Molecular Epidemiology of Clostridium difficile at a Medical Center in Taiwan: Persistence of Genetically Clustering of A−B+ Isolates and Increase of A+B+ Isolates

Introduction We investigated the changing trend of various toxigenic Clostridium difficile isolates at a 3 500-bed hospital in Taiwan. Genetic relatedness and antimicrobial susceptibility of toxigenic C. difficile isolates were also examined. Methods A total of 110 non-repeat toxigenic C. difficile isolates from different patients were collected between 2002 and 2007. Characterization of the 110 toxigenic isolates was performed using agar dilution method, multilocus variable-number tandem-repeat analysis (MLVA) genotyping, tcdC genotyping, and toxinotyping. Results Among the 110 toxigenic isolates studied, 70 isolates harbored tcdA and tcdB (A+B+) and 40 isolates harbored tcdB only (A−B+). The annual number of A+B+ isolates considerably increased over the 6-year study (P = 0.055). A total of 109 different MLVA genotypes were identified, in which A+B+ isolates and A−B+ isolates were differentiated into two genetic clusters with similarity of 17.6%. Twenty-four (60%) of the 40 A−B+ isolates formed a major cluster, MLVA-group 1, with a similarity of 85%. Seven (6.4%) resistant isolates were identified, including two metronidazole-resistant and five vancomycin-resistant isolates. Conclusions This study indicated a persistence of a MLVA group 1 A−B+ isolates and an increase of A+B+ isolates with diverse MLVA types. Moreover, C. difficile isolates with antimicrobial resistance to metronidazole or vancomycin were found to have emerged. Continuous surveillance is warranted to understand the recent situation and control the further spread of the toxigenic C. difficile isolates, especially among hospitalized patients.


Introduction
Clostridium difficile is an anaerobic, gram-positive, spore-forming bacillus. It is one of the most common nosocomial pathogens identified and is the primary cause of antibiotic-associated diarrhea. [1] C. difficile-associated disease (CDAD) encompasses diseases of a range of severity from uncomplicated mild diarrhea to toxic megacolon that can result in sepsis and even death. [1] CDAD has been an increasing problem in health care, especially because hypervirulent strains (ribotype 027, toxinotype III, and pulse-field NAP1) have emerged in North America and Europe over the past 10 years. [2,3,4] The pathogenicity of C. difficile is primarily based on the action of at least one of the two major exotoxins produced and secreted by the bacteria, i.e., toxin A (enterotoxin) and toxin B (cytotoxin), which are encoded by the tcdA and tcdB gene, respectively. [5,6] In addition, some C. difficile isolates also produce a binary toxin called CDT, which is an actin-ADP-ribosylating toxin. [7] Although the pathological role of CDT in CDAD remains unclear, CDT contributes to CDAD and has been associated with increased disease severity. [8,9].
Laboratory diagnosis of CDAD is currently achieved by isolation of toxigenic C. difficile isolates from stool samples and detecting the produced toxins. Several methods can be used to diagnose C. difficile infection. These methods included C. difficile culture, cell cytotoxicity assay from stool filtrates, latex agglutination for the detection of C. difficile-associated antigen in stools, and enzyme immunoassay for the detection of toxin A, toxin B or both from stool samples. [10,11] Recently, to distinguish toxigenic from non-toxigenic C. difficile isolates, a multiplex-PCR assay simultaneously amplifying tcdA and tcdB genes was developed. [12] PCRs for the detection of binary toxin and tcdC gene deletion were also studied. [13] A highly sensitive real-time PCR method for the rapid detection of toxigenic C. difficile in stool samples had also been used for diagnosing CDAD. [14,15,16].
The antibiotics metronidazole and vancomycin are frequently used to treat CDAD. Oral metronidazole is the drug of choice for initial CDAD therapy because of its lower cost and concerns regarding the proliferation of vancomycin-resistant nosocomial bacteria. Vancomycin is recommended for treatment in patients with severe infection because of faster symptom resolution and a significantly lower risk of treatment failure. [17] As previous reports have indicated, C. difficile clinical isolates were sensitive to metronidazole or vancomycin, [18] clinical laboratories do not routinely perform antimicrobial susceptibility tests on this organism. However, up to 6.3% of toxin-producing isolates with resistance to metronidazole, and 3% with intermediate resistance to vancomycin were reported. [19] Poor outcomes of metronidazole therapy in CDAD were also recently reported, [4,20] which suggests that the drug resistance pattern of C. difficile may be changing.
CDAD have been reported in Asia countries such as Japan, Korea, Singapore and Thailand. [21,22,23,24] In Taiwan, the incidence of CDAD has recently been reported as 45 cases per100,000 patient-days, and was highest in medical intensive care units. [25] Few systematic investigations have monitored the drug resistance pattern, prevalence of toxin genes, and bacterial strain clonality in clinical isolates.
Between 2002 and 2007, a total of 2,471 stool specimens were ordered for C. difficile cultures at Chang Gung Memorial Hospital, a 3 500-bed medical center in northern Taiwan. A total of 232 non-repeated C. difficile isolates from different patients were identified in the clinical microbiology laboratory. Of the 232 isolates, a total of 181 (78%) C. difficile isolates were retrospectively retrieved from the bacteria bank for toxin gene testing using the PCR amplification method. A total of 110 toxigenic C. difficile isolates were identified and subjected to antimicrobial susceptibility testing and genetic relatedness analysis using a multilocus variable-number tandem-repeat analysis. Further characterization of tcdC genotypes and toxinotypes was also performed.

