼ 58% identity between classes (Fig. S1). Bp CdiA proteins can be further divided into 10 different subclasses based on the variable CdiA-CT sequences (Fig. S1 and Table S1). In general, there is > 98% amino acid sequence identity between Bp CdiA-CTs within each subclass; however, the type X subclass is more variable with > 88% sequence identity in pairwise alignments. To determine whether these CDI ST are representative of Bp cdiA gene diversity, we analysed CdiA-CT sequences encoded by a collection of 60 unassembled genomes from clinical and environmental Bp isolates (Table S2). We also amplified and sequenced the cdiA-CT coding regions from several environmental isolates described by Chantratita et al. (Chantratita et al., 2008). The additional CdiA-CT sequences all corresponded to one of the ten ST identified in the assembled Bp strain genomes (Tables S1 and S2), suggesting that the list of CdiA-CT ST is comprehensive. Additionally, the predicted cdiI immunity genes found immediately downstream of cdiA are also variable and fall into 10 distinct ST that correspond to the CdiA-CT ST (Fig. 1). Intriguingly, some Bp CDI systems may encode multiple, distinct immunity proteins. For example, the presumed cdiI gene in Bp NCTC 13177 is followed by a short sequence encoding a protein related to the CdiI immunity protein from Enterobacter aerogenes KCTC 2190 (43% sequence identity, pairwise BLAST E-value = 1e−16), as well as CdiI immunity proteins from other bacteria. In general, the Bp CDI systems contain one to four potential cdiI immunity genes between cdiA and cdiB (Fig. 1). Taken together, these observations suggest that the Bp CdiA-CT and CdiI proteins constitute a family of variable toxin-immunity pairs.

Figure 1.

Genomic organization of B. pseudomallei cdi gene clusters. The genomic neighbourhoods of representative Bp CDI systems encoding the 10 different CdiA-CT/CdiI sequence types (CDI types I through X) are presented. The Bp strain and encoding chromosome (if known) for each representative CDI system is given in parentheses. Bp 1106a carries two predicted CDI systems on chromosome I (types VII and X); these loci are labelled i and ii to distinguish the two systems. The cdiA and cdiB genes are shown in olive and yellow (respectively); and the cdiA-CT regions (corresponding to the 3′ end of cdiA genes) and cdiI genes are depicted in various colours. Genes related to transposases are shown in orange and all other coding sequences, including rhs (rearrangement hotspot) genes are white. Some systems contain additional small coding sequences between the cdiI and cdiB genes (white triangles). The asterisk indicates a putative immunity gene that is related to cdiI from Enterobacter aerogenes KCTC 2190.

Bp CDI systems encode toxin/immunity pairs

To determine whether Bp CdiA-CTs possess growth inhibitory activity, we examined the type V CdiA-CT sequence encoded on chromosome II of Bp 1026b (hereafter referred to CdiA-CTII1026b). The region encoding CdiAII1026b residues Met2821 to Asn3122 was cloned into plasmid pCH450 under the control of the arabinose-inducible PBAD promoter and introduced into E. coli cells. Induction of CdiA-CTII1026b synthesis with L-arabinose significantly inhibited E. coli cell growth compared with control cells carrying only the plasmid vector (Fig. 2A). This growth inhibition was completely alleviated by coexpression of CdiIII1026b from the IPTG-inducible Ptrc promoter on a compatible pTrc plasmid (Fig. 2A). In contrast, the predicted CdiIE478 and CdiIE479 immunity proteins (from Bp isolates E478 and E479, respectively) did not rescue the growth of CdiA-CTII1026b inhibited cells when produced from the pTrc plasmid (Fig. 2A). These results indicate that CdiA-CTII1026b has growth inhibitory activity and suggest that CdiIII1026b confers specific immunity.

Figure 2.

B. pseudomallei CdiA-CTs inhibit cell growth. The CdiA-CT/CdiI toxin/immunity network was examined in E. coli cells using a two-plasmid strategy. CdiA-CTII1026b and the CdiA-CTE478/CdiIE478-DAS and CdiA-CTE479/CdiIE479-DAS complexes were produced from plasmid pCH450 upon induction with L-arabinose (indicated by the downward arrows). Each culture was also induced at 0 min with IPTG to produce Bp CdiI immunity protein (indicated in the inset legend) from a compatible pTrc plasmid. A. CdiA-CTII1026b production was induced at 30 min, and growth monitored by measuring the optical density at 600 nm (OD600). B. CdiA-CTE478/CdiIE478-DAS complex production was induced at 30 min and growth monitored by OD600. The CdiA-CTE478 is liberated in these cells through degradation of ssrA(DAS)-tagged CdiIE478 immunity protein. C. CdiA-CTE479/CdiIE479-DAS complex production was induced at 30 min and growth monitored by OD600. The CdiA-CTE479 is liberated in these cells through degradation of ssrA(DAS)-tagged CdiIE479 immunity protein. In each panel, the black curve represents the growth of control cells carrying empty pCH450 and pTrc plasmid vectors.

We next asked whether the same CdiA-CT regions from Bp E478 (CDI type I) and Bp E479 (CDI type II) also inhibit cell growth, but we were unable to clone the isolated cdiA-CTE478 and cdiA-CTE479 coding sequences without the linked cdiI genes, presumably due to CdiA-CT toxicity. Therefore, we introduced the cdiA-CT/cdiI modules into a pCH450 vector that fuses the ssrA(DAS) peptide coding sequence in-frame to the cdiI genes and found the resulting constructs were stable under conditions that repress transcription from the PBAD promoter. The ssrA(DAS) peptide targets the tagged CdiI proteins for degradation by the ClpXP protease (Wah et al., 2003; McGinness et al., 2006; Poole et al., 2011) and, upon induction with L-arabinose, both cdiA-CT/cdiI-DAS expression constructs inhibited the growth of E. coli cells (Fig. 2B and C). This growth inhibition was specifically blocked by the co-production of cognate CdiI from a compatible plasmid (Fig. 2B and C). Together, these results strongly suggest that Bp CdiA-CT/CdiI proteins constitute toxin/immunity pairs.

