Comparison of PCR-Ribotyping, Arbitrarily Primed PCR, and Pulsed-Field Gel Electrophoresis for Typing Clostridium difficile

Abstract

Clostridium difficile is now recognized as the major agent responsible for nosocomial diarrhea in adults. Among the genotyping methods available, arbitrarily primed PCR (AP-PCR), PCR-ribotyping, and pulsed-field gel electrophoresis (PFGE) have been widely used for investigating outbreaks of C. difficile infections. However, the comparative typing ability, reproducibility, discriminatory power, and efficiency of these methods have not been fully investigated. We compared the results of three methods—AP-PCR with three different primers (AP3, AP4, and AP5), PCR-ribotyping, and PFGE (with SmaI endonuclease)—to differentiate 99 strains of C. difficile that had been previously serogrouped. Typing abilities were 100% for PCR-ribotyping and AP-PCR with AP3 and 90% for PFGE, due to early DNA degradation in strains from serogroup G. Reproducibilities were 100% for PCR-ribotyping and PFGE but only 88% for AP-PCR with AP3, 67% for AP-PCR with AP4, and 33% for AP-PCR with AP5. Discriminatory power for unrelated strains was >0.95 for all the methods but was lower for PCR-ribotyping among serogroups D and C. PCR-based methods were easier and quicker to perform, but their fingerprints were more difficult to interpret than those of PFGE. We conclude that PCR-ribotyping offers the best combination of advantages as an initial typing tool for C. difficile.

Clostridium difficile is the etiologic agent of most cases of pseudomembranous colitis and of antibiotic-associated diarrhea (3). This microorganism is considered the major cause of nosocomially acquired diarrhea among adults (1). Different methods have been developed to analyze genetic relatedness among C. difficile isolates in nosocomial outbreaks. These methods include phenotypic marker methods, such as lysotyping (20), serogrouping (8), sodium dodecyl sulfate-polyacrylamide gel electrophoresis (21), and immunoblotting (13), and genotypic marker methods, such as plasmid analysis (18), restriction endonuclease analysis (17), ribotyping (5), pulsed-field gel electrophoresis (PFGE) (6, 15, 16), arbitrarily primed PCR (AP-PCR) (2) and, more recently, PCR-ribotyping (4, 12, 19). Phenotypic methods either require specific reagents (serotyping, phage typing) or lack standardization (sodium dodecyl sulfate-polyacrylamide gel electrophoresis). Genotype methods, such as AP-PCR, PCR-ribotyping, and PFGE, may overcome these problems and have been widely used to investigate epidemics during the past 10 years. AP-PCR is based on nonspecific random amplifications by PCR of the bacterial chromosome using a short primer under low-stringency conditions. PCR-ribotyping uses specific primers complementary to the 3′ end of the 16S rRNA gene and to the 5′ end of the 23S rRNA gene to amplify the variable-length intergenic spacer region. PFGE is based on the digestion of chromosomal DNA with a restriction endonuclease that cleaves infrequently and produces only a few high-molecular-weight fragments that can be separated under special conditions of electrophoresis. Despite several reports (7, 23), the typing ability, reproducibility, discriminatory power, and efficiency of these methods have not been fully evaluated and compared. We performed a comparison of AP-PCR with three different primers (AP3, AP4, and AP5), PCR-ribotyping, and PFGE (with SmaI endonuclease) using 99 strains (20 reference strains, 60 unrelated strains, and 19 related strains) of C. difficile that had been previously serogrouped.

(This work was presented during the 99th General Meeting of the American Society for Microbiology, May to June 1999, Chicago, Ill. (Poster L/U-10, Session 115.)

MATERIAL AND METHODS

Bacterial strains, serogroups, and culture conditions.

