Abstract
During an outbreak of diarrhea in a general hospital in 1992, 166 Clostridium difficile isolates from 102 patients were typed by restriction enzyme analysis (REA), arbitrarily primed PCR (AP-PCR), and protein profile analysis (PP) techniques. A total of 18 types and 5 subtypes were identified by REA, 32 types were identified by AP-PCR, and 9 types were identified by PP. Analysis of the data indicated the presence of a predominant strain among 76, 75, and 84% of the isolates by REA, AP-PCR, and PP, respectively. Subsequently, 45 C. difficile isolates which had been collected in 1990 from 33 patients in the same hospital following a significant increase in the number of cases of diarrhea caused by C. difficile were studied by REA, AP-PCR, and PP typing techniques. Thirteen types and one subtype were identified by REA, 12 types were identified by AP-PCR, and 5 types were identified by PP. As with the isolates from 1992, a dominant strain was identified. This strain was represented by 53, 64, and 70% of the total number of isolates when the strains were typed by REA, AP-PCR, and PP, respectively. Every isolate (210 of 211) from both 1990 and 1992 that was available for typing was typeable by all three methods. Furthermore, the same dominant strain was identified in both 1990 and 1992 by each method. This study demonstrates that each of the three typing methods can be useful in epidemiologic investigations of C. difficile outbreaks and that one strain can be dominant in an institution over a number of years.
Clostridium difficile is the most common cause of nosocomial diarrhea (13), and prolonged hospital outbreaks of diarrheal illness have occurred (4). It is generally accepted that patients acquire disease-causing strains of C. difficile from exogenous sources during hospitalization (10). However, epidemiologic investigations and infection control are hampered by the inadequacy of the typing methods used to identify causative strains. Currently, a wide variety of genomic typing methods (pulsed-field gel electrophoresis [PFGE], PCR, arbitrarily primed PCR [AP-PCR], restriction enzyme analysis [REA], plasmid profile analysis, and ribotyping) and nongenomic typing methods (protein profile analysis [PP], immunoblotting, bacteriocin production analysis, and bacteriophage sensitivity testing) are used (3, 22), but there have been few comparisons of methods in actual outbreak investigations. Identification and typing are infrequently performed because of the complexity, cost, and lack of comparability of the many typing methods used (22). Of course, failure to use adequate and reproducible typing methods precludes identification of C. difficile strains with enhanced virulence (22) or the ability to persist in the hospital environment, or both. In addition, the inability to recognize prevalent or persistent strains prevents infection control personnel from making rational choices in dealing with outbreaks or hospital-to-hospital spread (2). The recent identification of a C. difficile strain(s) which appears to be outbreak associated in geographically diverse hospitals (22) makes use of reproducible typing techniques an even more important public health consideration.
We investigated strains of C. difficile causing a prolonged outbreak of diarrheal disease in our hospital (21). C. difficile isolates collected from symptomatic, hospitalized patients in 1990 and 1992 were typed by REA, AP-PCR, and PP. We performed this study in order to determine the strain prevalence in our hospital, to determine the frequencies of relapse and reinfection, and to compare the utility of and concordance of results among the three typing methods.
(This work was presented in part at the 34th Interscience Conference on Antimicrobial Agents and Chemotherapy, 4 to 7 October 1994, Orlando, Fla. [1].)
MATERIALS AND METHODS
C. difficile culture acquisition and patient population.
A total of 166 stool samples obtained from 102 patients in the Stratton Veterans Affairs (VA) Medical Center in Albany, N.Y., and submitted to the clinical microbiology laboratory there for C. difficile toxin testing from 14 January 1992 through 12 November 1992 were found to be toxin positive by the MRC-5 cell cytotoxicity assay (11). These stool samples were frozen at −20°C. The frozen stools were later thawed and cultured for the presence of C. difficile at the Wadsworth Center, New York State Department of Health, Albany, N.Y. Following phenotypic characterization, all isolates were lyophilized and stored in anticipation of typing. Subsequently, 45 isolates (obtained from 33 patients) which had been isolated and lyophilized during a 1990 investigation (March to October 1990) of an increase in the incidence of C. difficile infection in our facility were studied in the same fashion. Lyophilized samples of isolates from both 1990 and 1992 were sent to the Nosocomial Pathogens Laboratory Branch, Centers for Disease Control and Prevention (CDC), Atlanta, Ga., and the Department of Anaerobic Microbiology, Virginia Polytechnic Institute and State University (VPI), Blacksburg, for typing. Relevant information about the patient population from which these samples were obtained is presented in Table 1.
TABLE 1.
