Microbial source tracking of fecal contamination in stormwater runoff

Fecal contamination can occur in a variety of water sources, including recreational water such as freshwater lakes and streams, along with stormwater canals and drainage ditches (Cao et al. 2017; Bakr et al. 2020; Kinzelman et al. 2020; Monteiro et al. 2021). This contamination poses a serious potential public health threat, as water contaminated with fecal matter can cause gastroenteric diseases (McBride et al. 2013; Oram 2014; Pal et al. 2018; Magana-Arachchi & Wanigatunge 2020). Stormwater poses many threats from instances of contamination from wastewater leaching into stormwater basins, canals, or ditches (Brownell et al. 2007; Bright et al. 2011; Hu et al. 2018). This untreated wastewater can pollute the stormwater supply via initial misconnection of pipes, leaking or overloaded septic systems, or damaged infrastructure, and any of these can lead to fecal contamination in natural bodies of water (Hu et al. 2018).

There has been a great deal of effort to prevent instances of fecal contamination, but there has been limited success, partially due to the inability of several current methods to accurately identify non-point pollution sources. The EPA defines non-point source pollution as pollution resulting from agricultural runoff, leaks in systems, and broken sewer lines, among others (EPA n.d.). Non-point source pollution is often challenging to identify, as the location where the fecal matter entered the water system can be unclear, especially when considering the number of non-point sources of pollution that can exist in an urban setting (Ribaudo 2001; Dressing et al. 2016). Numerous beaches and other recreational water sources are assessed weekly for the presence of fecal indicator bacteria (FIB), which is a commonly used method to assess water quality (Dada & Hamilton 2016; Nguyen et al. 2018). The reality of this method is that the FIB test is not source-specific and provides minimal information on the source of contamination (Nguyen et al. 2018; Zhang et al. 2019; Monteiro et al. 2021). The combination of FIB and advanced techniques such as microbial source tracking (MST) provides the potential to improve fecal pollution management efforts.

MST is a group of polymerase chain reaction (PCR)-based methods that identify an individual type of fecal contamination in water, traditionally quantified through concentrations of FIB in the sample (Wuertz et al. 2011; Steinbacher et al. 2021). Quantitative PCR (qPCR) is an especially useful type of MST, as the results can be rapidly produced (Ahmed et al. 2018). It is also a sensitive test, and relatively simple, making qPCR MST an excellent candidate for improved control strategies for fecal pollution (Ahmed et al. 2019a, 2019b, 2019c; Rajapaksha et al. 2019). qPCR MST methods have the distinct advantage of being able to accurately measure fecal pollution levels in addition to being able to identify the source of pollution from surface water (Sidhu et al. 2013; Ahmed et al. 2019a, 2019b, 2019c; Shanks & Korajkic 2020). While qPCR MST methods and associated markers differ considerably (Scott et al. 2002; Hagedorn & Weisberg 2009; Sidhu et al. 2013), species of the order Bacteroidales have been established as robust indicator microorganisms for adoption in MST efforts (Vadde et al. 2019; McKee et al. 2020).

The HF183 marker is a cluster of gene sequences from the genus Bacteroides, which has become a staple in tracking fecal pollution based on decades of research (Seurinck et al. 2005; Ahmed et al. 2008; Green et al. 2014; Ahmed et al. 2019a, 2019b, 2019c) This is mainly due to the abundance of Bacteroidales in the gastrointestinal tracts of mammals, as well as its host specificity and stability (Eckburg et al. 2005; Layton et al. 2006; Dowd et al. 2008; Kim et al. 2011). Particularly intriguing is the ability of the bacteria to co-evolve with the host, making HF183 a particularly useful marker for qPCR MST methods (Ahmed et al. 2016), targeted at environmental water, including stormwater (Ahmed et al. 2008).

