Abstract
Accurate strain typing is critical for understanding the changing epidemiology of Clostridium difficile infections. We typed 350 isolates of toxigenic C. difficile from 2008 to 2009 from seven laboratories in the United States and Canada. Typing was performed by PCR-ribotyping, pulsed-field gel electrophoresis (PFGE), and restriction endonuclease analysis (REA) of whole-cell DNA. The Cepheid Xpert C. difficile test for presumptive identification of 027/NAP1/BI isolates was also tested directly on original stool samples. Of 350 isolates, 244 (70%) were known PCR ribotypes, 224 (68%) were 1 of 8 common REA groups, and 187 (54%) were known PFGE types. Eighty-four isolates typed as 027, NAP1, and BI, and 83 of these were identified as presumptive 027/NAP1/BI by Xpert C. difficile. Eight additional isolates were called presumptive 027/NAP1/BI by Xpert C. difficile, of which three were ribotype 027. Five PCR ribotypes contained multiple REA groups, and three North American pulsed-field (NAP) profiles contained both multiple REA groups and PCR ribotypes. There was modest concordance of results among the three methods for C. difficile strains, including the J strain (ribotype 001 and PFGE NAP2), the toxin A-negative 017 strain (PFGE NAP9 and REA type CF), the 078 animal strain (PFGE NAP7 and REA type BK), and type 106 (PFGE NAP11 and REA type DH). PCR-ribotyping, REA, and PFGE provide different but overlapping patterns of strain clustering. Unlike the other methods, the Xpert C. difficile 027/NAP1/BI assay gave results directly from stool specimens, required only 45 min to complete, but was limited to detection of a single strain type.
INTRODUCTION
Clostridium difficile is a Gram-positive, spore-forming bacillus that causes a range of clinical syndromes ranging from mild to severe diarrhea, toxic megacolon, and, in some cases, sepsis and death (5, 21). C. difficile infections (CDI) continue to spread worldwide (4, 24, 27, 33). Some strains, such as the 027/NAP1/BI (where NAP is North American pulsed-field) “epidemic” strain (33), appear to show increased virulence, particularly in outbreak settings (3, 33, 34, 35). Accurate strain typing is critical for understanding the changing epidemiology of this organism and for determining outbreaks of infection in hospitals (3, 25, 38). However, clinical laboratories, particularly in the United States, have limited options for typing C. difficile isolates since virtually all typing methods require culturing of the stool sample to recover the isolate before typing can be performed, and cultures for C. difficile are performed rarely in the United States (16). Even when C. difficile culture methods are available in the laboratory, the organisms usually must be sent to a reference laboratory for typing, and results are often not available for days to weeks.
Multiple techniques have been used to study the epidemiology of C. difficile infections including pulsed-field gel electrophoresis (PFGE) (22), restriction endonuclease analysis (REA) of total DNA (7, 19), PCR-ribotyping, multilocus sequence typing (MLST), and multilocus variable-number tandem repeat assays (30, 46). Multilocus sequence typing is less discriminatory than the other methods and has been used primarily for population studies of C. difficile (30). Strain typing data are particularly valuable for investigations of hospital outbreaks but often are not available in real time to guide infection control efforts (4). Strain typing data that could be generated in parallel with the identification of the tcdB (toxin B gene), which is the basis for several PCR-based commercial C. difficile assays, could be of value to infection control efforts to reduce the spread of CDI in hospitals, as noted in two recent studies reported by Huang et al. (15) and Babady et al. (2).
The goals of this study were the following: (i) to compare the results of three C. difficile strain typing methods, i.e., PCR-ribotyping, REA, and PFGE, performed on toxigenic isolates available in pure culture, to determine how frequently the results were in agreement for the identification of common strains of C. difficile, such as the J strain (19), the toxin A-negative type 017 strain (38), the type 027/NAP1/BI hyperviruent/epidemic strain (26, 33), the type 078 “animal” strain (17, 39), and the type 106 United Kingdom epidemic strain (41); (ii) to determine the frequency of isolation of these strains at various study sites in the United States and Canada; and (iii) to assess the accuracy of the Xpert C. difficile assay for identifying 027/NAP1/BI strains directly in stool samples versus the strain types determined by the three typing methods on isolates obtained in pure culture.
MATERIALS AND METHODS
Bacterial isolates.
A total of 350 isolates of toxigenic C. difficile recovered from symptomatic patients from the eastern, midwestern, and western United States (including California, Illinois, Indiana, North Carolina, and Washington) and Canada (Quebec) were collected from November 2008 to January 2009, as previously described (43). Stool samples were collected, an aliquot was tested with the Xpert C. difficile assay on site (which included detection of tcdB [toxin B gene], cdtA [the binary toxin gene], and a single nucleotide deletion at base 117 in tcdC), and the remainder of the sample was transported to a central reference laboratory in anaerobic transport medium (Anaerobe Systems, Morgan Hill, CA). The samples were inoculated onto prereduced cycloserine-cefoxitin-fructose agar (CCFA) and into cycloserine-cefoxitin-mannitol broth with taurocholate-lysozyme-cysteine (CCMB-TAL) (both media were obtained from Anaerobe Systems, Morgan Hill, CA). The CCFA plate was incubated at 35°C to 37°C for 48 h; CCMB-TAL was incubated at 35°C to 37°C for 24 h and subcultured to CCFA for an additional 48 h. The presence of isobutyric, isocaproic, and isovaleric acids by gas-liquid chromatography as end products of glucose fermentation served as confirmation of C. difficile identification (18).
PCR-ribotyping.