Bp CdiA-CTs have distinct tRNase activities

The C-terminal 96 residues of CdiA-CTE478 share sequence identity with the cytotoxic tRNase domain of colicin E5 (56% identity, pairwise BLAST E-value = 5e−22), suggesting that nuclease activity is responsible for CdiA-CTE478 mediated growth inhibition. Colicin E5 specifically cleaves in the anticodon loops of tRNAHis, tRNATyr, tRNAAsn and tRNAAsp isoacceptors (Ogawa et al., 1999). Therefore, we asked if CdiA-CTE478 exhibits a similar tRNase activity. For these experiments, we used the controlled proteolysis system of McGinness et al. to activate the CdiA-CTE478 toxin in E. coli cells (McGinness et al., 2006). In this system, the CdiA-CTE478/CdiIE478-DAS complex was expressed constitutively in cells deleted for the sspB gene, which encodes an adaptor protein required to deliver ssrA(DAS)-tagged proteins to the ClpXP protease (Wah et al., 2002; 2003; Poole et al., 2011). Degradation of the ssrA(DAS)-tagged CdiIE478 immunity protein was then initiated by the production of SspB from a plasmid-borne PBAD promoter and total cellular RNA was isolated for analysis. Northern blot analyses revealed that tRNA1Tyr was cleaved upon activation of CdiA-CTE478 (Fig. 3). In contrast, full-length tRNA1Tyr persisted in control cells expressing SspB(Δ47), which binds the ssrA(DAS) peptide but delivers tagged proteins to ClpXP much less efficiently (Fig. 3) (Wah et al., 2003). Other substrates for colicin E5 (tRNAHis, tRNAAsn and tRNAAsp) were cleaved to a lesser degree, but non-substrate tRNA2Arg appeared unaffected by toxin activation (Fig. 3). These results indicate that CdiA-CTE478 exhibits tRNase activity resembling that of colicin E5.

Figure 3.

CdiA-CTE478 has tRNase activity similar to colicin E5. The CdiA-CTE478/CdiIE478-DAS complex was produced in E. coli ΔsspB cells and degradation of the ssrA(DAS)-tagged CdiIE478 was initiated at 0 h by induction of the SspB adaptor protein. RNA samples were taken at the indicated time points for Northern blot analysis. Cells expressing SspB(Δ47), which does not efficiently deliver CdiIE478-DAS to the ClpXP protease, were used as a control. Each gel lane contains 10 µg of total RNA, which was hybridized against oligonucleotide probes specific for E. coli tRNA1Tyr, tRNAAsn, tRNAHis and tRNA2Arg. The gel migration positions for full-length and cleaved tRNA are indicated.

Bioinformatic analysis of the CdiA-CT sequences from Bp 1026b and Bp E479 failed to reveal clues to their biochemical activities. Given that several characterized CdiA-CTs are nucleases (Aoki et al., 2010), we examined DNA and RNA from inhibited E. coli cells for evidence of nuclease activity. Gel analysis of nucleic acids isolated from cells inhibited by the CdiA-CTE479/CdiIE479-DAS complex failed to reveal obvious nuclease activity (data not shown). However, overexpression of a truncated cdiA-CTE479/cdiIE479 construct (encoding the C-terminal 160 residues of CdiA-CTE479) from a strong bacteriophage T7 promoter resulted in profound growth arrest and led to the degradation of most cellular tRNAs (Fig. 4). Northern blot analysis revealed a cleavage product for tRNA2Arg, but similar degradation intermediates were not observed with other tRNA species such as tRNA3Gly (Fig. 4). We used S1 nuclease protection to identify a cleavage between the conserved thymidine (T55) and pseudouridine (Ψ56) residues in the T-loop of tRNA2Arg (Fig. S2). This cleavage is predicted to inactivate tRNA and appears to initiate tRNA degradation – both processes could account for growth inhibition. These results also indicate that the N-terminal 156 residues of CdiA-CTE479 are not required for RNase activity.

Figure 4.

CdiA-CTE479 has tRNase activity. CdiA-CTE479/CdiIE479 production was induced by IPTG in E. coli cells, and RNA samples taken for gel and Northern blot analyses. RNA from cells expressing CdiA-CTE479 with the Asp285Ala mutation was also analysed. RNA was hybridized against oligonucleotide probes specific for E. coli tRNA2Arg and tRNA3Gly. The gel migration positions for ribosomal RNA (rRNA) and tRNA are indicated.

Nuclease activity was not readily observed by gel analysis of DNA and RNA isolated from CdiA-CTII1026b -inhibited E. coli cells (Fig. S3). Because the cleavage of one (or a few) tRNA species would be difficult to detect by gel analysis, we screened all 46 E. coli tRNA isoacceptors by Northern blot hybridization. This analysis revealed that a single tRNA, tRNA1BAla, was preferentially cleaved upon expression of CdiA-CTII1026b (Fig. S3). S1 nuclease protection analysis showed that this cleavage occurred in the aminoacyl acceptor stem near the 3′ end of the tRNA (data not shown). Cleavage at this position prevents tRNA charging and is presumably the underlying mechanism of growth inhibition. Together, these results show that the Bp CdiA-CTs have distinct tRNase activities, each providing a possible mechanism to inhibit cell growth.

