A total of eight water samples were collected from sites located along a creek in an urban watershed (Figure 1). Sampling sites were consecutively numbered, starting upstream (site 1) and moving downstream (site 4). Grab samples were collected at each site during a dry weather period (no precipitation during the previous week) on September 24, 2019 and ~12 h after the completion of a rain event (42.6 mm total precipitation during the previous 72 h) on October 23, 2019. The sliding 31-day precipitation averages for the study area on September 30 and October 31, 2019 were reported as 0.75 and 4.99 mm, respectively, with the greatest precipitation event (34 mm) occurring on October 16, 2019. Time series of the area-averaged daily precipitation rate for September and October 2019 were reported by Goddard Earth Sciences Data and Information Services Center Interactive Online Visualization and Analysis Infrastructure (Acker and Leptoukh, 2007). Surface water (2 L) was collected at each location using a sterile Nalgene carboy (Thermo Fisher Scientific, Waltham, MA, USA) treated previously with hydrochloric acid (10% v/v), ethanol (95% v/v), and autoclaved. Water samples were transported to the laboratory on ice. Temperature of the water samples was monitored to ensure that it did not reach above 8 °C during transportation by using a LogTag® single trip temperature alert indicator (LogTag Recorders, Auckland, New Zealand). Samples were processed as described below within 2 h of collection and following recommendations of EPA Clean Water Act Analytical Methods (EPA, 2019).

Escherichia coli and enterococci concentrations in the grab samples were enumerated, following manufacturer's instructions for most probable number (MPN) method per 100 mL of sample water using IDEXX Quanti-Tray System with commercial Colilert and Enterolert media (IDEXX, Westbrook, ME, USA), respectively, providing a culturable bacteria detection range between two and 2,491.6 MPN per 100 mL of water (additional details can be found in the Supplementary Material). During each sampling run, 100 mL of nuclease free water was prepared at site 4 with the Colilert and Enterolert media, respectively, and transported to the lab for processing as mentioned to serve as a trip sterility blank. Cultures of Enterococcus faecalis (ATCC® 29212^TM) and E. coli (ATCC® 29212^TM) obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) were prepared under standard growth conditions in Luria-Bertani broth at 37 °C overnight (16 h) with aeration, and 100 μl was added to 99.9 mL of nuclease free water containing IDEXX media to serve as positive controls for the Enterolert and Colilert assays, respectively.

A total of 600 mL of water from each sampling event was concentrated by using a combination of vacuum and syringe filtration. Water samples were first passed through six 0.6 μm pore size 25 mm polycarbonate Whatman Nuclepore Track-Etch Membranes (Millipore Sigma, St. Louis, MO, USA) by vacuum filtration, to remove trace minerals and other particulates and expedite subsequent filtration. Resulting filtrate was collected aseptically and passed consecutively through a single 0.22 μm pore size Sterivex^TM Filter Unit (Millipore Sigma, St. Louis, MO, USA) by using syringe filtration and six 0.1 μm pore size 47 mm polycarbonate Whatman Nuclepore Track-Etch Membranes (Millipore Sigma, St. Louis, MO, USA) by vacuum filtration to capture smaller microorganisms, including some viruses and phages. All filter membranes were stored at −80°C until DNA was prepared (<30 days).

For each sample, total DNA was isolated from the microbial biomass collected on the Sterivex^TM Filter Unit using the Qiagen DNeasy Power Water Sterivex Kit (Qiagen, Germantown, MD, USA), following manufacturer's instructions, to obtain a final elution volume of 60 μl. DNA was prepared from all 12 filter membranes using the Qiagen DNeasy Power Soil Kit (Qiagen, Germantown, MD, USA), with the following modifications for DNA extraction from the filter membranes. The 12 filter membranes for each sample were cut into ribbons ~2 mm by 10 mm and distributed evenly amongst six Power Bead Tubes included in the Qiagen Power Soil Kit (Qiagen, Germantown, MD, USA), and final elution volume for each of the six membrane preparations was 20 μl. Eluted DNA was pooled for each Sterivex^TM Filter Unit and the six membrane preparations to achieve 180 μl. DNA was purified using DNA Clean and Concentrator^TM-25 Kit (Zymo Research, Irvine, CA, USA), following manufacturer's instructions, providing a final elution volume of 50 μL.

