Selective Survival of Escherichia coli Phylotypes in Freshwater Beach Sand

Fecal pollution monitoring still relies on E. coli enumeration, despite the fact that this organism can survive for prolonged periods and has been shown to be easily transported from sand into surrounding waters through waves and runoff, thus no longer representing recent fecal pollution events. Here, we demonstrate experimentally that regardless of the host source, certain genetically distinct subgroups, or phylotypes, survive longer than others under conditions typical of Great Lakes beach sites.

ABSTRACT

Escherichia coli is used as an indicator of fecal pollution at beaches despite evidence of long-term survival in sand. This work investigated the basis for the survival of E. coli through field microcosm experiments and phylotypic characterization of >1,400 E. coli isolates recovered from sand, sewage, and gulls, enabling the identification of long-surviving populations and environmental drivers of their persistence. Microcosms containing populations of E. coli from each source (n = 176) were buried in the backshore of Lake Michigan for 45 and 96 days under several different nutrient treatments, including unaltered native sand, sterile autoclaved sand, and baked nutrient-depleted sand. The availabilities of carbon and nitrogen, as well as competition with the indigenous community, were major factors that influenced E. coli survival. E. coli Clermont phylotypes B1 and A were the dominant phylotypes surviving seasonally (>6 weeks), regardless of source and nutrient treatment, whereas cryptic clade and D/E phylotypes survived over the winter (>300 days). Autoclaved sand, presumably supplying nutrients through increased availability, promoted growth, and the presence of the indigenous microbial community reduced this effect. Screening of 849 sand E. coli isolates from four freshwater beaches demonstrated that B1 and D/E were the most common phylotypes recovered. Analysis by quantitative PCR (qPCR) for the Gull2, Lachno3, and HB human markers demonstrated that only 25% of the samples had evidence of gull waste and none of the samples had evidence of human waste. These findings suggest that the prevalence of E. coli in sand could be attributed more to long-term-surviving populations than to new fecal pollution.

IMPORTANCE Fecal pollution monitoring still relies on E. coli enumeration, despite the fact that this organism can survive for prolonged periods and has been shown to be easily transported from sand into surrounding waters through waves and runoff, thus no longer representing recent fecal pollution events. Here, we demonstrate experimentally that regardless of the host source, certain genetically distinct subgroups, or phylotypes, survive longer than others under conditions typical of Great Lakes beach sites. We found that nutrients were a major driver of survival and could actually promote growth and that the presence of native microorganisms modulated these effects. These insights into the dynamics and drivers of survival will improve the interpretation of E. coli measurements at beaches and inform strategies that could focus on reducing nutrient inputs to beaches or maintaining a robust natural microbiome in beach sand.

INTRODUCTION

Escherichia coli is commonly used in microbial source tracking and is a Gram-negative facultative anaerobe with a biphasic life cycle that alternates between the guts of warm-blooded animals and the external environment (1, 2). Its primary habitat is within the biofilms of the mucus layer of the digestive tract of a host (3, 4), and it is shed into secondary environments through fecal matter at a density of 10⁶ to 10⁹ CFU per g human waste (2). It has also been shown to be deposited at beaches through gull waste at a density of 10⁵ to 10⁹ CFU per g (5–7). Because E. coli is a common inhabitant of the gut in humans and animals, which can also harbor pathogens, the EPA has recommended the use of E. coli as a fecal indicator (8); thus, the enumeration of this organism is the most common method of determining recreational water quality at beaches across the Great Lakes (9). However, reliance on E. coli levels to demonstrate recent fecal pollution events does not account for the biphasic life cycle of E. coli, such that it can spend its life not only in a primary host-associated habitat but also in secondary environmental habitats, such as beach sand (2). Evidence suggests that some strains of E. coli are suited for life in their secondary habitat, since sand-associated populations have been identified through routine beach monitoring (10, 11) and have been determined to be genetically distinct from host-associated strains through repetitive extragenic palindromic (REP)-PCR DNA fingerprinting (12–14), multilocus enzyme electrophoresis (MLEE) and multilocus sequence typing (MLST) (15), and DNA microarrays (16).

E. coli has commonly been isolated at significant levels in beach sand, and it has been suggested that this organism has become part of the native community (17, 18). These sand-associated populations of E. coli have been shown to impact water quality, since they are transported into local water through wave action (19, 20), resulting in elevated levels of E. coli in recreational water without recent fecal pollution inputs. It has been suggested that the accuracy of E. coli as a fecal indicator at beaches would be improved if E. coli populations associated with long-term survival in sand could be differentiated (21). One method for differentiating E. coli is phylogenomics, which uses genetic distance to describe the substructures of populations. Early phylogenetic analysis of E. coli identified four main phylogroups, A, B1, B2, and D (22–24), with groups B2 and D at the basal position in the phylogenetic tree (25) and groups A and B1, which are considered sister groups, evolving later (26). Additional phylogenetic analysis identified group C as a sister group of B1 (27, 28) and group F as a sister group of B2 (27, 29). Phylogroup E is a sister group of D (30) and is not differentiated in the multiplex Clermont phylotyping method used in this study. Further phylotypic diversity was discovered by assessing the population genetics of E. coli isolates recovered from freshwater sand. Cryptic clades I to V were characterized, and clades III, IV, and V were suggested to be environmentally adapted, as evidenced by the recovery of a relatively higher frequency of these clade members from beach sand samples and their near-absence in host-associated populations (31, 32).

