Abstract
Escherichia coli serogroup O111 is among the six most commonly reported non-O157:H7 Shiga toxin–producing E. coli (STEC), which are emerging as important foodborne pathogens. We have assembled a collection of environmental and clinical strains of E. coli O111 from diverse sources and investigated various genotypic and phenotypic characteristics of these strains to gain a better understanding of the epidemiology and biology of this serogroup. Sixty-three percent of the strains (24/38) were of H-type 8, which dominated the environmental- and outbreak-strains group, whereas the sporadic-case strains were more heterogeneous in H-type. All of the environmental and outbreak strains harbored the Shiga toxin 1 gene (stx1), eae, and ehx, and a subset of these also carried the Shiga toxin 2 gene (stx2). Only 9 of 16 sporadic-case strains produced stx1 and/or stx2, and these were mostly of H-type 8 and 10. Pulsed-field gel electrophoresis analysis revealed a cluster of environmental, outbreak, and sporadic illness strains with high phylogenetic similarity. Strains in this pulsogroup were all of the H8 type and STEC pathotype, and carried eae and ehx. Smaller clusters of highly similar STEC O111 strains included outbreak and sporadic illness strains isolated during different time periods or from different geographical locations. A distinct aggregative behavior was observed in the cultures of all environmental and outbreak STEC O111 strains, but not in those of sporadic-case strains. Among environmental and outbreaks strains, aggregation was positively correlated with production of curli fimbriae and RpoS function, and negatively with cellulose synthesis, while the nonaggregative behavior of sporadic-case strains correlated (positively) only with cellulose production. Our results indicate that STEC O111 strains sharing high genotypic similarity and important phenotypic traits with STEC O111 outbreak strains are present in the agricultural environment and may contribute to the burden of foodborne disease.
Introduction
S
A large surveillance study for the prevalence of STEC in a major produce-growing region of the United States revealed that 11% of the samples obtained from surface water were positive for non-O157:H7 STEC (Cooley et al., 2014). STEC O111 was isolated from cattle (Cooley et al., 2013), multiple water sources, a feral pig, a coyote, and a crow (this study), suggesting that this serogroup may have multiple reservoirs in the agricultural environment. In order to gain insight into the epidemiology and ecology of this important serogroup, we investigated the genotypic and phenotypic characteristics of a collection of strains that were isolated from the environment, and from clinical patients in sporadic cases of illness and outbreaks.
Materials and Methods
Bacterial strains and growth conditions
A list of all E. coli O111 strains used in this study, and their source, is provided in Table 1. All O111 environmental strains collected in this study were isolated with a method previously described (Cooley et al., 2013). These were obtained in a large surveillance project for the prevalence of STEC and other enteric pathogens in the Salinas Valley of California (Cooley et al., 2007, 2014). Strain RM8755 was obtained in a similar study in Mexico (Amézquita-López et al., 2012). All other environmental and clinical strains were obtained from clinical laboratories and collections as listed in Table 1. All strains were cultured in LB-half salt broth (5 g NaCl L–1) (LB-HS) or on LB-HS agar and incubated at 28°C or at 37°C, as indicated.
Referred to as RM12788 in He et al., 2013.
Clinical case described if known.
BD, bloody diarrhea; CDC, Centers for Disease Control and Prevention; DECA, diarheagenic Escherichia coli collection; HUS, hemolytic uremic syndrome; STEC, Shiga toxin–producing Escherichia coli.
Serotyping
Bacterial H-typing was performed at the E. coli Reference Center (Pennsylvania State University, University Park, PA). The fliC gene was amplified by polymerase chain reaction (PCR), digested, and the restriction fragment length polymorphism pattern of the unknown sample was compared to known patterns of the 53 different H-types (Machado et al., 2000).
