Characterization of Clostridium difficile Strains in British Columbia, Canada: A Shift from NAP1 Majority (2008) to Novel Strain Types (2013) in One Region

Background. Clostridium difficile is a major cause of gastrointestinal illness. Epidemic NAP1 strains contain toxins A and B, a deletion in repressor tcdC, and a binary toxin. Objectives. To determine the molecular epidemiology of C. difficile in British Columbia and compare between two time points in one region. Methods. C. difficile isolates from hospital and community laboratories (2008) and one Island Health hospital laboratory (2013) were characterized by pulsed-field gel electrophoresis, PCR-ribotyping, toxin possession, tcdC genotype, and antimicrobial susceptibility. Results. In 2008, 42.7% of isolates had NAP1 designation. Hospital-collected isolates were associated with older patients and more NAP1 types. Unlike other isolates, most NAP1 isolates possessed binary toxin and a 19 bp loss in tcdC. All isolates were susceptible to metronidazole and vancomycin. A 2013 follow-up revealed a 28.9% decrease in NAP1 isolates and 20.0% increase in isolates without NAP designation in one region. Then, community-associated cases were seen in younger patients, while NAP types were evenly distributed. Isolates without NAP designation did not cluster with a PFGE pattern or ribotype. Conclusions. Evaluation of C. difficile infections within British Columbia revealed demographic associations, epidemiological shifts, and characteristics of strain types. Continuous surveillance of C. difficile will enable detection of emerging strains.


Introduction
Clostridium difficile infections (CDI) are considered a top priority by Canadian and American healthcare agencies based on risk assessments of pathogens associated with antimicrobial resistance [1,2]. CDI are characterized by diarrhea, fever, nausea, and abdominal pain and in severe cases can progress to toxic megacolon, sepsis, and death [3]. Risk factors for development of CDI include prior antimicrobial use, exposure to healthcare settings, underlying illness, and being over 65 years of age [4,5]. In the early 2000s, outbreakassociated fluoroquinolone-resistant strains classified as B1 by restriction enzyme analysis, North American pulsotype 1 (NAP1) by pulsed-field gel electrophoresis (PFGE), and

Clinical Samples.
CDI was defined by the presence of diarrhea and a positive clinical laboratory test. Hospital and community laboratories prospectively forwarded toxins A and/or B positive patient stool specimens or recovered C. difficile isolates to the British Columbia Public Health Laboratory for isolate confirmation, typing, and susceptibility testing. The initial study period was from March 1 to 31, 2008, and included 13 hospital laboratories, which serve mainly inpatients, and one province-wide community laboratory that serves only outpatients. These laboratories represent the majority of CDI testing within British Columbia. In 2008, all laboratories used enzyme immunoassays to detect glutamate dehydrogenase (GDH), a marker of C. difficile, and toxin A or B. Where confirmation was required (the clinical presentation was consistent with CDI but only the GDH result was positive or both markers of C. difficile were negative), C. difficile and toxins were detected following culture or by cell cytotoxin assay.
This study was repeated from March 3 to July 7, 2013, but only with one large hospital laboratory (LabA) that screens both hospital and community samples within the Island Health Authority for CDI by enzyme immunoassay (C. diff Quik Chek Complete; TechLab, Blacksburg, VA). This laboratory (LabA) routinely processes approximately 80% of all hospital samples and 40% of all outpatient samples originating from the region the authority covers, a region containing both rural and urban communities that represents approximately 17% of British Columbia's population. In 2013, toxin B PCR (GeneOhm Cdiff Assay; Becton, Dickinson and Co., Franklin Lakes, NJ) was used as a confirmatory test (versus cytotoxicity assay in 2008). During the 2013 study, medical record review discriminated between inpatients with CA-CDI and HA-CDI per established guidelines [27]. In the absence of other data, outpatients ( = 20) and inpatients characterized as having CA-CDI ( = 10) were collectively referred to as CA-CDI cases. Study approval was obtained from the University of British Columbia Ethics Review Board.