The stormwater canals in Jefferson Parish, Louisiana, are responsible for carrying runoff from precipitation, home water use, and agricultural overflow from the parish into Lake Pontchartrain, the designated discharge point for the parish. Lake Pontchartrain has suffered a decline in water quality over the past few decades, and the Lake Pontchartrain Basin Foundation (LPBF) has stated that broken sewer lines and wastewater overflows are some of the biggest threats to the quality of the water in Lake Pontchartrain (LPBF 2017). Higher rates of stormwater infiltration have also increased microbial loading in the lake (Jin et al. 2003). Because of broken sewer lines and wastewater overflows, the most significant issue therefore appears to be associated with the presence of pathogens in stormwater, possibly originating from human sewage contamination. Although the possibility of cross-contamination through faulty sewage networks has been highlighted in previous studies (Noble et al. 2006; Rajal et al. 2007; Sercu et al. 2009), there are still uncertainties regarding sources of contamination in the Jefferson Parish stormwater canal. The application of MST is promising in relation to stormwater fecal pollution management, in that human health risk assessment and remediation strategies for microbial contamination from stormwater can be more effectively implemented if sources of contamination are known. The objectives of this study are therefore (i) to apply MST markers for identifying the source of fecal contamination in stormwater runoff in Jefferson Parish and (ii) to determine whether existing Escherichia coli monitoring could be complemented with HF183 as a reliable marker for measuring fecal contamination, specifically from human sources.

Surface water samples were collected from each sampling site and put on ice immediately. Water samples were transported to the laboratory within 2 h for analysis. Physical and chemical water quality parameters, such as pH, temperature, dissolved oxygen (DO), salinity, and specific conductance were measured in situ using a YSI Pro2030 Meter (YSI, Yellow Springs, OH, USA).

The IDEXX method was used to ascertain the number of fecal bacteria (E. coli total coliforms) present in water samples gathered during the study. The samples were brought into the laboratory within 2 h of collection. The enumeration method used is as follows. One package of Colilert-18 reagent (Colilert^® IDEXX, Westbrook, ME, USA) was added to 100 mL of each collected sample, and after the reagent was completely dissolved, the mixture was transferred into the Quanti-trays and sealed to prevent contamination and leakage (Dichter 2011). Trays were stored at 35°C for 24 h to allow for proper growth (Dichter 2011). After this time, the trays were removed from incubation and read according to IDEXX Standard Methods (National Environmental Methods Index 2001). Any wells where the sample has turned yellow were counted as positive for total coliforms, and any wells that display blue fluorescence under a UV light were considered positive for E. coli (National Environmental Methods Index 2001). The Most Probable Number (MPN) is then calculated via the use of the provided table for quantification of the MPN per 100 mL of sample (National Environmental Methods Index 2001).

Briefly, 900 mL of water samples were filtered through an electronegative membrane (0.45 μm-pore-size, 90-mm-diameter, catalog no. HAWP-090-00, Millipore, Billerica, MA, USA) attached to a filter holder (Advantec, Tokyo, Japan) and washed using an acid rinse to elute viruses, as previously described (Xue et al. 2019). After filtration, sterile forceps were used to aseptically remove the membranes from the glass filter holder and fold each membrane into sterile Petri dishes, which were stored at −20 °C. Bacterial genomic DNA was purified from membrane filters using the PowerSoil^® DNA Isolation Kit (Mo Bio Laboratories Inc., Carlsbad, CA, USA) according to the manufacturer's instructions. To maximize the DNA extraction efficiency, membrane filters were cut into small pieces with sterile scissors and the DNA was quantified with a NanoDrop ND-2000 UV spectrophotometer (Thermo Scientific, Wilmington, DE, USA). Extraction eluents containing DNA were stored at −20 °C prior to molecular detection and enumeration.

SYBR Green-based qPCR assays for HF183 target the 82-bp fragment of human-associated Bacteroidales 16S rDNA gene (Bernhard & Field 2000a, 2000b), All qPCR assays were performed using the Applied Biosystems StepOne Real-Time PCR system (Applied Biosystems, NY, USA). SYBR Green-based qPCR reaction mixture (15 μL) contained 1× SsoAdvanced Universal SYBR Green Supermix (Bio-Rad, Hercules, CA, USA), 0.25 μM of each primer, and 2.5 μL of the template DNA. The amplification conditions consisted of a hold at 95 °C for 10 min, followed by 40 cycles of 95 °C 15 s, 60 °C 30 s, and 72 °C 30 s. For SYBR Green-based qPCR assays, plasmid DNA was obtained from Dr Feng's lab at Auburn University, Alabama. Sequencing results were confirmed by consulting the website NCBI using the nucleotide Basic Local Alignment Search Tool (BLAST). Plasmid DNA concentration was measured with a NanoDrop ND-2000 UV spectrophotometer and the gene copy numbers were calculated. A calibration curve with concentrations spanning the range from 10 to 10⁶ gene copies per reaction, with two replicates, was prepared. Duplicate no-template controls (NTCs) were included in each run. The amplification efficiencies (AEs) were calculated using the equation: AE = 10^(−1/slope)–1 (Shahin et al. 2022).