Strains were first serotyped by enzyme-linked immunosorbent assay with O111-specific antisera as described previously (Cooley et al., 2013), and confirmed by immunomicroscopy with anti-O111 monoclonal antibody (Rivera-Betancourt et al., 2000). For this method, culture samples were mounted on microscope slides, air dried, rinsed in distilled de-ionized (DD) water, and dried. The mounts were pre-incubated with 1% bovine serum albumin and 0.5% Tween-20 in phosphate-buffered saline. Slides were then incubated with mouse monoclonal antibody MigG3,15C4-C11 specific to the O111 lipopolysaccharide for 40 min, rinsed in DD water, air dried, and incubated again in Alexa-fluor 488 goat anti-mouse secondary antibody (Life Technologies, Grand Island, NY). The mounts were examined under a Leica DMRB microscope with a fluorescein filter to detect the signal of the O111-positive cells.
Detection of virulence determinants
The presence of genetic determinants of E. coli virulence factors listed in Table 2 was tested by PCR at the E. coli Reference Center (Pennsylvania State University, University Park, PA), based on the method described by DebRoy and Maddox (DebRoy and Maddox, 2001).
−, not detected by PCR; +, detected by PCR.
LT, heat-labile enterotoxin; STa, heat-stable enterotoxin a; STb, heat-stable toxin b; cnf 1, cytotoxic necrotizing factor 1; cnf 2, cytotoxic necrotizing factor 2; EHEC2, enterohemorrhagic E. coli category 2; STEC, Shiga toxin–producing E. coli; DEC, diarrheagenic E. coli; EPEC, enteropathogenic E. coli.
Pulsed-field gel electrophoresis (PFGE)
The standard non-O157 PFGE procedure was performed as described in PulseNet (
Aggregation in liquid cultures
Strains were grown in borosilicate glass KIMAX test tubes (Kimble Chase, Vineland, NJ) in 5-mL LB-HS and incubated to early stationary phase at 28°C with 25 rpm rotary agitation in a roller drum. Aggregation at the bottom of the culture tube was observed readily after incubation under rotary agitation and did not require stationary incubation to allow for settling of the aggregates over time.
Phenotypic tests on growth agar
Strains were screened for curli fimbriae and cellulose production based on previously described methods (Römling et al., 1998; Zogaj et al., 2001). Briefly, curli production was assessed by streaking strains on Congo Red Indicator (CRI) plates (LB–no salt agar with 40 μg mL–1 Congo red dye and 10 μg mL–1 Coomassie brilliant blue), followed by incubation at 28°C for 2 days. Red colonies indicated production of curli and were scored for color intensity, whereas white colonies indicated lack of curli production. Strains were screened for cellulose production on LB–low salt (0.5 g NaCl L–1) agar containing 0.025% Calcofluor white M2R with Evans Blue dye (Fluka, Sigma-Aldrich, St. Louis, MO) after incubation at 28°C for 2 days. Colonies were visualized under a Leica MZFL111 stereomicroscope and rated for their fluorescence intensity under ultraviolet light, indicating the presence of Calcofluor-bound cellulose.
Presence of a functional RpoS was detected by staining glycogen in the O111 colonies with iodine, based on a modified method by King et al. (King et al., 2004). Briefly, the strains were streaked onto LB-HS agar, incubated at 37°C overnight, and then at 4°C for 48 h. The colonies were exposed to iodine vapor by placing the inverted agar plate harboring the colonies onto the open lid of the petri plate onto which 30 g of iodine crystals (Mallinckrodt, St. Louis, MO) was spread. After a 5-min exposure to the iodine vapor, the colonies were rated for their brown or yellow color, which indicated an RpoS+ and RpoS− phenotype, respectively.
Results
Serotypes
Our collection comprised O111 strains with seven unique H-types. Twelve of the 16 environmental strains and all of the outbreak strains were H8. The strains from sporadic clinical patients were more heterogeneous in H-type: only 6 of 16 were O111:H8, and the remaining strains were H2, H10, H11, H12, and H21 (1, 2, 2, 4, and 1 of 16, respectively) (Tables 1 and 2).
