ABSTRACT
Fecal pollution at coastal beaches requires management efforts to address public health and economic concerns. Feces-borne bacterial concentrations are influenced by different fecal sources, environmental conditions, and ecosystem reservoirs, making their public health significance convoluted. In this study, we sought to delineate the influences of these factors on enterococcal concentrations in southern Maine coastal recreational waters. Weekly water samples and water quality measurements were conducted at freshwater, estuarine, and marine beach sites from June through September 2016. The samples were analyzed for total and particle-associated enterococcal concentrations, total suspended solids, and microbial source tracking markers (PCR: Bac32, HF183, CF128, DF475, and Gull2; quantitative PCR [qPCR]: AllBac, HF183, and GFD). Water, soil, sediment, and marine sediment samples were also subjected to 16S rRNA sequencing and SourceTracker analysis to determine the influence from these environmental reservoirs on water sample microbial communities. Enterococcal and particle-associated enterococcal concentrations were elevated in freshwater, but the concentrations of suspended solids were relatively similar. Mammal fecal contamination was significantly elevated in the estuary, with human and bird fecal contaminant levels similar between sites. A partial least-squares regression model indicated particle-associated enterococcal and mammal marker concentrations had the most significant positive relationships with enterococcal concentrations across marine, estuary, and freshwater environments. Freshwater microbial communities were significantly influenced by underlying sediment, while estuarine/marine beach communities were influenced by freshwater, high tide height, and estuarine sediment. Elevated enterococcal levels were reflective of a combination of increased fecal source input, environmental sources, and environmental conditions, highlighting the need for encompassing microbial source tracking (MST) approaches for managing water quality issues.
IMPORTANCE Enterococci have long been the federal standard in determining water quality at estuarine and marine environments. Although enterococci are highly abundant in the intestines of many animals, they are not exclusive to that environment and can persist and grow outside fecal tracts. This presents a management problem for areas that are largely impaired by nonpoint source contamination, as fecal sources might not be the root cause of contamination. This study employed different microbial source tracking methods for delineating the influences from fecal source input, environmental sources, and environmental conditions to determine which combination of variables are influencing enterococcal concentrations in recreational waters at a historically impaired coastal town. The results showed that fecal source input, environmental sources, and conditions all play roles in influencing enterococcal concentrations. This highlights the need to include an encompassing microbial source tracking approach to assess the effects of all important variables on enterococcal concentrations.
INTRODUCTION
Fecal contamination of coastal recreational waters is a significant public health concern, as fecal material, often from nonpoint sources, can harbor an array of different pathogens. The U.S. EPA has established regulations based on enterococcal bacteria as the indicator of feces-borne pollution to help manage water quality at estuarine and marine beaches (1). These organisms correlated well with predicted public health outcomes in several epidemiological studies that served as the basis for their adoption as the regulatory water quality indicator (2–5). The presence of human feces can present an elevated public health risk in recreational waters compared to that from nonhuman sources due to the lack of an “interspecies barrier” for diseases and the higher density of human pathogens that humans can carry (6–8). Although human pollution represents the greatest public health risk, other fecal sources that contain enterococci and possibly human pathogens can be chronic or intermittent sources of both, making beach water quality management and remediation efforts more complex.
The need to differentiate fecal sources in recreational waters led to the emergence of microbial source tracking (MST) methods in the early 2000s, most notably the PCR-based assays that target the 16S rRNA gene in Bacteroides spp. (9, 10). There are a wide range of species-specific genetic markers designed to identify human fecal sources and various domestic and wildlife fecal sources. These assays have been in use for well over a decade and are supported by numerous and rigorous laboratory evaluations and field applications (11–17). Initial field studies investigated the relationship between MST markers and fecal indicator bacterial (FIB) concentrations in recreational waters to better elucidate the potential sources of fecal pollution. Some studies have found strong relationships between the MST markers and enterococci (12, 18), while other studies have found either weak or no relationships (19–21), many of which are discussed in a review by Harwood et al. (22). One main factor affecting the relationship between enterococci and the relative strength of different sources of fecal contamination is that enterococci can persist and grow in the environment, which can significantly influence their concentrations in recreational water (23).
Due to the pervasiveness of enterococci in natural ecosystems, recent studies have been conducted to not only elucidate environmental parameters controlling their growth but also identify naturalized niches that can act as reservoirs for enterococci and the associated influence on water quality measurements. Specifically, enterococci have been shown to persist in fresh water sediments (24–26) and marine sediments (24, 27), and in some cases, their relative concentrations in sediments are several orders of magnitude higher than that in the overlying water (24, 28–30). In addition, enterococci persist in soils affected by anthropogenic activities (31) as well as in more natural soil environments (32–34). Thus, soil can act as a significant reservoir of enterococci that can, if eroded, confound the concentrations observed in recreational waters. Evaluating the influence of sediment and or soil on water quality has, in some studies, been conducted by measuring the total suspended solids as a surrogate for sediment-associated enterococci (27, 35, 36); however, this nonspecific approach does not indicate the specific type of the source(s) of the suspended solids. With the advent of next-generation sequencing, sources of sediment or soil bacteria can be fingerprinted via 16S rRNA sequencing, and programs such as SourceTracker can then determine the relative fractions of source-specific 16S fingerprints within a water sample (37).