CdiA-CT tRNase activities are required for growth inhibition

To determine whether the identified tRNase activities are responsible for growth inhibition, we inactivated each CdiA-CT with a single missense mutation and tested the resulting proteins for growth inhibition activity in E. coli. We altered residue Asp272 to Ala in CdiA-CTE478 (numbered from Gln1 of the (Q/E)LYN motif) based on previous work showing the corresponding residue in colicin E5 is important for nuclease activity (Lin et al., 2005). The resulting Asp272Ala protein lacked both tRNase and growth inhibitory activities (Fig. S4 and data not shown). The biochemistry and structures of CdiA-CTII1026b and CdiA-CTE479 are uncharacterized, so we used multiple sequence alignments with distant protein homologues to identify conserved residues for mutagenesis. An Asp214Ala mutation in CdiA-CTII1026b ablated tRNase activity in vivo (Fig. S3) and rendered the domain non-toxic in E. coli (Fig. S4). Similarly, mutation of either Asp280 or Asp285 to alanine in CdiA-CTE479 eliminated tRNase and growth inhibitory activities (Figs 4 and S4). Taken together, these results suggest that the tRNase activity of each CdiA-CT is required for growth inhibition.

CdiA-CTs bind to cognate CdiI immunity proteins

Our previous studies have shown that CdiI immunity proteins bind specifically to cognate CdiA-CTs to block nuclease activities (Aoki et al., 2010; Poole et al., 2011). To determine whether the Bp CdiA-CT/CdiI pairs also form specific complexes, we cloned each cdiA-CT/cdiI pair into an overexpression vector that appends a C-terminal His6 epitope tag onto the CdiI immunity protein and overproduced the proteins for biochemical analyses. The CdiA-CTII1026b/CdiIII1026b-His6 proteins were overproduced to high levels and co-purified by Ni2+-affinity chromatography, suggesting they form a stable complex. However, we found that the N-terminal region of CdiA-CTII1026b underwent extensive degradation during purification, so we generated a truncated construct that produced a stable CdiA-CTII1026 (beginning at residue Gly123 with respect to Gln1 in the QLYN motif) for subsequent analyses. This truncated CdiA-CTII1026b co-purified with His6-tagged CdiIII1026b (Fig. 5), indicating that the N-terminal region of CdiA-CTII1026b is not required to bind immunity protein. In contrast, the Bp E478 and Bp E479 CdiA-CT/CdiI-His6 could not be overproduced to high levels, and induction of these expression constructs inhibited cell growth. We reasoned that low levels of free CdiA-CT were degrading cellular tRNA and thus limiting protein overproduction. Therefore, we used the inactive Asp272Ala and Asp285Ala versions of the Bp E478 and Bp E479 CdiA-CTs (respectively) and obtained much higher levels of protein overexpression. As observed for the Bp 1026b CdiA-CT/CdiI, the untagged CdiA-CTs from Bp E478 and Bp E479 co-purified with their cognate His6-tagged CdiI proteins, suggesting complex formation. These results also indicate that the catalytically inactive CdiA-CT domains retain their native fold.

Figure 5.

B. pseudomallei CdiI immunity proteins bind specifically to cognate CdiA-CT. Purified CdiA-CTs and His6-tagged CdiI immunity proteins were incubated together, then subjected to Ni2+-affinity chromatography to bind CdiI-His6 and any associated CdiA-CT. All fractions were run on SDS-polyacrylamide gels and stained with Coomassie blue. The lanes labelled input show the unfractionated CdiA-CT/CdiI-His6 mixtures, the lanes labelled free show proteins that did not bind Ni2+ resin, and the lanes labelled bound show proteins that bound to the resin.

We next asked whether CdiA-CT/CdiI binding interactions are specific for cognate pairs. We isolated each CdiA-CT and CdiI-His6 protein by Ni2+-affinity chromatography under denaturing conditions, and then refolded the purified proteins for in vitro binding studies. CdiI-His6 immunity proteins were used as ‘bait’ to test for the co-purification of untagged CdiA-CT ‘prey’ by Ni2+-affinity chromatography. Each CdiI-His6 protein retained its cognate CdiA-CT on the Ni2+-NTA resin, demonstrating that cognate pairs can re-establish their binding interactions after the denaturation and refolding protocol (Fig. 5). In contrast, non-cognate CdiA-CTs did not co-purify with the His6-tagged CdiI immunity proteins (Fig. 5), demonstrating the specificity of CdiA-CT/CdiI binding interactions. These results are consistent with the specificity of CdiI-mediated immunity to growth inhibition in E. coli cells and suggest that CdiI blocks CdiA-CT nuclease activity.

CdiIII1026b specifically blocks CdiA-CTII1026b tRNase activity

We next tested whether purified CdiI immunity protein can block tRNase activity in vitro. Because we were unable to obtain appreciable quantities of active CdiA-CTE478 or CdiA-CTE479, we used the CdiA-CTII1026b/CdiIII1026b complex as a model. We assayed the in vitro tRNase activities of the truncated CdiA-CTII1026b and its Asp214Ala variant using B. thailandensis tRNA as a substrate. In contrast to the activity in E. coli cells, where only tRNA1BAla was cleaved (Fig. S3), we found that several tRNA species were cleaved in vitro (Fig. 6). The discrepancy in activities could reflect the different tRNA substrates used. However, CdiA-CTII1026b also cleaved most (if not all) E. coli tRNAs in vitro (data not shown), suggesting that conditions inside the cell limit tRNase activity. Substrate specificity notwithstanding, we found that the tRNase activity was specifically blocked by the addition of CdiIII1026b-His6, but not by either of the other two non-cognate immunity proteins (Fig. 6). Together, these results show that tRNase activity is contained within the C-terminal 175 residues of CdiA-CTII1026b and support a model in which CdiI blocks the toxin activity of CdiA-CT through direct binding.

Figure 6.

