Morphological and Genetic Diversity of Temperate Phages in Clostridium difficile

Abstract

Eight temperate phages were characterized after mitomycin C induction of six Clostridium difficile isolates corresponding to six distinct PCR ribotypes. The hypervirulent C. difficile strain responsible for a multi-institutional outbreak (NAP1/027 or QCD-32g58) was among these prophage-containing strains. Observation of the crude lysates by transmission electron microscopy (TEM) revealed the presence of three phages with isometric capsids and long contractile tails (Myoviridae family), as well as five phages with long noncontractile tails (Siphoviridae family). TEM analyses also revealed the presence of a significant number of phage tail-like particles in all the lysates. Southern hybridization experiments with restricted prophage DNA showed that C. difficile phages belonging to the family Myoviridae are highly similar and most likely related to previously described prophages φC2, φC5, and φCD119. On the other hand, members of the Siphoviridae phage family are more genetically divergent, suggesting that they originated from distantly related ancestors. Our data thus suggest that there are at least three genetically distinct groups of temperate phages in C. difficile; one group is composed of highly related myophages, and the other two groups are composed of more genetically heterogeneous siphophages. Finally, no gene homologous to genes encoding C. difficile toxins or toxin regulators could be identified in the genomes of these phages using DNA hybridization. Interestingly, each unique phage restriction profile correlated with a specific C. difficile PCR ribotype.

Clostridium difficile is now one of the leading bacterial pathogens responsible for nosocomial infections, and hypervirulent strains, such as NAP1/027 (North American pulsed-field type 1 and PCR ribotype 027), are emerging and are associated with increased morbidity and disease severity (33, 42, 43, 60). The development of C. difficile-associated diarrhea in individuals normally not at risk suggests that the epidemiology and virulence of C. difficile are changing (13, 42, 43). The main virulence factors are two exotoxins, TcdA and TcdB (for a review, see reference 59). In addition, about 6% of the C. difficile strains analyzed possess a binary toxin encoded by two genes, cdtA and cdtB (44, 52). The expression of this binary toxin may be associated with increased disease severity and morbidity (25) and was found to correlate with the presence of an 18-bp deletion in the tcdC gene (33). It is noteworthy that TcdC was recently shown to act as a negative regulator of toxin production (36). Various hydrolytic enzymes (48), as well as the capsule and flagella, also likely play a role in virulence (6, 14, 56). TcdA and TcdB are encoded in the pathogenicity locus (PaLoc) located on an ∼19.6-kb chromosome fragment of all toxigenic strains (59). Twenty-four variations in the PaLoc have been identified thus far, and they are used as markers for toxinotyping of C. difficile strains (28, 45).

For many years, bacteriophages have been associated with major pathogens, such as Vibrio cholerae, Escherichia coli O157:H7, or Clostridium botulinum. Often, these pathogens became virulent after acquisition of one or more prophages carrying powerful toxins or other virulence factors, like superantigens and various enzymes (for reviews, see references 7 and 8). With the sequencing of a number of microbial genomes, the role of prophages in bacterial evolution and virulence has become evident (2-4, 17, 40). It is noteworthy that numerous prophages, like those found in Streptococcus pyogenes, share extensive DNA similarity (10), which makes phage evolution possible through the exchange of functional modules (12, 26). In addition, genetically related prophages may serve as anchor points leading to bacterial genome reorganization (8, 9). Finally, they may also evolve through interactions with the virulent phage population (29).

Recent epidemics caused by hypervirulent strains, such as C. difficile NAP1/027, and the renewed interest in phage therapy (51, 54, 57) have prompted an increasing number of research groups to study phages of C. difficile. However, data on this group of phages are still scarce (15, 34, 35, 39, 49), and the possible impacts of these phages on the biology of their bacterial hosts are unknown. The first report of molecular characterization of temperate phages in C. difficile was published by Goh et al. in 2005 (22). In this study, 3 of 56 clinical isolates of C. difficile yielded four double-stranded DNA phages (φC2, φC5, φC6, and φC8) upon mitomycin C induction (22). Phages φC2, φC5, and φC8 were morphologically similar (family Myoviridae), and φC2 and φC5 were closely related, as revealed by DNA-DNA hybridization experiments (22). In contrast, phage φC6 belonged to the Siphoviridae family and shared only a little DNA similarity with the three other phages. In 2006, the first complete genomic sequence and annotation of a C. difficile temperate phage (φCD119, a member of the Myoviridae; 53,325 bp) were made available (23). The genome of a second temperate phage (φC2, a member of the Myoviridae; 56,538 bp) was also published recently (21). In addition, the multidrug-resistant C. difficile strain CD630 was found to harbor in its genome two highly related prophages whose sequences were identical in more than 60% of their entire genomes (47). These prophages were recently shown to be inducible by mitomycin C and are members of the Myoviridae family (21).