This study examined the coastal and estuarine beaches of Wells, ME (Fig. 1), where there has been historically elevated enterococcal levels, as reported by the Maine Healthy Beaches Program (38). Prior to this study, only a ribotyping-based MST study (39) that also involved other indicator tracking work had been conducted in this area. In that study, the two major freshwater inputs, the Webhannet River and Depot Brook, were found to be the major influences on water quality related to an array of fecal contamination sources. To investigate the potential sources of enterococci, we measured three major categories of variables (fecal source input, environmental conditions, and environmental sources) and then used a partial least-squares regression model approach to determine the most significant influences on the enterococcal concentrations in water samples.
Wells Maine study area and sampling sites. All water collection sites are marked with dark gray circles. Sites that correspond to fresh water are indicated with numerals. Soil and sediment sites are represented with stars, and estuarine sediment sites are shown with triangles. Map created with ArcGIS Online (using the Light Gray Canvas Map; sources: Esri, DeLorme, HERE, MapmyIndia).
RESULTS
Total and particle-associated enterococcal concentrations and total suspended solids in water.During this study, the total enterococcal concentrations were highest in freshwater sites, with concentrations significantly decreasing from there to the estuary and then the marine beach areas (Fig. 2). The geometric mean enterococcal concentrations were 197 and 40 CFU/100 ml at the Depot and Webhannet sites, respectively, with 71% of samples exceeding 104 CFU/100 ml at the Depot site compared to 21% at the Webhannet site. In contrast, the geometric mean enterococcal concentrations at the other sites were all <15 CFU/100 ml, and samples exceeded 104 CFU/100 ml 0% (at Wells Beach) to 25% of the time. In addition to measuring enterococcal concentrations in water samples, particle-associated enterococci (particle ENT) and suspended solid concentrations were measured to better understand the potential mode of transport of these bacteria within this coastal watershed. Throughout the study period (June to September 2016), the levels of total and particle-associated enterococci varied according to the site. The concentrations were lowest at the marine beach (Wells Beach) compared to those at other sites, with levels significantly higher in all estuary sites (W11 to W15) and freshwater sites (Depot and Webhannet) (Fig. 2).
Geometric mean concentrations of total and particle-associated enterococci and average total suspended solids (TSS) concentrations at the eight study sites. (A) Total and particle-associated enterococcal concentrations. Error bars are derived from variation from each site across the entire study. (B) Violin plots were used to represent TSS concentrations, and the color corresponds to the type of site, including marine beach (red), estuary beach (purple), estuary (green), or freshwater (blue).
Total and particle-associated enterococcal geometric mean concentrations were statistically similar at the estuary beach (W11, W12, and W13) and estuary (W14 and W15) sites. However, freshwater sites (Webhannet and Depot) had statistically higher enterococcal concentrations than other sites (P < 0.05) (Fig. 2). The ratio of total to particle-associated enterococci varied throughout the season, with an average (± standard deviation [SD]) of 36.3% ± 30% across all sites. Sites within the estuary beach showed the highest ratio (41% ± 32%); however, there were no significant differences observed between sites or types of sites. The average total suspended solids (TSS) concentrations were relatively low and similar for most sites, with an overall average of 2.9 mg TSS/liter (SD, ± 1.2 mg TSS/liter). However, the Webhannet freshwater site had a significantly lower average TSS concentration (1.2 ± 1.0 mg/liter, P < 0.05) (Fig. 2) despite, as previously mentioned, having higher enterococcal concentrations. The relationship between particle-associated enterococci and TSS was not significant (r2 = 0.0011), and significant rainfall events were seldom and sparse, with only one greater than 1 in. in 48 h prior to sampling. Overall, this study showed that enterococcal concentrations were significantly different by site and were ubiquitously associated with particles and independent of the concentration of suspended solids.
Presence of fecal sources in fresh, estuarine, and marine waters.The concentration of fecal pollution in this study area was determined using both PCR and quantitative PCR MST assays to identify and quantify the predominant sources of fecal contamination present in the water. The mammal fecal marker (Bac32) was detected via PCR at all sites 100% of the time throughout the study period (see the table in Section S1 in the supplemental material). The human fecal marker (HF183) was detected in 51% of all water samples, with the highest detection rate in fresh water (56%) and the lowest detection rate in marine beach water (46%). The differences in the percent detection of the gull fecal marker (Gull2) were most pronounced between freshwater (10%) and all other sites (>77%). The dog fecal marker (DF475) detection rate was highest in the estuary beach water (10/44 [23%]); however, 8 of the 10 positive samples were detected in July (8/13 [61%] samples positive at that site). For all other sites, an increase in the detection of the dog fecal marker also occurred during July, with 44% (16/36) detection, compared to 0% for August and September and <1% for June. Thus, most of the dog contamination at all sites was associated with unknown dog-related conditions during July.
Concentrations of mammal, human, and bird fecal contaminations.We used quantitative PCR (qPCR) to provide relative quantitative measures of mammal, human, and bird fecal contamination levels. Water at estuary and estuary beach sites contained significantly higher levels of mammal (AllBac) fecal marker copies, with an average of 1.54 × 107 compared to 2.62 × 106 in freshwater and 3.9 × 106 copies/100 ml in marine beach (P < 0.05). The average concentrations of human (HF183) and bird (GFD) fecal markers were not statistically different between sites; however, the concentrations of the human marker in individual samples varied from less than the limit of detection (LOD) to 2.04 × 104 copies/100 ml (Fig. 3), while bird fecal marker concentrations were relatively stable across all sites. No significant temporal trends were observed for any of the quantitative fecal marker levels. Compared with the presence/absence of the detection of fecal sources, quantitative measurements also did not show strong spatial patterns, except mammal marker levels showed significant increases at estuary and estuary beach sites compared to those at marine and freshwater sites.