Detection of virulence determinants
All the environmental and outbreak strains carried stx1 (Shiga toxin 1), eae (intimin), and ehx (hemolysin), and a considerable portion had also stx2 (Table 2). Only two strains in the entire collection carried the genes coding for toxins other than stx1 and stx2 that we tested in this study; both strains—one isolated from a feral pig colon (RM9907) and the other from stream water (RM10670)—had the gene encoding the heat-stable enterotoxin a (STa). Table 2 also shows that among the sporadic-case strains, 7 of 16 did not have any of the toxin genes tested; the others all had stx1 and 50% of those also carried stx2. The strains in this group that were positive for stx1 and/or stx2 were mostly of H-type 8 and 10. The eae gene was detected in 11 of 16 sporadic-case strains, and ehx was present only in a subset of these.
PFGE
The dendrogram comparing XbaI PFGE pulsotypes revealed 5 clusters of strains, all STEC pathotypes, with a similarity of at least 90% (Fig. 1). Cluster 1, composed solely of O111:H8 serotypes, included 3 water isolates with indistinguishable patterns and with 97.8, 92.7, and 90.3% similarity to isolates from a feral pig, water, and cow feces, respectively. Cluster 2, also composed solely of H8 types, comprised 3 strains isolated from various water sources in California that have 100% similarity to each other, 97.7% similarity to a sheep fecal isolate from Mexico, and 92.5% similarity to a strain from an outbreak in South Dakota. This cluster also included a sporadic-case strain from West Virginia and an Oklahoma outbreak strain, which had 93.3% similarity to each other. Cluster 3 comprised two sporadic-case strains and the Texas outbreak strain, all three of them H8 types. Cluster 4 included three H8 strains from cows and a crow that were all isolated in the Salinas Valley in California. The fifth cluster showed 90% similarity between a sporadic-case H10 strain from Canada and 2 H8 cow strains from Wisconsin and Pennsylvania. The remainder of the strains in the collection displayed high genotypic diversity.
Pulsed-field gel electrophoresis analysis of Shiga toxin–producing Escherichia coli O111 environmental strains, and clinical strains from sporadic cases and outbreaks. The dendrogram was generated from XbaI-restricted genomic DNA patterns. Genotypic relatedness and percent similarity are shown on the left side. The number of strains of environmental, sporadic clinical, and outbreak origin are marked in black, light gray, and dark gray text, respectively. Pulsogroups based on at least 90% similarity are delineated with brackets on the right.
Aggregation in culture
Aggregates were observed at the bottom of LB-HS cultures of a subset of O111 strains when incubated to a stationary phase of growth in glass tubes with agitation (Fig. 2, upper panel). Over time during incubation, the planktonic cells were depleted in the cultures and the number and size of aggregates increased. The formation of a very dense biofilm ring at the culture liquid–air interface during rotary agitation was visible. This aggregation phenotype was observed also in M9 glucose minimal medium (data not shown). This behavior appears to be different from that of enteroaggregative E. coli (EAEC), which did not produce aggregates under our culture conditions (Fig. 2, upper panel) but did form a biofilm coating the inner surface of the tube (not shown). Preliminary data indicate that the EAEC pAA-encoded aggregative determinants were not present in two of the aggregative O111 strains that we sequenced (unpublished data). Observation of the O111 aggregates by phase contrast microscopy revealed the presence of numerous large and dense cell clusters of various sizes (Fig. 2, lower panel), whereas only single cells were detected in the nonaggregative strains. All O111 environmental and outbreak strains in our study (24/38 strains) had the ability to aggregate, whereas none of the strains causing sporadic illness displayed this unique behavior (Table 3). Aggregation did not correlate with H-type (Tables 2 and 3).
Aggregative phenotype of Shiga toxin–producing Escherichia coli cells grown to stationary phase in LB-half salt broth glass tubes at 28°C for 24 h in a rotary shaker.
+++, ++, +, −, strong, medium, light, and no production, respectively, of curli, cellulose, and RpoS in colonies shown on indicator agar plates, as described in Methods. +/−, unusual colony phenotype on indicator agar plates indicating an unclear positive or negative reaction.
Agg, aggregative; Non-Agg, nonaggregative.
