Evidence for antibiotic induced Clostridium perfringens diarrhoea
- Department of Microbiology, The General Infirmary and University of Leeds, Old Medical School, Leeds, LS1 3EX, UK
- Dr Wilcox
- Accepted 13 March 2001
Clostridium difficile is a well documented cause of antibiotic associated diarrhoea in hospitalised patients, but may account for only approximately 20% of all cases. This leader reviews the current knowledge and understanding of the pathogenesis, epidemiology, and diagnosis of non-food borne Clostridium perfringens diarrhoea. Although enterotoxigenic C perfringens has been implicated in some C difficile negative cases of antibiotic associated diarrhoea, C perfringens enterotoxin detection methods are not part of the routine laboratory investigation of such cases. Testing for C perfringens enterotoxin in faecal samples from patients with antibiotic associated diarrhoea and sporadic diarrhoea on a routine basis would have considerable resource implications. Therefore, criteria for initiating investigations and optimum laboratory tests need to be established. In addition, establishing the true burden of C perfringens antibiotic associated diarrhoea is important before optimum control and treatment measures can be defined.
Clostridium difficile is the most commonly identified pathogen in hospital acquired infective diarrhoea. However, for most cases (in some series up to 80%) the organism responsible is undiagnosed. Staphylococcus aureus and Clostridium perfringens are the most frequently cited alternative causes of antibiotic associated diarrhoea.1Clostridium perfringens type A is an important cause of bacterial food poisoning world wide. There is mounting evidence to suggest that enterotoxigenic C perfringens strains can play a role in the aetiology of diarrhoeal disease distinct from food poisoning, including antibiotic associated diarrhoea and sporadic diarrhoea in humans.2–9Clostridium perfringens forms part of the normal human gut flora in small numbers—up to 103 colony forming units (cfu)/g. Ingestion of large numbers of vegetative cells (≥ 108 enterotoxin producing organisms) of C perfringens leads to diarrhoeal illness, as in cases of food borne disease. The organisms multiply in the small intestine and sporulate, releasing C perfringens enterotoxin (CPEnt), which is responsible for the classic symptoms of food poisoning. Faecal counts of ≥ 106C perfringens spores are seen in symptomatic cases, although similar counts can be obtained in patients without symptoms. Experimental diarrhoea has been produced in human volunteers after oral administration of CPEnt.10 More recently, CPEnt has been implicated in the pathogenesis of antibiotic associated diarrhoea and sporadic diarrhoea. The aim of this article is to summarise the available evidence for C perfringens as a cause of antibiotic associated diarrhoea and provide an overview of the laboratory methods for diagnosis of C perfringens diarrhoeal diseases.
Evidence linking C perfringens to non-food borne diarrhoeal disease
In 1984, Borriello et al demonstrated that some C difficile toxin negative cases of antibiotic associated diarrhoea were associated with enterotoxigenic strains of C perfringens .2 In their study, faecal specimens from patients with antibiotic associated diarrhoea were examined. Faecal specimens were screened by tissue culture assay for CPEnt, and the presence of enterotoxin was further confirmed by an in house enzyme linked immunosorbent assay (ELISA). All the cases of diarrhoea had high faecal counts (> 106 cfu/g) of C perfringens. Clostridium perfringens enterotoxin was detected in the stools of 11 patients with diarrhoea. The severity, duration, and sporadic nature of the disease was not characteristic of C perfringens food poisoning, and most (n = 10) of the patients developed diarrhoea after antibiotic treatment. Implicated antibiotics included penicillins, cephalosporins, trimethoprim, and cotrimoxazole. Furthermore, CPEnt was not detected in the stools of patients with inflammatory bowel disease (n = 29) and infective diarrhoea (n = 12), or in patients with normal stools (n = 16). These findings led the authors to conclude that enterotoxigenic C perfringens was a possible cause of antibiotic associated diarrhoea. Symptoms associated with non-food borne C perfringens diarrhoeal disease tend to be protracted and more severe than those of food poisoning.11 Common symptoms include abdominal pain and diarrhoea, which are often accompanied by blood and mucous in the faeces. Borriello and colleagues subsequently confirmed their earlier findings by extending the series to 50 patients.3
