DYNAMICS OF MICROBIAL AGENTS IN URBAN SEWERS, COMBINED SEWER OVERFLOW AND RECEIVING SURFACE WATERS By

ALESSIA ERAMO

A dissertation submitted to the

School of Graduate Studies

Rutgers, The State University of New Jersey

In partial fulfillment of the requirements

For the degree of

Doctor of Philosophy

Graduate Program in Civil and Environmental Engineering

Written under the direction of

Nicole L. Fahrenfeld, PhD

And approved by

______

______

______

New Brunswick, New Jersey

October 2018

1 ABSTRACT OF THE DISSERTATION

DYNAMICS OF MICROBIAL AGENTS IN URBAN SEWERS, COMBINED SEWER

OVERFLOW AND RECEIVING SURFACE WATERS

By ALESSIA ERAMO

Dissertation Director:

Dr. Nicole L. Fahrenfeld

Microbial pollution is one of the leading causes of surface water impairments. It is associated with wet weather events expected to increase due to climate change in many cities with outdated infrastructure. Of particular concern is the release of emerging microbial contaminants. Antibiotic resistance genes (ARGs) have been linked to elevated rates of antibiotic resistant infections, but significant gaps in understanding the environmental fate of ARG limit the ability to comprehensively assess threats to human health. The objective of this work was to investigate the treatment and fate of microbial agents in sewage, sewers, and wet weather flows towards a better understanding of the risk associated with their release into the environment.

A series of field and bench-scale studies were performed towards better understanding the water quality impact of wet weather flows and the performance of end- of-pipe treatment for combined sewer overflow effluent. (1) First, a field study was conducted to characterize the intra- and inter-storm variability in ARG and fecal marker

ii genes concentrations and the microbial community structures during CSOs. The partitioning of gene targets onto settleable particles during CSO events was monitored to provide insight into end-of-pipe treatment and fate upon release to surface water. ARG, fecal indicator and wastewater signature at the outfall varied both during and between storms and timing of peak concentrations targets did not necessarily coincide. The majority of ARG and fecal indicator concentrations were attached to particles rather than in the free phase, suggesting promise for treatment by enhanced sedimentation i.e. hydrodynamic separation. (2) Then, further insight into potential treatment for ARG and fecal indicator marker genes was obtained by investigating peracetic acid (PAA) disinfection kinetics in simulated CSO effluent. Using viability-based qPCR, PAA was found to be an effective disinfectant for reducing concentrations of ARG originating from viable cells but was unsuccessful in destroying the DNA. PAA disinfection resulted in significant shifts in the microbial community. However, further treatment would be needed to remove or destroy

ARG. (3) Finally, to complement the first two studies, the effectiveness of end-of-pipe treatment with removal of settleable particles and PAA disinfection was investigated with traditional cultivation-based methods. The potential for the treatment train to select for sul1 gene carrying E. coli and to promote regrowth of indicator organisms upon release to estuarine waters was investigated. The treatment train achieved >2 log removal of total coliform and E.coli with significant removal attributable to disinfection but not hydrodynamic separation, TSS removal by hydrodynamic separation may have enhanced disinfection. Incubation of surface water inoculated with treated CSO did not result in regrowth of fecal indicators. Although the proportion of E.coli carrying sul1was greater after disinfection, the concentration of E.coli CFU per 100 ml carrying sul1 decreased

iii significantly in disinfected samples and these targets were not observed after a seven-day incubation period.

Towards understanding the availability of ARG for proliferation via different mechanisms, a field study was conducted to investigate the relative proportion of ARGs in cells with intact membranes to total ARGs observed in wastewater treatment plant

(WWTP) effluent and receiving waters. ARGs in the effluent from three municipal wastewater treatment plants and the receiving surface waters was investigated using a viability-based qPCR technique (vPCR) with propidium monoazide (PMA). ARGs, fecal indicator marker genes BacHum, and 16S rRNA gene copies were found to be significantly lower in viable-cells than in total concentrations for WWTP effluent. Viable-cell and total gene copy concentrations were similar in downstream samples except for tet(G).

Differences with respect to season in the prevalence of nonviable ARG in surface water or

WWTP effluent were not observed. The results of this study indicate that qPCR may overestimate viable-cell ARG and fecal indicator genes in WWTP effluent but not necessarily in the surface water.

A final field study was conducted to characterize the factors that drive the loading of microbial agents in sewers. Sewer sediments represent an important source of contaminants released during overflow events. The amount of attenuation, growth and/or selection for antibiotic resistant microbes and other pathogens in this matrix is poorly understood. Sewer sediment and wastewater influent samples were collected from five wastewater collection systems over two seasons. ARG were more abundance in sewer sediments compared to wastewater. Differences in ARG concentrations between season

iv and sewer type (separate vs. combined) were observed, but correlations between ARG and heavy metals were generally not observed.

Overall the results presented provide new insights into the fate of microbial contaminants in sewers, sewage, wet weather flows, and end-of-pipe treatment systems.

The results of these studies will help inform future treatment for ARG from urban water sources and future risk assessments for these emerging contaminants.

v Acknowledgements

I would like to extend the most heartfelt gratitude to my advisor, Dr. Nicole

Fahrenfeld, for her guidance and advice over the last few years. The skills and education

I have learned from working with her are invaluable. I appreciate her helping me to grow as a researcher and for allowing me to explore opportunities to apply my interests in other areas, in particular for the opportunity to participate in the Eagleton Fellowship Program.

It was an insightful and worthwhile experience during my time at Rutgers. I also appreciate the opportunities to travel to numerous conferences and connect with the larger scientific community.

I would like to thank my committee, Dr. Guo, Dr. Mazurek, and Dr. Fennell, for serving on my committee, for their interest in my research, and for providing valuable suggestions and comments. Thanks also to the members of Dr. Fahrenfeld’s group who I have worked with over the past years for their suggestions, assistance, and friendship.

I would like to thank our many collaborators at wastewater treatment plants throughout New Jersey and beyond who made my research possible by providing my samples. In particular, the industry’s willingness to support progress in academic research with such flexibility, friendliness, and enthusiasm has been both inspiring and helpful.

Finally, I would like to thank all the cheerleaders I get to call my friends and family. I am blessed by endless support from great friends, a super sister, amazing parents, and the best husband.

vi Table of Contents ABSTRACT OF THE DISSERTATION ...... ii Acknowledgements ...... iv List of Tables ...... xi List of Figures ...... xiii 1. CHAPTER 1: Introduction ...... 1 1.1. Microbial water quality in urban surface waters ...... 1 1.2. Antibiotic resistance: a modern health crisis ...... 3 1.3. Application of new methods towards parameterizing the hazard posed by ARG ...... 4 1.4. Intersection of aging infrastructure and emerging contaminants ...... 6 1.5. Research Questions ...... 8 1.6. Annotated Dissertation Outline ...... 12 2. CHAPTER 2: Partitioning of antibiotic resistance genes and fecal indicators varies intra and inter-storm during combined sewer overflows ...... 15 2.1. Abstract ...... 15 2.2. Introduction ...... 16 2.3. Materials and Methods ...... 19 2.3.1. Sampling and Water Quality Analysis ...... 19 2.3.2. Biomolecular Analyses ...... 21 2.3.3. Statistical Analyses ...... 22 2.4. Results ...... 23 2.4.1. Partitioning of ARGs and Microbial Contaminants Across the Hydrograph ...... 23 2.4.2. Timing of Peak Microbial Agent Release ...... 30 2.5. Discussion ...... 33 2.5.1. Dynamics of Attached and Suspended ARG Concentrations ...... 33 2.5.2. Timing and Partitioning of Wastewater Signatures and Fecal Indicators in CSO Effluent 36 2.5.3. Implications for Design and Operation of Hydrodynamic Separators ...... 39 2.6. Conclusions ...... 40 2.7. Supplemental Information ...... 41 3. CHAPTER 3: Disinfection of microbial agents in combined sewer overflows using the green disinfectant peracetic acid ...... 51 3.1. Abstract ...... 51

vii 3.2. Introduction ...... 52 3.3. Materials and Methods ...... 55 3.3.1. Disinfection Experiment ...... 55 3.3.2. PAA Decay ...... 57 3.3.3. Viability Analysis ...... 57 3.3.4. Cultivation...... 58 3.3.5. Molecular Methods ...... 58 3.3.6. Data Analysis ...... 60 3.4. Results ...... 61 3.4.1. Disinfection of ARG-carrying cells ...... 61 3.4.2. Disinfection of Indicator Organism Markers and Cultivation ...... 64 3.4.3. PAA Decay Experiments ...... 65 3.4.4. Water Quality ...... 67 3.4.5. Total Bacterial Community Disinfection ...... 68 3.4.6. Kinetics ...... 69 3.5. Discussion ...... 71 3.5.1. PAA for Treating CSO ...... 71 3.5.2. PAA Degradation ...... 74 3.5.3. Mode of Action ...... 74 3.5.4. Gene to Gene Comparisons...... 77 3.6. Conclusions ...... 80 3.7. Supplemental Information ...... 82 4. CHAPTER 4: Settling and peracetic acid for end-of-pipe treatment of sul1-carrying indicator organisms and impact on receiving water ...... 85 4.1. Abstract ...... 85 4.2. Introduction ...... 85 4.3. Materials and Methods ...... 88 4.3.1. Sample Collection and CSO Simulation ...... 88 4.3.2. Treatment of Simulated CSO by Settling and PAA Disinfection ...... 90 4.3.3. Regrowth Experiment ...... 91 4.3.4. Water Quality Measurements and Fecal Indicator Enumeration ...... 91 4.3.5. PAA Analysis ...... 93 4.3.6. Biomolecular Analysis of sul1-Carrying E.coli ...... 93

viii 4.3.7. Statistical Analysis ...... 94 4.4. Results ...... 95 4.4.1. Removal of Settleable Particles, Disinfection, Selection ...... 95 4.4.2. Regrowth Upon Release to Surface Water...... 98 4.5. Discussion ...... 99 4.5.1. End-of-pipe Treatment Efficiency: Settling ...... 99 4.5.2. Disinfection Efficiency ...... 100 4.5.3. Water Quality Impacts on Disinfection...... 101 4.5.4. ARG Selection By Treatment and Impacts on Surface Water ...... 103 4.6. Conclusions ...... 104 4.7. Supplemental Information ...... 106 5. CHAPTER 5: Viability-based quantification of antibiotic resistance genes and human fecal markers in wastewater effluent and receiving waters ...... 108 5.1. Abstract ...... 108 5.2. Introduction ...... 108 5.3. Materials and Methods ...... 111 5.3.1. Field Sampling and Water Quality Analysis ...... 111 5.3.2. Cell Collection, Viability Cross-Linking and Biomolecular Analysis ...... 112 5.3.3. Matrix Spike Study ...... 114 5.3.4. Statistical Analysis ...... 115 5.4. Results ...... 115 5.4.1. Total and Viable-Cell Quantification of ARGs, BacHum Fecal Indicator Marker, and 16S rRNA in WWTP Effluent and Downstream Samples...... 115 5.4.2. Comparison of WWTP Effluent and Downstream Gene Concentrations ...... 119 5.4.3. Matrix Spike Experiment ...... 120 5.4.4. WWTP and Seasonal Impacts ...... 122 5.5. Discussion ...... 123 5.5.1. Total and Viable-Cell Gene Concentrations in WWTP Effluent and Downstream Samples 123 5.5.2. Comparison of WWTP Effluent and Receiving Water ...... 127 5.5.3. vPCR Versus eDNA Extraction Methods ...... 127 5.5.4. Implications for Understanding the Mechanisms Driving ARG Fate ...... 128 5.6. Conclusions ...... 129 5.6.1. Supplemental Information...... 131

ix 6. CHAPTER 6: Fate of microbial agents in wastewater collection systems ...... 134 6.1. Abstract ...... 134 6.2. Introduction ...... 134 6.3. Materials and Methods ...... 136 6.3.1. Sewer Sediment and Wastewater Sampling ...... 136 6.3.2. Chemical Characterization of Field Samples ...... 138 6.3.3. Biomolecular Analyses ...... 139 6.3.4. Statistical Analyses ...... 140 6.4. Results ...... 140 6.4.1. ARG in Sewer Sediment and Wastewater ...... 140 6.4.2. Metals in Sewer Sediment and Wastewater ...... 143 6.4.3. PCA Analysis ...... 144 6.5. Discussion ...... 145 6.5.1. Impact of Matrix, Season, and Sewer Type on ARG ...... 145 6.5.2. Sewer Sediments as Sources of Resistance ...... 148 6.6. Conclusions ...... 149 6.7. Supplemental Information ...... 150 CHAPTER 7: Conclusion ...... 153 References ...... 157 Appendix A: Shifts in microbial community structure and function in surface waters impacted by unconventional oil and gas waste water revealed by metagenomics ...... 171

x List of Tables Table 2.1 Example removal efficiency for minimum and maximum surrogate particles representing the attached and suspended fractions using an example hydrodynamic separator unit (i.e., Stromceptor STC 4800 with 3.7m diameter, 3.4m settling depth, and flow rate 0.051 m3/s). Hydrodynamic separators can sized during design to improve removal. For these estimates the following constants were used: kinematic viscosity water ν =1.12×10-6 m2/s, gravity 9.81m/s2, and water density ρ=999kg/m3...... 49 Table 3.1 Log removal of viable-cell ARGs, fecal indicator and 16S rRNA gene concentrations during treatment of simulated CSO with PAA. Results represent averages ± standard deviation (n=3) or ± maximum and minimum values (n=2). Bold results indicate a statistically significant decrease in genes originating from viable cells was observed from T=0 min (p<0.05)...... 65 Table 3.2 Average ± standard deviation for Chick-Watson disinfection coefficients (α) and R2 values of model fit combined for both 23% and 40% WW from different WWTP (n=3)...... 70 Table 3.3 Primers, annealing temperatures, and amplicon lengths...... 82 Table 3.4 Water quality data for source wastewater from disinfection experiments (Experiments 1 and 2) and PAA degradation experiment (Experiment 3)...... 84 Table 4.1 Concentrations of cultivable E. coli and TC measured in simulated CSO and after different treatments and log removal of the fecal indicator organisms attributable to treatment by removal of settleable particles (settling), disinfection of the suspended fraction of simulated CSO (supernatant), and disinfection of simulated CSO without settling. Log removal values are based on comparisons to starting concentrations in simulated CSO. The results represent average removals and standard deviations with statistically significant removals in bolded text...... 95 Table 4.2 Dilutions and volumes averaged for each treatment condition. At least two plates were tested for each condition and countable plates are included in the table. Countable plates contained 20-80 E.coli CFU and <200 TC CFU. The numbers outside parentheses represent the sample dilutions and the numbers inside parentheses represent the corresponding volumes filtered of the dilutions (milliliters)...... 106 Table 4.3 Dilutions and volumes used for each regrowth condition. At least two plates were tested for each condition and countable plates are included in the table. Countable plates contained 20-80 E.coli CFU and <200 TC CFU unless indicated. When all plates for a given sample resulted in E. coli CFU less than the target minimum of 20, the most nearly acceptable count was reported, in accordance with ASTM D5465-16 (ASTM International 2016). These plates are identified by the notation. The numbers outside parentheses represent the sample dilutions and the numbers inside parentheses represent the corresponding volumes filtered of the dilutions (milliliters)...... 107 Table 5.1 Average pH, conductivity and TSS (+/- standard deviation) collected from upstream, influent, effluent, and downstream samples from three wastewater treatment plants during the summer and winter seasons (n=2* or 3)...... 122

xi Table 5.2 Flow rate categories for three activated sludge wastewater treatment plants disinfecting with chlorine sampled and approximate distances between their discharge points and downstream sampling locations...... 131 Table 5.3 Field sampling details for wastewater treatment plants A, B, and C downstream and upstream samples collected during the summer and winter seasons. .. 131 Table 5.4 Primers, annealing temperatures, and amplicon lengths...... 132 Table 6.1 Field sampling details...... 138 Table 6.2 Chemical parameters in field samples...... 150

xii List of Figures Figure 2.1 Map of sampling locations for wet weather sampling in three waterways impacted by CSOs, location of CSO outfall sampled across precipitation event, and documented CSO outfalls...... 21 Figure 2.2 Storm II a. rainfall, total suspended solids (TSS), and conductivity, b. sul1, c. suspended bacteria by class, and d. attached bacteria by class ...... 24 Figure 2.3 Storm III a. rainfall, total suspended solids (TSS), and conductivity, b. sul1 flux, c. suspended bacteria by class, and d. attached bacteria by class ...... 25 Figure 2.4 Attached and suspended tet(G) gene copies/16S rRNA gene copies and rainfall across (a) Storm I, (b) Storm II, and (c) Storm III. Storm I results represent the average of duplicate samples with a relative difference of ±1.6-20.5% (log 16S normalized data)...... 26 Figure 2.5 sul1, sul2, tet(G), and tet(O) ARG concentrations normalized to 16S rRNA gene copies observed outside of a CSO outfall during three storm events (H1_s), the Hudson at a separate sampling location (H2), the Passaic (P), and Raritan (R) during baseflow (b) and wet weather (w)...... 28 Figure 2.6 Attached and suspended BacHum gene copies/16S rRNA gene copies and rainfall across (a) Storm I, (b) Storm II, and (c) Storm III. Storm I results represent the average of duplicate samples with a relative difference of ±1.7-19% (log 16S normalized data)...... 29 Figure 2.7 Non-metric multidimensional scaling analysis of bacterial community structures for wastewater influent concentrated by centrifugation (ww-p) or filtering (ww- p), and suspended (x) and attached (triangles) samples collected across three storms (I purple, II green, and III blue). The numbers above the storm sample symbols represent the order of samples (1 being the first sample collected, etc.). The overlay represents clusters with 85% similarity...... 32 Figure 2.8 Precipitation and mean daily air temperature data for nearest gage for the (a) daily for the Passaic (data from Newark, NJ), (b) hourly for Passaic during wet weather event (data from Newark, NJ), (c) daily for the Hudson at Jersey City (rainfall data from New York, NY), (d) hourly for the Hudson during wet weather event (data from New York, NY), (e) daily for Raritan Bay (data from Newark, NJ, and (f) hourly for wet weather event (data from Linden, NJ)...... 41 Figure 2.9 Hourly precipitation, sampling time, and time of high tide for the day of (a) Storm I, (b), Storm II, and (c) Storm III. (d) Daily cumulative precipitation across the study period. Precipitation data presented for Newark, NJ, the location of the nearest rain gauge...... 42 Figure 2.10 Storm I a. ARG flux, b. suspended bacteria by class and rainfall, c. attached bacteria by class. Sequencing results shown are for replicate samples collected at the same time...... 43 Figure 2.11 Attached and suspended tet(G) gene copies/16S rRNA gene copies and rainfall across (a) Storm I, (b) Storm II, and (c) Storm III...... 44 Figure 2.12 Attached and suspended BacHum gene copies/16S rRNA gene copies and rainfall across (a) Storm I, (b) Storm II, and (c) Storm III...... 45

xiii Figure 2.13 Cluster analysis for bacterial communities at the class level. Numbers refer to sample number (1 for first, etc.) for a given storm, “a” and “b” refer to replicates. Red branches connect samples with bacterial community structures that were not different via SIMPROF test...... 46 Figure 2.14 Rarefaction curves for (a) Storm I, (b) Storm II, (c) Storm III, and (d) simulated CSO. Numbers for storm samples indicate sample order. Attached samples are indicated by “att” and suspended samples are indicated by “plank.” Storm replicates are labeled as “a” and “b.” CSO samples were diluted wastewater (WW) concentrated by filtering or by centrifugation (“pellet”)...... 46 Figure 2.15 LEfSe analysis on wastewater, attached storm, and suspended storm (excluding storm samples clustering with wastewater) indicates biomarkers for each matrix at the Phylum and Class level...... 47 Figure 2.16 Attached and suspended calculated Enterobacteriales and rainfall across (a) Storm I, (b) Storm II, and (c) Storm III...... 48 Figure 2.17 Concentrations of ARGs normalized to 16S rRNA from composited sediment samples collected during base flow (dry) or wet weather conditions in the Passaic River (P) and Raritan Bay (R)...... 49 Figure 3.1 Comparison of total and viable cell gene copy numbers with different PAA treatment and exposure times for (a, b) sul1, (c,d) tet(G), (e,f) mexB, (g,h) BacHum, and (i,j) 16S rRNA genes. Experiments were performed with 23% wastewater from WWTPa or 40% wastewater from WWTPb to create the simulated CSO effluent. Experiments were performed for up to 60 min for WWTPa and 10 min for WWTPb. Error bars represent standard deviation of replicate (n=3, except as indicated in Table 3.1) samples...... 62 Figure 3.2 PAA concentration remaining (C) compared to initial PAA concentration (Co) during treatment of (a) 23% wastewater from WWTPa, (b) 11.5% wastewater from WWTPa, and (c) 20% or 40% wastewater from WWTPb and immediately after quenching. Treatment was performed with either 5 mg/L or 20 mg/L PAA. Error bars represent standard deviation of replicate (n=3) samples...... 67 Figure 3.3 (a) Cluster analysis and (b) relative abundance of Bacterial phyla for different PAA treatments and exposure times (Ct values in mg·min/mL). Samples connected by red bars on the cluster tree do not have significantly different structures. #’s represent replicate samples that were sequenced...... 70 Figure 3.4 Colony forming units (CFU) on LB agar from 40% WW treated with 0 mg/L or 5 mg/L PAA (n=2 or 3)...... 83 Figure 3.5 Water quality parameters (a) conductivity, (b) chemical oxygen demand (COD), (c) pH and (d) total suspended solids (TSS) in 40% WW treated with 5 mg/L PAA compared to no PAA controls...... 83 Figure 3.6 Rarefaction curves for samples treated with 20mg/L PAA or no treatment controls for 0 or 60 min. Viable indicates samples treated with propidium monoazide prior to submission for sequencing. # indicates replicate number...... 84 Figure 4.1 Map of surface water sampling location and combined sewer outfalls in the vicinity of the sampling location...... 89

xiv Figure 4.2 COD, TSS, and conductivity in the source waste stream (simulated CSO), after removing settleable particles (separated CSO), and after disinfection both with and without prior separation. Results represent the average of replicates [n=5 except for disinfected CSO (no separation) where n=2] and the error bars represent their standard deviation. TSS was not measured after disinfection...... 96 Figure 4.3 Fraction of E.coli CFU colonies (n=10) that contained the sul1 ARG at stages of treatment compared to surface water only and surface water spiked with treated effluent on day 0. E.coli CFU were not observed after 7 days of incubation. Results represent the average of experimental replicates and error bars represent their standard deviation (n=3)...... 97 Figure 4.4 E.coli and total coliform concentrations in reactors with (1) surface water spiked with a 1:10 dilution (v:v) of simulated CSO treated by settling and disinfection (1:10 treated; n=5), (2) surface water spiked with a 1:10 dilution (v:v) of simulated CSO with no treatment (1:10 untreated; n=2) and (3) surface water only (n=5). Error bars represent standard deviation of experimental replicates...... 99 Figure 4.4 Schematic of treatment and regrowth experimental conditions...... 106

Figure 5.1 Concentrations of ARGs (a.) sul1, (b.) tet(G), and (c.) blaTEM, (d.) fecal indicator marker BacHum, and (e.) 16S rRNA present overall (total) and in viable cells only (viable) measured in effluent and downstream samples from wastewater treatment plant A, B, and C during summer and winter seasons. Boxes represent upper and lower quartiles, whiskers extend to high and low data points excluding outliers, and dots indicate outliers...... 117 Figure 5.2 Ratios of viable-cell gene concentrations measured by vPCR to total gene concentrations measured by qPCR in effluent and downstream samples...... 118 Figure 5.3 Concentrations of 16S rRNA in centrifuge-concentrated E. coli culture at exponential phase, river samples only (water pellet), centrifuge-concentrated river samples with a live cell culture spike (water pellet + live), and centrifuge-concentrated river samples with an inactivated (heat-treated) cell culture spike (water pellet + dead). Samples were analyzed by qPCR (0 µM PMA) and vPCR (50 or 100 µM PMA), which represent total and viable-cell concentrations, respectively...... 121

Figure 5.4 Concentrations of ARGs (a.) sul1, (b.) tet(G), and (c.) blaTEM, and (d.) fecal indicator marker BacHum normalized to 16S rRNA concentrations present overall (total) and in viable cells only (viable) measured in effluent (n=17) and downstream (n=18) samples from 3 wastewater treatment plants during summer and winter seasons. Boxes represent upper and lower quartiles, whiskers extend to high and low data points excluding outliers, and dots indicate outliers...... 132 Figure 5.5 Concentrations of ARGs (a.) sul1 and (b.) tet(G) measured in upstream and downstream samples from 3 wastewater treatment plants during summer and winter seasons. Results represent the average of three days of sampling during each season and error bars represent the standard deviation...... 133 Figure 6.1 Relative abundance of sul1 in sewer sediment and wastewater samples collected from two separate and three combined collection systems (n=30)...... 142

xv Figure 6.2 Concentrations of ARGs sul1 and tet(G) in (a) sewer sediments and (b.) wastewater samples from separate (n=12) and combined (n=18) collection systems collected during two seasons...... 142 Figure 6.3 Principal component analysis of chemical parameters of (a.) sewer sediment and corresponding (b.) wastewater. Samples color coded by sewer type [separate (s) or combined (c)] and overlaid by season [winter (W) or summer (S)]...... 145 Figure 6.4 Principal component analysis of chemical parameters of (a.) sewer sediment and corresponding (b.) wastewater. Samples color coded by relative abundance of sul1 or tet(G) with the size of each half related to the relative abundance value. Samples were overlaid by season [winter (W) or summer (S)]...... 145 Figure 6.5 Cadmium concentrations in sewer sediment samples collected in two different seasons from combined or separate sanitary sewer systems. Three combined and two separate sanitary sewer systems were sampled (n=3 per plant and season, total N=30)...... 150 Figure 6.6 Copper concentrations in (a) sewer sediment and (b) wastewater influent samples collected in two different seasons from combined or separate sanitary sewer systems. Three combined and two separate sanitary sewer systems were sampled (n=3 per plant and season, total N=30)...... 151 Figure 6.7 Sieve analysis indicated the percentage of particles smaller than the aperture size listed. Results are shown for data currently available (three combined sewer systems: c1, c2, and c3, and one separate sanitary sewer: s1, for up to three winter samplings: 1, 2, 3 and up to three summer samplings: 4, 5, 6). Data is not available for c2_1, c2_3, s1_2, and s1_3...... 152

xvi 1

1. CHAPTER 1: Introduction

1.1. Microbial water quality in urban surface waters

In the urban environment, elevated bacterial levels from sewer overflows, leaky storm

sewers, runoff, and land disposal of waste impair surface water quality [Pepper et al.

(2018);(US Environmental Protection Agency 2011); others]. Point and nonpoint sources

release microbial agents such as pathogens, viruses, antibiotic resistance genes (ARGs),

antibiotic resistant bacteria (ARB), and indicator organisms to water environments [Rizzo

et al. (2013); Newton et al. (2013); ten Veldhuis et al. (2010)]. This poses a public health

concern because these areas are used for urban recreational activities including boating,

swimming, and fishing (Donovan et al. 2008), drinking water sources (Tondera et al.

2015), and bathing and washing clothes for homeless populations (Donovan et al. 2008).

While water quality in waterways in the US are regulated by the Clean Water Act, many

communities throughout the US struggle to meet safe quality standards towards the goal

of achieving water quality that is fishable and swimmable (US Environmental Protection

Agency 2018). The importance of microbial water quality is highlighted by the fact that

the USEPA identifies fecal indicator bacteria as one of the leading causes of impairments

to surface water (Arnone and Walling 2007).

Microbial water quality in urban areas is worsened during wet weather. Fecal

indicator concentrations resulting from runoff have been shown to be higher for high-

flow conditions during or after rainfall (Krometis et al. 2007). Furthermore, in cities with

combined sewer infrastructure, waterways are regularly impacted by raw sewage.

Combined sewers are antiquated wastewater collection systems that capture both

municipal sewage and stormwater runoff. During periods of wet weather when the 2 volume in the system exceeds wastewater treatment plant capacity, the combined stream overflows into neighboring waterways, bypassing treatment and releasing large quantities of potentially harmful contaminants. It has been shown that flood water contaminated with sewage or other bacterial sources could lead to disease outbreaks, especially in densely populated areas (ten Veldhuis et al. 2010). The frequency of flooding has increased due to accelerated urbanization and more intense rainfall events induced by climate change, posing new challenges for controlling microbial contamination in surface waters. Additionally, increasing temperatures will likely increase the risk of water-borne diseases in recreational waters (Sterk et al. 2015). This problem is exacerbated in cities with aging infrastructure where leaks, breaks, and combined sewer overflow (CSO) events are more frequent.

Fecal indicator organism counts are used to indicate if human wastes have contaminated water bodies (Shibata et al. 2004) and have been studied for many years to establish regulatory targets that protect human health for recreational waters (Wade et al.

2003). While regulations are built upon the fecal indicator organism paradigm, there is motivation for looking beyond these targets. Pathogens poorly correlate with indicators

(Harwood et al. 2005) and different microbes behave differently in the environment

(Suter et al. 2011), therefore, studying a broader range of microbial targets may be desirable. At the same time, emerging contaminants such as ARGs are highlighting a need to understand the fate of harmful DNA in the environment and for application of new techniques to better understand the hazard these contaminants pose. 3

1.2. Antibiotic resistance: a modern health crisis

Antibiotic resistance is one of our most serious public health threats with antibiotic resistant infections responsible for 23,000 deaths and over 2 million infections annually

(Centers for Disease Control and Prevention 2013). Microbial resistance to available antibiotics is developing faster than novel antibiotics can be developed (Spellberg 2010).

