The complete genome sequence of Clostridium difficile phage {phi}C2 and comparisons to {phi}CD119 and inducible prophages of CD630

Clostridium difficile has emerged as an important intestinal pathogen since the 1970s, continuing to plague hospital settings worldwide and causing recent epidemics in the USA (McDonald et al., 2005) and Canada (Loo et al., 2005; Warny et al., 2005). The major virulence factors of C. difficile are toxins A and B encoded by tcdA and tcdB respectively, which are located on a 19 kb genomic region termed the pathogenicity locus (PaLoc) (Braun et al., 1996; Hammond & Johnson, 1995). The toxins are positively regulated by TcdR (Mani & Dupuy, 2001; Rupnik et al., 2005) and are negatively regulated by TcdC (Hundsberger et al., 1997; Matamouros et al., 2006); they are encoded by tcdR and tcdC respectively, also on the PaLoc. Another toxin-associated gene, tcdE, appears phage related but its function is unknown (Tan et al., 2001). C. difficile acquires antibiotic resistance and virulence genes through plasmids and transposons (Bruggemann, 2005) shown to significantly contribute to genome plasticity (Sebaihia et al., 2006). It is possible that phages also contribute to variance in virulence-associated genes (Lemee et al., 2005) and to the emergence of outbreak strains (McDonald et al., 2005). However, the prevalence of phage genes within C. difficile genomes is not known. In comparison to phages of Escherichia coli, Staphylococcus aureus and Lactobacillus species, the study of clostridial phages is in its infancy. Only one phage specific for C. difficile, temperate phage φCD119, has been sequenced (Govind et al., 2006), while two putative prophage sequences were detected in the recently sequenced genome of C. difficile CD630 (Sebaihia et al., 2006).

Previously, we induced a temperate phage, φC2, from a clinical C. difficile isolate. The phage was partially sequenced and characterized (Goh et al., 2005b) and was shown to increase toxin B levels in lysogens (Goh et al., 2005a). Two other induced temperate phages, φC6 and φC8, were also shown to have this effect in C. difficile lysogens (Goh et al., 2005a). While C. difficile toxin production itself was not phage mediated, phages may have some other role in host physiology. In this study, we compared the φC2 genome with the genomes of φCD119 and C. difficile CD630 (Sebaihia et al., 2006), as well as the unfinished sequence of C. difficile QCD-32g58. We found φC2 genes to be prevalent in most of the clinical C. difficile isolates tested. Two phages, designated φC630-1 and φC630-2, were induced from CD630 in the course of this study.

Bacterial strains, phage and growth conditions.
Thirty-four clinical C. difficile strains from Singapore General Hospital, and Sir Charles Gairdner Hospital, Western Australia, were used in this study (Table 1). Reference strains used were CCUG 37782, CCUG 20309 and CCUG 16126, which were purchased from the Culture Collection, University of Göteborg, Sweden. VPI 10463 was generously provided by Dr M. Rupnik, University of Ljubljana, Slovenia, and CD630 was kindly provided by Dr P. Mullany, University College London, UK. For induction of temperate phage, an overnight culture of CD630 in Brain Heart Infusion broth (BHIB, Oxoid) supplemented with 5 % horse blood, 50 µg erythromycin ml¹ (Sigma) and 10 µg tetracycline ml¹ (Sigma) was induced with 3 µg mitomycin C ml¹ (Sigma) and incubated for another 8 h. The culture supernatant was filtered through a 2 µm membrane (Pall), then assayed for phage against 10 randomly chosen clinical C. difficile strains (Table 1) as previously described (Goh et al., 2005b). Propagation of φC630-1 and φC630-2 with a 4 h culture of CD843 was as previously described for φC6 (Goh et al., 2005b). φC2 was propagated on CD062 as previously described (Goh et al., 2005b). C. difficile strains were maintained in Cooked Meat Medium (Oxoid), from which 37 °C overnight cultures in BHIB were prepared. Exponential-phase cultures were prepared by subculturing 1 ml of an overnight culture in 9 ml BHIB and incubating at 37 °C for 4 h.