Analysis has shown that the distribution of phylogroups among E. coli sources is nonrandom. Members of phylogroup B1 were found to be more transient in humans than members of phylogroups A and B2 (33), suggesting that B1 members are more likely to be recovered from secondary habitats. In humans, phylogroups A (40.5%) and B2 (25.5%) were the most common, followed by B1 and D (17% each), whereas in animals, B1 is the most common phylogroup (41%), followed by A (22%), B2 (21%), and D (16%) (34, 35). Previous work captured a snapshot of E. coli populations at six freshwater beaches and demonstrated that B1 was the most common phylotype recovered from the sand (56%), followed by groups A and D (23% and 15%, respectively), while B2 (6%) was the least frequently recovered (15). This variance in the distribution of phylotypes among sources suggests that some phylotypes are associated with particular hosts and environmental niches.

In an effort both to explore in depth the ability of E. coli to survive in beach sand and to test the hypothesis that certain phylotypes survive preferentially over others, we conducted a series of microcosm experiments with and without the native microbial community, and with a range of nutrient conditions. Our objective was to understand the potential and time frame of survival for E. coli using isolates (i) obtained from the beach and already selected for in the secondary environment and (ii) from gulls, a common fecal pollution source at beaches, and using culture and Clermont phylotyping to assess shifts in the surviving E. coli population. We found that survival was associated with specific phylotypes rather than with isolation from primary or secondary habitats and was driven by nutrient availability and the absence of the native microbial community. Understanding which E. coli phylotypes can preferentially accumulate in the sand during long-term survival could be useful for identifying chronic E. coli reservoirs at beaches and for improving the interpretation of beach water quality that may be affected by the delivery of E. coli from beach sand.

RESULTS

Survival of E. coli populations from different sources.

E. coli isolates from each source and treatment were able to survive in the microcosm experiments for >6 weeks and, in some cases, nearly 1 year. Populations of E. coli representing a range of phylotypes were incubated in microcosms containing native sand (i.e., collected from the beach and unaltered). Cell densities were found to decrease steadily by 2 to 3 orders of magnitude over 45 days (Fig. 1), with exponential decay rates of −0.138, −0.104, and −0.122 day⁻¹ for sand, sewage, and gull isolates, respectively (see Table S2 in the supplemental material). In contrast, the same populations of E. coli incubated in autoclaved sand increased overall by 1 order of magnitude by the end of the incubation (96 days), with exponential decay rates of –0.007, 0.009, and –0.014. Throughout the initial incubation period, E. coli levels for each source remained within 1 order of magnitude of each other within each treatment, with initial concentrations of 1.38 × 10⁷, 1.27 × 10⁷, and 1.10 × 10⁷ CFU/100 g sand in autoclaved sand, sewage, and gull treatments, respectively, and concentrations of 1.70 × 10⁷, 1.23 × 10⁷, and 4.47 × 10⁷ CFU/100 g sand, respectively, after 96 days. Concentrations declined after the yearlong incubation, with 1.12 × 10⁶, 1.10 × 10⁶, and 4.0 × 10⁵ CFU/100 g sand for sand, sewage, and gull survivors, respectively.

FIG 1.

Temperature and E. coli levels in source microcosms. (Top) Internal temperatures in microcosm experiments, measured hourly with an in situ temperature probe. Temperatures fluctuated, ranging from 9 to 22°C. (Bottom) Levels of E. coli isolates recovered from native (solid lines) and autoclaved (dashed lines) sand inoculated with 176 E. coli isolates from each source. Microcosms were buried in sand at the shoreline and were continually wetted by wave action. Native sand treatments were deployed for 45 days and autoclaved sand treatments for 96 days. Microcosms were analyzed in triplicate for each source at each time point. Control microcosms with no seeded E. coli were negative. Isolate survival varied based more on treatment than on source.

In an effort to identify isolates that are capable of survival over the winter at the beach, autoclaved and native microcosms (in duplicate) were left buried for 320 (native treatment) and 355 (autoclaved treatment) days in microcosms constructed to allow for air and water flow. Isolates from each source survived in the autoclaved treatment with final average concentrations of 1.15 × 10⁶, 1.10 × 10⁶, and 4.00 × 10⁵ CFU/100 g sand in sand, sewage, and gull isolate collections, signifying a loss of E. coli of only 1 to 2 orders of magnitude. In contrast, E. coli in the native treatment had a 6-order-of-magnitude decrease over the yearlong experiment, with low numbers of residual isolates, 1.55 × 10¹ and 1.11 × 10¹ CFU/100 g sand, recovered from sewage and gull microcosms in native sand treatments, respectively (n, 34 isolates recovered; 6 combined microcosms).