For example, in 2016, an E.coli strain with a gene encoding for colistin-resistance was isolated for the first time in the US (McGann et al. 2016). For many years, antibiotics related to colistin were viewed as drugs of last resort and were used to treat complicated and oftentimes hospital-acquired infections (Watkins et al. 2016). Another pertinent example is the global movement of the NDM-1 enzyme. Microbes, or “superbugs,” possessing NDM-1 include various species and are resistant to many classes of antibiotics, such as colistin (Walsh et al. 2011). NDM-1 was originally isolated in India but has since made its way to China (Luo et al. 2013) and France (Diene et al. 2011).

Recently, it has also been detected in New Jersey wastewater samples by the Fahrenfeld

Lab. This highlights the importance of studying ARG in addition to ARB in environmental matrices.

ARGs are segments of DNA that encode for antibiotic resistance and are found on chromosomes or mobile genetic elements such as plasmids. ARGs are found at high concentrations in agriculture, domestic, hospital and industrial wastewater, and aquaculture (Pruden et al. 2013), but can also be found in drinking water (Khan et al.

2016). Moving towards a better understanding of the mechanisms of ARG proliferation and the risks are of interest because the incidence of antibiotic resistant infections in the clinical setting has been linked to environmental sources of ARG e.g.,(Casey et al. 2013). 4

Studying ARGs is different from studying the fate of host cells because ARGs can be shared between bacteria by horizontal gene transfer (HGT) mechanisms: conjugation, transduction, and transformation. Conjugation is the direct transfer of genetic material between neighboring bacteria, transduction is the transfer of such material by viruses, and transformation is the incorporation of extracellular DNA into competent bacteria. These mechanisms allow ARGs to proliferate in the environment at rates different from those that would be expected due to host cell growth and decay. Therefore, there is a non-zero risk to human health posed by ARG present in commensal organisms as well as from extracellular ARG (Ashbolt et al. 2013).

1.3. Application of new methods towards parameterizing the hazard posed by ARG

A comprehensive risk analysis that quantifies ARG uptake and human health effects should include health risks from the release of ARG to the environment (Ashbolt et al.

2013). The risk posed by ARG varies by the host and the novelty of the resistance mechanism (Martínez et al. 2015). This raises questions regarding what targets to monitor for establishing a human health risk assessment (Vikesland et al. 2017). While there is no consensus on which methods to apply to establish the risk posed by environmental ARG and antibiotic resistance (Vikesland et al. 2017), quantitative polymerase chain reaction (qPCR) has become standard practice and been shown to serve as a conservative proxy for ARG transformation (Chang et al. 2017). Traditional qPCR quantifies the number of target gene copies in samples regardless of whether these genes are intra- or extracellular. Extracellular DNA can be taken up by transformation and therefore poses its own unique challenges and concern, but methods to differentiate intra- and extracellular DNA are not standardized (Vikesland et al. 2017). This limits our 5 ability to measure HGT rates in the environment, which remain largely undefined

(Ashbolt et al. 2013).

To improve our understanding of the risk associated with ARG, methods beyond qPCR are needed that can help differentiate DNA originating from viable cells or extracellular or non-viable sources. In the environment, DNA is released mainly due to cell death and autolysis. Oxidants used in wastewater disinfection also lyse cells and release DNA into the water environment (Rizzo et al. 2013). This DNA can remain accessible to bacteria and persist particularly in sediment matrices, but the amount of

DNA released is unclear and largely depends on the conditions preceding cell death

(Nielsen et al. 2007). Few researchers have attempted to differentiate intra- and extracellular ARG in environmental matrices. Extracellular DNA measured using a intracellular/extracellular DNA extraction methods indicated that extracellular DNA generally accounted for less than 0.5% of total DNA in manure sludge samples (Zhang et al. 2013) while extracellular ARG were found in higher concentration than intracellular

ARG in river sediments including those receiving treated wastewater (Mao et al. 2014).

Viability-based molecular techniques may provide further insight into the availability of ARG for proliferation via different mechanisms. Viability-based qPCR (vPCR) method uses propidium monoazide (PMA), a dye that binds to extracellular DNA or

DNA originating from cells with compromised membranes, thereby inhibiting PCR

(Nocker et al. 2007a). Samples treated with PMA represent genes present in viable cells, while samples without PMA treatment represent the total gene copies. Cultivation methods are the current standard for measuring viability and meeting regulatory limits, but vPCR may provide opportunities for higher rates of sample processing due to the 6 rapid performance of PMA and ability to preserve samples once treated . The vPCR method, which has only been applied to one known study for the analysis of ARG

(Mantilla-Calderon 2017), combines the specificity of PCR in targeting specific genes across species and the ability of cultivation methods to measure survival kinetics under different conditions.

1.4. Intersection of aging infrastructure and emerging contaminants

CSO effluent is a widespread source of contaminants and degrades water quality. It contains a mixture of runoff, wastewater, and sewer sediments that introduce many chemical and microbial contaminants of concern (Gasperi et al. 2008). CSO plumes can persist for several days in the receiving water body concentrated at the surface and dispersing downstream (Eaton et al. 2013). Major water bodies throughout the world and in the US are affected by CSOs, including the Great Lakes and the Hudson River. The volume of water released during CSO events is significant. For example, 27 billion gallons of raw sewage from 460 overflows enter the NY Harbor alone each year

(Riverkeeper 2018). In New Jersey, 210 combined sewer outfalls distributed among 21 municipalities result in the release of seven billion gallons of untreated wastewater annually (Van Abs 2014).

The large volumes of sewage released to urban surface waters poses a public health risk. Superstorm Sandy in 2012 caused drinking water advisories to be issued in over 60

NY water treatment plants, some for up to two weeks, due to loss of power and poor receiving water quality (NY State Department of Health 2014). In New Jersey, incidental ingestion of CSO-plagued Passaic River water has been shown to present a high risk of gastrointestinal disease (Donovan et al. 2008), demonstrating one of the dangers posed 7 from bypassing conventional wastewater treatment. In the Hudson River, high levels of

ARB were associated with wet weather (Young et al. 2013). The rise in urban populations and climate change-induced extreme and wet weather events are expected to increase the number overflow events, especially in parts of the US with aging infrastructure.

The sewer contents that are flushed out along with sewage and runoff during CSO events impact receiving surface waters but are poorly characterized. Sewer sediments and other contents of the sewer environment may pose unique public health risks but is understudied for microbial contaminant growth and survival (Fahrenfeld and Bisceglia

2016). Sediment deposition in sewers is widespread (Ashley and Hvitved-Jacobsen

2003) and present in sanitary, storm, and combined sewer systems (Crabtree 1989) and a significant component of the solids load during CSO (Hannouche et al. 2014).

Wastewater solids may settle during conveyance and be eroded and re-suspended during high flow events [Ashley and Crabtree (1992); Ashley and Verbanck (1996)]. In addition, sediments can become cohesive and immobile and further bound by biofilm

(Michelbach 1995). In urban sewers, unique conditions and pressures will shape the sewer microbiome (McLellan et al. 2015). Combined systems may also be affected by variability in precipitation intensity (Mailhot et al. 2015) and flushing of the outfall pipe during rain events. Evidence of dechlorination of polychlorinated biphenyls in wastewater collection systems (Rodenburg et al. 2010) provides insight into unique microbial activity and redox conditions in sewers. Furthermore, the ability of the polio virus to persist in sewers after 800 million liters of wastewater passed through the system 8

(Hovi et al. 2001) suggests sewers should not be underestimated for their ability to harbor dangerous microbes, presenting a risk for sewage workers.

1.5. Research Questions

Research Question 1: How are total ARG in surface water impacted by wet weather and CSO? What is the microbiome of CSO impacted surface water? When are peak concentrations of microbial contaminants observed during wet weather?

CSO effluent has the potential to contribute significant quantities of microbial contaminants to receiving water bodies due to the presence of raw wastewater.

Researchers have monitored indicator bacteria during a CSO event (Passerat et al. 2011), but investigations of microbial targets other than fecal indicators are not known. While

ARB in the Hudson River has been associated with wet weather (Young et al. 2013), studies targeting the flux of ARG during actual CSO events have also not been conducted. The quality of CSO effluent has been reported to vary widely intra-storm, with wastewater accounting for 4–39% of the flow and storm water accounting for the remainder (Passerat et al. 2011). Monitoring for fecal indicators in stormwater runoff during wet weather indicated intra-storm variability in fecal indicators (Krometis et al.

2007). Notably, no data is available to define inter-storm variability for microbial water quality during CSOs. Studies of fecal indicators have evaluated the partitioning of these bacteria between the fraction attached to settleable particles and those in the free phase or attached to nonsettleable particles, which is important since these fractions will have different fates in the water environment (Krometis et al. 2007). But there is a need to understand the partitioning of a wider range of contaminants. Overall, little is known 9 about the concentration, timing, and partitioning of microbes beyond fecal indicators in

CSO, limiting our understanding of CSO as a source of these contaminants. These gaps in knowledge limit our potential to evaluate treatment options for CSO, discussed further in Research Question 2.

Research Question 2: What is the expected impact of end-of-pipe treatment on fecal indicators, antibiotic resistant fecal indicators, ARG, and microbial community?

CSO has been shown to degrade water quality and threaten human health but for many years, CSO-impacted surface water has continued to violate Clean Water Act guidelines primarily due to the high cost associated with upgrading sewer systems.

Treatment of the waste stream at the end of pipe and prior to discharge has been proposed as a lower cost strategy to mitigate CSO and bypass the need to make major changes to the collection system, estimated at $40.8 billion in the US (United States Environmental

Protection Agency 2016). But, little is known about the effect of end-of-pipe treatment systems on ARG and ARG carrying cells. Hydrodynamic separation is an end-of-pipe treatment strategy that targets the removal of settleable particles through extension of the particle flow path. It is typically used to target total suspended solids reduction in stormwater discharge but its performance for removal of a broader range of microbial contaminants has not been measured. Rapid disinfection at the end-of-pipe is another potential treatment strategy. Treatment of ARGs with conventional disinfectants such as chlorine and UV has been challenging because these disinfectants inactivate cells but do not necessarily destroy the DNA these cells carry at commonly applied doses (McKinney and Pruden 2012). Additionally, chlorine is not appropriate for treating CSO due to high 10 levels of organic material present in CSO effluent that can form carcinogenic disinfection byproducts when reacted with chlorine. Thus, peracetic acid (PAA), which does not create known hazardous byproducts (Kitis 2004), and has demonstrated rapid performance (McFadden et al. 2017), is appealing for this application without pretreatment. PAA’s effectiveness for inactivating indicator organisms has been reported but its ability to inactivate ARG-carrying cells and destroy ARG is unknown.

Additionally, the response of the microbial community to PAA disinfection has not been previously investigated. Limited research has been conducted on end-of-pipe treatment trains. One known study was a pilot study in New Jersey that targeted cultivable fecal indicators but did not measure ARB, ARG, or microbial community shifts post-treatment

(Bayonne Municipal Utilities Authority 2017). Environmental implications from the treatment of CSO are also not well characterized, such as the potential for CSO treatment to select for ARG. Furthermore, the regrowth potential of PAA-treated fecal indicators has been generally studied for short periods of time and without inoculation into surface water.

Research Question 3: What can viability-based qPCR methods teach us about the mechanisms of disinfection and availability of ARG for transformation in different environments?

The abundance of ARGs in environmental samples is often investigated by qPCR, which is specific for targeting DNA segments of interest, but lacks the ability to differentiate whether the genes are in viable cells or not. Whether these genes are present in viable cells is an important question for fate of ARG given that it has implications for 11 the mechanisms of ARG proliferation, As discussed in Research Question 2, disinfection studies typically rely on cultivation methods to measure disinfection performance, but they are unable to differentiate the relative proportion of viable and non-viable DNA sources following disinfection with oxidants and in receiving water bodies.

Differentiating DNA in viable versus non-viable sources is of interest for understanding the amount of extracellular ARG available for HGT by transformation in treated waste streams. WWTP effluent is often disinfected with oxidants prior to discharge and has been described as a hotspot for the spread of ARB and ARGs in the urban environment

(Rizzo et al. 2013). Questions remain about the relative importance of extracellular and nonviable sources of ARGs in the environment, including surface waters receiving treated wastewater effluent. vPCR methods have not been applied to provide insight into potential fate of ARGs released from WWTPs post-disinfection.

Research Question 4: What factors drive differences in the loading of microbial agents in sewers?

ARGs have the potential to accumulate in sewer sediments (Devarajan et al. 2015), but the ability of sewer sediments to serve as hotspots for ARG proliferation is unknown.

The factors driving the fate of microbial agents in sewer solids are a result of complex physical and biochemical processes and are poorly understood. Results of previous research indicates that the composition of sewer deposits will be related to the wastewater and runoff conveyed in the sewers, but that changes will occur within the microbial community upon deposition of sediments. However, little is known about the dynamics 12 of microbial agents within sewer sediments, the relationship between sewer sediments and wastewater, and factors driving contaminant concentrations in sewers.

1.6. Annotated Dissertation Outline

Chapter 1: Introduction

Chapter 2: Partitioning of antibiotic resistance genes and fecal indicators varies intra and inter-storm during combined sewer overflows

Field studies were conducted to investigate the intra- and inter-storm variability in ARG and fecal marker genes concentrations as well as the microbial community structure. The partitioning of both targets onto settleable particles during CSO events was monitored to provide insight into end-of-pipe treatment by vortex separation and fate upon release to surface water untreated. Research Questions 1 and 2 are addressed in this chapter.

Chapter 3: Disinfection of Microbial Agents in Combined Sewer Overflows Using the Green Disinfectant Peracetic Acid

PAA disinfection kinetics for end-of-pipe treatment was investigated for ARGs and indicator organism marker genes in simulated CSO effluent. These disinfection studies used a novel application of vPCR to determine ARG concentrations in viable cells and to determine shifts in the viable-cell microbial community structure following disinfection.

Research Questions 2 and 3 are addressed in this chapter.

Chapter 4: Settling and peracetic acid disinfection for end-of-pipe treatment of sul1 gene-carrying indicator organisms 13

To complement the work performed in Chapters 2 and 3, the effectiveness of end-of-pipe treatment with removal of settleable particles and PAA disinfection was investigated with traditional cultivation-based methods. This work builds on the available literature by tracking the potential for the treatment train to select for sul1 gene carrying E. coli and evaluates the potential for indicator regrowth upon release to estuarine waters. Research

Question 2 is addressed in this chapter.

Chapter 5: Viability-based quantitation of antibiotic resistance and fecal marker genes in wastewater effluent and receiving waters

A field study was conducted to investigate the relative proportion of ARGs in cells with intact membranes to total ARGs observed in WWTP effluent and receiving waters towards understanding their availability for proliferation via different mechanisms. The study design allowed for comparison between seasons and location. Research Questions

2, 3 and 4 are addressed in this chapter.

Chapter 6: Fate of microbial agents in wastewater collection systems

A field study was conducted to characterize the factors that drive differences in the loading of microbial agents in sewers, including season, sewer type and chemical/quality parameters. Research Question 4 is addressed in this chapter.

Chapter 7: Conclusion

Appendix A. Shifts in microbial community structure and function in surface waters impacted by unconventional oil and gas waste water revealed by metagenomics 14

UOG impacted surface water was analyzed in streams adjacent to a disposal facility and at background locations to determine the impact on the microbial community structure and function. Whether this site served as a hot spot of antibiotic resistance due to the addition of biocides to fracturing fluids was also explored. This study provides supplementary information about the factors driving microbial community structure and

ARGs, though not contained to the urban environment. Research Question 5 is addressed in this appendix. 15

2. CHAPTER 2: Partitioning of antibiotic resistance genes and fecal indicators varies intra

and inter-storm during combined sewer overflows

Published in Frontiers in Microbiology 2017, 8, 2024, 1-13.

Authors: Alessia Eramo, Hannah Delos Reyes, and Nicole L. Fahrenfeld

2.1. Abstract

Combined sewer overflows (CSOs) degrade water quality through the release of

microbial contaminants in CSO effluent. Improved understanding of the partitioning of

microbial contaminants onto settleable particles can provide insight into their fate in end-

of-pipe treatment systems or following release during CSO events. Sampling was

performed across the hydrograph for three storm events as well as during baseflow and

wet weather in three surface waters impacted by CSO. qPCR was performed for select

antibiotic resistance genes (ARGs) and a marker gene for human fecal indicator

organisms (BacHum) using methods to differentiate the partitioning of microbial

contaminants on settleable particles versus suspended in the aqueous phase. Amplicon

sequencing was performed on both fractions of storm samples to further define the timing

and partitioning of microbial contaminants released during CSO events. Samples

collected at the CSO outfall exhibited microbial community signatures of wastewater at

select time points early or late in the storm events. CSOs were found to be a source of

ARGs. The majority of ARGs at the CSO outfall were observed on the attached fraction

of samples: 64-79% of sul1 and 59-88% of tet(G). However, the timing of peak ARGs

and human fecal indicator marker gene BacHum did not necessarily coincide with

observation of the microbial signature of wastewater in CSO effluent. Therefore, unit 16 processes that remove settleable particles (e.g., hydrodynamic separators) operated throughout a CSO event would achieve up to 0.5-0.9-log removal of ARGs and fecal indicators by removing the attached fraction of measured genes. Secondary treatment would be required if greater removal of these targets is needed.

2.2. Introduction

Outdated sewer infrastructure results in the release of 26 million cubic meters of untreated wastewater to New Jersey (NJ) surface waters each year (New Jersey Future).

Combined sewer overflows (CSOs) degrade water quality and present a risk to public health because they discharge priority pollutants (Gasperi et al. 2008), nutrients (Gervin and Brix 2001), pathogens, and fecal bacteria (Donovan et al. 2008) into waterways.

Incidental ingestion of CSO-plagued Passaic River water was associated with an elevated risk of gastrointestinal disease (Donovan et al. 2008). Solutions are needed to remove or inactivate a broad suite of microbial contaminants in CSO effluent.

End-of-pipe treatment of CSO effluent is a potential solution for improving water quality at a lower cost than upgrading combined sewer systems to separate sanitary systems, which is estimated to cost $40.8 billion in the US (United States Environmental

Protection Agency 2016). End-of-pipe treatment systems for CSO effluent range from screening for floatables to more advanced unit processes such as rapid filtration and disinfection. Hydrodynamic or vortex separators are end-of-pipe treatment technologies that target the removal of settleable particles primarily through extension the particle flow path, allowing a longer time for gravitational forces to remove settleable particles

(Alkhaddar et al. 2001; Andoh and Saul 2003). Therefore, understanding of the timing and partitioning of microbial contaminants in CSO effluent on settleable particles is 17 needed for designing and operating end-of-pipe treatment systems. Quantifying the partitioning of microbial contaminants can also aid in determining the fate of microbial contaminants released to surface waters during storm events given that microbes attached to settleable particles will have different fates than those on non-settleable particles and planktonic cells (Krometis et al. 2007). For CSO (Passerat et al. 2011) and storm events

(Krometis et al. 2007), intra-storm variability in the proportion of attached and suspended microbes has been demonstrated for a limited set of indicator organisms. Understanding the timing and potential for removal of indicator organisms is pertinent for meeting water quality regulations of the Clean Water Act. However, because correlations between indicator organisms and pathogens are often weak (Harwood et al. 2005) and different microbes associate with particles at different rates (Suter et al. 2011), targeting a broader range of microbial contaminants is of interest.

Antibiotic resistance genes (ARGs) are emerging microbial contaminants of concern in the environment given that environmental sources of antibiotic resistance have been linked to clinical infections (Forsberg et al. 2012). High levels of antibiotic resistant bacteria in the CSO- impacted Hudson River were associated with wet weather (Young et al. 2013). Determining the relative importance of different sources of antibiotic resistance in our waterways is key for developing mitigation strategies. Yet, little is known about the concentration, timing, and partitioning of antibiotic resistant bacteria and ARGs in CSO (Fahrenfeld and Bisceglia 2016; McLellan et al. 2007), limiting our understanding of the importance of CSO as a source of these contaminants and our ability to evaluate end-of-pipe treatment technologies. The association of ARGs with colloidal material has been documented in wastewater (Breazeal et al. 2013) and elevated levels of 18

ARGs and antibiotic resistant bacteria have been observed in biofilm in various water environments (Zhang et al. 2009), indicating that the partitioning of ARGs onto settleable particles in CSO may be an important control on their fate.

The aim of this work is to expand our understanding of the intra- and inter-storm variability in the partitioning of a broader suite of microbial contaminants onto settleable particles during CSO events. To achieve this goal, water samples were collected across three CSO events and settleable and suspended bacteria were separated using previously validated methods (Characklis et al. 2005; Krometis et al. 2007). Analysis was performed for a human fecal indicator marker gene (BacHum) using qPCR to determine the timing of the release of fecal microbes during the CSO events and to allow for comparison to other studies of the partitioning of fecal indicators during CSO and storm events. Amplicon sequencing was performed to broaden our understanding of the variation of microbial communities released throughout a CSO event. To understand the role of CSO as a source of ARGs, select ARGs found in wastewater encoding for sulfonamide and tetracycline resistance were analyzed using qPCR during the CSO events and compared to those observed in local waterways in different flow conditions

(i.e., baseflow and wet weather). Finally, the intra- and inter-storm variability of the partitioning of ARGs observed in the CSO was quantified to provide insight into the potential for treatment processes that remove settleable particles to remove these emerging contaminants. 19

2.3. Materials and Methods

2.3.1. Sampling and Water Quality Analysis

CSO effluent was collected during three storm events in the surface water outside of a CSO outfall in northern NJ (Fig. 2.1, Fig. 2.8). The outfall sampled has a drainage area that is 98% urban and 57% impervious. The outfall had 78 discharges in 2014 and

0.25cm of rainfall were sufficient to cause CSO at another outfall in this system for which the dates of overflows were available (Van Abs 2014). During Storm I (2.4cm,

8/11/15), duplicate samples (1L) were collected at two time points. For Storm II (2.5cm,

9/10/15) and Storm III (3.6cm, 10/28/15), singlet samples (1L) were collected every fifteen to thirty minutes, based on the forecasted duration of the storm. All samples for biomolecular analysis were stored on ice until processing in the laboratory. Conductivity and pH were measured in the field with a multimeter (Orion Star A329, Thermo

Scientific). Aliquots (20mL) of samples from each time point were analyzed for total suspended solids (TSS), using Environmental Sciences Section Method 340.2 (Wisconsin

State Lab of Hygiene 1993). Field blanks consisting of autoclaved deionized water left open for the duration of Storm II and Storm III sampling and otherwise treated as field samples were analyzed for QA/QC. Historical rainfall data was collected from Weather

Underground (The Weather Company LLC).

To separate settleable and suspended microbes, methods previously validated by

Krometis et al. (2007) were applied. The two Storm I and seven Storm II samples were analyzed directly. The ten field samples from Storm III were composited in the lab to generate five representative samples from across the precipitation event. Briefly, samples

(~0.7-1L) were centrifuged at 1160×g for 10min at 4°C. The supernatant (top 70% v/v) 20 was collected by pipette and will be referred to as “suspended.” The suspended fraction was previously determined to contain 80% of non-settleable particles in stormwater

(Characklis et al. 2005). The remaining fraction represents the settleable particles and will be referred to as “attached.” After separating these two fractions, each sample was filter concentrated (0.22µm, nitrocellulose) and stored at -20°C prior to DNA extraction.

To determine the microbial community structure indicative of untreated wastewater, post- screening grab samples of wastewater influent were collected (10/26/16) from a municipal wastewater treatment plant (WWTP) outside of the studied sewer systems, diluted with sterile DI water (23% wastewater to simulate CSO effluent), and either pelleted by centrifugation at 4000×g for 15 minutes or filter concentrated (0.22µm, nitrocellulose).

To compare the concentration of ARGs in CSO effluent to that in local waterways under different flow and tidal conditions, water (1L) and composited bed sediment

(~50mL) samples were collected in triplicate during baseflow (April 6 or 16, 2015) and during or following wet weather (June 1, 2015, Fig. 2.9) in three surface waters (Hudson

River, Raritan Bay, and Passaic River) (Fig. 2.1). The mean daily temperature for baseflow and wet weather sampling events was ±2°C. Nearby rainfall gauges reported cumulative rainfall of 1.8cm for wet weather sampling in the Passaic River and Raritan

Bay and 4.2cm for the wet weather sampling on the Hudson River, all of which are sufficient to potentially trigger CSO. The Passaic and Hudson Rivers are tidally influenced at the sampling locations. Both the baseflow and wet weather sampling occurred during low tide in the Passaic River. The Hudson River baseflow sampling occurred during high tide, while the wet weather sampling occurred during low tide. The 21

Raritan Bay baseflow sampling occurred during low tide and wet weather sampling occurred during high tide. Aqueous samples (~500-900mL) were concentrated on 0.22-

µm nitrocellulose filters (Millipore Corporation, Billerica, MA) prior to DNA extraction.

Sediment samples were homogenized prior to DNA extraction and analyzed as described in Section 2.2.

Figure 2.1 Map of sampling locations for wet weather sampling in three waterways impacted by CSOs, location of CSO outfall sampled across precipitation event, and documented CSO outfalls.

2.3.2. Biomolecular Analyses

DNA was extracted from filter concentrated samples, cell pellets (0.5mL), or sediment (~0.5g) using a commercial kit (FastDNA Spin Kit for Soil, MP Biomedicals) following the manufacturer’s directions. To determine the concentration of ARGs across the precipitation event, qPCR was performed on Storms I-III samples for select ARGs

[sul1, tet(G)], BacHum as a human fecal indicator (Kildare et al. 2007) and 16S rRNA 22 gene copies for total bacterial population (Zhang et al. 2016). Baseflow and wet weather samples were subjected to qPCR analysis for select ARGs [sul1 and sul2 (Pei et al.

2006), tet(G) and tet(O) (Aminov et al. 2002)] and 16S rRNA gene copies as a surrogate for total bacterial population (Suzuki et al. 2000). A standard SybrGreen (5µL SsoFast

EvaGreen, BioRad, Hercules, CA) chemistry with 0.4µM forward and reverse primers, and 1µL diluted (1:100) DNA extract in a 10µL reaction was used for all genes except

BacHum. Probe chemistry (5µL SsoAdvanced Universal Probes Supemix, Biorad,

Hercules, CA) with 0.22µM of each primer, 0.07µM of probe, and 1µL diluted (1:100)

DNA extract was used for BacHum in a 10µL reaction. QA/QC on the qPCR was performed as previously described (Fahrenfeld et al. 2014). Amplicon sequencing

(Illumina MiSeq, 300bp, paired end) was performed on samples of WWTP influent and from the three storms targeting the V3-4 region of the 16S rRNA gene at a commercial lab (MrDNA, Shallowater, TX). Sequences were analyzed using mothur MiSeq Standard

Operating Procedure (accessed 9/2016) (Kozich et al. 2013).

2.3.3. Statistical Analyses

To test for differences in wet weather versus baseflow conditions, a Wilcoxon rank sum test was applied to ARG copies normalized to the 16S rRNA gene copies. A paired

Student’s t-test (normality of data confirmed by a Shapiro test) or a Wilcoxon rank sum test was applied to test for differences in attached and suspended log-normalized ARG or

16S rRNA gene copy numbers intra-storm. A Kruskal-Wallis rank sum test followed by a post-hoc pairwise t-test was performed on log-normalized total gene copy numbers to test for inter-storm differences. These statistical tests were performed in R (R Core Team

2013). A Bray-Curtis similarity matrix was calculated on log-normalized subsampled 23

(N=14,052 sequences) operational taxonomic unit data at the class level followed by cluster analysis with a SIMPROF test and non-metric multidimensional scaling (nMDS) in PRIMER 7. Rarefaction was performed in mothur. To determine which taxa were preferentially associated with a given sample type (settleable storm, suspended storm, wastewater) biomarker analysis was performed on class-level relative abundance data for samples excluding the storm samples that formed significant clusters with wastewater.

The linear discriminant analysis effect size (LEfSe) tool (Segata et al. 2011) was used to identify biomarkers using the default settings.

2.4. Results

2.4.1. Partitioning of ARGs and Microbial Contaminants Across the Hydrograph

Sampling during three storm events was performed to determine concentration of

ARGs in CSO effluent, the proportion of attached compared to suspended ARGs, and the intra- and inter-storm variability in this partitioning. Attached sul1 gene copies accounted for 64±6% of the total sul1 gene copies for Storm I (Fig. 2.10), 79±6% of the total sul1 gene copies for Storm II (Fig. 2.2), and 71±18% of the total sul1 gene copies for Storm III (Fig. 2.3). Across all storms, attached sul1 gene copies were greater than suspended sul1 gene copy concentrations (p=5.1×10-5). For tet(G), attached gene copy concentrations were greater than suspended for the three storms (p=8.4×10-3) with the attached fraction accounting for 88±9% of total tet(G) gene copies during Storm I,

77±13% of total tet(G) gene copies for Storm II, and 59±52% of total tet(G) gene copies for Storm III (Fig. 2.4). tet(G) was not detected in all Storm III samples or in both phases when it was detected. Normalizing sul1 or tet(G) concentrations to 16S rRNA gene copy 24

numbers resulted in no difference in the partitioning between attached and suspended

phases across the storms (p=0.07-0.35). a.

b.

c.

d.

