Saturation mutagenesis of a CepR binding site as a means to identify new quorum-regulated promoters in Burkholderia cenocepacia


E-mail; Tel. (+1) 607 279 3886; Fax (+1) 607 255 3906.


Burkholderia cenocepacia is an opportunistic pathogen of humans that encodes two genes that resemble the acylhomoserine lactone synthase gene luxI of Vibrio fischeri and three genes that resemble the acylhomoserine lactone receptor gene luxR. Of these, CepI synthesizes octanoylhomoserine lactone (OHL), while CepR is an OHL-dependent transcription factor. In the current study we developed a strategy to identify genes that are directly regulated by CepR. We systematically altered a CepR binding site (cep box) upstream of a target promoter to identify nucleotides that are essential for CepR activity in vivo and for CepR binding in vitro. We constructed 34 self-complementary oligonucleotides containing altered cep boxes, and measured binding affinity for each. These experiments allowed us to identify a consensus CepR binding site. Several hundred similar sequences were identified, some of which were adjacent to probable promoters. Several such promoters were fused to a reporter gene with and without intact cep boxes. This allowed us to identify four new regulated promoters that were induced by OHL, and that required a cep box for induction. CepR-dependent, OHL-dependent expression of all four promoters was reconstituted in Escherichia coli. Purified CepR bound to each of these sites in electrophoretic mobility shift assays.


The genus Burkholderia encompasses a fascinating collection of diverse β-proteobacteria (Coenye and Vandamme, 2003). This genus includes over 50 species, some of which are potentially useful in bioremediation, while other members are capable of forming nitrogen-fixing root nodules with legumes (Chen et al., 2003; Bontemps et al., 2010). Some members protect host plants against fungal pathogens, while others are themselves pathogenic against plants, animals and humans (Coenye and Vandamme, 2003; Jones and Webb, 2003). Seventeen pathogenic species are members of the Burkholderia cepacia complex, or BCC (Vandamme et al., 1997; Vanlaere et al., 2008; 2009), two of which are described by the Center for Disease Control as category B select agents (Godoy et al., 2003).

Burkholderia cenocepacia, previously known as B. cepacia genomovar III (Vandamme et al., 2003), is recognized as an opportunistic pathogen of humans and is a particular threat to cystic fibrosis (CF) patients (Vandamme et al., 1997; Mahenthiralingam et al., 2005). Colonization of the CF lung by B. cenocepacia (Vandamme et al., 2003) tends to occur in patients already infected with Pseudomonas aeruginosa, another opportunistic pathogen of the CF lung (Vandamme et al., 1997; Jones and Webb, 2003). An infection caused by both organisms can result in serious clinical complications. B. cenocepacia strains are resistant to most antibiotics, making them virtually impossible to eradicate (Nzula et al., 2002). Infections with B. cenocepacia may have variable clinical outcomes ranging from asymptomatic carriage to a sudden fatal deterioration in lung function (Mahenthiralingam et al., 2005).

Four strains of B. cenocepacia have been sequenced in their entirety, one of which is described in a publication (Holden et al., 2009). The Joint Genome Institute is currently sequencing nine additional strains ( All four sequenced isolates have three circular chromosomes that vary in size between 3.9 and 0.88 MB in length. Strains J2315 and HI2424 also have one plasmid, 93 KB and 165 KB in length respectively.

Many or possibly all Burkholderia spp. encode at least one regulatory system that resembles the LuxR and LuxI proteins of Vibrio fischeri, where LuxI synthesizes an acylhomoserine lactone (AHL)-type pheromone, also called an autoinducer, and LuxR is an AHL-dependent transcriptional regulator (Eberhard et al., 1981; Engebrecht and Silverman, 1984; Choi and Greenberg, 1992). Regulatory systems of this family are found in countless proteobacteria, where they are thought to allow individual bacteria to co-ordinate their physiology with their siblings. Collectively, these systems regulate diverse processes, including pathogenesis, biofilm formation, bacterial conjugation and the production of antibiotics and other secondary metabolites (Whitehead et al., 2001). In general, target genes are transcribed preferentially at population densities high enough to favour AHL accumulation (Eberhard et al., 1991), a phenomenon referred to as quorum sensing (Fuqua et al., 1994). Burkholderia thailandiensis has three such systems, one of which is implicated in cell aggregation, while another is required for antibiotic production (Chandler et al., 2009; Duerkop et al., 2009). A plant growth promoting isolate of Burkholderia ambifaria uses quorum sensing to regulate the production of the anti-fungal compound pyrrolnitrin (Schmidt et al., 2009).

LuxR-type proteins have two domains, an N-terminal pheromone binding domain and a C-terminal DNA binding domain (Pappas et al., 2004). Purified LuxR, TraR of Agrobacterium tumefaciens and LasR of P. aeruginosa, when complexed with their respective AHLs, bind with high specificity to recognition sequences (referred to as lux, tra or las boxes, respectively) that are found at target promoters (Zhu and Winans, 1999; Schuster et al., 2004; Urbanowski et al., 2004). LasR is also able to bind to sequences that have no obvious resemblance to canonical las boxes. A few members of this family bind DNA only in the absence of AHLs (Cui et al., 2005; Fineran et al., 2005; Minogue et al., 2005; Castang et al., 2006; Sjoblom et al., 2006).

Burkholderia cenocepacia J2315 encodes three LuxR homologues and two LuxI homologues (Lewenza et al., 1999; Malott et al., 2005; 2009). Among these, CepR and CepI appear to be well conserved within the BCC (Venturi et al., 2004). CepI synthesizes primarily octanoylhomoserine lactone (OHL), and lower levels of hexanoylhomoserine lactone (Lewenza et al., 1999; Gotschlich et al., 2001; Huber et al., 2001; Aguilar et al., 2003a). Null mutations in cepI or cepR increased the production of the siderophore ornibactin, and decreased the production of secreted lipases and metalloproteases ZmpA and ZmpB (Lewenza et al., 1999; Lewenza and Sokol, 2001; Sokol et al., 2003; Kooi et al., 2006). CepI and CepR are also required for swarming motility and biofilm formation (Huber et al., 2001) and for pathogenicity in several animal models (Kothe et al., 2003; Sokol et al., 2003). B. cenocepacia also expresses the CciI and CciR proteins, which are encoded on a genomic island called cci (cenocepacia island), that is associated with epidemic strains (Malott et al., 2005). The CepIR and CciIR systems extensively interact, in that CciR negatively regulates cepI, while CepR is required for expression of the cciIR operon (Malott et al., 2005). Transcriptional profiling studies indicate that CepR and CciR regulate many of the same genes, but do so in opposite ways (O'Grady et al., 2009). B. cenocepacia also encodes an orphan LuxR homologue called CepR2, which represses a cluster of genes that may direct production of an antibiotic or other secondary metabolite (Malott et al., 2009).