Previous studies have revealed some homology between genes in the C. difficile PaLoc and phage genes (20, 21, 23, 55). For example, the tcdA gene encoding toxin A was found to be related to a gene with an unknown function in phage φCT2 of Clostridium tetani (10). Furthermore, there is 55% identity between ORF22 of Lactobacillus casei phage A2 and the last 103 amino acids of TcdC found in C. difficile VPI 10463 (20). A low level of sequence identity was also noted between ORF41 of C. difficile phage φC2 and TcdB (21). Interestingly, ORF46 of φC2 and ORF41 of φCD119 shared similarity with a putative penicillinase repressor (Cdu1) encoded by a gene bordering the C. difficile PaLoc. Finally, the gene product of tcdE was shown to be structurally and functionally related to a phage holin, as revealed by bioinformatics analyses and in vivo expression in Escherichia coli (55). Accordingly, ORF34 of φCD119 was shown to be similar to TcdE (23), and DNA hybridization experiments revealed the presence of tcdE homologs in the genomes of C. difficile phages φC2, φC5, and φC8 (20, 21). Altogether, these results suggest that some of the genes in the PaLoc are genetically and, in the case of tcdE, functionally related to phage genes, thus supporting the assumption that the origin of PaLoc (or fragments of it) might have been an ancient prophage.

Complete genome sequencing and analysis of phages φCD119 and φC2, as well as the two prophages in strain CD630, failed to reveal any direct evidence of virulence factors (21, 23, 47). Moreover, none of the phages in C. difficile described so far could convert a nontoxigenic strain into a toxigenic strain. However, Goh et al. suggested that lysogenization of toxigenic strains of C. difficile with certain phages could lead to increased toxin A and/or toxin B production (20). Although tcdA transcription was shown to be significantly increased in a strain lysogenized with phage φC8, these authors could not find a correlation between the level of transcription and the level of toxin production in their lysogens, nor could they find a correlation with a specific phage. Nevertheless, this study suggests that some phages may play a role in the regulation of the expression of toxin genes or other genes in C. difficile. For example, the five putative transcriptional regulators identified in the genome of φCD119 may be involved in host gene regulation (23).

In this study, we performed a survey of the diversity of inducible prophages residing in the genomes of C. difficile clinical isolates. Morphological and genetic analyses were performed, and a preliminary classification of the phages was established. Screening for genes encoding toxins and toxin regulators in the PaLoc failed to identify homologous genes in the genomes of these phages.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

C. difficile isolates were kindly provided by Jacques Pépin of the Université de Sherbrooke (Québec, Canada). Strain CD630 was a kind gift from Julian Parkhill of the Wellcome Trust Sanger Institute (Cambridge, United Kingdom). C. difficile strains CD44 and CD52 were isolated from human fecal samples provided by Mirabelle Kelly of the Jean-Talon Hospital (Montréal, Québec, Canada). Bacterial strains were routinely cultured at 37°C under anaerobic conditions in brain heart infusion (BHI) broth (Oxoid, Nepean, Ontario, Canada) which had been prereduced under an anaerobic atmosphere for 24 h at 37°C prior to use.

Bacterial DNA extraction.

Genomic DNA was extracted using the following procedure. All centrifugations were performed at room temperature. Ten milliliters of an overnight culture was centrifuged in 15-ml conical tubes for 15 min in a clinical centrifuge. The cell pellets were suspended in 1 ml saline (0.85% NaCl), transferred to an Eppendorf tube, and centrifuged for 2 min at full speed in a microcentrifuge. Cells were then suspended in 1 ml of a 20% sucrose solution containing 120 mg/ml lysozyme and incubated at 37°C for 30 min. After centrifugation for 2 min, the pellet was suspended in 600 μl of 2% sodium dodecyl sulfate. The preparation was incubated at room temperature with frequent agitation until complete lysis, after which 300 μl of ice-cold potassium acetate (3 M, pH 7.0) was added. After mixing, the lysate was incubated on ice for 5 min and then centrifuged at full speed for 5 min. The cleared lysate was transferred to a new tube, and 650 μl of isopropanol was added. After mixing by inversion, the DNA was precipitated by centrifugation for 10 min. The resulting pellet was rapidly dissolved in 320 μl of distilled water, after which 200 μl of 7.5 M potassium acetate and 350 μl of a phenol-chloroform solution (1:1, vol/vol) were added. After vigorous mixing (no vortex), the phases were separated by centrifugation for 5 min at full speed. The upper aqueous phase was transferred to a new tube, and a second phenol-chloroform extraction was performed. The DNA was precipitated by addition of 1 ml of anhydrous ethanol and centrifugation for 15 min at full speed. The DNA pellet was dried for a few minutes and dissolved in 40 μl of 10 mM Tris-HCl-1 mM EDTA (pH 8.0).

