ABSTRACT
Clostridium perfringens enterotoxin (CPE) is responsible for the diarrheal and cramping symptoms of human C. perfringens type A food poisoning. CPE-producing C. perfringens isolates have also recently been associated with several non-food-borne human gastrointestinal (GI) illnesses, including antibiotic-associated diarrhea and sporadic diarrhea. The current study has used restriction fragment length polymorphism (RFLP) and pulsed-field gel electrophoresis (PFGE) analyses to compare the genotypes of 43 cpe-positive C. perfringensisolates obtained from diverse sources. All North American and European food-poisoning isolates examined in this study were found to carry a chromosomal cpe, while all non-food-borne human GI disease isolates characterized in this study were determined to carry their cpe on an episome. Collectively, these results provide the first evidence that distinct subpopulations ofcpe-positive C. perfringens isolates may be responsible for C. perfringens type A food poisoning versus CPE-associated non-food-borne human GI diseases. If these putative associations are confirmed in additional surveys, cpe RFLP and PFGE genotyping assays may facilitate the differential diagnosis of food-borne versus non-food-borne CPE-associated human GI illnesses and may also be useful epidemiologic tools for identifying reservoirs or transmission mechanisms for the subpopulations ofcpe-positive isolates specifically responsible for CPE-associated food-borne versus non-food-borne human GI diseases.
Clostridium perfringenstype A food poisoning currently ranks as the second most common foodborne disease in the United States (2). The diarrhea and cramps that comprise the typical clinical symptoms of this human illness are induced by a single 35-kDa polypeptide named C. perfringens enterotoxin (CPE) (20, 23, 24). Althoughcpe-positive isolates represent only a very small fraction of the global C. perfringens population (16, 19, 31, 32), recent epidemiologic studies (5-8, 10, 17, 22, 26) indicate that these bacteria can also cause non-food-borne gastrointestinal (GI) illnesses in humans. In particular, some surveys (5, 10, 26) have suggested that CPE-producing C. perfringens may be responsible for ∼10% of all cases of antibiotic-associated diarrhea (AAD) and 5 to 20% of all cases of sporadic non-food-borne diarrhea (SPOR). Some evidence implicating CPE-producing C. perfringens as a cause of AAD and SPOR includes the following. (i) Feces from many humans suffering from AAD (5, 7, 22) or SPOR (10, 17, 26) contain CPE at levels comparable to those found in feces from C. perfringens type A food-poisoning victims, while CPE is rarely, if ever, detectable in feces from either healthy individuals or individuals suffering from gastroenteritis caused by other enteropathogens (1, 3, 27); (ii) in the absence of other known enteropathogens, unusually high levels of C. perfringens cells or spores are present in feces from many individuals suffering from AAD or SPOR (7, 10); and (iii) many C. perfringens strains isolated from the feces of individuals suffering from AAD or SPOR are able to express CPE (7, 9), in contrast to the lack of expression of CPE by nearly allC. perfringens isolates found in feces from either healthy individuals or individuals suffering from gastroenteritis caused by other pathogens (27, 28).
Genotypic differences have recently been identified amongcpe-positive C. perfringens isolates. Results from two studies (13, 18) indicate that at least someC. perfringens type A food-poisoning isolates carrycpe on their chromosomes and that at least somecpe-positive C. perfringens veterinary isolates carry cpe on an episome (note that cpe-positiveC. perfringens isolates are also considered important causes of enteric disease in domestic animals [30]). However, these putative source-related genotypic associations remain tentative due to the relatively small number and limited (European) geographic origin of enterotoxigenic C. perfringens isolates genotyped to date.
To better appreciate the role that genotypically distinct subpopulations of enterotoxigenic C. perfringens may play in CPE-associated human (and veterinary) diseases, the present study has genotypically characterized >40 cpe-positive C. perfringens isolates. Notably, the isolates used in our study originated from considerably more diverse host, geographic, and disease sources than the cpe-positive isolates genotyped to date and include a sizeable number ofcpe-positive non-food-borne human GI disease isolates which had not yet been genotypically characterized. Results from the current study indicate that most (or all) cpe-positive food-poisoning isolates, regardless of their isolation date or geographic origin, carry a chromosomal cpe. Furthermore, our study also presents the first evidence that cpe is episomal in many (or all) CPE-associated non-food-borne human GI disease isolates. These new findings hold potential epidemiologic significance for suggesting that genotypically distinct subpopulations ofcpe-positive C. perfringens isolates may be responsible for human C. perfringens type A food poisoning and CPE-associated non-food-borne human GI diseases.
