|
|
| |||||||||||||||||||||||||||||||||||||||
Background Antibiotic-associated hemorrhagic colitis is a distinct form of antibiotic-associated colitis in which Clostridium difficile is absent. Although the cause is not known, previous reports have suggested a role of Klebsiella oxytoca.
Methods We studied 22 consecutive patients who had suspected antibiotic-associated colitis and who were negative for C. difficile. Patients underwent diagnostic colonoscopy, and among those who received a diagnosis of antibiotic-associated hemorrhagic colitis, stool samples were cultured for K. oxytoca. We isolated K. oxytoca strains and tested them for cytotoxin production using a tissue-culture assay. In addition, we also cultured stool samples obtained from 385 healthy subjects for K. oxytoca. An in vivo animal model for antibiotic-associated hemorrhagic colitis was established with the use of Sprague-Dawley rats.
Results Of the 22 patients, 6 had findings on colonoscopy that were consistent with the diagnosis of antibiotic-associated hemorrhagic colitis, and 5 of these 6 patients had positive cultures for K. oxytoca. No other common enteric pathogens were found in the five patients. Before the onset of colitis, all five were receiving penicillins, and two were also taking nonsteroidal antiinflammatory drugs (NSAIDs). All isolated K. oxytoca strains produced cytotoxin. K. oxytoca was found in 1.6% of the healthy subjects. In the animal model, K. oxytoca was found only in the colon of rats that received amoxicillin–clavulanate in addition to being inoculated with K. oxytoca. In these rats, infection with K. oxytoca induced a right-sided hemorrhagic colitis that was not observed in uninfected animals that received amoxicillin–clavulanate, indomethacin (an NSAID), or both.
Conclusions Our fulfillment of Koch's postulates for cytotoxin-producing K. oxytoca suggests that it is the causative organism in at least some cases of antibiotic-associated hemorrhagic colitis. Infection with K. oxytoca should be considered in patients with antibiotic-associated colitis who are negative for C. difficile.
The cause of antibiotic-associated hemorrhagic colitis is unknown. Several mechanisms have been suggested, including allergic reaction,3 mucosal ischemia,7 and infection with Klebsiella oxytoca.2,6 This gram-negative bacterium has been found in patients with antibiotic-associated hemorrhagic colitis, but it is also found in stools of healthy subjects.2 K. oxytoca strains isolated from patients with antibiotic-associated hemorrhagic colitis and half of the strains isolated from healthy subjects have been shown to produce a cytotoxin that is capable of inducing cell death in various epithelial-cell cultures.2,8,9,10 In contrast to C. difficile–induced antibiotic-associated colitis,11 however, there is no in vivo animal model for antibiotic-associated hemorrhagic colitis. Thus, it is uncertain whether K. oxytoca causes antibiotic-associated hemorrhagic colitis, since Koch's postulates have not been fulfilled.
We describe five patients with antibiotic-associated hemorrhagic colitis. A K. oxytoca strain isolated from one of these patients was used in an animal model of antibiotic-associated hemorrhagic colitis, which allowed us to fulfill Koch's postulates.
Methods
Patients and Controls
We studied 22 consecutive patients who had antibiotic-associated colitis but were negative for C. difficile. They underwent diagnostic colonoscopy between March 2001 and April 2004. The diagnosis of antibiotic-associated hemorrhagic colitis was made on the basis of the clinical history (use of antibiotics before the onset of diarrhea) and findings on endoscopy that are typical of segmental hemorrhagic colitis. Stool samples were examined for C. difficile, campylobacter, salmonella, yersinia, shigella, Escherichia coli O157, and C. difficile toxin A. In addition, we examined stool specimens from patients with antibiotic-associated hemorrhagic colitis for K. oxytoca using MacConkey agar plates, and bacterial identification was performed with the use of the API 20E test (bioMérieux). To determine the prevalence of K. oxytoca in normal bowel flora, we tested stool samples from 385 healthy subjects (139 employees and students of the Medical University of Graz and 246 recruits from the Austrian army) for K. oxytoca. The mean age of all healthy subjects was 24 years (range, 18 to 61). All experiments in humans were approved by the local institutional review board or the Austrian Ministry of Defense. Written informed consent was obtained from all patients and controls.