PCR ribotyping.

The PCR ribotyping method was used as an initial typing tool for classification of our C. difficile isolates (5). The 16S rRNA primer 5′-GTGCGGCTGGATCACCTCCT-3′ and the 23S rRNA primer 5′-CCCTGCACCCTTAATAACTTGACC-3′ described by Bidet et al. (5) were used. The PCR was performed with a PTC-200 thermal cycler (MJ Research, Waltham, MA) using a 50-μl (final volume) mixture containing 1× reaction buffer (Invitrogen, Burlington, Ontario, Canada), 125 μM of each deoxynucleoside triphosphate, 2 mM MgCl₂, 100 pmol of each primer, and 2.5 U of Taq DNA polymerase (Invitrogen). Two microliters of a 1/10 dilution of C. difficile DNA extract (see above) was added as a template. A similar amount of sterile water was used for the negative control. The cycling conditions were 3 min at 94°C, followed by 35 cycles of denaturation at 94°C for 45 s, annealing at 57°C for 45 s, and elongation at 72°C for 1 min. A final elongation step of 5 min at 72°C was included. The final reaction mixture was combined with 8 μl of 6× gel loading buffer (46), and 6 μl was separated using a 1.5-mm-thick 5% nondenaturing polyacrylamide gel in 1× Tris-acetate-EDTA buffer using a Mini Protean II apparatus (Bio-Rad Laboratories, Mississauga, Ontario, Canada). The gel was run for ∼45 min at 150 V and then stained with ethidium bromide before it was exposed to UV light. Recorded images were analyzed using the Quantity One software (Bio-Rad Laboratories).

PCR amplification of C. difficile CD630 prophage sequences.

PCR primers were designed to amplify specific regions of each of the two complete prophages predicted from the genomic sequence of C. difficile CD630 (accession number AM180355.1) (47). Primers CD630.1F (5′-CTGTATGACCTAACCTTTCTGAAACAAC-3′; nucleotide positions 1088267 to 1088294) and CD630.1R (5′-TTGCGACTAGGAGAAGCTACTGGA-3′; nucleotide positions 1088690 to 1088667) were specific for one of the prophages, while primers CD630.2F (5′-TTGTTTGCAGTCTTCGTTCCACCC-3′; nucleotide positions 3426729 to 3426706) and CD630.2R (5′-CCATCACCTCGTTGCTTATTTGGG-3′; nucleotide positions 3426338 to 3426361) were specific for the other prophage.

Prophage induction.

Bacterial cells were inoculated into 5 ml of BHI medium and grown at 37°C under anaerobic conditions until stationary phase (16 to 24 h). A 3% inoculum from an overnight culture was transferred into 10 ml of fresh BHI medium, and growth was monitored until the optical density at 600 nm (OD₆₀₀) reached 0.1. Then, depending on the strain, mitomycin C was added to a final concentration ranging from 0.5 to 5 μg/ml. Mitomycin-containing cultures were further incubated at 37°C under anaerobic conditions, and the OD₆₀₀ was monitored regularly. A significant decrease in the cell density after 3 to 5 h suggested that prophages were released. Crude lysates were then filtered (0.45 μm) and stored at 4°C until further analysis.

TEM.

Prior to observation, 1.5 ml of crude lysate was centrifuged for 1 h at 4°C and 24,000 × g. A fraction of the supernatant (approximately 1.4 ml) was gently discarded, and 1 ml of ammonium acetate (0.1 M, pH 7.5) was added to the remaining lysate, which was then centrifuged as described above. This step was performed twice. The washed phage particles were used to prepare transmission electron microscopy (TEM) grids as follows. Ten microliters of 2% uranyl acetate was deposited onto a 200-mesh Formvar/carbon-coated copper grid (Pelco International, Redding, CA). After 30 s, 10 μl of the phage preparation was mixed with the stain by pipetting the solution up and down a few times. After further incubation for 30 s, the liquid was removed from the grid by blotting with a piece of Whatman paper. To achieve good contrast, care was taken to leave a thin film of the uranyl acetate stain on the grid and to avoid complete blotting. The grids were dried for a few minutes and then observed at 80 kV with a JEM-1230 transmission electron microscope (JEOL, Tokyo, Japan) equipped with a DualVision digital camera (Gatan, Pleasanton, CA). Micrographs were recorded using the Gatan Digital Micrograph software (version 1.3), after the exposure time was set to 10 s.

Phage DNA analysis.