Effect of nutrient availability on the survival of sand isolates.

Additional microcosm experiments using nutrient-limited sand and E. coli isolates recovered from the sand confirmed that survival varied based on the sand treatment and nutrient availability. In this experiment, E. coli isolates originally collected from the sand were incubated in the nutrient-limited baked sand and survived poorly, with a 4-order-of-magnitude loss in cell density, whereas isolates seeded into native sand decreased by only 2 orders of magnitude. In contrast, isolates seeded into autoclaved sand grew and increased by 2 orders of magnitude (Fig. 2). The exponential die-off coefficients for the native, autoclaved, and baked treatments were −1.75, 2.26, and −3.33 day⁻¹, respectively (Table S1). The results were comparable to those of the source microcosm experiment, with survival (or growth) somewhat more robust during this deployment. The nutrient concentrations determined in each treatment mirrored the growth or decay dynamics (Table 1). Nitrogen concentrations were found to have the strongest correlation with decay rates, and higher total nitrogen levels correlated with slower decay. At the final time point, E. coli densities in the 90% baked–10% autoclaved treatment showed growth (decay constant, 1.93) and reached concentrations within an order of magnitude of those with the 100% autoclaved sand treatment (Fig. 2; Table S1). In contrast, the native sand amended with 10% autoclaved sand showed a reduction in cell densities and results nearly identical to those for the native sand treatment with no amendment.

FIG 2.

Temperature and population counts in nutrient microcosms. (Top) Internal temperatures in microcosm experiments, measured hourly with a deployable temperature probe. Temperatures fluctuated, ranging from 8 to 16°C. (Bottom) Population counts of E. coli isolates recovered throughout the 56-day nutrient microcosm experiment. Dashed lines indicate nutrient dose treatments. The same set of isolates originating from beach sand was used in each treatment. Nutrients appear to be correlated with E. coli population survival and growth.

TABLE 1.

Nutrient contents and Spearman’s rank correlation coefficients for nutrient treatments and E. coli survival in microcosm experiments

Sand treatment or correlation coefficient	Mean (SD) content (ppm)a
	Total carbon	Organic carbon	Total nitrogen	Total phosphorus
Native blank	17,600 (NA)	5,660 (NA)	19 (NA)	13.7 (NA)
Autoclaved blank	45,800 (NA)	23,700 (NA)	36.6 (NA)	18.6 (NA)
Baked blank	30,700 (NA)	14,900 (NA)	19.5 (NA)	16.6 (NA)
Native	17,700 (506)	151 (2.81)	13.7 (4.60)	9.77 (1.43)
Autoclaved	57,700 (19,800)	12,810 (17,900)	64.1 (21.3)	15.6 (0.46)
Baked	28,900 (803)	676 (134)	BLD (NA)	16.4 (1.73)
90% native, 10% autoclaved	21,000 (3,990)	12,800 (7,540)	12.7 (5.40)	12.8 (2.47)
90% baked, 10% autoclaved	34,500 (1,480)	8,470 (8,640)	23 (12.6)	13.9 (0.86)
Spearman’s rank correlation coefficientb	0.59	0.05	0.74	0.00

^aIn parts per million. Nutrients were measured in duplicate for inoculated samples and on single blank samples. NA, not applicable; BLD, below the limit of detection. In each sample, the organic nitrogen content was below the limit of detection of 10 ppm.

^bCalculated using nutrient measurements from the output of the experiment and the percentage of E. coli organisms surviving in each of the inoculated treatments.

Microbial community profiles within the microcosm experiments.

Analysis of 16S rRNA gene sequencing revealed a wide range of richness and diversity in the microbial communities within the sand treatments of the microcosm experiments. Simpson’s diversity index scores for samples collected from the beginning of the microcosm experiment were 0.99, 0.55, and 0.25 for native, baked, and autoclaved treatments, respectively; the scores shifted to 0.96, 0.89, and 0.86 by the end of the experiment, with a moderate shift in composition over the course of the experiment that included increases in the levels of Methylophilaceae, a common aquatic betaproteobacterium found in sediments in oligotrophic environments (36), and Pseudomonadaceae (37) (Fig. 3). The E. coli signal could be tracked in each microcosm in relation to the background microbial community within the 16S rRNA gene sequence data. The native sand, obtained from Bradford Beach, had residual culturable E. coli, but levels were approximately 4 orders of magnitude lower than the E. coli inoculum, and E. coli was not detected in the top 50 amplicon sequence variants (ASVs) from sequencing data. The average relative abundances of sequences annotated as Escherichia/Shigella in the community were 0.31% in the native sand input and 0.016% in the output, with a concurrent 2-order-of-magnitude reduction in cultured E. coli levels over the course of the experiment. In contrast, similar amounts of E. coli were introduced into the baked and autoclaved sand microcosms, but the E. coli signals were much larger, averaging 19% and 68% of the communities, respectively. The microbial communities for the baked and autoclaved sand treatments likely represent residual DNA, although the shifts in the communities demonstrate the regrowth of some members. The autoclaved sand treatments showed robust growth of E. coli with a 1- to 2-order-of-magnitude increase. However, by the end of the experiment, certain members of the microbial community increased more, as shown by the low relative abundance of Escherichia/Shigella sequences, which comprised 0.04 and 2.12% of the communities of the autoclaved and baked sand treatments, respectively.