Cytotoxin Assay and Antibiotic Susceptibility
We analyzed K. oxytoca isolates from patients for the production of cytotoxin using a previously established cell-culture assay.2,8 We used as a negative control two strains of K. oxytoca (American Type Culture Collection [ATCC] 13182 and 8724) that had previously been shown to be nontoxigenic.8,9,10 The cytotoxin-producing strain of K. oxytoca, MH 43-1, served as a positive control.8
To obtain a culture supernatant from K. oxytoca, tryptic soy broth was inoculated with a single bacterial colony and incubated at 37°C, with centrifugation at 150 rpm, for 20 hours. Bacterial cultures were then centrifuged at 20,000xg for 10 minutes at 4°C, and the supernatant was filtered through a membrane filter with a pore diameter of 0.2 µm (Iwaki).
Cytotoxin production was assessed with the use of monolayers of HEp-2 cells (ATCC CCL-23) in 96-well tissue-culture plates. After removal of the culture medium, 100 µl of a 1:1 dilution of the filtered supernatant from the culture of K. oxytoca strains with Dulbecco's phosphate-buffered saline was added to the wells. The plates were incubated in a 5% carbon dioxide atmosphere at 37°C for 48 hours and were examined under a microscope. Cytotoxicity was defined as cell rounding and cell death, as described previously.2,8,10 Antibiotic susceptibility testing of isolated K. oxytoca strains was performed with the use of Etest (AB Biodisk).
Experiments in Animals
For in vivo experiments in the animal model, we used female Sprague-Dawley rats that weighed 180 to 200 g and the K. oxytoca strain AHC (antibiotic hemorrhagic colitis) 1, isolated from the stool of Patient 1. Six groups of rats were studied: group 1 (11 rats) received amoxicillin–clavulanate and inoculation with K. oxytoca AHC 1; group 2 (8 rats), amoxicillin–clavulanate only; group 3 (10 rats), inoculation with K. oxytoca AHC 1 only; group 4 (18 rats), amoxicillin–clavulanate, indomethacin, and inoculation with K. oxytoca AHC 1; group 5 (11 rats), amoxicillin–clavulanate and indomethacin only; and group 6 (12 rats), indomethacin only. The experiments lasted for 5 days. Amoxicillin–clavulanate (Augmentin, GlaxoSmithKline) was given intraperitoneally twice daily on days 1 through 5 at a dose of 20 mg, and indomethacin (Liometacin, Chiesi Farmaceutici) was given subcutaneously once daily on days 3 and 4 at a dose of 1 mg. K. oxytoca AHC 1 was inoculated by delivering 1 ml of a 12-hour culture containing about 4x109 colony-forming units (CFU) per milliliter through an orogastric stomach probe on days 2 through 4 in groups 1, 3, and 4; control groups 2, 5, and 6 did not receive K. oxytoca.
On day 5 of the experiment, the animals were killed, and their intestines removed for histologic examination. Stool samples were obtained from the cecum and the distal colon for bacterial culture and testing for C. difficile toxin A. Two investigators, who were unaware of the treatment assignments, evaluated the bowel specimens for histologic changes, using an adapted scoring system12 for four histologic features: mucosal inflammation, epithelial alteration (defined as loss of goblet cells, anisonucleosis, and increased rates of mitosis and apoptosis), mucosal hemorrhage, and erosion. Possible scores for mucosal inflammation were as follows: 0, none; 1, infiltration of the mucosal stroma by neutrophils; 2, infiltration of the epithelial layer by neutrophils; and 3, formation of crypt abscess. Possible scores for epithelial alteration and erosion were as follows: 0, none; 1, mild; 2, moderate; and 3, severe. Possible scores for mucosal hemorrhage were as follows: 0, none; 1, mild, focal; 2, moderate, focal; and 3, diffuse. For comparison, colonic biopsy samples obtained from patients with antibiotic-associated hemorrhagic colitis were evaluated according to the same scoring system. The experiments in the rats conformed to animal-use regulations and were approved by the Austrian Ministry of Education, Science, and Culture.