PCR-ribotyping was performed as previously described by Stubbs et al. (40) with minor modifications (42). Analysis was performed using BioNumerics, version 5.1 (Applied Maths, Belgium). PCR-ribotyping patterns were compared to a database containing >3,000 clinical isolates including C. difficile reference strains obtained from the Culture Collection, University of Göteborg, Sweden, and from Ed Kuijper, Leiden University Medical Center, Netherlands (i.e., the Cardiff-European Centre for Disease Prevention and Control [ECDC] C. difficile collection).
PFGE.
For each isolate, DNA was prepared by in situ lysis of cells encased in agarose plugs and digested with SmaI, as described by Killgore et al. (22). Pulsed-field gel electrophoresis (PFGE) was performed using a Bio-Rad CHEF DR III System at 6 V/cm, 14°C, and 120° included angle, with switching from 5 to 15 s for 10 h, followed by switching from 15 to 60 s for 13 h. Images of gels stained with ethidium bromide or SYBR gold were archived using a Bio-Rad Gel Doc XR System. PFGE profiles were compared using BioNumerics with XbaI-digested Salmonella enterica serovar Braenderup H9812 DNA as a molecular size and gel normalization standard. PFGE patterns were categorized in comparison to known types using 80% similarity as determined by dendrogram analysis using Dice coefficients with the unweighted-pair group method using average linkages (UPGMA) as previously described by Killgore et al. (22). Analysis of PFGE patterns was performed using BioNumerics software, version 5.1 (Applied Maths, Belgium).
REA of chromosomal DNA.
Analysis of chromosomal DNA after restriction with the frequent-cutting enzyme HindIII was performed as previously described by Clabots et al. (7). DNA banding patterns were systematically compared by visual analysis in 1-mm segments to an extensive library of known REA groups. A similarity index was calculated, and patterns exhibiting ≥90% relatedness were grouped together. For this study, only the eight most frequently isolated REA groups were reported, i.e., BI, BK, CF, DH, G, J, K, and Y. The remaining patterns were designated the “unspecified” REA group.
Xpert C. difficile assay.
The Xpert C. difficile assay results were generated directly from stool samples at each of the seven study sites using the Xpert C. difficile cartridge, as described by the manufacturer (Cepheid, Sunnyvale, CA). Stool samples that were positive for the tcdB, cdtA, and a single nucleotide deletion at base 117 in tcdC (33) were reported as presumptive 027/NAP1/BI. Samples that were positive only for tcdB were reported as 027/NAP1/BI negative.
Analysis of strain typing results.
The congruence of the strain typing results generated by PCR-ribotyping, PFGE, and REA were determined by calculating the adjusted Rand and Wallace coefficients as described by Carrico et al. (6) and Pinto et al. (37) using the Comparing Partitions website (http://darwin.phyloviz.net/ComparingPartitions/index.php?link=Tool). Pairwise comparisons were made on data sets in which missing data (e.g., where one culture was contaminated and a strain type could not be determined) were deleted. Simpson's index of diversity was also determined for the three typing methods.
Statistical methods.
Fisher's exact test and a chi-square test were used for statistical analysis where appropriate.
RESULTS
PCR-ribotyping results.
All 350 toxigenic isolates obtained from symptomatic patients in six U.S. sites and one Canadian site were tested by PCR-ribotyping; 244 (69.7%) were assigned to 1 of 18 different PCR ribotypes (using the Cardiff-ECDC nomenclature). A total of 106 isolates did not match established ribotypes. Simpson's index of diversity for PCR-ribotyping was calculated to be 82.71 (95% confidence interval [CI], 80.02 to 85.40). The five most frequent ribotypes were 027 (92 isolates), 002 (18 isolates), 106 (17 isolates), 017 (15 isolates), and 078 (15 isolates) (Table 1). Isolates of all five of the most common ribotypes were recovered from eastern, midwestern, and western U.S. study sites, confirming the widespread dissemination of these strains.
Table 1.
Comparison of the most common PCR-ribotyping results to results of REA and PFGE
| PCR ribotype (no. of isolates) | Corresponding REA group; other identified group(s) (no. of isolates)a | Corresponding PFGE type; other identified type(s) (no. of isolates) |
|---|---|---|
| 001 (9) | J (7) | NAP2 (6) |
| 002 (18) | G (17); DH | Undefined types only |
| 014 (6) | Y (4) | NAP4 (4) |
| 017 (15) | CF (10); BK, Y | NAP9 (11) |
| 020 (8) | Y (8) | Primarily undefined types; NAP4 (1) |
| 027 (92)b | BI (86); BK | NAP1 (89) |
| 053 (13) | K (6); DH, Y | Undefined types only |
| 078 (15) | BK (14) | NAP7 (6) |
| 087 (4) | Unspecified group only | NAP12 (4) |
| 104 (9) | G (1); primarily unspecified group | NAP11 (8) |
| 106 (17) | DH (14) | NAP11 (13) |
Only eight major REA groups were identified in this study.
A total of 89 of the 92 isolates were identified presumptively as 027/NAP1/BI by the Xpert C. difficile assay.
Of the 92 PCR ribotype 027 strains, 89 were reported by PFGE as NAP1, and 86 were reported as BI by REA typing (Table 1 and Fig. 1). Five 027 isolates did not have REA types determined, and the other isolate belonged to the BK group. There were two undefined PFGE patterns, and the remaining culture was contaminated. PCR ribotype 002 correlated strongly with the REA G group (17 of 18 isolates), but there was no corresponding PFGE NAP type for this group of isolates. The 17 PCR ribotype 106 isolates were usually REA group DH and PFGE type NAP11 although most of the PCR ribotype 104 isolates were also NAP11 but were an unspecified REA group. Of the 15 PCR ribotype 017 isolates, 10 were REA group CF, and 11 were PFGE type NAP9. The PCR ribotype 078 isolates were usually REA group BK (14 of 15), but less than half had an identified PFGE type; of those that did, 6 of 15 were NAP7. Finally, among the nine PCR ribotype 001 isolates, seven belonged to the REA J group, and six of those were NAP2.
