Shiga-toxigenic Escherichia coli O157:H7 (EcO157) is responsible for over 96,000 diarrheal illnesses and 3200 hospitalizations annually in the United States (Scallan et al., 2011). One-fourth of the EcO157 associated outbreaks during 1998 to 2008 were linked with the consumption of contaminated leafy vegetables, fruits and nuts (Gould et al., 2013). Major outbreaks associated with produce indicate that pre-harvest contamination has occurred often, so it is critical to identify sources of pathogens (Mandrell, 2009). EcO157 responsible for 200 infections in a large multi-state outbreak (Mandrell, 2009) related to consumption of baby spinach was traced back to produce grown in central California coast (Jay et al., 2007). Indeed, EcO157 isolates that are genetically indistinguishable from the 2006 spinach outbreak strain were isolated from feral swine, cattle, surface water, and sediment samples near spinach fields (Jay et al., 2007). These results indicate wide-spread occurrence of this pathogen in numerous habitats, in the vicinity of produce production, each of which may affect survival of EcO157 differently. This produce growing region has been termed “The salad bowl of the US” since produce from this region are widely distributed to the entire US. Conversely, outbreaks have frequently been traced-back to the Salinas region. Therefore, environmental pathogens isolated from the Salinas region have the potential to impact public health throughout the US.

Although most EcO157 strains disappear rapidly from soil, manure and dairy wastewater environments (Avery et al., 2004; Berry and Miller, 2005; Ravva et al., 2006), a small proportion of cells remain viable for extended periods (Duffitt et al., 2011). Therefore, it is probable that most EcO157 cells within a population exposed to stressful environments outside the animal host fail to survive (Jones, 1999). Nevertheless, survival of some cells in water and soil results in transport of pathogen by surface water, irrigation and wind, possibly leading to produce contamination.

Survival of specific strains of EcO157 in sufficient numbers to cause infection is associated with their intrinsic fitness traits and genetic makeup (Topp et al., 2003). One such trait is the formation of biofilms, facilitated by the production of adhesins and exopolysaccharides (Cookson et al., 2002; Ryu and Beuchat, 2005). One such adhesin, curli, acts as a scaffolding agent in biofilms (Evans and Chapman, 2014) and is responsible for cell surface and cell to cell interactions (Reichhardt et al., 2015), immune activation, cell invasion and favor intestinal colonization (Saldaña et al., 2009; Biscola et al., 2011; Blanco et al., 2012; McWilliams and Torres, 2014). Curli are thin extra-cellular, proteinaceous aggregative fimbriae that act as a virulence factor by promoting attachment to eukaryotic cells (Uhlich et al., 2002; Kikuchi et al., 2005). Curli fimbriae, encoded on two divergently transcribed operons csgDEFG and csgBAC, are expressed in response to environmental stress factors such as low temperature, low oxygen, low osmolarity, and nutrient limitation (Barnhart and Chapman, 2006; Blanco et al., 2012).

Curli fibers are expressed at temperatures ranging from 25 to 37°C (Kikuchi et al., 2005; Barnhart and Chapman, 2006; Hung et al., 2014). Curli-expressing strains of EcO157 developed stronger association with leaf surfaces of produce (Patel et al., 2011; Fink et al., 2012; Macarisin et al., 2012) and E. coli strains isolated from plants appear to form significantly more curli compared to strains from humans or animals (Méric et al., 2013). In addition, curli producing strains significantly adhere to human epithelial cells compared to curli-deficient strain and form three-dimensional mature biofilms (Kikuchi et al., 2005). Biofilm formation by EcO157 isolates strongly correlated with curli production (Lee et al., 2011; Wang et al., 2012; Macarisin et al., 2014). Curli production in uropathogenic E. coli influences pathogenesis in vivo (Lim et al., 2014) and curli produced in the host provided a fitness advantage.