CdiIII1026 immunity protein specifically blocks CdiA-CTII1026b tRNase activity in vitro. Burkholderia thailandensis tRNA was incubated with purified CdiA-CTII1026b and analysed by gel electrophoresis and Northern blot hybridization. Different CdiI immunity proteins were included in reactions where indicated, and CdiA-CTII1026b containing the Asp214Ala mutation was also assayed for tRNase activity. RNA was hybridized against oligonucleotide probes specific for B. thailandensis tRNAICGArg and tRNAUGCAla. The gel migration positions of full-length tRNA are indicated.

Burkholderia CDI systems are functional

The data presented thus far show that B. pseudomallei CdiA-CTs are toxic when expressed inside E. coli cells, raising the possibility that these systems function in bacterial competition. To test this hypothesis, we examined the predicted CDI system on chromosome I of B. thailandensis E264 (Bt E264), which is very closely related to B. pseudomallei (Yu et al., 2006). We deleted the single cdiAIB gene cluster from Bt E264 and found that the resulting strain was inhibited by wild-type CDI+ bacteria when co-cultured together on Luria–Bertani (LB) agar plates (Fig. S5). Control experiments showed that co-culture of differentially marked wild-type Bt E264 cells resulted in no growth inhibition (Fig. S5). To determine whether this growth inhibition was due to CDI activity, we reintroduced the cdiIE264 immunity gene at the glmS locus (on chromosome I) of Bt E264 ΔcdiAIB cells using a Tn7-based vector (Choi et al., 2005). The resulting strain was no longer inhibited by wild-type Bt E264 (Fig. S5), consistent with the predicted immunity function of the CdiIE264 protein. In contrast, Bt E264 ΔcdiAIB cells carrying the cdiI genes from Bp 1026b, Bp E478 or Bp E479 were still inhibited by wild-type Bt E264 cells (Fig. S5). Taken together, these results strongly suggest that the Bt E264 cdiAIB genes are expressed and function in growth inhibition.

We next tested whether Bp CDI systems also mediate growth inhibition. We cloned the cdiAIB gene cluster from Bp 1026b (chromosome II) under control of the arabinose-inducible PBAD promoter on a low-copy pVS1-based shuttle plasmid (Heeb et al., 2000) and introduced the resulting construct into wild-type Bt E264. Bt E264 cells carrying the CDIII1026b system on a plasmid were co-cultured with Bt E264 target cells at a 100:1 ratio in both liquid media and on agar plates. Although no growth inhibition was observed in liquid culture (data not shown), viable target cell counts were reduced ∼ 10-fold after 6 h of co-culture on solid media (Fig. 7A). Target cells were further inhibited upon prolonged incubation, with ∼ 100-fold reduction in viable target cell counts after 48 h (Fig. 7A). These results indicated that the CDIII1026b system is functional in inter-species growth inhibition when expressed in Bthailandensis cells. To determine whether cell-to-cell contact is required for growth inhibition, we separated inhibitor and target cells with membranes containing pores that either allow (8 µm) or restrict (0.45 µm) cell passage (Fig. 7B). Target cell growth was arrested when cells could cross the membrane but not when cells were prevented from mixing (Fig. 7B). Taken together, these results indicate that inhibitor cells must be in close proximity to target cells to inhibit growth.

Figure 7.

Growth inhibition mediated by the CDIII1026b system is contact-dependent. A. Bt E264 inhibitor cells carrying the Bp CDIII1026b system or empty cosmid vector (mock inhibitors) were co-cultured with Bt E264 target cells as described in Experimental procedures. Cells were collected at the indicated time points, and viable counts for both inhibitor and target cells were determined as colony forming units per ml (cfu ml−1). B. A filter competition assay was carried out as described in Experimental procedures. Inhibitor cells expressing the CDIII1026b system, or mock inhibitor cells lacking the CDIII1026b system were separated from Bt6 target cells by filters with the indicated pore sizes, and viable target cell counts were determined after 24 h incubation. cfu ml−1 values were determined from three technical replicates and are expressed as the mean ± SEM.

Bp CDI systems are modular

The CdiA proteins of E. coli and related γ-proteobacteria are modular, capable of delivering a variety of CdiA-CT toxins when fused at the VENN peptide sequence (Aoki et al., 2010). Although Burkholderia CdiA proteins lack the VENN motif, the variability of Bp CdiA-CT/CdiI sequences suggests that these systems are also modular. Therefore, we replaced the cdiA-CT and cdiI coding sequences in the plasmid-borne CDIII1026b system with the corresponding sequences from Bp K96243 (CDI type I, 99.6% sequence identity to CdiA-CTE478) and Bp E479 to construct chimeric CdiA proteins carrying different CdiA-CTs fused at the (Q/E)LYN peptide sequence. Bt E264 expressing the CdiAII1026b-CTE479 chimera inhibited target Bt E264 cells ∼ 100-fold after 24 h of co-culture, but had no effect on target cells expressing the cognate CdiIE479 immunity protein (Fig. 8). Moreover, non-cognate CdiI proteins were unable to protect target cells from CdiAII1026b-CTE479 mediated growth inhibition (Fig. 8). These results indicate that the chimeric CdiAII1026b-CTE479 protein is functional and that the immunity function is specific for cognate CdiA-CT toxin. The CdiAII1026b-CTK96243 chimera also inhibited the growth of Bt E264 target cells, but cells expressing the cognate CdiIK96243 immunity protein were protected from inhibition (Fig. 8). Together, these results show that Bp CdiA proteins are modular and can deliver a variety of the CdiA-CT toxins when fused at the common (Q/E)LYN motif.

Figure 8.