Relative levels of mammal, human, and bird fecal sources at the different types of study sites. Box plots represent levels of microbial source tracking markers at marine beach (Wells Beach), estuary beach (W11, W12, and W13), estuary (W14 and W15), and freshwater (Webhannet and Depot) sites. Outlier data are represented by black diamonds.
Differences between water, soil, and sediment bacterial community compositions.16S amplicon sequencing was used to characterize the microbial communities present in water and other sample matrices (soil, sediment, and marine sediment), which was the nexus for the ensuing SourceTracker analysis. A total of 3,276,196 reads and 7,706 unique operational taxonomic units (OTUs) were obtained from the 177 samples of fresh, estuary, estuary beach, and marine beach water and soil, sediment, and marine sediment. The numbers of OTUs assigned and the Shannon diversity indices were significantly higher for soil, sediment, and marine sediment than for water samples (P < 0.05) (Fig. 4). Most taxa in the estuary and marine beach water samples were identified as Flavobacteriia, Alphaproteobacteria, and Gammaproteobacteria classes, which together accounted for 84% of the total assigned taxa. Cyanobacteria accounted for 34% of the taxa in marine sediment, and Betaproteobacteria was one of the top three most abundant taxa in fresh water, soil, and sediment (Fig. 4). A nonmetric multidimensional scaling (NMDS) ordination was used to determine if the bacterial communities from water and other matrices (soil and sediments) differed on the basis of their taxonomic compositions. Bacterial communities from the marine beach and estuary (all estuary) waters were similar but were statistically different from those from fresh water (P < 0.05) (Fig. 5). The bacterial communities associated with soil, sediment, and marine sediment were all distinct from each other and from those in water samples, indicating unique groups of OTUs (P < 0.05) (Fig. 5). The samples taken from different areas within the watershed (soil, estuarine water, freshwater, etc.) contained unique bacterial compositions, enabling downstream analyses with the SourceTracker software to discern the relative contributions of these different communities to the makeup of microbial communities in the different types of water samples.
16S taxon profiles and the top three most abundant bacterial classes in all source and sink samples. Stacked bar plots represent percentages of the class level composition of the microbial communities. Source corresponds to environmental sources that were fingerprinted with the SourceTracker program and then used to determine their presence within water (sink) samples. The table represents the top three classes for each group of samples; *, phylum level. For a complete list of all taxon assignments, refer to Section S4 in the supplemental material.
Differences between microbial communities from different source materials. Samples are color coded on the basis of the sample matrix (i.e., soil, fresh water, etc.). Percentages of variations explained are displayed on the x and y axes and the minimum stress of the ordination is shown in the top left corner.
Environmental source contribution to water samples.The fractions of freshwater, sediment, soil, estuarine sediment, and marine beach water source bacterial communities within estuary and estuary beach water samples were calculated using the Bayesian mixing model SourceTracker. The freshwater sample analysis showed a high probability of taxa originating from underlying sediment (74%) and much lower probability of taxa originating from the soil (2.6%). The initial results for the estuary and estuary beach samples indicated that marine beach water was the dominant source of bacteria (Table 1). However, given that the likely fecal sources are coming from the watershed, we excluded marine beach water as a potential source and included it as a sink and then reanalyzed the data. These second results showed that freshwater taxa had a high probability of being a significant fraction of estuary (73%), estuary beach (66%), and marine beach (35%) water communities, with a significantly higher percentage for the estuary locations than for the marine beach (P < 0.05) (Table 1), which was more influenced by ocean microbial taxa. Despite the significant percentage of freshwater taxon assignments in the estuary, estuary beach, and marine beach waters, there were no freshwater sediment or soil taxon assignments for these sites. The data for the percentage of unidentifiable taxa showed the opposite trend compared to the percentage of assigned freshwater taxa. Unidentifiable taxa in the marine beach were significantly higher (46%; P < 0.05), which is not surprising given that marine beach water communities would likely be most influenced by nonterrestrial sources. Estuarine sediment was the highest likely identified source in the water from the marine beach site (19%), and it was significantly higher than for the percentages calculated for all estuary sites (P < 0.05). The overall results showed that freshwater source-related taxa were a pervasive source throughout the estuary and marine beach, and while sediment source-related taxa were highly abundant in the freshwater, they were not observed within the estuary or marine beach.