Concentrations of genomic dsDNA were measured using Qubit® dsDNA High Sensitivity Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA) on an Invitrogen Qubit® 4.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). DNA libraries were prepared using Nextera XT DNA Library Prep Kit (Illumina Inc., San Diego, CA, USA). All DNA libraries were quantified, as previously mentioned, and sequenced using an Illumina HiSeq4000 Instrument (Illumina Inc., San Diego, CA, USA) with a 2 × 150 bp run. A negative sequencing control, consisting of nuclease-free water, and a sequencing standard, i.e., ZymoBIOMICS^TM Microbial Community Standard (Zymo Research, Irvine, CA, USA), were included for quality assurance of high-throughput sequencing. Metagenomic samples were sequenced with an average of 50.9 (min = 44.2; max = 56.8) million sequence read depth across samples.

Concentrations of cultural FIB were determined using the IDEXX/Quanti-Tray FIB assay, and results for each sample are shown in Figure 2A. Within dry weather sampling events, the concentrations of E. coli ranged between 117.8 and 248.1 MPN/100 mL, and enterococci ranged from 36.8 to 313 MPN/100 mL. During the post rainfall sampling event, variation in the concentration of E. coli and enterococci could only be assessed qualitatively because dilutions tested yielded results exceeding the upper limit of quantification; that is, concentrations of E. coli (sites 1 and 2) and enterococci (sites 1-3) in these samples were >2,491.6 MPN/100 mL. Thus, post rainfall, all samples exceeded the local single sample recreational water quality thresholds of 410 MPN/100 mL and 107 MPN/100 mL for E. coli and enterococci, respectively.

Culturable E. coli and enterococci were detected in all samples regardless of site (Figure 2A). Furthermore, all samples post rainfall exceeded the USEPA recommended water quality standards for single sample values of recreational freshwater watersheds in the study area (E. coli, 410 MPN/100 mL; enterococci, 107 MPN/100 mL) (EPA, 2012; DCMR, 2014). A similar trend was observed with FIB identified by qPCR where the indicators were detected in 95.1% of the samples. After rainfall, Enterococci and E. coli genetic marker weighted average log₁₀ concentrations were 1.26- and 1.52-log₁₀ copies per 100 mL times higher, respectively. These findings are in agreement with previous studies where precipitation was found to contribute to high concentrations of FIB in surface water (Lee et al., 2020). High FIB concentrations are commonly reported in many urban watersheds throughout the U.S. For example, mean E. coli concentration of 1,156 MPN/100 mL was reported for a tributary of the Des Moines River in Iowa, USA (Schilling et al., 2009).

Generally, when FIB were detected in post rainfall samples by culture and qPCR, WMS supported the increased RA of FIB in samples collected after rainfall (Table 3). Unlike culture and qPCR methods, taxonomic profiling of metagenomic sequencing reads associated with E. coli and enterococci indicated comparatively low numbers. The results suggest that WMS may not be suitable for FIB monitoring alone since results can be susceptible to shifts in occurrence associated with changes in RA of other community members. As a result, WMS is best applied as a complement to culture and qPCR by providing useful information toward understanding microbial diversity and occurrence of AR genes, presence of enteric microorganisms, as well as additional information on sources of contamination.

Identifying sources of fecal pollution in surface waters is a public health challenge, and qPCR is currently employed for surface water quality testing, namely to quantify specific genes in environmental samples (Dick et al., 2005; Shanks et al., 2008, 2009; Mieszkin et al., 2009; Green et al., 2012; Lee et al., 2013; Ohad et al., 2016; Rojas et al., 2017). In the study reported here, the weighted average fecal score ratio (log₁₀ copies per 100 mL) of the HF183/BacR287 genetic marker was −1.11 or ~12.8-fold greater after rainfall (Table 1, Figure 3). This indicates sewage and/or other sources of human waste, such as storm drain overflow that can contribute to water quality degradation in the urban watershed. In contrast, the avian-associated GFD genetic marker showed no significant difference between post rainfall and dry weather samples suggesting birds do impact water quality but perhaps independent of rainfall patterns. The ruminant-associated genetic marker (Rum2Bac) was detected in all samples but at very low incidence in this watershed. The canine-associated DG3 genetic marker was detected only in samples of water collected after a rainfall. Most probably surface runoff introduced canine excretory waste into the watershed.

The WMS and bioinformatic annotation strategy employed in this study identified fecal-associated microorganisms with finer taxonomic resolution; this has been observed in an earlier study (Brumfield et al., 2020b). Microbial species reported to be closely associated with the gut microbiota of specific animal groups (Puig et al., 1999; Buchan et al., 2001; Newton et al., 2013; Shanks et al., 2013; Harwood et al., 2017; Jebri et al., 2017) suggest WMS can be used to shed light on potential sources of fecal pollution. Viruses were detected in samples collected after a rainfall, with crAssphage exhibiting a strikingly similar pattern to the HF183/BacR287 qPCR genetic marker, showing a change between dry weather and post rainfall samples. CrAssphage was recently reported as a dsDNA Bacteroides bacteriophage tightly associated with human waste (Dutilh et al., 2014) and highly abundant in sewage (Korajkic et al., 2020). Future experiments comparing qPCR (Stachler et al., 2017) and WMS measurements of crAssphage in surface waters are planned to confirm this observation.