B. pseudomallei CDI systems are modular. Bt E264 target cells were co-cultured with inhibitor cells expressing the Bp CDIII1026b system or chimeric CDI systems that deploy the CdiA-CTs from Bp K96243 or Bp E479. Target cells were provided with Bp cdiI immunity genes as indicated. The competitive index was calculated as the inhibitor cell cfu/target cell cfu ratio at t = 0 h divided by the inhibitor cell cfu/target cell cfu ratio at t = 24 h. Competitive indices were determined from three technical replicates and are expressed as the mean ± SEM.

Dependence of CDI-mediated growth inhibition on CdiB and tRNase activity

By analogy with other two-partner secretion systems (Mazar and Cotter, 2006; Choi and Bernstein, 2010), the CdiB protein should be required for secretion of CdiA and growth inhibition of target cells. To test whether CdiB is required for CDI in Burkholderia, we deleted the cdiB gene from the plasmid-borne CDIII1026b system and found that this mutation abrogated growth inhibition (Fig. 9), consistent with the proposed role of CdiB in secretion. Finally, we tested whether the tRNase activities identified for the various Bp CdiA-CT proteins are required for cell-mediated growth inhibition. The Asp214Ala mutation that ablates CdiA-CTII1026b tRNase activity (Asp3039Ala in full-length CdiAII1026b) was introduced into the plasmid-borne CDIII1026b system, and the resulting strain was used in growth competition assays. Bt E264 carrying the CDIII1026b system lacking tRNase activity was unable to inhibit target cell growth (Fig. 9), indicating that the mutant system lost all inhibitory activity. We also tested CdiA chimeras carrying the Asp272Ala mutation in CdiA-CTK96243 and the Asp285Ala mutation in CdiA-CTE479 and found that both mutations also abrogated growth inhibition (data not shown). These results indicate that CdiA-CT tRNase activities are responsible for the inhibition of target cell growth during CDI.

Figure 9.

CDI-mediated growth inhibition requires CdiB and tRNase activity. Bt24 cells (B. thailandensis E264 ΔcdiAIB) carrying a plasmid-borne copy of the CDIII1026b system or its mutant derivatives that either lack the cdiB gene or encode the Asp3039Ala variant of CdiAII1026b (corresponding to the catalytically inactive Asp214Ala change in CdiA-CTII1026b) were incubated with Bt6 target cells. Bt7 target cells (expressing the cdiIII1026b immunity gene) were used as a negative control. Growth competition analysis was carried out as described in Experimental procedures. cfu ml−1 values were determined from three technical replicates and are expressed as the mean ± SEM.


The results presented here demonstrate that CDI systems of B. thailandensis and B. pseudomallei function in growth competition. Although the organization of the cdi gene clusters differs from that in most other bacteria, Burkholderia CDI systems retain tight linkage of the cdiA effector and cdiI immunity genes. Like other CDI systems, the C-terminal regions of Burkholderia CdiA proteins are variable and contain the growth inhibition activity. The Bp CdiA-CT toxins examined here all exhibit RNase activity and target distinct sites within tRNA molecules. Cleaved tRNAs are not functional in protein synthesis, and presumably this tRNA inactivation accounts for growth inhibition. However, we cannot exclude the possibility that cleavage of other critical RNAs contributes to inhibition. Molecular targets notwithstanding, it is clear that RNase activity accounts for the toxicity of Bp CdiA-CTs. Catalytically inactive Bp CdiA-CTs have no effect on growth when expressed in E. coli and do not inhibit B. thailandensis growth during cell-mediated CDI. These latter findings indicate that the CdiA-CT is the only toxin delivered by Bp CDI systems. The CdiA-CT from Bt E264 also has nuclease activity, but is a Mn2+-dependent DNA ‘nickase’ rather than a tRNase (K. Nikolakakis and C.S. Hayes, unpubl. results). Thus, all Burkholderia CdiA-CTs examined in this study possess nuclease activity capable of inhibiting bacterial growth. Additionally, we find that the predicted Burkholderia cdiI genes encode immunity proteins that neutralize the toxicity of corresponding cognate CdiA-CTs. Each CdiI immunity protein binds to its cognate CdiA-CT but not to heterologous toxins. This cognate binding interaction blocks CdiA-CTII1026b-mediated tRNase activity in vitro, suggesting that the other CdiI proteins also block nuclease activity upon binding cognate CdiA-CT. Based on these results, it seems likely that other Burkholderia CdiA-CT domains are toxins and are specifically neutralized by their cognate CdiIs. Our results indicate that CDI systems constitute a complex toxin/immunity network among Burkholderia species.

Intriguingly, each Bp CdiA-CT examined here appears to target tRNA to inhibit cell growth. The CdiA-CT deployed by several uropathogenic E. coli strains displays yet another tRNase activity that cleaves within the anticodon loops of many isoacceptors (Aoki et al., 2010). There are a number of possible reasons why tRNA molecules are common targets for CDI systems. First, tRNAs play a critical role in translation, and their inactivation provides an efficient means to interfere with protein synthesis. At least three bacteriocins (colicin E5, colicin D and pyocin S4) are known to cleave specific tRNA isoacceptors to kill susceptible bacteria (Ogawa et al., 1999; Tomita et al., 2000; Michel-Briand and Baysse, 2002). The CdiA-CTs from Bp E478 and Bp K96243 share significant sequence identity with the nuclease domain of colicin E5 (Aoki et al., 2010), implying exchange of the toxin between these systems. Additionally, the conserved architecture of tRNA molecules could facilitate the evolution of CdiA-CT toxin diversity. According to this model, the CdiA-CT domain would accumulate changes allowing it to recognize different tRNA isoacceptors, while retaining the same overall fold and intrinsic RNase activity. Of course, the associated CdiI immunity protein must also co-evolve to maintain cognate binding interactions and provide immunity to the new toxin. The CdiI proteins from related CDI systems typically exhibit lower sequence identity than the CdiA-CT regions, suggesting that the immunity proteins are evolving more rapidly than the toxins and could be the force driving diversification. Finally, because tRNAs are ubiquitous, these toxins have the potential to inhibit all cells provided they could be delivered into the cytosol. In accord, we find that the Bp CdiA-CTs inhibit growth when produced inside E. coli cells. In fact, the molecular targets for all characterized CDI systems are universal and include membranes, DNA and RNA (Aoki et al., 2009; 2010). This feature of CDI is consistent with the distribution of related toxin/immunity pairs in diverse bacteria (Aoki et al., 2010; Poole et al., 2011). For example, sequences related to CdiA-CTII1026b are found not only in other Burkholderia species (Burkholderia ambifaria IOP40-10, Burkholderia phymatum DSM 17167 and B. gladioli BSR3), but also in the CDI systems of other β- and γ-proteobacteria – including Ralstonia solanacearum strains, Serratia proteomaculans 568 and Pseudomonas aeruginosa PA7 (see Fig. S6). Additionally, two δ-proteobacteria (Myxococcus xanthus and Haliangium ochraceum) contain uncharacterized proteins with C-terminal domains related to the CdiA-CTII1026b nuclease. This wide distribution indicates that the tRNase domain functions in a variety of proteobacteria, where it presumably mediates intercellular growth competition.