Figure 2.2 Storm II a. rainfall, total suspended solids (TSS), and conductivity, b. sul1, c. suspended bacteria by class, and d. attached bacteria by class 25

a.

b.

c.

d.

Figure 2.3 Storm III a. rainfall, total suspended solids (TSS), and conductivity, b. sul1 flux, c. suspended bacteria by class, and d. attached bacteria by class 26

a. attached suspended rainfall s e i 0.0007 1.0 p

o Storm I c 0.0006 e

n 0.8 e g

0.0005 )

A m c N 0.6 (

0.0004 R l l r

a f S n 6 0.0003 0.4 i 1 a / R s

e 0.0002 i p

o 0.2 c

0.0001 ) G ( t 0.0000 0.0 e t 05:00:00 07:00:00 09:00:00 11:00:00 b. Time (hr:min:s)

attached suspended

s rainfall e i 0.0005 1.0 p Storm II o c

e

n 0.0004 0.8 e g )

A m c N 0.0003 0.6 (

R l l r

a f S n 6 0.0002 0.4 i 1 a / R s e i p

o 0.0001 0.2 c

) G ( t 0.0000 0.0 e t 17:00:00 18:00:00 19:00:00 c. Time (hr:min:s)

attached suspended rainfall s e i 0.0030 1.0 p

o Storm III c

e 0.0025

n 0.8 e g ) 0.0020 A m c N 0.6 (

R l l r

0.0015 a f S n 6 0.4 i 1 a / R

s 0.0010 e i p

o 0.0005 0.2 c

) G ( t 0.0000 0.0 e t 10:00:00 11:00:00 12:00:00 13:00:00 14:00:00 15:00:00 Time (hr:min:s)

Figure 2.4 Attached and suspended tet(G) gene copies/16S rRNA gene copies and rainfall across (a) Storm I, (b) Storm II, and (c) Storm III. Storm I results represent the average of duplicate samples with a relative difference of ±1.6-20.5% (log 16S normalized data). 27

Samples collected during baseflow and during or immediately following rainfall events in local waterways were analyzed to provide a broader understanding of ARGs during wet weather events in CSO impacted waters. The total (attached plus suspended)16S rRNA gene copy normalized sul1 and tet(G) concentrations observed during the three storm events were within the range observed during the baseflow and wet weather sampling in the Passaic River, Hudson River, and Raritan Bay (Fig. 2.5).

Significantly higher 16S rRNA normalized sul1 gene copy concentrations were observed in wet weather in water column from Passaic (p=0.041, both samples collected during low tide) and Hudson Rivers (p=0.013, potentially confounded by tides), but not Raritan

Bay (p=1.00). Significant differences were not observed in the water column for tet(G), tet(O) or sul2 between baseflow and wet weather in water or sediment at any site (Fig.

2.11). 28

Figure 2.5 sul1, sul2, tet(G), and tet(O) ARG concentrations normalized to 16S rRNA gene copies observed outside of a CSO outfall during three storm events (H1_s), the Hudson at a separate sampling location (H2), the Passaic (P), and Raritan (R) during baseflow (b) and wet weather (w).

Attached human fecal indicator marker gene BacHum accounted for 46±8% of total BacHum gene copies in Storm I and 61±29% in Storm II (Fig. 2.6). Intra-storm variability of BacHum was evidenced by observation of BacHum in only the suspended fraction early in Storm III then only in the settleable fraction later in the storm. Inter- storm variability in partitioning was observed for 16S rRNA gene copies where the attached portion accounted for 56±0% of total 16S rRNA gene copies for Storm I,

57±21% of total for Storm II, and 70±9% of total for Storm III (Fig. 2.12). During Storm

II, no differences were observed in attached compared to suspended concentrations of the

16S rRNA gene (p=0.38). Attached 16S rRNA gene copies were greater than suspended

16S rRNA gene copies (p=0.014) for Storm III. 29

attached suspended

s rainfall e i

a. p 0.004 1.0 o c

Storm I e n

e 0.8 g

0.003 ) A N m c

R 0.6 ( r

l l S

0.002 a f 6 n 1 i / 0.4 a s e R i

p 0.001 o

c 0.2

m u H

c 0.000 0.0 a

B 05:00:00 07:00:00 09:00:00 11:00:00 Time (hr:min:s)

attached b. suspended

s rainfall e i

p 0.0020 1.0 o Storm II c

e n

e 0.8 g

0.0015 ) A N m c

R 0.6 ( r

l l S

0.0010 a f 6 n 1 i / 0.4 a s e R i

p 0.0005 o

c 0.2

m u H

c 0.0000 0.0 a

B 17:00:00 18:00:00 19:00:00 Time (hr:min:s)

c. attached suspended

s rainfall e i

p 0.0006 1.0 o c

Storm III e

n 0.0005 e 0.8 g

) A 0.0004 N m c

R 0.6 ( r

l l S

0.0003 a f 6 n 1 i / 0.4 a s e 0.0002 R i p o

c 0.2 0.0001 m u H

c 0.0000 0.0 a

B 10:00:00 11:00:00 12:00:00 13:00:00 14:00:00 15:00:00 Time (hr:min:s)

Figure 2.6 Attached and suspended BacHum gene copies/16S rRNA gene copies and rainfall across (a) Storm I, (b) Storm II, and (c) Storm III. Storm I results represent the average of duplicate samples with a relative difference of ±1.7-19% (log 16S normalized data). 30

2.4.2. Timing of Peak Microbial Agent Release

Intra- and inter-storm variability was observed in the concentration of ARGs and fecal indicator marker genes. During Storm I, ARGs and indicator gene concentrations peaked in the sample collected at the end of the storm. During Storm II, concentrations of sul1 and BacHum peaked in the first and final samples, while concentrations of tet(G) peaked mid-storm event. For Storm III, sul1 concentrations increased across the storm event while the highest BacHum concentrations were observed in the middle of the storm. Inter-storm variability was observed with sul1 gene copy numbers significantly lower in Storm III than Storms I and II (both p<0.03). No significant differences were observed between Storms I and II for sul1 gene copy numbers (p=1). Differences between tet(G) and BacHum copies were not observed between the storms (p=0.2-1.0).

Amplicon sequencing was performed to determine when during the storm events

CSO effluent samples resembled wastewater and therefore was likely to contain microbial contaminants from wastewater. The bacterial community structure for the attached and suspended fractions observed at the CSO outfall across the storms (Fig.

2.2c-d, 3c-d, A3b-c) and wastewater concentrated two ways (filter concentrating versus cell pelleting) were compared using nDMS analysis (Fig. 2.7). Four storm samples had

89.4% similarity (Cluster analysis with SIMPROF test p=1×10-3, Fig. 2.13) with the wastewater. Rarefaction curves are included as Fig. 2.14. The storm samples with community structures resembling wastewater were marked by increased relative abundances of Bacteroidia (14+/-2% in wastewater-like storm samples versus 0.2+/-0.3% in other storm samples), Clostridia (5.1+-/0.6 in wastewater-like samples versus 0.2+/- 31

0.2 in other storm samples) and Fusobacteria (1.9+/-1.0% in wastewater-like samples versus 0.1+/-0.1% in other storm samples). Both the first attached and suspended samples from Storm II had microbial community structures resembling the wastewater.

For Storm III, the final attached and the second to last suspended sample resembled wastewater. Water quality was determined by measuring conductivity and TSS during the storm events. In general, peak conductivity and TSS (Fig. 2.2-3, S2.3) did not occur when CSO effluent microbial communities resembled wastewater.

Of particular interest is the timing of the release of bacterial groups that contain indicator organisms and pathogens. The order Enterobacteriales contains fecal indicators

(E. coli), several waterborne pathogens (e.g., E. coli, Salmonella, Klebsiella, Shigella), and many other bacteria that fit neither of these classifications. By finding the product of the relative abundance of Enterobacteriales and the concentration of 16S rRNA gene copies, one may estimate the concentration of Enterobacteriales in a sample (Fig. 2.16).

For Storm III, samples with community structures similar to wastewater had 101.7-101.8 greater Enterobacteriales (both >1 standard deviation above the average

Enterobacteriales observed in the other storm samples). This observation was not preserved for Storm II samples with community structures similar to wastewater: the attached and suspended Enterobacteriales were similar to other storm samples (107.9 gene copies in wastewater vs. 107.2±0.4 in storm attached, 104.2 gene copies in wastewater vs.

106.2±1.4 in storm suspended).

To better understand the relationship between the attached and suspended microbial communities in CSO effluent the two fractions were compared. The paired attached and suspended microbial communities clustered together without significant differences for 32

4/14 samples collected across the three storms (Fig. 2.7, Fig. S2.6). LEfSe analysis revealed 28 discriminating features serve as biomarkers for wastewater compared to the attached and suspended stormwater samples (Fig. 2.15). The suspended stormwater community was marked by phyla or members of the phyla Alphaproteobacteria,

Actinobacteria, Parcubacteria, and unclassified bacteria. The attached stormwater community was marked by Acidobacteria, Bacteroidetes, Chloroflexi, and three other minor phyla.

2D Stress: 0.09 Similarity 1 4 85 Storm-Phase I_susp 2 I_att II_susp II_att 7 1 3 III_susp 1 7 III_att 5 4 2 ww-p 3 5 ww-f

4 6 2 3 1 2a2a 1a 1a 5 2 6 5 4 2b2b 1b 1b 3

Figure 2.7 Non-metric multidimensional scaling analysis of bacterial community structures for wastewater influent concentrated by centrifugation (ww-p) or filtering (ww- p), and suspended (x) and attached (triangles) samples collected across three storms (I purple, II green, and III blue). The numbers above the storm sample symbols represent the order of samples (1 being the first sample collected, etc.). The overlay represents clusters with 85% similarity. 33

2.5. Discussion

2.5.1. Dynamics of Attached and Suspended ARG Concentrations

Storm samples collected for this study demonstrated that CSO are a source of

ARGs and that most observed ARGs in CSO effluent were attached to settleable particles. Removing the attached fraction of ARGs in CSO effluent from this study would result in 0.5-0.9 log removal of sul1 and tet(G) gene copies. The high percentages of settleable particle-associated ARGs and fecal indicator marker genes are likely due to

ARGs and fecal marker genes present in biofilm and cells that preferentially grow attached to particles. When ARG concentrations were normalized to 16S rRNA gene copies, no significant differences were observed between attached and suspended sul1 or tet(G) concentrations indicating the higher total bacterial densities in the attached fraction could explain the higher ARG concentrations observed on the attached phase.

The percentage of ARGs in the attached fractions observed outside a CSO outfall were on the higher end of ranges reported for particle-associated indicator organisms in other studies: 30-60% of total coliform, fecal coliforms, E. coli and Enterococci partitioned to settleable particles in a stormwater impacted river (Characklis et al. 2005), 40-50% of bacterial indicators were attached to settleable particles in a stormwater impacted stream

(Krometis et al. 2007), and 77% of the E. coli were attached in CSO effluent (Passerat et al. 2011). When comparing between these studies the differences in methodology for defining the attached and suspended fractions of water samples and for quantifying microbial contaminants should be considered. This study used the methods of Characklis et al. (2005) and Krometis et al. (2007) to separate attached and suspended fractions.

Characklis et al. (2005) determined that 80% of suspended particles would remain in the 34 supernatant using this centrifugation procedure. Passerat et al. (2011) used a filtration method (attached >5 µm; 5 µm>suspended>0.2 µm) and reported higher fractions of attached indicator organisms than the studies using centrifugation-based methods.

Another important distinction between this work and previous studies differentiating attached and suspended microbes is that the present study quantified gene copies with qPCR in contrast the analysis of samples by cultivation-based methods. qPCR is generally more sensitive than cultivation-based techniques due in part to the fact that it does not discriminate between genes from viable and non-viable sources (i.e., dead cells, extracellular DNA). For BacHum, qPCR can overestimate the concentration of fecal indicators by detecting DNA from non-viable sources and therefore can also overestimate risk. Because transformation of extracellular DNA into competent cells is possible,

ARGs from non-viable sources detected by qPCR are still potentially of interest for risk assessment.

The timing of peak ARG concentrations is of interest for CSO interventions including end-of-pipe treatment. sul1 peak concentrations were observed at the end of

Storm I, the beginning and end of Storm II, and at the end of Storm III. While peak sul1 concentration was observed early and/or late in the storm events, the timing of peak tet(G) concentrations was variable, indicating that treating more than the first and final flush may be desirable. Given that the timing of peak ARG concentrations did not necessarily coincide with the observation of microbial community signatures indicative of wastewater microbial communities, there are likely other sources of ARGs in the storm events (e.g., stormwater). The urban area investigated likely has runoff containing feces from domestic and wild animals that could contribute to the indicator organism and ARG 35 loads in CSO effluent. Domestic animals are receiving increasing antibiotic prescriptions and may serve as a reservoir for antibiotic resistant bacteria (Guardabassi et al. 2004).

There are several potential differences between the storms sampled that may contribute to the inter-storm variability in the timing and concentration of ARGs in CSO effluent including differences in rainfall intensity, timing of the storm with respect to diurnal flow, and, in particular, in the amount of rainfall preceding the targeted storm, as discussed further in 4.2. Sampling of more storm events could reveal common patterns not observed here given that only three storm events were captured.

Sampling for total ARGs performed during baseflow and wet weather conditions in the Passaic River, Hudson River, and Raritan Bay provided local comparison for the three storm samplings. The total ARGs (attached plus suspended) concentrations observed in the storm samples were within the range observed during baseflow and wet weather conditions in local water ways. During wet weather ARG concentrations in surface water would be lowered by dilution, an observation noted during historic flooding in Colorado (Garner et al. 2016), if rain and stormwater have lower ARGs. However, wet weather introduces sources including CSO effluent, stormwater, and resuspension of settled ARG. These phenomena may explain the variability in the concentrations of

ARGs observed in the river and bay samples during wet weather, all of which are impacted by CSO, compared to the end-of-pipe ARGs observed in the storm samples.

Compared to the maximum concentrations of ARGs observed in wastewater influent

(data not shown), ARG concentrations observed at the CSO outfall represent <1–40% of the sul1 and<1–5% of the tet(G) observed in wastewater influent. Dilution of ARGs may explain why the lowest sul1 concentrations were observed in Storm III, which had the 36 greatest cumulative rainfall. Concentrations of wastewater micropollutants such as hormones were found to decrease with increasing flow because of such dilution (Phillips et al. 2012).

2.5.2. Timing and Partitioning of Wastewater Signatures and Fecal Indicators in

CSO Effluent

Microbial community structures with high similarity to wastewater were observed at select time points early or late in the storm event outside of the CSO outfall during wet weather. This observation is consistent with other studies that reported peak concentrations of indicator organisms of fecal pollution early and late in the hydrograph for CSO effluent, CSO-impacted surface waters (Passerat et al. 2011; Puerta et al. 2002), and stormwater runoff (Krometis et al. 2007; Puerta et al. 2002). During CSO this first flush phenomena has been attributed to the increased shear stress sewer solids experience at the higher velocities occurring during wet weather events, which scour stationary bed solids that accumulated in combined sewers pipes during dry periods (Hvitved-Jacobsen et al. 2013). Sewer solids have been linked to the indicator organism loads in CSO effluent (Passerat et al. 2011). For Storm II, the microbial community signature of wastewater was observed at the same time for attached and suspended samples. Given that precipitation occurred earlier in the day before this storm (Fig. 2.8) the coincidence of the release of attached and suspended microbes resembling wastewater may be because settleable particles were recently deposited and therefore loose. A microbial community structure with a wastewater signature was also observed in the fourth suspended and fifth (final) attached sample from Storm III. The timing difference may be due to velocities below the critical velocity during the lull in rainfall, allowing for 37 flushing of suspended bacteria and for settling during lower flow followed by resuspension of solids when the rainfall resumed. More robust sampling of wastewater influent and sampling of sewer solids may have indicated a wastewater association with more samples and is recommended for future studies.

The timing of peak BacHum observations did not coincide with the microbial signatures of wastewater. Partitioning of BacHum to settleable particles was variable across the storms. BacHum is a highly sensitive but moderately specific human fecal indicator gene exhibiting some cross amplification with feces from deer, dogs, geese, and gulls (Ahmed et al. 2009; Boehm et al. 2013), which can be expected to be present in stormwater. Thus, the BacHum peak concentrations observed mid-storm may be representative of such cross amplification. The attached and suspended

Enterobacteriales were determined given that this group includes indicator organisms such as E. coli and select waterborne pathogens such as Salmonella, and Shigella.

Enterobacteriales concentrations were relatively consistent across the storms except for during Storm III where their concentrations peaked at times when the wastewater signatures were observed. The weak correspondence with wastewater signatures is likely because there are many commensal organisms in the Enterobacteriales order. Given the limitations of the short reads generated by amplicon sequencing, sequences were not identified below this taxonomic level.

Understanding the differences between the attached and suspended microbial communities is of interest for understanding potential removal of groups containing pathogens and for understanding the impact on surface water microbial ecology. The attached and suspended microbial communities observed in CSO effluent clustered 38 together without significant differences for 4/14 samples collected across the three storms

(Fig. 2.7, Fig. 2.13). LEfSe analysis revealed 28 discriminating features serve as biomarkers for wastewater and CSO resembling wastewater compared to the attached and suspended stormwater microbial communities (Fig. 2.15). The ability to identify biomarkers of attached compared to suspended stormwater microbial communities reinforces observations that attached to particles varies for different microbes (Suter et al.

2011). LEfSe analysis indicated that the phyla Firmicutes and Epsilonproteobacteria, which includes the order Enterobacteriales, were biomarkers of wastewater. Other researchers also observed genera belonging to Epsilonproteobacteria to be sewage biomarkers (McLellan et al. 2015). CSO events were also associated with increases in

Epsilonproteobacteria and Clostridia, as well as Gammaproteobacteria, Bacteriodia, and

Bacteriodetes in the receiving water (Newton et al. 2013). Differences in the sewage signatures observed between these studies are likely due to a combination of differences in targeted variable region known to impact amplicon sequencing results (Tremblay et al.

2015) and geographical and population differences which can alter the sewage microbiome (Newton et al. 2015). The storm samples that did not have wastewater signatures had higher relative abundances of Alphaproteobacteria and Actinobacteria compared to the storm samples clustering with wastewater. Other researchers looking for biomarkers of stormwater also observed elevated relative abundances of genera in the class Actinobacteria along with Gammaproteobacteria and Betaproteobacteria

(McLellan et al. 2015), with the same caveats as above on targeted variable region. 39

2.5.3. Implications for Design and Operation of Hydrodynamic Separators

Treatment by hydrodynamic separation removes settleable materials by gravity and extension of the particle flow path (Alkhaddar et al. 2001). Therefore, one cannot directly compare the centrifugal forces used in this study to separate settleable particles to the centrifugal forces in hydrodynamic separators, because that is a minor mechanism for particle removal. However, one can optimize these systems for removal of the settleable particles. Wilson et al. (2009) determined that Peclet number (a function of settling velocity, hydrodynamic separator dimensions, and flow rate) could be used to predict removal efficiency for hydrodynamic separator treatment devices. Particle settling velocity was determined using previously reported characteristics of the attached and suspended fractions separated by the centrifugation technique applied here (Krometis et al. 2007) (Table 2.1). As an example, using these settling velocities and example flow and dimensions provided by Wilson et al. (2009) for a select unit, one would estimate

0.3-35% removal of suspended particles. However, units can be sized to improve these removal efficiencies. Given that peak concentrations of ARGs and fecal indicators were observed at various times during the storm events, removing settleable particles throughout the storm rather than targeting first or final flush may be needed. Even with

100% removal of ARGs on settleable particles, further treatment may be desirable. Often disinfection is included post-separation [e.g., (Patoczka et al. 2016)] and disinfection has been shown to inactivate microbes carrying ARGs (Dodd 2012) and UV disinfection can help destroy ARGs (McKinney and Pruden 2012). 40

2.6. Conclusions

This study highlights the presence, partitioning, and intra- and inter-storm variability of ARGs, indicator marker genes, and microbial communities at a CSO outfall during wet weather. CSO were found to be a source of ARGs. Generally, ARGs were detected in higher concentrations in the attached compared to the suspended fraction of CSO effluent while BacHum was more variably associated with the different fractions. Higher total bacterial concentrations on the settleable particles could explain the difference between attached and suspended ARG concentrations. A microbial community structure with a wastewater signature was observed at select time points at the beginning or end of the storm but did not necessarily coincide with peak ARGs and BacHum concentrations indicating that treatment of more than the first and final flush may be desirable. Results of this study indicate further treatment would be needed after separation of settleable particles for ARGs given that ARGs would remain in the suspended fraction of water during CSO events.

Acknowledgements

This work was supported by the New Jersey Water Resources Research Institute, the

National Science Foundation (Award Number 1510461), a Mark B. Bain fellowship from

Hudson River Foundation to AE, and a Rutgers Aresty/Douglass Project research fellowship to HD. Sampling assistance was provided by Shirin Estahbanati. Thanks to the utility partner outside of the study area for providing access to WWTP influent. 41

2.7. Supplemental Information

10 30 2.0 a. Passaic at Kearny, NJ Rainfall b.Passaic at Kearny, NJ data from Newark, NJ Rainfall data from Newark, NJ 25 ) Temperature

8 C Sampling ( Sampling 1.5 e ) ) r

20 u m m t c c

6 a ( ( r

l l e l l

a 15 p a 1.0 f f m n n i i

4 e a a t

R 10 R n a 0.5 2 e 5 M

0 0 0.0 0 0 0 0 0 0 0 5 5 5 5 5 :0 :0 :0 :0 :0 :0 :0 /1 /1 /1 /1 /1 0 0 0 0 0 0 0 /6 0 /4 8 /1 :0 :0 :0 :0 :0 :0 :0 4 /2 5 /1 6 0 4 8 2 6 0 0 4 5 0 0 0 1 1 2 0

Date Time (hr:min:s)

10 30 2.0 Hudson at Jersey City Rainfall Hudson at Jersey City Rainfall data from New York, NY 25 ) Temperature data from New York, NY Sampling 8 C

c. ( d. Sampling 1.5 e ) ) r

20 u m m t c c 6 a ( ( r

l l l e l

a 15 p a 1.0 f f n m n i i

4 e a t a

R 10 R n a

e 0.5 2 5 M

0 0 0.0 0 0 0 0 0 0 0 5 5 5 5 5 :0 :0 :0 :0 :0 :0 :0 /1 /1 /1 /1 /1 0 0 0 0 0 0 0 /6 0 /4 8 /1 :0 :0 :0 :0 :0 :0 :0 4 /2 5 /1 6 0 4 8 2 6 0 0 4 5 0 0 0 1 1 2 0

Date Time (hr:min:s)

10 30 2.0 Raritan at Perth Amboy, NJ Rainfall Raritan at Perth Amboy Rainfall data from Newark, NJ 25 ) Temperature data from

8 C ( Sampling Sampling Linden, NJ e

) 1.5 r

e. ) f.

20 u m t c m

6 a ( r c

l ( l e

l l a 15 p f

a 1.0 f n m i n 4 e i a t

a

R 10 n R a 2 e 0.5 5 M

0 0 0.0

5 5 5 5 5 0 0 0 0 0 0 0 /1 /1 /1 /1 /1 :0 :0 :0 :0 :0 :0 :0 /6 0 /4 8 /1 0 0 0 0 0 0 0 4 /2 5 /1 6 :0 :0 :0 :0 :0 :0 :0 4 5 0 4 8 2 6 0 0 0 0 0 1 1 2 0

Date Time (hr:min:s)

Figure 2.8 Precipitation and mean daily air temperature data for nearest gage for the (a) daily for the Passaic (data from Newark, NJ), (b) hourly for Passaic during wet weather event (data from Newark, NJ), (c) daily for the Hudson at Jersey City (rainfall data from New York, NY), (d) hourly for the Hudson during wet weather event (data from New York, NY), (e) daily for Raritan Bay (data from Newark, NJ, and (f) hourly for wet weather event (data from Linden, NJ). 42

a. 1.0 1.0 b. Rainfall Rainfall ) 0.8 Storm I sampling ) Storm II Sampling

m 0.8 m

c High tide

c High tide ( (

l l l l a

0.6 a f 0.6 f n n i i a a r r

y 0.4

y 0.4 l l r r u u o o

H 0.2 H 0.2

0.0 0.0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 :0 :0 :0 :0 :0 :0 :0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 : : : : : : : 0 0 0 0 0 0 0 0 0 0 0 0 0 0 : : : : : : : :0 :0 :0 :0 :0 :0 :0 0 4 8 2 6 0 0 0 4 8 2 6 0 0 1 1 2 1 1 2 c. Time (hr:min:s) d. Time (hr:min:s)

1.0 ) 6

Rainfall m c Rainfall (

)

Storm III sampling l 0.8 l 5 Storm sampling m a

High tide f c n ( i

l 4 l a r a 0.6 f e n i v

i 3 a t r a

l

y 0.4 l u r 2 m u u o c H 0.2

y 1 l i a

0.0 D 0 0 0 0 0 0 0 0 :0 :0 :0 :0 :0 :0 :0 5 5 5 5 5 5 5 0 0 0 0 0 0 0 /1 /1 /1 /1 /1 /1 /1 :0 :0 :0 :0 :0 :0 :0 /3 7 1 4 8 2 6 0 4 8 2 6 0 0 8 /1 /3 /1 /2 /1 /2 1 1 2 8 8 9 9 0 0 1 1 Time (hr:min:s) Date (month/day/year)

Figure 2.9 Hourly precipitation, sampling time, and time of high tide for the day of (a) Storm I, (b), Storm II, and (c) Storm III. (d) Daily cumulative precipitation across the study period. Precipitation data presented for Newark, NJ, the location of the nearest rain gauge. 43 a.

b.

c.

Figure 2.10 Storm I a. ARG flux, b. suspended bacteria by class and rainfall, c. attached bacteria by class. Sequencing results shown are for replicate samples collected at the same time. 44

attached suspended rainfall s e i 0.0007 1.0 p

a. o Storm I c 0.0006 e

n 0.8 e g

0.0005 )

A m c N 0.6 (

0.0004 R l l r

a f S n 6 0.0003 0.4 i 1 a / R s

e 0.0002 i p

o 0.2 c

0.0001 ) G ( t 0.0000 0.0 e t 05:00:00 07:00:00 09:00:00 11:00:00 Time (hr:min:s)

attached b. suspended

s rainfall e i 0.0005 1.0 p Storm II o c

e

n 0.0004 0.8 e g )

A m c N 0.0003 0.6 (

R l l r

a f S n 6 0.0002 0.4 i 1 a / R s e i p

o 0.0001 0.2 c

) G ( t 0.0000 0.0 e t 17:00:00 18:00:00 19:00:00 Time (hr:min:s)

c. attached suspended rainfall s e i 0.0030 1.0 p

o Storm III c

e 0.0025

n 0.8 e g ) 0.0020 A m c N 0.6 (

R l l r

0.0015 a f S n 6 0.4 i 1 a / R

s 0.0010 e i p

o 0.0005 0.2 c

) G ( t 0.0000 0.0 e t 10:00:00 11:00:00 12:00:00 13:00:00 14:00:00 15:00:00 Time (hr:min:s)

Figure 2.11 Attached and suspended tet(G) gene copies/16S rRNA gene copies and rainfall across (a) Storm I, (b) Storm II, and (c) Storm III. 45

attached suspended

s rainfall e i a. p 0.004 1.0 o c

Storm I e n

e 0.8 g

0.003 ) A N m c

R 0.6 ( r

l l S

0.002 a f 6 n 1 i / 0.4 a s e R i

p 0.001 o

c 0.2

m u H

c 0.000 0.0 a

B 05:00:00 07:00:00 09:00:00 11:00:00 Time (hr:min:s) b. attached suspended

s rainfall e i

p 0.0020 1.0 o Storm II c

e n

e 0.8 g

0.0015 ) A N m c

R 0.6 ( r

l l S

0.0010 a f 6 n 1 i / 0.4 a s e R i

p 0.0005 o

c 0.2

m u H

c 0.0000 0.0 a

B 17:00:00 18:00:00 19:00:00 c. Time (hr:min:s) attached suspended

s rainfall e i

p 0.0006 1.0 o c

Storm III e

n 0.0005 e 0.8 g

) A 0.0004 N m c

R 0.6 ( r

l l S

0.0003 a f 6 n 1 i / 0.4 a s e 0.0002 R i p o

c 0.2 0.0001 m u H

c 0.0000 0.0 a

B 10:00:00 11:00:00 12:00:00 13:00:00 14:00:00 15:00:00 Time (hr:min:s)

Figure 2.12 Attached and suspended BacHum gene copies/16S rRNA gene copies and rainfall across (a) Storm I, (b) Storm II, and (c) Storm III. 46

80 Storm-Phase I_susp I_att II_susp II_att 85 III_susp III_att ww-p ww-f y t i r a l

i 90 m i S

95

100 4 5 1 1 7 2 3 5 2 3 7 2 1 4 3 4 6 a a 5 a a 6 b b 5 b 4 b 3 1 2 2 1 1 2 1 2 1 2 Samples

Figure 2.13 Cluster analysis for bacterial communities at the class level. Numbers refer to sample number (1 for first, etc.) for a given storm, “a” and “b” refer to replicates. Red branches connect samples with bacterial community structures that were not different via SIMPROF test. a. b.

c. d.