Curli, cellulose, and RpoS production
All 24 environmental and clinical outbreak strains were high curli producers since they formed deep red colonies on CRI plates. However, they did not produce cellulose, as concluded from the non-fluorescent colonies they formed on Calcofluor plates (Table 3). These strains also appeared to have a functional RpoS as assayed by iodine staining of the glycogen in their colonies (Table 3). In contrast, the sporadic-case strains all produced cellulose; however, they showed a broad range of curli production and RpoS function, with some strains lacking both.
Discussion
We have assembled a diverse collection of E. coli O111 strains of environmental and clinical origin, including strains associated with sporadic illness and epidemics. Most of the environmental strains were collected and isolated in 2009–2011 from a variety of sources in an important agricultural region of California, such as cow feces, coyote, crow, feral pig, and water from a trough, irrigation ditch, or a natural body of water (stream, river, or creek). We also obtained strains originally isolated from sheep in Mexico and from cow feces in Wisconsin, Pennsylvania, and Rhode Island. Sixteen sporadic-case and six outbreak strains were included in our analysis.
All of the outbreak strains, most of the environmental strains, and a few of the sporadic illness strains were O111:H8, the classic enterohemorrhagic E. coli O111 H-type associated with disease worldwide (WHO, 1998). The presence of stx1 and/or stx2, and eae and ehx in these strains suggests their high potential to cause human disease, with the exception of sporadic-case strain DEC 8E, which lacked ehx (Eklund et al., 2001; Hughes et al., 2006). All the isolates from cow feces, irrespective of their state of origin, were O111:H8 and carried stx, eae, and ehx, supporting the results of previous similar studies showing that E. coli O111 is commonly isolated from cattle, with most isolates positive for ehxA, eaeA, and stx1 and/or stx2 (Jeon et al., 2006; Lee et al., 2008; Joris et al., 2011).
In contrast, the O111 strains from sporadic illness patients formed a very diverse group and most were of an H-type other than 8, including H2, H12, and H21, which lacked stx, eae, and ehx, and are generally known to be prevalent in endemic and single cases of diarrhea (Campos et al., 1994; Alikhani et al., 2011). It is noteworthy that one of the strains isolated from stream water, RM9322, was of the H2 type and carried stx1, stx2, eae, and ehx, in contrast to our observations above. This strain formed a pulsogroup of 100% similarity by PFGE analysis with 2 other water isolates from the same region, 1 of which was H7 and the other H8, suggesting instability of the surface antigen in their ancestor strain. Also of interest is the carriage of the heat-stable enterotoxin a (STa) gene in two other strains phylogenetically related to the latter cluster, one originating from a feral pig (RM9907) and another from water (RM10670), all of which were isolated in the Salinas Valley of California. STa is frequently produced by strains that cause severe disease and watery diarrhea in humans (Nair and Takeda, 1998). Its apparent transfer between strains that possess stx and other virulence factors such as eae and ehx raises concerns about the emergence of highly virulent strains in this region.
Phylogenetic analysis by PFGE revealed several pulsogroups dominated by STEC O111:H8 strains, indicating a certain clonality in this serogroup. Our observations corroborate previous reports that STEC strains with the same H antigen are closely related (Whittam et al., 1993; Ju et al., 2014). This is well illustrated by the pulsogroup-cluster 2, which was composed of STEC O111:H8 strains isolated from water in California, sheep feces in Mexico (Amézquita-López et al., 2012), a sporadic-case patient in West Virginia, and patients in two different outbreaks: one in South Dakota in 2009 (Luna-Gierke et al., 2014) and another in Oklahoma in 2008 (Calderon et al., 2010). Additional clusters of highly related strains, including those from water and feral pigs (cluster 1) and those from cow and crow feces (cluster 4) suggested transmission in the agricultural environment as well as between livestock and wildlife in common habitats. The density of wild birds, such as European starlings, has been associated previously with the prevalence of E. coli O157:H7 in dairy herds (Cernicchiaro et al., 2012). Our results point to multiple reservoirs that may contribute to the cycling of STEC O111 in an ecological niche over large distances and long time periods. For example, in PFGE cluster 1, the three indistinguishable strains were collected in August 2009, November 2009, and April 2010 from two different watershed sites kilometers apart. The three indistinguishable strains in cluster 2 were collected in September 2009, March 2010, and March 2011 from three different sites (one strain from a watershed and two strains from water on two different ranches, all kilometers apart). This illustrates that STEC O111 strains are circulating over time in different locations, perhaps due to water run-off or movement through the watershed, and roaming wildlife. The use of higher-resolution genotyping approaches than PFGE, such as whole genome sequencing, will likely further enhance our understanding of O111 epidemiology.