Although CdiA-CTs can be exchanged between related CDI systems (Aoki et al., 2010), fusion of the Bp CdiA-CTII1026b at the VENN motif of the E. coli EC93 CdiA protein does not produce a functional chimera (S.K. Aoki and D.A. Low, unpubl. results). Because the toxic nuclease domain must be delivered into target cells, it is possible that specific sequences within the CdiA-CT are required for translocation across the target cell envelope. Comparative sequence analysis indicates that many CdiA-CTs are bipartite. For example, an alignment of Bp CdiA-CTII1026b with CdiA-CTs from other bacteria shows two distinct regions of homology (see Fig. S6). As described above, the C-terminal 130 residues of CdiA-CTII1026b share sequence identity with the C-terminal regions of CdiA proteins from other species. This extreme C-terminal region likely represents the actual toxic effector domain, because it contains the RNase activity and is sufficient to bind to immunity protein. The N-terminal region of CdiA-CTII1026b shares significant sequence identity with the corresponding regions from other Bp CdiA proteins, yet the C-terminal toxin sequences for each protein are distinct (see Fig. S6). These observations suggest that CdiA-CTs themselves are modular and built from C-terminal toxin domains and N-terminal regions of unknown function. Related N-terminal regions are generally restricted to phylogenetically similar bacteria, suggesting that these domains function in a limited number of species. The N-terminal region could contribute to cell-surface receptor binding, or perhaps help to mediate CdiA-CT translocation through the target cell envelope.

Are the polymorphic CdiA-CT regions of Bp CDI systems sufficient to account for environmental distributions? The results presented here clearly demonstrate that Burkholderia CDI systems mediate intercellular growth competition. Moreover, because the Bp 1026b CDIII system is capable of inhibiting B. thailandensis growth, it appears likely that CDI+Burkholderia cells are able to target other species within the genus. Our analysis shows that there are at least 10 CdiA-CT ST in Bp strains. There is additional CdiA-CT sequence diversity in the predicted CDI systems of B. ambifaria, B. glumae and B. phymatum. If all of these different CdiA-CTs can be delivered to every Burkholderia species, then the systems would constitute an elaborate toxin/immunity network that could perhaps shape species distribution in the environment. Of course, other factors are also likely to contribute to inter-strain competition. For example, bacterial type VI secretion systems have recently been shown to inject toxic effectors into competing bacteria to inhibit their growth (Hood et al., 2010; Schwarz et al., 2010). The proposed interactions are complex and it is not yet possible to predict the winner of a head-to-head competition between strains deploying different CDI and/or type VI systems. Do competing strains inhibit each other equally, such that no strain dominates? Are some toxins more potent than others (for example having a higher tRNase turnover rate), giving a decided advantage to some strains? These questions are further complicated by the fact that most Bp strains contain multiple CDI systems, increasing the complexity of any given cell–cell interaction. Moreover, there appear to be additional cdiI genes located in the cdiI–cdiB intergenic regions of some CDI systems, which could confer immunity to other bacterial strains.

Another unresolved question is whether the CDI mechanism can account for the observed distributions of Bp strains in the environment. Soluble CdiA fragments are not inhibitory and therefore each CDI+ inhibitor cell has a very limited ‘sphere of inhibition’. Therefore, micro-colonies of different Burkholderia strains should coexist provided they are segregated. CDI could influence the intersection zones of different strains. Burkholderia cells are motile (Tomich et al., 2002; Boonbumrung et al., 2006) and this property should allow different strains to encounter one another periodically. Additionally, the environment is subject to significant seasonal rains that could disrupt micro-niches and facilitate mixing of different Burkholderia species (Wuthiekanun et al., 2009), allowing CDI to play a role in niche competition. Alternatively, it is possible that CDI systems perform other functions that are distinct from growth inhibition. Bacteria typically exist as clonal micro-aggregates, and therefore CDI+ cells probably deliver toxins to their isogenic siblings. If CDI systems are solely dedicated to growth inhibition, then this delivery of CdiA-CT toxins to immune sibling cells would appear to be futile. However, the exchange of CdiA-CT domains between isogenic cells could serve an unappreciated signalling function. According to this model, the delivered CdiA-CT would have a distinct signalling function when it forms a complex with its cognate CdiI immunity protein.