Bacterial
Culture. C. difficile isolates were cultured on selective cycloserine cefoxitin agar containing 5% sheep blood following anaerobic incubation at 35 ∘ C for 48-72 hours. Where C. difficile was not initially recovered, alcohol shock treatment was performed. Typical colonies were purified on Columbia agar with 5% sheep blood (Oxoid, Nepean, ON) and C. difficile was identified by colony morphology, aerotolerance, and latex agglutination (C. difficile Latex Agglutination Kit; Microgen, Frederick, MD).

Molecular Typing.
Pulsed-field gel electrophoresis (PFGE) was performed by digestion using 40 U SmaI (New England BioLabs, Ipswich, MA) based on a previously described method [33]. Bacteria within plugs were lysed at 55 ∘ C overnight with lysis buffer (content description elsewhere [33]). Plugs were then washed at room temperature 4 times at 20 minutes each, prior to digestion for 2 hours at 25 ∘ C. DNA fragments were separated using a Chef-DR III mapper system (Bio-Rad, Hercules, CA) as per established run conditions [33]. Resultant patterns were visualized on a Molecular Imager Gel Doc XR System (Bio-Rad) following gel staining with ethidium bromide and analyzed with BioNumerics software version 6.1 (Applied Maths, Austin, TX) as described previously [34] using XbaI digested Salmonella braenderup as a reference. PFGE types were assigned based on a similarity of ≥80% to NAP1-12 isolates within the Canadian Nosocomial Infection Surveillance Program C. difficile Database.
PCR-ribotyping was performed with high-resolution capillary gel-based electrophoresis following amplification of the 16S-23S intergenic spacer (IGS) region as previously described [35]. Briefly, amplicon fragments were identified with an Applied Biosystems 3130xl Genetic Analyzer using a capillary array in reference to GeneScan 1200 LIZ Size Standard (all Life Technologies, Grand Island, NY), and peak profiles were analyzed with BioNumerics software version 5.1 (Applied Maths, Austin, TX). Ribotypes were assigned based on profile similarity to 106 reference ribotypes within the National Microbiology Laboratory C. difficile database.
For toxin gene detection, DNA was extracted with Insta-Gene Matrix (Bio-Rad). The presence of tcdA, tcdB, tcdC, and cdtB along with the housekeeping gene tpi was investigated with previously described primers [9]. PCR was carried out with illustra PuReTaq Ready-To-Go PCR Beads (GE Healthcare, Piscataway, NJ) and 3 mM MgCl 2 with an annealing temperature of 57 ∘ C on a Stratagene Robocycler (Agilent Technologies, Santa Clara, CA). Deletions in tcdC were detected by fragment size analysis following amplification with previously described C1 and C2 primers [36] (with C1 labeled with FAM) as described for toxin detection above. Denaturation occurred at 94 ∘ C for 5 minutes, followed by 30 cycles of 1 minute at 94 ∘ C, 1 minute at 56 ∘ C, and 1.5 minutes at 72 ∘ C. Final extension occurred for 10 minutes at 72 ∘ C. Amplicon size was determined using an Applied Biosystems 3130xl Genetic Analyzer in reference to GeneScan 1200 LIZ Size Standard. ATCC 9689 strain (locus tag B131 00660 on NCBI GenBank accession number NZ AQWV01000030) was used as a control.

Antimicrobial Susceptibility
Testing. Susceptibility to clindamycin, metronidazole, and vancomycin was evaluated using E-test strips (bioMérieux, Saint-Laurent, QC) on BBL Brucella agar with 5% horse blood, hemin, and vitamin K 1 (Becton, Dickinson and Co.) as previously described [37]. Briefly, bacteria were suspended in Brucella broth (Becton, Dickinson and Co.) prior to agar inoculation, and MICs were read after 48 hours of growth with Bacteroides fragilis ATCC 25285 and Staphylococcus aureus ATCC 29213 as controls. MIC interpretation for clindamycin and metronidazole was based on Clinical Laboratory Standards Institute (CLSI) breakpoints [38], while susceptibility to vancomycin was determined with European Committee on Antimicrobial Susceptibility Testing breakpoints [39] in the absence of established CLSI criteria.