Fig. 1.
Venn diagram displaying the results for the three typing methods (REA, PFGE, and PCR-ribotyping) used on isolated colonies and the Xpert C. difficile assay, which was performed directly on stool, for the 027/NAP1/BI strain. GX, Xpert C. difficile assay. Four of the 92 isolates designated PCR ribotype 027 did not have corresponding REA data and 1 isolate did not have a PFGE type due to strain contamination during shipment.
REA typing results.
A total of 224 (67.9%) of the 330 isolates available for REA typing were assigned to one of eight major REA groups including BI (88 isolates), Y (40 isolates), J (22 isolates), G (20 isolates), DH (19 isolates), BK (17 isolates), CF (10 isolates), and K (8 isolates) (Table 2). For this study, the REA subtypes within the major groups (e.g., BI7) were not used for analysis. Other isolates were assigned to a unspecified REA group. Simpson's index of diversity for REA typing was calculated to be 78.25 (95% CI, 75.50 to 81.00). All but 2 of the 88 isolates reported as belonging to the BI group were ribotype 027, and neither of these was designated NAP1. Two additional BI isolates were not NAP1, but both were PCR ribotype 027. Of the 40 Y group isolates, 26 were NAP4, and the remaining isolates were an undefined pattern. The PCR ribotypes from group Y isolates were diverse; eight were 020, four were 014, and the other isolates were types 017, 053, 087, or unknown. The 22 J group isolates also showed diverse PCR ribotypes and PFGE types. Only seven J group isolates were type 001, and of those only five were NAP2. Two other J group isolates were NAP2, but most were other unnamed PFGE types. Aside from one isolate each of types 046 and 056, the remaining PCR ribotypes for the J group isolates were unnamed. All but three of the 17 BK group isolates were PCR ribotype 078, but only six were NAP7 (the others were an undesignated PFGE type or, in one case, NAP1). All of the CF group isolates were PCR ribotype 017, and nine of those were NAP9. Of the 19 DH group isolates, 14 were PCR ribotype 106, and the remaining isolates were 002, 046, or 053. The 20 group G isolates included 17 PCR ribotype 002, 1 type 104, and 2 other types. The majority of the G group isolates did not have an identified PFGE type although there was one isolate each of NAP4 and NAP11.
Table 2.
Comparison of most common REA results to results of PCR-ribotyping and PFGE
| REA group (no. of isolates) | Corresponding PCR ribotype (no. of isolates); other identified type(s) | Corresponding PFGE type (no. of isolates); other identified type(s) |
|---|---|---|
| BI (88)a | 027 (86) | NAP1 (85) |
| BK (17) | 078 (14); 027 | NAP7 (6); NAP1 |
| CF (10) | 017 (10) | NAP9 (9) |
| DH (19) | 106 (14); 002, 046, 053 | NAP11 (12) |
| G (20) | 002 (17); 104 | Undefined types only |
| J (22) | 001 (7); 046, 056 | NAP2 (7) |
| K (8) | 053 (6) | Undefined types only |
| Y (40) | 020 (8); 014, 017, 053, 087 | NAP4 (26) |
A total of 85 of the 88 isolates were identified presumptively as 027/NAP1/BI by the Xpert C. difficile assay.
Pulsed-field gel electrophoresis.
Although all but 8 of the 342 isolates could be analyzed by PFGE, only 187 (54.7%) gave patterns that could be assigned to 1 of the 12 described NAP types. The patterns identified in the study were NAP1, NAP2, NAP4, NAP7, NAP9, NAP11, and NAP12 (Table 3). Simpson's index of diversity for PFGE was calculated to be 65.30 (95% CI, 61.29 to 69.32), which was the lowest of the three typing methods. Of the 91 isolates designated NAP1, 88 were PCR ribotype type 027, and 85 were REA group BI. Two isolates did not have a named PCR ribotype, while one was type REA type BK, one was an unspecified REA type, and four isolates were nontypeable or contaminated. Two NAP1 isolates were neither ribotype 027 nor REA group BI. Seven of the eight NAP2 isolates were the J group by REA, but only six were PCR ribotype 001. Of the 31 NAP4 isolates, 26 were REA group Y, but most did not have a named PCR ribotype. The four that did were PCR ribotype 014. The six NAP7 isolates all were PCR ribotype 078 and REA type BK. Of the 11 NAP9 isolates, all 11 were PCR ribotype 017, but only nine were REA group CF; the other two were group BK or Y. Isolates showing the NAP11 PFGE pattern encompassed both ribotypes 104 and 106, but only half were REA type DH (mostly those that also were type 106). The remaining isolates were not in a specified REA group.
Table 3.
Comparison of PFGE results to results of PCR-ribotyping and REA
| PFGE type (no. of isolates) | Corresponding PCR ribotype(s) (no. of isolates); other identified type(s) | Corresponding REA group (no. of isolates); other identified group(s) |
|---|---|---|
| NAP1 (91)a | 027 (88) | BI (84); BK |
| NAP2 (8) | 001 (6) | J (7) |
| NAP4 (31) | Primarily undefined types; 014 (1), 020 (1) | Y (26); DH, G |
| NAP7 (6) | 078 (6) | BK (6) |
| NAP9 (11) | 017 (11) | CF (9); BK, Y |
| NAP11 (24) | 106 (13); 104 | DH (12); G |
| NAP12 (7) | 087 (4); 015 | Unspecified group only |
A total of 88 of the 91 isolates were identified presumptively as 027/NAP1/BI by the Xpert C. difficile assay.