Table 1. Bacterial strains, phages and plasmids

Molecular cloning and DNA sequencing and analysis.
Phage purification by CsCl density gradient, DNA extraction by phenol/chloroform/isoamyl alcohol and purification using the Wizard DNA Clean-up System (Promega) was as previously described (Goh et al., 2005b). Phage DNA was also extracted directly from crude lysate using the Qiagen Lambda Midi Kit, according to the manufacturer's instructions. φC2 DNA was digested with AccI, HincII, HindIII or XbaI and cloned into pUC19. Positive clones were selected for by bluewhite colony screening, PCR using M13 primers and restriction enzyme digestion of recombinant plasmids. Sequencing was performed by the dideoxy chain-termination method with an automated ABI Prism 3100 DNA Sequencer (Applied Biosystems). Primers were designed and sequences were assembled using the Lasergene version 5.05 software (DNASTAR). A mean coverage of 3.53x (293 sequencing runs) and a minimum of 2x coverage (at least once in each strand) were obtained from sequencing the phage library. Gaps between contigs of φC2 were filled and ends of the genome were sequenced by primer walking using phage genomic DNA as template. Probable protein-encoding genes (ORFs) were predicted using GeneMark.hmm VIOLIN and GeneMark.hmm for prokaryotes version 2.4 programs (Besemer & Borodovsky, 1999). Predicted ORFs were searched for similarity to proteins in databases by BLASTP (Altschul et al., 1990). Nucleotide similarity between φC2 and C. difficile CD630 or QCD-32g58 was detected by BLASTN (Altschul et al., 1990) at and , respectively. Alignments between two sequences were carried out with bl2seq (Tatusova & Madden, 1999) and the genome was searched for tRNA genes using the tRNAscan web server (Lowe & Eddy, 1997). Cumulative GC skew was carried out with GenSkew at , and a hydrophobicity plot was generated using Hydrophobicity grapher using the KyteDoolittle scaling system (Kyte & Doolittle, 1982). Transmembrane regions and beta-turns in an ORF were predicted by TMPRED at and BTPRED (Shepherd et al., 1999), respectively. Tandem Repeats Finder (Benson, 1999) was used to detect direct or inverted repeats in the genome.

Protein expression.
Plasmid pQE-hol and pQE-AbiF contained φC2 putative ORF 36 and ORF 37, respectively, cloned between the BamHI and PstI sites of the pQE-30 expression vector (Qiagen). ORF 36 was amplified with Vent polymerase (NEB) using HolFBam (5'-CGCGGATCCATGGATAATTTAATAAG-3')/HolRPst (5'-AACTGCAGTTACTTTTCACCATCCT-3') with cycling conditions of 95 °C for 4 min, 30 cycles of 95 °C for 30 s, 53 °C for 30 s, 72 °C for 1 min and 72 °C for 10 min. ORF37 was amplified with Vent polymerase (NEB) and AbiFBam (5'-CGCGGATCCATGGTTGAAGTAAAAGA-3')/AbiFRPst (5'-AACTGCAGTTATTTAGCCAATATCTC-3') primers with cycling conditions of 95 °C for 2 min, 30 cycles of 95 °C for 30 s, 58 °C for 30 s, 72 °C for 2 min and 72 °C for 10 min. The PCR product was digested with BamHI and PstI and ligated into pQE-30 with complementary ends, then transformed into M15[pREP4] by electroporation (Bio-Rad Gene Pulser II). Recombinant pUC19 (NEB) plasmids harbouring φC2 DNA inserts were grown at 37 °C in LB (Invitrogen) supplemented with 100 µg ampicillin ml¹ (ICN Biomedicals), 62 ng X-Gal ml¹ (Bio-Rad) and 0.625 mM IPTG (Sigma). Recombinant pQE-30 (Qiagen) expression plasmids were grown at 37 °C in LB supplemented with 100 µg ampicillin ml¹ and 25 µg kanamycin ml¹ (Sigma). Protein expression in M15[pREP4] cells was induced with 1 mM IPTG when growth of culture at 37 °C with shaking at 250 r.p.m. had reached an OD₆₀₀ of 0.6.