Samuel et al studied the incidence of diarrhoea associated with CPEnt by screening diarrhoeal samples received from both general practice and hospital patients.4 In addition to screening for common enteric pathogens, samples were also tested for the presence of C perfringens and its enterotoxin. Of the 721 diarrhoeal specimens examined, 25 were shown to contain CPEnt accompanied by high clostridial counts (> 103 cfu/g). The control group samples, consisting of 120 normal stool specimens, were negative for CPEnt and had low faecal counts of C perfringens (≤ 103 cfu/g). The study demonstrated a clear association of C perfringens diarrhoea with elderly hospitalised patients, with 22 of 25 of the cases being hospital inpatients. Most (n = 18) of the patients had received antibiotics before the onset of diarrhoea. The antibiotics associated with the cases were penicillins, cephalosporins, augmentin, erythromycin, trimethoprim, and nitrofurantoin. The clinical symptoms developed on different days over a period of a few weeks and were not typical of food poisoning. A more recent study was designed to investigate the incidence of C perfringens as a cause of antibiotic associated diarrhoea.5 Hospitalised patients were assigned to one of the four groups on the basis of antibiotic usage and diarrhoeal disease. Patients with infective diarrhoea caused by other enteric pathogens were excluded from the study. The control group (n = 181) comprised patients who had or had not received antibiotics but had no diarrhoea. There was no predominant age group or sex. The stool samples were examined for the presence of C difficile and C perfringens toxins. CPEnt was detected in the faeces of 15 patients with antibiotic associated diarrhoea, but in none of the control groups. A further four patients with diarrhoea had detectable CPEnt, although there was no association with antibiotic treatment. The possibility of cross infection was suggested on the basis of geographical clustering, but isolates were not available for strain typing. However, an epidemiological investigation carried out elsewhere during an outbreak of C perfringens diarrhoea showed that infecting serotypes were present in the hospital environment and on the hands of infected patients.6Clostridium perfringens serotypes recovered from symptomatic patients were present in 59% of related environmental samples. Only 9% of samples from areas not associated with symptomatic patients were similarly positive. In a study from Japan, C perfringens isolates recovered from the faeces of elderly hospitalised patients with sporadic diarrhoea were DNA fingerprinted using pulsed field gel electrophoresis (PFGE).7 Of the 60 C perfringens isolates examined, 38 shared the same DNA PFGE pattern. Epidemiological and experimental data suggested that the diarrhoea was not related to a food borne outbreak, but was caused by a nosocomial spread of the bacterium.
Sporadic cases of diarrhoea in the community have also been caused by C perfringens, particularly in the elderly population.8,9 In a study by Mpamugo et al, faecal specimens from isolated sporadic cases of diarrhoea were examined using a reverse passive latex agglutination (RPLA) kit (see below) for enterotoxin detection, and positive results were confirmed by an in house ELISA.9 Enterotoxin was detected by ELISA in 43 of 45 specimens that yielded (in the RPLA kit) a difference of two wells or more between sensitised and control latex, and in 22 of 56 specimens with a one well difference. Hence, enterotoxin was detected in 65 of the 212 faecal specimens examined. Most of the enterotoxin positive cases were not associated with antibiotic treatment. In some of the cases, a single point source of food borne organisms could not be excluded. In addition, a higher proportion (79%) of the enterotoxin positive group than the enterotoxin negative group was resident in the community. The importance of these findings remains uncertain, but the surprisingly high prevalence of enterotoxin positive cases raises questions about the specificity of the detection methods used. In a recent large study of infectious intestinal disease in the community in England, CPEnt was found at a much lower prevalence than in the study by Mpamugo et al, but was more frequently identified in general practitioner patients (4%) than in the population cohort (1.2%).12
Thus, evidence supporting a causal role for C perfringens in non-food borne diarrhoeal disease includes the presence of enterotoxin in the faeces of patients with antibiotic associated diarrhoea or sporadic diarrhoea, as opposed to asymptomatic individuals or individuals suffering from infectious intestinal disease caused by other pathogens.13,14 It might be argued that patients identified as suffering from C perfringens antibiotic associated diarrhoea were in fact cases of food poisoning. Although it is difficult to eliminate the possibility of food poisoning completely, the clinical symptoms described above were not consistent with those of food poisoning. There were no case clusters in relation to time to suggest a common source food borne outbreak. In a recent study, Collie et al used restriction fragment length polymorphism (RFLP) and PFGE to genotype 43 enterotoxigenic C perfringens isolates.14 Results of their study showed that the enterotoxin gene (cpe) is extrachromosomal in C perfringens isolates associated with non-food borne diseases in contrast to the chromosomal enterotoxin gene present in food poisoning isolates.14 The importance of this observation is unclear, but it does strengthen the possibility that C perfringens antibiotic associated diarrhoea is a distinct entity. Furthermore, these findings may be useful for identifying reservoirs of genotypically distinct subpopulations of enterotoxigenic C perfringens isolates, which might help in instituting appropriate preventive measures.