Ninety-nine strains of C. difficile were used for this study. They included 20 reference strains kindly supplied by M. Delmée (University of Louvain, Brussels, Belgium) and corresponding to the 20 different serogroups (9) (A1 [ATCC 43594], A2, A3, A4, A5, A6, A7, A8, A9, A10, A11, B [ATCC 43593], C [ATCC 43596], D [ATCC 43597], F [ATCC 43598], G [ATCC 43599], H [ATCC 43600], I [ATCC 43601], K [ATCC 43602], and X [ATCC 43603]), 60 epidemiologically unrelated strains isolated from patients between 1992 and 1997, 11 epidemic strains isolated from patients hospitalized in an orthopedic department in April 1994 (11), and 4 pairs of epidemiologically related strains. Strains were cultured on TCCA plates (brain heart infusion agar supplemented with 5% defibrinated horse blood, 0.1% taurocholate, cefoxitin at 10 μg ml⁻¹, and cycloserine at 250 μg ml⁻¹) and incubated in an anaerobic atmosphere. Colonies were identified by use of an enzymatic profile from a RapID 32 A gallery (bioMérieux, La Balme les Grottes, France). Serogrouping was performed by use of an enzyme-linked immunosorbent assay according to the method described by Delmée et al. (9) with antisera A1, A5, A8, A9, A10, C, D, F, G, H, and K. The distribution of sporadic and related strains among the different serogroups is shown in Table 1.

TABLE 1.

Distribution of sporadic and related strains among serogroups

Strains (n)	No. of strains of serogroup:
	A1	A10	A5	A8	A9	C	D	F	G	H	K	NTa
Sporadic (60)	6	1	0	1	0	5	7	0	9	9	5	17
Related (19)						11	2		2	2		2

^aNT, nontypeable with the 11 antisera available. 

AP-PCR.

C. difficile DNA was extracted by the phenol-chloroform method according to the procedure described by Barbut et al. (2). Three different 10-mer primers were tested: AP3 (5′-TCA CGA TGC A-3′), AP4 (5′-TCA CGC TGC A-3′), and AP5 (5′-TCA CGC TGC G-3′). Amplification reactions were performed with a 100-μl volume containing 10 mM Tris-HCl (pH 8.8), 50 mM KCl, 4 mM MgCl₂, 200 μM each deoxynucleoside triphosphate (Pharmacia Biotech, Orsay, France), 25 pmol of primer, 2.5 U of Taq polymerase (Pharmacia Biotech), and 50 ng of DNA extract. Amplifications were carried out in a thermal cycler (Hybaid; Omnigen) for 1 cycle of 4 min at 94°C for denaturation; 45 cycles of 1 min at 94°C, 1 min at 36°C, and 2 min at 72°C; and a final extension cycle of 8 min at 72°C. For each primer, all the strains were assayed in two series of 50 strains with the same reaction mixture. Amplification products were fractionated by electrophoresis through 1.5% standard agarose (Eurobio, Les Ulis, France) for 2.5 h at 100 V in Tris-borate-EDTA buffer (Eurobio) and were analyzed on a UV table after ethidium bromide staining.

PCR-ribotyping.

DNA was extracted from three large C. difficile colonies by use of a Chelex resin-based commercial kit (InstaGene Matrix; Bio-Rad, Ivry, France) as recommended by the manufacturer. The primer sequences were 5′-GTG CGG CTG GAT CAC CTC CT-3′ (16S primer) and 5′-CCC TGC ACC CTT AAT AAC TTG ACC-3′ (23S primer) and corresponded to bases 1482 to 1501 of the 16S rRNA gene of C. difficile and bases 1 to 24 of the 23S rRNA gene of C. difficile, respectively (4). Amplification reactions were performed with a 100-μl volume containing 10 mM Tris-HCl (pH 8.8), 50 mM KCl, 1.5 mM MgCl₂, 200 μM each deoxynucleoside triphosphate (Pharmacia), 50 pmol of each primer, 2.5 U of Taq polymerase (Pharmacia), and 10 μl of DNA extract. Amplifications were carried out in a thermal cycler (Perkin-Elmer Cetus model 480) for 1 cycle of 6 min at 94°C for denaturation; 35 cycles of 1 min at 94°C, 1 min at 57°C, and 1 min at 72°C; and a final extension cycle of 7 min at 72°C. Amplification products were fractionated by electrophoresis through 3% Resophor agarose (Eurobio) for 6 h at 85 V in Tris-borate-EDTA buffer with a distance of 24 cm between electrodes (3.5 V/cm) and were analyzed on a UV table after ethidium bromide staining.

