Effect of phage infection on toxin production by Clostridium difficile

Clostridium difficile is a spore-forming anaerobe and the causative agent of pseudomembranous colitis and many cases of antibiotic-associated diarrhoea in hospital patients (Bartlett et al., 1978). Toxins A and B are major virulence factors causing enterotoxicity and cytotoxicity, respectively, to cells (Lyerly et al., 1985). The secretory mechanism of both toxins is unknown; the toxins lack signal peptides, suggesting that an alternative to the type II pathway of the general secretory pathway is present (Mukherjee et al., 2002). A toxin-associated gene, tcdE, which lies in the pathogenicity locus (PaLoc) between tcdA and tcdB, was shown to possess a phage holin-like function when expressed in Escherichia coli. As phage holins are involved in cell membrane disruption (Wang et al., 2003), TcdE was speculated to play a role in toxin release through formation of membrane lesions (Tan et al., 2001). We wished to investigate whether the presence of prophage would enhance toxin production in C. difficile. To date, production of C. difficile toxins has not been shown to be regulated or encoded by temperate phages, although the PaLoc possesses characteristics of a mobile genetic element (Braun et al., 1996; Hammond & Johnson, 1995).

Previous studies of the potential effects of phages on C. difficile toxin production are limited to the description of two phages that did not convert non-toxigenic strains to toxin production (Mahony et al., 1985). In our study, four temperate phages that were induced from three clinical C. difficile isolates were tested for toxin conversion without success, although tcdE was detected in phage DNA. Toxin levels of phage-infected strains (stable lysogens) were also examined, followed by an investigation of tcdA and tcdB transcription in lysogens.

Bacteria and bacteriophages.
Fifty-six C. difficile isolates were isolated from stool samples of diarrhoeal patients at Sir Charles Gairdner Hospital (SCGH), Western Australia, as previously described (Riley et al., 1994). Overnight broth cultures were induced with 3 µg mitomycin C ml¹ for phage, as previously described (Nagy & Foldes, 1991). Three of the 56 isolates harboured temperate phages as determined by plaque assay: φC2 from CD242, φC5 from CD578, and φC6 and φC8 from CD371 (Table 1). Isolates that produced a high number of plaques for each phage were determined through phage assay (see below) and used as propagating strains. All C. difficile isolates used in this study are shown in Table 1. Phages were purified as described by Sambrook et al. (1989) and determined to be Myoviridae (φC2, φC5 and φC8) and Siphoviridae (φC6) phages (unpublished data).

Table 1. C. difficile isolates used for phage induction, propagation and infection, and in toxin assays

Culture media, bacterial growth and bacteriophage propagation.
C. difficile isolates were cultured in brain heart infusion broth (BHIB; Oxoid) for overnight or 2.53 h (exponential phase) at 37 °C in 80 % N₂, 10 % CO₂ and 10 % H₂ (Don Whitley D3161). Cultures for toxin assays were incubated for 48 h. Anaerobe basal agar (ABA; Oxoid), either as base (1 % ABA) or overlay (0.74 % ABA, 0.01 M Ca²⁺ and 0.4 M Mg²⁺), was used for phage assays, culture of lysogens and viable counts. Phage was propagated and assayed on solid media as described previously (Adams, 1959). Briefly, a phage suspension (500 µl of 10⁶10⁷ p.f.u. ml¹) was mixed with an indicator culture (4 ml of 10⁸ c.f.u. ml¹) and overlaid onto 1 % ABA for phage propagation, or 10 µl phage was spotted onto plates seeded with 600 µl indicator for phage assay.

Isolation of stable lysogens of φC2, φC6 and φC8.
Five C. difficile isolates were infected with the four phages, producing six stable lysogens of φC2, φC6 and φC8 (Table 1). φC5 did not lysogenize these isolates. Centres of turbid plaques were stabbed with a sterile wire and cultured on blood agar (BA) for 48 h. Single colonies arising from phage-resistant cells within the plaque, due either to mutation or to resistance conferred by prophage, were induced with mitomycin C and assayed with the uninfected strain to determine the presence of prophage. Stability of lysogens after 70 °C storage was tested by mitomycin C induction and colony blots using phage genome probes labelled with digoxigenin (Roche), as previously described (Karcher, 1995). Stable lysogens were named by the parental isolate number followed by the phage type, e.g. lysogen CD382C2 was host strain CD382 lysogenized by φC2.