Southern hybridization and dot blot.
φC2 genomic probe was prepared by DIG-labelling of HincII-digested phage DNA according to the DIG High Prime DNA Labelling and Detection Starter Kit 1 (Roche) instructions. C. difficile and phage DNA were digested by HincII or HindIII and XbaI, separated in 1 % TBE agarose and transferred to a nylon membrane (Amersham) as previously described (Sambrook et al., 1989). Hybridization was carried out at 3739 °C for 1620 h, followed by stringent washing and colour development of membranes as in the system manual. Dot blots were carried out by spotting 120200 ng chromosomal DNA on nylon membranes, fixing the DNA by microwave on high power for 2.5 min (Angeletti et al., 1995) followed by hybridization to φC2 genomic probe, according to the DIG system manual (Roche).

CHEF electrophoresis.
Undigested phage DNA (150200 ng) was added to an equal volume of molten 1 % low-melting-point Agarose (Sigma), then loaded into a 1 % TBE Pulse Field Certified Agarose (Bio-Rad). Electrophoresis was carried out in 0.5x TBE running buffer using the CHEF-DR II Pulse Field Electrophoresis System (Bio-Rad) at pulse times of 313 s, 200 V for 22 h.

SDS-PAGE and N-terminal sequencing.
SDS-PAGE analysis of phage was carried out as previously described (Ford et al., 1998) with modifications. φC2 was purified through a CsCl density gradient as previously described (Goh et al., 2005b); 20 µl (10⁷ p.f.u. ml¹) was repeatedly frozen in liquid nitrogen and thawed at 37 °C three times, heated at 75 °C for 4 min, mixed with 5x sample buffer (0.255 M Tris pH 6.8, 50 %, v/v, glycerol, 5 % w/v SDS, 0.05 %, w/v, bromophenol blue, 0.25 M DTT) and heated again at 95 °C for 5 min before electrophoresis. Electrophoresis was carried out as previously described (Laemmli, 1970) using the Mini-PROTEAN II cell (Bio-Rad) and 12 % acrylamide gels, which were stained with BLUPRINT Fast-PAGE Stain (Gibco BRL). Proteins were electroblotted to PVDF membranes (Bio-Rad) as described by Sambrook et al. (1989) using cold tank blotting transfer buffer (25 mM Tris pH 8.3, 150 mM glycine, 20 % v/v methanol). The PVDF membrane was stained with Coomassie blue R250 and two major bands were excised from the membrane. Ten amino acids from the N-terminus of each protein were determined by an automated sequencer (Applied Biosystems 477 Protein Sequencer).

Determination of the attP region.
The attP region was predicted to be downstream of the integrase, and primer pair patt-1/int-1 was used to generate a 311 bp PCR product which was DIG labelled and used as an attP probe for hybridization at 39 °C. Primer sequences are: patt-1, 5'-GTAAAGATGATGAAGTGGATGAAG-3'; int-1, 5'-GCATTTTACAATAATTTGCCACCG-3'. The cycling conditions were 95 °C for 5 min, 30 cycles of 95 °C for 30 s, 50 °C for 30 s, 72 °C for 1 min and 72 °C for 10 min. Genomic DNA of CD242 (5 µg) was digested with AccI, HincII or HindIII and self-ligated with T4 DNA ligase. Inverse PCR (Ochman et al., 1988) was carried out with divergent primer pairs patt-0/Cdu1-1 and int-3/int-5. Primer sequences were as follows: patt-0, 5'-CTGTGAATGTAGTTGATTCATTC-3'; Cdu1-1, 5'-GGTTAGAGCGAAGGGAGTTT-3'; int-3, 5'-CAACCACTATGGACACATATTC-3'; int-5, 5'-CGCAAGCAATGAAAATTAAAG-3'. The PCR products were sequenced with the same primers. To determine attBP', an additional primer, gntR-F (5'-GGATTTAGAAGTAAATTCC-3') was used.

