Typing by sequencing the slpA gene of Clostridium difficile strains causing multiple outbreaks in Japan

It is well known that Clostridium difficile readily spreads nosocomially (Brazier, 1998; Johnson et al., 1999; Kato et al., 2001). It has been documented that specific strains cause multiple nosocomial outbreaks over the world. In the USA (Samore et al., 1997; Johnson et al., 1999), a single strain or two highly related types, i.e. restriction enzyme analysis (REA) type J9 and J7/arbitrarily primed PCR (AP-PCR) type 01/AP-PCR type A, have been shown to be involved in epidemics in a number of hospitals. The epidemic strain that caused an outbreak in New York was typed as serogroup G (H. Kato et al., 1993) and PCR ribotype gr (Kato et al., 2001). Brazier and others have reported that PCR ribotype 1 (Stubbs et al., 1999), which corresponded to serogroup G (Brazier et al., 1997), was responsible for most of the published outbreaks in the UK (Brazier, 1998). In addition, they typed a USA outbreak strain in Massachusetts as PCR ribotype 1 (Brazier, 1998). In Japan, one PCR ribotype, type smz, has been identified as frequently causing outbreaks (Kato et al., 2001). Interestingly, the epidemic strains in both the UK and the USA as well as in Japan, i.e. the type smz strain, were non-typable by PFGE typing because of DNA degradation (Samore et al., 1997; Corkill et al., 2000; Kato et al., 2001).

The typing results of epidemic strains of outbreaks previously documented are summarized in Table 1. To investigate the pathogenicity of these specific strains, it is necessary to compare the epidemic strains that are responsible for outbreaks across the world. Since it is not easy to exchange reference strains or antisera among different laboratories in different countries, typing systems in which results do not depend on banding-pattern analysis are required for a study of global epidemiology.

Table 1. Isolates used and their typing designation in comparison with epidemic strains reported from the USA and UK HK, designated by Kato et al. (2001); JSB, designated by Brazier et al. (1998); GEK, designated by Killgore et al. (1994); MS, designated by Samore et al. (1997).

Previous reports have documented that a surface layer protein (SlpA) varies among C. difficile isolates (McCoubrey & Poxton, 2001; Calabi et al., 2001) and typing results by RFLP analysis of the slpA gene correlate well with those by serogrouping (Karjalainen et al., 2002). In this study, we preliminarily identified the slpA gene of PCR ribotype smz strains recovered at a number of hospitals in Japan. One of our aims was to examine the variability of the slpA gene among smz strains; a second aim was to compare the sequence results with those of C. difficile previously available in databases as well as with that of the epidemic strain that caused an outbreak in the USA; a third aim was to apply slpA sequence typing to typing of C. difficile directly from stool specimens without culturing.

Bacterial strains and stool specimens.
Ten C. difficile isolates typed as PCR ribotype smz were obtained from patients at 10 different hospitals in diverse areas of Japan. The type smz strain was predominant at eight of these 10 hospitals, suggesting a nosocomial outbreak of the strain at those eight hospitals; a strain of a type other than smz was predominant at another hospital, but no dominant strains were found at the remaining hospital.

Included in the present study were one epidemic strain, strain 900233 (serogroup G/AP-PCR type 01/PCR ribotype gr), causing an outbreak in a hospital in upstate New York (H. Kato et al., 1993, 2001; Killgore & Kato, 1994), and a type smz strain, strain 900295, recovered during this New York outbreak (Kato et al., 2001). The epidemic strain from the same hospital in New York was reported to be AP-PCR type A/REA type J9 (Samore et al., 1997; Johnson et al., 1999).

Stool specimens were obtained with the informed consent of patients admitted to Kumiai Hospital, Gifu, Japan who were given the diagnosis of antibiotic-associated diarrhoea. The isolation of C. difficile from stool specimens and the identification of C. difficile were performed by methods described previously (Kato et al., 1998). Stool specimens were tested for toxin A using an immunoassay kit (Oxoid) and for toxin B using a Vero cell cytotoxicity assay with a neutralization test by anti-C. difficile toxin B serum (TechLab).