We observed an unusual type of bacterial aggregation by all STEC O111 environmental and outbreak strains, irrespective of H-type, but not by any of the sporadic-case strains. The aggregation that occurred over time in cultures grown in glass tubes is likely due to its initiation by attachment of the cells to the borosilicate surface at the liquid–air interface during incubation under agitation. This led to biofilm rings that grew thicker, then detached and dropped to the bottom of the tube. This process nearly depleted all of the planktonic cells from suspension. Given that this type of aggregation was not observed in EAEC cultures tested in our study, and that the genomes of two of our aggregative O111 environmental strains lack typical EAEC aggregative adherence fimbriae (AAF) genes (Okeke and Nataro, 2001) (unpublished data), this phenotype likely has a different molecular basis than that of EAEC. The presence of adhesins other than AAF that have been implicated in EAEC aggregation (Boisen et al., 2008), however, is worthy of exploring in future studies.
All of the aggregative STEC O111 strains in our study produced high levels of curli fimbriae but not cellulose, whereas all of the sporadic strains did not aggregate and produced cellulose, but disparate amounts of curli. These results suggest an interactive role of curli and cellulose in the outcome of their multicellular behavior. Co-expression of curli and cellulose operons in Salmonella enterica and nonpathogenic E. coli leads to biofilm formation (Zogaj et al., 2001). However, overproduction of cellulose in nonpathogenic E. coli was shown also to inhibit curli-mediated adhesion to surfaces (Gualdi et al., 2008). High curli-producing E. coli O157:H7 strains lacking any aggregation under our culture conditions (data not shown), and occasional nonaggregative sporadic-case strains that produce curli (in addition to cellulose), suggest that additional factors are involved in aggregation by STEC O111. This aspect of cell–cell interactions in STEC O111 is currently under investigation.
The multicellular behavior of all E. coli O111 environmental and outbreak strains, combined with a common curli, cellulose, and RpoS production phenotype, distinguishes them from the nonaggregative sporadic-case strain group, which displayed diverse, but overall, opposite phenotypes. Considering that curli is involved in the attachment of E. coli to a variety of biotic and abiotic surfaces and to eukaryotic cells (Barnhart and Chapman, 2006), and that RpoS enhances bacterial cell survival to multiple stresses (Battesti et al., 2011), we speculate that this combination of traits in STEC O111 strains, in addition to their virulence profile, contributes to the superior environmental fitness necessary to cause epidemics.
Conclusions
Characterization of a diverse collection of E. coli O111 revealed clusters of phylogenetically related strains isolated from the agricultural environment, and from sporadic clinical and outbreak patients. Based on their common virulence factors, serotype, surface characteristics, and aggregative behavior in vitro, the environmental and outbreak strains form a group phylogenetically distinct from the sporadic-case strains. Occurrence of outbreaks of STEC O111 linked to food, including produce, and the presentation of clinical symptoms indistinguishable from those of STEC O157:H7 infections, emphasize the need for heightened surveillance for the prevalence of this important serogroup.
Footnotes
Acknowledgments
The authors thank Yaguang Zhou, Marcia Sousa Oliveira, and Sung Im for technical assistance, and the many people who shared their strains. This research was supported by the United States Department of Agriculture, Agricultural Research Service CRIS projects 5325-42000-046 and 5325-42000-047.
Disclosure Statement
No competing financial interests exist.