Ethics Statement
The present study aimed to characterize C. difficile isolates using molecular methods. All isolates studied were retrieved retrospectively from the Bacteria Bank, Department of Laboratory Medicine, Chang Gung Memorial Hospital, Linkou. The clinical information of the patients was neither available nor required in this study. The, patient's informed consent was not required or collected because all microbial cultures were ordered by physicians due to the necessity of clinical management (none were collected purposely for this study). The design and procedure of the study had been approved by the Institutional Review Board of the Chang Gung Memorial Hospital, Linkou, in January 2009.

Bacterial Isolation and Identification
Between 2002 and 2007, a total of 110 non-repeat toxigenic C. difficile isolates from different patients were retrospectively retrieved from the bacteria bank for use in the present study. Clostridium difficile selective agar (Becton Dickinson, USA) was used for bacterial isolation. Isolates were identified using conventional physiological and biochemical tests and were confirmed using the rapid ID 32A system (BioMerieux, France). All specimens subjected to microbial cultures were ordered by physicians for clinical management.
Detection of Tcda, Tcdb, cdtA, and cdtb Genes DNA extraction was performed using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. To determine the presence of toxin genes tcdA and tcdB in the C. difficile isolates, PCR amplification was performed, as previously described. [12,26] Briefly, two primer pairs, NK9/ NK11 and NK104/NK105, were used to amplify the repeating domain of the tcdA gene and the non-repeating domain of the tcdB gene, respectively. PCR amplification produced the intact tcdA gene from A + B + isolates yielded a 1,200 bp DNA product. In comparison, shorter DNA fragments of 500 or 700 bp were amplified from the A 2 B + isolates. The binary toxin genes cdtA and cdtB were detected concurrently using PCR. [12] The primer pairs cdtApos/cdtArev and cdtBpos/cdtBrev were used to amplify a 375-bp fragment from cdtA, and a 510-bp fragment from cdtB, respectively. The amplified DNA products were separated by agarose gel electrophoresis and photographed under BioDoc-It system (UVP, USA).

Toxinotyping
The toxigenic C. difficile isolates were further characterized using toxinotyping according to the method of Rupnik et al. [27] Toxinotyping analyzed the restriction-fragment-length polymorphisms (RFLPs) of the genes encoding toxins A (tcdA) and B (tcdB) in a region of the C. difficile genome known as the pathogenicity locus (PaLoc). We used RFLP analysis for PCR fragments A3 and B1 because this typing assay can identify most of the toxinotypes, [27] in this study.

PCR Amplification and DNA Sequencing of tcdc Gene
To further investigate the tcdC gene, the toxigenic C. difficile isolates were analyzed using PCR with primers C1 and C2, as previously described. [28] A 718-bp fragment of the PaLoc encompassing the entire tcdC gene was amplified. PCR products were purified and subjected to sequencing with amplification (C1 and C2) primers from both directions using a 3100-Avant Genetic Analyzer (Applied Biosystems, USA). The sequences were analyzed and the amino acid sequences deduced using the Lasergene 7.0.0 software package (DNASTAR, Wisconsin, USA) were compared to the wild-type tcdC sequence from strain VPI10463 (GenBank accession number Y10689).