FIG 3.

Comparison of 16S rRNA gene compositions and E. coli counts in nutrient microcosm treatments. (Top) Concentrations of E. coli in microcosm treatments, determined through direct plate counts of samples. (Bottom) Top 50 ASVs in each sample, organized by treatment from initial blank to initial input, final output, and final blank.

Distribution of phylotypes among sources.

Clermont phylotyping of E. coli from freshwater beach sand (n = 879), sewage (n = 186), and gull waste (n = 361) revealed that most isolates tested fell into one of seven phylogroups—A, A/C, B1, B2, cryptic clade, D/E, or F—or the phylogroup was unknown. All phylotypes were recovered from each source; however, their distributions differed based on the isolation source. A chi-square goodness-of-fit test determined that the phylotype distributions did not differ significantly between isolates from sand and sewage, since they were both dominated by phylotype B1, averaging 42 and 53% relative abundance, respectively, with phylotype D/E averaging 18 and 20%, respectively (individual sources are detailed in Table S2). Phylotypes A, B2, and F were minor components (<20% collectively) of sand and sewage E. coli populations. This was in contrast to the phylotype distribution of gull isolates, which was more even and significantly different from those of the sand and sewage assemblages (chi-square P, <0.01; chi, 115 and 76.7, respectively) (Table S2).

Phylotype profile shifts in microcosm experiments.

Significant shifts in phylotype distributions (P < 0.01) were observed after 45 days in the native sand microcosms seeded with sand, sewage, or gull isolates, with phylotypes B1 and A increasing in relative abundance, and phylogroups A/C and F decreasing, regardless of the original source of the E. coli populations (Fig. 4a and b). The same isolates from sand, gull, and sewage collections incubated in autoclaved sand microcosms showed significant, yet not as pronounced, changes in phylotype distribution, with notable increases in the relative abundances of phylogroups A and B1 (Fig. S1d and e). However, in the nutrient microcosm experiment, only a slight shift in these phylotypes was observed in the sand isolate collections (Fig. 4c and d). Interestingly, the A/C phylotype remained relatively stable in the autoclaved sand microcosms, in contrast to the native sand microcosm, where it decreased. Characterization of isolates recovered at 1 year from the native sand treatments demonstrated that the sewage survivors (n = 29) comprised phylotypes B1, D/E, and B2, while the gull survivors (n = 15) comprised phylotypes clade, D/E, and (in very low numbers) B1 (Fig. 4c).

FIG 4.

Comparison of phylotype distributions in sources at the start and end of microcosm deployments with E. coli isolates from gulls, sand, and sewage. (a) Phylotype profiles of isolates from different sources in native sand treatments at the start of the deployment. (b) Phylotype profiles of isolates in native sand treatments at 45 days (n, 182 for gull or sand sources and 161 for sewage sources). (c) Phylotype profiles of isolates at 320 days (n, 29 from sewage and 15 from gulls). Panels a, b, and c are presented in Fig. S1 in the supplemental material but are shown here for reference. (d) Phylotype profile of sand isolates used as input in the nutrient microcosm experiment. (e) Phylotype profiles of the nutrient microcosm outputs.

Phylotype growth in minimal medium.

The growth of isolates from each phylogroup in the minimal medium R2A, R2A medium diluted 1:10, and a rich medium, MUG medium, demonstrated that there was no significant difference in the carrying capacity, growth rate constant, or generation times between the phylogroups within each medium type. However, carrying capacity differed among medium treatments (Fig. S2).

Distribution of phylotypes at freshwater beaches and detection of human and gull markers.

An extensive survey of E. coli isolates recovered from sand demonstrated that phylogroup B1 was the most common phylotype recovered, dominating 7 of 10 sand samples collected from May through October of 2008–2019, with phylogroups F and D/E also at notable frequencies. Clade/E, A/C, and B2 were more minor contributors to the populations (Fig. 5). A chi-square goodness-of-fit test determined that the distributions of phylotypes among urban and rural beaches differed significantly (chi-square P < 0.01; chi = 62.2). Although both types of beaches were dominated by B1, urban beaches had more members of phylogroups B2 and clade/E than rural beaches (Fig. 5). The E. coli isolates collected after the winter season in February (n = 49) and April (n = 118) of 2019 showed that the phylotypes were more diverse and that the February patterns were dominated by the F and D/E phylotypes rather than B1. Interesting, April patterns closely mirrored what has been recovered in gulls (Fig. 4). The E. coli levels in the winter samples were low, with 3 CFU/100 g sand in the February sample and 25 CFU/100 g in the April spring sample.