Demographic data for patients with C. difficile disease
| Characteristic | March– November 1990 | January– November 1992 |
|---|---|---|
| No. of patients | 33 | 102 |
| Males (%) | 97 | 97 |
| Females (%) | 3 | 3 |
| Age range (yr) | 23–86 | 28–88 |
| Mean age (yr) | 67 | 71 |
| No. (%) of patients at the following location: | ||
| Medicine | 12 (36)a | 48 (47)a |
| Surgery | 10 (30) | 21 (21) |
| Long-term care | 9 (27) | 26 (25) |
| Critical care | 6 (18) | 7 (7) |
| Outpatient | 0 (0) | 9 (9) |
| No. (%) of patients with the following source of infection: | ||
| Nosocomial | 33 (100) | 93 (91) |
| Community acquired | 0 (0) | 9 (9) |
| No. of patients treated with the following: | ||
| Vancomycin | 20b | 83b |
| Metronidazole | 5 | 48 |
| None | 2 | 24 |
| Unknown | 6 | 0 |
| Length (days) (range [mean]) of hospital stay | 8–511 (110) | 3–1,063 (127) |
| Acute care | 77 | 66 |
| Long-term care | 201 | 266 |
| No. (%) of patients with time in ICUc | 15 (45) | 38 (37) |
| Length (days) of ICU stay (range [mean]) | 1–45 (12) | 1–83 (11) |
| No. of patients who died during the following yr postinfection: | ||
| First yr | 10 | 33 |
| Second yr | 7 | 12 |
| Third yr | 1 | 5 |
| No. (%) of patients deceased within 3 yr of C. difficile diarrhea | 18 (55) | 50 (50) |
Percentages add to more than 100 because some patients were on more than one hospital unit.
Patient numbers add to more than 33 (1990) and 102 (1992) because some patients were treated with more than one antibiotic.
ICU, intensive care unit.
Isolation, identification, and characterization of C. difficile.
Stool specimens were cultured on cefoxitin-cycloserine-fructose agar (CCFA) (12). CCFA plates were examined daily for 4 days before being discarded. Specimens that were negative for the presence of C. difficile at the end of 4 days were shocked with alcohol and replated. Alcohol shock treatment consisted of mixing 1 ml of stool with 1 ml of absolute ethanol and agitating on a vortex mixer (9). The specimen was incubated at room temperature for 1 h and mixed on a vortex mixer at 15-min intervals. At the end of 1 h, the alcohol-stool mixture was placed in a reduced, cooked meat broth and incubated under anaerobic conditions for 48 h. This broth was then streaked onto a CCFA plate and examined for C. difficile-like colonies.
C. difficile-like colonies, whether they were from a CCFA plate of the original stool specimens or from cooked meat broth, were subcultured in reduced thioglycolate broth. After a 24-h incubation, this broth was used to inoculate media for conventional biochemical tests. These tests included tests for the fermentation of glucose, maltose, mannitol, lactose, sucrose, xylose, salicin, arabinose, glycerol, rhamnose, and trehalose. Isolates were tested for the production of catalase, indole, urease, hydrogen sulfide, lecithinase, and lipase. The isolates were also tested for motility, gelatin liquefaction, nitrate reduction, esculin hydrolysis, growth in bile, and the litmus milk reaction. Gas chromatography was used to determine the volatile short-chain fatty acids produced by the growth of the organisms (14).
Toxin production was tested at VPI and TechLab (Blacksburg, Va.) by a CHO K-1 cell cytotoxicity test following recovery of the C. difficile isolates from lyophilization. C. difficile toxin was considered present if ≥50% of the CHO cells exhibited cytotoxic effects upon exposure to C. difficile culture filtrate for 24 to 48 h. Production of specific C. difficile toxins was not evaluated.
REA typing.
REA typing was performed at the Wadsworth Center. C. difficile isolates were grown anaerobically in 5 ml of reduced brain heart infusion broth overnight at 37°C. DNA was extracted as described by Bowman et al. (1a). Cells were harvested by centrifugation, resuspended in 250 μl of TES (10 mM Tris, 1.0 M NaCl, 0.1 M EDTA) to which lysozyme had been added to a final concentration of 5 mg/ml, and incubated at 37°C for 1 h. Lysis was accomplished by the addition of 11 μl of 25% sodium dodecyl sulfate and incubation at 60°C for 10 min. Proteinase K (Boehringer Mannheim Biochemicals, Indianapolis, Ind.) was added to a final concentration of 0.5 mg/ml, followed by incubation at 55°C for 2 h. NaCl was added to a final concentration of 1 M. The DNA was extracted twice with phenol-chloroform-isoamyl alcohol (25:24:1) and once with chloroform and was precipitated by the addition of 2 volumes of cold ethanol. The DNA was pelleted by centrifugation and was resuspended in 200 μl of TE (10 mM Tris-HCl, 1 mM EDTA). DNA (approximately 1 μg) was restricted with 20 U of HindIII (New England Biolabs, Beverly, Mass.) according to the manufacturer’s instructions for 5 h at 37°C. The fragments were separated by electrophoresis in 0.7% LE (low electroendosmosis) agarose (FMC Bioproducts, Rockland, Maine) in 0.5× TBE (45 mM Tris, 45 mM boric acid, 1.0 mM EDTA [pH 8.0]) for 16 h at 2.5 V/cm with a model H1 (Bethesda Research Laboratories, Gaithersburg, Md.) apparatus. HindIII fragments of bacteriophage λ were used as size markers. Gels were stained in ethidium bromide and photographed under UV illumination.
Isolates whose HindIII-restricted DNA yielded indistinguishable banding patterns involving bands of 4.4 kb or larger (Fig. 1) were considered to be members of the same REA type. A new REA type was assigned to each band pattern for which there were reproducible differences in major bands. Subtypes were assigned on the basis of reproducible differences in faint bands. In order to ensure that band patterns (and therefore types) were reproducible, numerous randomly selected isolates were retyped.