Genotyping by Multilocus Variable-number Tandemrepeat Analysis
The genetic relatedness of the toxigenic C. difficle clinical isolates was investigated using multilocus variable-number tandem-repeat analysis (MLVA). MLVA was performed using seven C. difficile markers A6 Cd , B7 Cd , C6 Cd , E7 Cd , F3 Cd , G8 Cd , and H9 Cd as previously described, [29] with some modifications. Briefly, genomic C. difficile DNA was isolated using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The repeats were amplified with respective primer-pair using a single PCR protocol. The amplification reactions were performed in a 25-ml final volume containing 16PCR buffer, 0.2 mM of each deoxynucleoside triphsphate (GeneTeks BioScience, Taipei, Taiwan), 1 mM of each primer, 0.5 unit HotStar Taq DNA polymerase (Qiagen, Hilden, Germany), and 2.5 ml of DNA. An initial denaturation step at 95uC for 15 min was followed by 35 cycles of denaturation at 95uC for 30 sec, primer annealing at 52uC for 30 sec, and extension at 72uC for 30 sec. A final extension step at 72uC for 10 min was added, and the product was stored at 4uC until used. PCR fragments were analyzed using the QIAxcel DNA screening kit (Qiagen) on an HAD-GT12 eGene capillary electrophoresis system (Qiagen, Hilden, Germany) with an internal QX DNA size marker 15 bp-3 kb (Qiagen, Hilden, Germany). The size of each C. difficile marker was determined using the software supplied for the electrophoresis apparatus. To verify accurate repeat number assignment, each marker from a selected number of isolates was sequenced. Repeat numbers at each of the seven C. difficile markers were concatenated to generate an MLVA type for each isolates. Repeat numbers per locus were entered into BioNumerics software v6.0 (Applied Maths, Texas, USA) for cluster analysis. A dendrogram was constructed using the unweighted-pair group method with arithmetic mean clustering (UPGMA), using the Pearson correlation coefficient.

Antimicrobial Susceptibility
The minimum inhibitory concentration (MIC) of the 110 toxigenic C. difficile isolates against metronidazole and vancomycin was determined using the standard agar dilution method according to the Clinical and Laboratory Standards Institute (CLSI) guidelines. [30] The antibiotics and concentrations used were as followed: 0.25-16 mg/L for vancomycin and 0.5-32 mg/ L for metronidazole. The breakpoints for vancomycin and metronidazole were: susceptible, !2 mg/L; resistance, .2 mg/ L, according to the EUCAST breakpoints. [31].

Statistics Method
Statistical analyses were performed using Stata (version 11) software (StataCorp LP, USA). The annual number of various toxigenic C. difficile isolates by year was analyzed using the Cochran-Armitage test for trend. A P value of ,0.05 was considered statistically significant.

Annual Numbers and Proportions of Various Toxigenic C. Difficile Isolates
A total of 110 non-toxigenic C. difficile isolates from 110 different patients were studied. Of the 110 patients, 46 were female and 64 were male. Patient ages ranged from four months to 92 years old. Eleven (10%) patients were under the age of 10, and 81 (74%) patients were older than 60 years of age. Of the 110 toxigenic C. difficile isolates tested, 70 isolates harbored tcdA and tcdB (A + B + ) and 40 isolates harbored tcdB only (A 2 B + ). The relationship of coexistence between the toxin genes tcdA and tcdB and the binary toxin genes cdtA and cdtB was described in Table 1. Among these, 4 (3.6%) isolates were positive for all four genes ( Table 1). The annual numbers and proportions of the 110 toxigenic isolates during the period between 2002 and 2007 are shown in Fig. 1

Resistance Pattern of C. Difficile Toxin-producing Strains to Metronidazole and Vancomycin
The MICs of the 110 toxigenic C. difficile isolates against the antibiotics metronidazole and vancomycin were determined. The MIC range for metronidazloe was 0.5 to .32 mg/L and 0.25 to .16 mg/L for vancomycin. As shown in Table 2

Clonality of C. Difficile Isolates
To address whether the increasing number is due to clonal spreading of C. difficile isolates, the MLVA types of the 110 toxigenic C. difficile isolates were analyzed. A total of 109 different MLVA types were identified, in which A + B + isolates and A 2 B + isolates were differentiated into two clusters with a similarity of 17.6% (Figure 2). When isolates with 85% similarity were put into MLVA group, 22 different groups were identified among the 70 A + B + isolates, while 10 different MLVA groups were found among the 40 A 2 B + isolates (Table 3). Among the A 2 B + C. difficile isolates, a major MLVA group (group 1) was identified in 24 (60%) isolates and persisted throughout the study period. The highest numbers were identified in 2004 (8 isolates) and 2007 (8 isolates). In contrast, diverse MLVA types were identified in the 70 A + B + isolates, three major MLVA groups (groups 1, 2, and 3) were identified and accounted for 27 isolates (38%) ( Table 3).
Regarding to the four A + B + CDT + isolates, two toxinotype V/ tcdC-A isolates were genetic related and had a similarity of 97% (MLVA types 66 and 67); these isolates were identified in 2006 (Fig. 2). In contrast, two toxinotype III/tcdC-sc1 isolates identified in 2007 were genetically unrelated and belonged to MLVA types 15 and 27 (Fig. 2). All the four isolates were found to be genetically different to the hypervirulent ribotype 027 and ribotype 078 isolates in comparison with the MLVA results described previously [32,33]. Regarding the seven resistant isolates, the MLVA results indicated that the two metronidazole-resistant A + B + isolates identified in 2003 were genetically unrelated (MLVA types 14 and 55). In contrast, there were five vancomycin-resistant isolates, including three A + B + isolates identified in 2007 and two A 2 B + isolates identified in 2003 and 2006. Two of the three vancomycinresistant A + B + isolates were closely related and had a similarity of 94% (MLVA types 43 and 44). The other one A + B + isolate and two A 2 B + isolates were genetically unrelated (MLVA types 31, 102 and 109).