Experimental procedures

Bacterial strains and plasmids

All bacterial strains used in this study are presented in Table 1. Bacteria were grown in LB broth or M9 minimal media supplemented with 20 mM d-glucose. Antibiotics were used at the following concentrations: B. thailandensis– 500 µg ml−1 kanamycin (Kan); 300 µg ml−1 trimethoprim (Tp); 0.1% p-chlorophenylalanine (cPhe); 100 µg ml−1 polymyxin B (PxB); E. coli– 150 µg ml−1 ampicillin (Amp); 40 µg ml−1 kanamycin (Kan); 12.5 µg ml−1 tetracycline (Tet); and 50 µg ml−1 trimethoprim (Tp). All bacterial cultures were incubated at 37°C with aeration unless otherwise indicated. E. coli strain X90 and its derivatives were used for analysis of Bp CdiA-CT activities. All plasmids used in this study are listed in Table S3 and the details of all plasmid constructions are presented in the Supporting Information. All cdiA-CT/cdiI-DAS plasmid constructs were generated in either E. coli CH4180 or CH7157, and then transformed into strain X90 to assess CdiA-CT toxicity. The cdiA-CTII1026b and cdiA-CT/cdiI-DAS expression plasmids (for CdiA-CTE478 and CdiA-CTE479) were co-transformed into E. coli X90 along with compatible plasmids encoding the Bp CdiI immunity proteins. Overnight cultures of each strain were resuspended at OD600 = 0.05 in fresh LB medium supplemented with Tet, Amp and 1 mM IPTG to induce expression of CdiI immunity proteins. After 30 min, CdiA-CT (and CdiI-DAS) expression was induced by addition of L-arabinose to 0.2%, and culture growth monitored at 30 min intervals over the following 5 h. Overproduction of CdiA-CT/CdiI-His6 complexes was performed in strain CH2016 (Garza-Sánchez et al., 2006).

Table 1. Bacterial strains used in this study.
  1. KanR, kanamycin resistant.

X90F′lacIq lac′ pro′/araΔ(lac-pro) nal1 argE(amb) rifr thi-1 Beckwith and Signer (1966)
CH2016X90 (DE3) Δrna ΔslyD::kan, KanR Garza-Sánchez et al. (2006)
CH4180X90 ΔsspB Poole et al. (2011)
CH7157X90 ΔclpX ΔclpA::kan, KanR Poole et al. (2011)
Burkholderia thailandensis E264 Burkholderia thailandensis wild-type isolate Brett et al. (1998)
Bt6 B. thailandensis E264 attTN7::miniTn7T-KanThis study
Bt7 B. thailandensis E264 attTN7::miniTn7T-kan-cdiIII1026b, KanRThis study
Bt8 B. thailandensis E264 attTN7::miniTn7T-kan-cdiIK96243, KanRThis study
Bt9 B. thailandensis E264 attTN7::miniTn7T-kan-cdiIE479, KanRThis study
Bt23 B. thailandensis E264 ΔcdiAThis study
Bt24 B. thailandensis E264 ΔcdiAIBThis study
Bt36 B. thailandensis E264 ΔcdiAIB attTN7::miniTn7-kan, KanRThis study
Bt56 B. thailandensis E264 ΔcdiAIB attTN7::miniTn7-PS12-cdiIE264, KanRThis study

The cdiA (BTH_I2723) and cdiI/cdiB (BTH_I2722/BTH_I2721) genes were deleted sequentially from B. thailandensis E264 using allelic exchange as described (Kang et al., 2011). Regions upstream and downstream of cdiAE264 were amplified by PCR using primer pairs 2130/2265 and 2266/2267 (oligonucleotides used in this study are presented in Table S4), and the resulting PCR products combined by overlapping extension PCR (OE-PCR) using primers 2130 and 2267 (Aiyar et al., 1996). The resulting product was digested with HindIII/KpnI and ligated to pEX18Tp-pheS (Kang et al., 2011) to construct plasmid pDAL917. DNA regions encompassing the cdiI-cdiB DNA region were amplified by PCR using primer pairs 2130/2004 and 2005/2131, and the resulting products were combined by OE-PCR using primers 2130 and 2131. The resulting DNA product was digested with HindIII/KpnI and ligated to plasmid pEX18Tp-pheS to construct plasmid pDAL908. The cdiAE264 gene was first deleted using plasmid pDAL917, followed by deletion of the cdiIE264 and cdiBE264 genes using plasmid pDAL908. The cdiI immunity genes from B. pseudomallei and B. thailandensis were placed under control of the Bt E264 PS12 promoter and inserted into chromosome I of B. thailandensis E264 ΔcdiAIB cells using a Tn7-based integration vector (Choi et al., 2006; 2008). A cassette formed by oligonucleotides SpeI-S12-UP and S12-XmaI-DN was ligated into SpeI/XmaI-digested pUC18T-miniTn7T-Gm to construct plasmid pUC18T-miniTn7T-Gm-PS12. The cdiIII1026b immunity gene was amplified using primers CdiI1026b- XmaI-FWD and CdiI1026b-XhoI-REV and ligated to SmaI-digested pUC18T-miniTn7T-Gm-PS12 to construct pUC18T-miniTn7T-Gm-PS12-cdiIII1026b. Due to the intrinsic resistance of Bt E264 to gentamicin, a fragment containing the PS12-cdiIII1026b sequence was excised by HindIII/SpeI digestion and ligated to pUC18T-miniTn7-Kn-FRT (Choi et al., 2008). The resulting plasmid, pAMB7 (pUC18T-miniTn7T-Kan-FRT-PS12-cdiIII1026b) was transformed into E. coli DH5α, and the resulting strain used in a four-parent mating with SM10λpir/pTNS3 (Miller and Mekalanos, 1988; Choi et al., 2005), HB101 (pRK2013) (Ditta et al., 1980) and Bt E264. Bt E264 exconjugants were selected with PxB and Kan and clones screened for mini-Tn7 insertions by whole-cell PCR using primer pairs Tn7L/Tn7R and pglmS1/glmS1-DN. The cdiIK96243 gene was amplified with primers CdiIK96243-XmaI-UP/CdiIK96243-HindIII-DN, the product digested with XmaI/HindIII and ligated to pAMB7 to generate plasmid pAMB8. The cdiIE479 gene was amplified with primers CdiI479-KpnI-FWD/CdiI479-PstI-REV, the product digested with PstI and ligated to PstI/SmaI-digested pAMB7 to generate plasmid pAMB10. The cdiI gene from Bt E264 was amplified using primer pair 2536/2537 and the resulting product ligated pAMB7 between SmaI and KpnI to generate plasmid pDAL940. Plasmids pAMB8, pAMB10 and pDAL940 were introduced into E. coli DH5α and the resulting strains used in four-parent matings as described above. The cdi gene clusters from environmental Bp isolates were amplified using oligonucleotides cdiA-con and cdiB-con, and the resulting products sequenced using the initial primers and then by sequence walking.