FIG. 1.
REA banding patterns of C. difficile isolates. Molecular size markers are indicated on the left. Lanes: 1 to 2, type 7; 3, type 12a; 4, type 12; 5 to 7, 13, and 14, type 4 (marked with ▾); 8 to 9, type 17; 10 to 12, type 5.
AP-PCR typing.
AP-PCR typing was performed at CDC. One or two colonies from an anaerobic blood agar plate that had been incubated overnight at 37°C were suspended in 500 μl of TE buffer and heated in a boiling water bath for 5 min. Samples were cooled on ice and were used immediately or were stored at −20°C (16). The PCR conditions used were those of Williams et al. (26), with slight modification of some reactant concentrations. Amplification was conducted in a volume of 25 μl containing 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 2 mM MgCl2, 0.001% gelatin, dATP, dGTP, dCTP, and dTTP each at a concentration of 100 μM, 2 μM primer (primer ARB 11 [5′-CTAGGACCGC-3′]), 2 U of Taq polymerase, and 5 μl of sample. Samples were amplified on a Perkin-Elmer Cetus model 480 thermocycler for 45 cycles at 95°C for 1 min, 36°C for 1 min, and 72°C for 2 min.
Fifteen microliters of PCR product mixed with 3 μl of 6× loading buffer was applied to a 1.5% agarose gel in 0.5× Tris-borate-EDTA buffer cooled to 18°C. The gel was electrophoresed at 200 V until the bromophenol blue dye marker had migrated 20 cm. The gel was stained with ethidium bromide and photographed. The banding patterns of all lanes were visually compared, and isolates showing identical patterns or variations only in very faint bands were considered the same type (16). Because the reproducibility of the banding patterns can sometimes be problematic with AP-PCR typing (7, 18, 22, 24), the isolates from 1992 were all retyped along with the isolates from 1990. Types were assigned on the basis of these data. A representative example of the data obtained by AP-PCR is shown in Fig. 2.
FIG. 2.
AP-PCR banding patterns of C. difficile isolates. Lanes (unnumbered from left to right, respectively): 1, φX174 molecular mass markers; 2 to 4, type 05; 5 to 8 and 10 to 16, type 01; 9, type 06; 17 and 19, type 07; 18, type 08; 20, 1-kb molecular size marker. BP, base pairs.
PP typing.
PPs were performed at VPI and TechLab. Mini-dialysis tubes, a modified form of the brain heart infusion dialysis flasks that are used routinely for the large-scale production of clostridial toxins, were used in the study. The small-scale version allows a large number of strains to be screened in an economical fashion. The use of this type of device results in the production of high levels of toxin and other proteins that can easily be visualized by electrophoretic analysis.
One drop of overnight (18-h) brain heart infusion broth cultures of isolates was used to inoculate brain heart infusion mini-dialysis tubes. The mini-dialysis tubes were incubated at 37°C for 72 h, and the supernatant was collected by centrifugation (10,000 × g for 15 min at 4°C). The supernatants were filtered through 0.45-μm-pore-size membranes and stored at 4°C.
Samples (20 μl) of culture filtrates were mixed with glycerol-bromphenol blue tracking dye (5 μl) and were loaded onto 3 to 27% polyacrylamide gradient gels (Jules, Inc., New Haven, Conn.). Samples were electrophoresed at a constant current of 50 mA for 4 h at 10°C. The gels were stained with Coomassie blue R-250 and were destained for visualization of the protein banding patterns. The protein banding patterns determined with culture filtrates of the isolates were compared, and isolates displaying identical patterns were grouped together (Fig. 3). Very faint bands were not considered in assigning types. Representative isolates of each group were analyzed a second time to ensure that the protein profiles were reproducible.
FIG. 3.
PP banding patterns of C. difficile isolates. Lanes: 1 to 4, type I; 5, type V; 6, type II.
Calculation of concordance.
Agreement between typing methods was evaluated by calculating concordance, the percent agreement between the methods. In each comparison, the method yielding the least diversity was used as the reference. For example, since there are 11 PP types and 35 REA types and subtypes, PP was the reference method in the comparison of PP and REA. All reference types for which there was more than one isolate were compared.
RESULTS
Lyophilized specimens were sent from the Wadsworth Center to CDC and VPI for typing. One of 45 isolates from 1990 could not be regrown at either institution. Consequently, AP-PCR and PP typing were done with 44 isolates from 1990 and 166 isolates from 1992, for a total of 210 isolates. However, REA typing was done with all 211 isolates. All typing was done totally blinded. The investigators had access only to the accession numbers for the isolates until after the analyses were complete. Table 2 presents the numbers of C. difficile isolates of each REA, AP-PCR, and PP type from both 1990 and 1992.
TABLE 2.