Figure 2.14 Rarefaction curves for (a) Storm I, (b) Storm II, (c) Storm III, and (d) simulated CSO. Numbers for storm samples indicate sample order. Attached samples are indicated by “att” and suspended samples are indicated by “plank.” Storm replicates are labeled as “a” and “b.” CSO samples were diluted wastewater (WW) concentrated by filtering or by centrifugation (“pellet”). 47

Figure 2.15 LEfSe analysis on wastewater, attached storm, and suspended storm (excluding storm samples clustering with wastewater) indicates biomarkers for each matrix at the Phylum and Class level. 48

a.

b.

c.

Figure 2.16 Attached and suspended calculated Enterobacteriales and rainfall across (a) Storm I, (b) Storm II, and (c) Storm III. 49

Figure 2.17 Concentrations of ARGs normalized to 16S rRNA from composited sediment samples collected during base flow (dry) or wet weather conditions in the Passaic River (P) and Raritan Bay (R).

Table 2.1 Example removal efficiency for minimum and maximum surrogate particles representing the attached and suspended fractions using an example hydrodynamic separator unit (i.e., Stromceptor STC 4800 with 3.7m diameter, 3.4m settling depth, and flow rate 0.051 m3/s). Hydrodynamic separators can be sized during design to improve removal. For these estimates the following constants were used: kinematic viscosity water ν =1.12×10-6 m2/s, gravity 9.81m/s2, and water density ρ=999kg/m3.

Variables (units) Values Reference

Fraction attached, attached, suspended, suspended, (Krometis minimum maximum minimum maximum et al. Diameter, D (µM) 5 60 5 40 2007)

3 Particle density, ρs (kg/m ) 2.65E+03 2.65E+03 1.05E+03 1.05E+03

Settling velocity, Vs (m/s) 1.50E-05 2.09E-03 4.65E-07 2.97E-05 (Cheng 1997)

Peclet number, P (unitless) 3.7E-03 5.2E-01 1.1E-04 7.3E-03 Eq. 3 in (Wilson et al. 2009) 50

Removal efficiency, ɳ (%) 0.26 35 0.01 0.51 Eq. 7 in (Wilson et al. 2009) 51

3. CHAPTER 3: Disinfection of microbial agents in combined sewer overflows using the

green disinfectant peracetic acid

Published in Science Water Research and Technology 2017, 3, 1061-1072.

Authors: Alessia Eramo, William R. Morales Medina, and Nicole L. Fahrenfeld

3.1. Abstract

Combined sewer overflows (CSOs) degrade water quality and end-of-pipe treatment

is one potential solution for retrofitting this outdated infrastructure. The goal of this

research was to evaluate peracetic acid (PAA) as a disinfectant for CSOs using viability based

molecular methods for antibiotic resistance genes (ARGs), indicator organism marker gene

BacHum, and 16S rRNA genes. Simulated CSO effluent was prepared using 23-40% wastewater,

representing the higher end of the range of wastewater concentrations reported in CSO effluent.

Treatment of simulated CSO effluent (23% wastewater) with 100 mg∙min/L PAA (5 mg/L

PAA, 20 min) was needed to reduce viable cell sul1, tet(G), and BacHum (1.0±0.63-

3.2±0.25-log) while 25 to 50 mg•min/L PAA (5 mg/L PAA, 5-10 min) was needed to

reduce viable cell loads (0.62±0.56-1.6±0.08-log) in 40% wastewater from a different

municipal treatment plant. Increasing contact time after the initial decrease in viable cell

gene copies did not significantly improve treatment. A much greater applied Ct of 1200

mg∙min/L PAA (20 mg/L PAA, 60 min) was required for significant log reduction of 16S

rRNA genes (3.29±0.13-log). No significant losses of mexB were observed during the

study. Data were fitted to a Chick-Watson model and resulting inactivation constants for

sul1 and tet(G) > BacHum > 16S rRNA. Amplicon sequencing of the 16S rRNA gene

indicated the initial viable and total microbial communities were distinct and that 52 treatment with PAA resulted in marked increases the relative abundance of select phyla, particularly Clostridia which increased 1-1.5 orders of magnitude. Results confirm that membrane disruption is a mechanism for PAA disinfection and further treatment is needed to reduce total ARGs in CSO effluent.

3.2. Introduction

Combined sewer overflows (CSOs) degrade water quality and threaten human health by releasing viable fecal indicator bacteria, (Passerat et al. 2011) pathogen markers,

(Donovan et al. 2008) antibiotic resistant bacteria (ARB), (Young et al. 2013) and antibiotic resistance genes (ARGs) (Eramo et al. 2017a) into surface water bodies.

Combined sewer systems are designed to collect both storm water runoff and wastewater

(WW) and a portion of the untreated waste stream overflows into adjacent surface waters during heavy rainfall or snow melt when the flow exceeds plant capacity. The cost of upgrading combined sewer systems has been estimated at $40.8 billion for the

US.(United States Environmental Protection Agency 2016) Therefore, end-of-pipe treatment technologies for CSO may be attractive as lower cost alternatives to upgrading sewer infrastructure, with a variety of treatment trains under consideration.(Chhetri et al.

2016; Patoczka et al. 2016) Disinfection with peracetic acid (PAA) has been proposed for end-of-pipe treatment and is appealing because it has not been found to create toxic, mutagenic, or chlorinated by-products (Kitis 2004) reducing the need for pre-treatment to remove the high load of organic matter present in CSO (Passerat et al. 2011).

During CSO events water quality varies across the hydrograph with WW accounting for 4-39% of the flow and runoff accounting for the remainder (Passerat et al. 2011).

Effective disinfection (3.4-5.6-log removal) of simulated CSO effluent (5% WW) was 53 reported for E. coli with 5 mg/L PAA treatment for 5 minutes (Chhetri et al. 2014). More researchers have reported disinfection of total coliforms as measured by cultivation based techniques in WW influent indicating 20 mg/L for 10 min may be optimal for reducing target organisms (Sanchez-Ruiz et al. 1995). Given the cost of PAA, (Kitis 2004) PAA demand of organic matter in CSO effluent (González et al. 2012) and small footprint available for end-of-pipe treatment, understanding the impact of lower doses and shorter contact times for a range of CSO effluents is needed to provide economical recommendations for application.

CSOs are a source of ARGs (Eramo et al. 2017a) and elevated levels of ARB were observed in CSO impacted surface waters during wet weather (Young et al. 2013). In urban waters without significant agricultural impacts, WW is the dominant source of

ARGs during baseflow conditions (Storteboom et al. 2010). The ARG concentrations observed in CSO effluent were 2.5-100 times lower than observed for WWTP effluent

(Eramo et al. 2017a). However, untreated CSO effluent contributes viable ARB (Young et al. 2013) as opposed to the majority of the ARGs present in WWTP effluent that are extracellular DNA remaining after disinfection (Mao et al. 2014). ARGs in environmental matrices present a risk to human health (Ashbolt et al. 2013). The effect of PAA on antibiotic resistant organisms and ARGs has been investigated for secondary

WW but not CSO effluent. CSO effluent is unique in that it may receive no-pretreatment or treatment with different unit processes (e.g., hydrodynamic separation or rapid filtration) prior to disinfection as opposed to the settling and biological treatment that secondary WW effluent receives prior to disinfection. Two recent studies demonstrated that PAA is not effective at reducing the total concentrations of tet(A), blaTEM, qnrS, 54 ermB, sul1, and sul2 genes encoding resistance to tetracycline, beta-lactams, quinolone, erythromycin and two sulfonamide genes respectively, in secondary WW effluents.(Di

Cesare et al. 2016; Luprano et al. 2016) Flow cytometry analysis with live-dead staining demonstrated damage to cell walls indicating that a portion of detected ARGs were extracellular or present in nonviable cells (Di Cesare et al. 2016). Little data is available reporting disinfection kinetics for ARG-carrying cells in WW, (Biswal et al. 2014) and none is available for CSO effluent. Understanding the partitioning of ARGs between viable and non-viable sources is critical for understanding the mechanisms and therefore the risk of ARG proliferation in the environment. That is, defining the partitioning of

ARGs helps differentiate the risk from growth and horizontal gene transfer of ARGs in viable cells versus transformation (uptake of extracellular DNA) for extracellular ARGs and ARGs from non-viable cells. Viability based molecular techniques [i.e., propidium monoazide (PMA) qPCR also known as viability PCR (vPCR)] have been used for other disinfectants to demonstrate disinfection kinetics for indicator organism marker genes.(Varma et al. 2009) Evaluating disinfection kinetics for PAA treatment of ARG- carrying cells and shifts in microbial community following disinfection could further understanding of the risk associated with PAA treated CSO effluent.

The goal of this study was to determine PAA disinfection kinetics for end-of-pipe treatment by monitoring a broader range of microbial contaminants (ARGs and indicator organism marker genes) and using simulated CSO effluent representing higher percentages of WW (and therefore, risk) than have been previously investigated. The ability of PAA disinfection to interfere with the membranes of ARG-carrying bacteria and human fecal indicator organisms of the Bacteroidales order and to destroy ARGs in 55

CSO were investigated. Bacterial community analysis was performed to determine the impact of PAA on the “viable” microbial community structure. To differentiate between viable and total gene copies and better understand the mechanisms of PAA, viability- based qPCR targeting genes from cells only with sufficiently intact membranes was applied (Nocker et al. 2007). Data were modeled using Chick-Watson kinetic model to determine inactivation coefficients to help inform end-of-pipe disinfection. Results can be used to better understand the risk posed by ARGs in PAA disinfected CSO effluent.

3.3. Materials and Methods

3.3.1. Disinfection Experiment

Grab samples of municipal WWTP influent from two different utilities were collected during baseflow conditions October 26, 2015 (10:30 AM, WWTPa) and November 16,

2016 (10:30 AM, WWTPb) and stored on ice during transport then at 4°C until the start of the experiment. Total suspended solids (TSS) in influent measured according to

Environmental Sciences Section (ESS) Method 340.2(Wisconsin State Lab of Hygiene

1993) from WWTPa (228 ±109 mg/L) and WWTPb (63 ±31 mg/L) were within normally reported ranges for WW influent (Karvelas et al. 2003). Both WWTPa and WWTPb collect WW in separate sanitary sewer systems primarily from households with no major hospital or industrial inputs. Simulated CSO effluent was prepared by diluting municipal

WWTP influent with sterile deionized water (23 or 40% WWTP influent, v/v). Dilution with sterile DI water was chosen given that the microbial community structure was similar in WW influent diluted with sterile DI water and water collected outside a CSO outfall during wet weather (Eramo et al. 2017a). These dilutions were selected to represent the higher range for percentages of WW observed during CSO events,(Chhetri 56 et al. 2014; Passerat et al. 2011) but would not address microbial loads in runoff, generally from urban areas that makes up the remaining portion of the CSO effluent.

PAA (32% wt in acetic acid, Sigma Aldrich) was diluted to working stock solutions immediately prior to the disinfection experiments and concentrations were confirmed by colorimetric methods, described below. PAA is commercially available in the form of a quaternary equilibrium mixture containing acetic acid, hydrogen peroxide, PAA and water.(Kitis 2004)

In the first experiment, 23% WWTPa samples were treated with PAA (0, 5, and 20 mg/L) during rapid stirring on a stir plate in triplicate. Subsamples (170 mL) were collected at 0, 5, 10, 20, and 60 min. In the second experiment, 40% WWTPb samples were treated with 5 mg/L PAA while stirred, as in first experiment in triplicate and subsamples (250 mL) were collected at 0, 5 and 10 min, based on results from the first experiment indicating that incubations longer than 20 min often did not result in significant changes. Reactor volumes of 400 mL were used. Larger subsamples were collected in the second experiment to improve qPCR detection limits. The disinfection reaction in the subsamples was quenched with sodium thiosulfate (100 mg/L) and catalase (50 mg/L). For the second experiment, chemical oxygen demand (COD) was analyzed according to Hach Method 8000 with Hach COD vials (20-1500 mg/L range) and a DR2700 spectrophotometer (Hach, Loveland, CO). Aliquots (80mL) were analyzed for TSS. Conductivity and pH were measured with a calibrated multimeter

(Orion Star A329, Thermo Scientific) after quenching the reactions. Samples were analyzed using PAA test paper for 0-160 ppm (MicroEssential, Brooklyn, NY, reported precision +/- 10%) to evaluate PAA concentration before and after quenching. 57

3.3.2. PAA Decay

To estimate PAA decay in simulated CSO and the impact of COD, a third experiment was performed on grab samples of influent from WWTPa and b (collected during baseflow conditions on 7/27/2017 and 7/14/2017, respectively), in triplicate.

Samples from WWTPa (11.5 and 23% dilutions) were treated with PAA (5 or 20 mg/L for 0, 10, 20, and 60 min) and WWTPb (20 and 40% dilutions) were treated with PAA (5 mg/L for 0, 5, and 10 min). TSS, pH, and COD were measured prior to treatment, as described above. Samples were analyzed for PAA using a commercial kit (Peracetic

Acid Vacu-Vials, CHEMetrics, Midland, VA) at each sampling point and after quenching. PAA concentration was determined by measuring absorbance at 515 nm and using the calibration provided by the manufacturer to calculate PAA in mg/L.

3.3.3. Viability Analysis

Cells were recovered from the batch disinfection experiments by centrifugation at

4000×g for 15 min and reserving the bottom ~10 mL. Aliquots of the centrifuge concentrated cells (500 µL) were either treated with 50 µM propidium monoazide (PMA) or preserved for DNA extraction. PMA is a dye that inhibits PCR of DNA originating from non-viable cells and extracellular DNA by intercalation with double stranded nucleic acids. It has been reported that the photo-induced cross-linkage renders the DNA insoluble and results in its loss during DNA extraction.(Nocker et al. 2006) Samples treated with PMA allowed for quantification of genes from cells with intact membranes

(“viable cells”) only. Samples without PMA treatment allowed for quantification of total genes from cells with intact or compromised membranes along with extracellular DNA

(viable and nonviable). PMA treated samples were incubated at room temperature in the 58 dark for five minutes and then exposed in a PMA-Lite™ LED Photolysis Device

(Biotium, Hayward, CA) for 15 min to facilitate cross linking of the dye prior to storage.

Samples were stored at -20°C until DNA extraction. PMA methods were adapted from

Nocker et al. (Nocker et al. 2010; Nocker et al. 2007).

3.3.4. Cultivation

Heterotrophs were cultivated from subsamples from the second experiment to confirm the effectiveness of PAA treatments on reducing the concentration of viable cells. Aliquots of quenched triplicate subsamples were serially diluted (10-2-10-6, v:v), plated on LB Agar and incubated at 37oC until colony forming units (CFUs) were visible

(18 hrs). LB is a rich media that is used for cultivating E. coli and other heterotrophs.

Results were reported as CFU/mL.

3.3.5. Molecular Methods

DNA was extracted from centrifuge concentrated cells with and without PMA treatment (500 µL) using a Fast DNA Spin Kit for Soil (MP Biomedicals, Solon, OH).

DNA was diluted 1:50-1:100 to reduce inhibition and qPCR was performed to quantify the ARGs sul1,(Pei et al. 2006) tet(G), (Aminov et al. 2002) and mexB, (Yoneda et al.

2005) fecal indicator marker gene BacHum for Gram-negative Bacteroidales, (Kildare et al. 2007) and 16S rRNA gene copies (Muyzer et al. 1993). sul1 encodes for a dihydropteroate synthase of Gram-negative bacteria for resistance to sulfonamides. tet(G) is an efflux protein in Gram-negative bacteria for tetracycline resistance found on plasmid and chromosome. mexB is a subunit of an efflux pump conferring antibiotic resistance typically associated with Gram- negative Pseudomonas aeruginosa,(Sun et al.

2014) which has been expressed in Gram-positive organisms as well (Welch et al. 2010). 59

These ARGs were selected because they are commonly observed in wastewater and represent qPCR targets of different amplicon lengths. qPCR reaction mixtures for the analysis of sul1, tet(G), mexB and 16S rRNA gene copies consisted of 5 µL

SsoFastSuperMix (BioRad, Hercules, CA), 0.4µM forward and reverse primers, 2.4 µL molecular biology grade water, and 1µL diluted DNA extract. The qPCR reaction mixture for BacHum consisted of 5 µL SsoFastProbes SuperMix (BioRad, Hercules,

CA), 0.07 µM probe, 0.22 µM forward and reverse primers, 1 µL molecular biology grade water, and 1 µL diluted DNA extract. Thermocycler (BioRad CFX96 Touch,

Hercules, CA) conditions are summarized in Table 3.3. Samples were analyzed in triplicate with a seven-point standard curve (102-108 gene copies) and a no-template control with each qPCR plate. Melt curve analysis was performed for all ARGs and 16S rRNA gene analyses and the length of select PCR products was checked via agarose gel electrophoresis for all gene targets. The quantification limits for targeted genes were below 104 copies/mL. qPCR was also performed for ARGs associated with Gram positive microbes: vanA for vancomycin resistance (Clark et al. 1993) and mecA for methilicillin resistant Staphylococcus aureus (McKinney and Pruden 2012).

Amplification was not observed for these genes in the simulated CSO samples, despite amplification in positive controls.

Amplicon sequencing (300-bp, paired end) was performed at a commercial laboratory (MrDNA, Shallowater, TX) targeting the V3-4 regions of the 16S rRNA gene.

Sequences were analyzed using the mothur (Schloss et al. 2009) MiSeq Standard

Operating Procedure. Rarefaction curves are included in Supplemental Information. 60

Subsampling to obtain an equal number of sequences per sample resulted in 21,200 sequences/sample using a custom boot-strap.

3.3.6. Data Analysis

All statistical analyses were performed in R with an α=0.05 representing significant differences. A Kruskal-Wallis test was performed to test for differences (1) in total and viable log-normalized ARG concentrations for a given PAA treatment, (2) in the viable cell log-normalized ARG removal between different PAA treatments, (3) PAA concentrations across time for a given WW, and (4) PAA residual for the same treatment of different WW sources and dilutions. When significant differences were observed a post-hoc pair-wise t-test was performed with a Bonferroni correction for multiple comparisons. A Kruskal-Wallis test was also performed on TSS, COD, pH, CFU, and inactivation constants to test for differences between treatments with a post-hoc test as above. A student’s t-test (for parametric data) or Wilcoxon rank sum test (for nonparametric data) was performed to test for differences in log removal of a given gene with equivalent treatment (Ct=50 mg·min/L) for the two wastewaters tested. Gene log removal was calculated as

[ ] [ ]

[ ] − 1 − where [Gene]t=i is the gene concentration (copies/mL) at the time of interest i and

[Gene]t=0 is the initial gene concentration (copies/mL) for a given replicate. Ct values, expressed as mg·min/L, were calculated as [PAA applied (mg/L)]×treatment time.

Inactivation constants were calculated for data from both experiments combined for a given gene [except Ct=1200 mg·min/L PAA for BacHum, sul1, and tet(G)] using linear regression in Excel: 61

[ ] ln [ ] = − where α is the inactivation constant (L/mg·min), C is the concentration of PAA applied

(mg/L), n is the constant of dilution (taken to be 1 based on goodness of fit), and t is time

(min).

To compare microbial community structure for select total and viable samples, cluster analysis was performed in Primer7 (PrimerE, UK) on the Bray-Curtis dissimilarity matrix generated from log-normalized abundance data. To test for significant differences a SIMPROF test was performed.

3.4. Results

3.4.1. Disinfection of ARG-carrying cells

Treating simulated CSO effluent consisting of 23% WW with a Ct [nominal PAA dose (mg/L) × time (min)] of at least 100 mg·min/L PAA and 40% WW with 25 mg·min/L PAA resulted in significant decreases in sul1 concentrations from viable cells compared to the initial viable cell sul1 concentrations (all p<1.8x10-4, Fig. 3.1a, b).

Viable cell sul1 concentrations were lower than total sul1 concentrations observed at the same time for all tested Ct (all p<1.2x10-6) except for the 50 mg·min/L PAA treatment of either 23 and 40% WW (both p=1). Increased removal of sul1 concentrations from viable cells were not observed after 20 min with 23% WW and 5 mg/L PAA, after 10 min with 23% WW and 20 mg/L PAA, or after 5 min with 40% WW and 5 mg/L PAA

(p=1.0). Average log removals of viable cell sul1 ranged from 0.74 to 2.8 (Table 3.1).

Greater log removal was observed for sul1 with Ct of 50 mg·min/L PAA for 40% wastewater from WWTPb than for 23% wastewater with the same Ct from WWTPa

(p=0.036). 62

Figure 3.1 Comparison of total and viable cell gene copy numbers with different PAA treatment and exposure times for (a, b) sul1, (c,d) tet(G), (e,f) mexB, (g,h) BacHum, and (i,j) 16S rRNA genes. Experiments were performed with 23% wastewater from WWTPa or 40% wastewater from WWTPb to create the simulated CSO effluent. Experiments were performed for up to 60 min for WWTPa and 10 min for WWTPb. Error bars represent standard deviation of replicate (n=3, except as indicated in Table 3.1) samples. 63

Significant losses of viable cell tet(G) compared to initial viable cell tet(G) concentrations were observed for 23% wastewater with select Ct (100 and 300 mg·min/L) (p<0.02, Fig. 3.1c, d). No significant losses in viable cell tet(G) were observed for Ct 200, 400, and 1200 mg·min/L (p>0.16). A larger standard deviation observed between replicates at 200 mg·min/L may explain the lack of significant differences observed for that treatment. Viable cell tet(G) concentrations were lower than the paired total tet(G) concentrations for 23% WW with 1200 mg·min/L PAA (p=0.04).

Treatment of 23% WW resulted in 3.2 ± 0.30 log removal of viable tet(G) with 5 mg/L

PAA (Table 3.1). For the 40% WW experiment after 10 min both the control and experimental samples (Ct =50 mg·min/L) had significantly less viable and total cell tet(G) than at the start of the experiment (all p<2.0x10-3). No difference was observed in log removal with Ct of 50 mg·min/L PAA for 40% wastewater from WWTPb than for

23% wastewater from WWTPa (p=0.7).

There were no significant differences between the total and viable cell sul1 or tet(G) concentrations at the beginning of either experiment (all p=1). Changes in total

ARG concentrations were not observed with treatment compared to initial total ARGs during either experiment (all p>0.05). No significant change in ARG concentrations from total or viable cells was observed in untreated (0 mg/L PAA) controls (all p>0.30) except for tet(G) in 40% WW with no treatment and incubation of 10 min (p=2.3x10-5).

For mexB differences were not observed between total and viable cells at any time points

(all p>0.43, Fig. 3.1e, f) and changes in total or viable concentrations were not observed over time (all p>0.05). 64

3.4.2. Disinfection of Indicator Organism Markers and Cultivation

Treating simulated CSO effluent consisting of 23% WW for 20 min with 5 mg/L

PAA (Ct≥100 mg·min/L) and 20 mg/L PAA (Ct≥400 mg·min/L) resulted in significant decreases in BacHum concentrations from viable cells compared to the initial viable cell

BacHum concentrations (all p<0.02, Fig. 3.1g, h). Increasing contact time with 23%

WW beyond 20 min with either PAA concentration tested did not further reduce BacHum concentrations from viable cells (p>0.31). For 40% WW, 25 mg·min/L of PAA treatment decreased BacHum in viable cells compared to the initial viable cell concentrations (p<0.02), but concentrations in viable cells treated with 50 mg·min/L were not significantly different from initial viable concentrations (p=1.0). There was no significant difference in the total and viable cell BacHum concentrations at the beginning of either experiment (p>0.17). Changes in total BacHum concentrations were not observed with PAA treatment compared to initial total BacHum during either experiment

(p=1). No significant change in the total or viable cell BacHum concentrations was observed in untreated (0 mg/L PAA) controls across time during the experiment (p=1.0).

Log removals for viable cell BacHum ranged from 0.62 ± 0.56 to 1.8 ± 0.36 (Table 3.1).

No difference was observed in log removal with Ct of 50 mg·min/L PAA for 40% wastewater from WWTPb compared the same treatment for 23% wastewater from

WWTPa (p=0.2).

Cultivation was performed to confirm the loss of viability in cells with PAA disinfection demonstrated using vPCR for the 40% WW samples. Average starting concentrations in the 40% WW experiment were 1.0×105 CFU/mL (Fig. 3.4). CFU concentrations in treated samples (average concentration of 5.2×104 CFU/mL) were 65 significantly lower than control plates (average concentration 1.6×107 CFU/mL)

(p<0.03). Similar to qPCR results for ARGs and BacHum, significant losses of viable cell concentrations were observed with PAA treatments of 25 and 50 mg·min/L, resulting in 0.95±0.47 and 0.45±0.36 log removal of CFU, respectively.

Table 3.1 Log removal of viable-cell ARGs, fecal indicator and 16S rRNA gene concentrations during treatment of simulated CSO with PAA. Results represent averages ± standard deviation (n=3) or ± maximum and minimum values (n=2). Bold results indicate a statistically significant decrease in genes originating from viable cells was observed from T=0 min (p<0.05).

Treat ment WW [PAA] Time Ct Log Removal Fracti (mg∙ on min/ 16S (mg/L) (min) L) sul1 tet(G) mexB BacHum rRNA 5 10 50* 0.05±0.09 1.0±1.3 -0.17±0.24 0.14±0.03 0.27±0.83 5 20 100* 1.6±0.22 3.2±0.23 -0.38±0.07 1.0±0.63 0.02±0.54 5 60 300 2.0±0.12 3.2±0.25 0.50±0.06 1.8±0.36 0.03±0.36 23% 20 10 200 2.4±0.17 1.9 ±1.2 1.1±0.66 0.79±0.33 1.4±1.8 20 20 400 2.8±0.46 2.3±0.22 0.30±0.60 1.5±0.11 0.91±0.77 20 60 1200 2.8±0.03 2.2± 0.66 1.2± 0.10 1.1±0.35 3.29±0.13 5 5 25 0.90±0.36 0.32±0.15 - 0.62±0.56 0.34±0.21 40% 5 10 50 0.74±0.26 1.6±0.08 - 0.38±0.09 0.06±0.17 *n=2

3.4.3. PAA Decay Experiments

During Experiment 2, PAA test strips indicated the PAA concentration remained consistent over the course of treatment and disinfectant concentrations were 0 mg/L after quenching. PAA decay experiments were performed to better characterize the initial, final, and post quenching PAA doses and the impact of COD on PAA residual. For 66

WWTPa, the experiment was performed on 23% (75±18 mg COD/L) and 11.5% (39±13 mg COD/L) WW (summarized in Table S2). Nominal concentrations of 5 and 20 mg/L

PAA were comparable to the measured initial concentrations of 5.2±0.6 and 20±1.9 mg/L

PAA, respectively. No change in PAA concentrations were observed in the 11.5%

WWTPa reactor with treatment through 20 min (all p>0.18). At 60 min of treatment of

23% WWTPa, PAA concentration decreased compared to initial measured concentrations

(both p<0.04) with 59±28% of the 5mg/L PAA and 82±0.7% of the 20 mg/L PPA applied remaining as residual (Fig 3.2). For WWTPb, the experiment was performed on 40%

(158±34mg COD/L) and 20% (117±69mg COD/L) WW. The PAA dose applied in this experiment was 5 mg/L PAA for all samples and the measured initial concentrations were

4.7±0.1 mg/L PAA. Decreases in disinfectant concentrations were observed after 5 min or more of treatment for 40% WW (p<0.01) and 10 min for 20% WW (p=0.019, Fig.

3.2). After quenching, 0-3% of the initial PAA dose remained across Experiment 3.

Comparing between the two WW sources for Ct 50 mg·min/L PAA, the percentage of

PAA remaining was least for 40% WW from WWTPb (all p<0.02), which had the highest COD. The percent of PAA remaining at Ct of 50 mg·min/L PAA was slightly less for 23% WWTPb than 20% WWTPa (which had comparable COD) and for 11.5%

WWTPa than 20% WWTPb (both p<0.037). 67

Figure 3.2 PAA concentration remaining (C) compared to initial PAA concentration (Co) during treatment of (a) 23% wastewater from WWTPa, (b) 11.5% wastewater from WWTPa, and (c) 20% or 40% wastewater from WWTPb and immediately after quenching. Treatment was performed with either 5 mg/L or 20 mg/L PAA. Error bars represent standard deviation of replicate (n=3) samples.

3.4.4. Water Quality

Water quality was monitored throughout the 40% WW experiment (Experiment 2).

Conductivity in samples treated with 25 mg·min/L PAA was greater than the no 68 treatment controls (p=0.024) but not samples treated with 50 mg·min/L PAA (p=0.44,

Fig. 3.5a). COD increased significantly after 5 to 10 min of treatment with 5 mg/L PAA

(all p=8.5×10-6, Fig. 3.5b). COD in controls was unchanged across time (p=1.0).

Differences were observed in pH for select samples across time and with treatment +/-

0.4 pH units (Fig. 3.5c). TSS measurements were similar in treated samples and controls

(p=1.0, Fig. 3.5d).

3.4.5. Total Bacterial Community Disinfection

16S rRNA gene copies were quantified as a surrogate for total bacterial population. Treating 23% WW with 1200 mg·min/L PAA resulted in a significant decrease in viable cell 16S rRNA genes (p=8.9x10-5, Fig. 3.1i, j). Viable cell 16S rRNA gene concentrations were lower than total cell 16S rRNA concentrations in 40% WW with 25 mg·min/L PAA (p=4.3x10-5) but comparable to the starting viable cell 16S rRNA concentrations (p=0.05). Treating the 40% WW with 50 mg·min/L PAA resulted in no losses of viable or total 16S rRNA (p=1). No differences were observed between log removal with Ct of 50 mg·min/L PAA for 40% wastewater from WWTPb and 23% wastewater from WWTPa (p=0.91).