In addition to transcriptional profiling several other approaches have been used to identify genes whose expression is influenced by CepR and/or OHL. In one study, the proteome of a wild-type B. cenocepacia was compared with that of a cepR mutant. Fifty-five proteins were found to be differentially expressed in the two strains, approximately 10% of all detected proteins (Riedel et al., 2003). In a second study, fragments of a B. cepacia strain were cloned into a promoter trap plasmid and introduced into an E. coli strain that expressed CepR (Aguilar et al., 2003b). Twenty-eight promoter fragments were identified as being induced by OHL, and in all cases, induction required CepR. In a third study, a library of B. cenocepacia DNA fragments were introduced into a plasmid containing a promoterless luxCDABE operon (Subsin et al., 2007). That study identified 58 OHL-inducible promoters and 31 OHL-repressible promoters. Regulation of nine of these genes required CepR, while the others were not tested. Seven OHL-inducible genes were identified by screening a library of lacZ fusions (Weingart et al., 2005). Induction of all of these genes required CepR. Purified CepR-OHL complexes bound with high affinity and specificity to specific DNA sequences at two target promoters (Weingart et al., 2005). These binding sites contained a 16 nucleotide imperfect dyad symmetry and were centred approximately 44 nucleotides upstream of the transcription start sites. These two sites are to date the only experimentally confirmed CepR binding sites.

Most of the studies described above do not distinguish whether a target promoter is controlled by CepR directly or indirectly. CepR could regulate a promoter indirectly, for example, by directly regulating an unknown regulatory gene whose product directly regulates that promoter. Alternatively, a CepR mutation might perturb cellular physiology in such a way that various promoters are affected by secondary effects.

To date, the most comprehensive study attempting to define the optimal CepR binding site was done by Chambers, Sokol and colleagues (Chambers et al., 2006), who approached this question with an impressive combination of genetics and bioinformatics. Mutagenesis of the known CepR binding site within the cepI promoter completely abolished induction (Chambers et al., 2006). The promoters of six genes known to be induced by OHL were used to formulate a consensus CepR binding motif (Chambers et al., 2006). This information was used to test eight additional candidate promoters, six of which were CepR-regulated. Ultimately, 10 inducible promoters were used to refine the consensus sequence, and 57 possible CepR binding sites were identified upstream of various genes.

The consensus motif identified in the Chambers study included the sequence CTG-N10-CAG, which has dyad symmetry. However, several other bases in the consensus did not preserve this symmetry, and some of those non-symmetric bases were said to be highly conserved (Chambers et al., 2006). The partial dyad symmetry suggests that CepR binds DNA as dimer and that the two DNA binding domains have rotational symmetry. Although we have no proof of this, structural studies of a related protein support this idea (Vannini et al., 2002; Zhang et al., 2002). Several other LuxR-type proteins are thought to decode dyad symmetrical sequences (Whitehead et al., 2001; White and Winans, 2007; Antunes et al., 2008). In the present study, we tested the 10 putative CepR binding sites described above for the ability to bind purified CepR-OHL complexes. We also systematically resected and mutated a known CepR binding site, and use the resulting information to identify four new promoters that are regulated directly by CepR. All four promoters are regulated by CepR in vivo, require their binding sites for regulation, and bind with high affinity to CepR-OHL in vitro.


Binding of CepR to 10 known or putative CepR binding sites

As described above, another group identified 10 putative CepR binding sites that lay upstream of CepR-regulated promoters (Chambers et al., 2006). However, these DNA sequences were not tested for CepR binding, and were not shown to be required for CepR-dependent gene expression in vivo. We have had previous experience with CepR-OHL complexes in electrophoretic mobility shift assays (EMSA) and DNase footprinting (Weingart et al., 2005) and therefore assayed CepR binding to these 10 sequences. The 10 sites include the two that we had previously tested, one in the cepI promoter and the other in the aidA promoter. Synthetic double-stranded oligonucleotides 38 base pairs long were radiolabelled, combined with purified CepR-OHL complexes in the presence of a 10 000-fold excess (mass/mass) of calf thymus DNA, and size fractionated using native gel electrophoresis. CepR bound to the 38-mer containing the cepI binding site with an affinity of approximately 65 nM, in reasonable agreement with previous estimates (Fig. 1). It also bound to the fragment containing the aidA cep box, although significantly more weakly, also as expected (Weingart et al., 2005). To our surprise, of the eight other DNA fragments tested, only one was detectably shifted, even using very high concentrations of CepR (Fig. 1). The shifted fragment was that of the phuR promoter, which appeared to form two shifted complexes. We also tested the consensus sequence that was derived in the Chambers study (Chambers et al., 2006). Because this consensus was only 18 nucleotides in length, we used flanking sequences that were derived from the cepI cep box. This consensus sequence was bound by CepR with relatively high affinity (Kd = 210 nM, Fig. 1). We provide evidence that the sequences flanking the 18 nucleotide sequence contributed to binding affinity (see below).

Figure 1.

Electrophoretic mobility shift assays of putative CepR binding sites. Synthetic double-stranded DNA fragments 38 nucleotides long were end-labelled, combined with purified CepR-OHL complexes, and size-fractionated using native PAGE. In panel 1 and 11, CepR-OHL was added at the following concentrations (lanes 1–5): 0 nM, 21 nM, 65 nM, 210 nM and 650 nM. In all other panels, CepR-OHL was added at the following concentrations (lane 1–5): 0 nM, 650 nM, 2060 nM, 6500 nM and 20 600 nM. Free DNA is indicated with a filled triangle and shifted complexes are indicated using an arrow.