Numbers and types of C. difficile isolates by REA, AP-PCR, and PP
| REA
|
AP-PCR
|
PP
|
||||||
|---|---|---|---|---|---|---|---|---|
| Type | No. (%) of isolates
|
Type | No. (%) of isolates
|
Type | No. (%) of isolates
|
|||
| 1990 | 1992 | 1990 | 1992 | 1990 | 1992 | |||
| 1 | 2 | 01 | 28 (64) | 124 (75) | I | 31 (70) | 139 (84) | |
| 2 | 1 | 02 | 1 | 1 | II | 1 | 2 | |
| 3 | 4 | 03 | 1 | III | 10 | |||
| 4 | 24 (53) | 126 (76) | 04 | 1 | IV | 5 | ||
| 4a | 2 | 05 | 3 | 4 | V | 1 | ||
| 4b | 1 | 06 | 3 | 1 | VI | 1 | ||
| 4c | 1 | 07 | 1 | 2 | VII | 4 | ||
| 5 | 3 | 08 | 1 | 1 | VIII | 3 | 2 | |
| 6 | 1 | 09 | 1 | 3 | IX | 2 | ||
| 7 | 4 | 10 | 3 | X | 6 | |||
| 8 | 1 | 11 | 1 | XI | 3 | |||
| 9 | 1 | 12 | 1 | |||||
| 10 | 1 | 2 | 13 | 2 | ||||
| 11 | 1 | 14 | 1 | |||||
| 12 | 1 | 15 | 1 | |||||
| 12a | 2 | 16 | 2 | |||||
| 12b | 1 | 17 | 1 | |||||
| 12c | 1 | 18 | 1 | |||||
| 13 | 1 | 19 | 1 | |||||
| 14 | 1 | 20 | 2 | |||||
| 15 | 1 | 21 | 1 | |||||
| 17 | 5 | 22 | 1 | |||||
| 18 | 1 | 23 | 1 | |||||
| 19 | 1 | 24 | 1 | |||||
| 20 | 1 | 3 | 25 | 1 | ||||
| 21 | 5 | 26 | 1 | |||||
| 22 | 27 | 1 | ||||||
| 23 | 1 | 28 | 1 | |||||
| 24 | 1 | 29 | 1 | |||||
| 25 | 1 | 30 | 1 | |||||
| 26 | 2 | 31 | 1 | |||||
| 27 | 32 | 1 | ||||||
| 28 | 2 | 33 | 1 | |||||
| 29 | 1 | 34 | 1 | |||||
| 30 | 1 | 35 | 1 | |||||
| 31 | 1 | 36 | 1 | |||||
| 32 | 3 | 37 | 1 | |||||
| 38 | 1 | |||||||
| Total | 45 | 166 | 44 | 166 | 44 | 166 | ||
REA.
The banding patterns produced when randomly selected isolates were retyped by REA were compared to those obtained when the same isolates were first typed. No discernible differences were found. A total of 13 types and one subtype were identified among the isolates from 1990. Twenty-four (53%) of the 45 isolates from 1990 were type 4. Eighteen types and five subtypes were identified among the 166 isolates from 1992, of which 126 (76%) were type 4. Members of only two REA types (6%) were found in both 1990 and 1992.
AP-PCR.
AP-PCR typing was repeated with all isolates from 1992 when the isolates from 1990 were typed. This was done in order to avoid difficulties in interpretation of the banding patterns arising due to slight variations in minor bands. The final type assignments were based on the data obtained when all 210 available isolates were typed together. Thirteen AP-PCR types were identified among the 44 isolates from 1990. Twenty-eight (64%) of these were members of type 01. Thirty-two AP-PCR types were represented among the 166 isolates from 1992. Of these, 124 (75%) were type 01. Members of only 7 of the 38 AP-PCR types (18%) were found in both 1990 and 1992.
PP.
When representative isolates of each PP type were retyped, the results were consistent with those found originally. Data for all isolates from 1990 and 1992 were evaluated, and a typing scheme that included all isolates was devised. A total of 11 PP types, designated I through XI, were identified. Representatives of five of these types were identified among the isolates from 1990. Thirty-one (71%) of the 44 available isolates from 1990 were members of type I. Members of nine types were present among the 166 isolates from 1992, 138 (84%) of which were type I. Members of 3 of the 11 types (27%) were found among the isolates from both 1990 and 1992.
Comparison of typing methods.
Each of the 210 C. difficile isolates available to all three laboratories could be typed by every method. However, the discriminatory powers of the three methods were different. PP yielded only 11 types, while there were 29 REA types (and six subtypes) and 38 AP-PCR types. Although the discriminatory powers of the methods differed, each grouped together the majority of the isolates of the dominant type. These isolates were designated 4-01-I, indicating REA type 4, AP-PCR type 01, and PP type I, respectively. The dendrogram in Fig. 4 shows the relationships between the types identified by the three methods.
FIG. 4.
Dendrogram comparing typing methods. ·····, strains from 1990; ——, strains from 1992; ——————, endemic strain (1990 and 1992). The line connecting PP type I to REA type 4 represents 168 of the 210 isolates typed by PP and REA. The other 41 lines connecting PP and REA types represent the remaining 42 isolates. The line connecting REA type 4 to AP-PCR type 01 represents 146 of the 210 isolates typed by REA and AP-PCR. The other 48 lines connecting REA and AP-PCR types represent the remaining 64 isolates.