Discussion
The proportions of the 110 toxigenic C. difficile isolates over the 6-years differed from year to year, and an increasing number of A + B + isolates was observed over this period (Fig. 1). Further MLVA analysis of these toxigenic C. difficile isolates showed that an increasing number of genetic divergences between the C. difficile A + B + isolates (Table 3). In addition, the percentage of A 2 B + isolates reached 73.3% in 2004. One major MLVA group (MLVA-group 1) accounted for 60% of the 40 A 2 B + isolates. A clonal dissemination of the A 2 B + isolates similar to the outbreaks identified in Japan, Korea, Canada, and Poland was identified in our hospital. [34,35,36,37] After 2004, the proportion of A 2 B + isolates began to decrease and dropped to 23.9% in 2007. C. difficile A 2 B + isolates appeared to be an endemic strain to our hospital and is worthy of closely monitoring.
Characterization of the phylogenic relatedness of the C. difficile isolates by toxinotypes was achieved using PCR-RFLP for PaLoc and by directly DNA sequencing the tcdC gene. The results from both methods were essentially concordant. Of the studied isolates, toxinotypes III, V, and VIII were consistently associated with tcdC genotypes tcdC-sc1, tcdC-A and, tcdC-sc7, respectively (Table 1). In this study, all A-B+ isolates belonged to toxinotype VIII and tcdC-sc7 genotypes. The results were similar to those of previous reports. [38,39] Earlier reports indicated that the tcdC gene involves the negative regulation of tcdA and tcdB expression. [5] Many C. difficile isolates defective in tcdC were reported and grouped as ribotype 027 and toxinotype III. For example, the tcdC-A genotype which contains a nonsense mutation (C184T) and a 39-bp deletion from nucleotides 341 to 379, encodes a truncated 61-amino-acid TcdC protein. [28] Additionally, the tcdC-sc1 genotype, which contains a single deletion of nucleotide A117 and an additional 18-bp deletion from nucleotides 330 to 347, produces a truncated 65amino-acid TcdC protein. [40] These two strains with truncated tcdC generated nonfunctional TcdC and are responsible for the increased toxin production and virulence of C. difficile strains. [40] However, several studies reported contradictory results. Previous reports indicated that deletion or truncation of the tcdC gene was often found in toxigenic C. difficile but lacked association with disease severity. [41,42] In the present study, four isolates carrying all four tcdA, tcdB, cdtA and cdtB genes also harbored a 18-bp deletion (tcdC-sc1) or a 39-bp deletion (tcdC-A) (Table 1). However, no serious clinical symptoms were observed in the four patients (data not shown). Because the number of such isolates was too low to make any suggestions or conclusions, the relationship between the tcdC deletions and the development of more severe C. difficile diseases still awaits further investigation. In addition, the four isolates are genetically different to the hypervirulent ribotype 027 and ribotype 078 strains in comparison with the MLVA results of ribotype 027 and ribotype 078 strains from published references. [32,33] (data not shown) It appeared that there remains no evidence for the existence of the hypervirulent NAP1/027 strain in Taiwan.
Antimicrobial susceptibility testing identified seven resistant isolates (6.4%) including two metronidazole-resistant and five vancomycin-resistant isolates. Most isolates represented sporadic case, with the exception of two genetically related vancomycinresistant A + B + isolates indentified in 2007. Although the percentage of drug resistance among the C. difficile isolates was not high compared to previous reports, [19] emerging drug resistance in the toxin-producing isolates, especially in the increasing A + B + isolates, warrants concern. However, because the isolates for antimicrobial susceptibility testing were retrospectively retrieved from the bacteria bank, some metronidazole-heteroresistant populations may not be detectable using the agar dilution method. [19] Therefore, the number of metronidazole-resistant isolates may be underestimated.
In conclusion, the changing trend of various toxigenic C. difficile isolates was studied. Results indicated a persistence of MLVA group 1 A 2 B + isolates and an increase of A + B + isolates with diverse MLVA types between 2002 and 2007. Some C. difficile isolates with antimicrobial resistance to metronidazole or vancomycin have been identified. Continuous monitor is warranted to understand the developing situation and to control the further spread of such infections, especially among hospitalized patients.