Statistical Analysis.
Differences between years were determined by unpaired Student's -test. Differences among NAP types were determined by one-way analysis of variance (ANOVA) and Tukey's multiple comparisons test. Differences in the distribution of NAP types were determined by Pearson Chi-Square test. GraphPad QuickCalcs Software (La Jolla, CA) was used for analyses.  Table 1. Most (93.8%) NAP1 isolates possessed genes for toxin A, toxin B, and binary toxin CDT as well as a 19-bp loss in the tcdC gene, the latter suggestive of the described 18-bp deletion and associated single-base-pair deletion [33]. All NAP7 isolates possessed genes encoding toxins B and CDT and a 40 or 41 bp loss in tcdC. Of the other isolates, 105 (91%) possessed only toxins A and B genes and 55 (75.3%) had wild type size tcdC.

2008 British Columbia
All C. difficile isolates were susceptible to metronidazole and vancomycin, while 31.3% were susceptible to clindamycin ( Table 2). NAP1 isolates were least susceptible to metronidazole compared with isolates of other NAP types and those without NAP designation (with data from the latter combined, mean MIC 0.42 versus 0.13 g/mL) ( < 0.0001), while NAP2 isolates were least susceptible to clindamycin by the same analysis (mean MIC 189.2 versus 24.1 g/mL) ( < 0.0001).     Table 3. In 2013, most isolates ( = 30, 44.1%) had a PFGE pattern not matching a described NAP type. The proportion of this diverse set of isolates was high in CA-CDI (50.0%) and HA-CDI (39.5%) cases, and isolates were distributed among 22 different PFGE patterns (range 1-3 isolates/pattern). The most common pattern, 0037, was identified in only 3 (10.0%) cases. Three of these non-NAP type patterns were also identified in 2008 isolates from LabA. In 2008, isolates not matching a described NAP only accounted for 24.1% of isolates ( = 13) and were distributed among 9 different PFGE patterns (range 1-2 isolates/pattern). Of the NAP types, NAP1 ( = 8, 11.8%), NAP4 ( = 15, 22.1%), and NAP6 ( = 8, 11.8%) were most commonly identified in 2013. The proportions of NAP1, NAP4, and NAP6 isolates were similar between HA-CDI (together 47.4%) and CA-CDI (together 43.3%) cases. In 2013, the majority ( = 14, 93.3%) of infections with NAP4 isolates occurred in patients over the age of 65, whereas only HA-CDI NAP1 isolates ( = 4, 50.0%) were in patients over 65 years of age. In contrast, in isolates and a 20.0% increase in a diverse set of isolates without NAP designation (other differences were <10%). NAP designation (number) Ribotype (number) NAP1 (8) 027 (6), 153 (1), and none * (1) NAP4 (15) 011 (1) Other characteristics of isolates from 2013 were consistent with those from 2008 (Table 3). Most NAP1 isolates ( = 6, 75.0%) possessed ctdB for binary toxin CDT and a 19 bp smaller tcdC, while the NAP7 isolate did not have a detectable toxin A gene and showed a 40 or 41 bp loss in tcdC, and binary toxin ctdB was absent and tcdC was wild type for most isolates designated by other NAP types or without NAP designation ( = 49, 83.1%). All isolates were susceptible to metronidazole and vancomycin, while 22.1% were susceptible to clindamycin.