MATERIALS AND METHODS
Bacterial strains.The C. perfringens isolates used in this study are listed and described in Table1.
Summary of genotypic characterization results obtained in this study for cpe-positive C. perfringens isolates
Preparation of a DIG-labeled cpe-specific probe.A 639-bp digoxigenin (DIG)-labeled, double-stranded DNA probe corresponding to internal cpe sequences was prepared by a two-step PCR amplification method, as described previously (14).
Restriction fragment length polymorphism (RFLP) analysis.Starter vegetative C. perfringens cultures were grown overnight at 37°C in fluid thioglycolate broth (FTG) (Difco), as described previously (15). A 500-μl aliquot of each FTG starter culture was then used to inoculate 10 ml of TGY broth (3% Trypticase, 2% glucose, 1% yeast extract, 0.1% cysteine [14]), and these TGY cultures were incubated for 8 h at 37°C.
A previously described protocol (14) was used to isolate total C. perfringens DNA from each TGY culture. Samples of each isolated DNA were then digested to completion with eitherNruI (New England BioLabs) or EcoRI (Boehringer Mannheim), according to the manufacturer’s specifications. Eight micrograms of the digested DNA was then subjected to conventional electrophoresis on 0.8% agarose gels, transferred to positively charged nylon membranes (Nytran Maximum Strength; Schleicher & Schuell), and fixed with UV light (29). The 639-bp DIG-labeled cpe gene probe, prepared as indicated above, was hybridized to these blots, as described previously (14), and the blots were washed at high stringency (29). Hybridized probe was detected with a DIG chemiluminescence detection system with Lumi-Phos 530 substrate (Boehringer Mannheim).
PFGE studies. C. perfringens cultures were grown to the mid-exponential phase (i.e., for approximately 5 h) at 37°C in TGY broth. Cells from these TGY broth cultures were used to prepare genomic C. perfringens DNA in agarose plugs, as described previously (12). DNA in 40 μl of an agarose plug was then incubated overnight at 37°C with or without 1.6 U of I-CeuI (New England Biolabs) in 200 μl of the buffer solution recommended by the enzyme manufacturer. Following these incubations, the DNA samples were analyzed by pulsed-field gel electrophoresis (PFGE) with 0.7% agarose gels prepared with PFGE-grade agarose (Bio-Rad). PFGE was performed with a Bio-Rad CHEF-DR II apparatus, with pulse times ramped from 20 to 120 s over 12 h, followed by ramping from 60 to 100 s over 12 h (21). After PFGE, the gels were subjected to Southern analysis by using the same procedure described above for DNA samples subjected to conventional electrophoresis, except that the PFGE blots were washed under standard stringency conditions (4).
The rationale for these PFGE analyses, which have been used previously to genotype cpe-positive C. perfringens isolates (13, 18), is that, without any restriction enzyme digestion, unsheared C. perfringens chromosomal DNA is too large to enter a pulsed-field gel. However, because of its smaller size, at least some episomal DNA should enter a pulsed-field gel, even without any restriction enzyme treatment. Furthermore, since all I-CeuI sites are chromosomal in C. perfringensDNA (13, 18), I-CeuI digestion of samples should produce chromosomal DNA fragments that can enter pulsed-field gels, but treatment with this restriction enzyme should not effect the migration of episomal DNA. Consequently, if isolates with a chromosomalcpe are analyzed by this PFGE protocol, theircpe-containing DNA should remain in the gel wells when samples are not treated with I-CeuI, but somecpe-containing DNA from these isolates should enter the gel when samples are treated with I-CeuI. In contrast, PFGE analysis of isolates carrying an episomal cpe should show similar migration of cpe-containing DNA into gels, whether or not a sample is treated with I-CeuI.
RESULTS
Comparative RFLP analysis of cpe-positive isolates.A study (13) recently reported thatcpe is localized to 5-kb NruI and 10-kbEcoRI DNA fragments in several European enterotoxigenic food-poisoning isolates but is present in several European veterinary isolates on DNA of >20 kb following digestion with eitherNruI or EcoRI. Considering that only a limited number of isolates, all with similar origins, were examined in that previous study (13), in our current study we initially tested these putative source-related cpe RFLP patterns using our laboratory’s collection of cpe-positive C. perfringens isolates.