Systematic mutagenesis of a CepR-dependent promoter

In a previous study, a series of 4 nucleotide alterations were made in the CepR binding site of the cepI promoter, and all such mutations that disrupted any part of this binding site blocked promoter activation by CepR (Chambers et al., 2006). In the current study, we have extended those findings by altering single nucleotides within and beyond this site (Fig. 2). We altered each of these bases to its complement, fused the resulting promoters to lacZ, and measured their activity in vivo. We used a broad range of OHL concentrations, as low concentrations could unmask subtle phenotypes, while high concentrations might allow us to detect residual induction of strongly defective promoters.

Figure 2.

Mutagenesis of individual nucleotides within and adjacent to the CepR binding site of the cepI promoter. Bases are numbered relative to the centre of the axis of rotational symmetry, which is indicated using inverted arrows. Fusions were introduced into strain K56 I2, which lacks CepI. Strain K56 and its derivatives are naturally Lac-. Expression of each PcepIlacZ fusion was measured for β-galactosidase specific activity (Miller, 1972) in the presence of the indicated concentrations of OHL.

We were especially interested in the six conserved dyad symmetrical sequences (CTG-N10-CAG, positions −8 to +8, see Fig. 2). Alteration of bases −8, −7 or −6 strongly inhibited the activation of this promoter by CepR, especially at low or intermediate OHL concentrations (Fig. 2). The same was true of alteration of bases +6, +7 or +8. Alteration of most of the ten central residues (from −5 to +5) inhibited activation at low OHL concentrations but had less effect at moderate or high concentrations. The A-3T mutation was an exception, as it had little effect on expression at any OHL concentration.

Mutations at positions −9 or −10 caused moderate loss of induced expression, while mutations at +9 or +10 showed more severe defects. One explanation could be that mutations at +9 or +10 might alter sites required for RNA polymerase binding. The cep box is centred 44.5 nucleotides upstream of the transcription start site (Weingart et al., 2005), so any −35 motif should lie directly adjacent to the cep box. At positions +8 to +13 there is the sequence GTTACA, which resembles the E. coli consensus TTGACA at four positions. To learn whether these bases serve as a −35 sequence, we made several mutations to either weaken or strengthen the similarity to the E. coli consensus sequence. The mutation A+8T and T+10G were predicted to strengthen the −35 motif. However, neither single mutation nor a double mutation increased the basal or induced expression (Fig. 2). The mutation A+11C was predicted to decrease expression, but surprisingly, had the opposite effect. An A+13C was also predicted to weaken a −35 sequence, but had only a mild defect. These data strongly suggest that the sequence GTTACA is not a functional −35 site. However, the mutations in this region did have fairly strong phenotypes, and may therefore contribute in an undefined way to promoter expression.

Identification of the minimal CepR binding site

Because the data described above were obtained using whole cells, defects in gene expression could have a number of explanations. We wished to focus specifically on DNA sequences required for CepR-OHL binding, and therefore measured the ability of purified CepR-OHL complexes to bind altered cep boxes in vitro. We first synthesized complementary oligonucleotides that resected the cep box of the cepI promoter from the left (upstream) edge in three nucleotide intervals (Fig. 3A, only one strand of each DNA duplex molecule is shown). Fragments A1–A3 were tightly bound (Fig. 3A), while fragment A4 was more weakly bound, and fragment A5 was not detectably bound. The CTG-N10-CAG is present in fragments A1–A4 and partially absent in fragment A5. The observation that fragment A4 contains this sequence yet binds CepR rather poorly suggests that additional DNA sequences are required for high-affinity binding (see below). A similar set of resections were made starting from the right side of the CepR binding site, with similar results (Fig. 3B).

Figure 3.

Identification of the minimal DNA sequences sufficient for high-affinity binding of CepR-OHL. Synthetic oligonucleotides containing the indicated sequences were hybridized with complementary fragments, radiolabelled, combined with CepR, and size-fractionated by native PAGE. Part A: DNA resections from the left; Part B: DNA resections from the right and Part C: DNA resections from both sides. Numbers 1–5 in each gel correspond to numbers at the left of each oligonucleotide sequence. CepR was added to a final concentration of 0 nM, 90 nM, 360 nM, 1440 nM and 5780 nM in the five lanes of each gel. Free DNA is indicated with a filled triangle, non-specific control DNA lacking a cep box is indicated with an open triangle, and shifted complexes are indicated using an arrow. The core cep box is indicated using inverted arrows.

We used the information described above to define the smallest DNA fragment capable of binding CepR-OHL. We synthesized a series of complementary oligonucleotides of different lengths and tested each for CepR-OHL binding. Fragments C1 and C2 bound CepR with high affinity, forming well-defined shifted fragments (Fig. 3C). Fragment C3 also bound CepR, as little or no free DNA was detected in the presence of higher protein concentrations. However, most of the shifted DNA formed a smear rather than a well-defined complex, suggesting that binding was less stable than with longer fragments. Fragments C4 and C5 bound CepR poorly or not at all. All five fragments contain the 16 nucleotide sequence CTG-N10-CAG. These data suggest that CepR detects this 16 nucleotide sequence, but requires as many as five additional bases on each side for maximal affinity.

Contribution of each base to binding affinity in vitro

As a working hypothesis, we assumed that the CepR DNA binding domain probably has twofold rotational symmetry and that the DNA binding sites should therefore have dyad symmetry (see above). Identifying a fully symmetric binding site would greatly facilitate mutagenesis studies, as we could then construct duplex DNA molecules each containing two copies of the same oligonucleotide. We therefore set out to identify a high-affinity CepR binding site that was perfectly symmetric.