Relative contributions of different sources to the microbial communities in estuarine and marine watera
Relationships between environmental conditions, fecal source concentrations, environmental sources, and enterococcal concentrations.Two partial least-squares regression (PLSR) models were created to determine the relationships between enterococcal and fecal source concentrations, environmental sources, and environmental conditions (outlined in Materials and Methods). The first “freshwater” PLSR model indicated particle-associated enterococcal concentration, the concentration of mammal fecal marker, TSS concentration, percentage of sediment source, percentage of unknown source, and salinity were important variables (variable importance in projection [VIP] > 0.8) in resolving the variation in enterococcal concentrations (Table 1). A one-factor (single PLSR regression) model was deemed optimal (root mean predicted residual error sum of squares [PRESS] = 0.735) and showed that all variables (except salinity) had positive associations with enterococcal concentrations. The values for model performance (R2Y = 0.6, R2X = 0.5, and Q2 = 0.4) indicated that the model fit the data moderately well (R2X ≥ 0.5) but had poor predictive capability for enterococcal concentrations (Q2 < 0.5) (see Section S3 in the supplemental material). Of all the important variables, particle-associated enterococcus (particle ENT) concentrations showed the strongest relationship to total enterococcal concentrations (Table 2). The second PLSR model, a two-factor/two PLSR regressions model, was the best fit (root mean PRESS = 0.744) from the PLSR constructed for the estuary, estuary beach, and marine beach sites. The analysis identified particle-associated enterococcus concentration, mammal fecal source concentration, percentage of freshwater, unidentified and estuarine sediment sources, water temperature, and high tide height as significantly related to enterococcal concentrations. Factor one showed that all variables were positively associated, except for the percent unidentified and marine sediment sources. The second factor showed that mammal fecal sources, freshwater sources, and water temperatures were negatively related to enterococcal concentrations, which was the opposite of their associations for factor one. The high tide height and marine sediment were positively related to enterococcal concentrations for factor 2 of the PLSR (Table 2). Together, both factors explained 61.8% of the variation observed in enterococcal concentrations, and model performance (R2Y = 0.6, R2X = 0.5, and Q2 = 0.6) indicated better predictive ability with a similar fit to the data than for the freshwater model (see Section S3). Of all the potential variables measured (19 total) across three categories (fecal source input, environmental source contribution, and environmental conditions), particle-associated enterococcus and mammal fecal marker concentrations had the most significant relationships to enterococcal concentrations. The relationships between other variables and enterococcal concentrations were specific to freshwater and estuary/marine beach models, indicating ecosystem-specific relationships. However, the joint relationship of particle-associated and mammal fecal markers across freshwater and estuary/marine environments indicates their overarching importance in determining enterococcal concentrations.
Most significant relationships/contributions for all factors to enterococcal concentrationsa
DISCUSSION
Geometric mean enterococcal concentrations at the marine beach, estuary, and estuary beach sampling sites were all less than the State of Maine water quality standard of 35 CFU/100 ml, and the majority of single-sample concentrations were less than the State 104 CFU/100 ml single-sample maximum standard, indicating the water quality was typically considered acceptable for recreational use. Previous monitoring by the Maine Healthy Beaches Program in 2014 had shown the Wells Beach area was 1 of 7 beaches in Maine that had a greater than 20% exceedance rate, with suspicion that freshwater inputs were a significant source of contamination (38). Our findings confirmed that enterococcal concentrations were statistically higher at both major freshwater tributaries to the estuary, especially at the Depot Brook site, where levels were regularly above the 104 CFU/100 ml single-sample standard. The Depot Brook site is located in a watershed with a higher fraction of developed land (0.27 to 0.50) and more people per km2 (325 to 2,650 people) than the Webhannet site watershed, which has a lower developed fraction (0.13 to 0.25) and 150 to 325 people per km2 (40). This might help explain the difference in enterococcal concentrations between freshwater sites, as a more urbanized watershed can increase the transport of more pollution from the watershed to the freshwater tributary. However, the summer of 2016 was especially dry in this region (41), with just one event with >1 in. of rain (1.73 in., 28 June 2016) 48 h prior to the sampling time. This overall dry condition likely contributed to less fecal contamination transport (via freshwater discharge) from the watershed to the estuary and marine beach. This suggests that more typical rainfall conditions would probably have resulted in more freshwater discharge and higher enterococcal concentrations than we observed.
Enterococci were significantly associated with suspended particles of >3.0 μm diameter (R2 = 0.96, P < 0.05). On average, 36% (SD, ± 30%) of the total enterococcal concentrations were associated with particles, which suggests that particles are a potentially important transport mechanism. Other studies conducted in estuary and storm waters have found similar fractions of particle-associated enterococci, but they noted enterococci demonstrated a preference for a larger particle size of >30 μm (42–44). The large standard deviation for particle-associated enterococci might be attributed to the complex nature of particle interactions (sedimentation rate, electrostatic, hydrophobic, and other surface-surface interactions) and hydrogeological dynamics (salinity-driven turbidity maximum) (45). The mechanisms underlying enterococcus-particle interactions may also be related to the ionic strength in surface waters, as Enterococcus faecalis is negatively charged over a broad pH range (pH 2 to 8) and in the presence of different ion concentrations (46). The results for this study indicate that TSS and particle-associated enterococci had no linear relationship, indicating particle-associated enterococci were not dependent on the total amount of suspended material; thus, the association is likely due to other factors influencing cell-particle interactions.
A quantitative PCR assessment of several fecal sources is a potentially useful strategy to determine the relative significance of the different sources in a single sample and over time at sites of interest. PCR detection showed a chronic presence of a mammalian fecal source(s) (100% of samples), with a human fecal source(s) detected in approximately half of all samples; thus, qPCR analysis is useful for bringing context to the significance of these findings. For example, Mayer et al. (47) showed that wastewater effluent contains approximately 108 copies/100 ml of the AllBac mammal fecal marker, Sowah et al. (48) found that streams impacted by septic systems contained 105 to 107 copies/100 ml depending on the season, and Bushon et al. (49) determined that under storm flow conditions in an urban watershed, mammal marker copy numbers exceeded 108 copies/100 ml. Results for this study ranged from 105 to 8.6 × 107 copies/100 ml, values that are within previously reported ranges and likely concentrations reflective of a predominantly nonurbanized watershed and intermediate mammal source loading. The estuary and estuary beach area showed a statistically higher concentration of the mammal marker; however, there was no responsive increase in the concentrations of the human-associated fecal marker (HF183), which may indicate that humans are not the primary mammalian source for the increased fecal contamination.