There is a paucity of data relating to the pathogenesis of C perfringens antibiotic associated diarrhoea. Use of antibiotics in elderly hospitalised patients is considered the major risk factor for acquisition of the disease. It is not known whether antibiotic exposure primarily permits the proliferation of small numbers of resident C perfringens strains or allows acquisition of enterotoxigenic C perfringens strains—for example, as a result of impaired colonisation resistance. The symptoms are mediated by the production of CPEnt, which is a 35 kDa polypeptide.14,15 The cpe gene has been cloned, sequenced, and expressed in Escherischia coli.16 Further studies are required to assess the extent of genomic diversity among the food borne and non-food borne diarrhoeal disease isolates. These studies may also help to explain the differences in virulence between the genotypically distinct subpopulations of enterotoxigenic C perfringens isolates. It is widely believed that sporulation is essential for CPEnt production. However, Czeczulin et al reported that sporulation is not essential for CPEnt production, although it does lead to an increased yield of CPEnt.16 At present, regulation of CPEnt expression at the molecular level is not well understood. Studies using gene probe assays have shown that cpe is present in only 6% of global C perfringens isolates, whereas 60% of the isolates associated with food poisoning were found to be cpe positive by hybridisation.17,18 These findings were surprising considering the role of CPEnt in the pathogenesis of C perfringens food poisoning. Further studies identified sequence differences in cpe regions that had been used to design some of the DNA probes.16 These differences may well have interfered with probe hybridisation, resulting in false negative conclusions about cpe positive isolates. It is also possible that some of the isolates tested may have been non-enterotoxigenic normal flora C perfringens isolates because only single colonies were selected for testing.19 More recently, use of gene probe assays based on the cpe sequencing work of Czeczulin and colleagues16 found that only a small proportion of C perfringens isolates carried the cpe gene.20 Of 454 fresh C perfringens isolates from a variety of sources (animal and environmental), only 16 (3.5%) were polymerase chain reaction (PCR) positive for cpe.
The optimum laboratory investigations for diagnosis of C perfringens antibiotic associated diarrhoea have not been established, and the absence of a “gold standard” test hinders further progress. The tests that have been used by previous investigators involve the direct detection of faecal enterotoxin, quantitative stool cultures for C perfringens, and absence of other enteric pathogens in stool specimens. Culture of C perfringens alone does not prove the diagnosis, particularly in light of the high numbers of asymptomatic carriers in the institutional setting.21,22 However, stool cultures provide the isolates from which useful data can be obtained (see below).
Clostridium perfringens can be found in small numbers (up to 103 cfu/g) in the faeces of healthy individuals.4 However, there have been reports of high faecal counts of C perfringens (> 107 cfu/g) in specimens obtained from healthy geriatric patients.21,22 In a study carried out by Yamagishi et al,21 persistently high numbers of C perfringens were found in the faeces of Japanese geriatric patients (five of 30). Stringer et al studied faecal carriage of C perfringens in younger and elderly long stay hospital patients and found similar results in five of 21 elderly patients (counts > 107 cfu/g).22 However, the younger long stay patients in the same hospital had faecal counts in the range of 103 to 104 cfu/g. These findings are in contrast to the findings reported by Samuel et al.4 None of the control group samples (n = 120) in their study, including those from healthy elderly patients, had faecal counts of C perfringens ≥ 103 cfu/g. These conflicting results may be the result of the low numbers of patients recruited in the studies conducted by Yamagishi and colleagues21 and Stringer et al.22 Therefore, it is essential to detect enterotoxin in faecal samples when attempting to confirm C perfringens as the cause of illness.23 A large number of methods including ELISA, RPLA, and tissue culture assays have been used for the detection of C perfringens enterotoxin in faeces, but each has limitations.24,25 Tissue culture assay, using vero cells, is least sensitive (40 ng enterotoxin/g faeces) and reproducible among toxin detection methods.24 An in house ELISA available at the Public Health Laboratory Service central food hygiene laboratory has been suggested as the gold standard among toxin detection methods on the basis of its sensitivity (5 ng enterotoxin/g), specificity, and reproducibility.13,24 The RPLA kit (Oxoid, Basingstoke, UK) has been shown to be both sensitive and reproducible, but non-specific interference by faecal matter has been reported. In addition, some authors have reported that this method has