Precise Manipulation of the Clostridium difficile Chromosome Reveals a Lack of Association between the tcdC Genotype and Toxin Production

Abstract

Clostridium difficile causes a potentially fatal diarrheal disease through the production of its principal virulence factors, toxin A and toxin B. The tcdC gene is thought to encode a negative regulator of toxin production. Therefore, increased toxin production, and hence increased virulence, is often inferred in strains with an aberrant tcdC genotype. This report describes the first allele exchange system for precise genetic manipulation of C. difficile, using the codA gene of Escherichia coli as a heterologous counterselection marker. It was used to systematically restore the Δ117 frameshift mutation and the 18-nucleotide deletion that occur naturally in the tcdC gene of C. difficile R20291 (PCR ribotype 027). In addition, the naturally intact tcdC gene of C. difficile 630 (PCR ribotype 012) was deleted and then subsequently restored with a silent nucleotide substitution, or “watermark,” so the resulting strain was distinguishable from the wild type. Intriguingly, there was no association between the tcdC genotype and toxin production in either C. difficile R20291 or C. difficile 630. Therefore, an aberrant tcdC genotype does not provide a broadly applicable rationale for the perceived notion that PCR ribotype 027 strains are “high-level” toxin producers. This may well explain why several studies have reported that an aberrant tcdC gene does not predict increased toxin production or, indeed, increased virulence.

INTRODUCTION

Clostridium difficile causes a potentially fatal diarrheal disease through production of its principal virulence factors, toxin A and toxin B (20, 22). Understanding the genetic and molecular basis of C. difficile virulence will be a crucial step in combating the infection. However, Clostridium species are notorious for being genetically intractable. At present, insertional mutagenesis is the only form of genetic manipulation possible in C. difficile (5, 13, 14, 29). This can exert polar effects on genes near the site of insertion and does not permit the more refined genetic manipulations that are often required for robust functional genetic analyses and strain-engineering projects.

Precise genetic manipulation can be achieved via two-step allele exchange, in which both a positive selection marker and a counterselection marker are used (see Fig. S1 in the supplemental material). C. perfringens and C. thermocellum are the only Clostridium species for which counterselection markers have been described (2, 28, 35). However, these approaches employ genes with chromosomal homologues as counterselection markers, meaning that they can be used only in mutant background strains. In this work, the cytosine deaminase gene (codA) of Escherichia coli was developed as a heterologous counterselection marker for genetic manipulation of wild-type C. difficile strains. Cytosine deaminase (EC 3.5.4.1) catalyzes the conversion of cytosine to uracil, although its substrate specificity is sufficiently relaxed that it also converts the innocuous pyrimidine analog 5-fluorocytosine (FC) into the highly toxic 5-fluorouracil (FU). FU toxicity occurs via uracil phosphoribosyltransferase (EC 2.4.2.9), followed by a series of steps that result in irreversible inhibition of thymidylate synthase, a key enzyme in nucleotide biosynthesis, and misincorporation of fluorinated nucleotides into DNA and RNA (17, 21). CodA has been shown to confer FC sensitivity on eukaryotic cells (12, 25) and has been used in conjunction with uracil phosphoribosyltransferase (Upp) to generate unmarked gene deletions in Rhodococcus equi (36).

In this work, codA-mediated allele exchange was employed to systematically restore both the Δ117 frameshift mutation and the 18-nucleotide deletion that naturally occur in the tcdC gene of C. difficile R20291 (PCR ribotype 027) (see Fig. S2 in the supplemental material). Furthermore, the naturally intact tcdC gene of C. difficile 630 (PCR ribotype 012) was deleted and then restored with a silent nucleotide substitution, or “watermark,” so the resulting strain was distinguishable from the wild type. It has long been proposed that tcdC encodes a negative regulator of toxin production (18), and this notion has since been supported by in vitro protein interaction studies and qualitative functional genetic studies (4, 23). Therefore, increased toxin production, and hence increased virulence, is often inferred in strains of C. difficile with an aberrant tcdC genotype, particularly PCR ribotype 027 strains (4, 7, 23, 37). The notion that strains of C. difficile that produce more toxin are intrinsically more virulent is debatable (6, 24, 32, 39). However, to date, the limited capabilities of genetic tools have prevented a rigorous assessment of the exact influence the tcdC genotype has on the amounts of toxin A and toxin B produced by C. difficile. This has been addressed in the work described here, revealing a lack of association between the tcdC genotype and toxin production.