FIG 5.

Phylotype distributions of E. coli isolates collected from beaches along the western shores of Lake Michigan, 2008 to 2019. Isolates were collected from the berm of the beach and were characterized by Clermont phylotyping. Urban beaches included Atwater Beach (ATW) and Bradford Beach (BB) in metropolitan Milwaukee, Wisconsin. Rural beaches included Point Beach (PB) in Manitowoc, Wisconsin, and Kohler-Andrae Beach (KA) in Sheboygan, Wisconsin. The comparison of average phylotype distributions for urban and rural beaches excludes samples collected during the winter and those collected from BB and ATW in February and April 2019. The average phylotype distribution of E. coli isolates from gull waste (presented in Table S2 in the supplemental material) is included for the purpose of comparison.

To determine if recent fecal pollution could be detected in conjunction with the isolation of E. coli from sand, quantitative PCR (qPCR) was conducted on a subset of samples using fecal indicator markers such as human Lachnospiraceae (Lachno3) and human Bacteroides (HB), as well as the gull marker (Gull2).

Human markers were not detected in any of the samples analyzed. However, the gull marker was detected in 11/44 samples tested. This included 1/9 sites from each of Kohler-Andrae and Point beaches (collected 21 June 2016), 2/3 sites on only one of the four survey dates from Bradford Beach (30 May 18), and multiple sites from Atwater Beach on each of the four survey dates (30 May 2018, 11 July 2018, 16 July 2018, and 22 July 2019). The geometric mean of E. coli levels in samples positive for the gull marker was 4-fold higher than levels in samples with no marker detected (Table S3). Overall, the urban beach sites Atwater and Bradford showed evidence of fecal pollution in 23% of samples, and the rural sites, Kohler-Andrae and Point beaches, showed such evidence in 11% of samples. These results suggest that gull waste may be contributing significantly to the E. coli burden in the sand.

DISCUSSION

E. coli reservoirs and potential sources.

E. coli can be introduced into beach sand through many routes, including bird and animal waste, sewage, and runoff (10, 38, 39). While there is strong evidence that E. coli reservoirs in beach sand impact water quality, it is unclear how much might be attributed to repeated deposition as opposed to long-term survival (10). Previous work demonstrated that B1 was the most common phylotype isolated from freshwater beach sand (15), water (40), and soil (41), suggesting that E. coli within this lineage may become naturalized (42). However, the B1 phylotype is also found in animals, including birds (13, 35, 38). Gull waste has been identified as a major contributor of fecal pollution at Great Lakes beaches (5, 17, 43). We found E. coli present in sand in the absence of the Gull2 marker in 34 of 44 sand samples tested, with no samples positive for human markers, which is consistent with past reports by our laboratory (17). In this work, we also demonstrated that the phylotype profiles in gulls are different from what is recovered from sand in most cases, offering further support for the notion that the E. coli reservoirs are a result of long-term survival. We tested a range of isolates from the primary habitat of gulls, as well as isolates that had already been selected for in the secondary habitats of sand and sewage, in an effort to determine if prolonged survival was a uniform characteristic of certain phylotypes regardless of isolation sources. We further explored the decay rates under actual beach conditions and characterized the strains that were capable of surviving over the winter.

E. coli phylotypes in primary and secondary habitats and survival in beach microcosm experiments.

E. coli strains belong to one of eight phylotypes—A, B1, B2, C, D, E, F, and cryptic clade (I to V)—and have been shown to vary in behavior and lifestyle (32, 44). In each source tested (sand, sewage, and gull waste), all phylotypes were recovered. Populations from secondary habitats (beach sand and sewage) had similar phylotype distributions, dominated by phylotype B1, whereas gulls had a more even distribution of phylotypes and included higher proportions of A/C and B2; the latter is common in primary habitats, including human and animal hosts (34, 40, 45).

The results from the microcosm experiments support the hypothesis that some phylotypes are better adapted for secondary habitats than for primary habitats. After 6 weeks of incubation, the relative proportions of B1 as well as A phylotypes increased regardless of the original source, while members of phylotypes B2 and D/E, which are considered host-associated specialists (44), did not survive well in the microcosms. Previous work supports the idea that phylotypes B1 and A are strong survivors in the external environment, since they were the most common phylotypes recovered from dairy manure lagoons (46), and microcosms with filtered estuary water showed that phylogroup B1 persisted longer than the other phylogroups (47). Additionally, members of phylogroup A, isolated from raw sewage, were found to survive chlorine stress better than other phylogroups (48).