Statistical Analysis
Statistical analysis was performed with the use of two-way analysis of variance on the ranks for the histologic score and the variables of K. oxytoca in stool from the rats (presence or absence) and the administration of indomethacin (a nonsteroidal antiinflammatory drug [NSAID]) (presence or absence), since those data were not normally distributed. Multiple comparisons were performed with the use of the Student–Newman–Keuls test, and a P value of less than 0.05 was considered to indicate statistical significance.
Results
Clinical Data
Of the 22 patients evaluated, 6 received a diagnosis of antibiotic-associated hemorrhagic colitis, and K. oxytoca was isolated from the stool in 5 of these 6 patients (Table 1). K. oxytoca was found in the stools of 6 of the 385 healthy subjects 1.6%); none of these subjects reported gastrointestinal symptoms such as diarrhea.
|
|
|
During the observation period of 38 months (March 2001 through April 2004), we diagnosed 71 cases of infection with salmonella, 54 cases of infection with campylobacter, 2 cases of infection with shigella, and 121 cases of infection with C. difficile in addition to the 5 cases of antibiotic-associated hemorrhagic colitis induced by K. oxytoca at our hospital, which admits approximately 10,000 patients per year. There were no cases of infection with yersinia or enterohemorrhagic E. coli. The mean age of patients infected with C. difficile was 63 years (range, 20 to 94). Most cases of infection with C. difficile (68%) occurred during hospitalization, and 78% of infected patients had undergone antibiotic therapy.
Cytotoxin Production by K. Oxytoca
The supernatant from all cultures of K. oxytoca strains from patients with antibiotic-associated hemorrhagic colitis, as well as K. oxytoca MH 43-1, had a cytotoxic effect on monolayers of HEp-2 cells, as indicated by cell rounding and cell death (Table 2). In contrast, the supernatant from cultures of the negative control strains (ATCC 13182 and ATCC 8724) was not cytotoxic.
Experiments in Animals
Treatment with amoxicillin–clavulanate reduced the total number of bacteria and changed the bacterial spectrum in rat stool. The most prominent finding was the disappearance of lactobacillus after treatment. Klebsiella and C. difficile and its toxin A were not detected in the stool of rats before or after administration of amoxicillin–clavulanate alone (data not shown).
K. oxytoca AHC 1 colonized the colon of rats receiving amoxicillin–clavulanate in addition to K. oxytoca (groups 1 and 4). The average colonization rate was 7x109 CFU per gram of stool, with no difference between samples taken from the proximal colon and those from the distal colon. Isolated K. oxytoca strains had a pattern of biochemical and antibiotic resistance (the pattern of production of extended-spectrum beta-lactamase) that was identical to that of the inoculated AHC 1 strain. Group 3, inoculated with K. oxytoca but not treated with antibiotics, had no colonization.
We observed colitis (predominantly in the cecum) in rats that were inoculated with K. oxytoca AHC 1 and given amoxicillin–clavulanate (group 1), but not in the control groups receiving either amoxicillin–clavulanate only (group 2) or inoculation with K. oxytoca AHC 1 but no antibiotics (group 3) (Figure 2). The left colon and small bowel were not affected in any treatment group. Major histologic findings were mucosal inflammation, mainly with infiltration of the lamina propria by neutrophils, and epithelial damage. In some rats, a few neutrophils entered the epithelial layer, and there were no crypt abscesses. Epithelial alteration included the loss of goblet cells, marked anisonucleosis, and increased rates of mitosis and apoptosis. Patchy mucosal hemorrhage was a less prominent finding.
|
Histologic specimens obtained from rats with colitis resembled those from patients with antibiotic-associated hemorrhagic colitis (Figure 3), and the histologic scores were similar (Figure 2).
|
Five of six patients with antibiotic-associated hemorrhagic colitis had K. oxytoca in the stool, whereas stool cultures for other intestinal pathogens, including C. difficile, were negative. All K. oxytoca isolates from patients with antibiotic-associated hemorrhagic colitis produced cytotoxin on cell-culture assay.