MATERIALS AND METHODS

Bacterial strains and routine culture conditions.

Bacterial strains and plasmids used in this study are detailed in Table 1. E. coli was cultured aerobically (37°C; shaking at 200 rpm) in LB medium supplemented with chloramphenicol (25 μg/ml) where appropriate. C. difficile was routinely cultured in BHIS medium (brain heart infusion [Oxoid] supplemented with 5 mg/ml yeast extract [Oxoid] and 0.1% [wt/vol] cysteine [Calbiochem]) (33) supplemented with d-cycloserine (250 μg/ml), cefoxitin (8 μg/ml), and thiamphenicol (Tm) (15 μg/ml) where appropriate. FC and FU selections were carried out on C. difficile minimal medium (CDMM) (5, 19) to ensure there were no exogenous pyrimidines in the medium that could act as competitive inhibitors. All C. difficile cultures were incubated at 37°C in an anaerobic workstation (D. Whitley, Yorkshire, United Kingdom).

Table 1.

Bacterial strains and plasmids

Strain/plasmid	Characteristics	Reference(s)/source
Strains
E. coli
TOP10	F− mcrA Δ(mrr-hsdRMS-mcrBC) φ80lacZΔM15 ΔlacX74 nupG recA1 araD139 Δ(ara-leu)7697 galE15 galK16 rpsL(Strr) endA1 λ−	Invitrogen
CA434	E. coli HB101 [F− mcrB mrr hsdS20(rB− mB−) recA13 leuB6 ara-14 proA2 lacY1 galK2 xyl-5 mtl-1 rpsL20(Smr) glnV44 λ−] with plasmid R702	16, 38
C. difficile
R20291	Wild type; PCR ribotype 027 (Stoke Mandeville, UK, outbreak strain)	J. Brazier, Anaerobe Reference Laboratory, Cardiff, United Kingdom
CRG1635	R20291 ΔtcdC	This study
CRG1639	R20291 tcdC::117A	This study
CRG3111	R20291 tcdC::117A + 18	This study
CRG2545	R20291 630tcdC	This study
630	Wild type; PCR ribotype 012	41
CRG2207	630 ΔtcdC	This study
CRG3109	630 tcdC[MfeI]	This study
Plasmids
pMTL540FT-CD	Source of codA expression construct	11
pMTL82151	E. coli-C. difficile shuttle vector (pBP1 catP ColE1 traJ)	15
pMTL83151	E. coli-C. difficile shuttle vector (pCB102 catP ColE1 traJ)	15
pMTL84151	E. coli-C. difficile shuttle vector (pCD6 catP ColE1 traJ)	15
pMTL-SC7215	pMTL82151 with codA	This study (Fig. 1B)
pMTL-SC7315	pMTL83151 with codA	This study (Fig. 1C)

General molecular biology techniques.

E. coli was transformed by electroporation using a Gene-Pulsar (Bio-Rad), as recommended by the manufacturer. Plasmid DNA was isolated using the QIAprep spin miniprep kit (Qiagen). Genomic DNA was isolated from C. difficile using the DNeasy blood and tissue kit (Qiagen) after sequential treatment of cells with lysozyme (10 mg/ml in phosphate-buffered saline [PBS]) at 37°C for 30 min, followed by SDS (10% [wt/vol]) at 65°C for 30 min. PCRs were carried out using Phusion High-Fidelity DNA polymerase (New England BioLabs) in Failsafe PCR buffer E (Epicentre). The oligonucleotide primers used are detailed in Table S1 in the supplemental material. Thermocycling conditions were 95°C for 3 min, followed by 30 cycles of 95°C for 30 s, an appropriate annealing temperature for the primer pair being used for 30 s, 72°C for 1 min 30 s, and a final extension of 72°C for 10 min. Restriction enzyme digests, ligation reactions, and agarose gel electrophoresis were carried out according to standard protocols (31). DNA was purified from agarose gels using the QIAquick gel extraction kit (Qiagen). Sanger sequencing was carried out by Source Bioscience, United Kingdom.

Plasmid vectors and allele exchange cassettes.