Electron microscopy.
Carbon-coated copper Formvar grids (400 mesh, SPI supplies) were floated on 20 µl 0.01 % poly-L-lysine for 30 s and excess liquid was wicked off with a filter paper. The grid was floated on 20 µl phage suspension (10⁷10⁹ p.f.u ml¹) for 3 min and dried as above, then fixed with 20 µl 1 % glutaraldehyde for 1 min and negatively stained with 1 % phosphotungstic acid pH 7.4 for 3090 s. The grids were air-dried for 30 min and viewed under a Philips 2085 transmission electron microscope at 57 00089 000x magnification, operating at 100 kV. Dimensions of a minimum of four phage particles for each phage were measured, and results expressed as mean±SD.

General features of φC2 genome
The dsDNA linear genome was 59.7 kb as estimated by CHEF electrophoresis (Fig. 1a) while the unit genome length was 56 538 bp as determined by sequencing, indicating terminal redundancy of approximately 3.2 kb in the packaged genome. The G+C content was 28.72 mol%, which is identical to the recently sequenced φCD119 (Govind et al., 2006) and slightly lower than that of the C. difficile CD630 genome (29.06 %). A total of 84 putative ORFs were identified, of which seven (8 %) showed no homology to proteins in the NCBI and EBI databases; 37 ORFs (44 %) had sequence similarity to proteins with unknown functions, of which 13 ORFs (15 %) were found only in C. difficile phage/prophage genomes. A supplementary table detailing the predicted ORFs and similarities to other proteins is included with the online version of this paper. In general, the predicted ORFs in the left and right arms of the genome had sequence similarity to φCD119 and CD630 prophages 1 and 2, while ORFs in the middle section were similar to various bacterial or phage proteins. No programmed frameshift signals for translation were detected. The proteins of φC2 were separated by 12 % SDS-PAGE (Fig. 1b) and two major proteins of φC2 were identified as products of ORF 7, a putative capsid, and ORF 13, a putative tail sheath with the N-terminal methionine removed as expected (Ben-Bassat et al., 1987; Romero et al., 2004). Cumulative GC skew analysis of the genome sequence, included as a supplementary figure with the online version of this paper, revealed a putative origin of replication in the region of nt 37929 and a putative terminus of replication in the region of nt 30390.

(30K): Fig. 1. CHEF electrophoresis of undigested φC2 DNA (a) and SDS-PAGE of phage proteins (b). (a) The genome of φC2 was estimated to be 59.7 kb. M1, 5 kb molecular mass marker (Bio-Rad). (b) The major proteins were sequenced from the N-terminus for 12 residues. The sequences correspond to capsid and sheath proteins. M2, protein molecular mass markers (Fermentas).

Mosaic structure of φC2 genome
The genome is organized into six modules starting with a DNA packaging module from the left end of the genome, which has been assigned an arbitrary start from the putative terminase (Fig. 2). There was consistently high homology to ORFs of CD630 prophage 1 and 2, while sequence similarity to φCD119 was sporadic. Of interest within the tail assembly module is a cassette of genes found in phage EJ-1 of Streptococcus pneumoniae (Romero et al., 2004). The cell lysis module consisting of ORFs 36, 37 and 38 was atypical in the relative positions of the putative endolysin and holin, which are usually adjacent to each other (Young, 1992). Putative holins of the C. difficile phages had high sequence similarity to one another and good alignment with holin of another phage, despite low sequence homology (Fig. 3). A hydrophobicity plot (not shown) of ORF 36 revealed transmembrane regions separated by β-turns and outside N- and C-termini typical of type II holins (Wang et al., 2000). Constitutive expression of ORF 36 inhibited growth and induced expression of ORF 36 decreased turbidity in E. coli, while expression of ORF 37 did not have an effect on cell growth (Fig. 4). Similar results have been shown for other phage holins (Muyombwe et al., 1999; Sheehan et al., 1999) and for C. difficile tcdE (Tan et al., 2001), which had good sequence alignment with holins (Fig. 3).