PFGE analysis.

Strains were cultured in prereduced Schaedler broth for 17 h at 37°C and pelleted by centrifugation at 5,000 × g. Bacterial cells were embedded in agarose plugs and lysed for 18 h at 37°C with lysozyme-lysostaphin and overnight at 50°C with proteinase K by use of a GenePath Group 1 kit (Bio-Rad) according to the manufacturer's recommendations. Plugs were digested in a volume of 300 μl with 25 U of SmaI endonuclease overnight at 25°C. Electrophoresis was performed with a contour-clamped homogeneous electric field device (CHEF DRII; Bio-Rad) and a Gel kit (Bio-Rad) with program 14 (50 to 500 V, 19 h 7 min). Gels were stained with ethidium bromide and analyzed on a UV table.

Analysis of banding patterns.

Gel images were analyzed with Image Master software (Bio-Rad). Each gel migration was run with a reference strain (ATCC 43594 for PCR-ribotyping, ATCC 43599 for AP-PCR, and serogroup A2 for PFGE) in order to normalize different migrations. To facilitate comparison of patterns, computer-generated images of gels were obtained with Molecular Analyst software (Bio-Rad). For PFGE and PCR-ribotyping, strains presenting patterns differing by at least one band were assigned to separate groups. Patterns in AP-PCR differing by at least one band of high intensity or two faint bands were considered different.

Comparison of typing methods.

Typing ability was defined as the ratio between the number of strains producing a banding pattern and the number of strains tested. Two types of reproducibility were studied for PCR-based methods: reproducibility of repeated PCR assays with the same DNA extract (ATCC 43594 for PCR-ribotyping and ATCC 43599 for AP-PCR) and reproducibility of the patterns obtained from nine subcultures of strains ATCC 43593 and ATCC 43598 in the same PCR run (stability). Reproducibility of PFGE was studied with six subcultures of the serogroup A2 reference strain. Discriminatory power was calculated from the epidemiologically unrelated strains with the Hunter-Gaston index (14).

RESULTS

AP-PCR.

Patterns obtained with AP-PCR exhibited 1 to approximately 10 bands with various degrees of intensity (Fig. 1). Patterns obtained with primer AP3 displayed no common band; however, those obtained with primer AP4 shared two bands of about 300 and 550 bp, and those obtained with primer AP5 shared one band of about 550 bp, with two other bands often present at about 1.3 kbp and >2.5 kbp. All strains were typeable with primer AP3, whereas two strains were not amplified with primer AP4 and four were not amplified with primer AP5 (Table 2). For the 20 serogroup reference strains, AP-PCR with primer AP3 produced 14 different patterns: one corresponded to serogroups A1, A8, A10, C, and H, one corresponded to serogroups A3 and I, and one corresponded to serogroups B and F. AP-PCR with primer AP4 produced 16 different patterns with the reference strains: one corresponded to serogroups A3 and A8, one corresponded to serogroups B and F, and one corresponded to serogroups A2 and A9. Serogroup A4 was not typeable with primer AP4. AP-PCR with primer AP5 produced nine different patterns with the reference strains: one corresponded to serogroups A5 and F, and one corresponded to serogroups A1, A2, A4, A6, A8, A9, A10, A11, G, and H. Serogroup X was not typeable with primer AP5.

FIG. 1.

Patterns obtained by AP-PCR with primer AP3 for nine reference strains belonging to different serogroups (indicated above the lanes). Lane MW, molecular weight (100-bp ladder).

TABLE 2.