DNA extraction from C. difficile and phages.
An overnight broth culture (10 ml) of C. difficile was concentrated in 1 ml TE buffer, treated with 2 µg lysozyme ml¹ at 37 °C for 30 min followed by 1 % (v/v) SDS, 40 mM EDTA and 500 µg proteinase K ml¹ at 55 °C for 12 h, and 100 µg RNase A ml¹ at 37 °C for 30 min. DNA was extracted twice with equal volumes of phenol : chloroform : isoamyl alcohol (25 : 24 : 1) followed by chloroform, precipitated with 0.3 M sodium acetate (pH 5.2) and ethanol, and purified (Wizard Clean-up system by Promega). Extraction of phage DNA was carried out as previously described (Sambrook et al., 1989).

Detection of toxin A and B levels in uninfected hosts and lysogens.
Broth cultures of C. difficile were standardized to obtain 19x10⁵ c.f.u. ml¹ and filtered (0.22 µm; Acrodisc). Crude toxin in the filtrate was serially diluted twofold and toxin A and B were assayed by ELISA (Techlab) and cell culture, respectively. One toxin A unit was defined as the reciprocal of the ELISA end point dilution to logarithmic base 2. Toxin A assays were performed once only due to reagent costs. Toxin B was assayed by inoculating 20 µl crude toxin onto Vero cells with a seeding concentration of 50 000 cells per well (Mahony et al., 1989). Cells were observed after 48 h and the minimum cytotoxic titre (MCT), defined as the reciprocal of the dilution that resulted in 50 % of cell death, was determined by neutral red assay (Mahony et al., 1989). MCTs were compared by mean and variance of cytotoxic titre units, which was defined as the logarithmic base 2 of the MCT (Freeman & Wilcox, 2003; Ketley et al., 1984). Assays were done in triplicate and repeated twice.

Detection of toxin genes in phage DNA by PCR.
Primers for amplification of tcdA (NK 2/3) and tcdB (NK 104/105) were described previously (Kato et al., 1991). A PCR reaction of 30 µl consisted of 1 µl template DNA, 0.18 µM of each primer, 200 µM dNTP, 2.5 mM MgCl₂, 1x reaction buffer and 0.75 U Taq gold polymerase (Perkin Elmer). Cycling conditions were 94 °C for 9 min, 35 cycles of 95 °C for 20 s and 55 °C for 120 s, then 74 °C for 5 min. Primers for tcdE (KST 1/2) were also described previously (Tan et al., 2001). The 20 µl PCR reaction consisted of 7 µl template DNA, 0.2 µM of each primer, 200 µM dNTP, 2 mM MgCl₂, 1x reaction buffer and 0.75 U Taq gold polymerase (Perkin Elmer). Cycling conditions were 94 °C for 10 min, 45 cycles of 94 °C for 30 s, 55 °C for 30 s and 72 °C for 1 min, and 72 °C for 7 min. Primer sequences are shown in Table 2. Chromosomal DNA from four toxigenic and non-toxigenic isolates served as control for amplification of toxin genes.

Table 2. Oligonucleotides used in this study

Detection of toxin genes by Southern hybridization.
Phage and bacterial DNA was digested by HindIII and/or XbaI (Promega), separated in 1 % agarose, transferred to a nylon membrane (Amersham Pharmacia Biotech) (Sambrook et al., 1989) and fixed by heating in a microwave for 2.5 min (Angeletti et al., 1995). Southern hybridization was carried out with the DIG High Prime DNA Labelling and Detection Starter Kit I (Roche), according to the manufacturer's instructions. Purified (Wizard Clean-up system; Promega) undigested phage DNA was DIG-labelled for use as probes (Roche). Probes for tcdA, tcdB and tcdE were prepared from PCR-amplified products of C. difficile VPI 10463. Optimal hybridization temperatures were determined from dot blots using appropriate DNA controls. Southern hybridizations were repeated three times.

