Competition during enrichment of pathogenic Escherichia coli may result in culture bias

Infection with Shiga toxin-producing Escherichia coli (STEC) ranges from mild to severe and can include diarrhea, hemorrhagic colitis, hemolytic uremic syndrome, and death. Although E. coli O157:H7 has been subjected to international reporting for nearly two decades, the most prevalent non-O157 serogroups have only recently gained attention (Gill and Gill 2010). Over 435 STEC serotypes have been isolated from cattle (the main reservoir of these pathogens), of which 12–17% are associated with human illness (Mathusa et al. 2010; Kaspar et al. 2010).

STEC are often present at low ratios within the total bacterial population. To exceed the threshold of detection, isolation requires enrichment and selective media. Enrichment can increase cell numbers to achieve a detectable level in both culture and molecular-based assays (at least 100 cells/g food or feces) (Possé et al. 2008; Bai et al. 2012; Paddock et al. 2012). Standard detection procedures and selective media have been developed for O157-STEC; however, no gold standard for the cultivation and isolation of non-O157 STEC exists. Isolation of non-O157 STEC from complex matrices such as food or feces is often impaired due to high levels of background microbiota, growth inhibitors, and a lack of selective media. Additionally, samples hosting multiple pathogenic strains complicate detection measures and relevant strains may not be identified (Gorski 2012; Bettelheim 2007).

Several enrichment protocols for STEC have been reported, each utilizing different combinations of media, supplements, incubation temperatures, and assessment times. Yet no medium or protocol has proven appropriate for all serogroups or sample types (e.g., food, feces, and water) (Durso 2013). Among current protocols, typical incubation temperatures range from 35 to 42 °C, and duration from 6 to 24 h (Vimont et al. 2006; Sidari and Caridi 2011). Although higher enrichment temperatures may be more selective for some STEC strains, longer incubations are sometimes applied simply for practical reasons in the laboratory such as allowing incubations to run overnight to avoid removing cultures from the incubator during non-typical work hours (Vimont et al. 2006). Any enrichment should provide a balance between recovery of the target organism while avoiding the overgrowth of competing cells (Gorski 2012). Consequently, enrichment broths are frequently modified by the addition of antibiotics to suppress the growth of background microflora. However, O157 and non-O157 STEC are a diverse group, and the addition of antibiotics has been shown to inhibit the growth of some STEC strains (Sidari and Caridi 2011; Vimont et al. 2007).

Culture bias has been described between strains of the same species or subgroup for Listeria monocytogenes and Salmonella spp. (Gorski 2012). Studies using Salmonella strains have reported dissimilar recovery characteristics in enrichment media leading to isolation of certain strains over others (Gorski 2012; Singer et al. 2009). Similarly, variants naturally present in populations of E. coli O157:H7 show differences in stress resistance that affected phenotypes isolated after 6-h enrichment (Carter et al. 2011). Another study showed that numerous STEC serogroups are concurrently present within pooled cattle feces and that prolonged storage time prior to enrichment of feces significantly decreased the number of serogroups detected, indicating that the population of E. coli in cattle feces is constantly changing even without enrichment (Conrad et al. 2014). In the same study, numbers of STEC O26 inoculated into fresh feces did not increase over the course of enrichment (as quantified by PCR), suggesting that this strain lacked the ability to effectively compete with existing microbiota in fecal samples during enrichment (Conrad et al. 2014). Still, it is unknown whether a culture bias arises during enrichment of multiple STEC serogroups, which could result in certain strains being more (or less) likely to be isolated than others. The principles governing competitive growth of these bacteria are not well understood, and in vitro experiments are scarce. The objectives of this study were to determine: if serogroup- or strain-specific competition (culture bias) exists between STEC during controlled enrichment of mixed cultures; whether this competition changed over time in the course of six subsequent enrichments; and lastly, whether toxin and virulence gene profiles influenced a STEC strain’s ability to compete during enrichment.

