Escherichia coli O157:H7 in Feral Swine near Spinach Fields and Cattle, Central California Coast

We investigated involvement of feral swine in contamination of agricultural fields and surface waterways with Escherichia coli O157:H7 after a nationwide outbreak traced to bagged spinach from California. Isolates from feral swine, cattle, surface water, sediment, and soil at 1 ranch were matched to the outbreak strain.

R ecent experimental and epidemiologic studies suggest that domestic pigs are biologically competent hosts and a potential reservoir of Escherichia coli O157:H7 (1,2). Cattle are considered the primary reservoir of E. coli O157, but fecal shedding by other domestic livestock and wildlife has been described (3,4). E. coli O157 was isolated from a wild boar in Sweden, but there is limited information on its occurrence in feral swine in the United States (5). We report fi ndings from an environmental and laboratory investigation after a nationwide spinach-associated outbreak of E. coli O157 in which the outbreak strain was isolated from feral swine and other environmental samples.

The Study
In September 2006, an outbreak of E. coli O157 was linked to consumption of fresh, bagged, baby spinach, with 26 states and Canada reporting 205 cases of illness and 3 deaths (6). Contaminated product was traced to 1 production date ( August 15, 2006) at 1 processing plant and fi elds located on 4 ranches on the central California coast (7).
The outbreak strain was isolated initially from cattle feces collected on September 27, 2006, ≈1 mile from an implicated spinach fi eld on a ranch (ranch A) where numerous free-roaming feral swine were observed. We investigated potential involvement of feral swine in E. coli O157 contamination of spinach fi elds and surface waterways.
Feral swine were live-captured in traps or hunted and humanely killed during October-November 2006. Two feral swine corral traps were placed 1.4 km apart, and 1.7 km (trap 1) and 1.2 km (trap 2), respectively, from the implicated spinach fi eld ( Figure 1). Photographs from digital infrared remote-sensing cameras (Recon Outdoors, Huntsville, AL, USA) were used in combination with sightings and live-capture to ascertain the minimum number of individual feral swine present on the ranch (8). The average population density was calculated on the basis of an estimate of the area sampled by both traps and the estimated mean home range (1.8 km) for feral swine in mainland California by using ArcView version 9.2 (Environmental Systems Research Institute, Redlands, CA, USA) (8).
E. coli O157 was cultured from 45 (13.4%) of 335 samples, including cattle and feral swine feces, feral swine colonic feces from necropsy, surface water and sediment, and pasture soil ( Table 1). The eaeA, hlyA, and stx2 genes were present in all strains, and the stx1 gene was found in only 1 sample (subtype 5; Table 2, Figure 2). Isolates from 28 environmental samples at ranch A were indistinguishable from the major spinach-related outbreak strain by PFGE (Table 1). In contrast, E. coli O157 isolates from 3 other ranches implicated by traceback did not match the outbreak strain. Molecular typing by MLVA provided higher resolution discrimination between environmental strains ( Figure  2). Three major MLVA clusters from ranch A and the surrounding watershed were identifi ed. The cluster containing the outbreak strain (subtype E) is shown in Figure 2, and 16 other highly related subtypes were indistinguishable by PFGE (Table 2).
Ranch A is located in the central coast foothills of San Benito County, where the dominant habitat is coastal oak woodland interspersed with dense riparian vegetation near seasonal waterways (Figure 1). Approximately 2,000 range cattle were grazed on the ranch. Spinach and other leafy green vegetables were grown on a leased portion of the property that was separated from cattle pastures by wire mesh fence. Well water was used for irrigation. No evidence of cattle manure-based fertilizer application, runoff from cattle pastures, or fl ooding from surface waterways (based on topography) onto the implicated spinach fi eld was found during the investigation (7).
Feral swine were the most abundant wildlife observed on ranch A, and evidence of intrusion, including tracks, rooting, or feces in crop fi elds and adjacent vineyards, was documented ( Figure 1). Birds, black-tailed deer, cottontail rabbits, coyotes, and ground squirrels also were observed, but the population density of these species appeared lower, and their activity was confi ned mostly to rangeland areas according to visual observations. Swine visited the traps almost continuously from dusk until dawn with peak activity between 5:00 PM and midnight. An average of 3.6 swine/ trap/night were live-captured. The estimated population density was 4.6 swine/km 2 (95% confi dence interval [CI] 3.8-5.9), and the actual number of feral swine on ranch A was estimated to be 149 animals (95% CI 124-192) ( Figure  1). Feral swine used livestock rangelands and gained access to adjacent crop fi elds through gaps formed at the base of the fence by erosion and rooting. Cattle and feral swine had access to and congregated at surface waterways on the ranch (Figure 1). ‡Included feces from coyote (n = 1), deer (n = 4), dog (n = 1), horse (n = 2), sheep/goat (n = 3, composite), waterfowl (n = 2), unknown species (n = 11), and owl (n = 2). §Surface water (rivers, streams, ponds) was sampled by collection of 100-mL grab samples or placement of a modified Moore swab for 4-5 d. ¶Well water was sampled from 3 wells or sprinkler heads by collection of 100-mL or 1,000-mL grab samples or by concentration of 40,000 mL to 500 mL by using ultrafiltration (7).