(39K): Fig. 2. Genome of φC2. Direction of transcription and predicted ORFs are represented by block arrows which are labelled with corresponding putative functional assignments. Organization of the genome is broadly categorized into six genetic modules. Nucleotide homology to prophage 1 and 2 in CD630 and corresponding positions are shown with green and yellow bars, respectively. Orange and blue bars represent regions of nucleotide similarity to φCD119 and QCD-32g58, respectively. ORFs unique to φC2 are in red, ORFs common to φC2, φCD119 and prophages in CD630 are in brown.

(56K): Fig. 3. Multiple sequence alignment between the putative holin of φC2 (ORF36) and putative holins of φCD119, CD630 prophage 1 (ORF CD0971), CD630 prophage 2 (ORF CD2894A), Bacillus clarkii phage BCJA1c and TcdE of Clostridium difficile 88864. Identical amino acids are shaded in black while conserved substitutions are shaded in grey. Predicted transmembrane (TM), hydrophobic and hydrophilic regions are indicated, as well as their predicted topology in italics.

(14K): Fig. 4. Expression of putative phage holin gene and tcdE of C. difficile in M15[pREP4] cells was induced with 1 mM IPTG. OD600 of the cultures were measured over time. A decrease in OD600 was observed in the pQE-hol culture (•) up to 4 h post-induction. A similar trend was observed for tcdE expression (pQE-tcdE, ). Expression of another phage gene (abiF) did not result in cell growth retardation (pQE-abiF, ).

Surprisingly, ORFs 41 and 42 were homologous to ParA and ATPase/ParB of the Leptospira biflexa phage LE1 (Bourhy et al., 2005). These enzymes, together with a centromere-like sequence, parS, are required for maintenance of DNA stability (Austin & Abeles, 1983; Yamaichi & Niki, 2000); therefore ORFs 41 and 42 could be part of the lysogeny module. Possible parS sequences in the form of short direct or inverted repeats (Dam & Gerdes, 1994; Gallie & Kado, 1987; Radnedge et al., 1996) around ORFs 41 and 42 were not detected, as was the case for LE1 (Bourhy et al., 2005). Sequence similarities of ORF 41 to TcdB (22 % of 279 amino acids, E-value 1.1) and ORF 46 to Cdu1 (29 % of 125 amino acids, E-value 4x10¹²) are noteworthy. The φC2 lysogeny module is unusual in being extended compared to that of λ (Birge, 2000); there appear to be two sets of repressors (ORFs 49, 50, 58 and ORFs 51, 53, 61) and perhaps as a consequence, two antirepressors (ORFs 52 and 55). ORF 59 had low percentage sequence similarity to an excisionase, the first to be detected in C. difficile phages. Although not shown, homologues of ORFs 1014, 16, 2022 and 2426 have also been detected in the Clostridium sp. strain OhILAs sequence with high percentage sequence similarity. The attP of φC2 in CD242 was predicted to be between ORF 46 (transcriptional regulator/Cdu1 homologue) and ORF 47 (integrase), and was confirmed by Southern hybridization (Fig. 5a). The attP region was flanked by AccI, HincII and HindIII sites; digestion of lysogenic DNA with either enzyme was used for inverse PCR; however, only an attPB' PCR product was obtained (Fig. 5b, c). As the bacterial sequence of attPB' had high percentage similarity to a gntR transcriptional regulator in CD630 (92 % of 93 amino acids, E-value 1.4x10³⁸) and QCD-32g58, a forward primer specific for gntR (gntR-F) and int-3 were used to determine the sequence of attBP' (Fig. 5c). The φC2 attachment site sequence CTGTGAGAAAT is different from that of φCD119 (TTTATATGTGTTAT), CD630 prophage 1 (TAAAGATGA) and prophage 2 (TCCACTAGG). Interestingly, the translated 3' end of the integrated CD630 prophage 1 (nt 11433021143688) was similar to Cdu1 (36 % of 125 amino acids, E-value 3x10¹¹) and attP of φCD119 was 248 nt downstream of a Cdu1 homologue (Govind et al., 2006).