Subpopulations of EcO157 have been reported to adapt to harsh environmental conditions (Brzuszkiewicz et al., 2009) and environmental isolates appear to resist acidic conditions in comparison to clinical isolates from outbreaks (Oh et al., 2009). It is also known that curli-deficient variants of EcO157 survive better than curli producers under acidic conditions comparable to that of the stomach of the host (Carter et al., 2011). Curli-deficient strains and variant sub-populations persist longer in soil and resist protozoan predation compared to curli-expressing strains (Ravva et al., 2014a,b). Furthermore, the proportion of curli-expressing phenotype sub-populations of EcO157 decreased on repeated soil exposure while curli-deficient sub-populations increased. In fact, a curli-deficient strain of EcO157 was repeatedly isolated during a 45-day period from dry pasture soil indicating it's resilience and fitness in austere soil environment (Ravva et al., 2014a). In addition, the defense systems of certain plants like Arabidopsis recognize curli producing variants and adversely influence their survival and consequently curli-deficient EcO157 strain survive better on plants (Seo and Matthews, 2012). Thus, it is possible that environmental conditions aid in the enrichment of subpopulations with enhanced fitness.

Exposure to harsh environments may result in the development of stress resistant and/or virulent phenotypes. Our previous results indicated a possibility that soil exposure resulted in enriching a curli deficient subpopulation for survival (Ravva et al., 2014a). Consequently, it is likely other environmental conditions, particularly agricultural environments, also influence the survival of curli subpopulations. Here, we characterize phenotype differences in curli expression of 471 strains of EcO157 isolated from different environments (sample types) and associated with major outbreaks, sporadic infections and the fresh produce production environment of the central coast of California. These strains represent 328 phylogenetically distinct genotypes and were also compared to determine if genotypic differences influence the fitness selection of curli sub-populations.

Predicting the likely occurrence of foodborne outbreaks linked with the consumption of fruit and vegetable produce is extremely important but difficult in practice, as the survival, proliferation, and transport of the pathogen are all controlled by both multiple environmental and biological factors. Thus, survival of EcO157 has been reported to vary from nearly 2 years in manure piles (Kudva et al., 1998) to less than a day in dairy wastewater (Ravva et al., 2006). Likewise, it is intriguing that even a strain that survived on apples and in apple juice in sufficient number to cause an outbreak in 1996 also rapidly disappeared in a day in manure wastewater. These results indicate that EcO157 strains respond differently to different environmental conditions. However, strain specific factors are also important in fitness of EcO157 in human and animal host and in the environment. Strain specific differences in curli (Parker et al., 2012), biofilm formation (Wang et al., 2012), and shiga-toxin expression (Neupane et al., 2011) influence environmental fitness and may contribute to the likelihood of an outbreak.

Fecal isolates showed a uniform distribution of curli expression (Figure 1, Table 2). This situation was especially noticeable where multiple fecal strains from the same outbreak were analyzed. Some outbreaks associated with predominantly curli-deficient strains were caused by the consumption of contaminated water (2000), apple juice (Odwalla outbreak of 1996), lettuce (1999, 2002, and 2006), and ground beef (2002). Yet, in these cases, the sampling of available strains from each outbreak was rather small, limiting the conclusions that can be made. Significantly, the large collection of 2006 spinach outbreak strains showed nearly 50% curli expression. Likewise, the large collection of unrelated sporadic strains gave a similar result, supporting the suggestion that curli expression in these strains is more dependent on sample source than on genotype. Curli expression in bovine and feral pig fecal strains similarly showed a mixture of curli deficient and curli prevalent phenotypes, though the curli deficient group was at a significantly higher percentage, compared to human fecal strains. It is noteworthy, that the sampling methods for animals and humans were substantially different (unpublished results). In many cases, animal fecal samples were collected from the ground and, as such, were exposed to the environment for a period of time and may have experienced considerable stress. Consequently, these strains may be substantially more “environmental” than other fecal isolates collected directly from the gut.