Our initial cpe RFLP experiments confirmed a previous report (13) that European food-poisoning isolates NCTC 8239, NCTC 8798, and NCTC 10239 all carry cpe on 5-kb NruI and 10-kb EcoRI fragments (see Fig.1 for representative RFLP blot results for NCTC 8239 and NCTC 10239 and Table 1 for a summary of our results for all three of these strains). When similar RFLP analyses were extended to DNA from four European food-poisoning isolates and eight North American food-poisoning isolates that had not been previously genotyped, cpe was found to be similarly localized (see Fig.1 for representative RFLP blot results for previously unexamined strains NCTC 8238, NCTC 8359, NCTC 8799, FD 1041, and C-1841 and Table1 for a summary of our RFLP results for all 12 previously unexamined food-poisoning isolates) to 5-kb NruI and 10-kbEcoRI DNA fragments.
RFLP analysis of NruI-digested total DNA from selected C. perfringens disease isolates demonstrating presumptive evidence for episomal versus chromosomal distribution ofcpe in isolates from different disease sources. Southern blots were probed with a 639-bp DIG-labeled cpe-specific fragment. VET, veterinary. All samples except 8239/EcoRI were cut with NruI; 8239/EcoRI was cut withEcoRI and the results are shown for comparison. Molecular sizes (in kilobase pairs) of the DNA markers are given to the left of each blot. See Table 1 for full strain designations.
The cpe RFLP profiles reported for food-poisoning isolates in Table 1 (and elsewhere [13]) differ from those reported previously (13) for several veterinary isolates, all of which were from Europe. To confirm that the cpe RFLP profiles of cpe-positive veterinary isolates consistently differ from those of food-poisoning isolates, we next performedcpe RFLP analyses with 10 cpe-positive veterinary isolates that had not been previously genotypically characterized (Table 1). These RFLP analyses revealed that all nine of the North American veterinary isolates in our collection share similarNruI or EcoRI cpe RFLP patterns (see Fig. 1 for representative RFLP blot results for veterinary isolate 452 and Table 1 for a summary of our RFLP results for all nine North American veterinary isolates examined), with cpe always localizing to DNA of >20 kb following either NruI orEcoRI digestion. While the RFLP patterns of these North American veterinary isolates match those reported previously (13) for European veterinary isolates, our study has also identified the first veterinary isolate with an atypical cpeRFLP pattern. European veterinary isolate 29, which was reputedly obtained from a diarrheic pig, was observed to carry its cpeon a 5-kb NruI and a 10-kb EcoRI fragment (Table1).
Having confirmed, with the single exception of isolate 29, that food-poisoning and veterinary enterotoxigenic C. perfringensisolates exhibit consistent differences in their cpe RFLP patterns, we next performed the first cpe RFLP analyses of enterotoxigenic isolates originating from patients with CPE-associated non-food-borne human GI diseases. Interestingly, these RFLP analyses (see Fig. 1 for representative RFLP blot results for isolates B11, B45, F4591, and F4969 and Table 1 for a summary of our RFLP results for all non-food-borne human GI disease isolates examined) demonstrated that, following either NruI or EcoRI digestion,cpe is present on DNA fragments of >20 kb for all 7 AAD isolates and all 11 SPOR isolates in our collection. Therefore, these results demonstrate that the non-food-borne human GI isolates in our collection consistently exhibit cpe RFLP patterns different from those of human food-poisoning isolates.
Finally, the specificity of the cpe probe used in the current RFLP studies has been confirmed by control studies demonstrating (Table 1) that this probe does not hybridize to eitherNruI- or EcoRI-digested DNA from C. perfringens ATCC 3624, a strain previously shown to becpe negative (14, 25).
PFGE confirmation of episomal versus chromosomal locations forcpe in selected isolates.The RFLP results presented above supporting source-related genotypic differences betweencpe-positive C. perfringens isolates are particularly interesting in light of recent studies (13, 18) showing that cpe is chromosomal in several European food-poisoning isolates carrying cpe on 5-kb NruI and 10-kb EcoRI fragments but is present on an episome in several European veterinary isolates carrying cpe on DNA fragments of >20 kb after NruI or EcoRI digestion. Therefore, the RFLP results shown in Fig. 1 and Table 1apparently suggest that all food-poisoning isolates in our collection carry a chromosomal cpe and that all non-food-borne human GI disease and veterinary (except isolate 29) isolates in our collection carry cpe on an episome.
