Packaged meats were purchased from 3 national-chain grocery stores in the Tucson, Arizona, area on 3 occasions at 1-month intervals from January to April 2007. Products sampled were both uncooked (ground beef, ground pork, ground turkey, pork sausage, and pork chorizo) and ready to eat (beef summer sausage, pork braunschweiger) (Table). Pork chorizo was produced and distributed locally; all other samples were national brands. Products with different sell-by dates (a surrogate for production date) were sampled for each meat type. Samples were not representative of all meat products in each grocery store.

For each sample, 1 g of meat was added to two 10-mL tubes of prereduced brain heart infusion (BD, Franklin Lakes, NJ, USA), which had been supplemented with 0.5% yeast extract (BD), 0.05% DL-cysteine (Sigma-Aldrich, St. Louis, MO, USA), and 0.1% taurocholate (MP Biomedicals, Solon, OH, USA). One tube was heat shocked (80°C, 10 min), and both were then incubated anaerobically at 37°C for 72 h. Aliquots were subcultured onto taurocholate cycloserine cefoxitin fructose agar (TCCFA) (10) and incubated anaerobically for 24–72 h at 37°C. Colonies were subcultured onto anaerobic blood agar, TCCFA (with or without antimicrobial agents), and confirmed as C. difficile by p-cresol odor, yellow-green fluorescence under UV illumination, a positive L-proline aminopeptidase reaction, and negative indole reaction.

Isolates were characterized by PCR ribotyping (11), toxinotyping (3), and pulsed-field gel electrophoresis (PFGE) (12). Presence of tcdA, tcdB, cdtB (binary toxin), and deletions in tcdC was determined by PCR (2).

MICs were determined by Etest (AB Biodisk, Solna, Sweden) on Brucella blood agar with vitamin K and hemin (Remel, Lenexa, KS, USA) that was incubated anaerobically at 35°C. Reference interpretive criteria for C. difficile susceptibility to clindamycin and moxifloxacin were used; MICs for levofloxacin and gatifloxacin were interpreted by using criteria for moxifloxacin (13). Bacteroides fragilis ATCC 25285, B. thetaiotaomicron ATCC 29741, C. difficile ATCC 700057, and Enterococcus faecalis ATCC 29212 were included as controls.

Proportions were compared by χ² or Fisher exact test. Thirty-seven (42.0%) of 88 retail meats yielded C. difficile, including 42.4% of beef, 41.3% of pork, and 44.4% of turkey products (Table). Ready-to-eat products were more commonly culture positive (11/23; 47.8%) than were uncooked meats (26/65; 40.0%), although the difference was not significant (p = 0.34). The highest percentages of C. difficile isolates were recovered from pork braunschweiger (62.5%) and ground beef (50.0%). Culture-positive results came from both heat-shocked and non–heat-shocked cultures, whereas culture-negative specimens were negative in both types of culture, and no specimen was positive by both methods (not shown). No association was found with the meat processor, the sell-by date, the store, or the month sampled (not shown). Multiple independent cultures from 2 braunschweiger samples yielded indistinguishable isolates in the same meat sample (10/10 from 1 package and 12/12 from another; not shown), which suggests that a single strain may predominate when C. difficile is present. Our percentage of recovery of C. difficile from retail meat products is higher than that reported (20%) in a similar study of Canadian ground beef (9), possibly because of differences in culture methods, the meats sampled, or national or geographic variation.

Isolates were grouped into ribotype 078/TT V (27/37, 73.0%) and ribotype 027/TT III (10/37, 27.0%). Strain types were not specific to meat type, store, or sampling month (Table). All isolates were PCR positive for binary toxin (cdtB), tcdA, and tcdB. Characteristic 18-bp and 39-bp deletions in tcdC were present in 027/TT III and 078/TT V isolates, respectively (2,12). PFGE divided 027/ TT III isolates into NAP1 (>80% related to human NAP1) and NAP1-related (78% related to human NAP1) groups and 078/TT V isolates into NAP7 and NAP8 groups (Figure).

Ribotype 027 isolates are described almost exclusively in context of the current human epidemic strain, NAP1/027/TT III (2). In this study, we also found 027/TT III isolates that were only 78% similar to NAP1 (i.e., NAP1-related). Ribotype 078 strains were previously uncommon causes of healthcare-associated CDI in humans (12), but now they are emerging in pigs and calves with diarrhea ( 7; J.S. Weese, pers. comm.) and in persons with CDI (12). Two epidemiologically unrelated 078/TT V isolates from human CDI patients are indistinguishable by PFGE from pig isolates (12).

The 078/TT V isolates were uniformly susceptible to levofloxacin, moxifloxacin, and gatifloxacin. Like human TT V isolates (12), most 078/TT V meat isolates were nonsusceptible to clindamycin (56% resistant, 41% intermediate). This may not be surprising given the widespread use of tylosin, erythromycin, virginiamycin, and lincomycin in food animals and the potential for selection of macrolide-lincosamide-streptogramin resistance (14).

NAP1 isolates have demonstrated high-level resistance to levofloxacin, moxifloxacin, gatifloxacin (>32 μg/mL), and clindamycin (>256 μg/mL), consistent with current human strains (2). NAP1-related isolates were susceptible to levofloxacin, moxifloxacin, and gatifloxacin but resistant to clindamycin, similar to the pattern of historic NAP1 strains (2).

Fluoroquinolones are widely used in human therapy, and the current epidemic strain may have emerged because of its resistance to these agents. Fluoroquinolone use is limited in food animal production (14), with the exception of enrofloxacin for treatment of bovine respiratory disease (now approved for use in swine).

The source of C. difficile in retail meats may involve antemortem deposition of spores in the animal’s muscle or other tissues, fecal or environmental contamination of carcasses, or contamination during processing. Spores could persist in packing plants, resulting in contamination of carcasses or food products during processing. Contamination may also occur in retail meat markets.

Direct or indirect human-to-human transmission is responsible for most healthcare-related CDIs (15) and most likely contributes to CA-CDI. Therefore, stopping such transmission remains the critical control point for preventing most human CDIs. Nonetheless, our findings highlight the potential both for selection of virulent or resistant strains in animals and interspecies transmission through the food supply. Our data do not prove transmission of C. difficile from foods to humans but highlight the need for studies to characterize risks posed by this organism in the human food supply.