Discussion
Evaluation of the molecular epidemiology of C. difficile within British Columbia in 2008 and in 2013 revealed demographic associations, recent epidemiological shifts within one region, and characteristics of non-NAP1 C. difficile strain types. Overall, most cases of CDI were in patients over 65 years of age (67.0%), consistent with this elderly population being at risk [4,5]. Patients presenting to community laboratories were significantly younger than hospital patients, while patients with CA-CDI were significantly younger than those with HA-CDI. These findings align with previous reports of patients with CA-CDI being younger, likely healthier, and having fewer traditional risk factors for CDI than those with HA-CDI [28][29][30][31][32]. NAP1 designation was most commonly identified across British Columbia in 2008. NAP1 strains were endemic in several countries during the 2000s [6], although NAP distribution varied by region with NAP2 being predominant from 2005 to 2007 in the Canadian provinces of Manitoba [34], Ontario [40], and Prince Edward Island [41]. NAP2 and NAP4 isolates were also identified in high frequency in British Columbia in 2008. Consistent with previous reports, the presence of binary toxin, toxin A gene, and tcdC genotype correlated with designated NAP types [34,[40][41][42][43], although outliers did exist. NAP1 isolates were least susceptible to metronidazole [17][18][19][20][21].
NAP distribution shifted between 2008 and 2013 in one region in British Columbia: an increase in a diverse set of isolates without NAP designation and a decrease in NAP1 isolates were observed in the Island Health Authority region. C. difficile strains are naturally diverse and can evolve rapidly into those with increased pathogenicity [44,45]. Recently, several countries have reported the emergence of non-BI/ NAP1/027 C. difficile stains within healthcare settings, including those of ribotype 078 (analogous to NAP7) [46], ribotype 244 [47,48], ribotype 012 [49], and 018 [50]. In the present study, isolates without NAP designation were diverse and did not cluster with a particular PFGE pattern or ribotype, suggesting absence of a dominant emerging strain.
Isolates collected in 2008 from hospital sites had a higher proportion of NAP1 types compared with community sitecollected isolates. While NAP1 strains were found to be more common among HA-CDI than CA-CDI in the US in 2011 [51], this difference was not seen within the one region of British Columbia that was studied here in 2013. Taken together, the high proportion of a diverse set of isolates not matching described NAP types among the 2008 community sitecollected isolates and the 2013 CA-CDI isolates could relate to the transmission of strains acquired from community sources to the healthcare setting. Further analysis involving methods more discriminatory than PFGE, such as whole genome sequencing, could elucidate such a transmission pattern. Indeed, using analyses of patient interactions and whole genome sequencing, a recent study of nonoutbreak cases in a hospital setting in the UK found that distinct sources other than symptomatic CDI cases are involved in C. difficile transmission [52]. Therefore, the observed shift in NAP distribution between 2008 and 2013 in the Island Health Authority region could reflect the steady increase in the proportion of CA-CDI inpatient cases identified, from 11.7% in 2008 to 35.6% in 2013. The observed shifts in NAP distribution could also relate to interventions in infection control or antibiotic stewardship within healthcare settings. Indeed, such interventions did occur in the Island Health Authority between the two study periods; however their descriptions are beyond the scope of the current paper.
Our findings of population shift and strain diversity emphasize the need for continuous surveillance and characterization of circulating C. difficile to aid in detection and control of emerging strains. Our analysis was limited by the lack of case definition as either HA-or CA-CDI in 2008 and inclusion of only a single laboratory in 2013. Conclusions on changes in molecular epidemiology were therefore drawn from comparisons with only a subset of 2008 cases. Furthermore, we were unable to determine true cases of CA-CDI according to established criteria [27] in the absence of patient history (modified definitions were used) and CDI detection methods differed slightly between study years. Study findings are also limited by sampling periods, as CDI incidence can vary by season [53]. Comprehensive surveillance defining CA-and HA-CDI cases will enable investigations of sources of acquisition and transmission. Description of C. difficile isolates in nonoutbreak settings will allow for a more comprehensive understanding of the pathogenicity and associated markers of strains. Characteristics of strains that result in extensive transmission along with increased morbidity and mortality remain to be fully elucidated.

Disclosure
Preliminary results from this study were presented at the Canadian Association for Clinical Microbiology and Infectious Diseases-Association of Medical Microbiology and Infectious Diseases Canada Conference; April 2-5, 2014, Victoria, British Columbia.