Statistical analyses were run using SPSS statistical software (SPSS Inc., Chicago, IL, USA). Concentrations of indicators did not follow a normal distribution and were log-transformed prior to analysis. Kruskal–Wallis tests were performed to determine whether there were significant differences in the concentrations of E. coli, HF183, and physicochemical parameters across sites and sampling time points. Spearman's rank correlation tests were used to determine the relationships between all parameters. Mann–Whitney tests were performed to determine whether there were significant differences in the concentrations of HF183 marker and physicochemical parameters during conditions of water quality standard (WQS) exceedance for E. coli and during varying antecedent rainfall conditions. Concentrations of E. coli and physicochemical parameters were also compared during detection and non-detection of the HF183 marker. As previously described (Ahmed et al. 2016), the non-parametric Spearman rank correlation with a two-tailed p-value was also used to establish the relationship between E. coli, HF183, and physicochemical parameters in water samples. In general, r > 0.7 was designated a strong positive correlation, r > 0.4 but < 0.7 a moderate correlation, and r > 0.2 but < 0.4 a weak correlation.

Apart from the Napoleon Avenue, which generally showed less variability in temperatures, water temperature was highly variable among the five sites (ranging from 12.9 to 21.3°C, Figure 2(e)). Among all the physicochemical parameters measured at the five stormwater sites, significant differences between locations were not observed for temperature (Kruskal–Wallis, χ²(5) = 3.38, p = 0.4957). Specific conductivity was observed to be highest at the Bonnabel Boat Launch (ranged from 0.04 to 3,318 S/cm) but was not significantly higher than at other sites (p-values > 0.55). Salinity across all five tested sites was generally low and ranged from a minimum of 0.00 ppt to a maximum of 1.67 ppt at the site closest to the lake (Bonnabel Boat Launch). Significant differences were, however, not observed for salinity between the sites (Figure 2(f)), as was generally reflective of the observations made for specific conductivity. Although observed pH generally ranged between 7.03 (Napoleon Avenue) and Metairie Road (8.50), there were no statistically significant differences in the pH observed across sites.

During the 1-year sampling period, DO across all tested sites ranged from 3.53 to 10.2 mg/L (Figure 2(d)) and generally reflected a gradient of pollution across the sites tested. For instance, apart from the notable exception of Metairie Road, DO increased across the sites tested. Generally, the lowest DO values were recorded at the Napoleon Avenue, the site with the consistently highest E. coli concentrations, but values gradually increased across Esplanade, Poplar Street, and Bonnabel Boat Launch (Figure 2(f)). DO at Bonnabel Boat Launch was significantly higher than that observed at Esplanade (Kruskal–Wallis, p = 0.0063, Figure 2(d)) and Napoleon Avenue (Kruskal–Wallis, p < 0.0001, Figure 2(d)).

Across all tested sites, HF183 was detected in 62% of the water samples collected during the stormwater sampling (37/60). HF183 was detected most frequently at Metairie Road (92%, 11/12) and Napoleon Avenue (75%, 9/12). At Esplanade and Bonnabel Boat Launch, HF183 was detected 50% of the time (i.e. 6/12) while it was only detected 5 of 12 times in samples collected from the Poplar Street stormwater site. Comparatively, Metairie Road and Napoleon Avenue are situated in the most densely populated part of the studied locations. HF183 concentrations in genome copies per liter ranged from 0.31 to 5.45, 0.5 to 6.17, 0.23 to 3.16, 0.29 to 3.05, and 0.27 to 7.29 log GC/L, respectively at the Metairie Road, Napoleon Avenue, Esplanade, Poplar Street, and Bonnabel Boat Launch stormwater sites (Figure 2(b)) However, when averaged across all sampling time points, Kruskal–Wallis tests indicated that the human-associated HF183 marker was not present at significantly different concentrations in these locations (χ²(5) = 2.01, p = 0.7340, Figure 2(b)).