Phage DNA was isolated from crude lysates obtained after mitomycin C induction using a rapid miniprep protocol described elsewhere (38). Phage DNA was digested with restriction enzymes (Roche Diagnostics, Laval, Québec, Canada) used according to the manufacturer's recommendations. The digestion mixture was heated at 70°C for 10 min to disrupt any possible cohesive end joining and immediately run through a 0.8% agarose gel. Gels were stained with ethidium bromide, exposed to UV, and photographed using a gel documentation system (Bio-Rad Laboratories).

Southern hybridization.

Phage DNA was digested with restriction enzymes, and the resulting fragments were separated on a 0.8% agarose gel before they were transferred onto positively charged nylon membranes (GE HealthCare, Baie D'Urfé, Québec, Canada) using standard protocols (46). Phage DNA (∼1 μg) was labeled using a digoxigenin (DIG) High-Prime kit from Roche Diagnostics according to the manufacturer's recommendations. The labeling reaction was performed for 16 h at 37°C. Probes corresponding to genes from the PaLoc (tcdA, tcdB, tcdC, tcdE, and tcdR) and ctdA encoding the binary toxin were also amplified by PCR with DIG-11-dUTP incorporation. The primer sequences used for generating the probes are shown in Table 1. Prior to hybridization, the labeled DNA probes were quantified, denatured by boiling for 10 min, diluted in DIG Easy-Hyb buffer (Roche Diagnostics) to obtain a final concentration of ∼25 ng/ml, and filtered (0.45 μm) to reduce the background. Hybridization was carried out at 42°C, and the membranes were washed and detected as recommended by Roche Diagnostics. The hybridized membranes were then exposed to Kodak Biomax Light film.

TABLE 1.

Primers used to generate probes for toxin genes and toxin regulators

Primer	Sequence (5′-3′)a	Target gene	Probe size (bp)
LCF-139	TTAATTAATTTTCTCTACAGCTATCC	tcdC	722
LCF-140	TCTAATAAAAGGGAGATTGTATTATG
LCF-149	TGAACCTGGAAAAGGTGATG	cdtA	376
LCF-150	AGGATTATTTACTGGACCATTTG
LCF-160	TGGTGATATGGAGGCATCACCACT	tcdB	501
LCF-161	TGAGCTTTAGCTCTTGCATCGTCA
LCF-162	ACATTATGAAGAGGGAGAAACAGA	tcdR	385
LCF-163	AGCCTTATTAACAGCTTGTCT
LCF-164	CCTAGGAGGCGTTATGAATATGACA	tcdE	366
LCF-165	GGTACTGGTAATCCACATAAGCAC
LCF-166	AGCTCTATTGGACTAGACCGTTGGGA	tcdA	514
LCF-167	ACGCATAAGCTCCTGGACCACTTA

^aMost sequences were based on the genomic sequence of C. difficile strain CD630 (accession number NC_009089); the exceptions were primers LCF-149 and LCF-150, which were based on the genomic sequence of C. difficile strain QCD-32g58 (accession number AAML00000000).

RESULTS

PCR ribotyping.

First, 45 isolates of C. difficile were obtained from various sources; these strains included toxigenic strain CD630 obtained from the Sanger Institute. All the strains except CD630 were isolated from human stools during the C. difficile outbreak in 2003 and 2004 in the province of Québec (Canada). In this outbreak, one strain, strain NAP1/027, was found to be predominant, (25, 33). This strain is also referred to as QCD-32g58, and its whole genome is currently being sequenced (accession number AAML00000000).

The 45 isolates were then classified by PCR ribotyping in order to subsequently analyze the prophage contents of different strains. For the 45 isolates analyzed, 11 distinct ribotypes were obtained, 6 of which were later shown to contain inducible prophages (see below). The latter six ribotypes are shown in Fig. 1A. Note that isolates CD5, CD6, and CD44 are most likely clones of the same strain as they have identical PCR ribotype profiles (Fig. 1A). In fact, 31 isolates (69%) had the same ribotype as CD5, CD6, and CD44 (data not shown). These 31 isolates also belonged to toxinotype III, were binary toxin positive, and harbored an 18-bp deletion in the tcdC gene (J. Pépin and É. Frost, personal communications). Based on the findings described above, we concluded that these 31 isolates represented hypervirulent strain NAP1/027 (25, 33).

FIG. 1.

(A) PCR ribotyping of C. difficile isolates. The designations above the lanes are the designations of the C. difficile isolates analyzed by PCR ribotyping. Isolates CD5, CD6, and CD44 have identical ribotypes and correspond to hypervirulent strain NAP1/027. Lane 100-bp contained a DNA marker. (B) Comparison of HindIII DNA restriction profiles of phage DNA. DNA was purified from the lysates of mitomycin C-induced cultures of various C. difficile clinical isolates. The phage designations above the lanes refer to the C. difficile isolate designations in panel A. Lanes 1-kb contained a DNA marker.