RNA analysis

RNA was isolated from E. coli and B. thailandensis cells as previously described (Garza-Sánchez et al., 2006). Northern blot hybridization was performed as described (Hayes and Sauer, 2003), using 5′ radiolabelled oligonucleotides Bt tRNAICGArg and Bt tRNAUGCAla probes. S1 nuclease protection assays were conducted essentially as described (Hayes and Sauer, 2003). Oligonucleotide tRNA2Arg S1 probe was radiolabelled at the 5′ end with polynucleotide kinase to map 3′ cleavage fragments and at the 3′ end with terminal transferase to map 5′ cleavage fragments. The marker oligonucleotides were radiolabelled at the appropriate termini and used as gel migration standards.

In vitro analysis of CdiA-CT/CdiI proteins

CdiA-CT/CdiI-His6 complexes were purified by Ni2+-affinity chromatography and the proteins separated under denaturing conditions as described (Aoki et al., 2010; Poole et al., 2011). Isolated CdiA-CTs were dialysed against 10 mM acetic acid and stored at 4°C. All CdiI-His6 proteins were overproduced in the absence of CdiA-CT and purified by Ni2+-affinity chromatography. CdiI-His6 proteins were dialysed against storage buffer [50 mM Tris-HCl (pH 7.5), 100 mM NaCl, 10 mM β-mercaptoethanol (β-ME)]. CdiA-CT/CdiI-His6 binding studies were conducted in binding buffer [50 mM Tris-HCl (pH 7.5), 400 mM NaCl, 10 mM β-ME]. Purified CdiA-CT (5 µM) was incubated with CdiI-His6 (5 µM) for 30 min at 23°C and an aliquot removed for gel analysis. Ni2+-NTA resin was then added and incubated for 1 h. The resin was then collected by centrifugation and the supernatant removed as the unbound or free fraction. The resin was washed with binding buffer, and the bound proteins eluted in binding buffer supplemented with 250 mM imidazole. Samples were analysed by SDS-PAGE and staining with Coomassie R-250.

CdiA-CTII1026b activity was examined in vitro using RNA isolated from Bt E264 cells. Purified CdiA-CTII1026b or the inactive Asp214Ala variant (at 1 µM final concentration) were incubated with 2 µg of total cellular RNA (isolated from Bt E264) in storage buffer supplemented with 1 mM MgCl2 for 1 h at 37°C. Where indicated, purified CdiI-His6 proteins (at 2 µM final concentration) were also included. Reactions were extracted with guanidinium isothiocyanate-phenol as described (Garza-Sánchez et al., 2006), and 0.5 µg of each RNA sample was analysed by polyacrylamide gel electrophoresis and Northern blot hybridization as described above.

CDI growth competition

Wild-type B. thailandensis E264 was used in competitions with various target cells derived from Bt E264. Target cells (Bt6, Bt7, Bt8, Bt9, Bt36 and Bt56) were grown to early log phase (OD600 ≈ 0.2–0.5) in LB medium. Cultures were adjusted to OD600 = 0.2, mixed at an inhibitor to target cell ratio of 1000 to 1, and 150 µl of the co-culture was plated onto LB agar followed by incubation at 37°C. At the indicated time points, cells were harvested from the plates with 1 ml of M9 salts, and viable cell counts determined. For competitions with the plasmid-borne CDIII1026b system, the Bp cdiAIB gene cluster was induced with 0.2% L-arabinose in M9 minimal medium (M9-Ara) before co-culture with target cells at a 100:1 inhibitor to target ratio. After different periods of incubation, cells were collected and viable cell counts determined by plating on selective media. Proximity dependence of growth inhibition was determined using a modified filter assay (Aoki et al., 2005). Inhibitor cells (Bt E264 carrying plasmid pJSW1-6) or mock inhibitor cells lacking the CDIII1026b CDI system (Bt E264 carrying plasmid pJSW2) were plated onto M9-Ara agar and nitrocellulose filters placed on top of the cells. Bt6 target cells were then plated on top of the filter, and the plate incubated at 37°C. After 24 h of incubation, the cells were recovered and plated onto LB agar supplemented with Kan to quantify viable target cells. Three technical replicates were conducted for each growth competition time point, and all experiments were repeated (biological replicate) at least once.


We are grateful to Herbert Schweizer and Tung Hoang for plasmids and strains, Bruce Braaten for technical assistance, and Sanna Koskiniemi and Bruce Braaten for helpful comments on the manuscript. This work was supported by the National Institutes of Health through Grant U54 AI065359 (A.T., P.S.K, C.S.H and D.A.L.) and Tri-Counties Blood Bank Postdoctoral Fellowship (S.K.A.). The content is the sole responsibility of the authors and does not necessarily represent the official views of the National Institute of Allergy and Infectious Diseases or the National Institutes of Health.