INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Clostridium perfringens is a common cause of bacterial food-borne illness (2). It was identified as the etiologic agent in 15 to 20% of bacterium-related food-borne illness from 1979 to 1982 but caused only ~3% of outbreaks from 1985 to 1987 (3). The decrease in the incidence of food-borne illness caused by C. perfringens-related illness may be a result of a shift in priorities for state health departments rather than a true decrease in the incidence of diarrheal illness due to C. perfringens (3). Currently, C. perfringens is identified in only approximately 2% of all outbreaks of bacterium-related diarrheal disease (2). Cases likely are underreported because many people do not seek medical treatment due to the mildness of symptoms and quick recovery (25, 27). Even when medical treatment is sought, the etiologic agent is not identified and does not attract the attention of local state health departments unless a large number of cases occur in a single outbreak.

C. perfringens is ubiquitous in nature; it forms heat-resistant endospores, multiplies rapidly, and produces a cytotoxic enterotoxin (C. perfringens enterotoxin [CPE]) (19). These factors contribute to the potential of C. perfringens to cause food-borne disease. Gastrointestinal illness is most likely due to the presence of CPE produced in the intestine during sporulation of ingested vegetative cells. Identified outbreaks of C. perfringens are usually due to improperly cooked and/or mishandled meat and meat products and are most often associated with restaurants, institutions, or catered events (2, 11). Outbreaks in the United States reported since 1989 involved over 100 cases per outbreak; one outbreak in 1990 had 700 cases (2). The ability of C. perfringens to cause hundreds of cases of food-borne disease at a single event signifies the need to investigate and identify outbreaks due to C. perfringens in order to implement corrective actions to prevent recurrence.

Recommended laboratory criteria used in association with clinical presentation and epidemiologic evidence to implicate C. perfringens in food-borne disease are (i) >10⁵ vegetative cells per g of food, (ii) >10⁶ spores per g of stool, (iii) identification of the same serotype of C. perfringens in stools from different patients, and (iv) identification of the same serotype in both food and stool isolates (14). Identification of C. perfringens as the etiologic agent of a food-borne outbreak is complicated by its presence in stools (10³ to 10⁴ cells per g) from healthy individuals and nonimplicated foods (10¹ to 10⁴ cells per g). In addition, only 2 to 6% of C. perfringens isolates contain the CPE-producing gene (cpe), and so simple cell or spore counts may overestimate the involvement of C. perfringens in food-borne disease (20). The presence of CPE in stools, when detected in at least some of the patients in an outbreak, is a reliable indicator for implicating C. perfringens as the etiologic agent in food-borne disease (4). CPE is produced in the intestine during sporulation of vegetative cells and so is undetectable in suspect foods. Because CPE cannot be detected directly in foods and it is difficult to make some strains of C. perfringens sporulate in the laboratory, it is difficult to link patient isolates with food sources by toxin detection. Serotyping has been used successfully in the United Kingdom in the investigations of gastrointestinal illness due to C. perfringens (5, 29). This method has been less successful in the United States. Even when used routinely, the reagents available in the United States identified the serotype in only 40% of the isolates (13).

Pulsed-field gel electrophoresis (PFGE) has been used successfully to subtype over 50 different bacterial species (30). Furthermore, PFGE has been shown to be a valuable technique in outbreak investigations of a variety of food-borne diseases (1, 17). Carnard and Cole (6) demonstrated the feasibility of using PFGE to characterize the genome of C. perfringens. Recently, the potential of PFGE in evaluating clonal relationships between C. perfringens isolates was demonstrated (7, 23).

In this study, we used 62 C. perfringens isolates from food-borne-outbreak investigations and culture collections to develop a PFGE method that discriminated among isolates from unrelated investigations but demonstrated a clonal relationship among isolates that were epidemiologically related. This study also provided a direct comparison between serotyping, the only currently recommended typing method for C. perfringens (14), and molecular subtyping with PFGE. The PFGE protocol incorporated a computer-based, DNA fragment band analysis program to more objectively evaluate similarities between isolates. This protocol will serve as the basis for establishing a standardized protocol to be included in The National Molecular Subtyping Network (PulseNet), which currently contains standardized protocols for subtyping Escherichia coli 0157:H7, Listeria monocytogenes, and several common Salmonella serotypes (28).