(29K): Fig. 5. Integration of φC2 DNA into the C. difficile genome. (a) Southern hybridization of attP probe to phage and lysogenic DNA digested with HincII. Although cross-hybridizations of the attP probe to other fragments of φC2 were observed, the brightest signal was a 2.4 kb band, which was absent in lysogen CD242. Two new observed bands of 1.1 kb and 5.7 kb may contain the attPB' and attBP'. M, DIG-labelled/HindIII marker (Roche). (b) φC2 DNA before integration, where the attPP' sequence (bold) was between ORF46 (black arrow) and ORF47 (grey arrow). The attPP' aligned with a homologous attBB' sequence (bold) in a gene encoding a gntR-like transcriptional regulator, present in the host genome. Divergent PCR primers patt-0/Cdu1-1 and int-3/int-5 were used to amplify and sequence the attPB' and attBP', respectively. Primers patt-1/int-1 were used to amplify the attPP' region to use as a probe for Southern hybridization in (a). (c) Right side (attPB') of integrated phage DNA in the lysogenic C. difficile genome by inverse PCR and left side (attBP') by PCR with gntR-F and int-3 primers. φC2 integration resulted in a disrupted gntR. C. difficile sequences are in lower case; φC2 sequences are in upper case; the homologous att site is in bold.

φCD119 and prophage 1 and 2 homologues were mostly found in the DNA replication, recombination and modification module of the φC2 genome. ORF 65, common to the C. difficile phages, is likely an essential recombination function (Erf) protein that is a member of a superfamily of single-strand annealing proteins involved in phage genome circularization via homologous recombination following DNA entry (Iyer et al., 2002). Also common was ORF 78, with high sequence similarity to RusA, an enzyme thought to have phage origins (Sharples et al., 2002). RusA is a DNA endonuclease that resolves Holliday junctions in DNA replication, recombination and repair (Mahdi et al., 1996).

Relatedness of φC2, φC630-1, φC630-2 and φCD119 and prevalence of prophage genes in C. difficile isolates
Mitomycin C induction of CD630 resulted in two plaque types which corresponded to phage particles of the same morphology but having slightly different head sizes. Since the genome of prophage 1 (55 850 bp) in CD630 is larger than prophage 2 (49 178 bp), it is likely that prophage 1 produced larger particles (φC630-1) compared to prophage 2 (φC630-2). φC630-1 particles measured 31.7±0.7 nm in head diameter and 62.4±5.1 nm in tail length, while φC630-2 particle dimensions were 28.1±1.3 nm (head) and 39.5±5.8 nm (tail) (Fig. 6). BLASTN identified regions of nucleotide sequence similarity throughout the φC2 genome to regions of CD630 that indicated the φC630-1 and φC630-2 prophages have similar genome organization (Fig. 2). Dot plots (not shown) of φC2 and the hypothetical genome sequences of φC630-1 and φC630-2 were virtually identical. Hence, common frameshift regions in φC2 were detected at nt 2629026996, nt 3293333469, nt 4735050177 (φC630-1) and nt 4254653713 (φC630-2). Nucleotide comparison of φC2 and φCD119 genomes by dot plot (not shown) revealed some regions of similarity (Fig. 2) but were a less similar pair compared to φC2/φC630-1 and φC2/φC630-2. Dot-plot comparisons of φCD119 to φC630-1 and φC630-2 revealed sequence similarity in parts of the cell lysis, lysogeny control and DNA replication and modification modules of the three phages.