Treatment of samples with the DNA intercalating dye PMA resulted in a significant shift in microbial community structure between total and viable samples even without PAA treatment (p=0.001, Fig. 3.3a). The total community was dominated by

Proteobacteria throughout the experiment (Fig. 3.3b). The viable community with 0 mg·min/L PAA was also dominated by Proteobacteria but had lower relative abundances of Fusobacteria, Actinobacteria, Clostridia, and Bacilli. The viable cell community observed with 1200 mg·min/L PAA treatment was least like the other samples (81% 69 similar, p=0.001). This community was dominated by Clostridia and marked by increases in Bacilli, Actinobacteria, and Erysipelotrichia and characterized by losses of

Flavobacteria and Bacteriodia. No significant differences were observed in the microbial community structure for any treatment among the samples representing the total community. Likewise, no significant differences were observed in the viable community across time with no PAA treatment. Replicate samples that were sequenced

(total, Ct=1200 mg·min/L PAA) were 92.7% similar.

3.4.6. Kinetics

Data were fitted to the Chick-Watson model and inactivation coefficients are listed in Table 3.2. R2 values for the Chick-Watson model fit ranged from moderate to substantial (0.45±0.17-0.76±0.21). Data were pooled across the experiments for the kinetic analysis given that reasonable model fits were observed with the combined data

(note: comparable log removals were observed for the same Ct for four of the five genes tested with the WW from different sources and dilutions). Inactivation rates were greatest for sul1 and tet(G), which were not significantly different from one another

(p=1). The inactivation coefficient for BacHum was significantly less these two ARGs

(both p<0.001). The inactivation coefficient for the 16S rRNA gene was significantly less than all other genes tested (all p<0.024). Correlation was not observed between qPCR amplicon length and inactivation coefficient. 70

a.

b.

Figure 3.3 (a) Cluster analysis and (b) relative abundance of Bacterial phyla for different PAA treatments and exposure times (Ct values in mg·min/mL). Samples connected by red bars on the cluster tree do not have significantly different structures. #’s represent replicate samples that were sequenced.

Table 3.2 Average ± standard deviation for Chick-Watson disinfection coefficients (α) and R2 values of model fit combined for both 23% and 40% WW from different WWTP (n=3). BacHum tet(G) sul1 16S rRNA amplicon 89 134 163 202 length (bp) Chemistr TaqMan SybrGreen SybrGreen SybrGreen y α (L/mg 1.1×10-2 1.9×10-2 1.8×10- 6.1×10- min) ±1.7×10-3 ±1.4×10-3 2±1.3×10-3 3±1.2×10-3 71

R2 0.76±0.21 0.45±0.17 0.67±0.19 0.67±0.14

3.5. Discussion

3.5.1. PAA for Treating CSO

PAA treatment resulted in 0.62-3.3 log removal of the fraction of sul1, tet(G), and

BacHum in viable cells in simulated CSO effluent with 25 mg·min/L PAA treatment or greater. The results of this study targeting viable cell genes can be compared to log removal of cultivable organisms in the literature given that PAA disinfection studies (and regulations) generally rely on cultivation or most probably number (MPN) evaluations.

In a similar study with simulated CSO effluent, there was 3.4-5.6 log removal of E.coli compared to <2 log removal of Enterococcus in 5% WW with nominal PAA Ct of 25-

300 mg∙min/L (Chhetri et al. 2014). The greater log removal observed with 5% WW may be due to there being less oxidizable organic matter than in the current study using

23-40% WW. Lower PAA doses were applied in a CSO treatment pilot study that achieved E.coli, fecal coliform, and Enterococcus log reductions of 1.7-2.3 with 2 mg/L

PAA for 3 min (nominal Ct=6 mg∙min/L).(Patoczka et al. 2016) However, for field studies where PAA was not reported to be quenched that have PAA, residual may result in potentially a longer incubation time than reported given the allowable hold time for these indicator organism techniques. Viable cell BacHum losses observed here (0.62-

1.8-removal with 25-1200 mg∙min/L PAA) are lower than previously reported 3-log removal of cultivable total coliforms, fecal coliforms and fecal Streptococcus observed with a nominal Ct of 100 mg∙min/L PAA (PAA=10 mg/L) in disinfected wastewater effluent.(Lazarova et al. 1998) Other reports of differences in chlorine disinfection efficiencies between vPCR and cultivation based techniques for the same organism have 72 suggested viable but not cultivable organisms may play a role in this discrepancy.(Varma et al. 2009)

Having a low Ct is of interest when treating CSO effluent because (1) the footprint available for end-of-pipe treatment systems is often limited and (2) PAA was previously reported to be been more expensive than chlorination (estimated at five times the cost for treating wastewater in 2004),(Kitis 2004) although more recent case studies

(2012) suggest that the cost may be comparable.(Graham et al. 2012; United States

Environmental Protection Agency 2012) Results of our study provide insight into a potential treatment for ARG-carrying cells at CSO discharge locations. Further reductions in viable cell ARG concentrations were not observed for treatments longer than 10 min and tet(G) and 10 to 20 min for treatment of sul1. Thus, the rapid activity of

PAA in reducing select viable ARG concentrations could minimize the treatment infrastructure requirements if PAA is supplied at sufficient dose (5-20 mg/L with residence time of 5 min).

The CSO effluent source may impact the PAA dose required for significant removal of viable cell ARGs and BacHum given that lower doses were required for significant removals for the 40% compared to 23% WW. For example, while the starting sul1 concentration was higher in the 40% WW compared to 23% WW (p=0.02), significant removal of viable cell sul1 was observed with a lower Ct for 40% WW (25 mg∙min/L) compared to the 23% WW (100 mg∙min/L). Greater log removal of sul1 was observed with the same Ct (50 mg∙min/L) for the 40% WW compared to 23% WW, but notably not the other genes analyzed. Measuring PAA residual and COD could help explain these observations, as was performed for Experiment 3 demonstrating that 73 generally higher COD WW had a higher PAA demand, however these results are not directly comparable given the sampling was performed at a different time. Others have observed that PAA concentrations up to 6 mg/L for 60 min (nominal Ct=360 mg∙min/L) could not consistently meet target fecal coliform levels in disinfected WW effluent due to day-to-day variability in effluent quality.(Gehr et al. 2003) Gehr et al.’s observation of variability in PAA performance may be due in part to the differences in PAA available for treating cells that would result from the demand of other organic constituents being oxidized by PAA. In contrast, WW concentration had little effect on disinfection efficiency of Enterococcus in 5, 15 and 40% WW treated with up to 30 mg/L for 10 min.

(Chhetri et al. 2014) Considering the variability in effluent quality across a storm,(Eramo et al. (in review)) testing a range of CSO effluents is likely a pertinent consideration for end-of-pipe treatment.

The impact of PAA treatment on water quality was determined by monitoring pH, conductivity, COD, and TSS. Of these parameters, pH was slightly lower in treated samples, unlikely of practical significance, and COD was higher in treated samples. The increase in COD with PAA treatment is expected to result from reactions with transition metals, suspended and dissolved solids, and organic species in the wastewater.(Block

2016) Others have reported increases in biological oxygen demand (BOD) and COD after disinfection with oxidants thought to be due to the oxidant causing modifications to more recalcitrant organic matter resulting it in being more readily oxidable by the COD test.(El-Rehaili 1995) 74

3.5.2. PAA Degradation

PAA decay experiments demonstrated that 59-87% of the initial PAA dose was present after 10-60 min of treatment. Extended exposure times of 60 min (nominal

Ct=1200 mg min/L) were needed before significant decreases in PAA were observed for

11.5 and 23% WWTPa. However, significant PAA decay was observed after only 5 min

(nominal Ct=25 mg min/L) with 40% WW and only 10 min (nominal Ct=50 mg min/L) with 20% WW from WWTPb. The generally greater PAA residual remaining with lower dilutions of WW and lower COD are consistent with greater PAA consumption by higher

WW concentrations and COD observed by others.(Chhetri et al. 2014; Gehr et al. 2003)

PAA concentrations measured after quenching indicated that the concentrations of sodium thiosulfate and catalyze were sufficient to reduce the remaining amount of disinfectant. For 23% WWTPa and 5 or 20 mg/L PAA, compared to nominal doses of

PAA multiplied by time of 50, 100, 200, 300, 400 and 1200, observed values were

51±9.0, 91±14, 189±7.7, 199±36, 351±15 and 927±32 mg∙min/L, respectively. For 40%

WWTPb treated with 5mg/L PAA the nominal dose of PAA multiplied by time of 25 and

50 resulted in observed Ct of 19±0.49 and 33±0.47 mg∙min/L, respectively.

3.5.3. Mode of Action

PAA was effective at reducing the fraction of sul1 and tet(G) from viable cells present in simulated CSO effluent. This result is consistent with reports that PAA did not enhance overall ARG removal (i.e., reduce total ARG concentrations measured by qPCR) in post-secondary treated WW effluent after disinfection. (Di Cesare et al. (2016);

(Luprano et al. 2016)) The application of vPCR in this study incorporates the specificity of qPCR and overcomes the limitation of qPCR to differentiate between DNA originating 75 from viable and nonviable sources. PAA’s mode of action is presumed to be oxidization of sulfhydryl and sulfur bonds in proteins, enzymes and other metabolites, including those in microbial membranes (Kitis 2004). Because treatment of cells with PMA reduces the qPCR signal from extracellular DNA and DNA from cells with compromised membranes, it can be reasoned that PAA was effective at disrupting the membranes of cells carrying sul1 and tet(G) genes resulting in significant decreases in viable cell ARG concentration. Thus, following treatment, the majority of the sul1, tet(G), and BacHum gene concentrations present in simulated CSO samples originated from nonviable cells. Live-dead staining with flow cytometry was previously used to demonstrate the membrane destroying mechanism of PAA for WW disinfection (Di

Cesare et al. 2016). At the beginning of treatment in the current study, there were no differences between total and viable ARG concentrations indicating ARGs were predominantly in viable cells, consistent with reports that bacterial cells with intact membranes comprised approximately 70% of the bacterial community in municipal wastewater influent (Di Cesare et al. 2016). A previous study differentiated intra- and extracellular ARGs in environmental matrices using a washing technique, observing extracellular DNA generally made up only less than 0.5% of total DNA in animal sludge samples (Zhang et al. 2013). In contrast, extracellular ARGs were found in higher concentration than intracellular ARGs in river sediments including those receiving treated wastewater (Mao et al. 2013).

The viability-based method applied here may overestimate the dose required for inactivation of ARG carrying cells given that it is possible for cells with intact membranes to not be viable (i.e., if DNA is sufficiently damaged). Although the use of 76

PMA is a more conservative method to measure loss of viability because it relies only on the permeability of the membrane, linear correlations were observed between reduction in cultivability and of PMA-qPCR signals in samples treated with various disinfectants

(Nocker et al. 2007b). While UV damage to cells can be repaired, there are not known reports of repairs of membrane damage from oxidants. Nonetheless, the method does allow for better estimation of the risk of extracellular DNA versus intracellular DNA, which would be exchanged between cells using different mechanisms, compared to qPCR.

PAA was not effective at reducing total ARG concentrations in treated simulated

CSO effluent consistent with a recent investigation for its use for disinfection of WW (Di

Cesare et al. 2016). ARGs from non-viable sources are of interest given that the rate of

ARG propagation may be expected to be different for ARGs in viable cells (which can spread via growth and horizontal gene transfer) compared to extracellular ARGs/ARGs in cells with compromised membranes (which can spread via transformation) (Ashbolt et al.

2013). The ARGs sul1, tet(G), and mexB investigated here may be found on chromosomal DNA or plasmids.(Chopra and Roberts 2001; Skold 2000; Sondergaard et al. 2012) Transformation of extracellular ARGs (plasmids carrying genes for kanamycin- resistance) and expression was previously demonstrated with sediment cells in a controlled experiment (Mao et al. 2014). Given that the primer sets used in our study targeted PCR inserts (134-244 bp) that were not the full length of the ARG, it is possible that disinfection with PAA resulted in ARG damage to the given gene outside of the target amplicon and therefore not detectable with our protocols. PAA was reported to reduce the concentration of nucleic acids as measured by optical density in a study 77 looking for the mechanisms of fungicidal effects (Tutumi et al. 1973). However, the observation that PAA disinfection did not reduce total ARGs in treated simulated CSO is consistent with other disinfection studies with oxidants (Dodd 2012). For example,

Fahrenfeld et al. (2013) observed that chlorination did not reduce total sul1, tet(G), or tet(O) concentrations in secondary WW effluent (losses of sul2 were observed post- chlorination). In contrast, UV disinfection reduces total ARG concentrations at higher doses of UV than is required for loss of viability of antibiotic resistant bacteria

(McKinney and Pruden 2012). Therefore, treatment combining PAA to damage cell membranes inactivating microbes and providing better exposure for UV to disrupt DNA may be a strategy for mitigating the risk of ARG propagation in CSO impacted environments.

The Cts reported to result in significant losses of antibiotic resistant cultivable organisms in treated WW were comparable or higher than minimum found to reduce vPCR signals from ARG-carrying cells in simulated CSO effluent in this study. PAA disinfection significantly decreased detection of cultivable ARG carrying uropathogenic

E. coli isolates (detected via microarray following cultivation) in treated WW effluent at a dose of 0.9-2 mg/L and a contact time of 30 min to meet a goal of 200 CFU/100 mL

(Biswal et al. 2014). Exposure to PAA was associated with 1.6-3.7 log reductions in uropathogenic E. coli concentrations with 30, 50 and 60 mg∙min/L PAA (applied dose × time) (Biswal et al. 2014).

3.5.4. Gene to Gene Comparisons

There was gene-to-gene variation in the required disinfectant dose for removal of viable cell gene copies. For example, 50 mg∙min/L PAA resulted in significant loss of 78 viable cell tet(G) for 40% WW compared to 25 mg∙min/L PAA for significant loss of viable cell sul1. However, the inactivation coefficients for these two ARGs were comparable. In contrast, mexB concentrations did not change with PAA treatment, although at the highest PAA treatment viable cell concentrations were almost different compared to the initial viable cell concentration (p=0.05), thus behaving most similarly to the 16S rRNA gene. There were significant differences observed between the inactivation coefficients with sul1 and tet(G) > BacHum >16S rRNA (inactivation coefficients were not evaluated for mexB given that significant losses in viable cell concentrations were not observed across the experiment). Notably, 16S rRNA genes originating from viable cells exhibited a significant decrease only after 60 min of treatment at the highest Ct tested (1200 mg∙min/L PAA). qPCR product lengths and temperature have were previously reported to influence PMA-qPCR(Contreras et al.

2011) and could contribute to these differences. Short DNA fragments (<200 bp) were not found to have completely reduced signals for certain heat-killed strains (Banihashemi et al. 2012). However, the 16S rRNA amplicon length was greater than those for sul1 and tet(G) where viable cell gene copy losses were observed with lower PAA Ct (and all genes quantified using SybrGreen chemistry). mexB had the longest targeted amplicon length and its concentrations were unchanged following PAA disinfection. Further,

BacHum had the shortest amplicon (quantified with TaqMan probe chemistry) and behaved similarly to sul1 and tet(G). Correlation between inactivation coefficient and amplicon length were not observed. Therefore, factors other than amplicon length (e.g., temperature, target cell, reaction chemistry) are relevant for understanding PAA disinfection efficiency. 79

The impact of membrane structure (Gram-positive versus Gram-negative) is of interest for PAA disinfection. The ARGs sul1 and tet(G) measured in this study are associated with Gram-negative organisms and the action of PAA on outer membrane lipoproteins may facilitate its effectiveness against Gram-negative cells (Kitis 2004).

Bacteriodales, the host order for BacHum genes, are also Gram-negative organisms. The lower Ct values required for removal of viable sul1 and tet(G) and fecal indicator compared to universal 16S rRNA gene may be partially explained by the membrane composition of cells associated with these primer targets given that both Gram-negative and Gram-positive cells contain 16S rRNA. There is evidence that much longer PAA treatment times (240 and 360 min) were required for 3 and 4-log removal of cultivable

Enterococcus, a Gram-positive organism, compared with E.coli, a Gram-negative organism requiring only 10 min treatment time for similar removals (Chhetri et al. 2014).

This would suggest that large increases in treatment time regardless of PAA concentration may be required for specific targets, as observed here for 16S rRNA and mexB. Other researchers have reported that PMA may need to be optimized for different cell types and enhancing kits are available for targeting Gram-positive bacteria. In fact, in the PMA and PAA treated sample for amplicon sequencing, marked increases in

Clostridia, Actinobacteria, and Bacilli were observed. All three of these phyla are Gram- positive, thus supporting that either or both PAA and PMA were less effective at reducing vPCR signals from Gram-positive bacteria using our protocol. Given that these phyla had higher relative abundances in the total than the viable populations with Ct of 0 mg∙min/L PAA, differences in PAA’s ability to disrupt membranes between Gram- positive and -negative bacteria may contribute. The initial bacterial community was 80

>90% Gram-negative at the start of the experiment therefore differences in membrane structure are unlikely the only contributing factor. ARGs associated with Gram-positive microbes (vanA and mecA) were tested for but not observed in the samples. However, mexB, which has mostly been associated with both Gram-negative bacteria, was not impacted by PAA, highlighting the potential for specific disinfection efficiencies across targets.

In the current study, differences in microbial community structures for samples treated with PMA were observed compared to untreated samples from the initial Ct of 0 mg∙min/L. Community shifts have been observed in response to PAA treatment of drinking water biofilms, (Roeder et al. 2010) UV disinfection of biologically treated wastewater (Hu et al. 2016) and in the viable community of ozonated municipal wastewater sludge investigated by PMA-modified Miseq sequencing (Tian et al. 2017).

Similar to this experiment, Proteobacteria was a dominant phylum in the ozonation study although there was a decrease in the relative abundance in the viable population.(Tian et al. 2017)

3.6. Conclusions

This work provides insight into the partitioning of ARGs between cells with intact membranes versus cells with compromised membranes or extracellular DNA from PAA treated CSO effluent. The viability method applied allowed for demonstration that PAA was effective at reducing viable cell sul1, tet(G), and fecal indicator with at least 25 or

100 mg∙min/L PAA and 16S rRNA genes with 1200 mg∙min/L PAA. The PAA disinfection efficiency for sul1 carrying cells varied by wastewater source with a higher nominal Ct required for more dilute simulated CSO effluent indicating that treatment of 81 select targets may be a greater function of source water quality than dilution factor.

Inactivation coefficients for sul1 and tet(G) were greater than for BacHum and inactivation coefficients for these ARGs and BacHum were greater than for 16S rRNA.

These inactivation coefficients may be used to estimate removal of viable concentrations of these gene targets in CSO effluent. Overall, results indicate that further steps are necessary to remove total ARGs in PAA treated CSO effluent [e.g., hydrodynamic separation (Eramo et al. 2017a), UV disinfection (McKinney and Pruden 2012)].

Acknowledgements

Laboratory assistance was provided by Hannah Delos Reyes, Sophia Blanc, and Reba

Oduro. Thanks to our utility partners for providing access to influent samples. Funding for this project was provided by grants from the New Jersey Water Resources Research

Institute, the National Science Foundation (#1510461), and a Mark B. Bain Fellowship from the Hudson River Foundation to AE. 82

3.7. Supplemental Information

Table 3.3 Primers, annealing temperatures, and amplicon lengths. Gene Primer sequence Ta Amplicon Source (°C) length (bp) sul1 CGCACCGGAAACATCGCTGCAC (Pei et al. 65 163 TGAAGTTCCGCCGCAAGGCTCG 2006) tet(G) GCAGAGCAGGTCGCTGG (Aminov 68 134 et al. CCYGCAAGAGAAGCCAGAAG 2001) BacHum TGA GTT CAC ATG TCC GCA TGA (Kildare CGT TAC CCC GCC TAC TAT 60 81 et al. CTA ATG 2007) /56-FAM/TCC GGT AGA CGA TGG GGA TGC GTT /36-TAMSp/ 16S rRNA CCTACGGGAGGCAGCAG (Muyzer 65 202 et al. ATTACCGCGGCTGCTGG 1993) MexB GTGTTCGGCTCGCAGTACTC (Yoneda 60 244 et al. AACCGTCGGGATTGACCTTG 2005) 83

Figure 3.4 Colony forming units (CFU) on LB agar from 40% WW treated with 0 mg/L or 5 mg/L PAA (n=2 or 3).

Figure 3.5 Water quality parameters (a) conductivity, (b) chemical oxygen demand (COD), (c) pH and (d) total suspended solids (TSS) in 40% WW treated with 5 mg/L PAA compared to no PAA controls. 84

Figure 3.6 Rarefaction curves for samples treated with 20mg/L PAA or no treatment controls for 0 or 60 min. Viable indicates samples treated with propidium monoazide prior to submission for sequencing. # indicates replicate number.

Table 3.4 Water quality data for source wastewater from disinfection experiments (Experiments 1 and 2) and PAA degradation experiment (Experiment 3).

Experiment Experiment 1 Experiment 3 2 WWTPa WWTPb WWTPa WWTPb Sampling date 10/26/2015 11/16/2016 7/14/2017 7/27/2017 Percent WW 23% 40% 23% 11.50% 40% 20% 228 ±109 63 ±31 - 280 (100% WW) TSS (mg/L) (100% WW) (100% WW) COD (mg/L) - 79±8 75±18 39±13 158±34 117±69 pH - 7.7±0.01 6.88 6.59 85

4. CHAPTER 4: Settling and peracetic acid for end-of-pipe treatment of sul1-carrying

indicator organisms and impact on receiving water

In review for the ASCE Journal of Environmental Engineering Authors: Alessia Eramo, Sophia Blanc, Nicole L. Fahrenfeld

4.1. Abstract

Combined sewer overflows (CSOs) negatively impact water quality during wet weather. An

end-of-pipe treatment train with settleable particle removal and peracetic acid (PAA) disinfection

was investigated for treatment of E. coli, sul1 gene-carrying E. coli, and total coliform (TC) in

simulated CSO effluent. Settling reduced chemical oxygen demand by 49±11% and total

suspended solids by 78±7%. Overall removals of 3.1±0.14 and 2.5±0.37-log were achieved for

E. coli and total coliform (TC), respectively, with settling and disinfection, but these removals

were not significantly different than removals with disinfection only. After disinfection, the

fraction of E.coli carrying the antibiotic resistance gene (ARG) sul1 increased. Treated samples

were spiked into estuarine water to determine regrowth potential of target bacteria following

release. After seven days, E.coli was not detected in reactors with treated CSO and TC

concentrations decreased significantly. This research provides insight into a potential end-of-pipe

treatment and suggests that disinfection rather than settling is more effective for microbial

treatment of the wastewater component of CSO effluent.

4.2. Introduction

Combined sewer systems collect both wastewater and storm water runoff and during

sufficiently large wet weather events the untreated waste stream overflows into adjacent surface

water. This releases untreated fecal contamination in waterways across the United States and

globally (Eaton et al. 2013), posing a public health risk (Donovan et al. 2008), and limiting the

growth potential of communities (Van Abs 2014). The sewage component of combined sewer

overflows (CSOs) results in significantly higher surface water E.coli concentrations compared 86 with stormwater only events (McLellan et al. 2007). Upgrading combined sewer systems in the

US has been estimated to cost $40.8 billion (United States Environmental Protection Agency

2016). Therefore, lower cost gray and green infrastructure upgrades are of interest (Copeland

2014), including end-of-pipe treatment.

Understanding the impact of different treatments trains on CSO effluent quality is needed to help inform effective full-scale end-of-pipe treatment approaches. Different proposed end-of- pipe treatment technologies include combined treatment via hydrodynamic separation units followed peracetic acid (PAA) or ultraviolet disinfection (Bayonne Municipal Utilities Authority

2017), chlorine disinfection (New York City Department of Environmental Protection 2014), performic acid disinfection (Chhetri et al. 2015), and constructed wetlands (Meyer et al. 2012).

Removal of settleable particles via hydrodynamic separation is well studied for end-of-pipe treatment of total suspended solids (TSS) in storm water (Andoh and Saul 2003). During wet weather, many fecal indicators (Krometis et al. 2007) and harmful DNA fragments [antibiotic resistance genes (ARGs)] (Eramo et al. 2017a) are attached to particles, a portion of which could be removed by properly sized and operated hydrodynamic separation units (Wilson et al. 2009).

We previously estimated that up to 88% of ARGs and 70% of 16S rRNA gene copies in samples collected at the end-of-pipe during wet weather were attached to settleable particles (Eramo et al.

2017a). The remaining bacterial load would remain in the effluent and therefore either direct or secondary treatment by disinfection is of interest.

PAA is an attractive disinfectant for CSOs because untreated wastewater has a high organic content and PAA has not been shown to form hazardous disinfection byproducts, in contrast to chlorine (Kitis 2004). PAA differs in its behavior to chlorine in that the active oxygen produced by PAA targets sulfhydryl and sulfur bonds in cell membranes, while hypochlorite is non- selective and rapidly reacts with both organic and nonorganic compounds (McFadden et al. 2017) and creates several known disinfection by-products. We previously demonstrated that PAA 87 disinfection resulted in 2.8-log inactivation of cells carrying select ARGs (Eramo et al. 2017b).

In comparison to hypochlorite, PAA was shown to require a lower CT value (concentration × time) to begin disinfection (McFadden et al., 2017), an attribute that may make it more appropriate for rapid treatment at CSO outfalls.

Beyond the focus on fecal indicators of regulatory interest, there is motivation to look for a broader range of microbial agents given that antibiotic resistant bacteria were found to be prevalent during wet weather in CSO impacted waters (Young et al. 2013). Of particular relevance for end-of-pipe treatment is the observed selection for ARGs over a four-day incubation of PAA-disinfected wastewater effluent (Di Cesare et al. 2016) even though PAA disinfection of uropathogenic E. coli was found to generally decrease ARGs (Biswal et al., 2014).

PAA damages cell membranes but does not destroy DNA (Di Cesare et al. 2016; Eramo et al.

2017b) and therefore non-viable sources of ARGs may still pose a hazard in the environment if

ARGs are released then transformed (i.e., incorporated into the genome) by other competent bacteria (Dodd 2012). To the authors’ knowledge, there have not been other studies investigating the potential for PAA treatment of CSOs to promote the selection for/regrowth of ARG-carrying fecal indicators.

Understanding the potential for regrowth of fecal indicators, especially those carrying ARGs, is of interest for understanding the impact of end-of-pipe treated CSO effluent on surface water quality. Regrowth following PAA treatment may be a concern because the acetic acid already present in commercial PAA mixtures (Gehr et al. 2003) and resulting from the breakdown of

PAA may serve as an easily assimilable carbon source (Kitis 2004). In other studies, regrowth of coliform bacteria disinfected with PAA was not observed after short incubation periods of five hours (Rossi et al. 2007), six hours (Mezzanotte et al. 2003), and 29 hours (Antonelli et al. 2006), but was observed after one day (Lefevre et al. 1992; Sanchez-Ruiz et al. 1995) and four days (Di

Cesare et al. 2016). Most of these studies measured regrowth in treated secondary effluent 88 without inoculation in environmental waters. At the particular location where surface water was obtained for our regrowth study, it is recommended that contact with surface water is avoided at least 72 hours after rain (Comi and Mickley 2017). However, Lazarova et al. (1998) observed a slow increase in culturable fecal coliform after seven days of incubation of PAA disinfected wastewater effluent diluted 1:10 in seawater. Thus, this longer incubation period was selected for assessing regrowth in the current study.

The objectives of our study were (1) to determine the effect of removing settleable particles and PAA disinfection on reducing the concentration of cultivable fecal bacteria, including those carrying a sulfonamide resistance ARG (sul1), and (2) to assess the potential for regrowth upon release to saline surface water. To achieve these goals, a bench-scale experiment was performed using simulated CSO effluent made with diluted wastewater. The simulated CSO was treated by either removing settleable particles (Krometis et al. 2007) then disinfecting with PAA or directly disinfecting with PAA. Water quality parameters, E. coli, sul1 gene-carrying E. coli and total coliform (TC) concentrations were measured prior to and after each treatment. Next, treated effluent was spiked into natural river/bay water to evaluate the potential for indicator organism regrowth, particularly sul1 gene-carrying E. coli. The ARG sul1 was monitored in this study because it is commonly observed in wastewater and is associated with mobile genetic units including plasmids and class I integrons, which promotes its efficient dissemination among E.coli and other potentially pathogenic microbes (Skold 2000). Results of this study are compared to recent reports on PAA disinfection and can provide insight into the fate and risk associated with release of a treated CSO waste stream to the environment.

4.3. Materials and Methods

4.3.1. Sample Collection and CSO Simulation

Approximately two liters of 24-hr composite wastewater influent was collected from a municipal treatment plant that receives wastewater from a sanitary sewer system. Surface water 89 grab samples were collected from Perth Amboy, NJ, where the Raritan River meets the Raritan

Bay (Fig. 4.1) during baseflow conditions. Mixing of water in the bay is affected by a large counter-clockwise gyre and salinity ranges from a low of about 12 parts per thousand (ppt) near the mouth of the Raritan River, which is closer to the sampling location, to about 32 ppt off of

Sandy Hook (USGS 2018). All water samples were collected in sterile sampling bottles, transported to the laboratory at 4 oC, and processed upon arrival.