Prophage induction.

Addition of mitomycin C to growing cultures, including cultures of C. difficile strains, is well known to induce excision of prophages (22, 23, 49). Mitomycin C was added to early-log-phase cultures of representative isolates of each of the 11 PCR ribotypes. The optical density was monitored, and a sudden decrease in the OD₆₀₀ after 3 to 5 h of growth following mitomycin C addition suggested that there was cell lysis due to the induction of prophages and possibly the release of viral particles. The latter suggestion was confirmed by TEM observations and extraction of whole phage DNA from the crude lysates. With this approach, C. difficile isolates CD5, CD8, CD24, CD38, CD52, and CD630 representing the six different ribotypes (Fig. 1A) were found to release complete viral particles from which DNA could be purified, whereas the isolates representing the other five ribotypes did not contain inducible prophages under the conditions tested. Interestingly, isolate CD5, which is representative of the predominant and hypervirulent strain, was found to contain an inducible prophage. Previous computational analyses by our group and other groups failed to identify a complete prophage in the genome of this strain, most likely because the sequencing project is incomplete (21). Thus, to our knowledge, this is the first report of the presence of an inducible prophage in the genome of the NAP1/027 strain.

Finally, we also noticed that some C. difficile isolates showed a decrease in the OD₆₀₀, but we could not find intact phage particles by TEM observation or isolate phage DNA from the crude lysates. Instead, numerous phage tail-like particles (PT-LPs) that were 130 by 18 nm were found (Table 2), and these particles were morphologically similar to those previously identified by Nagy and Foldes in mitomycin C-induced lysates of C. difficile (39). Similar PT-LPs were also observed in Vibrio cultures (19). It is noteworthy that the amount of these PT-LPs varied greatly from one C. difficile isolate to another.

TABLE 2.

Features of C. difficile temperate phages

^aMeans of at least 5 to 10 measurements obtained with different viral particles.

^bThe magnification is ×200,000 and the bars are 100 nm long except for PT-LPs. For PT-LPs the magnification is ×300,000 and the bars are 50 nm long.

TEM observations.

From the six C. difficile isolates (with distinct PCR ribotypes) that contained inducible prophages, eight phages could be differentiated based on TEM observations (Table 2). Two distinct phage morphotypes were observed in the lysates of C. difficile strains CD8 and CD38, whereas only one morphotype was found in the other four lysates (CD5, CD24, CD52, and CD630).

Three phages (φCD5, φCD52, and φCD630) had an isometric capsid and a contractile tail (morphotype A1). These three phages were very similar in morphology and size (Table 2). Phage particles with contracted sheaths were also observed in several lysates containing these phages (data not shown). Based on criteria of the International Committee on Taxonomy of Viruses, they belong to the Myoviridae family (1). Our results are in agreement with a recent article showing that the inducible C. difficile prophages φCD630-1 and φCD630-2 belong to the Myoviridae family (21). However, contrary to the findings of Goh et al. (21), we observed only one morphotype in our CD630 lysates, even though two distinct phages were present as revealed by PCR and partial sequencing (data not shown). In addition, it was previously reported that phage φCD630-1 had a larger capsid (31.7 ± 0.7 nm versus 28.1 ± 1.3 nm) and a longer tail (62.4 ± 5.1 nm versus 39.5 ± 5.8 nm) than phage φCD630-2 (21). It is noteworthy that our φCD630 phage dimensions (capsid diameter, 62 nm; tail length, 126 nm) are significantly larger than those previously reported for these inducible prophages (21). On the other hand, our measurements are consistent with the usual size of capsids that are capable of packaging 49-kb (φCD630-2) and 55-kb (φCD630-1) genomes (16, 21).

The remaining five phages (φCD8-1, φCD8-2, φCD24, φCD38-1, and φCD38-2) had the B1 morphotype, which corresponds to an isometric capsid and a long noncontractile tail, and thus belong to the Siphoviridae family (1). Interestingly, C. difficile strains CD8 and CD38 both released two types of Siphoviridae phages (Table 2). Phage φCD8-1 had a larger capsid than φCD8-2 (diameters, 90 and 46 nm, respectively) and a longer tail (432 and 338 nm, respectively). Similarly, φCD38-1 had a larger capsid than φCD38-2 (diameters, 77 and 58 nm, respectively), as well as a longer and larger tail (402 by 15 and 262 by 12 nm, respectively). φCD38-1 is morphologically similar to the virulent Lactococcus lactis phage 949 (16), although its capsid is larger than that of phage 949 (90 nm versus 70 nm). The size of φCD24 was midway between the sizes of the four other phages belonging to the Siphoviridae family; the capsid diameter was 66 nm, and the tail size was 302 by 10 nm. Detailed observation of φCD24 morphology revealed the presence of six fibers that were about 60 nm long at the tip of its tail (Table 2). Each of these fibers also had a globular structure at its extremity. This is the first time that such a tail appendage has been described for a Clostridium phage.