Presumptive identification of 027/NAP1/BI strains using Xpert C. difficile.
A total of 96 (27.4%) C. difficile isolates were designated presumptive 027/NAP1/BI by the Xpert C. difficile assay. Of these, 83 were PCR ribotype 027, REA group BI and PFGE type NAP1. Eight additional isolates were PCR ribotype 027, of which five were also NAP1 but not BI (four had REA groups other than the eight major groups reported), and two were BI but not NAP1. The remaining isolate did not have a PFGE or REA group reported due to culture contamination. One additional isolate was type 053/NAP undefined/K. The four remaining Xpert presumptive 027/NAP1/BI-positive isolates (of the 96 total) were neither 027, BI, nor NAP1. None of the four isolates, which came from three geographically diverse locations, had a known PCR ribotype, NAP type, or REA group identified. The two from the same location were isolated several months apart. Overall, 91 of 96 (94.8%) of the Xpert C. difficile presumptive 027/NAP1/BI designations were confirmed by at least two typing methods as being in the 027/NAP1/BI cluster. There were no significant differences between the sensitivities, specificities, or positive or negative predictive values of the results obtained with the Xpert C. difficile test and the results of PCR-ribotyping, REA, or PFGE (P > 0.05 for all comparisons) (Table 4). The presumptive identification of the 027/NAP1/BI strain using the Xpert cartridge was available within 45 min of initiating the testing of the stool sample.
Table 4.
Results of Xpert C. difficile 027/NAP1/BI test compared to results of standard typing methods
| Parametera | No. of isolates of the indicated type/no. of isolates tested (%)b |
P valuec | ||
|---|---|---|---|---|
| PCR ribotype 027 (n = 350) | PFGE type NAP1 (n = 342) | REA type BI (n = 330) | ||
| Sensitivity | 91/92 (98.9) | 88/91 (96.7) | 85/88 (96.6) | 0.5739 |
| Specificity | 253/258 (98.1) | 244/251(97.2) | 236/242 (97.5) | 0.8343 |
| PPV | 91/96 (94.8) | 88/95 (92.6) | 85/91 (93.4) | 0.8287 |
| NPV | 253/254 (99.6%) | 244/247 (98.8) | 236/239 (98.7) | 0.5779 |
PPV, positive predictive value; NPV, negative predictive value.
Typing was determined by the indicated method. The denominator changes because not all isolates were available for typing by each method. n, total number of isolates.
Calculated using Fisher's exact test.
Distribution of known C. difficile strains among study sites and analysis of congruence.
The results of the three typing methods (PCR ribotype, PFGE type, and REA group) were combined to give a composite strain type for five well-characterized strains previously identified in the literature in an effort to better understand their geographic distribution. This included the J strain (001/NAP2/J) (19, 47), the toxin A-negative toxin B-positive ribotype 017 strain (017/NAP9/CF) (38), the hypervirulent/epidemic 027/NAP1/BI strain (33), the PCR ribotype 078 animal strain (078/NAP7/BK) (13, 17, 39), and the PCR ribotype 106 United Kingdom outbreak strain (106/NAP11/DH) (Table 5) (41). The 027/NAP1/BI strain was the most widely disseminated of the five strain types among the study sites and represented an overall mean of 25% of isolates recovered at each site (range, 11 to 54%) (data not shown). With the exception of the 001/NAP2/J strain, which was not reported by the East Coast study site, the other strains were present all across the United States but not in Canada. The 017/NAP9/CF results formed the tightest cluster and showed the most overlap of the three typing methods (after the 027/NAP1/BI cluster), while the J strain had many non-001 PCR ribotype results and non-NAP2 PFGE results, forming the loosest clustering of results.
Table 5.
Geographic distribution of commonly recognized strains by three typing methods
| Strain (PCR ribotype/PFGE/REA) | No. of isolates | Locations |
|---|---|---|
| 001/NAP2/Ja | 6 | U.S. Midwest and West |
| 017/NAP9/CF | 9 | U.S. East Coast, Midwest, and West |
| 027/NAP1/BIb | 84 | U.S. East Coast, Midwest, Pacific Northwest, and West; Canada |
| 078/NAP7/BKc | 5 | U.S. East Coast, Midwest, and West |
| 106/NAP11/DHd | 11 | U.S. East Coast, Midwest, and West |
The adjusted Rand coefficients, which indicate the overall congruence between the results of two typing methods, were 0.546 for PCR-ribotyping and PFGE, 0.584 for PFGE and REA, and 0.686 for PCR-ribotyping and REA. This indicates a slightly better matching of PCR-ribotyping and REA results (the closer the number is to 1, the higher the congruence of results), but none of the pairs of results generated by two typing method was highly correlated. The Wallace coefficients, which indicate the likelihood that a strain type from one method can predict the strain type from a second method or, alternatively, that adding a second method will provide additional strain discrimination, are presented in Table 6. The high coefficients for PCR-ribotyping and PFGE (0.902) and for PCR-ribotyping and REA (0.834) indicate that the addition of either PFGE or REA groups to PCR-ribotyping adds little additional strain discrimination. However, adding PCR-ribotyping to PFGE results (0.488) has better utility (since the number is closer to zero) than adding PCR-ribotyping to REA (0.658).
Table 6.