The clinical features of antibiotic-associated hemorrhagic colitis in our patients were similar to those described previously.3,4,6,13 The condition was observed mainly in young and otherwise healthy outpatients after brief treatment with penicillin derivatives, whereas C. difficile–associated diarrhea occurred mainly in older, hospitalized patients. The observed colitis was segmental and was mainly present in the right or transverse colon. Whereas sigmoidoscopy is thought to be sufficient for the diagnosis of antibiotic-associated colitis induced by C. difficile, colonoscopy is required for the diagnosis of antibiotic-associated hemorrhagic colitis, since sigmoidoscopy alone would miss many cases.
To establish a causal link between antibiotic-associated hemorrhagic colitis and toxigenic K. oxytoca according to Koch's postulates, we developed an in vivo animal model. In rats receiving antibiotics to suppress the physiologic colonic flora, oral inoculation of a K. oxytoca strain isolated from Patient 1 resulted in colonization of the colon in high concentrations and induced right-sided hemorrhagic colitis. In the absence of treatment with antibiotics, inoculation with K. oxytoca resulted in neither colonization of the colon nor colitis in rats. The histologic features and location of colitis (the right colon) in this animal model were similar to those in patients with antibiotic-associated hemorrhagic colitis. Thus, Koch's postulates for antibiotic-associated hemorrhagic colitis have largely been fulfilled: an organism (K. oxytoca) was identified in most patients with a disease that is clinically well described, the organism was cultured and caused disease (right-sided colitis) in animals, and K. oxytoca could again be cultured from the diseased colon in the animals.
Previous data suggesting K. oxytoca as the cause of antibiotic-associated hemorrhagic colitis were limited to descriptive clinical observations of patients with antibiotic-associated hemorrhagic colitis and data from in vitro studies using a cytotoxin isolated from K. oxytoca. Several case reports and small series have demonstrated that K. oxytoca is present in the stool of most patients with antibiotic-associated hemorrhagic colitis.2,4,5,6,8,14 The prevalence of K. oxytoca among our healthy subjects was 1.6% in Austria, as compared with 9% in France,2 suggesting that the prevalence varies by geographic region. Two case reports, in which K. oxytoca was the presumed cause of colitis in patients who had not received antibiotic therapy,15,16 deserve special attention, since they suggest that K. oxytoca may be pathogenic even in the absence of antibiotics under certain circumstances.
In experiments performed by two groups of researchers in Japan, K. oxytoca strains isolated from patients with antibiotic-associated hemorrhagic colitis were shown to be capable of producing a cytotoxin that was toxigenic in cultures of HEp-2, Vero, Chinese hamster ovary, and HeLa cells.8,10 Furthermore, this cytotoxin, a low-molecular-weight organic substance, induced mucosal damage in isolated loops from rabbit ileum.17 However, neither the isolated cytotoxin nor the toxin-producing strain of K. oxytoca (OK-1) induced inflammation or damage in the colonic mucosa of rabbits.17 The cytotoxic effect was also observed by Beaugerie et al.: 14 of 17 K. oxytoca strains (82%) obtained from patients with antibiotic-associated hemorrhagic colitis, and 42 of 90 (47%) obtained from healthy controls, were cytotoxic to HEp-2 cells in their study.2
Histologic findings in patients with antibiotic-associated hemorrhagic colitis are similar to those in patients with other toxin-induced forms of colitis, such as colitis due to enterohemorrhagic E. coli or shigella.7,18,19 The histologic findings in patients with these forms of colitis may resemble the findings in patients with ischemic colitis18,19; the resemblance with ischemic colitis has also been described for antibiotic-associated hemorrhagic colitis.7 Furthermore, abnormalities are mainly located in the right colon both in patients with hemorrhagic colitis due to enterohemorrhagic E. coli18 and in those with antibiotic-associated hemorrhagic colitis. Crypt abscesses, which are commonly found in patients with infectious colitis due to invasive bacteria, were absent in both humans and rats with antibiotic-associated hemorrhagic colitis in our study. We therefore assume that K. oxytoca induces colitis, probably by means of the cytotoxin produced by certain strains.