All overnight cultures used in inoculations reached 10⁸ CFU/mL and optical densities for 10-fold dilutions of overnight pure cultures ranged from 0.104 to 0.158 (10⁷ CFU/mL), with the exception of O121-1 (0.028–0.033; 10⁶ CFU/mL). When the glycerol stock of O121-1 was first plated, colonies appeared beige (indicating nonlactose fermenting). Following the first overnight incubation in EC broth and re-plating of serial dilutions, approximately 15% of O121-1 colonies appeared pink indicating a shift toward lactose fermentation. PCR was used to confirm that both pink and beige colonies were O121-1 and not the result of contamination.

After the first 6 h of incubation, all isolates reached 10⁶–10⁷ copies per 2 µL extracted DNA (relation of copy number/mL to CFU/mL has been described previously by Conrad et al. (2014)) across all competitions, with the exception of O121-2 (Fig. 2). Although the initial OD reading was similar to all other isolates, O121-2 reached 10⁴ copies per 2 µL extracted DNA in all competitions (Fig. 2), a difference of 2 log₁₀ or more compared with other isolates from all other serogroups.

There were 57 individual competitions in each media (114 in total; Fig. 3). Differences in copy number ≥2 log₁₀ were apparent in 12%, 38%, and 52% of competitions after 30, 54, and 78 h of consecutive enrichments, respectively (C. Conrad, personal observation, 2015). For remaining enrichments, copy numbers of the two strains were similar with ≤2 log₁₀ difference between 30 and 78 h of incubation.

The isolates with FE from each competition for the three trials, after 78 h of sequential enrichments, are given in Fig. 3. The serogroup-specific copy numbers between competing isolates differed from less than 1 log₁₀ to 8 log₁₀ copies per 2 µL DNA.

For the majority of competitions, the FE strain was the same in both concentrated and diluted media. In cases where the dominant serogroup differed between media, one of the serogroups was often O121. The isolates with the highest number of competitions prevailed (as percentage) were O45-1, O26-2, and O145-1 (Fig. 4), with O45-1 and O26-2 prevailed more competitions than O26-1, O121-1, O45-2, and O157-1 (P < 0.05). Serogroup per se did not influence the outcome of competitions. Rather, the number of competitions prevailed was impacted by the individual isolate (P < 0.01). In other words, two isolates may share the same serogroup yet one isolate may be a strong competitor, and the other may be weak.

For trial 1, each isolate prevailed in a similar percentage of competitions irrespective of media, with the exceptions of O121-1, which prevailed in 60% more competitions in diluted EC broth (Fig. 4a), and O157-1, which was least able to compete, prevailing in less than 20% of competitions. For trial 2, isolate O26-2 prevailed in the greatest number of competitions, and O45-2 prevailed in the least (Fig. 4b).

The Shiga toxin profile of isolates had a significant impact on the outcome of the competition monitored by copy number (P < 0.01), whereas serogroup, media, and other virulence genes had no significant impact in all trials. If both isolates shared similar stx gene profiles (e.g., stx1 vs. stx1, stx2 vs. stx2, both toxin genes vs. both toxin genes), copy numbers for each were superior to those of stx1 matched against stx2-only isolates (P < 0.001), but did not significantly differ from copy numbers when other combinations of stx genes were present (e.g., stx1 vs. both, stx2 vs. both). A tendency for interaction between the media and the stx profile was demonstrated, as stx1-only isolates prevailed in 80% of competitions against those with both toxin genes (trial 3) in EC broth whereas isolates with both Shiga toxin genes prevailed in 60% of competitions against stx1 isolates in the diluted broth (Fig. 4c). Two isolates contained stx2 only: O45-1 and O157-2. The O45-1 isolate prevailed 100% of competitions in both media when competing against isolates with both Shiga toxin genes. However, when O157-2 competed against isolates with stx1 only, it prevailed 50% of competitions in both media (Fig. 4d).