Phylogroups A and B1 have been shown to cluster mostly on one branch when assessed by MLST, and phylogroups B2, D/E, and F cluster on a separate branch, while cryptic clade sequence types are set apart on their own branch deep within the tree (32). Interestingly, A/C, which is a sister group to A, did not survive as well as A and B1. Our results demonstrated that original isolation from primary or secondary habitats did not affect the relative enrichment of A and B1, suggesting that there is ongoing flux between the host and the secondary habitat, as opposed to long-term naturalization of select strains within a lineage. This is consistent with past research showing that E. coli B1 sequence types from water and sediment could not be distinguished from B1 isolated from various hosts by use of MLST (14, 27). Because A and B1 are clearly phylogenetically distinct from other phylotypes, they may possess unique phenotypic traits that make them better suited for the external portion of the biphasic E. coli lifestyle.

Phylotypes B1 and A survive seasonally, but not over the winter.

Here, we show that on a seasonal scale (∼6 to 8 weeks), phylotypes B1 and A survive preferentially over others, but not on an annual basis. At the ∼1-year time point, isolates from each source were easily recovered from the autoclaved sand treatment, since levels were reduced only 1 order of magnitude from the original inoculum. From the native sand treatments, there were minimal residual strains in only two of the three source microcosms. Interestingly, the majority of these phylotypes were B2, D/E, and clade/E; the B1 phylotype was found only in the sewage source microcosm (Fig. 4). The results from the winter sampling of E. coli at beaches mirrored the results of the yearlong microcosm experiments. While it is difficult to draw conclusions from a limited number of strains, these results, taken together, support the idea that strains within B2, cryptic clade, and D/E can survive for >1 year; these lineages are distinct from the A and B1 branches of the phylogenetic tree (15). Therefore, it may be that select strains of these host-associated groups have the ability to become naturalized, i.e., to become long-term replicating members of the sand microbial community, rather than B1 and A, which appear to be suited for seasonal survival.

Factors that influence the survival of E. coli in beach sand.

Previous work has demonstrated that high nutrient levels at beach sites are associated with high E. coli burdens (49). Here, results from the native and autoclaved sand microcosms suggested that the ability to survive is dependent on both nutrient availability and competition with the native community. Interestingly, the E. coli in native sand and the E. coli in native sand amended with 10% autoclaved sand (providing a nutrient source) had nearly identical decay rates, while the E. coli in baked sand amended with 10% autoclaved sand grew nearly to the levels found in microcosms with 100% autoclaved sand. While these experiments cannot distinguish if the native community was competing for nutrients or if exclusion or predation played a role, it is clear that the indigenous microbial community in sand is a major modulator of E. coli survival. Previous work showed that predation and competition have negative effects on E. coli survival, while increased nutrients are associated with increased survival (50), and that E. coli not only survived but grew in sand amended with nutrient sources such as stormwater runoff, plankton, and autoclaved sand (12, 51, 52). It has been observed that moisture content can affect bacterial growth and survival and also that the bacterial load is highest in the moist sand of the berm (12, 53, 54). The microcosm experiments performed here are distinct from those reported previously (except for the study of Alm et al. in 2006 [52], which tested a specific strain of E. coli), where experiments were carried out under actual beach conditions for long periods and were matched with phylotyping that demonstrated a shift to phylotypes B1 and A under an array of conditions. It is unclear what critical nutrients the autoclaved sand provided to promote growth (i.e., carbon, nitrogen, or both). However, knowledge of specific nutrients that drive prolonged survival would be important to facilitate better understanding of these dynamics.

By evaluating growth in various minimal and rich media, we showed that members of all phylogroups can grow at low nutrient concentrations. Previous work also shows that the growth rate of E. coli does not vary based on phylogenetic association (55). These findings suggest that the levels of E. coli found in beach sand are a function of traits involved in survival—for example, resistance to stressors or metabolic traits involved in homeostasis—rather than traits associated with growth under suboptimal conditions.

Phylotype patterns at beaches and implications for beach monitoring.

The phylotype patterns recovered from beach sand sampled during the swimming season included a high relative proportion of phylotype B1 and resembled a mixture of gull profiles and the sand profiles we originally characterized in the microcosm experiments. The phylotype distribution differed between urban and rural beaches: phylogroups B2 and D/E were more common at urban beaches and in general, and sites that tested positive for the gull marker generally contained more B2 or D/E phylotypes. The April collection also closely resembled what was recovered from gulls. This suggests that urban beaches may have more constant inputs. However, we noted that the B2, D/E, and cryptic clade phylotypes were enriched in the beach profiles relative to the gull profiles, and these were the same groups that survived, albeit at low levels, over the winter. Taken together, our findings suggest that the major culturable burden may come from seasonal survivors but that over an annual cycle, very-long-term survivors can add to the overall beach burden, raising questions about their ability to take up enough nutrients to actually grow rather than simply survive. Understanding the population dynamics and survival characteristics can improve the interpretation of beach monitoring results, since high levels of B1 and A may indicate contamination from seasonal survival in beach sand rather than from recent fecal input. The time scale of the microcosm experiments suggests that while there is survival that could impact water quality results during the beach season, E. coli reservoirs in sand do not appear robust from season to season.