Simulated CSO was prepared by diluting the wastewater with sterile deionized water to create a 40% wastewater solution by volume. This dilution was selected to represent the higher end of the range of wastewater concentrations observed in CSO events (Chhetri et al. 2014; Passerat et al. 2011). Although the CSO waste stream collected prior to discharge into surrounding surface water was not accessible and therefore not used for this study, the microbial community structure in wastewater was found to be similar to that observed at select time points for water collected outside of a CSO outfall (Eramo et al. 2017a). Additionally, wastewater is a major source of fecal indicator organisms in CSO (Passerat et al. 2011).

Figure 4.1 Map of surface water sampling location and combined sewer outfalls in the vicinity of the sampling location. 90

4.3.2. Treatment of Simulated CSO by Settling and PAA Disinfection

Settleable particle removal and PAA disinfection were performed in series to determine the removal of regulatory target organisms: culturable E.coli and total coliform (TC). Experiment

1 was performed in triplicate with samples collected on three separate days over a two-week period in September 2017. To separate settleable and suspended microbes in the simulated CSO effluent, a centrifugation method previously validated by Krometis et al. (2007) was applied.

Briefly, four 500-mL aliquots of simulated CSO effluent were centrifuged at 1160 ×g for 10 min

(Eppendorf Model 5810, Hauppauge, NY). The top 70% volume fraction was collected by pipette and represented the fraction containing suspended cells. To determine the effectiveness of

PAA disinfection after removing settleable particles, a 400-mL aliquot of the suspended fraction was treated with 5 mg/L PAA for five minutes during rapid stirring on a stir plate. The nominal

PAA dosage and contact time were selected because they were found to be effective at significantly reducing viable-cell ARG concentrations in simulated CSO effluent with 40% wastewater (Eramo et al. 2017b). An aliquot of the disinfected sample was removed for the regrowth study described below. Then, the disinfection reaction was quenched with 100 mg/L sodium thiosulfate and 50 mg/L catalase before cultivation.

Given that significant removal of fecal indicator organisms was not observed by settling in the first experiment, a second experiment was conducted to determine if removal of settleable particles improved disinfection efficiency and/or reduced the PAA demand. Experiment 2 was performed in duplicate with samples collected approximately one week apart in June and July

2018. Samples were collected from the same locations as for Experiment 1 using the same methods. As previously described, samples of the simulated CSO effluent were prepared, treated by settling then PAA disinfection, and an aliquot of the suspended fraction was removed for the regrowth study and the remaining suspended sample was quenched before cultivation. Separate 91 aliquots of the simulated CSO effluent were treated directly with PAA (5 mg/L, 5 min). The reactors were quenched before cultivation. Finally, no-treatment controls were performed using simulated CSO effluent. The controls were stirred for 5 min with no PAA addition then an aliquot was removed for the regrowth study. A schematic of the treatment conditions and regrowth experiment described in the next section is provided as Fig. 4.5.

4.3.3. Regrowth Experiment

Aerobic reactors (400 mL) were prepared to assess the potential for E. coli and TC regrowth after the release of separated and disinfected simulated CSO. For two-step treated samples from both experiments, a 40-mL subsample was removed after PAA disinfection but before quenching and mixed with surface water for a 1:10 (v:v) dilution of treated sample to surface water. Control reactors were prepared during Experiment 2 with a 1:10 (v:v) dilution of untreated simulated CSO to surface water. Surface water control reactors were prepared for both experiments without addition of CSO effluent. All reactors were incubated with stirring for seven days at room temperature under aerobic conditions.

4.3.4. Water Quality Measurements and Fecal Indicator Enumeration

Select water quality parameters and fecal indicator organisms were measured over the course of the treatment and regrowth experiments. TSS was measured in the simulated CSO effluent, simulated CSO effluent following removal of the settleable fraction, and in surface water samples. pH, conductivity, COD, E.coli, and TC were analyzed in simulated CSO prior to treatment, following removal of settleable particles, and at the end of PAA disinfection. In the no-treatment control, E.coli and TC were measured after stirring for 5 min. Fecal indicator organisms, COD, pH, and conductivity were measured in regrowth reactors on days 0 and 7 of incubation.

TSS were measured according to Environmental Sciences Section (ESS) Method 340.2

(1993). Wastewater and surface water pH and conductivity were measured in the lab with a 92 calibrated benchtop meter (pH 700, Oakton) and a multimeter (Orion Star A329, Thermo

Scientific), respectively. Chemical oxygen demand (COD) was analyzed according to Hach

Method 8000 with Hach COD vials (20-1500 mg/L range) and a DR2700 spectrophotometer, using the manufacturer’s protocol.

E. coli and TC were enumerated using membrane filtration and MI agar following EPA

Method 1604 (2002). After 24 hours of incubation at 35 oC, blue colonies that also fluoresced under UV light were characterized as E. coli. The total number of colonies (blue and non-blue) that fluoresced under UV were characterized as TC. As described by EPA Method 1604, countable plates were identified as those with 20-80 E. coli colony forming units (CFUs) per plate

(ideal range) and less than 200 TC CFU per plate. For each experimental condition, several volumes of dilutions were plated to achieve countable CFU ranges. When more than one countable plate was obtained for a given sample, the CFU/100 mL concentrations calculated from those plates were averaged in accordance with ASTM D5465-16 standard practice for determining microbial colony counts (2016). Volumes and dilutions counted for each experimental condition are summarized in Tables 4.2 and 4.3.

To confirm the E. coli characterization, ten E. coli colonies from the plates were selected for confirmation testing according to EPA Microbiological Methods (1978). Briefly, the centers of ten well-isolated blue colonies were inoculated into 10 mL lauryl tryptose broth (tryptose from

Sigma Aldrich, St. Louis, MO; remaining ingredients from VWR, Radner, PA) made according to the USDA Bacteriological Analytical Manual (U.S. Food and Drug Administration 2018) using a sterile pipette tip and incubated for 24 hours at 35 oC. Culture tubes with 10 mL of EC broth

(Thermo Fisher, Lenexa, KS) were subsequently inoculated with growth from lauryl tryptose tubes and incubated for 24 hours at 44.5 oC. Gas production in all tubes confirmed the presence of E.coli. 93

4.3.5. PAA Analysis

In samples that were treated with PAA, PAA concentrations were analyzed at the beginning of disinfection to measure initial PAA consumption, at the end of the 5 min contact time to measure residual PAA, and after quenching to ensure disinfection was effectively terminated. The initial concentrations of PAA in the regrowth experiment reactors were also measured immediately after reactors were established. PAA was measured using a commercially available kit (Peracetic Acid Vacu-vials, CHEMetrics, Midland, VA) according to the manufacturer’s instructions.

4.3.6. Biomolecular Analysis of sul1-Carrying E.coli

The fraction of E.coli carrying the sul1 gene was determined as a measure of the potential for treatment to select for this ARG. During Experiment 1, ten E. coli colonies from each treatment and regrowth condition (simulated CSO effluent, post-settled CSO effluent, settled then disinfected CSO effluent, and regrowth reactors on day 0) were preserved for analysis of the sulfonamide resistance gene su11 by PCR. Colonies were collected from plates with a sterile pipette tip, deposited in 50 µL of sterile molecular biology grade water, heated at 100 oC for 10 min to extract the DNA (Woodman 2008), and stored at -20 oC until analysis. Prior to analysis, tubes with eluted colony DNA were centrifuged and 1 µL of the supernatant was added to a PCR reaction, modified from Woodman (2008). The PCR reaction mixture consisted of 5 µM forward and reverse primers targeting the sul1 gene (Pei et al. 2006), 10 µL 5X buffer, 1.5 µL 25 mM

MgCl2 solution, 1 µL 10 mM dNTPs, and 0.25 µL GoTaq G2 Flexi DNA Polymerase according to the manufacturer’s instructions (Promega, Madison, WI) for a total reaction volume of 25 µL.

The thermal cycler protocol consisted of an initial denaturation step at 95 oC for three minutes, 40 cycles of 95 oC at 15 sec, 55.9 oC annealing for 30 sec, and 72 oC for 30 sec, followed by a final extension step at 72 oC for seven minutes. Presence of the sul1 gene was determined by agarose gel electrophoresis (1.5% agarose). Duplicate PCR results were obtained for at least 10% of 94 samples to confirm results. A no-template control and positive control for the sul1 gene were performed with each PCR reaction for QA/QC.

4.3.7. Statistical Analysis

All statistical analyses were performed in R (http://www.r-project.org) with an α < 0.05 representing significant differences. A Shapiro test was used to test for normality of TSS concentrations and then a paired student's t-test was performed to test for differences before and after removal of settleable particles. Given that the data did not meet the requirements for parametric statistics, a Kruskal-Wallis test was applied to log-normalized E.coli and TC data, pH, conductivity, and COD to test for differences between the different treatments (i.e., simulated

CSO effluent, CSO effluent with settleable particles removed, settled and disinfected CSO effluent, disinfected CSO effluent without settling, and the no treatment control). When differences were detected, a post-hoc pairwise t-test with a Bonferroni correction was performed.

The same approach was used to test for differences in log-removal values from the initial simulated CSO effluent after settling only, disinfection only, settling and disinfection, and no treatment. Finally, the same approach was used to test for differences between PAA concentrations at the beginning of disinfection, at the end of five minutes of disinfection, and after quenching in simulated CSO samples with and without the settleable fraction removed prior to disinfection. For the regrowth study, this approach was used to test for differences in log- normalized E.coli and TC concentrations, pH, conductivity, and COD between the beginning and end of incubation in reactors inoculated with treated CSO, reactors inoculated with untreated

CSO, and reactors with surface water only, as well as between the three different reactor types on day 0 and day 7 of incubation. One-way proportion tests were conducted to test for differences between fractions of E. coli CFU with sul1 ARG during different treatment steps and in different matrices. 95

4.4. Results

4.4.1. Removal of Settleable Particles, Disinfection, Selection

A treatment train consisting of removal of settleable particles followed by PAA disinfection was applied to simulated CSO effluent to determine its impact on water quality.

Removal of settleable particles did not reduce E. coli and TC concentrations compared to untreated simulated CSO effluent (p>0.95) or affect conductivity (p=0.23). TSS decreased by an average of 78±7.4% (p=2.1×10-3) and COD decreased by 49±11% (p=2.3×10-3) after the removal of settleable particles. Following settling, disinfection of the CSO supernatant significantly decreased both target organism CFU concentrations (both p<5.7×10-10; Table 4.1). Increased

COD concentrations (p=1.1×10-2), similar to pre-treatment levels (p=1.0.; Fig. 4.2), were observed after disinfection and quenching. As expected, concentrations of fecal indicators remained consistent in the no-treatment controls compared to the initial concentrations observed in the simulated CSO (p=1).

Table 4.1 Concentrations of cultivable E. coli and TC measured in simulated CSO and after different treatments and log removal of the fecal indicator organisms attributable to treatment by removal of settleable particles (settling), disinfection of the suspended fraction of simulated CSO (supernatant), and disinfection of simulated CSO without settling. Log removal values are based on comparisons to starting concentrations in simulated CSO. The results represent average removals and standard deviations with statistically significant removals in bolded text. 96

Figure 4.2 COD, TSS, and conductivity in the source waste stream (simulated CSO), after removing settleable particles (separated CSO), and after disinfection both with and without prior separation. Results represent the average of replicates [n=5 except for disinfected CSO (no separation) where n=2] and the error bars represent their standard deviation. TSS was not measured after disinfection.

PAA concentrations were measured to determine whether any appreciable decay of the disinfectant occurred during treatment and to calculate CT values. Average peracetic acid concentrations measured at the beginning of the disinfection experiments were 4.8±0.1 mg/L or

97% of the target nominal PAA concentration of 5 mg/L in simulated CSO with prior settling.

PAA concentrations decreased by 8.4±0.7% to an average concentration of 4.4±0.1 mg/L

(p=3.1×10-3) following treatment.

At each step in the treatment process, the proportion of E. coli carrying the sul1 ARG was measured to determine whether treatment selected for culturable organisms carrying this gene.

The removal of settleable particles did not significantly change the fractions of sul1-carrying

CFU in simulated CSO (p=0.39), nor were differences observed between the supernatant post- settling and the post-disinfection samples (p=0.13). However, comparing the untreated simulated

CSO effluent and disinfected CSO supernatant, the proportion sul1-carrying fraction increased

(p=2.3×10-3; Fig. 4.3). 97

Figure 4.3 Fraction of E.coli CFU colonies (n=10) that contained the sul1 ARG at stages of treatment compared to surface water only and surface water spiked with treated effluent on day 0. E.coli CFU were not observed after 7 days of incubation. Results represent the average of experimental replicates and error bars represent their standard deviation (n=3).

Given that significant removal of fecal indicator organisms was not observed by settling, treatment by direct disinfection was performed to determine if not removing settleable particles decreased the disinfection efficiency and/or increased the PAA demand. Directly disinfecting simulated CSO effluent samples resulted in significant decreases of E.coli and TC (p<1.7×10-7;

Table 4.1). Comparing the treatment train and direct disinfection, log-removals were the similar for both E.coli and TC (both p>0.12). As observed for the treatment train, after direct disinfection the conductivity increased (p=8.6×10-4), pH remained stable (p=0.56), and COD was the same as initial concentrations in the simulated CSO effluent (p=1.0). Initial PAA concentrations were similar in simulated CSO without prior separation (4.8±0.09 mg/L) compared to the post settling disinfection samples (p=1.0). After direct disinfection, PAA decreased 11±2.9% to an average concentration of 4.3±0.22 mg/L (p=0.01). No difference in

PAA demand was observed for samples with and without settling prior to disinfection (p=1.0). 98

4.4.2. Regrowth Upon Release to Surface Water

The potential for regrowth of fecal indicators after a seven-day period was assessed for

(1) simulated CSO effluent treated by settling and disinfection spiked into surface water, (2) untreated simulated CSO effluent spiked into surface water, and (3) a surface water control. The disinfected effluent was removed from the disinfection reactors prior to quenching given that the disinfected effluent is expected to be directly discharged into surface water during end-of-pipe treatment for CSOs. On day zero, E. coli and TC concentrations were higher in surface water spiked with untreated simulated CSO effluent compared with the other two reactor types

(p<1.9×10-3; Fig. 4.4). After a week of incubation, E. coli CFU were not detected in either the treated simulated CSO effluent spiked in surface water or surface water control samples. E. coli were detected in reactors with untreated wastewater spiked into surface water after one week.

E.coli concentrations in all three reactor types decreased (p<0.02) over the seven-day incubation period, while TC concentrations remained stable (p>0.09). At the end of incubation, TC concentrations in the reactors spiked with untreated simulated CSO effluent remained significantly higher than the other two reactor types (p<1.9×10-7), while the reactor spiked with treated CSO exhibited TC concentrations comparable to the surface water control (p=1). Thus, evidence for regrowth of fecal indicators in treated CSO effluent was not observed after seven days while fecal indicators remained elevated in the surface water spiked with untreated simulated CSO effluent. No differences were observed in COD, pH, and conductivity in spiked or control reactors during the incubation period (all p>0.05; data not shown).

The proportion of E. coli carrying the sul1 resistance gene was measured in the reactors containing treated CSO effluent spiked into surface water and surface water controls to determine if sul1 gene carrying E. coli were more likely to regrow and/or proliferate after end-of-pipe treatment. The treated CSO effluent had a higher fraction of E.coli CFU with sul1 compared with the receiving water at the beginning of the experiment (p=5.5×10-8), but after the seven-day 99 incubation period, E. coli CFU were no longer detected in either the treated spike reactor or in the surface water control, thus E.coli CFU containing sul1 were not observed.

Figure 4.4 E.coli and total coliform concentrations in reactors with (1) surface water spiked with a 1:10 dilution (v:v) of simulated CSO treated by settling and disinfection (1:10 treated; n=5), (2) surface water spiked with a 1:10 dilution (v:v) of simulated CSO with no treatment (1:10 untreated; n=2) and (3) surface water only (n=5). Error bars represent standard deviation of experimental replicates. 4.5. Discussion

4.5.1. End-of-pipe Treatment Efficiency: Settling

The results of a two-step end-of-pipe treatment train for simulated CSO effluent suggest that after removing the majority of TSS and an insignificant portion of fecal indicators by settling, most of the observed loss of fecal indicator organisms was achieved by disinfection. Similar to the present study, primary settling was not found to have an impact on the concentration of fecal indicators in wastewater treatment plant influent (Turolla et al. 2018), which was used to simulate

CSO effluent in the present study. Previously, the removal of settleable particles was found to help reduce the total load of a marker gene for fecal indicator organisms (BacHum) in surface water impacted by CSOs by 0-61% (Eramo et al. 2017a). This difference in the potential for treatment by settling could be due to there being a greater bacterial load present in the settleable portion of samples collected in the field study (demonstrated by TSS concentrations 2.0-4.8 times 100 higher than average TSS measured before separation of simulated CSO in this study) or the focus on the different targets (cultivable fecal indicators here versus marker genes in the field study).

The qPCR procedure used in the field study measured both intra- and extracellular DNA, which may have different tendencies to associate with particles from cultivable organisms. It is also possible that any significant removal of the target microbial agents by settling in actual CSO is from these organisms present in the storm water as opposed to wastewater component of the overflow effluent. The E.coli and fecal coliform loads attached to settleable particles in storm water were found to comprise 40% of the overall load during storm events (Krometis et al. 2007).

4.5.2. Disinfection Efficiency

The losses of viable fecal indicators in this study were primarily due to disinfection, which achieved 99.1-99.9% inactivation of fecal indicator targets. Neither disinfection efficiency nor PAA demand were improved by settling prior to disinfection compared to simulated CSO samples which did not undergo this treatment prior to disinfection. The efficiency of the PAA disinfection observed here can be compared to other studies investigating PAA disinfection for simulated and actual CSO effluent. Here, disinfection with a 5mg/L PAA dose achieved 2.9±0.14 log-removal of E.coli and 2.3±0.36 log-removal of total coliform after five minutes. In experiments using simulated CSO effluent with 5% wastewater (compared to 40% wastewater in the current study), a higher E. coli log-removal of approximately 3.75 was observed with a 5 mg/L PAA dose for 10 min (Chhetri et al. 2014). In 5% wastewater, Chhetri et al. (2014) observed 32 times lower initial COD than the average COD of simulated CSO measured here, therefore their disinfectant demand for oxidizing organics was likely lower. Comparable results to this study were observed in a pilot-scale study where CSO effluent treated with hydrodynamic separation followed by PAA contact times of 3-6 min and concentrations of 0.56-6.9 mg/L resulted in average log reductions of 2.3 for E. coli and 2.0 for fecal coliform (Bayonne

Municipal Utilities Authority 2017). Considerably more work has been performed investigating 101

PAA disinfection of wastewater effluent, which can be compared to the results of this study with some caveats. While various studies measured higher removals up to 4 or 5 log units for fecal indicators in secondary wastewater effluent with PAA dosages similar to this study (Kitis 2004), this could be expected due to lower COD and organic material in the secondary effluent compared untreated wastewater in this study. Overall, this study and others suggest PAA is highly effective in inactivating E.coli.

Results of this study can also be compared to research applying biomolecular methods for determining CSO disinfection efficiency. The same wastewater dilution, PAA nominal concentration, and contact time was evaluated with biomolecular (viability-based qPCR or vPCR) techniques (Eramo et al. 2017b). We previously observed lower log inactivations of viable cells carrying the ARG sul1 and fecal indicator marker genes BacHum of 0.90 and 0.62-log, respectively. This may be due to differences in the wastewater for the two studies: wastewater quality have been shown to have large impact on PAA performance (Sanchez-Ruiz et al. 1995), as discussed below. Differences in observed removals between vPCR and cultivation-based methodologies after chlorine disinfection has also been suggested to be caused, at least in part, by viable but not cultivable organisms (Varma et al. 2009) or possibly nonviable cells without sufficiently compromised membranes to be differentiated by vPCR. The vPCR method captures all viable cells carrying sul1, not exclusively E.coli, perhaps providing a broader picture of the overall disinfection efficiency of cells carrying this gene.

4.5.3. Water Quality Impacts on Disinfection

An important question with respect to PAA disinfection efficiency in CSO is the relationship, if any, between required dose of disinfectant and pre-treatment CSO effluent water quality. CSO will contain organic and inorganic contaminants from wastewater and from storm water runoff (Gasperi et al. 2008) that could exert a PAA demand and therefore interfere with disinfection of microbial agents (Domínguez Henao et al. 2018). Generally, the presence of TSS 102 and oxidizable organic matter are expected to hinder disinfection with oxidants, but in this study, disinfection efficiency was unchanged by the removal of settleable particles. This is supported by the observation that PAA was much less sensitive to the presence of large particles >10 and 100

µm in wastewater than hypochlorite, but PAA disinfection was still slightly more effective against solids <10 µm compared to larger particles (McFadden et al. 2017). A portion of the large particles that can interfere with PAA disinfection may have been removed during the centrifugation in the current study because the method was designed to remove settleable particles 5-60 µm in diameter (Krometis et al. 2007), but this did not improve our observed disinfection efficiency. PAA effectiveness was also reported to decrease at pH greater than 7.5

(De Luca et al. 2008), but the treated samples had pH values lower than that threshold in this study.

COD may be an important parameter for understanding the initial disinfectant demand and appropriate dosage. Higher PAA demand was observed for higher COD source wastewater and greater cell inactivation by PAA was observed using vPCR with more dilute wastewater

(Eramo et al. 2017b). Results of a recent pilot study indicated that correlations were observed between the reduction of indicator organisms and the PAA dose applied normalized to starting

COD concentration in the waste stream (Bayonne Municipal Utilities Authority 2017). This group reported that a PAA dose of 0.01 mg/L of PAA per mg/L of COD typically resulted in over

3-log reductions for E. coli. Considering the COD averaged 116 and 257 mg/L prior to disinfection with and without prior separation, respectively, in the present study, the required

PAA dose would be 1.16 or 2.57 mg/L (1.9-4.3 times lower than applied here). Thus, according to the pilot study’s regression analysis, a lower PAA dosage could have resulted in similar E.coli removals to those observed here. Given that only 0.41±0.07 mg/L of PAA was consumed during disinfection after settling in our bench scale study, this may be the case. It should be noted, however, that pilot studies samples were real rather than simulated CSO effluent and not reported 103 to have been quenched prior to cultivation in the same way as this study. Also, while an increase in COD following PAA disinfection is expected from the formation of acetic acid, this is only expected to result in a COD of 14 mg/L for a 5 mg/L PAA dose (Kitis 2004). The higher COD increase following disinfection observed in this study may be due to the fact that the samples were quenched prior to COD measurement and the addition of catalase increased the oxygen demand.

4.5.4. ARG Selection by Treatment and Impacts on Surface Water

Evidence for the impact of disinfection on the selection for cells carrying antibiotic resistance genes is of interest to limit to the spread of ARGs. In this study, the proportion of

E.coli carrying sul1was greater after disinfection, but the overall concentration of E.coli CFU per

100 ml carrying sul1 significantly decreased. Further, while disinfection may have selected for sul1-carrying E. coli, these cells did not survive after seven days of incubation in the surface water batch reactors. PAA was previously reported to reduce the presence and also the proportion of uropathogenic E.coli carrying a broad array of ARGs in wastewater as determined by microarray (Biswal et al. 2014). Biswal et al. reported that while the overall percentage of resistance classes decreased, which is promising because ARG carriage is considered of greatest risk in pathogens (Martínez et al. 2015), but the decrease for sulfonamide resistance classes was not significant. Likewise, others found an overall decrease in the amount of culturable ampicillin-resistant E.coli and total heterotrophic bacteria after disinfection (Turolla et al. 2018).

The potential for regrowth of E. coli particularly the sul1-carrying cells was investigated in surface water to better understand the impact of directly releasing treated simulated CSO effluent. In other studies where regrowth in PAA-treated samples was investigated, the treated waste stream was monitored directly, without simulating the release to environmental matrices

(Di Cesare et al. 2016; Lefevre et al. 1992). Regrowth after PAA treatment is of concern given that acetic acid from the breakdown of PAA can serve as an energy source for regrowth of 104 regulated targets (Kitis 2004). In surface water reactors spiked with treated CSO effluent, E. coli was not detected, and TC concentrations decreased to concentrations comparable with surface water control reactors after seven days. The observation of others that concentrations of fecal coliform exhibited a slow increased in seawater spiked with PAA-treated wastewater effluent after seven days, was not observed here. In that study, the increased concentrations were not attributed to regrowth per se, but to the portion of bacteria with decreased enzyme activity that was respiring and not cultivable immediately after PAA treatment (Lazarova et al. 1998).

In contrast to reactors spiked with treated CSO, in surface water reactors spiked with untreated CSO, E.coli persisted after seven days albeit at decreased levels while TC concentrations did not change. This provides a strong argument for treatment of CSO prior to release to environmental surface waters, particularly if the effluent is not sufficiently diluted after release. Various factors may have promoted the removal and prevented the regrowth of E. coli and TC in reactors spiked with treated CSO including the PAA residual, salinity of the receiving water (12 ppt) (Anderson et al. 1979; Conter et al. 2001), aerobic conditions (Conter et al. 2001), and competition with the natural microbial flora in surface water (Flint 1987).

4.6. Conclusions

Bench-scale treatment of simulated CSO effluent achieved up to 3.1±0.14 log-removal of

E.coli and 2.5±0.37 log-removal of TC by a treatment train of removing settleable particles followed by PAA disinfection. While simulated hydrodynamic separation was effective at reducing TSS and COD, it did not significantly reduce the concentrations of E. coli or TC in the effluent. PAA disinfection efficiency was not found to improve by being performed after removing settleable particles, however there may be other water quality benefits for removing settleable particles (e.g., removal of heavy metals in storm water). Disinfection was responsible for inactivating cultivable fecal indicators in the simulated CSO effluent. This work did not account for the storm water chemistry or its microbial loads and it is notable that previous field 105 studies indicated that many microbes or genetic markers are attached to settleable particles during wet weather flows (Eramo et al. 2017a; Krometis et al. 2007) that could be removed by a properly sized hydrodynamic separator. Direct release of the disinfected effluent without quenching to well-mixed estuarine water did not result in regrowth of E. coli, sul1-carrying E. coli, or TC. The

E.coli and TC concentrations did not increase after release of untreated CSO to surface water but persisted over a seven-day period at levels greater than those observed in surface water spiked with treated CSO effluent. While this treatment train is promising for improving some water quality parameters, it is also important to acknowledge that CSO effluent disinfected at the end- of-pipe would likely release residual PAA into the receiving water. Thus, while PAA is expected to have a short half-life in water and has not been reported to create chlorinated disinfection by- products, the impact of releasing treated effluent with a PAA residual on the microbial and other life in the receiving water should be considered.

Acknowledgements

Thanks to our utility partner for providing access to influent samples. Funding for this project was provided by a Mark B. Bain Fellowship from the Hudson River Foundation to AE and a grant from the National Science Foundation (#1510461). Additional funding support was provided by an Eagleton Fellowship Program to AE. Thanks also to William Morales Medina for his assistance in the lab in support of this project. 106

4.7. Supplemental Information

Figure 4.5 Schematic of treatment and regrowth experimental conditions.

Table 4.2 Dilutions and volumes averaged for each treatment condition. At least two plates were tested for each condition and countable plates are included in the table. Countable plates contained 20-80 E.coli CFU and <200 TC CFU. The numbers outside parentheses represent the sample dilutions and the numbers inside parentheses represent the corresponding volumes filtered of the dilutions (milliliters). 107

Table 4.3 Dilutions and volumes used for each regrowth condition. At least two plates were tested for each condition and countable plates are included in the table. Countable plates contained 20-80 E.coli CFU and <200 TC CFU unless indicated. When all plates for a given sample resulted in E. coli CFU less than the target minimum of 20, the most nearly acceptable count was reported, in accordance with ASTM D5465-16 (ASTM International 2016). These plates are identified by the notation. The numbers outside parentheses represent the sample dilutions and the numbers inside parentheses represent the corresponding volumes filtered of the dilutions (milliliters). 108

5. CHAPTER 5: Viability-based quantification of antibiotic resistance genes and human

fecal markers in wastewater effluent and receiving waters

In review for Science of the Total Environment Authors: Alessia Eramo, William R. Morales Medina, Nicole L. Fahrenfeld

5.1. Abstract

Antibiotic resistance is a public health issue with links to environmental sources of

antibiotic resistance genes (ARGs). ARGs from nonviable sources may pose a hazard

given the potential for transformation whereas ARGs in viable sources may proliferate

during host growth or conjugation. In this study, ARGs in the effluent from three

municipal wastewater treatment plants (WWTPs) and the receiving surface waters were

investigated using a viability-based qPCR technique (vPCR) with propidium monoazide

(PMA). ARGs sul1, tet(G), and blaTEM, fecal indicator marker BacHum, and 16S rRNA

gene copies were found to be significantly lower in viable-cells than in total

concentrations for WWTP effluent. Viable-cell and total gene copy concentrations were

similar in downstream samples except for tet(G). Differences with respect to season in

the prevalence of nonviable ARGs in surface water or WWTP effluent were not

observed. The results of this study indicate that qPCR may overestimate viable-cell

ARGs and fecal indicator genes in WWTP effluent but not necessarily in the surface

water >1.8km downstream.