We constructed two DNA fragments, designated L-L′ and R′-R, which are fully symmetrical, and designed using the left half and right half of the wild-type cep box of the cepI promoter respectively (Fig. 4). Fragment L-L′ bound CepR with threefold higher affinity than the wild-type sequence, and appeared to form well-focused complexes rather than smears (Fig. 4, middle panel). This finding provides further evidence that CepR binds DNA as a rotationally symmetric dimer. In contrast, fragment R′-R bound CepR weakly (Fig. 4, bottom panel). This indicates that sequences within the left (promoter-distal) half-site of the wild-type cep box contribute more to the overall binding affinity than do the sequences within the right (promoter-proximal) half-site. Fragment L-L′ was therefore used as a starting point for systematic mutagenesis.

Figure 4.

CepR binding affinity for fully symmetrical CepR binding sites. The sequence designated L-L′ is composed of sequences of the native cep box of the cepI promoter from positions −11 to −1 followed by the inverse complement of these sequences. Fragment R′-R is similar but contains sequences of the native cep box from positions +1 to +11, and their inverse complements at positions −11 to −1. DNA fragments were radiolabelled and combined with CepR-OHL in the following concentrations (lanes 1–10) 0 nM, 0.6 nM, 1.8 nM, 6 nM, 18 nM, 60 nM, 180 nM, 580 nM, 1830 nM and 5780 nM. Complexes were size-fractionated by native PAGE. Kd values were calculated by determining the amount of CepR-OHL required to shift half of the DNA fragments.

We constructed a total of 33 variations of fragment L-L′, in which each nucleotide of the two half-sites was altered to each of the three other possible bases. This is exemplified in Fig. 5, in which nucleotide −11 and +11 of the sequence are simultaneously altered, preserving the overall symmetry of the DNA molecules. Each of these fragments was self-annealed and the resulting duplex molecules were tested for CepR binding by EMSA. The alteration from A−11 to G and T+11 to C caused very little if any change in affinity, while the other two alterations caused modest decreases in binding.

Figure 5.

Alteration of nucleotides −11 and +11 of fragment L-L′ to all possible bases. Oligonucleotides containing altered sequences were self-annealed, incubated with CepR and size-fractionated. The concentrations of CepR-OHL were the same as in Fig. 4.

Alterations at nucleotides −10 and +10 showed moderate effects on binding affinity, depending on the specific alteration (Figs 6 and 7A). Two alterations of nucleotides −9 and +9 had severe effects on CepR binding while the third alteration caused only a subtle defect. All alterations of positions −8 and +8, positions −7 and +7 and positions −6 and +6 completely blocked detectable binding. All alterations at positions −5 and +5 also caused severe defects.

Figure 6.

Systematic alteration of each nucleotide of fragment L-L′ to all possible bases and their effects on CepR binding affinity. Each of the nucleotides is fully self-complementary and therefore can form duplex DNA molecules. Each was radiolabelled, and analyzed for CepR-OHL binding affinity exactly as shown in Fig. 5. n.d. indicates that binding was not detected.

Figure 7.

Summary of binding affinities of cep boxes having the indicated alterations. Only the left half-site is shown. Wild-type binding affinity is indicated using a dashed horizontal line.
A. Affinity of wild type and mutant cep boxes for CepR-OHL.
B. Affinity of wild type and mutant tra boxes for TraR of A. tumefaciens is shown for comparison (White and Winans, 2007).

Alterations of positions −4 and +4 were interesting in that two of the three alterations actually increased CepR binding affinity (Figs 6 and 7A). Apparently, the residues A−4 and T+4 of fragment L-L′ were not optimal, and that T or C residues at position −4 (and A or G residues at position +4) are preferred. Alterations at positions −3 and +3, positions −2 and +2 and positions −1 and +1 caused highly variable effects, depending on the incorporated base.

CepR causes a DNA bend at its binding site

As described above, the critical bases required for CepR binding extend from nucleotide −9 to −5 (and +5 to +9 on the opposite half-site, see Fig. 7A). This is in some ways quite different from the sequences decoded by TraR of A. tumefaciens (White and Winans, 2007), where nucleotides −6 to −4 (and +4 to +6) were essential (Fig. 7B). Structural studies showed hydrogen bonding between TraR and these bases (Vannini et al., 2002; Zhang et al., 2002). Therefore, the critical bases detected by CepR are located considerably further from the dyad axis than those bases detected by TraR. Binding each of the half-sites of the cep box might require that the DNA recognition helices of CepR be further apart from each other than their counterparts in TraR. Alternatively or additionally, CepR might impart a higher angle DNA bend to a cep box than does TraR to a tra box, in effect bringing the two half-sites of DNA closer together. TraR induces a 30° bend in this DNA sequence (Vannini et al., 2002; Zhang et al., 2002; Pappas and Winans, 2003), and this model predicts that CepR might impart a greater bend.

To determine whether CepR causes a DNA bend at its binding site, we introduced a DNA sequence containing a consensus cep box into plasmid pBEND3, a plasmid that facilitates the study of intrinsic or protein-induced DNA bending (Zwieb and Brown, 1990). This plasmid allows the creation of a set of DNA fragments that are the same length and circularly permuted. The cep box will therefore lie near one end of some fragments and closer to the centre of other fragments. If CepR causes a DNA bend, then complexes having this bend near the centre of the fragment will migrate in electrophoretic gels more slowly than complexes whose bend is closer to one end of the fragment (Zwieb and Brown, 1990). We digested the resulting plasmid individually with three different restriction endonucleases, added sufficient CepR to shift these fragments and size-fractionated these complexes by native PAGE (Fig. 8). For comparison, a similar analysis was done using the TraR binding site and purified TraR.

Figure 8.

Assays for a CepR-directed DNA bend.
A. The optimal CepR binding site, as determined in Fig. 5 (ACCCTGTCAGATCTCACAGGGT) was introduced into plasmid pBEND3, and the resulting plasmid was digested individually with restriction endonucleases BamHI (lane 1, 2), EcoRV (lane 3) or MluI (lane 4), combined with CepR-OHL (lanes 2–4), and size-fractionated by native PAGE.
B. A similar plasmid was constructed containing the consensus tra box, cut with BamHI (lane 5, 6) EcoRV (lane 7) or MluI (lane 8). Fragments were combined with TraR-OOHL (lanes 6–8), and size-fractionated by native PAGE.
C. A map of the multiple cloning site of the pBEND3 derivatives containing a cep box or a tra box.