MATERIALS AND METHODS

Experimental overview.

Experiments were designed to characterize E. coli bacteria capable of long-term survival at the beach and to determine the factors that allow for their survival. Microcosm experiments were employed to compare the survival of E. coli organisms from various sources under different nutrient conditions, including unaltered native sand, autoclaved sand, and nutrient-depleted baked sand. Phylotyping was used to characterize shifts in populations after incubation. Microbial communities in each treatment were assessed through 16S rRNA gene sequencing and culturing for E. coli. Surveys were conducted at four beach sites to determine if phylotype profiles matched the patterns found in microcosms, with gull and human fecal pollution markers measured concurrently with E. coli levels. Taken together, these experiments were able to characterize E. coli strains capable of long-term survival in beach sand.

Isolate collection.

E. coli was isolated from freshwater beach sand, gull waste that was observed to be freshly deposited, and wastewater influent from multiple dates (see Table S2 in the supplemental material). Gull samples were included because they are documented to be a potential contributor to fecal pollution at beaches along the Great Lakes (10, 17, 56) and around the world (20, 57). E. coli organisms isolated from freshwater beach sand were collected from the berm of the beach, where the sand receives intermittent moisture through wave action, at three to four sites 10 m apart per beach (Table S3). Samples were collected from top 6 cm of the berms of Bradford and Atwater beaches in Milwaukee, WI, Kohler-Andrae Park in Sheboygan, WI, and Point Beach in Manitowoc, WI. Additional samples were collected from Bradford and Atwater beaches in February (n = 49) and April (n = 118) of 2019; the February sample was recovered from under the snow and ice buildup on the beach in the approximate area of the berm. Due to low E. coli concentrations, winter samples from each beach were combined to form composite samples. The samples were collected in sterile Whirl-Pak bags, transported on ice, and processed using the method described by Getrich and colleagues (58), in which the samples were then eluted in sterile water (1:10), shaken by inversion for 2 min, and then filtered onto a 0.45-µm-pore-size nitrocellulose filter (Millipore, Billerica, MA). Filters were transferred to modified mTEC agar plates, and incubated according to U.S. EPA method 1603 for 18 h (9). After incubation colonies were streaked out onto mTEC agar plates and incubated for 18 h. Isolated colonies were then picked into 96-well microtiter plates and stored in a 25% glycerol–75% MUG medium solution at −80°C until needed.

Microcosm experiments.

Microcosms were used to conduct in situ studies on the survival of E. coli in beach sand. The microcosm design was adapted from the work of Alm et al. (52). Microcosms were constructed from pieces of polyvinyl chloride pipe 9 cm long and 5 cm in diameter and were sealed with polyvinyl chloride knockout test caps that had approximately 30 1-mm-diameter holes drilled through them. The end caps were lined with two 0.22-µm filters on the inner sides, which prevented bacteria and other microbes from entering or leaving the microcosms while allowing for oxygen and moisture to enter. The microcosms were filled with various sand treatments, seeded with E. coli, sealed with silicone sealant, and buried 0.5 m deep in the backshore sand of Lake Michigan in Milwaukee, WI, USA. The native sand treatment mimicked the conditions at the beach and consisted of sand collected directly from the beach, with its native microbial community and available nutrient content intact. The autoclaved treatment consisted of moist sand collected from the beach that was autoclaved for 1 h, which inactivated the native microbial community, releasing nutrients from biological constituents and making organic nutrients highly available (Table 1). Last, the baked sand treatment created a nutrient-limited environment; it consisted of sand that was baked in a muffle furnace for 3 h at 550°C, then washed with sterile Milli-Q water, and autoclaved to sterilize it. Isolates used in the experiments were grown for 18 h in MUG medium, washed three times with DNA-free sterile water, and then diluted to a final concentration of 10⁶ cells/ml by using optical density (OD) values.

The source microcosm experiment entailed pooling populations of 176 isolates (each) from gull waste, sewage, and beach sand by source and then inoculating the pooled isolates into native and autoclaved sand. The nutrient microcosm experiment entailed inoculation with the beach sand isolate collection. Native sand microcosms were buried for 45 days, and autoclaved microcosms were buried for 96 days, with an additional set of microcosms left for 320 and 360 days to test long-term survivability. The following summer season, a nutrient microcosm experiment was conducted in which the same population of beach isolates (n = 176) was seeded into native, autoclaved, and baked sand, and microcosms were buried at the beach for 56 days. Included in this experiment were two nutrient dose treatments, in which native and baked sand were seeded with 10% (by weight) autoclaved sand, which acted as a nutrient source. The nutrient-spiked treatment was compared to the nonspiked treatment in order to determine the effects of competition on E. coli survival in beach sand. Isolates from the end of each experiment were recovered for phylotyping and were stored in 96-well microtiter plates in a 25% glycerol–75% MUG medium solution in a −80°C freezer.