5.2. Introduction

Antibiotic resistance is a major modern public health issue. In the United States at

least two million illnesses and 23,000 deaths are caused by antibiotic-resistant bacteria

annually (Centers for Disease Control and Prevention 2013). Given that antibiotic 109 resistant infections in humans have been linked to environmental sources of antibiotic resistance (Casey et al. 2013), fully addressing the risk posed by ARGs requires understanding environmental hotspots and determining the relative importance of ARGs from viable and nonviable sources for different mechanisms of proliferation.

Wastewater treatment plants (WWTPs) have been described as hotspots for the spread of antibiotic resistant bacteria and ARGs in the urban environment (Rizzo et al.

2013) and correlations have been observed with respect to land use and ARG concentrations (Pruden et al. 2012). The oxidizing disinfectants applied in secondary treatment at WWTPs, such as chlorine, can inactivate antibiotic resistant bacteria but do not necessarily destroy ARGs (Fahrenfeld et al. 2013; Yuan et al. 2015). This release of extracellular ARGs and nonviable cells containing ARGs to the environment from activities such as wastewater treatment presents a potential risk because of the possibility for these genes to be incorporated into competent host genomes by transformation

(Chang et al. 2017; Mao et al. 2014). Understanding the relative proportion of ARGs that are present in viable versus nonviable cells or as extracellular DNA (eDNA) can provide insight into their availability for different mechanisms of ARG proliferation (i.e., via growth and/or conjugation versus transformation) in a given environmental compartment.

eDNA, including ARG-carrying plasmids, persists in the environment after the host bacteria dies, and is therefore potentially available for transformation (Mao et al. 2014).

Likewise, the total DNA of fecal indicators has been shown to persist long after the decay of culturable E. coli and E. faecalis, which occurs within days in surface water

(Gutierrez-Cacciabue et al. 2016). In contrast to ARGs, 16S rRNA genes from fecal indicators are not generally considered a hazard if the host cell is not viable. The 110 viability qPCR (vPCR) method used in this study reduces the qPCR signal from DNA originating from nonviable sources in environmental samples (Bae and Wuertz 2009;

Eramo et al. 2017b; Nocker et al. 2007b). This is achieved using propidium monoazide

(PMA), a dye that irreversibly intercalates extracellular DNA or DNA in cells with compromised membranes, thereby preventing amplification by qPCR. Recent applications of this method for targeting ARGs include (1) determining disinfection efficiency of combined sewer overflow effluent with peracetic acid (PAA) (Eramo et al.

2017b) and (2) measuring microbial decay rates in aerobic and anaerobic sludge

(Mantilla-Calderon 2017).

An extraction procedure (Corinaldesi et al. 2005) has also been used to differentiate

ARGs in extracellular DNA (eDNA) and intracellular DNA (iDNA) in the sludge of livestock waste management operations (Zhang et al. 2013) and surface water (Mao et al.

2014). To the authors’ knowledge, the vPCR and extraction methods have not been compared on paired samples for ARGs, but have the potential to provide slightly different information: extraction is dependent on physically separating iDNA and eDNA, vPCR targets eDNA and the fraction of iDNA in cells with compromised membranes. While there is no consensus on which methods to apply to establish the risk posed by environmental ARGs and antibiotic resistance (Vikesland et al. 2017), the published literature applying either extraction or vPCR provides a basis for defining which environments have or do not have significant concentrations or fractions of extracellular/nonviable cell ARGs [i.e., livestock waste management structures (Zhang et al. 2013), surface water sediments (Mao et al. 2014), PAA-treated combined sewer overflow (Eramo et al. 2017b)]. But, given that these studies applied the different 111 methods across different environments, questions remain about the relative abundance of extracellular and nonviable sources of ARGs in environmental hotspots, including surface waters receiving treated wastewater effluent.

The objective of this study was to investigate the relative proportion of ARGs in cells with intact membranes to total ARGs observed in WWTP effluent and receiving waters towards understanding their availability for proliferation via different mechanisms. The partitioning of ARGs in WWTP effluent and receiving waters was evaluated by differentiating viable-cell ARGs by vPCR from total gene copies by qPCR, the latter of which has been shown to serve as a conservative proxy for ARG transformation (Chang et al. 2017). Other bacterial genes (BacHum, a human fecal marker, and 16S rRNA genes) were evaluated because these target genes are not considered functional genes and therefore may have different fates with respect to transformation compared to ARGs.

The sampling design also allowed for determination of whether the vPCR results varied by season or WWTP. Overall, the results presented have implications for the understanding the relative pool of ARGs available for horizontal gene transfer (HGT) via transformation in different aqueous environments.

5.3. Materials and Methods

5.3.1. Field Sampling and Water Quality Analysis

Samples were collected on three days in the summer (7/30 or 8/5, 8/24, and

8/31/2015) and three days in the winter (1/25 or 3/2, 3/7, and 3/14/2016) from the treated effluent of three WWTPs (WWTP-A, B, and C in the US mid-Atlantic region) and from surface water upstream and downstream of the WWTPs’ discharge locations. WWTP flow rates, upstream, and downstream sampling distances are summarized in 112

Supplementary Table 5.2. Surface water and treatment plant effluent samples were collected on the same day for as many sampling events as possible. Variation from the paired sampling design was due to precipitation that occurred before sampling could be completed on 7/30/2015 for WWTP-A downstream samples, a plant issue on 8/24/2015 preventing effluent sampling of WWTP-B, and inaccessibility to field sites due to snow and ice on 1/25/2016. Composite effluent samples (500-mL each) were collected at 8 am, 10 am, and 12 pm. All samples were collected during baseflow conditions in sterile

500-mL Nalgene bottles, transported to the lab on ice and stored at 4 oC prior to processing. WWTP-A and WWTP-B serve municipalities with major hospitals and/or medical facilities within the sewershed, while WWTP-C does not have major hospitals and/or medical facilities within the sewershed. Surface water sampling locations were chosen based on availability of public access. Further details including sampling times for the surface water samples are presented in Supplementary Table 5.3.

pH and conductivity of the downstream water samples were measured in the field with an Orion Star A329 multimeter (Thermo Scientific, Waltham, MA). Downstream and composite effluent samples were analyzed for total suspended solids (TSS) using

Environmental Sciences Section Method 340.2 (Wisconsin State Lab of Hygiene 1993).

Field blanks consisting of autoclaved deionized water were carried in the field on each sampling date for QA/QC.

5.3.2. Cell Collection, Viability Cross-Linking and Biomolecular Analysis

Bacteria from 210 mL aliquots of effluent composite and downstream samples were harvested by sequential centrifugation at 4,000×g for 15 min and removing the 113 supernatant except for the last 10 mL, as demonstrated by others (Banihashemi et al.

2012; Contreras et al. 2011; Luo et al. 2010; Nocker et al. 2007a). Aliquots of centrifuge-concentrated cells were either treated with 50 µM propidium monoazide

(PMA) or preserved for DNA extraction. The former allowed for quantification of the

“viable” cell fraction of DNA in the samples, defined as DNA in cells with intact membranes, and the latter for quantification of total DNA in the samples. A PMA concentration of 50 µM was selected because it was sufficient to suppress qPCR signals from heat-inactivated cells without exhibiting toxic effects. Additional preliminary testing showed that 20 µM PMA was not sufficient (data not shown). This PMA concentration has been used by others for cells from wastewater, estuarine benthic mud, and marine sediment (Nocker et al. 2007a). PMA-treated samples were incubated in the dark for 5 min and photoactivated for 15 min using a PMA-LiteTM LED Photolysis

Device (Biotium, Fremont, CA) to facilitate cross linking of the dye with DNA. PMA methods were adapted from Nocker et al. (2007a) and Nocker et al. (2010). Upstream and downstream samples (~200-700 mL) were also filter concentrated (0.22 µm nitrocellulose) and preserved for DNA extraction. Samples were stored at -20 °C until

DNA extraction.

DNA was extracted from centrifuge-concentrated cells (500 µL) with and without

PMA treatment and filter concentrated cells using a FastDNA® SPIN Kit for Soil (MP

Biomedicals, Solon, OH). Concentrations of select ARGs [sul1 (Pei et al. 2006), tet(G)

(Aminov et al. 2002), blaTEM (Narciso-da-Rocha et al. 2014)], fecal indicator marker

BacHum (Kildare et al. 2007) and 16S rRNA gene copies for total bacterial population

(Muyzer et al. 1993)] were quantified by qPCR. A standard SybrGreen (5 µL SsoFast 114

EvaGreen, BioRad, Hercules, CA) chemistry with 0.4 µM forward and reverse primers, and 1 µL diluted DNA extract (1:50 dilution to reduce inhibition) in a 10 µL total reaction volume was used for all targets except BacHum. Hydrolysis probe chemistry (5

μL SsoFastProbes SuperMix, BioRad, Hercules, CA) for BacHum consisted of 0.07 μM probe, 0.22 μM forward and reverse primers, and 1 μL diluted DNA extract (1:50).

Thermocycler (BioRad CFX96 Touch, Hercules, CA) conditions are summarized in

Table S3. Samples and standards were analyzed in triplicate (technical replicates). A seven-point calibration curve and negative control were included in each run. Average amplification efficiency across all assays was 89±15 (mean ± SD) and average R2 was

0.99±0.01. Standards were generated from 10-fold serial dilutions of cloned and sequenced genes originating from environmental samples. A melt curve and/or gel electrophoresis were used to verify the specificity of qPCR products. Based on the lowest standard on the curve and factoring in dilution implementing during sample processing the limits of quantification were 4.1-4.9 log gene copies for the genes tested.

5.3.3. Matrix Spike Study

To explore whether environmental matrix effects were interfering with the PMA performance in the downstream samples, centrifuge-concentrated river water was spiked with heat-inactivated E. coli cells (100ºC, 10 min). E. coli were plated on LB agar prior to the heat inactivation to confirm viability and after heat treatment to confirm inactivation. Samples were split and treated with PMA (0, 50, or 100 µM) or preserved for DNA extraction followed by vPCR or qPCR for 16S rRNA gene copies, as described above. 115

5.3.4. Statistical Analysis

All statistical analyses were performed in R (http://www.r-project.org). Except for blaTEM and 16S rRNA, qPCR data were Box-Cox transformed prior to analysis given that ≥20% samples had ARGs below detection. A paired Wilcoxon rank sum test was used to determine differences (α< 0.05) in absolute ARG (copies/mL) or 16S rRNA gene normalized ARG concentrations between the viable-cell and total gene copies in samples and between effluent and downstream viable-cell to total ratios for all WWTPs. A

Kruskal Wallis test was used to determine differences between effluent and downstream concentrations and seasonal differences in viable or total gene copies for a given WWTP.

A post-hoc pairwise t-test with a Bonferroni correction for multiple comparison was then applied. All sample results were included in the analyses including zeros and outliers.

The same approach was used to compare pH, TSS, and conductivity data. A one-way test for equal means was performed on data from the matrix spike experiment, followed by a post-hoc pairwise t-test with a Bonferroni correction for multiple comparison.

5.4. Results

5.4.1. Total and Viable-Cell Quantification of ARGs, BacHum Fecal Indicator

Marker, and 16S rRNA in WWTP Effluent and Downstream Samples

A viability-based approach was used to quantify viable-cell gene copies of ARGs sul1, tet(G), and blaTEM, human fecal marker BacHum, and 16S rRNA genes in WWTP effluent and surface water to better understand the potential availability of these genes for growth/conjugation versus transformation. Total concentrations of the target genes were quantified by qPCR and concentrations of viable-cell copies were quantified by first pretreating the samples with PMA, a membrane impermeable dye. There were 116 significantly lower concentrations of all genes tested in the viable-cell fraction compared to the total measurements in the paired WWTP effluent samples (all p<0.02), as would be expected for WWTP effluent disinfected with chlorine (Fig. 5.1). ARGs sul1 and tet(G) were detected in both the viable-cell and total fractions of 76% of effluent samples

(n=17), where the average ratio of viable-cell to total concentrations (vPCR/qPCR) was

0.43 ± 0.52 for sul1 and 0.46 ± 0.29 for tet(G) (Fig. 5.2). In the other 24% of samples,

ARGs were either not detected in the viable-cell fractions but present in the total ARG measurement or were not detected in either measurement in treated effluent (Fig. 5.1a, b).

The ARG blaTEM was detected in both the viable-cell and total measurement in 100% of effluent samples at a ratio of viable-cell to total ARGs of 0.69 ± 0.46 (Fig. 5.1c).

In addition to the ARG vPCR and qPCR, two other genes were quantified with both methods: BacHum human fecal marker and 16S rRNA genes. These genes would presumably be less likely to be transformed and maintained in the host cell’s genome compared to ARGs which serve as functional genes. In contrast to the ARGs, the

BacHum marker was detected in the viable-cell fraction of only one effluent sample, where a ratio of 0.51 viable to total gene copies per mL (vPCR/qPCR) was observed.

BacHum was below detection in 53% of effluent samples and observed in the total measurement but not viable cells in the remaining 41% of samples (Fig. 5.1d). The average ratio of 0.24 viable-cell to total 16S rRNA gene copies in effluent samples suggests, as expected, that most bacteria are inactivated with a membrane sufficiently damaged to allow PMA dye entry prior to leaving the WWTPs sampled in this study (Fig.

5.1e). 117

Figure 5.1 Concentrations of ARGs (a.) sul1, (b.) tet(G), and (c.) blaTEM, (d.) fecal indicator marker BacHum, and (e.) 16S rRNA present overall (total) and in viable cells only (viable) measured in effluent and downstream samples from wastewater treatment plant A, B, and C during summer and winter seasons. Boxes represent upper and lower quartiles, whiskers extend to high and low data points excluding outliers, and dots indicate outliers. 118

Figure 5.2 Ratios of viable-cell gene concentrations measured by vPCR to total gene concentrations measured by qPCR in effluent and downstream samples.

Towards understanding the fate of environmental ARGs, vPCR and qPCR were next applied in the receiving surface water. In contrast to the WWTP effluent, viable-cell concentrations of sul1, blaTEM, BacHum, and 16S rRNA gene copies via vPCR were not different from the total measurement of these genes via qPCR in the downstream samples

(all p>0.18). sul1, blaTEM, BacHum, and 16S rRNA targets were detected in both the viable-cell and total gene copy concentrations in 33%, 83%, 12%, and 100% of downstream samples (n=18), respectively. sul1 was below detection in both the total and viable-cell fraction for 50% of samples and observed only in total cells in the remaining three (17%) samples. In the 3/18 samples where viable-cell blaTEM was not observed, the gene was detected in the total measurement via qPCR. 119

tet(G) was the only gene tested where viable-cell gene copies were significantly lower than total gene copies in the downstream samples (p<4.7×10-3). In 76% of downstream samples where tet(G) was detected in both the viable-cell fraction and the total measurement via qPCR, the ratio of viable-cell to total gene copies was 0.71 ± 0.35.

In the remaining samples, tet(G) was below detection in both the viable-cell fractions and the total measurement or only detected in the total measurement.

5.4.2. Comparison of WWTP Effluent and Downstream Gene Concentrations

Concentrations of viable-cell and total gene copies were also compared between effluent and downstream sampling locations to further understand the fate of ARGs and fecal markers released from WWTPs. Effluent concentrations of total sul1 were higher than the receiving water where the total sul1 gene copies were below detection limits in many (9/17) downstream samples (p=0.02; Fig. 5.1a). When above detection limits, total sul1 gene copies downstream were on average 98 ± 6% less than the total sul1 concentration detected in the WWTP effluent. Both the 16S rRNA normalized concentrations of the viable-cell and total sul1 gene copies (viable-cell sul1/viable-cell

16S rRNA genes and total sul1/total 16S rRNA genes), were higher in WWTP effluent samples compared to the downstream samples (all p<1.1 ×10-3; Supplementary Fig. 5.4a).

The concentrations of the other genes measured in the viable-cell fractions via vPCR and total measurement via qPCR were similar in effluent and downstream samples, both on a volume concentration (gene copies/mL) and a 16S rRNA gene copy normalized basis (all p>0.06). For most sampling events, if gene targets were detected in the viable fraction of cells downstream, they were also detected in viable cell fraction in the corresponding effluent sample. However, on the first winter sampling event, sul1 was detected in the 120 viable-cell fraction downstream but not in the viable-cell fraction of effluent at WWTP-

C, and tet(G) at WWTP-A. (Effluent and downstream samples were not paired for this event given sampling access issues.) Detection in viable-cells downstream but not in effluent also occurred on the second day of winter sampling for sul1 and tet(G) at

WWTP-C and WWTP-B, respectively. BacHum exhibited this phenomenon on the last summer day of sampling at WWTP-B.

Comparing the ratio of viable-cell to total gene copies observed in WWTP effluent and the downstream surface water, the ratio of viable-cell to total gene concentrations (vPCR/qPCR) was greater in surface water samples compared to effluent for sul1, tet(G), and 16S (all p<0.05; Fig. 5.2). The ratio of viable-cell to total concentrations was <0.5 for these genes after chlorination, but this impact of wastewater disinfection on viable ARG-carrying microbes was no longer observed in downstream surface water because there was generally no difference between the viable-cell and total concentrations. Similar viable-cell to total concentration ratios were observed for both blaTEM and BacHum in effluent and downstream samples (p>0.08). In contrast to the other genes, blaTEM was the most frequently detected ARG in viable cells and had the highest ratio of viable-cell to total concentrations leaving the WWTP, while BacHum was detected infrequently in both effluent and downstream samples. These results indicate some gene-to-gene differences in ARG fate.

5.4.3. Matrix Spike Experiment

Given that differences between vPCR and qPCR results were observed less frequently in downstream samples, a separate experiment was performed to determine if matrix effects were interfering with the PMA performance in the downstream samples. 121

PMA was found to successfully suppress 16S rRNA qPCR signals in centrifuge- concentrated river water cells spiked with inactivated (heat-treated) cell cultures, which represented free DNA and cells with compromised membranes (Fig. 5.3). These samples exhibited the same 16S rRNA gene copy concentrations as concentrated water only controls (p=0.30). Concentrated water samples spiked with the same culture but not heat- treated served as a second control. As would be expected, samples spiked with viable cultures were greater than river water (p=5.9×10-3) because PMA does not interfere with qPCR for cells with intact membranes. Plates with the live spike (not heat-treated) were too numerous to count but plates with heat-inactivated cultures had no growth, confirming that heat treatment inactivated the E. coli culture.

Figure 5.3 Concentrations of 16S rRNA in centrifuge-concentrated E. coli culture at exponential phase, river samples only (water pellet), centrifuge-concentrated river samples with a live cell culture spike (water pellet + live), and centrifuge-concentrated river samples with an inactivated (heat-treated) cell culture spike (water pellet + dead). Samples were analyzed by qPCR (0 µM PMA) and vPCR (50 or 100 µM PMA), which represent total and viable-cell concentrations, respectively. 122

5.4.4. WWTP and Seasonal Impacts

No differences in total or viable-cell concentrations between the two seasons sampled were observed in effluent and downstream samples for the genes tested (all p>0.06). Downstream surface water pH, conductivity and TSS and effluent TSS were also consistent between seasons (p>0.45; Table 5.1). Thus, seasonality did not impact concentrations of genes or water quality measurements observed in this study.

Table 5.1 Average pH, conductivity and TSS (+/- standard deviation) collected from upstream, influent, effluent, and downstream samples from three wastewater treatment plants during the summer and winter seasons (n=2* or 3).

Comparisons were made among the three WWTPs to determine if there were plant-to-plant differences. While all WWTP have similar treatment trains, WWTP-A and

B have larger influent flow rates (>100 MGD) compared to WWTP-C (<100 MGD;

Supplementary Table 5.2). In addition, WWTP-A and WWTP-B serve municipalities with combined collection systems while WWTP-C receives water from a separate sanitary collection system. Stormwater did not impact the sampling because all samples were collected during baseflow conditions. There were no differences observed in the effluent concentrations of the viable-cell via vPCR or the total measurement via qPCR of 123

tet(G) and blaTEM between the three plants sampled (both p>0.59). In contrast, WWTP-A and WWTP-B had higher concentrations of viable-cell and total sul1 compared to

WWTP-C (all p<0.02). Additionally, WWTP-A had higher BacHum marker and 16S rRNA gene copy total concentrations than WWTP-C (p<4.0×10-3). Higher concentrations of TSS were also observed at WWTP-A compared to WWTP-C (p=0.02), but no other differences between plants were observed (p>0.27; Table 5.2). The select inter-plant effluent differences observed were not preserved downstream: no differences were observed between the three surface water bodies sampled for any of the genes tested

(all p>0.27).

5.5. Discussion

5.5.1. Total and Viable-Cell Gene Concentrations in WWTP Effluent and

Downstream Samples

A viability-based qPCR method was applied demonstrating that the ARGs and other target genes were significantly lower in the viable-cell fraction of effluent samples compared to total gene concentrations. This was expected given that all three WWTPs sampled in this study disinfected by chlorine, an oxidant whose disinfectant mechanisms includes the disruption of cell membranes by reactive oxygen (McFadden et al. 2017).

Cultivation-based methods have also demonstrated that some viable antibiotic resistant bacteria are released from wastewater treatment to surface water (Huang et al. 2012; Li et al. 2015). While the vPCR method used in this study reduces signal from extracellular

DNA and cells with compromised membranes, this method may be expected to overestimate the viable fraction of cells given that cells may be inactivated without having damaged membranes, which may explain the observations of vPCR/qPCR values 124 above zero. The same viability-based method applied here was recently used to demonstrate the disinfection efficiency of PAA on ARG-carrying cells in simulated combined sewer overflow (Eramo et al. 2017b). With PAA disinfection, the fractions of viable-cell sul1 and tet(G) concentrations compared to total sul1 and tet(G) concentrations reached as low as 5.3×10-3 ± 3.5×10-3, or less than 1% (Eramo et al.

2017b), while in this study, the percentage was slightly less than 50%. Differences in disinfectant, concentration, contact time, and the presence of organic matter can affect the disinfection efficiency and may explain the differences in these observations.

The lack of observed differences between vPCR and qPCR results for sul1 and blaTEM in the water column downstream of WWTP is consistent with other reports of

ARGs in river water samples where the majority of DNA was intracellular DNA (iDNA) rather than eDNA (Mao et al. 2014). Those researchers accounted for the loss of eDNA in the water column through observations that eDNA was much higher in river sediment samples, suggesting deposition was a loss mechanism for extracellular ARGs in the water column (LaPara et al. 2015; Mao et al. 2014). Sediment samples were not collected in the present study to confirm this result, but it is worth noting that vPCR may not perform well at differentiating nonviable sources in sediment due to matrix interferences (Kim et al. 2014). In addition to deposition, the lack of differences for viable-cell and total gene copies in the water column may also be explained by degradation (LaPara et al. 2015), dilution (LaPara et al. 2015), cellular repair (Huang et al. 2011), and/or transformation of extracellular gene copies or those associated with nonviable cells in WWTP effluent

(Mao et al. 2014). In the Mississippi River, it was reported that the flow rate relative to

WWTP effluent flow resulted in a large dilution effect on ARGs, minimizing the 125 measurable impact on receiving water (LaPara et al. 2015). In this study, no differences were observed between total ARG concentrations in surface water samples collected upstream and downstream of the three WWTPs during both the summer and winter seasons (p=1.0; Fig. 5.5). ARGs in these samples were measured by filtering of water samples through 0.22 µM nitrocellulose filters followed DNA extraction from filters, rather than collection of cells by centrifugation applied for the total and viable-cell concentrations. The results of upstream and downstream sampling support the observation that ARG concentrations present in WWTP effluents do not significantly impact downstream concentrations and indicate that dilution and/or other loss/decay mechanisms could be responsible for this phenomenon. The lack of gages in the vicinity of sampling locations made it difficult to estimate a dilution rate for the WWTP effluent in this study. Another potentially confounding factor, tide direction, did not result in differences in ARG concentrations (Table 5.3).

BacHum human fecal marker and 16S rRNA genes were also measured in effluent and downstream samples to provide information on the fate of non-functional genes to compare to the ARG data. Their fate was expected to be different than that of

ARGs given than ARGs can provide a selective advantage and therefore may be more likely to be transformed. BacHum was observed in only one viable-cell effluent sample.

This observation is consistent with other studies where viable fecal indicator organisms were not detected in secondary effluent (Li et al. 2014) or were detected at very low concentrations by vPCR (Varma et al. 2009). Viability qPCR methods have been used to target pathogens and fecal indicators with positive correlations observed between vPCR and cultivation methods (Li et al. 2014; Varma et al. 2009). Thus, the utility of this fecal 126 marker gene as an indicator gene for ARGs is not clear, consistent with our previous observations during combined sewer overflows (Eramo et al. 2017a). The BacHum marker was also observed infrequently in the water column downstream from WWTPs.

Fecal indicator genes not detected in viable cells may have settled, degraded, or, most likely, been too dilute for detection. The presence of BacHum suggests human fecal pollution, which is most likely to originate from WWTPs during dry weather sampling of urban environments without septic tanks or from cross amplification with select other non-human fecal sources (Kildare et al. 2007). As expected, the low ratio of viable-cell to total BacHum marker downstream does not provide evidence of growth or HGT.

To explore whether environmental matrix effects were interfering with the PMA performance in the downstream samples, concentrated river water was spiked with heat- inactivated cells and analyzed by vPCR and qPCR. PMA was found to successfully suppress 16S rRNA qPCR signals in centrifuge-concentrated river water cells spiked with inactivated (heat-treated) cell cultures, which represented free DNA and cells with compromised membranes. This result is attributed to PMA suppression of DNA amplification from the heat-treated culture. During PMA treatment, the light exposure inactivates any excess PMA that has not entered cells, so that it cannot affect DNA when it is released during DNA extraction (Nocker et al. 2010). Also, upon light treatment, the free DNA that binds to PMA is rendered insoluble and lost during subsequent DNA extraction (Nocker et al. 2006). This result indicates that matrix interference does not explain the different results observed in WWTP effluent compared to surface water. The lower concentrations of viable-cell tet(G) concentrations compared to total concentrations 127 in this study also supports that differences in viability are detectable in surface water samples.

5.5.2. Comparison of WWTP Effluent and Receiving Water

While the volume-based concentrations of viable-cell sul1 gene copies (gene copies/mL) in the WWTP effluent and downstream samples were similar, sul1 gene carrying cells comprised a smaller fraction of the viable-cell microbial community downstream. The sul1 viable-cell concentration in WWTP effluent was similar to both the viable-cell and the total concentration of this gene in downstream samples. This suggests that the nonviable cells carrying sul1 and extracellular sul1 genes were, on net, not remaining suspended in the water column. Community analysis was not performed in this study therefore the shifts in bacterial community composition of ARG hosts between

WWTP effluent and receiving water that could help explain the fate of the investigated genes are not known.

5.5.3. vPCR Versus eDNA Extraction Methods

Other methods have been used to investigate nonviable sources of ARGs within aqueous samples by qPCR. The eDNA/iDNA extraction method previously referenced

(Corinaldesi et al. 2005; Mao et al. 2014; Zhang et al. 2013) allows for simultaneous separation of the fractions. However, Liu et al. (2018) noted that eDNA concentrations were too low harvest in treated water and utilized an adsorption-elution method. This method does not allow for simultaneous comparison of the two DNA fractions but allows for observation over time or treatments. The benefits of the vPCR method used in this study are that it allows for paired analysis of viable-cell and total gene concentrations and that it is simple to apply directly to collected cells in ~15min. eDNA/iDNA extractions 128 procedures require multiple washing and pelleting steps, while PMA pretreatment only requires collecting bacterial cells and particles, here accomplished by centrifuging to generate a pellet. However, the vPCR method may not perform well in sediment (Kim et al. 2014) which represents a major limitation compared to the extraction methods which were developed for sediment. Even in aqueous samples it is recommended that the PMA dye used for vPCR should be titrated to determine the correct concentration: if too much is added it can be toxic to cells resulting in artificially low vPCR results, and if too little is added one would expect results artificially similar to the qPCR. In this study we tested

20-100 µM PMA and found 20 µM was insufficient and that results were similar for 50 or 100 µM. In this study select samples, mostly in surface water, had vPCR to qPCR ratios (copies/mL : copies/mL) of greater than two up to ten (Fig. 5.2), which was not expected. These anomalies may be due to the variability in the environmental split samples for vPCR and qPCR. As stated above, the eDNA extraction and vPCR methods have not been previously applied to paired samples, to the authors knowledge, but could provide further insight into the fate of eDNA and DNA from nonviable cells.

5.5.4. Implications for Understanding the Mechanisms Driving ARG Fate

A critical question with respect to nonviable sources of ARGs measured in this study is determining the potential pool of DNA available for HGT via transformation in different environments. A risk assessment requires accounting for the relative importance of different mechanisms of HGT (Vikesland et al. 2017). Once these environments are identified, kinetic investigations can be performed to determine the rates of HGT. At the downstream sampling locations in this study located more than 1.8 kilometers away from the WWTPs, qPCR and vPCR results were similar for sul1, blaTEM, BacHum marker, and 129

16S rRNA genes indicating nonviable DNA was not present in high abundance, unlike in the WWTP effluent, and thus no longer available in high proportion for transformation into viable cells in the water column. For tet(G), the nonviable concentrations were a significant proportion of the ARGs in the water column. These gene-to-gene differences may be driven by differences in the fate of the host cells, although, sul1, tet(G), and blaTEM all encode for resistance to antibiotics used to control Gram negative bacteria.