Typing of recovered isolates.
The toxin-producing type of isolates was determined by a PCR method described previously (Kato et al., 1998, 1999). PCR ribotyping was performed as described previously (Stubbs et al., 1999; Kato et al., 2001). The variable region of the slpA gene was amplified by the primers slpAcom19 (sequence 5'-GTTGGGAGGAATTTAAGRAATG-3') and slpAcom22 (sequence 5'-GCWGTYTCTATTCTATCDTYWCC-3'). The PCR assay was performed in the same manner as described for the PCR detection of the toxin genes (Kato et al., 1998) with a modification of the thermal profile, i.e. 35 cycles comprising 95 °C for 20 s and 55 °C for 180 s. The PCR product was purified and sequenced directly with an ABI Prism dye-terminator cycle sequencing ready reaction kit (PE Applied Biosystems) as described previously (Kato et al., 1999). Both strands of the amplified products were sequenced.

Direct typing on stool specimens.
DNA was extracted from stool specimens using the QIAamp DNA stool mini kit (Qiagen). The primers used for a nested PCR detecting the non-repeating region of the toxin A gene (tcdA) from DNA extracted from stool specimens were HK5 (sequence 5'-AACTTATCCTGGGAAGTTGC-3') and HK6 (sequence 5'-GTGAATTATATTTTTAACAGGCTC-3') for the outer primer set, and NK3 (sequence 5'-GGAAGAAAAGAACTTCTGGCTCACTCA GGT-3') and NK2 (sequence 5'-CCCAATAGAAGATTCAATAT TAAGCTT-3') for the inner primer set (N. Kato et al., 1993).

The primers used for sequence typing of the slpA gene directly on DNA extracted from stool specimens were slpAcom19-slpAcom22 for the outer primer set, and slpAcom13 (sequence 5'-TAGGTGATGGAR AWTAYGTWG-3') and slpAcom12 (sequence 5'-CATAWBBTT TAGCTAAATYTTBWGC-3') for the inner primer set. The primers slpAcom19, slpAcom22, slpAcom13 and slpAcom12 were selected from the conserved regions of the slpA gene among five PCR ribotype strains frequently recovered in Japan, including a type smz strain.

The PCR assay on DNA extracted from the specimens was performed in a reaction volume of 50 µl consisting of 2.5 mM MgCl₂, 50 mM KCl, 10 mM Tris/HCl (pH 9.0), the four deoxynucleoside triphosphates (200 µM each), 12.5 pM of each primer, 1.25 U of Taq DNA polymerase (Promega), 5 µg of BSA and 5 µl of extracted DNA (or PCR product for the second PCR). The thermal profile was 35 cycles comprising 95 °C for 20 s and 55 °C for 180 s for the primer pair slpAcom19-slpAcom22, and 35 cycles comprising 95 °C for 20 s and 55 °C for 120 s for the other primer pairs. The PCR product from the second PCR with primers slpAcom13-slpAcom12 was sequenced in the same manner as described for the slpA sequence typing of isolates.

Results and Discussion
The slpA gene of a type smz strain, GAI 97660 (Kato et al., 2001), recovered from a patient with pseudomembranous colitis, was amplified by primers slpAcom19 and slpAcom22, and the resultant PCR product was sequenced. The deduced amino acid sequence of slpA in GAI 97660 showed the highest identity of 76 % with that of strain 9354 (GenBank accession no. AF448120), which was characterized as serogroup A, subgroup unknown (Karjalainen et al., 2002).

Among 10 type smz strains from different areas of Japan, subtypes were identified (type smz-1, smz-2 and smz-3) that differed from each other by one nucleotide, involving a single deduced amino acid that differed for each one. The slpA gene of three strains, GAI 97660 (type smz-1), MRY 04-0409 (type smz-2) and MRY 04-0410 (type smz-3), was sequenced after 10 subcultures on Brucella HK agar plates (Kyokuto Seiyaku). In each of the three isolates, the sequence results following serial subcultures agreed with those before subculturing. The numbers of strains typed as slpA sequence types smz-1, -2 and -3 were five, four and one, respectively. Strains of all three slpA sequence types were detected from outbreaks, and strains of smz-1 and smz-2 types from sporadic cases.

The slpA sequence of the New York outbreak strain, 900233, showed 38 % homology to that of GAI 97660 and was 100 % identical to that of strain R8366 (GenBank accession no. AJ300676), which was serogrouped as G by Delmee's group (Karjalainen et al., 2002) and typed as PCR ribotype 1 by Brazier's group (Stubbs et al., 1999). During the outbreak at the New York hospital, PCR ribotype smz strains were recovered from three patients but, unlike the PCR ribotype gr strain, they did not appear to be involved in the epidemic. One of the three smz isolates, 900295, was typed as smz-2 by slpA sequence typing.