The average concentration of the human marker was 1,500 copies/100 ml across all sites (geometric mean, 167 copies/100 ml), with the highest concentration being 20,364 copies/100 ml (Webhannet, 22 June 2016). Boehm et al. (50) showed that 4,200 copies/100 ml of HF183 in recreational waters contaminated by raw sewage is the cutoff for where gastrointestinal (GI) illnesses exceed the EPA acceptable risk level of approximately 30/1,000 for swimmers (1) Another recent study also established a copy number cutoff for the HF183 assay using a threshold of 36 predicted GI illnesses per 1,000 swimmers as a benchmark. Their findings showed that at 3,220 copies/100 ml in untreated sewage and at 3,660 copies/100 ml of HF183 in secondary treated sewage, the predicted GI illness exceeds the 36/1,000 swimmers threshold (51). It is important to note that both studies simulated risk on the basis of either raw sewage or secondary treated sewage contamination and did not account for differential decay between pathogens and molecular markers. In our study, the primary source of contamination appears to not be sewage; thus, a direct application of these thresholds to our study might not convey the same risk. However, we chose to compare our HF183 copy numbers to the threshold published by Boehm et al. (50) due to their use of the EPA acceptable risk level of 30/1,000 threshold and to give a general context for the potential health risk for this study. On average, the sites in this study did not exceed this benchmark level; however, there were 10 occasions when sites were above the 4,200/100 ml threshold (7 different sites across 4 sampling dates), indicating that sporadic events or conditions can cause elevated human fecal contamination and potential public health concerns (see Section S4 in the supplemental material). Boehm et al. also showed that at the limit of quantification (LOQ) for most assays, 500 copies/100 ml or 1,000 copies/100 ml, there is still a predicted GI illness of 4 or 8 cases per 1,000 swimmers, suggesting positive detection at the LOQ is indicative of a low-level health risk (50). For this study, the LOQ was 250 copies/100 ml for the HF183 assay, and 67 of 117 samples (57%) tested positive at or above this limit, suggesting that more than half of the collected water samples indicated the presence of a low-level health risk. Although there were no statistical differences between sites for human fecal contamination, site W11 did contain the highest geometric mean (493 copies/100 ml) (Section S4). This may be reflective of the location of the site, as it is where drainage from the Webhannet and Depot watersheds meets and is also directly downstream from a boat marina with the harbor sewage pump station, which might be a possible point source of contamination. Nonetheless, even though the sites on average were below the published thresholds, the detection of human contamination even at low concentrations is a concern.
Although human fecal sources are the greatest public health concern (6, 7, 22, 52), we did not observe any relationship between human fecal contamination and enterococcal concentrations. We did however observe a positive relationship between mammal fecal contamination and enterococcal concentrations through PLSR analysis, suggesting another mammalian fecal source(s) is more influential in explaining the variation observed in this study. It is important to note that the mammal marker is at a higher copy number in sewage/feces than the human marker, making it by default easier to detect. Also, the mammal and human molecular markers could decay at different rates in the environment, thus masking any potential relationship between human fecal contamination and enterococcal concentrations. Interestingly, gull fecal sources were detected in 77% or more of the samples in the estuary and marine beach area; however, only 10% of the samples were positive within the fresh water (Section S1) despite there being no decrease in the bird fecal marker concentration, suggesting the presence of different bird sources in these areas. Anecdotally, Canada geese were observed upstream of both the Webhannet and Depot freshwater sites periodically throughout the season, which might be a significant source of bird fecal contamination in the freshwater locations (53).
One of the unique findings of this study was the relative contribution of different sources to the bacterial community in the estuarine water. The bacterial community in estuarine water primarily originated (>90%) from marine beach water, which is not surprising for a well-flushed estuary such as the study site. Because the study period was minimally influenced by rainfall and the associated runoff of freshwater, we expected that the influence of freshwater sources would be low. In ensuing analyses, we chose not to include marine beach water as a potential source for a variety of reasons. First, the samples were always collected during low tide, before the ebb when the estuary water was draining and water was moving from the watershed toward the marine beach. Second, we had already shown that the OTU compositions for the marine beach and estuary samples were very similar, increasing the possibility of a type I error (false positive) for identifying marine beach as the likely source of enterococci. Lastly, fecal pollution sources most likely come from the watersheds and not from marine water; thus, excluding the marine beach water helps to enhance the determination of watershed influences. Our second analysis (marine beach source excluded) showed that freshwater was a significant source of bacteria to the estuary (>65% assignment) compared to soil, sediment, and estuarine sediment. This implicates freshwater as a major conduit for bacterial transport, as well as the major source of enterococci to the estuary. Overall, this finding highlights the importance of freshwater discharge as a controlling factor in transporting contamination from the watershed to the coast. The specific percent assignment of freshwater source could be an overestimate; however, the trend observed is a likely scenario given the rational discussed.