Conclusions
We describe the fi rst, to our knowledge, isolation of E. coli O157 from feral swine in the United States. The percentage of specimens positive for E. coli O157 among feral swine (14.9%) and cattle (33.8%) and the density (4.6 swine/km 2 ) were high compared with results of previous ecologic studies (Table 1) (2)(3)(4)(5)8,14,15). Molecular typing of isolates by PFGE and MLVA showed possible dissemination and persistence of the outbreak strain in multiple environmental samples as long as 3 months after the outbreak (Tables 1, 2). MLVA is more reproducible than PFGE and better at discriminating between closely related E. coli O157 isolates (10,12,13). Recovery of related E. coli O157 subtypes by both methods suggested swine-to-swine transmission, interspecies transmission between cattle and swine, or a common source of exposure such as water or soil (Table 2, Figure 2).
Mechanisms of in-fi eld contamination of leafy greens for this and previous outbreaks remain unclear, but hypotheses have emerged. A relatively high density of feral swine near cattle and spinach fi elds could represent a risk factor for E. coli O157 contamination. Wildlife may be sentinels for E. coli O157 in the produce production environment, or they may be vectors involved in the contamination of plants directly by fecal deposition or indirectly by fecal contamination of surface waterways or soil. Notably, baby spinach is harvested with a lawn mower-like machine that could pick up fecal deposits in the fi eld and thereby contaminate large volumes of product during processing. Fecal loading of surface waterways by livestock and wildlife with subsequent contamination of wells used for irrigation represents another possible route of transmission to plants in the fi eld. Although E. coli O157 was not detected in irrigation water, older agriculture wells at ranch A appeared vulnerable to contamination by surface water (R. Gelting, pers. comm.). Unrecognized environmental and management practices during preharvest and postharvest processing also could have contributed to amplifi cation and dissemination of E. coli O157 in raw spinach.
In summary, E. coli O157 contamination of spinach and other leafy greens is likely a multifactorial process. Additional research is needed to develop and implement effective risk assessment and management practices. For example, studies are needed to determine colonization potential of and levels of fecal shedding by feral swine, and the importance of interspecies transmission to other vertebrate or invertebrate (e.g., fl ies) populations near agricultural fi elds.   (Table 2). Numbers between circles represent summed tandem-repeat differences between MLVA types (10). The shaded areas (red, green, and blue) denote genetically related clusters with MLVA differences <3. Red circles indicate types comprising isolates that were indistinguishable from the spinach-related outbreak strain (subtype E) by pulsed-fi eld gel electrophoresis (PFGE).