Phage genome comparison.

Phage DNA was extracted from the crude lysates of C. difficile isolates representing the six PCR ribotypes (Fig. 1). The HindIII endonuclease was used to generate DNA restriction profiles, which allowed easy discrimination of the induced prophages. As shown in Fig. 1B, each unique C. difficile ribotype had a distinct phage DNA restriction profile. Interestingly, isolates with identical PCR ribotypes, such as CD5, CD6, and CD44 (and others not shown here), had identical phage DNA restriction profiles. The profiles obtained for lysates of CD8 and CD38 were rather complex, with numerous bands and high-molecular-mass fragments. These findings are consistent with phages with a large capsid and/or with the presence of distinct phages in the lysates.

We performed a series of DNA-DNA hybridization experiments in order to determine the relatedness of the phage genomes shown in Fig. 1B. This methodology has been found to be very effective for classifying phages of L. lactis (16) and Streptococcus thermophilus (32). Genomic DNA from each phage preparation was DIG labeled and used as a probe to detect homologous DNA sequences. Hybridization results obtained with probes for phages φCD5, φCD24, φCD38-1/φCD38-2, and φCD52 are shown in Fig. 2. Using the φCD5 and φCD52 probes, we noted that the three phages belonging to the Myoviridae family were related and that phage φCD52 was more similar to φCD630 (Fig. 2A).

FIG. 2.

Southern hybridization analysis of phage DNA. (A) An ethidium bromide-stained gel containing HindIII-digested genomic phage DNA is at the top. Autoradiographs of Southern blots are below this gel, and the corresponding phage probes are indicated on the left. The phage designations above the lanes refer to the C. difficile isolates used for mitomycin C induction. (B) Same procedure as in panel A, but a different agarose gel which included φCD8 DNA was used. Lanes M contained a 1-kbp DNA marker.

Greater genomic diversity was observed among the phages belonging to the Siphoviridae family. First, it was rather obvious that φCD24 was quite different from the other phages, since the φCD24 probe strongly hybridized mainly with the φCD24 genome. Limited similarity was observed between φCD24 and φCD5 (Fig. 2A), and according to the results obtained with the DIG labeling and detection methods used, DNA regions as short as 200 bp with ≥60% DNA identity could be detected across the membrane. Thus, complete genome sequencing of φCD24 and φCD5 is required to precisely determine the extent of DNA identity between these two phages.

Finally, the φCD38-1/φCD38-2 genomic probe hybridized with DNA fragments from the corresponding phage lysate and also from φCD8-1 and φCD8-2 (Fig. 2B). Since the CD8 lysate contained two phages (φCD8-1 and φCD8-2) and the φCD38-1/φCD38-2 probe contained two genomes (φCD38-1 and φCD38-2), it is not possible at this time to know which phage probe hybridized to which phage from the lysate. Nonetheless, it is clear that these siphophages are different from the other phages (e.g., φCD24, φCD5, φCD52, and φCD630).

Correlation between the type of prophage and PCR ribotype.

Our TEM and restriction profile analyses suggested that there was a unique prophage for each individual PCR ribotype. This was exemplified by phages φCD5, φCD6, and φCD44; all three of these phages were identical in morphology (not shown) and HindIII restriction profiles (Fig. 1B and 2), and they were induced from C. difficile isolates with identical PCR ribotypes (Fig. 1A). We investigated a total of seven different isolates corresponding to epidemic strain NAP1/027 and found that they all released identical prophages following the addition of mitomycin C (data not shown). A similar observation was made with two isolates of the CD8 ribotype and two isolates of the CD38 ribotype (data not shown). Thus, at least in our strain collection, the PCR ribotype correlates with the inducible prophage content.

Screening for potential toxin genes within C. difficile genomes.