As a part of this study, 45 lyophilized isolates from 33 patients obtained prospectively during a 1990 C. difficile outbreak investigation at the Stratton VA Medical Center were examined retrospectively. Analysis of the typing data for these isolates revealed that a dominant organism, type 4-01-I, was present. Among the isolates from 1990, 53% were REA type 4, 64% were AP-PCR type 01, and 70% were PP type I. Twenty-two of the 44 isolates typed by every method (50%) were identified as type 4-01-I, and there was concordance between the three methods for 24 of 43 isolates (56%) of PP types represented by more than one isolate. Complete concordance data are found in Table 3.
TABLE 3.
Percent concordance of C. difficile typing methods
| Typing method | % Concordance for strains of the indicated type isolated in the following yr:
|
||||||||
|---|---|---|---|---|---|---|---|---|---|
| 1990
|
1992
|
1990 and 1992
|
|||||||
| 4-01-Ia | Other | Total | 4-01-I | Other | Total | 4-01-I | Other | Total | |
| PP and REA | 74 | 33 | 63 | 91 | 60 | 86 | 88 | 51 | 81 |
| PP and AP-PCR | 81 | 17 | 63 | 89 | 36 | 81 | 88 | 30 | 77 |
| REA and AP-PCR | 96 | 75 | 86 | 98 | 52 | 90 | 97 | 56 | 89 |
| PP, REA, AP-PCR | 71 | 18 | 56 | 88 | 24 | 79 | 85 | 22 | 74 |
Strains were identified as type 4 by REA, type 01 by AP-PCR, and type I by PP.
Analysis of the typing data for 166 isolates obtained from specimens collected prospectively for this study from 102 patients during 1992 revealed that the same dominant strain (type 4-01-I) identified among the isolates from 1990 was still present. By REA typing, 126 of 166 isolates (76%) were type 4; by AP-PCR typing, 124 of 166 isolates were (75%) type 01; and by PP typing, 139 of 166 isolates (84%) were type I. There was substantial agreement between the three methods in the typing of members of the dominant strain, with 123 of 166 isolates (74%) being identified as type 4-01-I. The continued dominance of this strain over 2 years suggested that it was endemic. Concordance among all three methods was found for 129 of the 164 isolates (79%) of PP types represented by more than one isolate.
Type changes among serial isolates from individual patients.
C. difficile isolates of different types were obtained from serial specimens acquired from 11 patients. Sequential isolates in which changes in type (determined by at least two methods) had occurred were obtained from five of these patients. From one patient a type 2-02-IV strain was isolated initially and a type 4-01-I strain was isolated 34 days later. From a second patient a type 19-14-IV strain was isolated initially and a type 4-01-I strain was isolated 81 days later. From a third patient a type 4-01-I strain was isolated initially, followed by isolation of a type 12-12-III strain 27 days later and a type 12a-11-I strain 56 days after that. The endemic strain type, type 4-01-I, was never isolated from two of these five patients. From one of these two patients, a type 3-05-VII strain was isolated first, followed by isolation of a type 6-03-I strain. From the other patient a type 7-07-I strain was isolated initially, followed by isolation of a type 7-08-II strain.
Toxin testing.
A total of 210 C. difficile isolates, 45 from 1990 and 166 from 1992, were tested for C. difficile toxin production after shipment and recovery from lyophilization (210 of 211 isolates were successfully regrown from lyophilized stocks). All were toxin positive.
DISCUSSION
Typing methods designed to distinguish between bacterial strains of the same species are often used in epidemiologic investigations of nosocomial infections (19, 20). Several such methods, both phenotypic and genotypic, have been used to distinguish strains of C. difficile (2, 22, 27). However, we have found no published reports of studies in which several molecular typing techniques for C. difficile have been used to study an outbreak over an extended period of time.
It is important that each of the 210 strains available for typing in all three laboratories could be typed by every method. Some other methods, such as plasmid profile analysis and PFGE, have occasionally proven to be inadequate because too many strains are untypeable or because data are too difficult to interpret (6, 8, 15, 22, 23, 25, 29). In fact, the endemic strain identified in this study by REA, AP-PCR, and PP is untypeable by PFGE (22). Use of such methods can result in the inability to acquire useful epidemiologic data, since information about isolates critical to the investigation may be unavailable. All three typing methods used in this study revealed the presence of a dominant strain in the outbreak under investigation. Furthermore, there was 85% concordance between the three methods used to identify this strain. That is, 145 of the 170 isolates identified as type I by PP were identified as type 4 by REA and type 01 by AP-PCR.
Each typing technique used in this study has advantages and disadvantages, including the ease and rapidity of the procedure, its discriminatory power, and the reproducibility of the results. Of the three methods, REA is arguably the most technically demanding and time-consuming. Because so many bands are produced by REA typing, it is difficult to be sure that the DNA has been completely digested. With problematic strains, it may be necessary to titrate the DNA against a fixed amount of enzyme to be sure that complete digestion has occurred. If the same band pattern is seen as the DNA concentration decreases but the bands simply get lighter, one can be confident that digestion is complete. However, this process is very time-consuming, and therefore, it is not practical for routine use. Furthermore, the presence of hundreds of bands makes REA patterns difficult to interpret. Although various methods have been devised for analysis of C. difficile REA data (5, 17, 24, 27, 28), all require a great deal of skill and are tedious and time-consuming. Compared to REA, PP and AP-PCR require relatively little hands-on time. However, PP requires about the same total amount of time as REA. Of the three methods, results can be obtained most rapidly by AP-PCR. Interpretation of AP-PCR and PP results is generally uncomplicated, because fewer numbers of bands are produced by AP-PCR and PP and they are more discrete than those produced by REA. Since REA typing is technically more demanding than the other methods, requires a larger investment in bench time, and yields results that are more difficult to interpret, we do not recommend it for routine use in epidemiologic investigations except by those experienced in C. difficile REA typing and in situations in which time is not critical.