Phylotypic characterization.

Phylotyping was used to characterize the populations of E. coli in each of the sources tested and to assess the populations of E. coli that survived the microcosm experiments. The Clermont phylotyping method was used to classify samples into one of seven groups: A, A/C, B1, B2, D/E, F, or cryptic clade (30). The template for this reaction was prepared by growing cultures for ∼18 h, then diluting 1:20 with sterile water, and placing in a thermocycler at 100°C for 10 min to lyse the cells. The template was then used in a multiplex PCR as described by Clermont et al. in 2013 (30), and the PCR product was analyzed on a 2% agarose gel, which was stained with 500 ml of 0.5-mg/ml ethidium bromide in 1× Tris-acetate-EDTA (TAE) for 15 min and was then visualized under UV light at 320 nm.

Nutrient analysis.

The levels of total carbon, total nitrogen, organic carbon, organic nitrogen, and total phosphorus were measured in each treatment of the nutrient microcosm. Each sand sample was dried in a drying oven at 60°C overnight prior to processing. Carbon and nitrogen measurements were made with a Carlo-Erba NA-1500 CNS elemental autoanalyzer (Haake Buchler Instruments, Saddle Brook, NJ), and levels were determined with an acetanilide standard. Total phosphorus for each sample was processed by the methods described by Ruban et al. (59). Briefly, 200 g of sand was combusted with Mg(NO₃)₂ for 2 h, followed by a 16-h digestion in 1 N HCl. Sand extracts were then diluted and analyzed using the ascorbic acid phosphomolybdate method.

Nutrient growth experiment.

A nutrient growth experiment was conducted in an effort to compare the effects of nutrient availability on various phylotypes. A subset of isolates from each phylotype (n, 8 to 12) was grown in MUG medium for ∼18 h and was then inoculated into media under three different nutrient conditions in triplicate: an optimal growth medium, MUG medium (Thermo Fisher Scientific, NH); a nutrient-limited medium, R2A medium (60); and an extremely nutrient-limited medium, 1:10-diluted R2A medium. The growth of the isolates was then tracked by measuring the optical density of each culture at a wavelength of 600 nm on a BioTek (Winooski, VT, USA) plate reader for 36 h.

16S rRNA gene sequencing and qPCR analysis.

Each sample was analyzed in duplicate. DNA was extracted from nitrocellulose filters that processed 100 ml of sand eluent that had been frozen at −80°C and crushed manually using a sterile spatula. DNA extraction was completed with a FastDNA spin kit for soil according to the manufacturer’s instructions (MP Biomedicals, Solon, OH). The microbial communities of sand samples from within each microcosm experiment were assessed by 16S rRNA gene sequencing of the V4–V5 hypervariable region, using protocols developed at the Josephine Bay Paul Center at the Marine Biological Laboratory, Woods Hole, MA (61). The qPCR assays were carried out on sand samples from the beach surveys using an ABI StepOne real-time PCR system with TaqMan hydrolysis probe chemistry (Applied Biosystems, Foster City, CA). The qPCR assays conducted were based on those run by Cloutier and McLellan in 2016 (17) and included human-associated Lachnospiraceae (Lachno3) (62), human-associated Bacteroides (HF 183) (63, 64), and gull-associated Catellicoccus marimammalium (Gull2) (65). Samples were run in duplicate, and standard curves, made by six serial dilutions of a 1:10-diluted (1.5 × 10⁶ copies per reaction) linearized plasmid containing a target sequence, were run in triplicate. PCR cycling was performed as follows: 2 min at 50°C, 10 min at 95°C, and 40 cycles of 95°C for 15 s and 60°C for 1 min, except the Lachno3 assay, which had an extension temperature of 64°C (17). The copy number was then converted to the copy number per 100 g of sand.

Data analysis.

All statistical analysis was conducted in R, version 3.5.1 (66), using R core packages. Phylotypic characterization of habitats was carried out, and phylotype shifts during microcosm experiments were characterized by chi-square analysis, with a P value of <0.01 considered an acceptable significance level. The Growthcurver package was used to assess growth in the nutrient limitation experiment (67), and analysis of variance (ANOVA) employed in assessing the results, with a P value of <0.05 considered an acceptable significance level. Spearman’s rank correlation coefficient was used to determine any correlations between the nutrients measured and the final population concentrations, with an r^s of <0.6 considered an acceptable significance level. Lastly, the 16S rRNA gene sequence data were processed; the reads were trimmed using Cutadapt, v2.10 (68); forward and reverse reads were merged using PEAR, v1.10.12 (69); taxonomic determination was conducted with DADA2, v1.16 (70), with phyloseq, v1.22.3, used for visualization (71); and statistical analysis was completed with the R program vegan, v2.5.6.

Data availability.

The community 16S rRNA gene sequence data have been submitted to GenBank, and the accession numbers can be found via BioProject record number PRJNA680350.