Although curli fimbriae promote adherence to surfaces (Barnhart and Chapman, 2006) and help to colonize animal tissues (Uhlich et al., 2002; Rosser et al., 2008), it is remarkable to find more than half of clinical strains from outbreaks and sporadic cases of infections are curli-deficient. Nevertheless, it is possible that curli plays a critical role in biofilm formation in different environments (Saldaña et al., 2009; Wang et al., 2012). For instance, curli-deficient variants are known to be acid-resistant (Carter et al., 2011) and it is likely that the acid sensitive curli-producers are eliminated in the stomach of the animal host. Thus, the results presented here indicate that a lower percentage of curli expressing strains are found in animal feces and environmental samples. Consequently, selection for curli expressing strains may occur in the gut of cattle and humans, and perhaps, to a lesser degree in other species. Additionally, some plants appear to recognize curli and induce defense systems. Hence, pathogens expressing curli survive poorly on plant surfaces (Seo and Matthews, 2012). However, csgA (a gene encoding curli biosynthesis) mutants of an EcO157 strain showed no reduction in their ability to bind to alfalfa sprouts, suggesting that the pathogen may have more than one mechanism for binding to plant surfaces (Jeter and Matthysse, 2005). These results indicate that curli is part of a collection of various fitness capabilities selected to survive multiple, stressful environments.

Except for some strains from bagged-spinach and food isolates, 82% of environmental strains are curli-deficient (Figure 1, Table 3). The prevalence of curli-deficient strains from the environment was predicted as some curli-deficient strains survived longer in soils and some of the same strains also resisted predation by protozoa in dairy waste water (Ravva et al., 2014a,b). Furthermore, curli-deficient clonal sub-populations of both clinical and environmental strains survived longer in soil compared to the curli-positive clones of the same strains (Ravva et al., 2014a). These conclusions are further supported here with additional strains from soil, but also strains from water and sediment (Figure 1, Table 3). Thus, curli-deficient strains and curli-deficient sub-populations have fitness traits for survival in harsh produce production environments.

We observed intra-genotype differences in proportion of curli phenotype expression, possibly based on environmental exposure or residence in human or animal host. The 2006 spinach outbreak strains, indistinguishable from each other by MLVA genotyping, differ substantially in phenotypic expression. Half of the clinical strains of MLVA type 163 failed to express curli while isolates of the same genotype from cow feces exhibited phenotypes ranging from no expression of curli to 100% clones that are curli (Figure 1, Table S1). Several other MLVA genotypes showed variation in curli expression, but these genotypes failed to cluster, possibly indicating that variation of curli expression is not linked to specific genotypes. However, one branch of the minimal spanning tree of environmental strains (Figure 1) showed preferential curli expression. It is also noteworthy that the sample locations of 8 of these strains were within a 31 km region in Salinas and during a 4 month period (data not shown). Consequently, their close phylogenetic and temporal/spatial relationship may indicate transport and variation of a strain from a single point in the environment. Importantly, this instance of potential genetic control of curli may be only apparent within the background of relatively invariant environmental isolates. Since, the majority of fecal strains show variable curli expression, the genotype effects may have been masked by influence of the environment in the gut. Although it needs further validation using many more genotypes, from this same cluster and others, one can predict that certain genotypes may preferentially express curli.

In summary, these results expand on previous observations contrasting curli expression in fecal and more austere environmental isolates, to include a variety of domestic and wild animal fecal strains and a large group of water, sediment, and soil isolates from a major agricultural region of California. We have demonstrated that curli expression of clonal subpopulations depends primarily upon the type of environmental exposure and the isolation source, although genetic differences also contribute somewhat to clonal variation in curli expression. Further experiments will be necessary to quantify these effects as it relates to degree and duration of exposure to these environments. Nevertheless, the curli-deficient phenotype may be a selective trait for survival and proliferation in austere produce production environments.

SR designed and conducted experiments, analyzed, and wrote the manuscript. CS and MC conducted experiments, analyzed, and contributed to the manuscript preparation.