Congruence of strain typing results as indicated by Wallace coefficients
| Method | Wallace coefficient (95% CI)a |
||
|---|---|---|---|
| PCR-ribotyping | PFGE | REA | |
| PCR-ribotyping | 0.902 (0.852–0.951) | 0.834 (0.795–0.872) | |
| PFGE | 0.488 (0.460–0.516) | 0.588 (0.547–0.629) | |
| REA | 0.658 (0.618–0.698) | 0.792 (0.733–0.852) | |
Data were generated using the Comparing Partitions website (http://darwin.phyloviz.net/ComparingPartitions/index.php?link=Tool). CI, confidence interval.
DISCUSSION
C. difficile strains can be differentiated from one another by a variety of different phenotyping and genotyping methods. PCR-ribotyping, PFGE, and REA are among the most commonly used genotypic methods for outbreak analysis and epidemiologic studies of C. difficile strains. MLST is reserved primarily for population studies as its level of strain discrimination is too low for studies of outbreaks (30). Although Killgore et al. (22) examined the ability of multiple genotypic methods to cluster isolates of the hypervirulent/epidemic 027/NAP1/BI strain from around the world accurately, the study was limited to 42 isolates and did not rigorously explore the accuracy of the methods for other C. difficile strains. Nonetheless, Wilson et al. used the report of Killgore and colleagues as a basis for extrapolating REA results to PCR ribotypes (45). In that study, REA group DH strains were presumed to be ribotype 106, and group J strains were presumed to be ribotype 001. Our data indicate that group DH strains could be ribotypes 002, 046, and 053 in addition to 106, while group J strains could be ribotype 046 in addition to ribotype 001, potentially altering the interpretation of the data. Martin and colleagues typed 1,080 C. difficile isolates from Ontario, Canada, by multiple methods, including PCR-ribotyping, but performed PFGE and toxinotyping only on selected isolates (32). While the study did not provide adequate information about the congruence of the typing methods, the authors did note that NAP1 isolates could be assigned one of three ribotypes and that NAP2 isolates could be divided into four distinct ribotypes. This prompted our interest in whether one typing method, such as PCR-ribotyping, could be used to identify strains that were initially described using another typing technique, such as the J strain (representing REA grouping) (19) or the toxinotype V/NAP7 strain from animals (identified by PFGE), which would later be identified as PCR ribotype 078 (17).
Among the 350 isolates we examined by three methods, PFGE, which had the lowest calculated index of diversity at 65.30, yielded an established pattern for only 54% of the isolates tested while PCR-ribotyping (index of diversity of 82.71) provided a named ribotype for approximately 70% of the isolates. Both typing methods usually generated readable banding patterns, but many of the patterns had not yet been assigned a strain type. This is less of a problem with REA, which has more extensively defined patterns in its database (7), although for simplicity only the eight major patterns were used in this study (yielding an index of diversity of 78.25). Aside from the 027/NAP1/BI clone, the three methods showed the best congruence for strains that were 017/NAP9/CF. Knowing that PCR ribotype 017 isolates are typically PFGE type NAP9 and REA group CF may help identify these isolates in PFGE or REA databases and illuminate the epidemiology of toxin A-negative/toxin B-positive isolates. None of the other intersecting strain types, including 001/NAP2/J, 078/NAP7/BK, or 106/NAP11/DH were as tightly linked, and trying to predict a strain type based on one method (such as PFGE) using the other method (such as REA) may be problematic. This was reflected in the adjusted Rand coefficients for the pairwise comparisons of typing results, which ranged from 0.546 to 0.686, where values close to 1.0 indicate high congruence of results between methods. That said, the linkages of the three types provided here may still be useful as a starting point for examining the global spread of these common C. difficile lineages, recognizing, for example, that not all NAP11 isolates will be either PCR ribotype 106 or REA group DH.
Of the three methods used in this study, REA has the longest history as a molecular typing method for C. difficile (28) but is used primarily by a single laboratory in the United States and relies on visual interpretation of the data. This technique involves digesting total bacterial DNA with a frequent-cutting restriction endonuclease and separating the DNA fragments by agarose gel electrophoresis. REA was critical in defining a new epidemic strain of C. difficile (the J strain) in 1999 (19). It is interesting that of the 22 J group isolates in this study, only seven were both PCR ribotype 001 and NAP2, suggesting that REA typing tends to define a broader group of isolates than PFGE or PCR-ribotyping. This also seems to be true for REA groups BK, DH, G, and Y (Table 2). This may be consistent with the broader sampling of genomic information afforded by REA, in contrast to the more restricted location or number of chromosomal sites sampled by PCR-ribotyping and PFGE, respectively. However, as noted above, there appears to be a closer link between REA group CF strains, PCR ribotype 017, and NAP9, suggesting that this lineage is more clonal. Subtypes of the REA groups have also been described (e.g., BI6, BI8, and BI17) but those extended subtypes were not used in this study for strain comparisons as it made comparisons too complex. The discrepancies among REA, PFGE, and PCR-ribotyping have been noted previously (17, 22).
PCR-ribotyping is a more recent molecular typing technique that is playing a key role in defining the epidemiology of C. difficile infections in Europe, North America, and the Far East (24, 29, 40). There are now more than 200 PCR ribotypes described. Although this presumably provides higher levels of strain discrimination than PFGE, Martin et al. argue that if all PFGE patterns are considered and not just those patterns that are named, PFGE may provide a higher level of strain discrimination (32). Kuijper and colleagues established the link between the BI/NAP1 epidemic strain of C. difficile in North America and the PCR ribotype 027 in Europe in 2006, which helped advance our understanding of the epidemiology of this globally disseminated strain (27). Nonetheless, as noted by Killgore et al., not every ribotype 027 strain is REA group BI, nor is every NAP1 isolate ribotype 027 (22). However, of the three typing techniques, PCR-ribotyping is more readily implemented in a clinical laboratory.