These sequence results indicate that the type smz strain clearly differs from the strain causing the New York outbreak not only by PCR ribotyping (Kato et al., 2001) but also by slpA sequence typing. The region of slpA examined was found to be conserved among clinical smz isolates, including the type smz strain which was found at the New York hospital but not involved in the epidemic there. Since strains typed as PCR ribotype smz are non-typable by PFGE because of a DNA degradation problem (Kato et al., 2001), the results of slpA sequence typing cannot be compared with those by PFGE. While three slpA sequence types were found among type smz strains, no predominant subtype was detected in this study.

slpA sequence typing was applied to direct typing on DNA extracted from stool specimens obtained from patients at a hospital. A total of 22 stool specimens were examined for toxins A and B, C. difficile, and direct PCR detection of tcdA. Of these specimens, 17 were PCR-positive for tcdA (results of other tests on these 17 are summarized in Table 2). The remaining five specimens, which were PCR-negative for tcdA, were also negative for the detection of fecal toxins or C. difficile culture.

Table 2. Detection of toxins A and B, culture for C. difficile, and typing results from direct PCR of 17 stool specimens where direct PCRs were positive for detection of tcdA SlpA S type, slpA sequence type; A+B+, toxin A-positive, toxin B-positive; ND, not done.

Nested PCR amplification of the slpA gene was performed on the 17 stool specimens that were directly PCR-positive for tcdA. The second PCR for slpA generated a PCR product approximately 550 bp in size in all 17, and the resultant PCR product was sequenced. Two slpA sequence types, smz-1 and smz-2, were identified among the 17 specimens from the hospital patients, suggesting a nosocomial spread of two subtypes at that hospital. C. difficile was recovered from 12 of these 17 specimens, and all 12 isolates were typed as type smz by PCR ribotyping and slpA sequence typing, the results of which completely agreed with those of direct sequence typing in all 12 specimens. The nested PCR led to success in the slpA sequence typing of C. difficile in five stool specimens with negative results on culturing. It was suggested that both of the nested PCRs detecting tcdA and slpA directly from stool specimens used in this study may be more sensitive than culturing C. difficile, although the number of stool specimens tested here was limited.

The typing results of banding-pattern analyses, such as those by AP-PCR, PCR ribotyping, REA or PFGE, may be difficult to compare in many isolates and to share among different laboratories. Typing by sequencing has the advantage that the typing results can be compared among numerous laboratories via the Internet without the need for exchanging reference strains or antisera. Recently, multilocus sequence typing (MLST) has been used for a long-term epidemiological study of a variety of bacteria, and its application to C. difficile has been reported (Lemee et al., 2004). Like MLST analysis, the use of slpA sequence typing may be feasible for a global epidemiological study by sharing typing data from many clinical isolates identified at a number of laboratories. Moreover, this method can be applied to direct typing from clinical materials. In the present study, five culture-negative specimens were PCR-positive for slpA followed by slpA sequence typing. Such typing is also useful for analysing a specimen containing small numbers of C. difficile or one that is inappropriate for culturing, such as one stored at ambient temperature for prolonged periods before laboratory testing. While this method needs to be further evaluated for its typing ability on a large number of clinical isolates and stool specimens and for its specificity to the slpA gene of C. difficile, slpA sequence typing appears to be a reproducible, sensitive and reliable typing method for C. difficile.

For their valuable help in the collection of strains and specimens, the authors would like to thank T. Yamamoto and K. Suzuki (Nagoyashi Koseiin Geriatric Hospital), S. Ishigo (Ogaki Municipal Hospital), M. Kunihiro and I. Nakamura (Yamaguchi Prefectural Hospital), T. Oguri and S. Misawa (Juntendo University Hospital), K. Yoshimoto (Shirasagi Hospital), M. Komatsu and M. Aihara (Tenri Hospital), H. Kato (Toyokawa City Hospital), H. Sato and C. Sakai (Chiba Cancer Center Hospital), and G. E. Killgore (Centers for Disease Control and Prevention). The technical assistance of Y. Yoshimura is also gratefully acknowledged. This work was supported by a grant-in-aid for Scientific Research (C) from the Ministry of Education, Culture, Sports, Science and Technology, Japan.