In this study REA and AP-PCR were clearly more discriminatory than PP. Examination of Fig. 4 shows that there is significant divergence of PP types when they are compared to either REA or AP-PCR types but that there is relatively little divergence or convergence between REA and AP-PCR types. For example, there is 63% concordance between PP and REA for the isolates from 1990, while there is 86% concordance between REA and AP-PCR for the same isolates (Table 3). Comparison of concordance for nonendemic isolates is especially informative, since inclusion of the endemic isolates, which constitute the vast majority of the total, tends to obscure the degree of agreement between the methods for other types. For nonendemic isolates from 1990 the rates of concordances between PP and REA and between PP and AP-PCR are only 33 and 17%, respectively, while there is 75% concordance between REA and AP-PCR (Table 3). The greater overall level of concordance between REA and AP-PCR may in part be because REA and AP-PCR are both genotypic methods, while PP is a phenotypic method. In any case, all three methods successfully identified the endemic strain, demonstrating that although typing by PP lacks the discriminatory power of typing by REA and AP-PCR, it is still a useful epidemiologic tool. Whether or not the differences in discriminatory power revealed in this study are an issue depends upon the nature of the outbreak under investigation. If the isolates examined all fall into one or just a few types by PP typing, it would be advisable to retype the organisms by REA, AP-PCR, or some other method more discriminatory than PP. However, if PP provides enough discrimination to allow useful epidemiologic information to be obtained, no further analysis may be required. As far as their discriminatory powers, either REA or AP-PCR would probably be useful as the sole typing method in most investigations. In this outbreak investigation, both methods grouped the endemic strain successfully but also revealed strain divergence that was not seen by PP typing.
With regard to reproducibility, difficulties can be encountered with each of these methods. AP-PCR is probably the most problematic, since type assignments depend in part on light or minor bands that are not always reproducible (7, 16, 22, 24). In this study, the AP-PCR patterns of some isolates from the first group were not reproducible when those isolates were retyped during the initial typing of the second group. Therefore, all 210 available isolates were typed together and types were assigned on the basis of these data. We recommend that with AP-PCR typing of C. difficile, all isolates under investigation be typed at one time. In this way, experimental variables can be minimized.
Since it may not be feasible to use more than one typing technique for routine investigation of outbreaks in clinical settings, the selection and consistent use of one typing technique is advisable, with the understanding that if that technique is PP, it may occasionally be necessary to supplement it with REA, AP-PCR, or some other more discriminatory method. Unlike for many other bacteria, no single typing method is clearly superior for C. difficile typing.
The presence of a dominant strain in 1990 (type 4-01-I), as well as its persistence and markedly increased prevalence in 1992, requires examination, especially since this strain has recently been identified in outbreaks in two other parts of the United States (22). There are two obvious explanations for the dominance and persistence of this strain. First, it may have a selective advantage in vivo. Interestingly, however, for isolates from the five patients from whom serial isolates were obtained and whose sequential isolates changed type by at least two of the three typing methods, the strain type changed from that of the endemic strain for one patient and the strain type changed to that of the endemic strain for two patients. The endemic strain was never isolated from the other two patients from whom serial C. difficile isolates with different types were obtained. These results do not indicate that type 4-01-I has an in vivo advantage over other C. difficile strains. Another possibility is that type 4-01-I C. difficile has a survivability advantage in the hospital environment. However, since no environmental sampling was done as part of this study, that hypothesis cannot yet be addressed. Certainly, the persistence and increased prevalence of this strain over 2 years in the Stratton VA Medical Center, as well as its appearance in other parts of the country, indicate that it warrants further investigation.
Among the cases of C. difficile infection investigated in 1990, none was community acquired. In contrast, 11 patients (18 isolates) were identified as having community-acquired C. difficile infections in 1992. Two of these 18 isolates were the endemic type (type 4-01-I). It is not clear what the source of C. difficile infection was in the patients from whom these isolates were acquired. The patients had not previously been admitted to the Stratton VA Medical Center, but they may have been patients in other area hospitals. To our knowledge, C. difficile isolates obtained from patients in other area hospitals have not been typed; thus, the possible involvement of these institutions as reservoirs for C. difficile type 4-01-I cannot currently be assessed.
The death rate for the patients in this study within 3 years after the detection of C. difficile infection was 55 for the patients from 1990 and 50% for the patients from 1992. The greatest number of deaths occurred within the first year after the organism was detected. The poor prognosis for these patients may correlate with the seriousness of their primary diseases, with C. difficile infection, or with both factors. In any case, it is clear that the development of diarrhea caused by C. difficile may be correlated with a poor prognosis for the patient.