Plasmid pMTL84151::codA was constructed to test codA-mediated sensitivity to FC in C. difficile. The codA gene of E. coli was cloned as an EcoRI fragment from pMTL540FT-CD (11) into similarly digested pMTL84151 (15). Two different vectors were constructed for carrying out allele exchange in C. difficile: pMTL-SC7215 (Fig. 1B) and pMTL-SC7315 (Fig. 1C). These vectors were used in C. difficile R20291 and C. difficile 630, respectively. They are identical except for their Gram-positive replicons. Plasmid pMTL-SC7215 was constructed by removing the pCD6 Gram-positive replicon of pMTL84151::codA as an AscI/FseI fragment and replacing it with the pBP replicon from similarly digested pMTL82151 (15). Likewise, pMTL-SC7315 was constructed by removing the pCD6 Gram-positive plasmid replicon of pMTL84151::codA as an AscI/FseI fragment and replacing it with the pCB102 replicon from similarly digested pMTL83151 (15).

Fig 1.

Features of plasmid vectors used in this study. (A) Schematic of the codA translation initiation region. The SD sequence is boxed, and the GTG start codon of codA is in boldface. The solid and dashed brackets indicate spacer regions of 12 nucleotides for the codA open reading frame and 10 nucleotides for a competing open reading frame that encodes a 39-amino-acid peptide with no known function, respectively. (B) Plasmid vector pMTL-SC7215, used for allele exchange in C. difficile R20291 (GenBank accession no. JQ040260). (C) Plasmid vector pMTL-SC7315, used for allele exchange in C. difficile 630. Note that pMTL-SC7215 and pMTL-SC7315 are identical except that they have the pBP1 and pCB102 Gram-positive replicons, respectively (between AscI and FseI restriction sites). Half-arrows indicate annealing regions of primers used in this study.

The decision to use different Gram-positive plasmid replicons for allele exchange in R20291 and 630 was based on previous work in our laboratory in which pBP1 was used successfully to deliver heterologous sequence to the chromosome of C. difficile R20291 (5) and pCB102 was used successfully to deliver heterologous sequence to the chromosome of C. difficile 630 (14). Note that the low frequency of DNA transfer achieved by conjugation from E. coli into C. difficile means that it is not feasible to use suicide vectors (i.e., replication-defective plasmids) for delivery of heterologous sequence into the chromosome of C. difficile. Furthermore, no conditional plasmid replicon has been described for C. difficile to date. Therefore, “pseudosuicide” vectors have been used (5, 14, 29). “Pseudosuicide” is a term coined in our laboratory to describe a plasmid that replicates in the host cell but is segregationally unstable because it replicates at a rate that is lower than that of the host chromosome (5). Therefore, under selection for a plasmid-borne antibiotic marker, integrant clones can be enriched (rather than selected, as would be the case for a suicide vector) from a population of transconjugant cells (see Fig. S1 in the supplemental material). This is because the growth rate of integrant clones is not limited by the rate at which the pseudosuicide vector can replicate and segregate into daughter cells (as is the case for nonintegrant transconjugant cells).

Allele exchange cassettes were designed to have approximately 500 bases of homology to chromosomal sequence in both the up- and downstream regions flanking the sequence to be altered. They were synthesized by DNA2.0 and cloned individually into the PmeI site of pMTL-SC7215 or pMTL-SC7315, as appropriate. Four separate allele exchange cassettes were used in C. difficile R20291 (i) to generate an in-frame deletion in the tcdC open reading frame (ORF) by deleting nucleotides 61 to 653, inclusive, out of 680 in total (including the TAA stop codon) (ΔtcdC) (see Fig. S3A in the supplemental material); (ii) to repair the single-nucleotide deletion at position 117 (tcdC::117A) (see Fig. S3B in the supplemental material); (iii) to repair both the single-nucleotide deletion and the 18-nucleotide deletion (tcdC::117A + 18) (see Fig. S3C in the supplemental material); and (iv) to replace the whole tcdC ORF with that of C. difficile 630 (630tcdC) (see Fig. S3D in the supplemental material). Two separate allele exchange cassettes were used in C. difficile 630 (i) to generate an in-frame deletion in the tcdC ORF by deleting nucleotides 61 to 672, inclusive, out of 699 in total (including the TAA stop codon) (ΔtcdC) (see Fig. S4A in the supplemental material) and (ii) to subsequently restore the in-frame deletion and introduce the silent base substitution A420G as a “watermark” so the resulting strain would be distinguishable from wild-type 630 (tcdC[MfeI]) (see Fig. S4B in the supplemental material). This created an MfeI restriction site in the tcdC open reading frame without affecting the encoded amino acid sequence.