The mobility of the EcoRV-generated fragment was considerably lower than fragments generated by the other two enzymes (Fig. 8A). This is diagnostic of a DNA bend. An identical analysis using TraR and tra box DNA showed a somewhat similar result (Fig. 8B). However, the differences in mobility were less pronounced, indicating a lower angle DNA bend. Using the equation of Thompson and Landy (Thompson and Landy, 1988), we estimate the bend angle to be approximately 45° for CepR and 30° for TraR. Structural studies of TraR-DNA complexes agree with this estimate (Vannini et al., 2002; Zhang et al., 2002). These data indicate that CepR causes a higher angle DNA bend than does TraR, confirming our predictions.

Identification of new CepR-regulated promoters

The enoLOGOS web server (Workman et al., 2005) was used to obtain a pictorial representation of the most favoured bases in a canonical cep box. The dissociation constants in Fig. 6 were used as input, and the few mutant sequences that were not detectably bound by CepR were arbitrarily assigned a dissociation constant of 1 mM, which is eightfold weaker than the weakest detected binding. The resulting cep box logo is shown in Fig. 9. The EnoLOGOS web server also converted these dissociation constants into a log-likelihood matrix (Table 1). This matrix was used to obtain a similarity score between the cep box logo and every 22 nucleotide sequence in the B. cenocepacia genome. This was done using the MOODS algorithm (Korhonen et al., 2009), which provided a list of 237 possible cep boxes and scored their similarity to the consensus sequence (Table S1).

Figure 9.

A cep box logo, derived from the experimental dissociation constants for mutant cep boxes and enoLOGOS.

Table 1.  Log-likelihood matrix.
PositionACGTOptimal sequenceScore for Optimal SequnceLeast optimal sequenceScore for least optimal sequencecepI.ScoreaidAScore
  1. The log-likelihood matrix (columns 1–4) were obtaining using the enoLOGOS web server, and the following parameters: Weight type: Probabilities (Ka = s); GC ratio: 67%.

Totals−38.73 159.2 −5.89 −1.315

Many of the DNA sequences that resemble the cep box logo lie far from any predicted promoter. However, 142 of these sites lie within 300 nucleotides upstream of a predicted translation start site (Table S1). The expression of most of these genes is not affected by a CepR mutation (O'Grady et al., 2009). However, 43 genes that have possible nearby cep boxes are differentially expressed at least twofold by a mutation in cepR (Table S1). Of these 43 genes, 13 were chosen for further analysis (Table 2). These 13 were chosen for a variety of reasons. Two were cepI and aidA, which served as positive controls. We also chose the cepR promoter, which was said to be autorepressed in one study (Lewenza and Sokol, 2001) but not in another (O'Grady et al., 2009). BCAM0188 encodes CepR2, which if induced would show a quorum cascade. Two other genes are adjacent to BCAM0188. BCAM1869 was chosen because it lies adjacent to and divergent from cepR. BCAM1413a and BCAM1414 were chosen because they encode three AidA homologues. The promoter of pBCA055 was found to be induced in an earlier study from our lab (Weingart et al., 2005).

Table 2.  Induction of B. cenocepacia genes by OHL.
GeneSequenceScorecep boxNo OHLOHL (1 nM)RatioOHL (1 µM)Ratio
  1. nd, not determined.

BCAL0510CGCCCGCCAGAATTGACAGGCC6.338Full138 ± 2469 ± 1173.5747 ± 735.4
Half351 ± 27167 ± 400.5303 ± 160.9
BCAM0186 (NRPS)ACCCTGTGATTTGATGCCGGTC9.398Full43 ± 797 ± 62.3771 ± 6717.9
Half63 ± 669 ± 31.1762 ± 3412.1
BCAM0188 (cepR2)ATCCTGTTCAAAAGGACAGTTT−1.875Full543 ± 134743 ± 1411.41452 ± 1052.6
Half669 ± 23876 ± 271.31678 ± 2232.5
BCAM0189 (NRPS)ATCCTGTTCAAAAGGACAGTTT−1.875Full180 ± 17150 ± 170.8360 ± 162
Half170 ± 5123 ± 90.7329 ± 401.9
BCAM1413a (aidC)TACCTGTCAGGTTTGATGGGGG6.26Full134 ± 6145 ± 61.1408 ± 263.1
Half173 ± 16201 ± 81.1217 ± 291.2
BCAM1414TACCTGTCAGGTTTGATGGGGG6.26Full80 ± 2383 ± 81.03175 ± 522.1
Half101 ± 893 ± 90.9114 ± 101.1
BCAM1868 (cepR)ACGCTGTCATACTTGTCAGGTT−8.188Full205 ± 6193 ± 40.94224 ± 161.1
BCAM1869ACGCTGTCATACTTGTCAGGTT−8.188Full179 ± 131598 ± 1938.93536 ± 12519.7
Half121 ± 12153 ± 31.3609 ± 455
BCAM1870 (cepI)ACCCTGTAAGAGTTACCAGTTA−5.885Full60 ± 3655 ± 2510.92489 ± 16241.5
Half40 ± 1639 ± 30.9133 ± 423.3
BCAS0155-0153ATACTGTTAAAACCGGCAGGTT−9.521Full67 ± 472 ± 61.1176 ± 442.6
Half87 ± 761 ± 40.7114 ± 101.3
BCAS0156ATACTGTTAAAACCGGCAGGTT−9.521Full110 ± 22518 ± 644.71399 ± 14412.7
Half101 ± 784 ± 60.8116 ± 141.1
BCAS0293 (aidA)AAGCTGTAAAAGTAAACAGGTC−1.315Full109 ± 9416 ± 183.81648 ± 6115.1
Half118 ± 8108 ± 70.9174 ± 321.6
pBCA055-054 (bqiCD)CCACTGTCAAATCTACGAGGGC2.799Full149 ± 49775 ± 1935.23126 ± 105220.9
Half167 ± 6167 ± 11272 ± 1061.6

Each of the 13 putative promoters was fused to lacZ on a multicopy plasmid. Each fragment extended upstream just far enough to include the putative CepR binding site plus 6–8 extra nucleotides. We also constructed 13 similar fragments that included only the downstream half of the predicted binding sites. All 26 plasmids were introduced into the cepI mutant strain K56-2I and tested for induction of β-galactosidase in the absence of OHL or in the presence of two OHL concentrations, one that causes approximately half-maximal induction of cepI and aidA (Weingart et al., 2005), and one that causes maximal expression.