relatively poor sensitivity (50–100 ng enterotoxin/g).25 Recently, a new C perfringens enterotoxin ELISA kit (produced by TechLab; distributed by BioConnections, Leeds, UK) has become commercially available. However, data comparing this ELISA with the methods referred to above are not available. In one study,5 detection of enterotoxin alone using the new ELISA kit was used as the laboratory diagnostic criterion for inpatients with antibiotic associated diarrhoea. A study undertaken at our hospital using the commercial ELISA kit found that 15% of faecal specimens from patients with antibiotic associated diarrhoea were CPEnt positive (NJ Ransome et al, Abstracts of 38th Interscience Conference on Antimicrobial Agents and Chemotherapy, 1998, San Diego. Abstract K 126). This figure is very similar to the results of a previous study.5 Ten percent of isolates from ELISA negative faecal specimens were cpe positive by PCR (NJ Ransome et al, Abstracts of 38th Interscience Conference on Antimicrobial Agents and Chemotherapy, 1998, San Diego. Abstract K 126). It is possible that cpe positive C perfringens isolates produced enterotoxin in vivo to cause disease, but there were insufficient faecal concentrations of CPEnt to be detected by ELISA, thus generating false negative results. This could be compounded by the collection of samples late after the onset of illness, because there is evidence that maximum excretion of enterotoxin occurs early in the food borne diarrhoeal illness.13 Further studies are required to confirm that these findings hold true in cases of non-food borne diarrhoeal illness, and to determine the importance of these preliminary findings. In addition 13 of 19 presumptive C perfringens isolates from ELISA positive faecal specimens were cpe negative (NJ Ransome et al, Abstracts of 38th Interscience Conference on Antimicrobial Agents and Chemotherapy, 1998, San Diego. Abstract K 126). It is possible that these specimens contained very low numbers of cpe positive organisms and consequently tested isolates were PCR negative. Furthermore, C perfringens isolates can lose the enterotoxin gene during subculture.26
Gene probe or PCR assays can be used to determine the enterotoxigenicity of C perfringens isolates and thereby avoid the problems associated with culture, and subsequent enterotoxin production in vitro, and the inherent limitations of toxin detection methods. In an attempt to overcome these potential problems, direct detection of cpe in primary faecal specimens by PCR was undertaken in our hospital, using a PCR protocol based on that of Kokai-Kun et al.20 However, all tested faecal samples, including the specimens spiked with enterotoxigenic C perfringens reference strains, produced negative results, despite universal bacterial DNA primers yielding positive amplification reactions. It is assumed that the high concentration of background bacterial DNA in stool samples may have interfered with amplification. However, elsewhere PCR assays have been used successfully for the direct detection of enterotoxin genes in faeces.27,28 For example, Saito et al successfully used the supernatant of sporulation cultures of faeces as template DNA to detect cpe by PCR.28 Further work is needed to overcome the problems associated with the extraction and amplification of C perfringens DNA from stool samples. Resolution of this problem could render PCR an important tool for understanding more about C perfringens diarrhoeal diseases.
It has been 17 years since Borriello et al first demonstrated the role of enterotoxigenic C perfringens in the aetiology of antibiotic associated diarrhoea or sporadic diarrhoea, but only limited further research has followed. Current knowledge and understanding of the pathogenesis, epidemiology, and diagnosis of C perfringens non-food borne diarrhoeal illness is based on the results of a few published studies, and partly on the extrapolation of data obtained from C perfringens food poisoning outbreaks. Clostridium difficile is a well documented cause of antibiotic associated diarrhoea in hospitalised patients, but may account for only approximately 20% of all cases.29 Although enterotoxigenic C perfringens has been implicated in some of these C difficile negative cases of antibiotic associated diarrhoea, CPEnt detection methods are not part of the routine laboratory investigation of such cases. The criteria for initiating investigations and optimum laboratory tests need to be established. In addition, any diagnostic method used should be rapid, easy to perform, and inexpensive. The performance of the commercial ELISA kit needs to be further evaluated against other methods before being adopted by routine diagnostic laboratories. Testing for C perfringens enterotoxin in faecal samples from patients with antibiotic associated diarrhoea and sporadic diarrhoea on a routine basis would have considerable resource implications. In addition, establishing the true burden of C perfringens antibiotic associated diarrhoea is important before optimum control and treatment measures can be defined.