Plasmid transfer into C. difficile.

Plasmids were transferred into C. difficile by conjugation as described previously (5). Briefly, 1 ml of an E. coli CA434 overnight culture harboring the plasmid to be transferred was pelleted (4,000 × g for 1 min) and washed in PBS. The pellet was transferred to an anaerobic workstation, resuspended in 150 μl of an overnight C. difficile culture, and spotted onto BHIS agar. After 24 h of incubation, the conjugation mixture was harvested in 500 μl of PBS and spread across five BHIS plates, each supplemented with cycloserine, cefoxitin, and thiamphenicol. Transconjugant colonies were picked and restreaked onto the same medium when they became visible after 72 h of incubation.

Allele exchange protocol.

C. difficile transconjugants were restreaked on BHIS supplemented with cycloserine, cefoxitin, and Tm to enrich for and identify faster growing single-crossover integrant clones. Single-crossover clones were restreaked to purity on the same medium and then confirmed by PCR analysis (as detailed in Results). To allow rare second recombination events to occur, single-crossover clones were restreaked onto nonselective BHIS medium and incubated for 96 h. All growth was harvested in 500 μl PBS, serial dilutions were made (10⁻¹ to 10⁻⁶), and 100 μl of each dilution was plated onto CDMM supplemented with 50 μg/ml FC. After 48 h of incubation, FC-resistant clones were patch plated onto BHIS supplemented with cycloserine, cefoxitin, and Tm to screen for plasmid loss. Fluorocytosine-resistant (FC^r) thiamphenicol-sensitive (Tm^s) clones were analyzed by PCR to distinguish double-crossover recombinant clones from wild-type revertant clones (as detailed in Results), and Sanger sequencing was used to confirm the expected genotype.

Cell growth and toxin production assays.

Strains were cultured in tryptose-yeast (TY) medium (3% [wt/vol] Bacto tryptose, 2% [wt/vol] yeast extract, and 0.1% [wt/vol] thioglycolate, adjusted to pH 7.4) (10) to measure cell growth and production of toxin A and toxin B over a 72-h time course. Care was taken to ensure that cultures were inoculated with cells in the exponential phase of growth. To do this, C. difficile was cultured from storage at −80°C by streaking onto TY medium supplemented with 0.1% (wt/vol) sodium taurocholate (Sigma). After 16 h of incubation, three colonies were picked into 1 ml of TY medium and incubated for 8 h. Serial dilutions of this culture were made into fresh TY medium (10⁻¹ to 10⁻⁸) and incubated for 16 h. The culture setup with the most dilute inoculum that showed visible growth was then diluted 1 in 100 into fresh TY medium to start the assay. Typically, this was the 10⁻⁶ dilution for 630 and the 10⁻⁵ dilution for R20291. CFU were determined at 0, 3, 6, 9, 12, 24, 48, and 72 h by making serial dilutions and plating onto BHIS agar supplemented with 0.1% (wt/vol) sodium taurocholate (Sigma). At the same time points, samples of culture supernatant were harvested by centrifugation (12,000 × g for 1 min) and filtration (0.2-μm pore size), and toxin A and toxin B were measured by HT29 and Vero cell cytotoxicity assays.

HT29 and Vero cell cytotoxicity assays.

Toxin A and toxin B were measured by titrating out cytotoxic activity on HT29 (human colon carcinoma) and Vero (African green monkey kidney) cell monolayers. HT29 and Vero cells are reported to be more sensitive to toxin A and toxin B, respectively (34). Cell monolayers were prepared by seeding each well of a 96-well plate with 100 μl of cell suspension at a density of 2 × 10⁵ cells/ml. HT29 cells and Vero cells were cultured in McCoy's 5A medium and Dulbecco's modified Eagle's medium (DMEM), respectively, each supplemented with 10% (vol/vol) fetal calf serum and 1% (vol/vol) penicillin/streptomycin. The plates were incubated for 48 h (37°C; 5% CO₂) to let monolayers form before addition of C. difficile culture supernatants. Fourfold serial dilutions of each supernatant sample were made in PBS, and 20 μl of each dilution was added separately to the 100 μl of medium above HT29 and Vero cell monolayers (resulting in a further 1-in-6 dilution of each sample). After 24 h of incubation (37°C, 5% CO₂), the monolayers were examined by light microscopy (Nikon Eclipse TS100) to determine the toxin endpoint titer for each supernatant sample. The endpoint titer was defined as the first dilution in the series where cell morphology was indistinguishable from the negative controls. The results were expressed as “1/toxin endpoint titer” because there is an inverse relationship between the amount of toxin in a sample and the endpoint titer. Cell supernatants from isogenic toxin A⁻ B⁻ mutants were used as negative controls (20). Pure toxin A and toxin B (List Biological Laboratories) were used as positive controls, so that the detection limit of each assay could be determined in terms of the toxin concentration. The detection limit of the HT29 cell cytotoxicity assay was 100 pg/ml toxin A, and the detection limit of the Vero cell cytotoxicity assay was 25 pg/ml toxin B.