Of the 13 plasmids containing full CepR binding sites, 12 were predicted to be induced by OHL, as one (cepR) was predicted to be repressed. Of these twelve, seven were induced at least fourfold by 1 µM OHL (Table 2). Of these seven, induction of six was abolished or severely reduced in isogenic plasmids containing only half of the putative cep box (Table 2). The exception was the BCAM0186 promoter, which was strongly induced even with only half of its cep box. This gene lies close to BCAM0188 (cepR2), and may be regulated by the CepR2 protein (Malott et al., 2009). Of the six promoters whose OHL induction required a cep box, two were previously characterized (cepI and aidA), while four were new [BCAL0510, BCAM1869, BCAS0156 and pBCA055-4 (bqiCD)]. Three of these four genes were previously shown to be induced by CepR, and to be unaffected by mutations in CciR or CepR2 (Malott et al., 2009; O'Grady et al., 2009), although those studies did not show whether CepR acted directly or indirectly. The fourth is BCAS0156, which was not previously reported to be induced.

The four newly identified CepR-regulated promoters are divided among the four replicons. BCAL0510 lies on the largest chromosome and its product resembles a group of hypothetical proteins (data not shown). BCAM1869, lies on chromosome 2 and is adjacent to and divergent from cepR (Fig. S1). The cep box lies 114 nucleotide upstream of the BCAM1869 translation start site. Induction was reduced, although not abolished, by removing the promoter-distal half of the cep box. BCAM1869 and cepR are adjacent in many species of Burkholderia, providing suggestive evidence that their proteins are in some way functionally linked. A protein homologous to BCAM1869 was recently shown to play a role in transcription regulation (Mattiuzzo et al., 2010). The cepR gene was previously identified as being autorepressed in one study (Lewenza and Sokol, 2001), although in another study it was found to be unregulated by CepR (O'Grady et al., 2009). Our data support the latter study (Table 2).

BCAS0156, which lies on chromosome 3, resembles a family of β-lactamases and other penicillin-binding proteins (pfam00144). pBCA055 (bqiC) and pBCA054 (bqiD) lie on the 93 KB plasmid and appear to be expressed as an operon (Fig. S2). pBCA055 is a multidomain protein whose central domain resembles GGDEF proteins (COG2199) and whose C-terminal domain resembles EAL proteins (pfam00563), and may therefore synthesize and or degrade c-di-GMP (Romling and Amikam, 2006). pBCA054 has a C-terminal domain that resembles the DNA binding domains of LuxR proteins (pfam00196), suggesting that it may be a transcription factor.

The fact that these six promoters required OHL and a putative CepR binding site for induction suggested that CepR might directly activate them. To provide additional evidence, we determined whether CepR and OHL could activate these promoters in a heterologous host lacking any other LuxR-type protein. We introduced into E. coli strain MC4100 plasmids containing each of the six inducible promoters as well as plasmids containing the same promoters but containing only part of the CepR binding site. These strains also contained a second plasmid expressing CepR or a vector control. Strains containing CepR were cultured in the presence or absence of OHL and assayed for β-galactosidase. All six strains containing CepR and a full cep box showed strong induction of β-galactosidase by OHL (Table 3). Strains whose plasmids lacked the full cep boxes were either uninduced or weakly induced by OHL, as expected. Strains lacking CepR were not induced by OHL, also as expected. We conclude that for each promoter, induction by OHL requires CepR and a CepR binding site. The reconstitution of CepR-dependent induction of these promoters in E. coli provides further evidence that each is directly regulated by CepR.

Table 3.  Activation of CepR-regulated promoters in the heterologous host E. coli.
Genecep box+CepRRatio−CepRRatio
No OHL(1 µM)
BCAM1870 (cepI)Full4 ± 0.2491 ± 6.21284 ± 0.7120
Half4 ± 0.212 ± 2.734 ± 1.13
BCAS0293 (aidA)Full4 ± 0.2176 ± 17.7444 ± 0.743
Half4 ± 0.18 ± 1.023 ± 0.33
BCAL0510Full4 ± 0.149 ± 3.3134 ± 0.014
Half4 ± 0.111 ± 0.435 ± 0.92
BCAM1869Full3 ± 0.1211 ± 6.9654 ± 0.352
Half4 ± 0.229 ± 0.774 ± 0.18
pBCA055-054 (bqiCD)Full4 ± 0.1159 ± 13.4443 ± 0.248
Half4 ± 0.14 ± 0.513 ± 0.01
BCAS0156Full4 ± 0.241 ± 1.3103 ± 0.412
Half4 ± 0.13 ± 0.113 ± 0.21
(No insert) 4 ± 0.33 ± 0.114 ± 0.21

We used EMSA to test whether purified CepR could bind these sites in vitro. Synthetic duplex oligonucleotides 26 base pairs in length were combined with highly purified CepR-OHL complexes and size-fractionated by native PAGE (Fig. 10). All of the six inducible promoters were shifted, indicating that they can be bound by CepR. These data are therefore consistent with the hypothesis that these four genes are directly regulated by CepR-OHL complexes.

Figure 10.

CepR binding affinity for all known cep boxes. Duplex DNA molecules 26 nucleotides in length containing the sequences shown in Fig. 10 were radiolabelled, combined with purified CepR-OHL, and the resulting complexes were size-fractionated by native PAGE. For the cepI cep box (upper left panel), CepR-OHL was added at the following concentrations (lanes 1–5): 0 nM, 21 nM, 65 nM, 210 nM and 650 nM. In all other panels, CepR-OHL was added at the following concentrations (lane 1–5): 0 nM, 650 nM, 2060 nM, 6500 nM and 20 600 nM.