To test whether our phages could potentially encode toxins or toxin regulatory proteins related to those encoded in the PaLoc, we performed a series of DNA-DNA hybridizations using probes for tcdA, tcdB, tcdC, tcdE, and tcdR. The probes for tcdA and tcdB were amplified from the genome of strain CD630 and covered a portion of the nonrepeating region of both toxin genes corresponding to nucleotide positions 910 to 1423 for tcdA and nucleotide positions 1089 to 1589 for tcdB. We also included a probe for cdtA, one of the two genes encoding the binary toxin. All probes were PCR amplified and labeled with DIG-11-dUTP; the PCR primers used to generate the probes are listed in Table 1. Genomic DNA from strain CD630 was used as the template for all PCR probes except the probe for cdtA. Indeed, cdtA and cdtB were found to be highly altered pseudogenes in the genome of strain CD630 (47). Thus, for amplification of cdtA, we used DNA from strain CD6, corresponding to the hypervirulent strain, which was previously shown to encode the binary toxin (33). Primers were designed based on the available genomic sequence of strain QCD-32g58 (accession number AAML00000000). Although all our probes allowed easy detection of the corresponding genes in the genomic DNA of our C. difficile isolates, no signal could be detected with our phage genomes, despite extended autoradiography exposures (data not shown). It is worth noting that in these experiments the number of copies of each phage genome was significantly greater than the number of copies of whole bacterial DNA. Hence, if a positive signal was detected with bacterial DNA but not with phage DNA, we concluded that the phages do not contain genes homologous to genes in the PaLoc or to the cdtA gene encoding the binary toxin.

Indicator strains.

We tried to identify C. difficile strains that could be used as indicator strains to propagate the induced prophages. Ten microliters of each lysate was deposited on soft agar overlays containing 500 μl of mid-log-phase cultures (OD₆₀₀, 0.5 to 0.8) of each C. difficile isolate from our collection. Plates were incubated anaerobically for 48 h. Unfortunately, none of the 45 C. difficile isolates available allowed lytic amplification of the phages under the conditions tested. The difficulty of finding proper indicator strains has been reported previously by several authors, although a limited number of indicator strains have been found for other C. difficile phages, including φCD630-1 and φCD630-2 (21, 23, 34). We are currently screening a larger collection of C. difficile isolates using different experimental conditions in order to identify such strains.

DISCUSSION

So far, only two prophages of C. difficile, φCD119 and φC2, have been studied at the molecular level (21-23). Hence, our knowledge of the genetic diversity of these phages and their possible role in the biology of C. difficile is still very limited. Here, we describe morphological characterization and the DNA relatedness of eight other temperate phages of C. difficile.

Ribotyping of C. difficile is based on PCR amplification of the 16S-23S spacer region within the rRNA gene cluster. The number of alleles and the length of the spacer region vary from one strain to another, which results in a specific banding pattern upon gel electrophoresis of the PCR products (5, 53). Considering the conserved nature of ribosomal gene sequences and the mobility of phage elements (24), we initially thought that the ribotype may not be indicative of the prophage content. Moreover, bacteria tend to gradually delete from their genomes DNA sequences that represent a threat to their survival (30). However, our observations suggest that there is a correlation between the PCR ribotype and specific prophages. For each PCR ribotype, we found one prophage with a unique restriction profile. In the case of hypervirulent strain NAP1/027, seven different isolates, all collected in the province of Québec during the outbreak in 2003 and 2004, were shown to contain identical prophages. We cannot exclude the possibility that identical restriction profiles could in fact be slightly different due to small insertions or deletions or to point mutations. Nonetheless, to our knowledge, the association between ribotyping and prophage content is reported for the first time here and could be valuable for selecting strains to study their prophage contents. Since over 150 different C. difficile ribotypes have been described so far (27), there could potentially be many different prophages.

Electron microscopy observations revealed the presence of two phage morphotypes in our lysates, morphotypes A1 and B1 (1). Three of the phages (φCD5, φCD52, and φCD630) had the A1 morphotype, which is a characteristic of the members of the Myoviridae family. The morphology and size of these myophages fell in the morphology and size ranges of previously described C. difficile myophages, like phage 56 (34) and φCD119 (23), as well as φC2, φC5, and φC8 (21, 22). Nagy and Foldes also reported TEM observations for six similar myophages in both toxin-positive and toxin-negative C. difficile clinical isolates (39). The Southern hybridization data reported here revealed that phages φCD5, φCD52, and φCD630 are closely related as they share significant DNA similarity. Likewise, Goh et al. (22) showed that the highly related phages φC2 and φC5 (Myoviridae) were also related to the prophage φC8, but to a lesser extent. Comparative bioinformatics analyses of the genomes of C. difficile prophages φC2 and φCD119, as well as the genomes of φCD630-1 and φCD630-2 (all Myoviridae phages), also confirmed this relatedness (21-23). In fact, it has been suggested that the prophages found in the genome of strain CD630 are a mosaic of φC2 and φCD119 (23). Based on our DNA hybridization experiments, we deduced that phages φCD5 and φCD52 described in this study are similar to φCD630 and hence are genetically related to phages φC2, φC5, and probably φC8 described by Goh et al. (21, 22). Together, the Myoviridae prophages isolated so far from C. difficile make up a genetically closely related group.