The autochthonous microbiome of surface water is important in sustainability of natural ecosystems, and potential microbial shifts during periods of dry weather and post rainfall can provide useful information to microbial ecology and sources impacting their occurrence. Previous investigation has associated high taxonomic diversity of lotic ecosystems in urban areas with low fecal contamination in those waterways (Paruch et al., 2019). In the current study, increased fecal pollution appeared to be paralleled with higher alpha diversity following rainfall (Figure 2B), and work is underway to confirm the impact of stormwater runoff on fecal pollution and microbial diversity.

Dominant bacterial phyla detected in the urban watershed (Figures 4, 5) were similar to those detected in surface water of other freshwater aquatic systems (Newton et al., 2011; Staley et al., 2013, 2014; Brown et al., 2015; Hamner et al., 2019), i.e., Proteobacteria comprised more than 60% RA in both dry weather and post rainfall samples. Actinobacteria, a common soil microbe often present in pristine waterbodies (Jenkins et al., 2009; Ghai et al., 2011), was also prominent in the microbiome of all samples examined in this study. The majority of viruses detected in the urban watershed were bacteriophages belonging to three major bacteriophage families, i.e., Myoviridae, Siphoviridae, and Podoviridae (Figures 4, 5), a finding in agreement with other reported freshwater viromes (Mohiuddin and Schellhorn, 2015). Thaumarchaeota spp., a group of ammonia-oxidizing archaea detected in urban areas demonstrating high rates of nitrification (Reisinger et al., 2016; Epp Schmidt et al., 2019), were present in most samples in this study. Nitrospirales were detected predominantly in samples collected following rainfall. These archaea are considered to be important players in recovery of microbial species composing soil microbiomes following disruptive flooding events (Wang et al., 2019). Clavaria fumosa, a fungus, and the protozoan Pseudoperonospora cubensis, reported by Lee and colleagues (Lee et al., 2020) who used WMS to define the microbial species composition of residential urban stormwater runoff, were detected in this study.

Most watersheds research currently employs FIB identified by culture methods, and a few investigations have adopted qPCR, primarily for enterococci (USEPA, 2012), to link traditional water quality criteria standard definitions to host-associated qPCR. Rapid qPCR for enterococci and E. coli offers a shorter sample processing time (<3 h), compared to traditional FIB culturing (> 18 h). Host-associated genetic markers reliably track common fecal pollution sources, e.g., human, avian, dog, and ruminant, as well as provide quantitative information useful for water quality management. However, these methods were all designed and optimized to detect and quantify a known genetic marker, usually requiring an individual assay to detect each specific gene, without being able to identify uncharacterized microorganisms, i.e., any potential pathogens not yet recognized. Here, use of metagenomic analysis includes thousands of targets including those not yet identified by culture or qPCR expanding the breadth of information regarding the water quality and our knowledge as it relates to the changes in the microbial population.

WMS has been used to explore AR trends in anthropogenically impacted environments (Karkman et al., 2019) and to detect wastewater-associated enteric pathogens (Stamps et al., 2018), including viruses (Lee S. et al., 2019). Stormwater is considered to promote transmission of AR among bacteria (Di Cesare et al., 2017), as well as introduce enteric pathogens (Ahmed et al., 2018), suggesting occurrence of AR and enteric microorganisms can vary between dry weather and after rainfall in our watershed of interest. In this study, we observed an increase in AR genes detected after a rainfall (Figure 2C). At site 4, which is a location near where the creek debouches into a larger river, there was an abundance of tetracycline resistance genes that were not detected in the dry weather samples. It is possible that the additional AR genes detected at this location were impacted by microorganisms introduced to the creek from the larger river during periods of dry weather. However, the relative abundance of various AR classes detected at each site varied only slightly between sampling events (Figure 2D). These observations are the result of a single grab sample at each location, and additional observations are needed to establish a resistome baseline prior to determining the overall impact that rainfall may have on AR gene composition.