RESULTS

A counterselection marker for C. difficile.

Genome sequence analysis revealed that C. difficile does not have a homologue of codA but does have a homologue of upp, the gene encoding uracil phosphoribosyltransferase (C. difficile 630 GeneID, CD3479). This indicated that C. difficile would be inherently resistant to FC but sensitive to FU, a prerequisite for using codA as a counterselection marker. MIC assays on CDMM confirmed this, revealing that both C. difficile R20291 and 630 were resistant to a high concentration of FC (MIC, 1 mg/ml) but highly sensitive to FU (MIC, 1 μg/ml).

Plasmid pMTL84151::codA was constructed to assess the effect of codA expression on C. difficile sensitivity to FC. Initial attempts to clone codA into an expression context failed. Therefore, a previously described construct was used (11). Sequence analysis revealed it has a suboptimal spacer of 12 nucleotides between the Shine-Dalgarno (SD) sequence and the GTG start codon of codA. In addition, a competing ORF was identified (encoding a 39-amino-acid peptide with no known function) that has a more favorable spacer of 10 nucleotides between the SD sequence and an ATG start codon (Fig. 1A). Introduction of the codA construct into C. difficile reduced the MIC of FC on CDMM from 1 mg/ml (for the wild-type and pMTL84151 empty-vector control strains) to 10 μg/ml for both R20291 and 630. This confirmed that codA could be employed as a counterselection marker in C. difficile.

Systematic restoration of tcdC in C. difficile R20291.

Using codA as a counterselection marker, two-step allele exchange was carried out to systematically restore the mutated tcdC ORF of C. difficile R20291. In total, four tcdC recombinant strains were constructed; CRG1635, a tcdC-null in-frame deletion mutant (ΔtcdC); CRG1639, in which the single-nucleotide deletion at position 117 was repaired (tcdC::117A); CRG3111, in which both the single-nucleotide deletion and the 18-nucleotide deletion were repaired (tcdC::117A + 18); and CRG2545, in which the native tcdC ORF was replaced with that of C. difficile 630 (630tcdC) (Table 1).

Recombinant alleles were introduced into wild-type R20291 on the pseudosuicide vector, pMTL-SC7215 (Fig. 1B). Thiamphenicol-resistant (Tm^r) transconjugant colonies took 72 h to grow and were isolated at a mean frequency of 3.1 × 10⁻⁹ per CFU of E. coli donor (n = 8, two for each of the four recombinant tcdC alleles). Single-crossover integrants were easily identified in subsequent restreaks, as they yielded visibly larger colonies than transconjugants after 16 to 24 h on medium supplemented with Tm (Fig. 2A). This is because the growth of single-crossover integrants is not limited by the rate at which the pseudosuicide plasmid can replicate and segregate into daughter cells (see Materials and Methods for a detailed explanation). The maximum number of serial restreaks required to identify single-crossover integrant colonies from a transconjugant clone was three (for pMTL-SC7215::630tcdC). Integration of pMTL-SC7215 into the chromosome was confirmed by PCR, with one primer annealing to plasmid sequence (SC7-F) and the other annealing to chromosomal sequence (tcdA-Rs1), to obtain a product of 1.8 kb (Fig. 2B and C). The purity of single-crossover integrant clones was confirmed by the absence of a 1.8-kb PCR product when primers annealing to chromosomal sequence on either side of the site of plasmid integration were used (cdd-Fs1 and tcdA-Rs1) (Fig. 2B and D).

Fig 2.