Members of our laboratory have previously used genetic, biochemical and structural approaches to study interactions between another LuxR-type protein, TraR, and its DNA binding site. TraR binds DNA as a dimer, and the two DNA binding domains of each dimer have twofold rotational symmetry. The DNA binding site also has twofold rotational symmetry, and complexes between the TraR-CTD and DNA also have rotational symmetry. This type of symmetry is found in the binding sites of other LuxR-type proteins, and of a great number of other DNA binding proteins. The imperfect dyad symmetry of two CepR binding sites leads us to believe that CepR would follow a similar pattern. It was therefore interesting that the consensus sequence previously described had several highly conserved asymmetric bases (Chambers et al., 2006). Unfortunately, only three of the ten putative bindings sites was detectably bound by purified CepR-OHL. Our studies suggest that a core CTG-N10-CAG is critical for binding. Loss of one of these six bases may be tolerated, but loss of any two or more probably is not. Of the 10 DNA sequences compiled in the Chambers study, three had all six bases of this consensus and were bound by CepR-OHL, while seven sequences lacked between one and four of these bases, and were not bound. The consensus sequence identified previously (Chambers et al., 2006) also has all six of these bases, and was bound. In the Chambers study, there seemed to be an implicit assumption that all CepR-inducible genes would be induced directly by CepR. In light of the present data, it seems more likely that induction of some genes could occur indirectly.

In order to study specifically how CepR decodes its binding site, we felt it was important to move to an in vitro system using purified components. The finding that 26 nucleotides are needed for full binding affinity was somewhat surprising, as this sequence would extend 2.5 helical turns, or 1.25 turns per half-site. It is far from clear why such a long sequence would be required. CepR does not strongly discriminate specific sequences at positions −11, −10, +10 or +11 (Figs 6 and 7), suggesting that the need for bases far removed from the dyad centre may not be sequence-specific.

We also tested a set of 22 nucleotide, perfectly symmetric sequences based upon the left half-site of the cepI cep box. Mutation of any of the bases from −9 to −5 (and the corresponding bases at +5 to +9) either abolished or strongly impaired binding affinity. From these data, we conclude that the core CepR binding site could be amended from the 16 nucleotide sequence CTG-N10-CAG to the 18 nucleotide sequence CCTGT-N8-ACAGG. We had previously discounted the bases at positions −9, −5, +5 and +9 as they are not part of the dyad symmetry of the two cep boxes previously aligned (Weingart et al., 2005). However, these bases are somewhat conserved in a new set of CepR binding sites (see Fig. 11).

Figure 11.

Alignment of experimentally confirmed CepR binding sites of the indicated promoters. In the top panel, the entire sites are aligned, while in the lower panel, the left half-sites are aligned with each other and with the inverse complement of all the right half-sites. The occurrence of each base at each position is tabulated and a consensus sequence for the full site and for the half-site is indicated.

The effects of mutations within the central spacer (−4 to +4) had more variable effects on binding affinity. Mutations from A to C or T at position −4 (and from T to G or A at position +4) enhanced affinity, indicating that the original L-L′ sequence was not optimal at this position. Many of the newly identified cep boxes contain the bases C or T at position −4 and a G or A at position +4 (Fig. 11). We denote the cep box residues from −4 to +4 as the central spacer, and predict that there are no sequence-specific protein–DNA contacts in this region, as was shown for TraR (Vannini et al., 2002; Zhang et al., 2002). It is well established that non-contacted bases can have large effects on protein affinity, generally via effects on the helical pitch or on the flexibility of the DNA, or by imparting a sequence-directed DNA bend (Sarai and Kono, 2005). This phenomenon is sometimes referred to as “indirect readout”, while sequence decoding by direct proteinBDNA interactions is called “direct readout”.

The identification of the bases essential for CepR binding facilitated a search for new genes that could be regulated directly by CepR. Of the thirteen promoters that were tested, six were significantly induced by OHL and required the putative cep box for this induction. These six CepR-regulated genes are distributed across all three chromosomes and the 92 KB plasmid (Figs S1 and S2).

BCAS0293 (aidA) was reported previously to be OHL-regulated (Aguilar et al., 2003a; Riedel et al., 2003; Weingart et al., 2005; Chambers et al., 2006). This gene is in a two-gene operon, and the downstream gene, aidB, is homologous to aidA. This homology extends to three additional genes that we designate aidC, aidD and aidE (BCAM1413a, BCAM1412, and BCAM1414, respectively, see Fig. S3), none of which was induced more than 2–3 fold by OHL (Table 2). BCAM1413a and BCAM1414 are divergent and flank a cep box, although the score of this site is weak (Table 2). The roles of these proteins are unknown, although AidA was previously reported to play a role in the slow killing of the nematode Caenorhabditis elegans (Huber et al., 2004). All five proteins are members of the PixA protein family (pfam12306).

In this study, we also examined the regulation of gene BCAM0186. This gene was unusual in that it was strongly induced by OHL and had a possible cep box, yet induction was cep box-independent. In hindsight, this should not have been surprising, as the similarity between this putative cep box and the consensus is rather weak, and the cep box lies almost 600 nucleotides upstream of the BCAM0186 translation start site. Its expression was reported to be inhibited by the product of the nearby BCAM0188 (cepR2) (Malott et al., 2009). We hypothesize that BCAM0186 may be directly repressed by CepR2, and that repression might be blocked by OHL.

Experimental procedures

Bacterial strains and growth conditions

Strains used in this study are described in Table S2. As needed, B. cenocepacia was cultured in 100 µg ml−1 of trimethoprim or 300 µg ml−1 of tetracycline. E. coli strains were cultured with 15 µg ml−1 of tetracycline, 100 µg ml−1 of streptomycin or 100 µg ml−1 of ampicillin.

Plasmid pYWN302 was constructed by digesting pKP302 (Pappas and Winans, 2003) with NsiI, and inserting an NsiI fragment containing the tet gene of pBBR-MCS3. This tet gene was created by PCR amplification using oligonucleotides described in Table S3. Plasmid pAFM113 was constructed by cloning the cepR gene into the pSRKKm broad-host range vector that had been digested with NdeI and HindIII.