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Bacterial strains. C. perfringens strains previously isolated during outbreak investigations were obtained from the Centers for Disease Control and Prevention culture collection (Table 1). Isolates from several individuals (or in some cases, multiple isolates from one individual) were available from each of three investigations (California, 1983 [CA 1983]; Hawaii, 1995-A [HI 1995-A]; and New Jersey, 1997 [NJ 1997]). At least one isolate was available from only one individual from each of four other investigations (Georgia, 1997 [GA 1997]; Hawaii, 1995-B [HI 1995-B]; Hawaii, 1997 [HI 1997]; and Illinois, 1994 [IL 1994]). Control strains (ATCC 43402, NCTC 8798, FDA 57, FDA 1041, FDA 1, NFRI 2498, and NFRI 1090) from the Amherst Chenoweth Laboratory collection were kindly provided by R. Labbe, University of Massachusetts. All strains were cultured by established procedures (8). The identities of the C. perfringens strains were confirmed by anaerobic growth, Gram stain, lecithinase activity, milk hydrolysis, indole reaction (negative), gelatinase activity, motility (nonmotile), and nitrate reduction. Serotype was determined as described previously with a panel of 93 antisera (8).

PFGE. PFGE was performed by the method of Barrett et al. (1) with modifications. Isolates of C. perfringens were grown anaerobically on egg yolk agar overnight at 37°C (8). Cells were suspended in 75 mM NaCl-25 mM EDTA (pH 8.0) (SE) buffer to an approximate absorbance of 1.3 at 610 nm. Bacterial cells were embedded in 1.2% chromosomal-grade agarose (SeaKem Gold; FMC BioProducts, Rockland, Maine) by mixing equal volumes (0.5 ml) of the cell suspension and melted agarose equilibrated to 55 to 65°C. In some preparations, bacterial cells were pretreated with formalin to reduce the interference by endogenous DNase activity before they were mixed with agarose (10). Plugs were solidified at 4°C in 1.5-mm-thick molds (Bio-Rad Laboratories, Hercules, Calif.). The agarose-embedded cells were lysed by incubation of the plugs overnight at 55°C with gentle shaking in lysis buffer (50 mM Tris-HCl [pH 8.0], 50 mM EDTA [pH 8.0], 1% N-laurylsarcosine, and proteinase K [1 mg per ml]). The plugs were washed at 50°C with vigorous shaking three times for 15 min each in sterile reagent-grade water and then three times for 15 min each in 10 mM Tris-1 mM EDTA, pH 8.0 (TE). Plugs were stored at 4°C in TE buffer. Plugs were cut into 3- to 4-mm slices and equilibrated with appropriate restriction buffer containing 20 mg of bovine serum albumin (Boehringer Mannheim Corp., Indianapolis, Ind.) per ml. Six restriction enzymes and enzyme-appropriate conditions recommended by the manufacturer were evaluated for their ability to distinguish between unrelated strains following incubation for 4 h: SmaI, ApaI, FspI, MluI, KspI, and XbaI (Boehringer Mannheim). Restriction fragments were separated by electrophoresis through a 1% agarose gel composed of SeaKem Gold in a 0.5× solution of Tris-borate and EDTA (10× TBE buffer [Life Technologies, Inc., Grand Island, N.Y.]) at 14°C in a contour-clamped homogeneous electric field MAPPER XA PFGE apparatus (Bio-Rad). The run time was 20 h with a voltage of 6 V per cm and a linearly ramped pulse time of 0.5 to 40 s.

Analysis. A C. perfringens strain (CPerf1), restricted with SmaI, was calibrated with both a lambda ladder (Bio-Rad) and yeast chromosome DNA (Saccharomyces cerevisiae; Bio-Rad) and used as the reference standard. The size of the largest DNA fragment of C. perfringens was estimated by altering the run conditions to 28 h and a linearly ramped pulse time of 47 s to 3 min. After electrophoresis, the gel was stained with GelStar (FMC Bioproducts, Inc.) and imaged with the Gel Doc 1000 system (Bio-Rad). Analysis of the banding patterns was done with the Molecular Analyst Fingerprinting Plus software (Bio-Rad) with a 1% tolerance for fragment shifts. Similarities of band patterns between strains were determined by the software with the Dice coefficient. In general, strains were considered clonal if they showed 100% similarity. Band patterns were clustered to deduce a dendrogram by the unweighted pair group method with arithmetic averages.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Restriction enzymes and PFGE. Six restriction enzymes (SmaI, ApaI, FspI, MluI, KspI, and XbaI) were tested for their usefulness in subtyping outbreak-related human isolates of C. perfringens. SmaI consistently produced 11 to 13 well-resolved fragments. XbaI did not restrict the isolates tested, and KspI produced too few fragments to differentiate between unrelated isolates (data not shown). ApaI, FspI, and MluI produced too many fragments, which inhibited comparison of PFGE patterns (data not shown). On the basis of the above results, the enzyme SmaI was chosen for further studies.