To test this hypothesis, selected isolates in our collection were analyzed by the same PFGE approach (see Materials and Methods) used in previous studies (13, 18) to establish chromosomal or episomal locations for cpe in selected European veterinary and food-poisoning isolates. When we performed these PFGE analyses with food-poisoning isolate NCTC 10239, no migration ofcpe-containing DNA into pulsed-field gels was detected in the absence of I-CeuI treatment (data not shown). However, an ∼360-kb cpe-containing restriction fragment was observed when DNA from this food-poisoning strain was treated with I-CeuI prior to PFGE (data not shown). These results are in complete agreement with previous PFGE characterization of this strain (13, 18). When the same PFGE analyses were extended to North American food-poisoning isolates FD 1041 (see Fig.2 and Table 1) and C-1841 (Table 1), which had not previously been examined by PFGE, the same migration patterns were observed for the cpe-containing DNAs of these two strains, as reported above for cpe-containing DNA from NCTC 10239. Therefore, our PFGE results confirm that all three of these food-poisoning isolates carry a chromosomal cpe.
PFGE evidence supporting the localization ofcpe to episomes in isolates from some disease sources. PFGE and Southern hybridization analysis of undigested and I-CeuI-digested genomic DNA from select C. perfringens disease isolates. (A) pulsed-field gel stained with ethidium bromide. (B) Southern blot of the gel shown in panel A. The blot was probed with a 639-bp DIG-labeled cpe-specific fragment. FP, food poisoning; VET, veterinary; U, undigested, intact genomic DNA; C, I-CeuI-digested DNA. The gel was calibrated with Saccharomyces cerevisiae chromosomal DNA. Molecular sizes of the DNA markers are given in the center. See Table 1 for full strain designations.
The same PFGE analyses described above were also performed with DNAs from North American veterinary isolates 222, 452, and 458, which also have not been previously analyzed by PFGE, in order to confirm our RFLP results suggesting that these isolate carry an episomal cpe. PFGE analyses of DNAs from these three North American veterinary isolates showed that their cpe-containing DNA migrates into pulsed-field gels in the absence of I-CeuI treatment. Further supporting the fact that cpe is episomally localized (see Materials and Methods) in these three isolates, we observed no change in the migration of cpe-containing DNA from isolates 222, 452, and 458 with I-CeuI treatment (see Fig. 2 for representative PFGE blot results for isolate 452 and Table 2 for a summary of PFGE conclusions for all three veterinary isolates examined). During these studies we also attempted to perform similar PFGE analyses with DNA from isolate 29, the unusual veterinary isolate that exhibits an RFLP profile suggesting a chromosomal cpegene. Unfortunately, this strain was not amenable to PFGE analyses, yielding DNA smears on PFGE gels even in the absence of treatment with I-CeuI (data not shown). Similar PFGE smearing has been observed (Table 1) (13) for food-poisoning strain C. perfringens NCTC 8239 and is believed to result from the fact that some C. perfringens strains produce unusually large amounts of a very stable endogenous nuclease(s) (12).
The final series of experiments conducted in the current study involved extending PFGE analyses to eight representative non-food-borne human GI disease isolates from our collection (including four AAD and four SPOR isolates, none of which have previously been examined by PFGE). Interestingly, DNAs from these eight non-food-borne human GI disease isolates were found to exhibit the same PFGE behavior described above for DNAs from veterinary isolates 222, 452, and 458, which carry an episomal cpe. That is, we observed the entrance ofcpe-containing DNA species from all eight of these AAD or SPOR isolates into pulsed-field gels in the absence of I-CeuI treatment, and the migration of thesecpe-containing species was unaffected by I-CeuI treatment (see Fig. 2 for representative PFGE blot results for isolates B11 and F4969 and Table 2 for a summary of PFGE results for all 8 non-food-borne disease isolates examined in this study).
Finally, the specificity of the cpe probe used in our PFGE studies was confirmed by demonstrating that this probe did not hybridize to either nondigested (Fig. 2) or I-CeuI-digested (data not shown) samples of DNA from cpe-negative strain ATCC 3624 that had been run on pulsed-field gels.