Pathogens can be found in stormwater runoff, which is subsequently transported to environmental water bodies through sewer overflows, and urban and agricultural runoff (Ahmed et al. 2019a, 2019b, 2019c). Previous studies have assessed levels of fecal contamination in stormwater (Marsalek & Rochfort 2004; Parker et al. 2010; Sauer et al. 2011; Madoux-Humery et al. 2013). This study focused on applying MST markers for identifying the source of fecal contamination in stormwater runoff in Jefferson Parish. This study successfully used qPCR methods to enumerate HF183 while comparing its detection and concentration with traditional FIB (E. coli and coliforms). Total coliforms were consistently very high, and higher than the IDEXX maximum limit of > 2,419.6 MPN/L, through the entire period of the study. Hence total coliform concentration offered no substantial value with respect to discussions on the source of fecal contamination. Although in our study E. coli was detected 100% of the time, concentrations were variable across the 1-year sampling period. The implication of this observation is that fecal contamination was always present at varying levels during the study period. E. coli concentrations were also found to be variable across the sampling sites in this study, despite their proximity to each other, suggesting that reliance on mere grab samples, particularly at the Metairie Road, Napoleon Avenue, and Esplanade sites, may not be robust enough to inform stormwater management policies aimed at addressing fecal contamination. This conclusion is consistent with previous studies (Converse et al. 2011). While a gradient of E. coli concentration across the first site far inland (i.e. the Metairie Road site) through to the site closest to Lake Pontchartrain (i.e. Bonnabel Boat Launch) would have been expected, it appears that this is not the case. However, it was observed that E. coli contamination was generally highest at the Napoleon Avenue storm water site, which is situated close to the multiple cross-over road network in the densely populated part of the watershed (Figure 3). As the WQS exceeded 9 of the 12 times when samples were collected at this site, it is logical to deduce that pollution was chronic at this site, compared to the low level of WQS exceedance observed at other sites. A previous study (Selvakumar & Borst 2006) also reported that pathogen concentrations from high-density residential areas were higher than those associated with low-density residential and landscaped commercial areas. Selvakumar & Borst (2006) concluded that these elevated microbial concentrations may have been from feces of domestic animals and wildlife, since the outfalls were free of sanitary wastewater cross-connections.

HF 183 markers were detected in our study with concentrations as high as 7.29 log GC/L. High concentrations of HF183 in environmental water (and specifically stormwater) have been attributed to human development and faulty infrastructure (Mallin et al. 2000; Ahmed et al. 2018). The detection rate of approximately 6 of every 10 samples, with HF183 markers detected in at least 5 of the 13 sampling sessions at each of the tested stormwater sites seem to point to a likely chronic issue stemming from urban runoff across various seasons covered in the sampling regime. In a recent study (Mika et al. 2014), HF183 was present in 58% of the samples and was detected at least once at every sample site. However, it appears that the Napoleon and Metairie Road sites are hotspots for the HF183 markers as these two sites account for nearly three-quarters of the HF183 detection in the five stormwater sites. Despite the lack of statistically significant differences in HF183 concentrations across tested sites, the HF 183 approach still adds value to efforts aimed at stormwater-related fecal pollution management, in that meaningful conclusions are still derivable based on the varying HF183 detection rates observed across sites.

The use of the HF183 marker approach to support fecal pollution management has been reported in previous studies, as its reliability has been increasingly demonstrated over the years. For instance, host sensitivity and specificity of the HF183 marker have been extensively tested and found to lie between 50 and 100% and between 85 and 98%, respectively (Carson et al. 2005; Ahmed et al. 2008, 2009, 2016, 2018, 2019a, 2019b, 2019c). Ahmed et al. (2018) found that monitoring sewage-associated markers, from leakage or other sources, was a useful method of detecting fecal markers (Ahmed et al. 2018). HF183 marker detection in our study adds value to the study of stormwater pollution in the Jefferson watershed. Without further testing for HF183 markers, the conclusion from this study would have been that fecal contamination was always present at varying levels during the study period. It would have been difficult to decipher the source of the fecal contamination; that is, whether the fecal contamination in the stormwater canals comes primarily from human or non-human sources, including ducks, amphibians, and other wildlife. Analysis of HF183 markers therefore adds another layer of conclusions derivable based on E. coli concentrations alone. For instance, a 100% detection level of varying E. coli concentrations and low rates of HF183 detection, particularly at the Esplanade, Poplar Street, and Bonnabel Boat Launch sites, the sites closest to the lake outlet, throughout the study period, indicate that fecal contamination at these stormwater canals comes primarily from non-human sources, including ducks, amphibians, and other wildlife. However, the Metairie Road and Napoleon Avenue sites, which have the highest HF183 detection rates, on top of chronic pollution by other non-human sources, are also influenced by human fecal pollution, potentially attributable to human development and faulty infrastructure (Panasiuk et al. 2015). Understandably, these sites are the farthest inland from the lake and are situated in the most densely populated part of the studied locations. Future work, including studies that focus on other animal markers, is needed to improve upon our findings.