An analysis of environmental reservoirs of enterococci (soil, sediment, etc.) and their presence within water samples using SourceTracker revealed a variety of source contributions to freshwater, estuary, and marine waters. To date, there have been limited studies using SourceTracker to identify soil and sediment-associated taxa within water samples, and none of these studies have focused on a coastal watershed with the potential for freshwater, estuarine, and marine sources. One study conducted in the upper Mississippi River identified up to 14% of sediment and 1.4% of soil sources of the taxa within the river water (54). However, this study showed that the sediment source was much more abundant in freshwater (74%), indicating a greater degree of mixing between the freshwater and underlying sediment communities. The amount of sediment and soil sources within water samples may be related to site-specific characteristics such as relief or soil texture, which have shown TSS fluxes on a global scale (55). Thus, the degree to which the underlying sediment community mixes with the overlaying water is likely site specific. Interestingly, even though freshwater contained a significant amount of sediment source taxa, no sediment source was observed at the estuary and marine beach sites through the SourceTracker analysis. This difference might indicate that rapid sedimentation happens during transit to and within the estuary and at the estuarine turbidity maximum zone (56). TSS concentrations and the ratio of particle-associated to total enterococcal concentrations, however, showed no differences between freshwater and estuary/marine sites. This could be related to the separate and quite different hydrodynamics within these different water systems. The percentage of sediment source in the freshwater samples observed here might also be an overestimate/overfit from SourceTracker given the limited number of potential sources used, but the results consistently showed an elevated presence of sediment in all freshwater samples in this study. The SourceTracker analysis also revealed that the freshwater source was significant (35% or more) in estuary and marine beach water samples, suggesting that freshwater is a significant conduit for microbial and fecal contamination and transport from the watershed to the estuary and marine beach.
The use of predictive models for water quality has been a focus in the field in parallel with the adoption of bacterial indicator organisms as the gold standard for water quality determination. The goal of this research was to identify significant influences on enterococcal concentrations by measuring a wide variety of variables. To distill this information, we used a PSLR model, which has been shown to outperform similar multiple linear regression and principle components regression analyses (57) and has gained popularity in the water quality field (58, 59). Results from the PLSR analysis in this study showed that particle-associated enterococci and concentrations of mammal fecal sources were the driving force behind the variation in enterococcal concentrations, as described by both PLSR models constructed. Other factors were found to influence enterococcal concentrations; however, these differed between the freshwater and estuary/marine beach models. For example, TSS concentration, as well as the percentage of both freshwater sediment and unknown sources, positively influenced enterococcal concentrations at freshwater sites. This indicates that sediment is a likely source of enterococci that influences the concentrations measured in the water. The positive influences from the unidentified source taxa suggest either that there is an alternative source (not measured in this study) within the watershed that also influences enterococcal concentrations or that SourceTracker could simply not resolve all the potential sources we used. This finding is not surprising given the vast number of potential sources of fecal pollution within a watershed and that fecal sources were not a part of the SourceTracker analysis. The results from the estuary and marine beach model returned a two-factor regression, with each factor essentially being the inverse of the other. Specifically, it highlighted freshwater as a major conduit for microbial transport to and through the estuary. Negative influences from the unknown source reaffirm this finding, along with positive influences from the previous high tide height. The second factor explained approximately 15% of the variation in enterococcal concentrations; therefore, its importance must be weighed proportionately to that for factor one, which explained almost 50% of the variation. However, positive loadings from previous high tide height and the percentage of estuarine sediment indicate estuarine sediment could be a source of enterococci whose influence is dependent on tide height. The negative loadings from a mammal fecal source(s) may indicate that enterococci originating from the estuarine sediment are not from mammal fecal sources.
Overall, the results from this study demonstrated that the concentrations of enterococci in the coastal estuarine/marine beach study area were largely controlled by particle-associated enterococci and mammal fecal source input. The influence of these factors is likely universal across freshwater and estuarine environments; however, other ecosystem factors likely play a role as well. For freshwater portions of the coastal watershed, sediment may act as a significant enterococcal reservoir that is frequently resuspended within the water column. Freshwater itself might act as a major conduit for bacterial transport to an estuary and marine beach area where other environmental factors (water temperature and high tide height) can influence enterococcal concentrations as well. These findings highlight the dynamic nature of enterococci in natural aquatic ecosystems outside the mammalian fecal tract and that concentrations within freshwater and estuary/marine beach water are influenced by a variety of factors.
MATERIALS AND METHODS
Site description.This study was conducted in Wells, ME, USA (Fig. 1). Eight different sites were used to monitor water quality (n = 2 freshwater, n = 2 estuary, n = 3 estuary beaches, n = 1 marine beach) as well as 12 soil, 12 freshwater sediment, and 4 estuarine sediment sampling sites. The data for air temperature and rainfall amount for the 48 h prior to sampling were obtained from Weather Underground (https://www.wunderground.com/weather/us/me/wells) and the characteristics of tides during sampling were obtained from US Harbors (http://me.usharbors.com/).
Water sampling.Surface water samples were collected weekly from June to September 2016 (n = 117). Sampling started 2 h before low tide to maximize the potential impacts of freshwater pollution sources, and samples from all estuary and marine beach sites were collected before the slack tide. Water samples were collected in autoclaved 1-liter Nalgene wide-mouth lab-quality polypropylene copolymer (PPCO) bottles (Thermo Fisher Scientific, Waltham, MA, USA), and environmental parameters were measured with a YSI Pro2030 dissolved oxygen, conductivity, and salinity instrument (YSI Incorporated, Yellow Springs, Ohio, USA). A field replicate was collected at a different site for each sampling event.