PFGE is a third molecular technique used in this study for typing isolates of C. difficile. Initial attempts to type C. difficile by PFGE were largely unsuccessful due to degradation of DNA in the gels (23). The PFGE method was later optimized by incorporating thiourea into the gels and was used in conjunction with REA in early epidemiologic studies to define the most recent epidemic strain of C. difficile, designated BI/NAP1/027 (33). While the BI/NAP1/027 epidemic strain is classified as toxinotype III (38), the multicenter typing study reported by Killgore and colleagues showed that PFGE classified two toxinotype IX isolates as NAP1 (22). These two isolates did not contain the classic deletion in tcdC (a single base deletion resulting in a frameshift at position 117), indicative of the BI/NAP1/027 epidemic strain. Thus, as a typing method, PFGE is known to include within the NAP1 designation some strains of C. difficile that do not have the characteristics of the 027/NAP1/BI epidemic strain. However, only two NAP1 strains that did not contain the characteristic tcdC deletion of the epidemic 027/NAP1/BI strain were seen in our study. Thus, like REA, some NAP strain designations appear to encompass more isolate types than are recognized by the other typing methods. Martin and colleagues have made this same observation (32). For a laboratory wishing to initiate typing of C. difficile isolates, the Simpson's index of diversity scores and Wallace coefficients from our data set indicate that PCR-ribotyping should be the primary method of strain differentiation. Adding PFGE results provided better strain discrimination than adding REA typing as a second method (Wallace coefficient of 0.902 versus 0.834) and is likely to be easier to implement in the laboratory.
The Xpert C. difficile test provided a presumptive identification of a single strain, 027/NAP1/BI, in parallel with the detection of the tcdB gene for C. difficile identification, which may be of value particularly for infection control activities in hospitals (2, 15). The ability to identify rapidly a cluster of patients in a hospital that is infected with the 027/NAP1/BI strain could be the key to implementing infection control efforts to halt transmission in the hospital, including restriction of fluoroquinolones (15, 20, 35). The 027/NAP1/BI isolates have the potential to produce high levels of spores and require enhanced environmental cleaning efforts for eradication (1, 35). In addition, the 027/NAP1/BI strain may cause more severe disease in patients than other ribotypes (11, 14, 31, 36), and patients infected with this strain may disseminate the strain more broadly in health care settings due to the enhanced volume of stool produced and subsequent environmental contamination. However, therapeutic decisions should be based on disease severity and not on organism strain type (9) as not all C. difficile infections caused by PCR ribotype 027 isolates are severe (8, 21).
Overall, the Xpert C. difficile results for presumptive identification of 027/BI/NAP1 agreed well with the results of PCR-ribotyping, REA, and PFGE. There were only four instances out of 96 results where the positive presumptive identification of an isolate by the Xpert C. difficile test was not type 027, NAP1, or BI. Of the techniques tested in this study, only the Xpert C. difficile assay could be run directly on stool samples and provide presumptive evidence of type 027/NAP1/BI in 45 min. Confirmation of the strain type would require culturing of the stool sample to recover the C. difficile isolate and then sending the isolate to a reference laboratory for typing using one of the three methods discussed above. There are relatively limited options available for strain typing C. difficile isolates in the United States since few state public health laboratories provide this service. However, testing the isolate for moxifloxacin resistance by the Etest method could be used to support the identification of the isolate as 027/NAP1/BI (33). The unique combination of the three genetic markers (tcdB, cdt, and the tcdC deletion at nucleotide 117) make the presumptive identification of the 027/NAP1/BI strain feasible. Similar loci are not readily available for identification of other C. difficile strains.
In conclusion, PCR-ribotyping, REA, and PFGE are all useful for typing C. difficile isolates although they each provide different patterns of strain clustering. One must be cautious about predicting a strain type based on any single method (i.e., not all NAP1 strains can be assumed to be ribotype 027 or REA group BI). Using more than a single method increases strain differentiation and may better illuminate the changing epidemiology of C. difficile strains, but few laboratories perform more than a single typing method. The Xpert C. difficile assay was the only method that reported strain data directly from specimens and yielded the information at the same time that the toxin B gene data were available. Rapid identification of the 027/BI/NAP1 strain type may be useful for tracking in-hospital or between-hospital outbreaks of C. difficile in real time, which may be of particular value for improving infection control interventions in conjunction with antimicrobial stewardship programs to decrease the duration of C. difficile outbreaks (12, 20, 35, 44).
ACKNOWLEDGMENTS
We thank Ingela Alefjord for her technical expertise and Joseph Whitmore for assistance with statistical analysis. We also thank Grit Ackermann for her advice on detection of 027 strains.
F.C.T., E.W., A.W., and D.H.P. are employees and shareholders of Cepheid. T.B. and A.M.J. are employees of Cepheid A.B. D.N.G., R.V.G., and T.A. were investigators on a research grant from Cepheid that funded this study.
D.N.G. was supported by the U.S. Department of Veterans Affairs.
Footnotes
Published ahead of print on 9 March 2011.