(114K): Fig. 6. Particle morphology of CD630 phages. (a) φC630-1, (b) φC630-2 with contracted sheath. Bars, 50 nm.

To evaluate the prevalence of φC2-related genes in clinical C. difficile strains, dot blots and Southern blots were carried out with a φC2 genomic probe. The results showed that out of 37 strains tested (Table 1), only six (CD13, CD57, CD62, CD062, CD843 and CCUG 37782) did not contain prophage genes; 31 strains including four typed reference strains exhibited homology to varying degrees. Differences in dot-blot signals were further analysed by Southern hybridization (not shown). The most common homologous bands found in almost all strains correlated to ORFs 713 of the head structural module and ORFs 7884 of the DNA replication, recombination and modification module, respectively. Regions representing modules of lysogeny control, lysis, head and DNA methylase were also found in some strains. To determine whether φC2-related phage genes were present in another sequenced C. difficile strain, QCD-32g58, a BLASTN search was carried out. Only three regions of high sequence similarity were found between QCD-32g58 and φC2 (Fig. 2). However, a whole prophage genome was not detected in QCD-32g58. In the last few years there has been an exponential increase in sequenced phage genomes made available for comparison. This has resulted in the currently accepted view of divergent phages being related by virtue of genetic modules (Hendrix, 2002; Juhala et al., 2000) and co-evolution of bacterial hosts and phage (Kwan et al., 2005; Pedulla et al., 2003). An aim of this study was to compare the genome of our previously partially characterized phage φC2 to other phage genomes to provide an insight into phages of C. difficile.

The genome organization of φC2 is typical of phages with a lysogenic cycle (Canchaya et al., 2003). Its genome length was previously underestimated by addition of restriction fragments separated under normal electrophoresis (Goh et al., 2005b). It was previously shown not to possess cohesive ends (Goh et al., 2005b) and genome sequencing revealed it to have terminally redundant ends. The packaging mechanism of a terminase may be predicted from the large subunit amino acid sequence (Casjens et al., 2005); the φC2 terminase large subunit had high percentage sequence similarity to SPP1 (26 % of 426 amino acids, E-value 3x10²⁵) of the P22-like headful subgroup. This suggests the ends of φC2 were likely to be generated by a headful packaging mechanism. Based on the number of protein homologues and unique hypothetical proteins, φC2 is closely related to the other C. difficile phages, φCD119, φC630-1 and φC630-2, and demonstrates modular mosaicism (Casjens, 2005). Gene divergence appears greatest within the lysogeny control module, followed by tail structural proteins, which is common for tailed temperate phages as a method of diversifying infectable hosts (Casjens, 2005). Tail structural components of φC2 may have been derived from an ancestral phage of EJ-1, while a putative LysM within tail-associated proteins may indicate lytic enzymes are used for local cell wall degradation and hence penetration of host wall for injection of phage DNA, similar to tailspike proteins of P22, Sf6 (Freiberg et al., 2003) and T4 (Kanamaru et al., 2005) and tail fibre proteins of anti-K1 phages (Muhlenhoff et al., 2003). Homologues of extrachromosomal replicative proteins ParA and ParB found close to the lysogeny module and their relative direction of transcription suggest their expression is associated with lysogenic conversion. Interesting possibilities include φC2 switching to an LE1-like replicative prophage state for stability and perhaps having a pseudolysogeny phase (i.e. genome does not integrate into host chromosome but rather exists as a circular intermediate), or that it was once a replicative prophage. Low percentage sequence homology between ORF 41 and TcdB suggests phage origins of the toxin and may explain the genetic variability in tcdB, which has been observed in C. difficile toxinotypes (Rupnik et al., 2001), more commonly in tcdB than tcdA (Rupnik et al., 1998). Sequence similarity of ORF 46 to Cdu1, which borders the PaLoc, and the holin having a similar effect on E. coli as TcdE also point toward the PaLoc as a collection of genes transferred by phages, which have evolved to become the current virulence genes of C. difficile. ORF 46 (Cdu1 homologue) and its downstream non-coding region appear to be involved in integration because the Cdu1-attP-integrase arrangement is conserved in φC2, φC630-1 and φCD119. The attPs of φC2 and φCD119 are 197 nt and 248 nt downstream of a Cdu1 homologue (Govind et al., 2006), respectively, while cdu1 contains an attP for φC630-1 (Sebaihia et al., 2006). There does not appear to be a preferred site for the integration of this group of phages in C. difficile, as the attachment sites are different for each phage. The presence of an AbiF protein on the phage genome is unusual; it is normally carried on a plasmid and confers phage resistance to bacteria, resulting in an abortive phage infection at the level of phage DNA replication (Garvey et al., 1995). The advantage of carrying abiF is unknown and has not been found on other phage genomes; one possibility could be to prevent superinfection of C2 lysogens by unrelated phages that are susceptible to AbiF.