Example of PCR screening for single-crossover integrant clones. (A) Single-crossover clones (some of which are indicated by arrows) were identified among transconjugant colonies based on their higher growth rate. (B) Schematic illustrating the chromosomal arrangement in a transconjugant versus a single-crossover integrant clone. The native chromosomal allele is presented as |X|Y|Z| and the recombinant allele as |X|*Y*|Z|. The half-arrows indicate primer-annealing regions. Note that primers cdd-Fs1 and tcdA-Rs1 anneal outside sequence with homology to the allele exchange cassette. (C) PCR screening of two independent single-crossover integrant clones using primers SC7-F and tcdA-Rs1. (D) PCR screening of the same samples using primers cdd-Fs1 and tcdA-Rs1. Control lanes: P, pMTL-SC7215; wt, wild-type C. difficile R20291 genomic DNA.

Pure single crossovers were streaked onto BHIS agar without selection to allow rare second recombination events to occur and the resulting plasmid to be lost from cells. In the subsequent step, FC^r colonies were isolated on CDMM at a frequency ranging from 2.3 × 10⁻⁷ to 1.4 × 10⁻⁴ (across the 8 experiments). To confirm loss of the plasmid-borne catP marker, FC^r colonies were patch plated onto medium supplemented with Tm. Between 27% and 100% (depending on the experiment) of FC^r clones were Tm^s, indicating that they were either double-crossover recombinants or wild-type revertants. The remaining FC^r clones were still Tm^r, suggesting that pMTL-SC7215 was still present. PCR and sequencing analysis of 10 of these FC^r Tm^r clones (selected at random) revealed that the plasmid was still integrated into the chromosome at the tcdC locus and that mutations were present in either the plasmid-borne codA gene or the chromosomal upp gene (data not shown). This follows, as loss-of-function mutations in either codA or upp would relieve FC-mediated counterselection.

FC^r Tm^s double-crossover recombinant tcdC clones were identified and characterized by PCR and sequencing analysis. PCR with primers cdd-Fs1 and tcdA-Rs1, which anneal to the chromosome outside regions with homology to the allele exchange cassettes, confirmed excision of pMTL-SC7215 from the chromosome in each recombinant strain (Fig. 3A and B). As expected, the product for CRG1635 (ΔtcdC) was 1.2 kb compared with 1.8 kb for the other tcdC recombinants and the 630 and R20291 control strains (Fig. 3B). Sequencing of these PCR products confirmed that each recombinant had the expected tcdC genotype (see Fig. S5 in the supplemental material). Allele-specific PCR yielded a product of 354 bp for CRG1639 (tcdC::117A), CRG3111 (tcdC::117A + 18), and CRG2545 (630tcdC), confirming that the Δ117 tcdC frameshift mutation had been corrected in these strains (Fig. 3C). PCR with primers 18F and 18R yielded a product of 191 bp for CRG1639 (tcdC::117A), confirming that the strain still had an 18-bp deletion in the tcdC ORF, but a product of 209 bp was amplified for both CRG3111 (tcdC::117A + 18) and CRG2545 (630tcdC), confirming that the 18 bp had been restored in these strains (Fig. 3D). Amplification of a 295-bp product from the CDR20291_0242 gene of R20291, which is absent in 630, confirmed that all four tcdC recombinant strains were indeed derived from C. difficile R20291 (Fig. 3E). Finally, PCR with M13F and M13R primers, which amplify a 1.7-kb product comprising the plasmid-borne codA gene (Fig. 1B), confirmed that pMTL-SC7215 was no longer present in any of the tcdC recombinant clones (Fig. 3F).

Fig 3.

PCR analysis of recombinant C. difficile R20291 strains. (A) Schematic of the tcdC locus in wild-type R20291. The half-arrows indicate primer-annealing regions. Note that primers cdd-Fs1 and tcdA-Rs1 anneal outside sequence with homology to the allele exchange cassettes. The naturally occurring Δ117A deletion and 18-bp deletion are shown above the arrow depicting the tcdC ORF. The triangle below the ORF indicates the deletion made in CRG1635 (ΔtcdC). (B to F) Lanes: 1, Plasmid pMTL-SC7215; 2, C. difficile 630; 3, C. difficile R20291; 4, CRG1635 (ΔtcdC); 5, CRG1639 (tcdC::117A); 6, CRG3111 (tcdC::117A + 18); 7, CRG2545 (630tcdC). PCR analysis was carried out with primer pairs cdd-Fs1 and tcdA-Rs1 (B); 117AS-F and 117AS-R, which specifically amplify a product of approximately 350 bp if the Δ117A deletion has been restored or is absent (C); 18F and 18R, which amplify a 191-bp product if the 18-bp deletion is present or a 209-bp product if the 18- bp deletion has been restored or is absent (D); CDR20291_0242-F1 and CDR20291_0242-R1, which amplify a 295- bp sequence that is present in R20291 but absent in 630 (E); and M13F and M13R, which amplify the codA gene in pMTL-SC7215 (F) (Fig. 1B).