Results of AP-PCR, PCR-ribotyping, and PFGE assays for the typing of 99 C. difficile strains

Method	Enzyme or primer	% of strains typed	No. of:			Band size range (kbp)	% Reproducibility	Discriminatory power
			Types	Bands	Common bandsa
AP-PCR	AP3	100	40	1–>10	0	0.4–>2.5	88	0.96
	AP4	98	44	1–>10	2	0.25–>2.5	67	0.95
	AP5	96		1–>10	1	0.25–>2.5	33
	AP3 and AP4b	98	70					0.99
PCR-ribotyping	16S and 23S primers	100	41	7–15	0	0.22–0.7	100	0.98
PFGE	SmaI	90	58	5–10	0	48–>500	100	0.99

^aNumber of bands common to all the patterns. 

^bCombined results of AP-PCR with primer AP3 and AP-PCR with primer AP4. 

The reproducibility of AP-PCR patterns with the same extract as well as with different extracts of subcultures was poor, especially with primer AP5. Based on the results of the nine subcultures, reproducibilities were 88, 67, and 33% for primers AP3, AP4, and AP5, respectively. Forty patterns were discriminated with AP3, and 44 patterns were discriminated with AP4. Patterns obtained with AP5 were not analyzed. Discriminatory powers with AP3, AP4, and a combination of both primers were 0.96, 0.95, and 0.99, respectively.

PCR-ribotyping.

All the strains were typeable by PCR-ribotyping (Table 2). Among the 99 strains in the study, 41 PCR-ribotypes were discriminated. Patterns exhibited from 7 to 15 bands ranging from 220 to 700 bp, with no band common to all the profiles (Fig. 2). PCR-ribotyping produced 19 different patterns with the 20 reference strains, serogroup H and serogroup A8 sharing the same pattern. The reproducibility of the patterns was 100% with the same DNA extract or with different DNA extracts of the nine subcultures. The discriminatory power of the method was 0.96. All strains of serogroup D displayed the same pattern, and all strains of serogroup C but one belonged to the same PCR-ribotype.

FIG. 2.

Patterns obtained by PCR-ribotyping. Lanes 1, 4, 7, 8, and 9, epidemic strains of serogroup C displaying the same PCR-ribotype; lanes 2, 3, 5, 6, and 10, sporadic strains; lane MW, molecular weight (100-bp ladder).

PFGE.

Nine of the 99 strains of the study, including the reference strain ATCC 43599, were nontypeable by PFGE due to DNA degradation during the extraction step (90% were typeable) (Table 2). All those strains belonged to serogroup G and to three close PCR-ribotypes. Three other strains of serogroup G were typeable, producing three different pulsotypes. Except for the serogroup G strain, each reference strain exhibited a different pattern. The typeable strains of the study displayed 58 different pulsotypes. Patterns displayed 5 to 10 bands ranging from 48 kbp to more than 500 kbp, without any band common to all the profiles (Fig. 3). All the subcultures of the serogroup A2 reference strain produced the same pattern (reproducibility, 100%).

FIG. 3.

Patterns obtained by PFGE. Lanes 1 to 13, sporadic strains; lanes 4, 5, 8, and 12, nontypeable strains of serogroup G; lanes A2, serogroup A2 reference strain.

Concordance among methods.

Good concordance was observed among PFGE, PCR-ribotyping, and serogroups. Except for six strains, all the strains of a given pulsotype belonged to the same PCR-ribotype and, except for two strains, all the strains of a given PCR-ribotype belonged to the same serogroup. In contrast, correlations between the results of those methods and of AP-PCR were not as strong.

Classification of epidemiologically related strains.

All 11 epidemic strains from the orthopedic department belonged to serogroup C and exhibited the same PCR-ribotype. All but one of those strains belonged to the same pulsotype. With AP-PCR, the 11 strains displayed four different patterns with primer AP3 and two different patterns with primer AP4, forming five combinations of patterns with both primers. The four pairs of related strains were correctly grouped by all the methods (one couple of serogroup G being nontypeable by PFGE).