DISCUSSION
The current study offers several significant contributions to our understanding of the recently discovered (13, 18) subpopulations of cpe-positive C. perfringens. First, the RFLP results presented in this study for three previously examined European food-poisoning isolates (NCTC 8239, NCTC 8798, and NCTC 10239) and four previously unexamined European food-poisoning isolates (NCTC 8235, NCTC 8238, NCTC 8359, and NCTC 8799), which are supported by PFGE results for NCTC 10239, significantly strengthen the hypothesis (13, 18) that most, if not all, European food-poisoning isolates carry a chromosomal cpe. Our RFLP results confirming that NCTC 8235 is cpe positive are particularly interesting. Even though NCTC 8235 was originally isolated as an enterotoxigenic strain, a recent study (13) could not detect cpe in ATCC 12922, which is reputedly the same strain as NCTC 8235. Collectively, these results for NCTC 8235/ATCC 12922 could be indicating that the cpe gene disappeared from some subcultures of NCTC 8235. If so, these subcultures may have lost a mobile cpe-containing genetic element that had been integrated into the NCTC 8235 chromosome, since a putativecpe-carrying transposon has recently been identified in the chromosome of NCTC 8239, another European food-poisoning strain (11).
The current study also reports the first genotypic characterization ofcpe-positive North American food-poisoning isolates. Our RFLP results for eight North American food-poisoning isolates, supported by PFGE results for North American food-poisoning isolates FD1041 and C-1841, indicate that all these isolates (which originated from three separate food-poisoning outbreaks) also carry a chromosomalcpe. When these results are coupled with similar results for European food-poisoning isolates ((13, 18); this study), it is now clear that cpe has a chromosomal location in most (or all) C. perfringens food-poisoning isolates, regardless of the isolate’s geographic origin or isolation date (note that the food-poisoning isolates examined in our current study and previous studies [13, 18] were isolated over a period extending from the 1950s to the 1990s).
Additionally, the PFGE and RFLP results presented in the current study provide the first evidence that most (or all) cpe-positive North American veterinary isolates carry an episomal cpe. When combined with previous genotyping results (13, 18) for European cpe-positive veterinary isolates, it now appears that most cpe-positive veterinary isolates, regardless of geographic origin or date of isolation, carry an episomalcpe. However, the current study has also identified the first anomalous cpe-positive veterinary isolate (isolate 29) whose DNA does not fit the standard genotypic pattern for these veterinary isolates, i.e., RFLP analyses suggest that isolate 29 is an unusual cpe-positive veterinary isolate carrying a chromosomal cpe. Since isolate 29 is a porcine isolate, this observation could provide an insight into the epidemiology of C. perfringens type A food poisoning by suggesting that some cases ofC. perfringens type A food poisoning, which almost always results from consumption of a contaminated meat product (22), may be caused by ingesting meats originating from animals that had been contaminated with atypical C. perfringens veterinary isolates (like isolate 29) carrying the chromosomal cpe linked to food poisoning.
The current study’s most significant finding is presentation of the first evidence that cpe is episomal in many or allcpe-positive isolates originating from CPE-associated non-food-borne human GI diseases. While this association is based upon PFGE analysis of eight different non-food-borne human GI disease isolates (and is further implied by the similar RFLP patterns shared by these 8 isolates analyzed by PFGE and the 10 other non-food-borne human GI disease isolates examined in the current study), caution should still be exercised before extrapolating the overall involvement of isolates with episomal cpe in CPE-associated non-food-borne human GI diseases from the current data. Since the AAD and SPOR isolates used in our current study all share a similar (European) geographic origin, it will be important to conduct additional surveys with isolates from other geographic locations to confirm the putative association between episomal cpe isolates and CPE-associated non-food-borne human GI diseases suggested by our study. In this respect, it should be noted that reliably characterized North American enterotoxigenic AAD and SPOR C. perfringens isolates were unavailable for our current study, reflecting the limited research attention CPE-associated non-food-borne human GI illnesses have received to date in the United States.