View larger version (33K): [in this window] [in a new window]   FIG. 1.   Dendrogram of PFGE subtype patterns of all C. perfringens control strains and outbreak isolates. Analysis range was 40 to 1,400 kb. CPerf1 was used as the reference standard.

View larger version (99K): [in this window] [in a new window]   FIG. 2.   PFGE of C. perfringens strains associated with a single outbreak (NJ 1997). Lanes 1 and 8, reference standard Cperf1; lanes 2 through 7, one isolate from each individual in the outbreak (subtype pattern L).

View larger version (101K): [in this window] [in a new window]   FIG. 3.   PFGE of C. perfringens associated with a single outbreak with possible multiple infecting strains (CA 1983). Lanes 1 and 10, reference standard Cperf1; lanes 2 through 4 and 7 through 9, individuals with subtype pattern A; lane 5, individual with subtype pattern G; lane 6, individual with subtype pattern N.

Serotyping. Of the outbreak isolates, only the CA 1983 isolates had type-specific agglutination with the available serotyping reagents. All the isolates of PFGE pattern type A were serotype PS35. One isolate from PFGE pattern type G was serotype PS34; one autoagglutinated. The CA 1983 isolates with PFGE pattern type K were serotype PS21. Although the number of isolates that were typeable with the serotyping reagents was limited, the serotype corresponded to the PFGE pattern types. One CA 1983 isolate that gave no PFGE pattern was serotype PS70; the other CA 1983 isolates that gave no PFGE pattern were not serotypeable as well. Of the 62 isolates tested, 63% (n = 39) did not agglutinate with any of the available typing sera.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Recently, the potential of PFGE to establish clonal relationships between C. perfringens strains isolated in association with disease was shown (7, 23). We further demonstrated the utility of this method in evaluating clonal relationships in the outbreak setting with seven control strains and 55 isolates from seven food-borne outbreaks. In three outbreaks, isolates from different individuals gave identical patterns. In contrast, four unique patterns were obtained with 18 isolates from one outbreak (CA 1983). This outbreak may have been caused by more than one strain. The isolates that had PFGE pattern type A were undoubtedly related to the outbreak since the identical pattern was obtained with isolates from 13 different individuals. The presence of a low-intensity small fragment was variable in one isolate, giving a lower (90%) similarity between this and the other isolates with pattern type A. All subtype pattern A isolates were serotype PS35. Two control strains (FDA 1041 and NCTC 8798) also had patterns with 90% similarity, but no available information on these isolates suggested that they were related. In addition, the NCTC 8798 isolate consistently lacked a band at ~160 kb and also had an observable band shift at ~375 kb compared to the FDA 1041 isolate. Control strain NCTC 8798 was serotype Hobbs 9 and FDA 1041 was nontypeable, also suggesting that the isolates were unrelated. The CA 1983 isolates that had PFGE pattern types G and K, although different from those with pattern type A (<60% similarity), were also likely related to the outbreak since identical patterns were obtained from two different individuals for each subtype. It is unknown whether the isolate with PFGE pattern type N or the isolates that gave no PFGE patterns were related to the outbreak. None of the isolates were tested for the ability to produce CPE, so it is unknown whether these five isolates were capable of producing symptoms of food-borne disease.

The isolates from HI 1995-B demonstrated the necessity of typing multiple isolates from a single individual. In this outbreak, three of four isolates from a single individual had the same PFGE pattern and most likely represented the outbreak strain. The remaining isolate may have been part of the normal intestinal flora or may represent a second outbreak strain. Selecting multiple isolates from each individual or food item for PFGE subtyping improves the chances of identifying the C. perfringens strain that caused disease. The use of a CPE gene (cpe) detection assay (7, 15, 23) in conjunction with PFGE would most likely answer the question of whether isolates with different PFGE patterns within the same outbreak are related to the food-borne illness caused by C. perfringens.

Traditionally, the laboratory criteria described by Hauschild (14) have been used in conjunction with epidemiologic data to investigate food-borne diseases thought to be caused by C. perfringens. The laboratory criteria depend on the enumeration of C. perfringens in stools and/or food sources and subsequent determination of serotype of isolates. However, there are several problems associated with this approach (16). It is difficult to isolate vegetative organisms from implicated foods since the viability of the organism declines rapidly in leftover food stored at low temperatures. In addition, some populations, such as institutionalized well patients, have fecal spore counts of >10⁶ per g of stool.