Studies have shown that enrichment can severely distort the apparent diversity or taxonomic profile within populations due to culturing biases (Dunbar et al. 1997; Pettengill et al. 2012). In one study, three agar media (Rainbow, Chromagar, and Blood agar) were compared for recovery of STEC from over 4000 enrichments (tryptic soy broth) of a variety of sample types (feces, plant, soil, and water). Out of 880 positive samples, no more than 56 were positive on all three media (Cooley et al. 2013). Furthermore, 377, 112, and 110 samples were positive on only one of the three media, suggesting a potential culture bias (Cooley et al. 2013).

Interestingly, reports of STEC prevalence in cattle feces show substantial variation, ranging from 0.2% to 70% (Hussein 2007; Kaspar et al. 2010). Such variance could be due to differences in climate, ecology, farming practice, and variation in sampling (Hussein 2007). However, such differences could also, in part, be a reflection of culture biases and differences in methodologies used for pathogen isolation. Enrichment of complex samples (food, feces, soil, etc.) could result in a biased culture due to the nature of the microbiota present, production of bacteriocins, bacteriophage interactions, differences in relative growth rates, and presence of growth inhibitors (Martínez-Castillo and Muniesa 2014; Muniesa et al. 2005; Pettengill et al. 2012). The aim of the present study was to utilize a simplified scenario to examine intra-species competition between STEC following enrichment using solely two bacterial strains in a broth medium. Competitive trends were examined between STEC strains based on molecular serogroup characteristics specific to O157 and the top six non-O157.

Numerous observations reported that E. coli cells in stationary phase can contain multiple chromosomes (Boye and Lobner-Olesen 1991; Akerlund et al. 1995). Despite this, qPCR was chosen to measure approximate cell number within mixed cultures due to speed and sensitivity of the technique. In comparison to PCR, traditional culturing methods are not specific enough to determine qualitatively and (or) quantitatively individual serogroups in mixed cultures. In the light of increasing chromosomes per cell, higher copy numbers are indicative of high cell concentrations, representing an adequate way to differentiate cell number among the used strains.

During the initial 6 h of enrichment, no significant differences were observed among the cell numbers of the competing strains within the same medium (diluted or undiluted). However, many of the subsequent enrichments (30, 54, and 78 h) produced biased cultures. Although enrichments longer than 30 h are not typical, the intent was to demonstrate the effect of prolonged storage of a sample prior to enrichment and analysis. Often, samples are collected days before subsequent analysis in the laboratory. In general, samples are shipped and held at 4 °C, but sometimes they are shipped under suboptimal storage conditions. Our results suggest that competition could be ongoing from the moment of sample collection, and delaying analysis could mean the results obtained are not reflective of the original sample.

The results from our test group of 14 strains do not suggest that any of the “Top 7” serogroups dominate all other serogroups during enrichment. Conversely, our results indicate that intra-species STEC competition is strain-specific. A study using Salmonella found similar results, which indicated that different selective media gave dissimilar patterns of strain dominance in both pure culture and fecal enrichment cultures (Gorski 2012). No serotype was most fit, but strains of certain serogroups were more likely to dominate than others.

The stx gene profile may in some way influence the ability of STEC strains to proliferate and coexist in the tested medium. Overall, strains possessing one stx toxin gene prevailed a greater number of competitions compared with those harboring both toxin genes. The majority of single toxin isolates contained stx1 only; however, two isolates possessed solely stx2, with one of these isolates prevailing 100% of its competitions. In undiluted broth, isolates with solely stx1 prevailed the greater number of competitions compared with isolates with both stx genes. However, when the nutrients within the media were restricted by 50%, isolates with both toxin genes prevailed nearly 20% more competitions than stx1-only isolates. This suggests that STEC strains may differ in their competitive ability when nutrient concentrations are limited, although statistical analyses did not show a significant influence of media, likely due to lack of replication.