In conclusion, our study resulted in the detection of an endemic C. difficile strain, type 4-01-I, which remained endemic during the 3 years of the study. This strain was detected equally well by REA, AP-PCR, and PP. Among the 210 isolates typed by all three methods, AP-PCR detected the largest number of types (n = 38 types), followed by REA (n = 29 types and 6 subtypes) and PP (n = 11 types). In comparing typing methods, the greatest concordance was seen with REA and AP-PCR. The endemic strain, C. difficile type 4-01-I, was dominant in 1990 and became even more dominant in 1992. Consequently, changes in isolate type were observed for isolates from only 5 of the 47 patients from whom multiple isolates were obtained. Because the endemic strain was so dominant, representing 69% of all 210 isolates typed by all three methods, it was not possible to distinguish reinfection from recurrence in infected patients. The fact that approximately half of these patients died within 3 years after C. difficile was first detected in their stools emphasizes their vulnerability and, therefore, the importance of using molecular typing methods to acquire the information needed to develop strategies for the prevention of the nosocomial spread of this organism.
ACKNOWLEDGMENT
We thank Laurie Neville for expert assistance in the performance of PP and cytotoxin analysis.
REFERENCES
- 1.Baltch A L, Rafferty M E, Smith R P, Tenover F, Killgore G, Wilkins T, Lyerly D M, Schoonmaker D, Hannett G, Shayegani M. Program and abstracts of the 34th Interscience Conference on Antimicrobial Agents and Chemotherapy. Washington, D.C: American Society for Microbiology; 1994. A comparison of three methods for typing of Clostridium difficile isolates during an outbreak of diarrhea in a general hospital, abstr. J32; p. 58. [Google Scholar]
- 1a.Bowman R A, O’Neill G L, Riley T V. Non-radioactive restriction fragment length polymorphism (RFLP) typing of Clostridium difficile. FEMS Microbiol Lett. 1991;79:269–272. doi: 10.1016/0378-1097(91)90097-t. [DOI] [PubMed] [Google Scholar]
- 2.Brazier, J. S., M. E. Mulligan, M. Delmee, S. Tabaqchali, and the International Clostridium difficile Study Group. 1997. Preliminary findings of the International Typing Study on Clostridium difficile. Clin. Infect. Dis. 25(Suppl. 2):S199–S201. [DOI] [PubMed]
- 3.Brazier, J. S. 1995. An international study on the unification of nomenclature for typing Clostridium difficile. Clin. Infect. Dis. 20(Suppl. 2):S325–S326. [DOI] [PubMed]
- 4.Cartmill T D I, Shrimpton S B, Panigrahi H, Khanna V, Brown R, Poxton I R. Nosocomial diarrhoea due to a single strain of Clostridium difficile: a prolonged outbreak in elderly patients. Age Aging. 1992;21:245–249. doi: 10.1093/ageing/21.4.245. [DOI] [PubMed] [Google Scholar]
- 5.Clabots C R, Johnson S, Bettin K M, Mathie P A, Mulligan M E, Schaberg D R, Peterson L R, Gerding D N. Development of a rapid and efficient restriction endonuclease analysis typing system for Clostridium difficile and correlation with other typing systems. J Clin Microbiol. 1993;21:1870–1875. doi: 10.1128/jcm.31.7.1870-1875.1993. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Clabots C R, Peterson L R, Gerding D N. Characterization of a nosocomial Clostridium difficile outbreak by using plasmid profile typing and clindamycin susceptibility testing. J Infect Dis. 1988;158:731–736. doi: 10.1093/infdis/158.4.731. [DOI] [PubMed] [Google Scholar]
- 7.Collier M C, Stock F, DeGirolami P C, Samore M H, Cartwright C P. Comparison of PCR-based approaches to molecular epidemiologic analysis of Clostridium difficile. J Clin Microbiol. 1996;34:1153–1157. doi: 10.1128/jcm.34.5.1153-1157.1996. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Devlin H R, Au W, Foux L, Bradbury W C. Restriction endonuclease analysis of nosocomial isolates of Clostridium difficile. J Clin Microbiol. 1987;25:2168–2172. doi: 10.1128/jcm.25.11.2168-2172.1987. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Dowell V R, Jr, Hawkins T M. DHEW publication no. (CDC) 74-8272. Atlanta, Ga: Center for Disease Control; 1974. Laboratory methods in anaerobic bacteriology; pp. 1–96. [Google Scholar]
- 10.Fekety R, Kim K-H, Brown D, Batts D H, Cudmore M, Silva J. Epidemiology of antibiotic-associated colitis. Am J Med. 1981;70:906–908. doi: 10.1016/0002-9343(81)90553-2. [DOI] [PubMed] [Google Scholar]
- 11.Florin I, Thelestam M. Intoxication of cultured human lung fibroblasts with Clostridium difficile toxin. Infect Immun. 1981;33:67–74. doi: 10.1128/iai.33.1.67-74.1981. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.George W L, Sutter V L, Citron D, Finegold S. Selective and differential medium for isolation of Clostridium difficile. J Clin Microbiol. 1979;9:214–219. doi: 10.1128/jcm.9.2.214-219.1979. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Gerding D N. Disease associated with Clostridium difficile infection. Ann Intern Med. 1989;110:255–256. doi: 10.7326/0003-4819-110-4-255. [DOI] [PubMed] [Google Scholar]