Colony hybridization.
Lysogens grown on BA in a matrix were transferred onto a nylon membrane and lysed as previously described (Karcher, 1995). DNA was fixed and hybridized to phage probes as described above.

Extraction of bacterial RNA and real-time RT-PCR.
RNA extraction from a 10 ml overnight culture of C. difficile was carried out using reagents supplied in Totally RNA (Ambion) according to the manufacturer's instructions. Relative quantitative real-time RT-PCR was performed using the ABI Prism 7000 Sequence Detection System (Applied Biosystems) on RNA extracted from uninfected (parental) and infected (lysogenic) C. difficile isolates. ToxA-283/360 and ToxB-23/101 primers were designed to flank their TaqMan probes TOXA-315 and TOXB-48 (Table 2), respectively. TaqMan probes were labelled 5' with the reporter dye 6-carboxyfluorescein, and 3' with minor groove binder groups for increased sequence specificity (Kutyavin et al., 2000). The reaction volume was 20 µl, which consisted of 1x Superscript reaction mix (Invitrogen), 2.8 mM MgCl₂, 10 mM DTT, 0.3 U Superscript III RT/Platinum Taq Mix (Invitrogen), 10 U RNaseOUT (Invitrogen), 0.3 µM of each primer, 0.2 µM TaqMan probe and 40 ng template RNA. RT-PCR reactions were carried out at 50 °C for 30 min, 94 °C for 5 min, and for 45 cycles of 94 °C for 10 s, 55 °C for 10 s, and 68 °C for 30 s. A sample was used as calibrator RNA, where serial dilutions resulting in 2.44, 9.77, 39 and 156.25 ng RNA per reaction were amplified. The C_T values were used in generating a standard curve for relative quantification of samples (ABI PRISM 7000 Sequence Detection System User Guide). All reactions were carried out in triplicate.

Phage did not convert non-toxigenic strain to toxin production
C. difficile strain CD382 was toxin B negative by cell culture and a common host to all four phages. CD382 was infected with phages for isolation of stable lysogens, which were tested for toxin B production through cell culture. The presence of tcdA and tcdB in DNA was also determined through PCR. CD382C2, CD382C5, CD382C6 and CD382C8 did not have a cytotoxic effect on Vero cells. DNA of three single colonies of each lysogen was extracted for PCR amplification with NK 2/3 (tcdA) and NK 104/105 (tcdB) primers. All samples were negative for tcdA and tcdB.

Detection of toxin genes in phage DNA
tcdA and tcdB were not detected in phage DNA by PCR or Southern hybridization (not shown). Probes were chosen for the non-repeating region of tcdA and tcdB in the interest of detecting them separately. The non-repeating nucleotide regions in tcdA and tcdB encode the enzymic domain (Barroso et al., 1994; Faust et al., 1998). Although the phages did not harbour these sections of the toxin genes, it is possible that sequences related to the repeating units were present. Nucleotide repeating sequences at the 3'-end of both genes (Barroso et al., 1994; Dove et al., 1990) encode repeating units responsible for carbohydrate binding in TcdA and receptor binding and translocation in TcdB (Pfeifer et al., 2003).

Despite a lack of association between the toxin genes and phage DNA in the above experiments, there are two intriguing reports of homology between tcdA and genomes of phages belonging to other species. The toxin A gene was reported to be homologous to an unknown gene in phage φCT2 of Clostridium tetani (Canchaya et al., 2002). Lactobacillus casei phage A2 ORF22 was hypothesized to have a sie function (Proux et al., 2002). Homology between ORF22 and toxin A has been reported although no details of its extent were provided (Proux et al., 2002). Alignment of ORF22 (GenBank accession no. NC_004112) to TcdA of VPI 10463 (GenBank accession no. X9298) and variant strains (GenBank accession nos AJ011301, AF279457, AF217291, AJ132669 and Y10689) did not show significant similarity. However, alignment of ORF22 to TcdC of VPI 10463 showed 55 % identity over the last 103 aa of TcdC with an e-value of 2e22, but no significant similarity to the truncated TcdC of the variant strain 8864 (GenBank accession no. Y10689). Such homology between TcdA and TcdC to phage proteins suggests phage origins of toxin genes. It is interesting to note that ORF22 was absent in the variant strain with a truncated TcdC. Homology of phage genes to tcdB has not been described; however, part of the φC2 genome, sequence 584 (GenBank accession no. AY522336), had low homology to tcdB (unpublished data). Matching bases were distributed between nucleotide positions 227 and 630 of tcdB, which is the enzymic domain or the non-repeating section of the gene, but outside of the region amplified by the PCR primers. No significant similarities or alignments were identified between a putative ORF in sequence 584 and TcdB to suggest any phage function in toxin B production. This may explain the lack of hybridization between the tcdB probe and φC2 DNA.