The mechanisms and evolutionary forces governing the emergence of new STEC types are not well understood. Phylogenetic analyses have revealed that the gain and loss of virulence elements has occurred several times and in parallel in separate lineages (Franz et al. 2014). Such convergent evolution suggests the accumulation of specific combinations of virulence factors offers a selective advantage to the bacterium (Franz et al. 2014). Competitive advantages of E. coli possessing Shiga toxin genes have been reported. As one study demonstrates, Shiga toxins may act as a bacterial defense against Eukaryotic predators (Lainhart et al. 2009). Shiga toxins have also been implicated in the promotion of adherence and colonization of O157:H7 to human intestinal epithelial cells in comparison to stx-negative strains (Liu et al. 2010). However, selective advantages are more likely to operate in the bovine reservoir or external environment than in the human gastrointestinal tract as they are considered “accidental hosts,” and STEC/enterohemorrhagic E. coli (EHEC) are unlikely to have evolved specifically to cause human infection (Franz et al. 2014).

Laboratory-cultured bacteria are assumed to behave like those in nature. However, bacterial strains have been reported to rapidly transform during laboratory domestication as demonstrated by the O121 isolate in the present study, which alternated in ability to ferment lactose. Therefore, to decrease the likelihood of in-house domestication, potentially producing biased cultures, shorter and fewer incubations are recommended (Eydallin et al. 2014). Enrichment protocols other than the one in this study (different media, duration, and (or) temperatures) would likely affect the dominance order of STEC found in the present study. Different concentrations and combinations of multiple strains (more than two) should be examined to more realistically reflect the natural variation present in feces.

Further studies are necessary to fully determine the influence of environmental conditions including limiting nutrient concentrations on competitive ability of STEC with various stx and virulence gene profiles in comparison with stx-negative strains. A review of 90 STEC outbreaks from around the world, occurring between 1982 and 2006, found the vehicle of transmission was unknown in nearly 30% of outbreaks (Snedeker et al. 2009), suggesting that traceability of these pathogens is limited. Furthermore, the inability to detect certain serotypes may result in an underestimation of pathogenic non-O157 STEC. A culture bias could contribute to an underestimation of disease-associated serotypes and difficulty in traceback investigations if the pathogens are out-competed by other E. coli or bacteria during enrichment. Future studies should examine STEC culture bias during enrichment of environmental, as well as food and feces, samples, with diverse native microbiota, spiked with known concentrations of additional STEC.

The initial intent of this study was a proof of concept to examine competition causing potential bias between a mix of solely two (out of 200) different O-serogroups, strains with differing toxin gene profiles, but not considering any other (sub)type characteristics (e.g., H-type and (or) toxin subtype). For simplicity, only one enrichment media was used. However, no standard media exist for STEC detection and several different protocols are available each utilizing a different enrichment media. For example, the U.S. Department of Agriculture Food Safety and Inspection Service and the U.S. Food and Drug Administration procedures recommend tryptic soy broth and buffered peptone water, respectively. STEC enrichment bias using these broths has not been examined; however, we suspect that each enrichment medium could potentially produce biased cultures. Therefore, we suggest that future studies include a variety of media to further examine the level of potential STEC enrichment bias. Importantly, introduction of such a bias would critically influence outbreak trace-back investigations. Bias during outbreak investigations could have several outcomes, including the relevant pathogen being missed and the transmission vehicle being unidentified, or outbreaks attributed to the wrong pathogen/strain due to competition during enrichment (Pettengill et al. 2012). With continued emergence of novel pathogens, advances in STEC detection can potentially improve food safety monitoring systems, decreasing circulation of compromised food products and minimizing risks to human health.

The authors would like to thank Susanne Trapp and Yidong Han for their assistance in the laboratory. We thank Shaun Cook for expertise and valuable discussion. This research was supported by a grant from Alberta Innovates BioSolutions.