Five prophages (φCD8-1, φCD8-2, φCD24, φCD38-1, and φCD38-2) had the B1 morphotype and belonged to the Siphoviridae family. Phage φCD24 is very similar in size and morphology to C. difficile phages 41 (34) and φC6 (22) and to one phage identified by Nagy and Foldes in C. difficile strain 25 (39). Nagy and Foldes also isolated two siphophages with tails that were 300 and 370 nm long, which is in the size range of the tails of phages φCD8-1, φCD8-2, and φCD38-1 (338 to 432 nm), but the capsid sizes were smaller than the capsid sizes of the phages described here (39). Previous Southern hybridization experiments showed that the siphophage φC6 was only distantly related to Myoviridae phages (22). Likewise, we showed that phage φCD24 is distantly related to the myophages characterized in this study. The other Siphoviridae phages (φCD8-1, φCD8-2, φCD38-1, and φCD38-2) reported here are also different from the other phages as no or only very limited DNA similarity was detected in Southern hybridization experiments. For example, no hybridization signal could be detected between phage φCD24 and φCD38-1 or φCD38-2. Thus, these phages constitute two genetically distinct groups of Siphoviridae phages. Whole-genome sequencing of these genetically unrelated phages should provide valuable data on the genetic diversity of C. difficile temperate phages.

One of the most intriguing findings of this study was the presence of PT-LPs in all our C. difficile lysates. Induction of some C. difficile isolates led to the production of large amounts of these PT-LPs, which resemble contractile tails from Myoviridae phages. Nagy and Foldes also reported detection of highly similar PT-LPs in 7 of 18 strains induced with mitomycin C (39). Several other examples of bacteria producing PT-LPs have been reported, including the R-type pyocin of Pseudomonas aeruginosa (31) and the PT-LPs of Budvicia aquatica (50), Pragia fontium (50), and Photorhabdus luminescens (18), as well as the PT-LPs of mitomycin C-induced cultures of Vibrio sp. (19). The latter PT-LPs are of particular interest as they possess flower-like appendages similar to those observed in this study (19). Some of these PT-LPs are known to have bactericidal activity (58), although such activity remains to be demonstrated for the PT-LPs of C. difficile. According to a current hypothesis, genes encoding PT-LPs were derived from partially deleted prophages and they provide competitive advantages to their hosts (11). For example, the pyocin of P. aeruginosa was derived from an ancestor of phage P2 (Myoviridae) (41). Interestingly, in some cases, the different amounts of released PT-LPs depended on the host's physiological status and growth environment (19). Whether modifying the growth conditions of C. difficile strains influences or does not influence the production of these PT-LPs remains to be determined. Nonetheless, the presence of the PT-LPs in numerous C. difficile isolates suggests that they may have a functional role in the biology of this pathogen.

Previous studies have revealed some degree of similarity between phage genes and tcdA, tcdB, tcdC, and particularly tcdE in the PaLoc of C. difficile (20, 21, 23, 55). This finding is supported by the results of Southern blot and PCR experiments with whole phage DNA, functional assays in E. coli, and bioinformatics analyses (20, 55). Our experiments using whole phage DNA and probes against tcdA, tcdB, tcdC, tcdE, tcdR, and the binary toxin gene cdtA failed to detect any significant similarity between these bacterial genes and the induced prophages. Although we didn't find any DNA similarity, this result does not exclude the possibility that these phages could impact the expression of the toxins, as observed by Goh et al. (20), or could contribute somehow to the virulence of C. difficile. In fact, the identification of virulence factors in phage genomes is not always straightforward. A good example of unsuspected virulence factors was reported recently for Streptococcus mitis (37). The authors found that the prophage tail proteins PblA and PblB were released upon induction from a subset of the bacterial population. Unexpectedly, subsequent attachment of PblA and PblB to the surface of intact surrounding S. mitis cells was shown to increase the virulence of this pathogen in an animal model of endocarditis (37).

In summary, this study provides new data regarding the prophage diversity within C. difficile strains. Our results suggest that particular prophages are associated with specific PCR ribotypes. Over 150 different C. difficile ribotypes have been reported; it will be interesting to assess if this is representative of the prophage diversity within C. difficile. Also, our data enabled us to distinguish at least three genetically distinct phage groups. The first group is composed of genetically related Myoviridae phages, including φCD5, φCD52, φCD630-1, φCD630-2, φC2, φC5, φC8, and φCD119. The other two groups, composed of Siphoviridae phages, are more heterogeneous. The siphophage φCD24 is the only member of the second group, whereas the third group includes the Siphoviridae phages φCD8-1, φCD8-2, φCD38-1, and φCD38-2. Finally, an inducible prophage was found within the genome of the hypervirulent strain NAP1/027 (QCD-32g58), which was shown to be responsible for outbreaks in North America and Europe. We failed to detect any DNA similarity between this phage and genes in the PaLoc or encoding the binary toxin. Thus, whether this prophage increases the fitness of its host remains to be determined.