The tcdE gene was detected in φC2, φC5 and weakly in φC8 by PCR (not shown). To verify these results, Southern hybridizations of tcdA, tcdB and tcdE probes to phage genomes were performed. The tcdE probe hybridized to a band of expected size (5.1 kb) in VPI 10463, as well as to three other bands (18.6, 12.6 and 7 kb) not predicted to lie within the PaLoc (Fig. 1). These could be prophage sequences with homology to tcdE, as phage sequences have been shown to be present in VPI 10463 (unpublished data). The tcdE gene was detected in φC2, φC5 and weakly in φC8. Four bands were common to φC2 and φC5, while the two bands in φC8 were unique (Fig. 1). Interestingly, tcdE was present only in the Myoviridae phages (φC2, φC5 and φC8) and may encode an essential protein such as holin. In this regard, φC6 of the Siphoviridae family may possess a different holin that is dissimilar to the tcdE product. φC2 and φC5 were closely related, and both were moderately related to φC8 but not to φC6, as determined by phage immunity assays and Southern hybridization (Goh et al., 2005). This level of relatedness was reflected in the patterns of hybridization of tcdE to phage DNA. The presence of tcdE in φC2, φC5 and φC8 is interesting because it supports the idea that phage may have some role in toxin secretion through a holin-like pathway, as suggested in two studies (Mukherjee et al., 2002; Tan et al., 2001). Holins are small membrane proteins that disrupt the bacterial membrane together with endolysins (Wang et al., 2000). To date, little is known about C. difficile toxin secretion, and although the type II pathway of the general secretory pathway is present in clostridia it is unlikely to be involved (Bruggemann et al., 2003; Mukherjee et al., 2002). Phage-infected toxigenic strains of C. difficile may be able to release toxins through a fully functional holin pathway resulting in cell lysis. Alternatively, the existing tcdE-associated holin pathway may be complemented in lysogens by phage holin genes and function at a higher efficiency to cause membrane disruption compared to uninfected toxigenic strains that rely only on the tcdE-associated holin pathway. It is also possible that the presence of a holin-like tcdE in the PaLoc is a coincidence. Putative holin sequences have been found in bacterial chromosomes, probably as remnants of cryptic phages (Wang et al., 2000).

(67K): Fig. 1. Detection of tcdE in phage DNA. (a) Separation of HindIII-digested C. difficile and phage DNA in 1 % agarose. Lanes: 1, λ/HindIII marker; 2, VPI 10463; 3, CD33; 4, φC2; 5, φC5; 6; φC6; 7, φC8. (b) Southern hybridization of (a) to DIG labelled tcdE probe carried out at 36 °C. There was strong hybridization of the tcdE probe to VPI 10463, φC2 and φC5, and weak signals from φC8.

Toxin A and B assays of lysogens and parents
Cytotoxicity assays were carried out on five toxigenic parental isolates and six lysogens (Table 1). Since an absorbance standard for quantification of toxin A was not provided with the ELISA kit (TechLab), the minimum end point titre was used as a measure of toxin A levels. A twofold increase in toxin A was measured in CD1017C6, while the remaining lysogens were unaltered, suggesting phage had little effect on toxin A production (Fig. 2a). Significant increases in toxin B were observed in all lysogens except for CD594C8 and CD727C6, which had unaltered levels (Fig. 2b). This apparent lack of coordination between toxin A and toxin B production may be due to methodology and should be confirmed by repeating the toxin A assays, and using an ELISA method for toxin B. Phage had a positive effect on toxin B production in CD839C2, CD1017C6, CD1017C8 and CD6938C6 (Fig. 2b) despite the apparent absence of tcdB in phage DNA. Hence increased toxin production was not through acquisition and transcription of an additional toxin gene. However, as discussed earlier, toxin B levels may have been increased through a phage-mediated process of toxin secretion. Alternatively, toxin expression may have been up-regulated in these lysogens.