REFERENCES
- 1. Akerlund T., et al. 2008. Increased sporulation rate of epidemic Clostridium difficile type 027/NAP1. J. Clin. Microbiol. 46:1530–1533 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Babady N. E., et al. 2010. Evaluation of the Cepheid Xpert Clostridium difficile Epi assay for diagnosis of C. difficile infection and typing of the NAP1 strain at a cancer hospital. J. Clin. Microbiol. 48:4519–4524 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Barbut F., et al. 2007. Clinical features of Clostridium difficile-associated infections and molecular characterization of strains: results of a retrospective study, 2000–2004. Infect. Control Hosp. Epidemiol. 28:131–139 [DOI] [PubMed] [Google Scholar]
- 4. Barbut F., et al. 2007. Prospective study of Clostridium difficile infections in Europe with phenotypic and genotypic characterisation of the isolates. Clin. Microbiol. Infect. 13:1048–1057 [DOI] [PubMed] [Google Scholar]
- 5. Bartlett J. G. 2008. Historical perspectives on studies of Clostridium difficile and C. difficile infection. Clin. Infect. Dis. 46(Suppl. 1):S4–S11 [DOI] [PubMed] [Google Scholar]
- 6. Carrico J. A., et al. 2006. Illustration of a common framework for relating multiple typing methods by application to macrolide-resistant Streptococcus pyogenes. J. Clin. Microbiol. 44:2524–2532 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Clabots C. R., et al. 1993. Development of a rapid and efficient restriction endonuclease analysis typing system for Clostridium difficile and correlation with other typing systems. J. Clin. Microbiol. 31:1870–1875 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Cloud J., Noddin L., Pressman A., Hu M., Kelly C. 2009. Clostridium difficile strain NAP-1 is not associated with severe disease in a nonepidemic setting. Clin. Gastroenterol. Hepatol. 7:868–873 [DOI] [PubMed] [Google Scholar]
- 9. Cohen S. H., et al. 2010. Clinical practice guidelines for Clostridium difficile infection in adults: 2010 update by the Society for Healthcare Epidemiology of America (SHEA) and the Infectious Diseases Society of America (IDSA). Infect. Control Hosp. Epidemiol. 31:431–455 [DOI] [PubMed] [Google Scholar]
- 10. Debast S. B., et al. 2009. Clostridium difficile PCR ribotype 078 toxinotype V found in diarrhoeal pigs identical to isolates from affected humans. Environ. Microbiol. 11:505–511 [DOI] [PubMed] [Google Scholar]
- 11. Deneve C., Janoir C., Poilane I., Fantinato C., Collignon A. 2009. New trends in Clostridium difficile virulence and pathogenesis. Int. J. Antimicrob. Agents 33(Suppl. 1):S24–S28 [DOI] [PubMed] [Google Scholar]
- 12. Dumford D. M., III, Nerandzic M. M., Eckstein B. C., Donskey C. J. 2009. What is on that keyboard? Detecting hidden environmental reservoirs of Clostridium difficile during an outbreak associated with North American pulsed-field gel electrophoresis type 1 strains. Am. J. Infect. Control. 37:15–19 [DOI] [PubMed] [Google Scholar]
- 13. Goorhuis A., et al. 2008. Emergence of Clostridium difficile infection due to a new hypervirulent strain, polymerase chain reaction ribotype 078. Clin. Infect. Dis. 47:1162–1170 [DOI] [PubMed] [Google Scholar]
- 14. Hookman P., Barkin J. S. 2009. Clostridium difficile associated infection, diarrhea and colitis. World J. Gastroenterol. 15:1554–1580 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Huang H., Weintraub A., Fang H., Nord C. E. 2009. Comparison of a commercial multiplex real-time PCR to the cell cytotoxicity neutralization assay for diagnosis of Clostridium difficile infections. J. Clin. Microbiol. 47:3729–3731 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Jarvis W. R., Schlosser J., Jarvis A. A., Chinn R. Y. 2009. National point prevalence of Clostridium difficile in US health care facility inpatients, 2008. Am. J. Infect. Control. 37:263–270 [DOI] [PubMed] [Google Scholar]
- 17. Jhung M. A., et al. 2008. Toxinotype V Clostridium difficile in humans and food animals. Emerg. Infect. Dis. 14:1039–1045 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Johnson E. A., Summanen P., Finegold S. M. 2007. Clostridium, p. 889–910 In Murray P. R., Baron E. J., Jorgensen J. H., Landry M. L., Pfaller M. A. (ed.), Manual of clinical microbiology, 9th ed. ASM Press, Washington, DC [Google Scholar]
- 19. Johnson S., et al. 1999. Epidemics of diarrhea caused by a clindamycin-resistant strain of Clostridium difficile in four hospitals. N. Engl. J. Med. 341:1645–1651 [DOI] [PubMed] [Google Scholar]
- 20. Kallen A. J., et al. 2009. Complete restriction of fluoroquinolone use to control an outbreak of Clostridium difficile infection at a community hospital. Infect. Control Hosp. Epidemiol. 30:264–272 [DOI] [PubMed] [Google Scholar]
- 21. Kelly C. P., LaMont J. T. 2008. Clostridium difficile—more difficult than ever. N. Engl. J. Med. 359:1932–1940 [DOI] [PubMed] [Google Scholar]
- 22. Killgore G., et al. 2008. Comparison of seven techniques for typing international epidemic strains of Clostridium difficile: restriction endonuclease analysis, pulsed-field gel electrophoresis, PCR-ribotyping, multilocus sequence typing, multilocus variable-number tandem-repeat analysis, amplified fragment length polymorphism, and surface layer protein A gene sequence typing. J. Clin. Microbiol. 46:431–437 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Klaassen C. H., van Haren H. A., Horrevorts A. M. 2002. Molecular fingerprinting of Clostridium difficile isolates: pulsed-field gel electrophoresis versus amplified fragment length polymorphism. J. Clin. Microbiol. 40:101–104 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Koh T. H., Tan A. L., Tan M. L., Wang G., Song K. P. 2007. Epidemiology of Clostridium difficile infection in a large teaching hospital in Singapore. Pathology 39:438–442 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Kuijper E. J., et al. 2008. Update of Clostridium difficile infection due to PCR ribotype 027 in Europe, 2008. Euro Surveill. 13:18942. [PubMed] [Google Scholar]