In addition to AR, WMS makes possible detection of multiple wastewater-borne enteric microorganisms without a priori knowledge. For example, Cryptosporidium spp., detected at three of the four sites after rainfall (Table 2), currently is the leading cause of reported cases of diarrhea linked to human parasites in contaminated water in the U.S. (Gharpure et al., 2019) and are also known to infect animals. Similarly, Aeromonas hydrophilia, an opportunistic pathogen causing gastroenteritis and blood infections was detected in all samples and at slightly increased in RA after rainfall (Table 2). These findings may add support to other reports concluding that Aeromonas hydrophilia is omnipresent at low abundance in surface water (Poffé and de Beeck, 1991), but at high concentrations in raw wastewater (Shannon et al., 2007).

The source of microbial contamination of urban watersheds is often varied and can originate from farms, animal feed lots, septic tanks, combined sewer overflow, among others, and modes of delivery can introduce multiple fecal sources to the waterway which can make mitigation difficult. A microbial signature approach, to identify fecal pollution in waters off an urban coast of Lake Michigan, USA, used 16S rRNA sequencing to classify sequences to three wastewater-associated bacterial genera (Acinetobacter, Arcobacter, and Trichococcus) and five fecal-associated bacterial families (Bacteroidaceae, Porphyromonadaceae, Clostridiaceae, Lachnospiraceae, and Ruminococcaceae) (Newton et al., 2013). Newton and colleagues determined the RA of sewer and fecal signatures increased to >2% of the measured surface water bacterial communities following sewer overflow. Similarly, during a metagenomic survey of wastewater in the United Kingdom, the genera Arcobacter and Aeromonas were identified as predominant fecal pollution indicators (Acharya et al., 2020). In this study, with the exception of Trichococcus, these sewer and fecal signatures were detected at varying RA between dry and post-rainfall sampling events (Table 3, Figures 4, 5).

Metagenomic sequencing can be used to detect and identify a wide range of microorganisms, including bacteria, viruses, fungi, and protists, in some cases to sub-species level. However, viability or infectious potential of detected microorganisms requires additional analyses of metabolic activity. Metagenomic analysis via short read lengths limits the information available within a single read, and characterization of sophisticated genomic structures requires assembled genomes for analysis (Ayling et al., 2020). However, the coverage of microorganisms present at lower RA is often not sufficient to obtain de novo metagenomic assembled genomes. Until novel metagenomic assembly tools are developed, the taxonomic annotation of unassembled sequencing reads remains a viable option for metagenomic data analysis to conserve as much of the less abundant species sequence as possible (Lax et al., 2012; Ponnusamy et al., 2016; Roy et al., 2018; Connelly et al., 2019; Brumfield et al., 2020a). Direct testing of sediments and potential fecal pollution sources harboring pathogens near the study area could also be helpful to index and further describe the microbial communities associated with stormwater runoff and to differentiate the transient stormwater population from the indigenous microbial population.

This pilot study demonstrates the feasibility and application of combining traditional and non-conventional techniques to characterize microbial communities in ambient waters and identify potential sources of pollution. FIB by culture, qPCR amplification of FIB and host-associated genetic markers, and WMS to detect, identify, and enumerate bacteria, archaea, fungi, protists, and viruses were employed. This comprehensive strategy provided useful insights of the microbial constituents present during dry weather and post rainfall in the surface water of an urban watershed and could potentially improve water quality management in the U.S. and globally (United Nations, 2018). Trends were characterized with a limited number of grab samples and are illustrative of a single event representing temporal shifts in microbial communities before and after rainfall. Further assessments are warranted to identify the naturally occurring community ecology and establish a comparative baseline that can be used to verify microbial shifts between dry and rain event conditions, explore the influence of antecedent dry period and stream flow, as well as characterize potential public health significance of metagenomic DNA sequence findings. Future stormwater investigations could also shed light on potential shifts in WMS and qPCR results in response to storm size and duration, soil and particulate content, season, groundwater height, elevation, and land use. It could also be useful to compare sample processing, DNA sequencing, and bioinformatic approaches used here with other strategies. It is worth noting that WMS data reporting is typically limited to shifts in the relative abundance of sequence reads. While this practice can provide a wealth of new information, it does not conclude absolute quantification of specific genomic targets of interest, viability, or infectious potential of the detected microorganisms. Furthermore, the use of WMS for monitoring microbial communities can be more expensive and time consuming than traditional techniques and require advanced bioinformatics software for analyses. Future research directions could also include the exploration of inter-method correlation between different microbial water quality indicators (Acharya et al., 2019) and compare different WMS strategies for routine integration of these three methods for water quality monitoring. Until the above future research directions are fully characterized and considered in the context of water management, WMS is best applied as a complement to established culture and qPCR practices.