(29K): Fig. 2. Toxin A and B of lysogens compared to parental strains. (a) Toxin A titres, expressed as absorbance units, were determined by ELISA. The end point dilution was the reciprocal of the minimum dilution that resulted in an A450 0.09. (b) Minimum cytotoxic titre (MCT) of toxin B was determined by neutral red assay. Data are expressed as mean ± SEM. *, P values < 0.01; na, the strain was not a host of the phage, or a stable lysogen was not isolated.

Toxin A and B expression in lysogens and parents (real-time RT-PCR)
A relative quantitative method of real-time RT-PCR was employed to compare toxin transcripts in parental isolates and lysogens. Coordinated transcription of tcdA and tcdB after phage infection was observed in all lysogens except CD594C8 and CD6938C6, in which significant changes in the amounts of tcdA transcripts were observed (Fig. 3). This is unusual as tcdA and tcdB transcription is co-regulated (Hammond et al., 1997; Hundsberger et al., 1997) through tcdD (Mani & Dupuy, 2001), which is responsive to growth phase, constituents of medium, and temperature (Karlsson et al., 2003; Mani et al., 2002). Interestingly, phage infection increased tcdA transcription in CD594C8 but not TcdA production, while phage infection did not affect tcdB transcription but increased TcdB production in four lysogens (Table 3). This suggests that phage infection has a greater effect on toxin production or release than transcription of tcdA and tcdB, perhaps through the action of holins. In lysogens, functional holins may result in channels that are only large enough for TcdB (266 kDa) to pass through but not TcdA (308 kDa), hence extracellular TcdA levels are unaltered. However, a recent study on the size of membrane lesions formed by the λ holin showed that a protein of 480 kDa passed through the lesions. The authors suggested that holins may not form channels of a constant diameter, but rather cause general disruption of the cell membrane (Wang et al., 2003).

(35K): Fig. 3. Level of tcdA and tcdB transcripts in lysogens compared to parental strains by real-time RT-PCR. Coordinated transcription of tcdA and tcdB was observed in all lysogens except CD594C8 and CD6938C6. (a) Toxin A transcription was significantly increased in CD594C8 but significantly decreased in CD6938C8 compared to the corresponding parental strain. (b) The overall trend of toxin B transcription in lysogens is similar to that in (a). *, P values < 0.01. Data are expressed as mean ± SEM. Input RNA was determined using a calibrator sample from which a standard curve was constructed.

Table 3. Summary of toxin levels and toxin gene transcription in lysogens of φC2, φC5 and φC8 +, Toxin level in lysogen is significantly higher than parental strain; 0, no change or insignificant change in toxin level of lysogen compared to parental strain; , toxin level in lysogen is significantly lower than parental strain.

The PaLoc of C. difficile has characteristics of a mobile element; non-toxigenic strains lack this virulence element (Braun et al., 1996). Perhaps the homology of tcdE, tcdA and tcdB to phage sequences discussed so far indicates that the PaLoc was carried by phages that integrated at a specific site and conferred toxigenicity to ancestral C. difficile strains. However, functional prophage genes have disintegrated severely and the present PaLoc is a remnant of cryptic phage genes. The inability of contemporary phages to convert contemporary non-toxigenic C. difficile strains to toxin production could be due to phage genomes having evolved to play some other role in the biology of C. difficile. In summary, toxin production was not always directly correlated with toxin transcription in lysogens. Differences in toxin levels appeared to be dependent on the parental strains rather than the infecting phage, since these changes were not associated with a specific phage, and occurred in some, but not all, lysogens of φC2, φC6 and φC8. Variable toxin levels may be related to the toxinotype of parental strains, given the putative relationship of PaLoc and phage genes. Certain toxinotypes might also be more readily lysogenized and/or exhibit greater stability of lysogenic phages. Identification and sequencing of the holin genes in the phages, as well as site-directed mutagenesis of tcdE, will be necessary for determining the role of phages in toxin production or secretion in C. difficile.