K. oxytoca strains isolated from three of our patients were resistant to the antibiotics that were administered before the onset of colitis. K. oxytoca isolates in two patients receiving amoxicillin–clavulanate were, however, sensitive in vitro to this beta-lactam antibiotic. Previous studies investigating the effect of orally administered amoxicillin–clavulanate on fecal microflora revealed that fecal enterobacteria that are sensitive to amoxicillin–clavulanate may persist in the intestinal microflora during antibiotic therapy.20,21 This may be due to very low fecal concentrations of amoxicillin–clavulanate.20 We assume that, in our study, low amoxicillin–clavulanate concentrations, especially in the right colon, were sufficient to suppress most bacteria in the intestinal microflora but were too low to inhibit the growth of K. oxytoca.
We suggest the following pathogenetic model for antibiotic-associated hemorrhagic colitis. Cytotoxin-producing K. oxytoca strains are present at least temporarily in the colon of some people, in whom antibiotic therapy can lead to overgrowth of toxigenic K. oxytoca in the colon, owing to selection of this organism. Since klebsiella is nearly always resistant to penicillins because it expresses a beta-lactamase, antibiotic-associated hemorrhagic colitis is particularly seen during therapy with penicillin derivatives. Overgrowth of K. oxytoca in the colon finally results in high cytotoxin concentrations that induce mucosal damage.
We observed frequent use of NSAIDs in our patients with antibiotic-associated hemorrhagic colitis, as have other investigators.6,7 Use of NSAIDs has also been reported to be associated with colitis and diarrhea of various causes.22,23,24 In our experiments in rats, concomitant administration of indomethacin and inoculation of K. oxytoca tended to cause more severe colitis and to induce erosion. Thus, NSAIDs appear to aggravate colitis caused by potentially pathogenic bacteria in the intestinal lumen.
Our study suggests that toxigenic K. oxytoca should be considered in the differential diagnosis of potential intestinal pathogens. When a patient with antibiotic-associated colitis is negative for C. difficile, stool samples should be cultured for K. oxytoca. In these patients, it will probably suffice to discontinue antibiotics and NSAIDs and to begin symptomatic therapy. We consider it important to draw attention to toxigenic K. oxytoca, since relatively young patients may have received a misdiagnosis of ischemic colitis with spontaneous resolution in the past, and since the diagnosis of antibiotic-associated hemorrhagic colitis has consequences for the care of such patients, who could be harmed by unnecessary treatment with drugs.
Supported by a grant from the Medical University of Graz.
No potential conflict of interest relevant to this article was reported.
We thank Prof. Gregor Högenauer, Prof. Friederike Turnowsky, Georg Waeg, Ph.D., and Gebhard Feierl, M.D., for their helpful ideas and advice; Hans Hosbein, Andrea Fuchsbichler, Christina Strempfl, and Waltraut Eberharth for their technical help with the experiments; Babak Yazdani-Biuki, M.D., Col. Hans-Christoph Stenzel, M.D., and First Sgt. Thomas Scheucher for their help in obtaining samples from healthy subjects; Dr. T. Chida of the Department of Microbiology, Medical and Dental University, Tokyo, for kindly providing the cytotoxin-producing K. oxytoca strain MH 43-1 that was used as a positive control; and Eugenia Lamont, B.A., for editorial assistance.
Source Information
From the Department of Internal Medicine (C.H., R.K., N.G., G.J.K., T.A.H.), the Institute of Pathology (C.L., G.G.), and the Institute of Experimental and Clinical Pharmacology (E.B., I.T.L., R.S.), Medical University of Graz, Graz, Austria.
Address reprint requests to Dr. Högenauer at the Department of Internal Medicine, Medical University of Graz, Auenbruggerplatz 15, A-8036 Graz, Austria, or at christoph.hoegenauer{at}meduni-graz.at.
References
2 microglobulin transgenic rats. J Clin Invest 1996;98:945-953. [ISI][Medline]
| |||||||||||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2008 Massachusetts Medical Society. All rights reserved. |
Previous