Effect of tcdC genotype on C. difficile R20291 cell growth and toxin production.

Cell growth and toxin production were measured in TY medium for each tcdC recombinant strain over a 72-h time course. As expected, there were no differences in cell growth between strains (Fig. 4A). Surprisingly, however, there were also no differences in toxin production between strains (Fig. 4B and C).

Fig 4.

Cell growth and toxin production by recombinant C. difficile R20291 strains. The strains were cultured in TY medium for 72 h. (A) CFU were determined by serial dilution and plating. (B and C) Production of toxin A and toxin B was determined by titrating out cytotoxicity on HT29 cells (detection limit, 100 pg/ml toxin A) (B) and Vero cells (detection limit, 25 pg/ml toxin B) (C). The data points represent the means of three independent experiments (n = 3). The results were highly reproducible. The error bars are omitted for clarity.

Deletion and restoration of tcdC in C. difficile 630.

To ensure that the unexpected observations with C. difficile R20291 were not due to an unknown strain defect, the study was verified using wild-type C. difficile 630, which naturally has an intact tcdC ORF. Therefore, a tcdC in-frame deletion mutant was constructed and designated CRG2207 (Table 1). The ΔtcdC allele exchange cassette was designed to yield a deletion equivalent to that which had been made in C. difficile R20291. In anticipation of observing a difference in toxin production, a second round of allele exchange was carried out to restore the tcdC ORF in CRG2207. This was done instead of complementing in trans from a plasmid. The resulting strain was designated CRG3109 (Table 1). To distinguish it from wild-type 630, a “watermark” was introduced into the tcdC ORF by making a single-nucleotide substitution (A420G). The resulting ORF encodes exactly the same amino acid sequence as wild-type tcdC but has an MfeI site that is not present in the wild-type sequence (see Fig. S4 in the supplemental material).

Recombinant alleles were introduced into C. difficile 630 on pMTL-SC7315 (Fig. 1C). Otherwise, the same allele exchange procedure was followed as for R20291. Tm^r transconjugants were isolated at a mean frequency of 2.7 × 10⁻⁸ per CFU of E. coli donor (n = 4, two for each recombinant allele). Again, single-crossover recombinants were isolated easily by restreaking onto BHIS supplemented with Tm and identifying visibly larger colonies after 16 to 24 h of incubation. Subsequently, FC^r clones were isolated on CDMM at frequencies ranging from 4.2 × 10⁻⁷ to 7.1 × 10⁻⁵ (across the 4 experiments). Between 48% and 100% of these (depending on the experiment) were identified as double-crossover clones, based on sensitivity to Tm. FC^r Tm^s double-crossover clones were screened by PCR using primers cdd-Fs1 and tcdA-Rs1, which anneal to the chromosome outside regions with homology to the allele exchange cassette (Fig. 5A). As expected, CRG2207 (ΔtcdC) yielded a 1.2-kb PCR product, whereas wild-type 630 and CRG3109 (tcdC[MfeI]) yielded a 1.8-kb PCR product (Fig. 5B). Sequencing the PCR products confirmed that each strain had the expected tcdC genotype (see Fig. S6 in the supplemental material). In addition, restriction analysis of the same PCR products confirmed that the tcdC locus of CRG3109 has an MfeI site that is absent in wild-type 630 (Fig. 5C).

Fig 5.

PCR analysis of recombinant C. difficile 630 strains. (A) Schematic of the tcdC locus in wild-type 630. The half-arrows indicate annealing regions of primers. Note that primers cdd-Fs1 and tcdA-Rs1 anneal outside sequence with homology to the allele exchange cassettes. The triangle below the arrow depicting the tcdC ORF indicates the deletion made in CRG2207 (ΔtcdC). The approximate location of the MfeI site introduced into CRG3109 (tcdC[MfeI]) is shown above the ORF. (B) PCR analysis was carried out with primer pairs cdd-Fs1 and tcdA-Rs1. (C) The resulting products were purified, digested with MfeI, and resolved again on a 1% agarose gel. Lanes: 1, plasmid pMTL-SC7315; 2, C. difficile 630; 3, CRG2207 (ΔtcdC); 4, CRG3109 (tcdC[MfeI]).