- 26. Kuijper E. J., et al. 2007. Update of Clostridium difficile-associated disease due to PCR ribotype 027 in Europe. Euro Surveill. 12:E1–E2 [DOI] [PubMed] [Google Scholar]
- 27. Kuijper E. J., Coignard B., Tull P. 2006. Emergence of Clostridium difficile-associated disease in North America and Europe. Clin. Microbiol. Infect. 12(Suppl. 6):2–18 [DOI] [PubMed] [Google Scholar]
- 28. Kuijper E. J., Oudbier J. H., Stuifbergen W. N., Jansz A., Zanen H. C. 1987. Application of whole-cell DNA restriction endonuclease profiles to the epidemiology of Clostridium difficile-induced diarrhea. J. Clin. Microbiol. 25:751–753 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Kuijper E. J., et al. 2006. Clostridium difficile ribotype 027, toxinotype III, the Netherlands. Emerg. Infect. Dis. 12:827–830 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Lemee L., Dhalluin A., Pestel-Caron M., Lemeland J. F., Pons J. L. 2004. Multilocus sequence typing analysis of human and animal Clostridium difficile isolates of various toxigenic types. J. Clin. Microbiol. 42:2609–2617 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Loo V. G., et al. 2005. A predominantly clonal multi-institutional outbreak of Clostridium difficile-associated diarrhea with high morbidity and mortality. N. Engl. J. Med. 353:2442–2449 [DOI] [PubMed] [Google Scholar]
- 32. Martin H., et al. 2008. Characterization of Clostridium difficile strains isolated from patients in Ontario, Canada, from 2004 to 2006. J. Clin. Microbiol. 46:2999–3004 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. McDonald L. C., et al. 2005. An epidemic, toxin gene-variant strain of Clostridium difficile. N. Engl. J. Med. 353:2433–2441 [DOI] [PubMed] [Google Scholar]
- 34. McDonald L. C., Owings M., Jernigan D. B. 2006. Clostridium difficile infection in patients discharged from US short-stay hospitals, 1996–2003. Emerg. Infect. Dis. 12:409–415 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Muto C. A., et al. 2007. Control of an outbreak of infection with the hypervirulent Clostridium difficile BI strain in a university hospital using a comprehensive “bundle” approach. Clin. Infect. Dis. 45:1266–1273 [DOI] [PubMed] [Google Scholar]
- 36. O'Connor J. R., Johnson S., Gerding D. N. 2009. Clostridium difficile infection caused by the epidemic BI/NAP1/027 strain. Gastroenterology 136:1913–1924 [DOI] [PubMed] [Google Scholar]
- 37. Pinto F. R., Melo-Cristino J., Ramirez M. 2008. A confidence interval for the wallace coefficient of concordance and its application to microbial typing methods. PLoS One 3:e3696. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Rupnik M. 2008. Heterogeneity of large clostridial toxins: importance of Clostridium difficile toxinotypes. FEMS Microbiol. Rev. 32:541–555 [DOI] [PubMed] [Google Scholar]
- 39. Rupnik M., Widmer A., Zimmermann O., Eckert C., Barbut F. 2008. Clostridium difficile toxinotype V, ribotype 078, in animals and humans. J. Clin. Microbiol. 46:2146. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Stubbs S. L., Brazier J. S., O'Neill G. L., Duerden B. I. 1999. PCR targeted to the 16S-23S rRNA gene intergenic spacer region of Clostridium difficile and construction of a library consisting of 116 different PCR ribotypes. J. Clin. Microbiol. 37:461–463 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Sundram F., et al. 2009. Clostridium difficile ribotypes 027 and 106: clinical outcomes and risk factors. J. Hosp. Infect. 72:111–118 [DOI] [PubMed] [Google Scholar]
- 42. Svenungsson B., et al. 2003. Epidemiology and molecular characterization of Clostridium difficile strains from patients with diarrhea: low disease incidence and evidence of limited cross-infection in a Swedish teaching hospital. J. Clin. Microbiol. 41:4031–4037 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Tenover F. C., et al. 2010. Impact of strain types on detection of toxigenic Clostridium difficile: comparison of molecular diagnostic and enzyme immunoassay approaches. J. Clin. Microbiol. 48:3719–3724 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Weiss K., et al. 2009. Multipronged intervention strategy to control an outbreak of Clostridium difficile infection (CDI) and its impact on the rates of CDI from 2002 to 2007. Infect. Control Hosp. Epidemiol. 30:156–162 [DOI] [PubMed] [Google Scholar]
- 45. Wilson V., et al. 2010. Predictors of death after Clostridium difficile infection: a report on 128 strain-typed cases from a teaching hospital in the United Kingdom. Clin. Infect. Dis. 50:e77–81 [DOI] [PubMed] [Google Scholar]
- 46. Zaiss N. H., et al. 2009. Typing Clostridium difficile strains based on tandem repeat sequences. BMC Microbiol. 9:6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Zaiss N. H., Witte W., Nubel U. 2010. Fluoroquinolone resistance and Clostridium difficile, Germany. Emerg. Infect. Dis. 16:675–677 [DOI] [PMC free article] [PubMed] [Google Scholar]