Comparative DNA sequence analysis of φC2, φC630-1, φC630-2 and φCD119 showed the degree of pairwise relatedness to be φC630-1/φC2>φC2/φC630-2>φC630-1/φCD119>φCD119/φC630-2>φCD119/φC2. This indicates that φC630-1 and φC630-2 are intermediates of φC2 and φCD119 in the evolutionary sense. The brief sampling of randomly chosen clinical C. difficile isolates revealed that φC2-related prophage genes are prevalent in C. difficile. Although these may represent only phage remnants, the apparently low prevalence of inducible phage (Goh et al., 2005b) may simply be due to the lack of appropriate indicator strains. Isolates possessing both tcdA and tcdB (e.g. CD61) were as likely to be devoid of φC2-related genes as isolates possessing truncated versions of tcdA and/or tcdB (e.g. CCUG 20309, CD843), or lacking either or both toxin genes (CCUG 37782, CD55). In general, there was no correlation between the presence of tcdA and tcdB and prevalence of φC2-related prophage genes in C. difficile strains. Therefore, the current role of φC2 is not in generating genetic diversity within the PaLoc but perhaps in other areas of the host genome related to virulence. The following ORFs may affect host fitness: ORF 19 (putative TerD), ORF 9 (sequence similarity to the Alkaliphilus metalliredigenes sigma-54 interaction region, 28 % of 114 amino acids, E-value 0.054) and ORF 15 (low sequence similarity to Clostridium thermocellum S-layer protein). φC2 disruption of a gntR-like transcriptional regulator potentially affects expression of host genes, which may lead to altered fitness. In CD630, gntR was upstream of genes encoding the mannose-specific phosphotransferase system (PTS) (Sebaihia et al., 2006). The mannose PTS is involved in sugar transport and global regulation of gene expression, in a number of Gram-positive genera (Abranches et al., 2003; Arous et al., 2004; Chaillou et al., 2001; Reizer et al., 1999), including the regulation of energy metabolism and virulence genes in Streptococcus mutans (Abranches et al., 2006). Hypothetically, integration of φC2 into CD630 could lead to significant changes in the C. difficile phenotype through mannose PTS deregulation. The contribution of φC2 and related temperate phages to the physiology of C. difficile and their potential roles in gene transfer and as genetic tools for this species are worthy of further investigation.

We thank Mark Schreiber for bioinformatics advice and support throughout the course of this work and Josephine Howe for help with electron microscopy.

Edited by: P. R. Herron

Footnotes

^,†, Peh Fern Ong¹, Keang Peng Song¹ †Present address: Department of Cell and Molecular Biology, Programme for Genomics and Bioinformatics, Karolinska Institutet, Berzelius väg 35, SE-171 77 Stockholm, Sweden. ‡Present address: School of Arts and Sciences, Monash University Malaysia, No. 2 Jalan Kolej, Bandar Sunway, Petaling Jaya 46150, Selangor Darul Ehsan, Malaysia.