Systematic mutagenesis of the cepI promoter

Single-site mutations in the cepI promoter were constructed by site-directed mutagenesis using oligonucleotides listed in Table S3 (IDT, Coralville, Iowa). PCR fragments containing the desired mutations were cloned into the KpnI and PstI sites of pYWN302, introduced into E. coli strain DH5α by transformation, checked by automated DNA sequencing (Cornell Biotechnology Resource Center), and then introduced into B. cenocepacia strain K56 I2 by electroporation (Cangelosi et al., 1991). Transformants were cultured in LB medium at 37°C to mid log phase (OD600 of 0.3–0.4), then diluted 20-fold into LB medium containing 0, 0.1 nM, 1 nM or 1 µM OHL. Cultures were incubated at 37°C with aeration until an OD600 of approximately 0.5, and then assayed for β-galactosidase specific activity (Miller, 1972). B. cenocepacia is naturally Lac-. Experiments were performed in triplicate with three different isolates of each strain.

Identification of the minimal CepR binding site

Synthetic oligonucleotides containing the sequences indicated in Fig. 2 were hybridized with their complementary sequences by heating to 95°C for 15 min and cooling to 40°C over 4 h in steps of 5°C. Double-stranded oligonucleotides were radiolabelled using gamma-P32-ATP and T4 DNA Kinase, and combined with purified CepR-OHL complexes at the indicated concentrations. CepR-OHL was purified as described (Weingart et al., 2005). Binding reactions contained CepR-OHL in the indicated concentrations and 10−12 M DNA fragments in a buffer containing 10 mM Tris-HCl (pH 7.9), 1 mM EDTA, 1 mM DTT, 60 mM potassium glutamate, 30 µg ml−1 calf thymus DNA, 20 µg ml−1 BSA and 10% glycerol, and were incubated for 30 m at 4°C. EMSA were conducted at 4°C using 8% PA gels containing TAE buffer, and visualized using a Storm Phosphoimager (Model 840, Molecular Dynamics). Self-complementary oligonucleotides were self-hybridized, and radiolabelled using the same procedures. Dissociation constants were calculated by determining the amount of CepR required to shift 50% of the DNA. We confirmed that this calculation is independent of the amount of added DNA (data not shown).

Measurement of DNA bending by CepR

Synthetic oligonucleotides containing a consensus cep box or tra box are described in Table 2. They were allowed to self-anneal by heating to 95°C for 5 min and then gradually cooled to room temperature, then digested simultaneously with XbaI and SalI, and introduced into plasmid pBEND3 (Zwieb and Brown, 1990) after digestion with the same enzymes. The resulting plasmids, pGR110 (containing a cep box) and pGR111 (containing a tra box), were digested individually with BamHI, EcoRV or MluI to permute the position of the box with respect to the DNA termini, radiolabelled using T4 polynucleotide kinase (New England Biolabs) and [gamma-P32]-ATP and purified using Centri-Spin columns (Princeton Separations). Binding reactions (10 µl) contained 800 ng DNA and clarified lysates containing overexpressed CepR-OHL or TraR-OOHL in a buffer containing 10 mM Tris-HCl (pH 7.9), 1 mM EDTA, 1 mM DTT, 60 mM potassium glutamate, 30 mg ml−1 calf thymus DNA, 2 × 0 mg ml−1 BSA, and 10% glycerol, and were incubated for 30 min at 4°C, size-fractionated using a 14% polyacrylamide gel in TAE buffer.

Identification of new CepR-regulated promoters

The enoLOGOS (Workman et al., 2005) web server was used to construct a log-likelihood matrix from a set of EMSA-derived Kd values for set of 22 nucleotide sequences shown in Fig. 6. The Kd value for the reference sequence and each of its variants (Table S4) were converted to Ka = s (association constants, Table S5), which were entered into the programme. EnoLOGOS returned a log-likelihood matrix (Table 1), calibrated for a GC content of 67%, and scaled by a factor of −1 to conform to the usual sign convention for binding energies. DNA sequences that resemble the cep box logo were identified using the log-likelihood matrix and the MOODS algorithm (Korhonen et al., 2009). This algorithm created a set of over eight million 22 nucleotide sequences derived from the B. cenocepacia genome, each overlapping its nearest neighbours by 21 nucleotides. Each of these sequences was compared with the canonical cep box using the log-likelihood matrix. The optimal and least optimal sequences received scores of approximately −39 and 160 (Table 1), respectively, while the two confirmed CepR binding sites, upstream of the cepI and aidA promoters, received scores of −5.9, and −1.3 respectively.

Promoters containing suspected CepR binding sites were PCR amplified using oligonucleotides shown in Table S5. For each promoter, a fragment containing a complete CepR binding site was amplified, as well as a similar fragment containing only the promoter-proximal half of the site. These PCR fragments contained a KpnI site at the promoter-distal end, and a PstI site at the promoter-proximal end. The PCR fragments were purified by using QIAquick Gel Extraction Kits (Qiagen) and digested with KpnI and PstI (New England Biolabs). The digested fragments were cloned into the promoter probe plasmid pYWN302, generating transcriptional fusions between each promoter and lacZ. The resulting plasmids were introduced into B. cenocepacia strain K56-I2 or E. coli strain MC4100 by electroporation. For assays of β-galactosidase, B. cenocepacia and E. coli strains were cultured in LB medium supplemented with either 300 µg ml−1 of tetracycline or 100 ug ml−1 kanamycin and 12 ug ml−1 tetracycline, respectively, overnight at 37°C. Each culture was diluted 1:100 into LB medium containing the indicated concentrations of OHL, and incubated with aeration at 37°C to an OD600 of 0.4, and assayed for β-galactosidase specific activity (Miller, 1972). Experiments were performed in triplicate with three different isolates of each strain.


We thank the members of our lab for helpful discussions and critical review of this manuscript. We also thank Anatol Eberhard for providing acylhomoserine lactones. This study was supported by a grant from the NIGMS (GM042893).