Serotyping is the only laboratory criterion (14) currently recommended to assist in linkage of isolates (patient and/or food) within an outbreak. This method has been used successfully in the United Kingdom. Almost 80% of C. perfringens strains isolated from 1970 to 1978 were typeable with 75 antisera. Sixty-six percent of 524 outbreaks were confirmed serologically (29). Continuous efforts of the Central Public Health Laboratory in London to prepare antisera to unknown strains resulted in the identification of 193 different serotypes in 1,525 outbreaks (5). Although serotyping of C. perfringens on occasion has been useful in epidemiologic investigations in the United States (12, 31), only 35% of outbreaks from 1967 to 1978 were confirmed by serotype results (13). Similarly, only 35% of the strains used in our study agglutinated with the available typing sera. Interestingly, these serotypeable strains were isolated before 1984. None of the strains in our study isolated from outbreak investigations in the 1990s were serotypeable. In contrast to the United Kingdom, efforts to prepare additional serotyping reagents in the United States ceased in the early 1980s, and serotyping for confirmation of food-borne outbreaks due to C. perfringens was essentially abandoned in 1983 (25). Clearly, new laboratory criteria that can reliably discriminate between pathogenic and nonpathogenic C. perfringens isolates and that can establish a relationship among exposed individuals or between ill patients and potential food sources are needed.

Several methods have been evaluated for their ability to discriminate between C. perfringens strains. Chemical methods such as gas-liquid chromatography (21) and more recently nuclear magnetic resonance spectroscopy (26) of the capsular polysaccharides, while highly discriminatory, are impractical for routine use in the clinical laboratory. Strains have been distinguished by bacteriocin typing (18), but as in serotyping, the discriminatory power is limited to the available panel of bacteriocins. Plasmid profiling (9, 18), multilocus enzyme electrophoresis (22), ribotyping (24), and PFGE (7, 23) have all been used to discriminate between isolates of C. perfringens. Plasmid profiling was of limited use in one study (18), since the predominant outbreak strain lacked a plasmid, and plasmids were absent in strains from 33% of outbreaks in another study (9). PFGE may be more useful and reproducible than plasmid profiling and more discriminatory than ribotyping or multilocus enzyme typing (30).

In our study, PFGE clearly discriminated among unrelated isolates of C. perfringens and established a clonal relationship between related strains. Although unrelated isolates showed similarities of only 40 to 70%, there were numerous fragments that comigrated between isolates, similar to those reported by Carnard and Cole (6), suggesting that a number of sequences in the genome are conserved. However, based on earlier studies (6, 23) it was not surprising to observe different patterns between unrelated isolates. The use of the computer analysis program to measure the similarity between isolates provided a more objective assessment of clonality. Visual inspection of subtype patterns A and B might lead one to suspect they were related (one band difference) when in fact the isolates were obtained in different parts of the United States and differed by a time factor of 12 years. Computer analysis showed the isolates with subtype patterns A and B to have 80% similarity. A similarity value of 100% may provide laboratory confirmation of epidemiologic links during outbreak investigations. Similarity values of <100% suggest that isolates are epidemiologically unrelated.

Our study showed PFGE to be superior to serotyping, which is the currently recommended method for providing laboratory links to epidemiologic data. In addition, since the identification of type is not dependent on the availability of typing reagents, a link between patient and/or food isolates can be made relatively quickly during epidemiologic investigations. Although 8% of the isolates tested gave no PFGE pattern, this is much lower than the 25 to 70% nontypeability reported with other typing methods. No criterion used to identify the isolates as C. perfringens correlated with an inability to obtain PFGE subtype patterns. The inability to subtype some C. perfringens strains by PFGE was also observed in ~5% of isolates in one study (23) and in 10% of isolates in another (6). The reason generally given for this inability to subtype isolates by PFGE is the presence of strong nucleases. Although the actual presence of stable nucleases in these isolates has not been proven, it is the most likely explanation of the presence of degraded DNA observed with these small numbers of nontypeable isolates. Although other factors, such as physical shearing, may produce the same observation in gels, it is unlikely that only a small percentage of isolates would be adversely affected by techniques which are used for all of the isolates.

PFGE provides reliable laboratory criteria that can be used in conjunction with epidemiologic data to establish C. perfringens as the etiologic agent in food-borne disease. This study will provide the basis for a standardized PFGE method for subtyping C. perfringens isolates associated with food-borne disease. A standardized protocol and reference standard will be provided to participants of PulseNet (28). PulseNet is a unique federal-state partnership composed of an integrated network of laboratories formed to quickly detect and respond to outbreaks of food-borne disease; international laboratories are in the process of being added to the network. Inclusion of a C. perfringens standardized molecular subtyping protocol in PulseNet will ensure comparability of DNA fragment patterns between participating laboratories.