Soil, sediment, and marine sediment collection.Environmental sources were collected twice throughout the sampling season to build source libraries that were “fingerprinted” with 16S sequencing and SourceTracker analysis. Six soil and sediment samples were collected upstream of both freshwater sites (Webhannet and Depot) (Fig. 1). Soil samples were collected at the crest of the stream embankment, where a 10-cm-by-10-cm plastic square template was placed down and all soil (O-horizon) within the template at a 2-cm depth was collected. Samples were sieved (USA standard no. 5) to remove any loose-leaf litter and roots to only sample smaller soil particles and their microbes. Underlying stream sediments were collected using a Van Veen sediment sampler from depositional sites chosen on the basis of the presence of fine-grain sediments. One grab sample was collected for each site, and then the top 2 cm of sediment was subsampled for analysis. Sediments were sieved (USA standard no. 45) to remove coarse-grain and gravel-size particles. Four estuarine sediments were collected during low tide, when intertidal sediments were exposed, using the Van Veen sampler, and the top 2 cm was again collected for analysis.
Enterococci and total suspended solids quantification.Total and particle-associated enterococci were enumerated using the EPA method 1600 membrane filtration protocol (60), and particle-associated enterococci were determined via filtration through a 0.47-mm-diameter 3.0-μm-pore-size polycarbonate filter (Millipore, Darmstadt, Germany) as first reported by Crump et al. (61). The filters were rolled onto plates containing mEI agar and incubated at 41°C ± 0.5°C; representative colonies were counted at 24 ± 2 h. Total suspended solids (TSS) were measured using EPA method 160-2, where 500 ml of the water sample was used to determine TSS concentrations (62).
DNA extractions.DNA extraction from all matrices was performed with the PowerSoil DNA extraction kits (Mo Bio Laboratories, Carlsbad, CA, USA), with modifications to the manufacturer's protocol needed to optimize the extraction from water sample filters. For water samples, 500 ml collected water sample was filtered through a 0.47-mm-diameter 0.45-μm-pore-size polycarbonate filter (Millipore, Darmstadt, Germany), which was stored in a sterile 2-ml cryotube at −80°C for at least 24 h. Prior to DNA extraction, frozen filters were crushed into small pieces with an ethanol-sterilized razor blade, a practice commonly used to maximize DNA recovery (63–65). To minimize additional DNA loss during the extraction process, solutions C2 and C3 (from the manufacturer's protocol) were halved in volume and combined into a single step (as per communication with the manufacturer). DNA extractions from soil, freshwater sediment, and marine sediment were conducted per the manufacturer's protocol.
Microbial source tracking PCR and qPCR assays.Microbial source tracking (MST) PCR assays that target mammals (Bac32 [65]), humans (HF183 [9]), gulls (Gull2 [66]), dogs (DF475 [10]), and ruminants (CF128 [9]) were used to determine the presence of fecal sources in water samples (Table 3). Positive-control plasmids were created for each PCR assay from fresh fecal samples that came from each target organism (human, gull, dog, and cow). The TOPO TA Cloning kit was used (Invitrogen, Carlsbad, CA, USA), with a blue/white screen of Escherichia coli transformants on kanamycin (50 μg/ml) selection tryptic soy agar (TSA) plates. Positive E. coli colonies were screened with their respective PCR assay, and PCR-positive colonies were then grown in tryptic soy broth (TSB) and extracted with the PureLink Quick Plasmid Miniprep kit (Invitrogen, Carlsbad, CA, USA). PCR assays were run on a T100 thermal cycler (Bio-Rad, Hercules, CA, USA) with the GoTaq Green master mix (Promega, Madison, WI, USA). Cycling conditions and amplification protocols for each assay targeted the different source-specific markers according to protocols delineated by different studies: Bac32 (67), HF183 (67), CF128 (68), DF475 (69), and Gull2 (66). Quantitative PCR assays were also run to determine fecal source strength for mammals (AllBac [70]), humans (HF183 [71]), and birds (GFD [72]); the primer and probe sequences can be found in Table 3. All qPCR assays were run on an Mx3000P cycler (Agilent Technologies, Santa Clara, CA, USA), TaqMan assays used the Perfecta FastMix II (QuantaBio, Beverly, MA, USA) master mix, and the SYBR Green assay used the FastSYBR Green master mix (Applied Biosystems, Foster City, CA, USA). A standard curve ranging from 106 to 102 copies (mammal assay) or 105 to 101 copies (human and bird assay) was also run for each experimental run with the limit of quantification (LOQ) being 100 copies (mammal) or 10 copies (human and bird) per PCR. The threshold cycle (CT) values and amplification efficiency, slope, and R2 values for each standard curve were compared to previously run standard curves to ensure satisfactory performance before being used to calculate copy numbers for that run. Each environmental sample was diluted 1:10 and run in triplicate, and the reaction mixture (25 μl) contained a final concentration of 0.2 mg/ml bovine serum albumin (BSA). Amplification/cycling conditions were preformed per published protocols for AllBac (53), HF183 (53), and GFD (16). TaqMan assays were run with an internal amplification control (73) with a downshift of 1 cycle considered inhibition. Samples spiked with a plasmid containing 104 copies of the GFD amplicon were used as inhibition controls for the SYBR assay, with a recovery of less than 104 copies (100%) considered inhibition. For a list of primers, probes, and standard curve performance, see Section S1 in the supplemental material.