- 14.Holdeman L V, Cato E P, Moore W E C. Anaerobe Laboratory Manual. 4th ed. Blacksburg: Virginia Polytechnic Institute and State University; 1977. Gas chromatography: an identification of the organisms; pp. 134–137. [Google Scholar]
- 15.Kato H, Kato N, Watanabe K, Ueno K, Ushijima H, Hashira S, Abe T. Application of typing by pulsed-field gel electrophoresis to the study of Clostridium difficile in a neonatal intensive care unit. J Clin Microbiol. 1994;32:2067–2070. doi: 10.1128/jcm.32.9.2067-2070.1994. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Killgore G E, Kato H. Use of arbitrary primer PCR to type Clostridium difficile and comparison of results with those by immunoblot typing. J Clin Microbiol. 1994;32:1591–1593. doi: 10.1128/jcm.32.6.1591-1593.1994. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Kristjansson M, Samore M H, Gerding D N, DeGirolami P C, Bettin K M, Karchmer A W, Arbeit R D. Comparison of restriction endonuclease analysis, ribotyping, and pulsed-field gel electrophoresis for molecular differentiation of Clostridium difficile strains. J Clin Microbiol. 1994;32:1963–1969. doi: 10.1128/jcm.32.8.1963-1969.1994. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Martirosian G, Kuipers S, Verbrugh H, Van Belkum A, Meisel-Mikolajczyk F. PCR ribotyping and arbitrarily primed PCR for typing strains of Clostridium difficile from a Polish maternity hospital. J Clin Microbiol. 1995;33:2016–2021. doi: 10.1128/jcm.33.8.2016-2021.1995. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Maslow J N, Mulligan M E, Arbeit R D. Molecular epidemiology: the application of contemporary techniques to typing bacteria. Clin Infect Dis. 1993;17:153–164. doi: 10.1093/clinids/17.2.153. [DOI] [PubMed] [Google Scholar]
- 20.Mazurek G. Modern typing methods in the investigation of nosocomial infections. Curr Opin Infect Dis. 1993;6:538–543. [Google Scholar]
- 21.Nelson D, Auerbach E, Baltch S B, Desjardin A L, Beck-Sague E, C, Rheal C, Smith R P, Jarvis W R. Epidemic Clostridium difficile-associated diarrhea: role of second- and third-generation cephalosporins. Infect Control Hosp Epidemiol. 1994;15:88–94. doi: 10.1086/646867. [DOI] [PubMed] [Google Scholar]
- 22.Samore M, Killgore G, Johnson S, Goodman R, Shim J, Venkataraman L, Sambol S, DeGirolami P, Tenover F, Arbeit R, Gerding D. Multicenter typing comparison of sporadic and outbreak Clostridium difficile isolates from geographically diverse hospitals. J Infect Dis. 1997;176:1233–1238. doi: 10.1086/514117. [DOI] [PubMed] [Google Scholar]
- 23.Samore M H, Venkataraman L, DeGirolami P C, Arbeit R D, Karchmer A W. Clinical and molecular epidemiology of sporadic and clustered cases of nosocomial Clostridium difficile diarrhea. Am J Med. 1996;100:32–39. doi: 10.1016/s0002-9343(96)90008-x. [DOI] [PubMed] [Google Scholar]
- 24.Tang Y J, Houston S T, Gumerlock P H, Mulligan M E, Gerding D N, Johnson S, Fekety F R, Silva J., Jr Comparison of arbitrarily primed PCR with restriction endonuclease and immunoblot analyses for typing Clostridium difficile isolates. J Clin Microbiol. 1995;33:3169–3173. doi: 10.1128/jcm.33.12.3169-3173.1995. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.van Dijck P, Avesani V, Delmee M. Genotyping of outbreak-related and sporadic isolates of Clostridium difficile belonging to serogroup C. J Clin Microbiol. 1996;34:3049–3055. doi: 10.1128/jcm.34.12.3049-3055.1996. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Williams J G K, Kubelik A R, Livak K J, Rafalski J A, Tingey S V. DNA polymorphisms amplified by arbitrary primers are useful as genetic markers. Nucleic Acids Res. 1990;18:6531–6535. doi: 10.1093/nar/18.22.6531. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Wolfhagen M J H M, Fluit A C, Torensma R, Jansze M, Kuypers A F A, Verhage E A E, Verhoef J. Comparison of typing methods for Clostridium difficile isolates. J Clin Microbiol. 1993;31:2208–2211. doi: 10.1128/jcm.31.8.2208-2211.1993. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Wren B W, Tabaqchali S. Restriction endonuclease DNA analysis of Clostridium difficile. J Clin Microbiol. 1987;25:2402–2404. doi: 10.1128/jcm.25.12.2402-2404.1987. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Wust J, Sullivan N M, Hardegger U, Wilkins T D. Investigation of an outbreak of antibiotic-associated colitis by various typing methods. J Clin Microbiol. 1982;16:1096–1101. doi: 10.1128/jcm.16.6.1096-1101.1982. [DOI] [PMC free article] [PubMed] [Google Scholar]