Pending results from additional surveys, the fact that all 18 non-food-borne human GI disease isolates surveyed in the current study apparently carry an episomal cpe still holds considerable significance. First, this result provides important new evidence that enterotoxigenic C. perfringens isolates are, in fact, responsible for non-food-borne human GI diseases. Although this association has been strongly suggested by classical epidemiologic studies (see Introduction), these previous studies have not totally eliminated the possibility that individuals putatively identified as suffering from CPE-associated AAD or SPOR were, instead, victims of isolated, unrecognized cases of C. perfringens type A food poisoning. However, if these individuals had been sickened by C. perfringens type A food poisoning, then it would be expected that the “non-food-borne human GI disease” isolates in our collection should have exhibited the same genotypic patterns as isolates causing recognized cases of C. perfringens type A food poisoning.
Second, genotyping results for our non-food-borne disease isolates now strongly suggest that both episomal and chromosomalcpe-positive isolates can be enteropathogenic for humans. However, there may be differences in virulence between these subpopulations of cpe-positive C. perfringensisolates, based upon epidemiologic reports (5, 10, 17, 22, 26) indicating that the diarrheal symptoms of CPE-associated non-food-borne human GI diseases are generally more severe and of longer duration than the typical symptoms of C. perfringenstype A food poisoning. Since the basis for these differences in symptomatology between CPE-associated human GI diseases remains unclear, future studies should compare the virulence phenotypes of chromosomal versus episomal cpe-positive isolates. Also, note that the identification of isolate 29 (which apparently carries a chromosomal cpe and which originated from a diarrheic pig) in the current study strongly suggests that both chromosomal (isolate 29) and episomal (13, 18; this study)cpe-positive isolates can be enteropathogenic for domestic animals.
Third, by implying that enterotoxigenic isolates from CPE-associated non-food-borne human GI diseases often (or always) differ genotypically from C. perfringens type A food-poisoning isolates, the current results also provide the first evidence that distinct subpopulations of cpe-positive C. perfringensisolates may be linked to particular types of CPE-associated human GI diseases. This insight could have potentially important diagnostic and epidemiologic implications. For example, if the associations suggested by this and previous genotyping studies are verified in further surveys, RFLP or PFGE assays for cpe would become useful for differentially diagnosing cases of CPE-associated food-borne versus non-food-borne diarrhea. In this regard, note that results with >20 isolates (13; this study) indicate that RFLP assays for cpe are very reliable predictors of whether an isolate carries an episomal versus chromosomal cpe. Therefore, RFLP appears to be a useful alternative for the presumptive identification of chromosomal versus episomal cpe isolates in laboratories lacking PFGE equipment (it also deserves mention that existingcpe detection-based assays other than PFGE and RFLP cannot distinguish chromosomal versus episomal cpe-positive isolates).
Similarly, cpe RFLP or PFGE analyses could prove to be useful for epidemiologic surveys for the identification of reservoirs for each genotypic subpopulation of cpe-positive isolates. Identification of these reservoirs, which are poorly understood at present, may allow for the introduction of measures specifically aimed at preventing the introduction of chromosomal cpe isolates into foods or preventing the spread of CPE-associated non-food-borne diseases, which may be exogenous infections (5, 7). These epidemiologic surveys should also better determine the relative prevalence of chromosomal versus episomal cpe-positive isolates, both globally and in specific ecologic niches. Although a previous study (18) claims that chromosomal cpeisolates are less common in nature than episomal cpeisolates, this conclusion appears premature since it was based upon surveys (13, 18) that predominantly examined veterinary isolates, which mostly carry an episomal cpe (13, 18; this study).
Finally, by strongly suggesting that many (or most) CPE-associated non-food-borne human GI disease isolates carry an episomalcpe, the current study also provides important additional support for the growing appreciation (18) of the important role that extrachromosomal genetic material plays in the virulence of clostridia involved in human infections. A future challenge will be to determine whether the cpe-carrying episome(s) present in most, or all, of our non-food-borne human GI disease (and most veterinary) isolates is a large plasmid or a phage capable of extrachromosomal replication.
ACKNOWLEDGMENTS
We thank Saleem Khan for assistance with PFGE experiments and helpful discussions. We also thank Peter Borriello, M. M. Brett, Ronald Labbe, David Mahony, J. Glenn Songer, and D. J. Taylor for generously providing us with many of the isolates used in this study.
This work was supported by Public Health Service grant AI 19844-15.
FOOTNOTES
- Received 5 May 1997.
- Returned for modification 1 August 1997.
- Accepted 1 October 1997.
- Copyright © 1998 American Society for Microbiology