Primers and probes for MST PCR and qPCR assays
16S library preparation.The V4 region of the 16S rRNA gene, using the 515F-806R primer-barcode pairs, was used for amplicon sequencing (74). The Earth Microbiome Project protocol was used for the amplification and pooling of samples, with minor modifications (75). The Qubit double-stranded DNA high-sensitivity (dsDNA HS) assay was used to quantify sample concentrations, and 500 ng of DNA was pooled per sample. The pool was then run on a 1.2% low-melt agarose gel to separate primer-dimers from acceptable product, and bands between 300 to 350 bp were cut and extracted as described above. The final DNA sample was then run on the Agilent Technologies 2200 TapeStation system (Santa Clara, CA, USA) to determine final size, quality, and purity of the sample. Each library was sent to the Hubbard Center for Genome Studies at the University of New Hampshire to be sequenced (2 × 250 bp) on the Illumina HiSeq 2500 (San Diego, CA, USA).
Quality filtering and operational taxonomic unit picking.QIIME 1.9.1 was used to perform all major quality filtering and OTU picking (76). Forward and reversed reads were quality trimmed (μ P25) and Illumina adapters removed via Trimmomatic (77). Any reads that were less than 200 bp were discarded, and reads were merged with the QIIME joined_paired_ends.py, using a minimum overlap of 10 bp and a maximum percent difference of 10%. Paired-end data were analyzed using the QIIME open-reference OTU picking strategy with UCLUST for de novo picking and the Greengenes 13_8 database (78) for taxonomic assignment. Alternative OTU picking strategies were also tested to determine the best workflow; for the performance of difference strategies, refer to Section S2.
SourceTracker analysis.Samples from 4 source types (fresh water, soil, sediment, and marine sediment) and 4 sink types (fresh water, estuary water, estuary beach water, and marine beach water) were analyzed by the open-source software SourceTracker v1.0 (37). Default parameters were used (rarefaction depth, 1,000; burn-in, 100; restart, 10; alpha [0.001] and beta [0.01] dirichlet hyperparameters) in accordance with previously published literature (53, 79). A “leave-one-out” cross validation was performed to assess the general performance of the model, and source samples were iteratively assigned as sinks to assess how well a known sink would be assigned (i.e., source = soil and sink = soil). The percent assignments from SourceTracker are the result of the Gibbs Sampler assigning OTUs from an unknown sample to sources in a random and iterative fashion and then calculating the likelihood of that OTU originating from said source. The final output can be interpreted as the percentage (or likelihood) of OTUs present in an unknown sample originating from the sources used in the analysis.
Partial least-squares regression model.A partial least-squares regression (PLSR) model was used to determine the most important and significant variables affecting enterococcal concentrations (80). Two models were created, one for the estuary, estuary beach, and marine beach sites and one for the freshwater sites. Particle-associated enterococci, environment variables (water temperature, air temperature, dissolved oxygen, salinity, height of previous high tide, and rainfall in previous 48 h), fecal source strength (mammal, human, and bird), and percentage of environmental source (fresh water, soil, sediment, and marine sediment) were used as explanatory variables for the nonfreshwater model. The same parameters, except height of previous high tide and percentage of freshwater source, were used for the freshwater model. All data except the percent assignments from SourceTracker were log(x + 1) transformed before performing the analysis. A KFold cross validation (K = 7) with the NIPALS method was used to determine optimal factors and variable importance (VIP > 0.8) for each model. Models were then rerun with only explanatory variables that were determined to be significant. To see the model validation and diagnostic plots, refer to Section S3.
Routine statistical analysis and data visualizations.All routine statistical analyses were performed in R v3.4.0, Python 3.6.1, or JMP Pro13, while multivariate analyses were performed with PC-ORD v6. Graphing was performed in IPython notebook with matplotlib, seaborn, pandas, and numpy packages. All pairwise comparisons were done using the Kruskal-Wallis nonparametric method, with Dunn's nonparametric multiple comparisons run post hoc using Bonferroni's correction.
Data availability.Data for all sequenced samples are publicly available in the NCBI BioProject database under accession number PRJNA431501.
ACKNOWLEDGMENTS
We thank Meagan Sims and Keri Kaczor at the Maine Healthy Beaches program and Sean Smith at the University of Maine for their guidance and help with general knowledge of the Wells, ME, area and planning of field sampling. Field work, sample processing, and molecular work were assisted by Christine Bunyon, Alexandra Bunda, Jackie Lemaire, and Audrey Beresnson, and Joseph Sevigny assisted in optimizing bioinformatic workflows.
This work was funded by a National Science Foundation New Hampshire EPSCoR IIA-1330641 grant.
FOOTNOTES
- Received 30 April 2018.
- Accepted 23 June 2018.
- Accepted manuscript posted online 13 July 2018.
- Address correspondence to Stephen Jones, Stephen.jones{at}unh.edu.
Citation Rothenheber D, Jones S. 2018. Enterococcal concentrations in a coastal ecosystem are a function of fecal source input, environmental conditions, and environmental sources. Appl Environ Microbiol 84:e01038-18. https://doi.org/10.1128/AEM.01038-18.
Supplemental material for this article may be found at https://doi.org/10.1128/AEM.01038-18.
REFERENCES
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- Copyright © 2018 American Society for Microbiology.