Turbidity and DO were observed to be highly variable in our study, consistent with previous studies (Birch et al. 2004; Métadier & Bertrand-Krajewski 2012; Khamis et al. 2017). However, turbidity levels did not strongly align with observed E. coli concentrations as only weak positive correlations were observed when considering the pooled data. Previous studies have also reported positive correlations between E. coli concentrations and turbidity (Paule-Mercado et al. 2016). However, at individual sites, the relationship between turbidity and E. coli was varied in our study. For instance, turbidity levels were low at the Napoleon Avenue site despite the elevated E. coli concentrations observed at this site. A gradient of increasing turbidity from the Esplanade site through to the site closest to the lake outlet (i.e. Bonnabel Boat Launch) was observed. A similar gradient was observed for DO, potentially indicating increasing nutrient enrichment towards the lake. A positive correlation between DO and turbidity was observed in this study. Positive correlations between DO and chlorophyll, an indicator for nutrient-enriched water that support algal growth (Yu et al. 2012), has also been previously reported by Zang et al. (2011). Observed positive DO-turbidity correlations therefore call for more studies that pay particular attention to nutrient levels in the stormwater sites. In our study we observed strong positive correlations between pH and DO. This agrees with several previous authors who have reported significant positive correlations between DO and pH (e.g. Zhang et al. 2011). Unlike the weak positive correlation observed between E. coli and turbidity in our study, HF183 was moderately correlated with turbidity. This is consistent with previous studies (McCarthy et al. 2007; Bonkosky et al. 2009; Eichmiller et al. 2013; Jardé et al. 2018; Shahin et al. 2022). Albeit weakly, both HF183 and E. coli were positively correlated with temperature in our study. This is consistent with previous studies (Hathaway et al. 2010; Paule-Mercado et al. 2016; Islam et al. 2017) that have focused generally on environmental water and specifically on storm water. Research in this area has yielded highly variable responses from the scientific community. Some studies have reported that E. coli had the highest survival rates at 8 °C and the lowest survival rates at 25 °C (e.g. Wang & Doyle 1998). Others have asserted that E. coli concentrations are found to be higher at warmer temperatures (Van Elsas et al. 2011; Galfi et al. 2016). Some others (e.g. Christian et al. 2020) have, however, found no correlation between temperature and E. coli survival. Understanding the relationship between physical parameters and FIB levels can be both complex and daunting. As argued by Harmel et al. (2016), an often-overlooked component of these monitoring activities is the inherent random and systematic uncertainty present in measured stormwater data. Consequently, future studies of this nature would therefore require a more targeted yet multiple discrete sampling regime aimed at specific storm events in a way that allows understanding of highly variable intra-event characteristics, including initial fecal indicator and physicochemical parameter concentrations, and other variables as documented by McCarthy et al. (2012), such as peak concentrations, maximum rate of change, and relative confidence interval, in a way that adequately captures the influence of the catchment area on pollutant characteristics. Notwithstanding these limitations, this study has been able to show the value in complementing the existing traditional indicator bacteria approach to stormwater fecal pollution management with the advanced approaches that MST offers, specifically in the use of HF183 markers. In the current study, chronic human fecal pollution at two of the five tested sites would have been missed had we relied on E. coli alone. While generally lower rates of HF183 detection are a positive indicator that existing measures put in place to protect stormwater canals and limit infiltration may be effective, particularly at the Esplanade, Poplar Street, and Bonnabel Boat Launch sites, more efforts will still need to be channeled towards addressing human waste infiltration at Metairie Road and Napoleon Avenue.

The present study aimed at using MST markers for identifying the source of fecal contamination in stormwater. IDEXX and qPCR methods were used to track and assess concentrations of E. coli and HF183 in stormwater canals in Jefferson Parish, Louisiana. Our study results showing a 100% detection level of varying E. coli concentrations and low rates of HF183 detection, particularly at Esplanade, Poplar Street, and Bonnabel Boat Launch sites, the sites closest to the lake outlet, indicate that fecal contamination at these stormwater canals comes primarily from non-human sources, including ducks, amphibians, and other wildlife. However, the Metairie Road and Napoleon Avenue sites, which have the highest HF183 detection rates, on top of chronic pollution by other non-human sources, are also influenced by human fecal pollution, possibly because of human development and faulty infrastructure. This study highlights the advantages of the use of MST methods to complement traditional indicator bacteria-based efforts aimed at addressing fecal pollution in surface water. Further work, including studies that focus on other animal markers, is needed to improve upon our findings.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.