DISCUSSION

The identification of C. difficile as a major nosocomial enteropathogen has led to the application of numerous methods for typing of this bacterium. The development of molecular biology in most microbiology laboratories has focused interest on genotyping methods, such as PFGE and the PCR-based methods AP-PCR and PCR-ribotyping. The aim of the current investigation was to compare different characteristics of these methods for a large sample of C. difficile strains. In view of our results, PCR-ribotyping seems to offer the best combination of advantages as an initial typing tool for C. difficile.

PCR-based methods were quite easy to perform. For PCR-ribotyping, we used a simplified Chelex-based DNA extraction method which gave results identical to those given by phenol-chloroform extraction, as shown in a previous study (4). However, in order to improve standardization, phenol-chloroform extraction was used for AP-PCR. PFGE was more time-consuming, and culture times had to be strictly respected because of the sensitivity of C. difficile strains to DNA degradation. Results with PCR-based methods can be obtained in 2 days, while at least 4 days are needed with PFGE. Interpretation of the patterns was very easy for PFGE with SmaI. Analysis of PCR-ribotyping patterns was hampered by the closeness of the molecular weights of the fragments, and that of AP-PCR was hampered by the lack of reproducibility of faint bands.

Indeed, the major problem of AP-PCR was the lack of reproducibility, as previously reported by several authors (6, 7, 10). This was particularly true with primer AP5, so that any classification of patterns obtained with this primer seemed totally illusory, and the results were discarded. For PFGE, the problem encountered was that several strains, all from serogroup G, remained nontypeable due to early degradation of their genomic DNA, as previously reported (7, 15, 23).

The three methods demonstrated good discriminatory powers, PFGE being the most discriminatory as far as strains that were typeable. The discriminatory power of PCR-ribotyping was slightly higher than that of AP-PCR with various primers and lower than that of AP-PCR with the combination of primers AP3 and AP4. However, the lack of reproducibility of AP-PCR may have led to an overestimation of the discriminatory power of this method. Collier et al., studying 49 strains of various origins, concluded that PCR-ribotyping was slightly more discriminatory than AP-PCR and that a higher concordance was observed between PFGE and PCR-ribotyping than between PFGE and AP-PCR (7). However, discriminatory powers were not calculated, serogroups were not determined, except for the reference strains, and the reproducibilities of the different methods were not studied.

For some serogroups, PCR-ribotyping appeared to be poorly discriminatory, especially for serogroup D (only one pattern) and serogroup C (two patterns). For serogroup C strains, which are often responsible for epidemic bursts, the lack of discrimination of PCR-ribotyping could raise difficulties in differentiating epidemic from sporadic strains. Van Dijck et al. (23), studying outbreaks of strains belonging to serogroup C, found, as we did, only two patterns with PCR-ribotyping, while PFGE and AP-PCR were able to differentiate some groups of strains. Serogroup D strains are likely genetically closely related because patterns obtained by PFGE and PCR-ribotyping were similar and all the strains have been shown to be nontoxinogenic.

Classification of epidemic strains is a key criterion for evaluating a typing method: the 11 epidemic strains from the orthopedic department belonged to serogroup C and were all placed in one group by PCR-ribotyping and PFGE (one strain whose pattern differed by only one band may be related to the others [22]). In contrast, AP-PCR produced at least two patterns per primer without clear correspondence between them. This surprising result in a well-documented nosocomial outbreak leads us to question whether AP-PCR methods are not too discriminatory (or not reproducible enough) to disclose outbreaks among small samples of strains.

In conclusion, it appears that PFGE, although displaying the highest discriminatory power, is hampered by the inability to type most strains belonging to serogroup G and is far more time-consuming than PCR-based methods, while AP-PCR raises many problems of interpretation due to its lack of reproducibility. For the purpose of investigating outbreaks of C. difficile, PCR-ribotyping appears to be the best method to use initially, being reproducible, relatively quick and easy to perform, and sufficiently discriminatory. However, for continuous epidemiological surveys, where a more precise characterization of isolates is expected and the delay in results is less important, PFGE has the advantage of very high discriminatory power together with perfect reproducibility. Being a well-standardized method, it allows interlaboratory comparison of pulsotypes. PCR-based methods offer an alternative for nontypeable strains of serogroup G.