Effect of the tcdC genotype on C. difficile 630 cell growth and toxin production.

Again, cell growth and toxin production were measured in TY medium for each tcdC recombinant strain over a 72-h time course. As expected, no differences in cell growth were observed between strains (Fig. 6A). However, in keeping with the R20291 results, there were also no differences in toxin production between strains (Fig. 6B and C).

Fig 6.

Cell growth and toxin production by recombinant C. difficile 630 strains. The strains were cultured in TY medium for 72 h. (A) CFU were determined by serial dilution and plating. (B and C) Production of toxin A and toxin B was determined by titrating out cytotoxicity on HT29 cells (detection limit, 100 pg/ml toxin A) (B) and Vero cells (detection limit, 25 pg/ml toxin B) (C). The data points represent the means of three independent experiments (n = 3). The results were highly reproducible. The error bars are omitted for clarity.

DISCUSSION

In this study, the codA gene of E. coli was successfully developed as a heterologous counterselection marker for C. difficile. Initial attempts to clone codA into an expression context failed. The codA gene is known to be tightly regulated in E. coli (8, 27, 30). Therefore, it was reasoned that deregulated expression of codA caused a lethal shift in the cellular pyrimidine pool from cytosine to uracil. Sequence analysis of the codA expression construct used revealed it has a suboptimal translation initiation region with a 12-base spacer between the Shine-Dalgarno sequence and a GTG start codon, as well as a competing ORF (Fig. 1A). Therefore, expression of codA in this context must be appropriately balanced to avoid a lethal shift in the cellular pyrimidine pool but to still confer sensitivity to FC. Interestingly, codA-mediated sensitivity to FC was relieved in both C. difficile R20291 and C. difficile 630 when rich medium (BHIS) was used in place of minimal medium (CDMM) (data not shown). This is likely to occur because free cytosine in the medium can act as a competitive inhibitor of CodA-mediated conversion of FC to FU and/or because expression of key enzymes required for FU toxicity is sufficiently downregulated in rich medium to nullify the toxic effect.

Using codA as a counterselection marker, two-step allele exchange was employed to manipulate the tcdC alleles of two C. difficile strains, R20291 and 630. Four independent recombinant strains were constructed in C. difficile R20291. In C. difficile 630, the naturally intact tcdC ORF was deleted in frame and subsequently restored. This ORF restoration approach is superior to complementing a mutant in trans using a multicopy plasmid, as it maintains genomic context and negates the need for an empty-vector control strain. Furthermore, it demonstrates that the codA counterselection marker and the positive selection marker (catP in this case) can be recycled, meaning that multiple genetic alterations can be made in the same strain of C. difficile.

Until now, the limited capabilities of genetic tools have prevented a rigorous test of the exact influence the tcdC genotype has on the amounts of toxin A and toxin B produced by C. difficile. Intriguingly, no association between the tcdC genotype and toxin production was found in this work in either C. difficile R20291 or C. difficile 630. Any question that this finding arises due to some artifact of the codA allele exchange method can be ruled out because in-frame deletion of the master regulator of sporulation (spo0A) in C. difficile R20291 results in the expected Spo⁻ phenotype (see Fig. S7 in the supplemental material). At first, our finding seems to be at odds with qualitative functional genetic studies that have found TcdC to be a negative regulator of toxin production (4, 23). However, it may be that current thinking simply requires slight modification and that TcdC may act as a “safety catch” to guard against inappropriate expression of toxin, rather than to affect the amount of toxin produced per se. It is also possible that our results arise because the tcdC ORF is not expressed correctly in either R20291 or 630. While this seems unlikely, we cannot rule it out. Either way, there is certainly no association between the tcdC genotype and production of toxin A and toxin B by either C. difficile R20291 or C. difficile 630 under the conditions tested in this work (i.e., in TY medium). This may well explain why several studies have reported that an aberrant tcdC genotype does not predict increased toxin production or, indeed, increased virulence (4, 6, 26, 32, 39). Notwithstanding this, it should be emphasized that the Δ117 tcdC frameshift mutation still provides a convenient basis upon which to make a presumptive identification of PCR ribotype 027 strains (1, 3, 9, 40), even though this itself does not provide a broadly applicable rationale for the perceived notion that PCR ribotype 027 strains are “high-level” toxin producers.