Microbial community analysis was not performed in this study therefore the shifts in bacterial community composition of ARG hosts between WWTP effluent and receiving water that could also explain the fate of the genes investigated are not known. Testing for a broader range of ARGs may also provide further insight if this approach were repeated with high throughput qPCR arrays targeting a broader range of ARGs.

5.6. Conclusions

Overall results from this study confirm that qPCR is a conservative proxy for analyzing viable-cell ARGs in WWTP effluent. qPCR may represent a reasonably accurate measure of sul1 and blaTEM concentrations 1.8-3.9 km from these WWTP outfalls and overestimate viable-cell concentrations of tet(G). How close to the WWTP outfalls this observation would be consistent is not clear but likely is a function of dilution factor and the background prevalence of these genes in the receiving waterbody.

Therefore, it would be advisable to sample the water column closer to the WWTP outfalls in future research to look for evidence of significant fractions of eDNA and nonviable cells with sul1 and blaTEM in the water column that would be available for transformation

(the lack of public access points prevented that in this study) or to focus on bed 130 sediments, which were found by others to have significant amounts of eDNA (Mao et al.

2014).

Acknowledgements

Laboratory assistance was provided by Hannah Delos Reyes and Sophia Blanc. Thanks to our utility partners for providing access to treatment plant samples. Funding for this project was largely provided by a grant from the New Jersey Water Resources Research

Institute, with additional support provided by the National Science Foundation

(#1510461), a Mark B. Bain Fellowship from the Hudson River Foundation, and NIH

Bridges to the Doctorate Program (R25). 131

5.6.1. Supplemental Information

Table 5.2 Flow rate categories for three activated sludge wastewater treatment plants disinfecting with chlorine sampled and approximate distances between their discharge points and downstream sampling locations.

Table 5.3 Field sampling details for wastewater treatment plants A, B, and C downstream and upstream samples collected during the summer and winter seasons. 132

Table 5.4 Primers, annealing temperatures, and amplicon lengths.

Amplicon Ta Gene Primer sequence length Source (°C) (bp) CGCACCGGAAACATCGCTGCAC Pei et al. sul1 65 163 TGAAGTTCCGCCGCAAGGCTCG (2006) GCAGAGCAGGTCGCTGG Aminov et tet(G) 68 134 CCYGCAAGAGAAGCCAGAAG al. (2002) TTCCTGTTTTTGCTCACCCAG Xi et al. bla 60 247 TEM CTCAAGGATCTTACCGCTGTTG (2009) TGA GTT CAC ATG TCC GCA TGA CGT TAC CCC GCC TAC TAT CTA ATG Kildare et BacHum 60 81 /56-FAM/TCC GGT AGA CGA TGG GGA TGC GTT al. (2007) /36-TAMSp/ 16S CCTACGGGAGGCAGCAG Muyzer et 65 202 rRNA ATTACCGCGGCTGCTGG al. (1993)

Figure 5.4 Concentrations of ARGs (a.) sul1, (b.) tet(G), and (c.) blaTEM, and (d.) fecal indicator marker BacHum normalized to 16S rRNA concentrations present overall (total) and in viable cells only (viable) measured in effluent (n=17) and downstream (n=18) samples from 3 wastewater treatment plants during summer and winter seasons. Boxes represent upper and lower quartiles, whiskers extend to high and low data points excluding outliers, and dots indicate outliers. 133

Figure 5.5 Concentrations of ARGs (a.) sul1 and (b.) tet(G) measured in upstream and downstream samples from 3 wastewater treatment plants during summer and winter seasons. Results represent the average of three days of sampling during each season and error bars represent the standard deviation. 134

6. CHAPTER 6: Fate of microbial agents in wastewater collection systems

In preparation for submission

Authors: Alessia Eramo, Nicole L. Fahrenfeld

6.1. Abstract

The sewer environment is a potential hotspot for the proliferation of antibiotic

resistance genes (ARGs) and other hazardous microbial agents. Attenuation and growth

of harmful microbes in sewers is a concern for sewage workers and in the event of sewer

overflows. To better understand this understudied environment, a field study was

conducted to understand the controls on ARG and heavy metals, which may select for

resistance, in sewers and sewage. The relative abundance of sul1 and tet(G) ARG

between the two matrices were compared and were found to be similar. Both sample

types exhibited higher ARG concentrations in winter, potentially from higher excreted

resistance rates in municipal wastewater in winter. The sewer type (combined or

separate) was not found to influence ARG concentrations. The presence of metals was a

more strongly related factor in this study. Metals correlated with ARG and the silt

fraction in sediment and also drove principle component analysis. This work provides

novel insight into the complex sewer microbiome and the potential hazard posed by

different sewer matrices.

6.2. Introduction

Sewers are understudied and complex environments containing hazardous microbial

agents. The potential for the proliferation of antibiotic resistant bacteria in sewers has not

been well studied and the associated public health threat is not defined (Fahrenfeld and 135

Bisceglia 2016). In cities with combined sewer infrastructure, the releases of sewer contents to adjacent water ways during overflow events are point sources of contamination that degrade waterways and can contribute to waterborne-disease outbreaks (ten Veldhuis et al. 2010). Sewer overflows present a risk to public health by serving as a source of antibiotic resistant bacteria (Eramo et al. 2017a; Young et al. 2013) and pathogens (Donovan et al. 2008) in surface water. Antibiotic resistance is one of the largest public health threats and a comprehensive risk assessment requires an understanding of the fate of ARG from environmental hotspots (Ashbolt et al. 2013) including sewage collection systems, particularly given the risk posed by separate sanitary and combined sewer overflows (CSO).

Understanding the potential for growth and selection for antibiotic resistant microbes in sewers is of interest. Microbes are present in wastewater and other sewer matrices including biofilms (Crabtree 1989) and wastewater solids that settle during conveyance

(Ashley and Crabtree 1992). These sediments may become cohesive or may erode and be re-suspended during high flow events (Ashley and Verbanck 1996). Sewer sediments flushed out during CSO events introduce attached bacteria and other contaminants from the collection system into the environment (Passerat et al. 2011). The observation of polychlorinated biphenyls dechlorination in wastewater collection systems (Rodenburg et al. 2010) is an indication of high levels of microbial activity in sewers and provides motivation for better understanding the fate of microbial agents in sewer sediments.

Microbial agents in sewer sediments will be affected by various factors including their deposition and resuspension in the system and their interactions with biofilm and wastewater matrices. ARG loading in the sewer may be affected by characteristics of the 136 wastewater and sediment as well as by season and by sewer type. The temperature, carbon source, and presence of selecting agents may promote growth and horizontal gene transfer within the microbial community. In combined sewer systems, the stormwater component of CSO will periodically enter the sewer environment. The transport of microbial agents can be attenuated during conveyance, as demonstrated by the ability of the polio virus to persist in sewers even after 800 million liters of wastewater had passed through the system (Hovi et al. 2001)In order to mitigate the associated risk, a better understanding of factors driving microbial growth and selection is sewers is necessary.

To address the gap in knowledge surrounding the potential for growth and attenuation of antibiotic resistant microbes in sewers, a field survey was conducted to understand the loading of these microbial agents in sewer deposits and wastewater conveyed by the sewer system. Water quality parameters including the presence of heavy metals were monitored in sanitary and combined sewers across two seasons. Results can provide insight into the hazard posed by the sewer environment and by the release of different matrices during overflow events.

6.3. Materials and Methods

6.3.1. Sewer Sediment and Wastewater Sampling

Sewer sediments and post-screen wastewater influent were collected from five different sewer systems in the Northeast Atlantic region during two sampling campaigns

(Table 6.1). To capture seasonal variation, sampling was performed for Fall/Winter between September 2016 and January 2016 and for Summer between June 2017 and

September 2017. Sewer sediment sampling locations were selected based on the presence of appreciable sewer deposition and accessibility. Although it was assumed that 137 sediment deposition was ubiquitous, sediment was not always encountered in the field.

Numerous manhole locations (systems C1 and C2) were rejected because sediment was not found at the bottom of the pipe, while on the same day in the vicinity accumulation was appreciable. Samples were collected from a variety of accumulation points in the system as described in Table 6.1. When possible, different points within the systems were sampled. Samples from each system were collected during baseflow conditions at least one week apart. To provide anonymity to collaborating utilities, systems were labeled C1-C3 for the three combined systems and S1-S2 for the two separate systems sampled. On the same day of sediment sampling, a 24-hr composite wastewater influent sample (two liters) was also collected from the corresponding downstream wastewater treatment plant. In one exception, combined system C3 wastewater was collected one day after sediment samples due to precipitation that may have influenced the 24-hr composite sample. Samples were preserved in sterile sample bottles on ice during transport to the lab where they were immediately processed. Field blanks consisting of autoclaved deionized water left open for the duration of sediment sampling were collected during each sampling season. 138

Table 6.1 Field sampling details.

WWTP Sewer Fall/Winter Sampling Summer Sampling Sewer sample Sampling location ID Type Dates Dates location/type frequency 3 different sampling locations during each season; 2/3 locations C1 combined October/November June/July Bottom of sewer pipe; were sampled during each collected from manhole season Bottom of sewer pipe; 6 different sampling manhole C2 combined October/November July/August collected from manhole locations

CSO detention tank; composite samples collected collected from C3 combined October/November June/July from stockpile during each day accumulated sediment of sampling stockpile

Pump/metering stations; 3 different sampling locations collected deposited during each season; 2/3 locations S1 separate September June/July sediment channel or were sampled during each flume season S2 separate December/January August/September Wet well One sampling location

6.3.2. Chemical Characterization of Field Samples

Sediment samples were sieved < 2 mm and subsamples were analyzed for moisture content, pH, conductivity, sieve analysis, and metals. Moisture content was measured by drying aliquots to constant weight (ASTM International 2005). Sediment pH and conductivity was measured according to standard methods (United States Department Of

Agriculture 1954; US Environmental Protection Agency 2004). Analyses were conducted in duplicate for 20% of samples for QA/QC. Particle size analysis was conducted by a sieve method. For each sample, approximately 200-650 g of sediment was dried at ~100 ºC overnight to achieve constant mass. Samples were homogenized with a mortar and pestle and sieved through a series of stacked stainless-steel U.S.

Standard sieves numbered 35, 60, 120, and 230 (ASTM E-11 Certified), which correspond to aperture sizes 500, 250, 125 and 63 µm. The stack was placed on a mechanical shaker for approximately 10 minutes and the mass passing through each sieve 139 was measured. The fraction > 63 µm was classified as sand and the fraction < 63 µm was classified silt/clay. Biomass would be expected to associate with the latter fraction.

Conductivity, pH and redox potential in wastewater samples were measured with a multimeter (Orion Star A329, Thermo Scientific) according to standard methods.

Chemical oxygen demand (COD) was analyzed according to Hach Method 8000 with

Hach COD vials (20-1500 mg/L range) and a DR2700 spectrophotometer. Total suspended solids (TSS) and volatile suspended solids (VSS) were measured according to

Environmental Sciences Section (ESS) Method 340.2 (Wisconsin State Lab of Hygiene

1993). Sediment and wastewater samples were submitted to an outside lab (TestAmerica,

Edison, NJ) for analysis of total arsenic, cadmium, copper and nickel according to EPA

Method SW846 6010C. Metals results were reported on a dry weight basis.

6.3.3. Biomolecular Analyses

Wastewater samples (~150 mL) and field blanks were concentrated on 0.22 µm nitrocellulose filters (Millipore Corporation, Billerica, MA). Filters and sieved sediment aliquots (~0.5 g wet weight) were added to DNA lysing tubes and stored at -20 °C prior to DNA extraction. DNA extractions were conducted using a commercial kit (FastDNA

Spin Kit for Soil, MP Biomedicals) following the manufacturer’s directions. qPCR was performed for select ARG [sul1 (Pei et al. 2006), tet(G) (Aminov et al. 2002)] and 16S rRNA gene copies for total bacterial population (Muyzer et al. 1993). A standard

SybrGreen (5 µL SsoFast EvaGreen, BioRad, Hercules, CA) chemistry with 0.4 µM forward and reverse primers, and 1 µL diluted (1:100) DNA extract in a 10 µL reaction was used. QA/QC on the qPCR included a no-template control on each plate, a 7 point calibration curve, and melt curve and/or gel electrophoresis to verify the specificity of 140 qPCR products. Mass concentrations of ARG were determined on a dry weight basis to compare results between different days and locations and for comparisons to metals data.

Field blanks were also analyzed.

6.3.4. Statistical Analyses

Statistical tests were performed in R (R Core Team 2013). qPCR data were log- normalized for analysis. Comparisons between ARG concentrations for different matrices, seasons, and sewer types were conducted by a student’s t-test (normality of data confirmed by a Shapiro test) or a Wilcoxon rank sum test if data was not normal.

Comparisons between metals observed in different sewer types were also conducted in this way. Multiple comparisons between ARG and sampling location or season for each sewer type were conducted by a one-way ANOVA if data was normal or a Kruskal-

Wallis rank sum test if data was not normal with a post-hoc Tukey or pairwise t-test with a Bonferroni correction, respectively. Correlations between ARG and metals and ARG or metals and silt/clay were tested using a Spearman rank test. To analyze sediment and water quality relationships, principal component analysis (PCA) was applied to normalized environmental data using Primer-E 7 (Plymouth, United Kingdom).

6.4. Results

6.4.1. ARG in Sewer Sediment and Wastewater

The settling of solids during wastewater conveyance contributes to sediment that may collect at joints or other discontinuities in the sewer system. To determine the relationship between sediment samples and their corresponding wastewater, normalized ARG concentrations between the two matrices were compared. The relative abundance of sul1

(gene copies / 16S rRNA gene copies) was 1.1 times higher in wastewater compared to 141 sediment (p=0.03) (Fig. 6.1). Differences in tet(G) abundances were not observed between the two matrices (p=0.16).

The sewer environment may serve as a reservoir of antibiotic resistance which is a concern if/when there are releases of contents during sewer overflow events. Thus, it is important to understand the environmental factors potentially associated with selective pressure for antibiotic resistance in this environment. ARG concentrations were analyzed to determine their relationship to season, sewer type, and sampling location. Seasonal variation in ARG was observed: sul1 and tet(G) concentrations were higher in the fall/winter compared to summer in both sediment and wastewater samples on both a mass or volume concentration and 16S rRNA gene normalized basis (all p<0.03; Figure 2). As expected given that samples were collected during baseflow conditions, wastewater influent exhibited no differences in ARG concentrations or relative abundance between plants belonging to separate and combined systems (all p>0.12). Separate sewer sediment samples had higher mass concentrations of ARG compared to combined systems (both p<3.0×10-4), but this was likely due to these samples having higher biomass given that differences were not observed on a 16S rRNA gene copy normalized basis. To rank the risk associated with these environments, comparisons were made between sewer types for different seasons. In sediments collected in the summer, mass concentrations of sul11 and tet(G) were higher in separate systems (p<0.02). In fall/winter, sul1 had higher mass concentrations in separate systems (p=7.7×10-3). These differences again are likely due to differences in biomass given that differences were not observed when data were normalized to 16S rRNA gene copies (p>0.12). 142

Given the availability of sediment sampling locations varied by system (Table 6.1), concentrations between the five systems were compared. A notable difference is that sediment collected at system C3 was not submerged in wastewater while sediment from other systems were. At C3, the CSO sediment stockpile was only in periodic contact with CSO effluent. Despite the potential for availability of sampling sites, inter-system differences in ARG relative abundances were not detected (all p>0.25).

Figure 6.1 Relative abundance of sul1 in sewer sediment and wastewater samples collected from two separate and three combined collection systems (n=30).

Figure 6.2 Concentrations of ARGs sul1 and tet(G) in (a) sewer sediments and (b.) wastewater samples from separate (n=12) and combined (n=18) collection systems collected during two seasons. 143

6.4.2. Metals in Sewer Sediment and Wastewater

The heavy metal content of sewer sediments and wastewaters were evaluated to provide insight into their potential to select for / be associated with elevated antibiotic resistance. Cadmium concentrations were higher in combined systems (p=1.2×10-4; Fig.

6.5), ranging from 0.44-11.6 mg/kg, than separate systems where it was not detected in the sediment. Copper was detected in all sediment samples and was higher in separate systems (p=7.6×10-3; Fig. 6.6a.). Differences in arsenic and nickel concentrations were not observed between sewer types. Metals were less frequently detected in wastewater samples than sediment samples. Arsenic, cadmium, and nickel were each detected in two samples (<10 % of wastewater samples collected). Copper was detected consistently at concentrations between 39 and 191 µg/L; Fig. 6.6b. No differences in aqueous copper concentrations were observed based on sewer type, despite being higher in sediment samples from the separate collection systems.

Given that antibiotic resistance has been correlated with heavy metals in other environments, the relationship between ARG and heavy metals were tested for sewer sediments and wastewater. On a mass basis, tet(G) correlated positively to copper

(Spearman’s r = 0.45; p=0.01) and total metals (Spearman’s r = 0.40; p=0.03). However, negative correlations were observed between mass concentrations of sul1 and cadmium

(Spearman’s r = -0.53; p=2.9×10-3) and arsenic (Spearman’s r = -0.38; p=0.04) and between tet(G) and arsenic (Spearman’s r = -0.36; p=0.05). For wastewater, correlations were not observed between ARG concentrations and any metal (all p>0.09).

Given that sediments may be released during sewer overflows, the association of heavy metals with different sediment fractions (Fig. 6.7) was explored to provide insight 144 into treatment. Correlations between heavy metals and fractions of silt/clay in sediment were tested to provide insight into sorption likelihood and potential for sediments to serve as sources of metals in the environment. Positive correlations were observed between fractions of silt/clay (<63 µM) and the four metals analyzed (Spearman’s r = 0.46-0.76; p<0.04). Regression analysis indicated a strong linear relationship between arsenic and the silt/clay fraction (R2=0.84) but not for the other metals (R2=0.17-0.22).

6.4.3. PCA Analysis

PCA illustrated how wastewater and sediment physiochemical characteristics were related to each other in each matrix. For sediment, of the total physiochemical data variance, 29.7% was explained by PC1 and 51.2% by PC2 % (Fig. 6.3a.). Samples loosely grouped together by sewer type with some outliers in each group. Clear distinctions were not evident for sewer or season. For wastewater, of the total physiochemical data variance, 37.3% was explained by PC1 and 52.7% by PC2 (Fig.

6.3b.). Clustering by sewer or season was also not observed but overall samples were more closely related by PC2. Metals (arsenic. nickel, and cadmium) and ORP drove

PC2, underscoring the importance of metals in the sewer environment. The lack of sewer and season clustering suggest samples were not related by those variables and sewer conditions were more important. Relative abundances for resistance genes overlaid on the PCA plots do not show strong trends in resistance prevalence (Fig. 6.4). 145

Figure 6.3 Principal component analysis of chemical parameters of (a.) sewer sediment and corresponding (b.) wastewater. Samples color coded by sewer type [separate (s) or combined (c)] and overlaid by season [winter (W) or summer (S)].

Figure 6.4 Principal component analysis of chemical parameters of (a.) sewer sediment and corresponding (b.) wastewater. Samples color coded by relative abundance of sul1 or tet(G) with the size of each half related to the relative abundance value. Samples were overlaid by season [winter (W) or summer (S)].

6.5. Discussion

6.5.1. Impact of Matrix, Season, and Sewer Type on ARG

Given that sewers contain microbial agents and high concentrations of selecting agents (i.e., antibiotics and heavy metals) there is the potential for selection among 146 attenuated microbes which could present a risk to sewage workers, skew sewage epidemiology data, and degrade surface water quality in the event of sewer overflows.

The impact of matrix, season and sewer type on ARG were evaluated. The relative abundance of ARG in paired sediment and wastewater was similar despite there being significant differences in the microbial community structures between these matrices

(data not shown). This may be due to selective agents in sediments such as antibiotics present in wastewater becoming highly bound to sediments and therefore less bioavailable for stimulating resistance (Jarnheimer et al. 2004). Because quantitative models for HGT and selection rates in environmental systems are lacking (Ashbolt et al.

2013), whether the residence time in sewers is sufficient for the expected kinetics of gene transfer is not known. Bacteria associated with newly settled solids may not have acquired new resistance elements, while biofilm that grows on older sediment may be involved in more genetic exchange and therefore be more resistant to selective pressure than free phase microbes (Stewart 2002).

Seasonal differences between ARGs were observed with overall higher concentrations for fall and winter across all samples. Seasonal variations are not thought to be due to changes in sewer water temperature, which were not previously shown to impact the sewage microbiome (Vandewalle et al. 2012). Higher levels of antibiotic resistance bacteria were observed in wastewater in winter and fall. Antibiotic use is up to three times higher in winter (Coutu et al. 2013) and may be an ultimate cause of observed differences given that antibiotics use has been correlated to clinical antibiotic resistant infections (Sun et al. 2012). It appears the microbial risk from contact with sewer overflows is higher during fall and winter compared to summer. However, in the 147 summer, downstream factors such as warmer water temperature in summer can still promote outbreaks of water-borne bacteria (Sterk et al. 2015) and should be taken into consideration to assess the overall threat from sewer overflows.

The sediments in combined and separate systems are exposed to different waste streams with different sources. In addition to different composition, combined systems are also subject to sporadic variations in flow rate. Higher variability in deposition rates and selective pressures in sediments would be expected in combined systems.

Concentrations of sul1 and tet(G) genes in separate systems were higher than in combined systems, but not on a 16S rRNA normalized basis. Thus, while more ARGs are present in separate sewer sediments, the different environments to not appear to be preferentially selecting for these ARG carrying bacteria compared to the community.

Increased flows during wet weather, any associated impacts on of sediments, and the storm water microbial community were not found to impact resistance profiles in comparison to separate systems. In both combined and separate systems erosion may be limited by cohesion observed in sewer sediment (Berlamont and Torfs 1996). More mobility would be expected closer to sewer outfalls.

The variety of sediment sampling locations across collections systems in this study allowed for comparisons to understand if certain types of sediment or accumulation locations served as more significant hotspots. Sediments had similar sul1 and tet(G) concentrations. Even samples collected from the sediment stockpile in the CSO detention tank, which is sometimes dry, exhibited consistent ARG abundances with the other plants. Thus, there is evidence that ARG persist in sewer sediments even without constant exchanges with the mobile bed load and wastewater matrix. 148

6.5.2. Sewer Sediments as Sources of Resistance

Combined sewer sediments were shown to contain ARG and heavy metals that could be released to the environment during wet weather flows. The detection of ARGs at abundances similar to wastewater highlights the that sewer sediments can be a source of microbial contaminants released during CSO events. The presence of heavy metals has been implicated in the coselection for antibiotic resistance and metals resistance. Co- resistance occurs when the genes specifying resistant functionalities are grouped on the same mobile genetic elements such as plasmids (Baker-Austin et al., 2006). In this study, tet(G) correlated positively with copper concentrations in sediment. Copper was the most frequently detected heavy metal. The positive correlation could be due to coselection because plasmids carrying both copper and tetracycline resistance have been reported

(Gullberg et al., 2014). Other metals were not detected in all sediment samples and this may have contributed to the lack of additional observed correlations. The PCA results for both sediment and wastewater further highlight the role of metals in driving relationships between samples.

Metals were detected less frequently in water than sediment samples, indicating that metal concentrations in sediments were likely deposited over time and could be remobilized in the mobile bed load and transported with the wastewater. Sewer type did not correspond with the overall metal accumulation rates. Correlations between metals and silt/clay indicate that sorption may be an important mechanism for metals retained in sewer sediment. This has implications for fate and transport. The sorption of heavy metals to sewer sediments allow them to be retained in the system and if they are bioavailable they have the potential to then serve as a coselective agent. The association 149 of metals with solids also indicates that end-of-pipe treatment methods that remove settleable solids has the potential to remove these contaminants from the effluent during

CSO events.

6.6. Conclusions

This research provides valuable insight towards understanding an understudied potential hotspot for ARGs in the environment. The results of sewer sediment and wastewater sampling revealed similar relative abundances of two ARG between the two matrices. Metals were determined to be an important factor in the sewer environment because they exhibited correlations with ARG and exhibited relationship by PCA.

Metals were also correlated positively with the silt fraction of sediments. Previous investigations into selection in sewers had been conducted. This work will help inform mitigation strategies for sewer overflows and preventative sewer maintenance.

Acknowledgements

Thanks to our utility partners for providing wastewater samples and access to sewer sampling locations. Funding for this project was provided by a grant from the National

Science Foundation (#1510461). Additional support was provided by the Eagleton

Fellowship Program and NIH Bridges to the Doctorate Program. 150

6.7. Supplemental Information

Table 6.2 Chemical parameters in field samples.

Wastewater Influent WWTP Conductivity pH COD (mg/L) TSS (mg/L) VSS (mg/L) ORP (mV) ID (uS/cm) C1 7.3 ± 0.1 891 ± 93 697 ± 160 274 ± 107 220 ± 126 278 ± 90 C2 7.4 ± 0.2 2020 ± 522 603 ± 330 179 ± 56 150 ± 35 253 ± 127 C3 7.0 ± 0.3 863 ± 257 759 ± 220 420 ± 121 344 ± 108 228 ± 71 S1 7.3 ± 0.3 895 ± 158 516 ± 61 277 ± 70 223 ± 52 296 ± 159 S2 7.1 ± 0.1 784 ± 119.6 737 ± 139 443 ± 163 326 ± 115 350 ± 104 Sewer Sediment WWTP Conductivity Moisture pH ID (uS/cm) Content (%) C1 7.0 ± 0.6 119 ± 45 25 ± 8 C2 7.1 ± 0.5 575 ± 702 35 ± 17 C3 6.9 ± 0.8 659 ± 466 30 ± 9 S1 6.6 ± 0.7 466 ± 423 26 ± 16 S2 6.3 ± 0.5 224 ± 79 65 ± 7

Figure 6.5 Cadmium concentrations in sewer sediment samples collected in two different seasons from combined or separate sanitary sewer systems. Three combined and two separate sanitary sewer systems were sampled (n=3 per plant and season, total N=30). 151

Figure 6.6 Copper concentrations in (a) sewer sediment and (b) wastewater influent samples collected in two different seasons from combined or separate sanitary sewer systems. Three combined and two separate sanitary sewer systems were sampled (n=3 per plant and season, total N=30). 152

Figure 6.7 Sieve analysis indicated the percentage of particles smaller than the aperture size listed. Results are shown for data currently available (three combined sewer systems: c1, c2, and c3, and one separate sanitary sewer: s1, for up to three winter samplings: 1, 2, 3 and up to three summer samplings: 4, 5, 6). Data is not available for c2_1, c2_3, s1_2, and s1_3. 153

CHAPTER 7: Conclusion

Microbial contaminants are widespread in the urban water environment and pose threats to human health particularly during wet weather in cities with outdated combined sewer infrastructure. As climate change brings more wet weather to areas with many older cities with combined sewer systems, gray and green infrastructure solutions including end-of-pipe treatment are needed to preserve urban water quality. This dissertation described field and laboratory studies conducted to provide insight into the factors driving abundance of microbial agents and their fate in the urban water environment. Microbial agents including ARGs, human fecal markers, and fecal indicator organisms, and microbial community composition were monitored under various treatment and environmental conditions towards understanding their relationship and developing remediation solutions.

Novel insights provided by this research include:

Treatment by settling

 The majority of ARG at the CSO outfall were observed on the attached fraction of

samples ranging from 59-88% for tested ARGs. The timing of peak ARG did not

coincide with the human fecal indicator marker.

 Microbial community signatures in surface water at a CSO outfall were similar to

wastewater at select early or late storm time points.

 Settling reduced chemical oxygen demand by 49±11% and total suspended solids by

78±7% but did not reduce fecal indicator concentrations.

Treatment by PAA disinfection 154

 PAA disinfection kinetics of cells carrying ARGs were determined. Treatment of

simulated CSO effluent (23% wastewater) with 100 mg∙min/L PAA (5 mg/L

PAA, 20 min) was needed to reduce viable cell sul1, tet(G), and BacHum

(1.0±0.63-3.2±0.25-log) while 25 to 50 mg•min/L PAA (5 mg/L PAA, 5-10 min)

was needed to reduce viable cell loads (0.62±0.56-1.6±0.08-log) in 40%

wastewater from a different municipal treatment plant. Inactivation constants for

measured genes were sul1 and tet(G) > BacHum > 16S rRNA.

 Increasing contact time after the initial decrease in viable cell gene copies did not

significantly improve treatment. Under tested conditions, PAA did not bring

viable-cell concentrations below detection or decrease total ARG concentrations.

 Amplicon sequencing showed that treatment with PAA resulted in marked

increases the relative abundance of select phyla, particularly Clostridia which

increased 1-1.5 orders of magnitude. After PAA disinfection, the fraction of E.coli

carrying the antibiotic resistance gene (ARG) sul1 also increased.

 With settling and disinfection, removals of 3.1±0.14 and 2.5±0.37-log were achieved for

E. coli and total coliform (TC), respectively, similar to removal with disinfection only.

Disinfection was more effective than settling for the microbial treatment of the

wastewater component of CSO effluent.

 Regrowth of E.coli and TC was not detected from PAA-treated CSO was not observed

after seven days.

Treatment by chlorination

 In chlorinated wastewater treatment effluent, ARGs, human fecal indicator

marker, and 16S rRNA gene copies were significantly lower in viable-cells than

overall concentrations. These differences were not preserved in downstream 155

samples, except for tet(G). The results of this study suggest extracellular ARGs

are present in WWTP effluent but may not be available for transformation in the

surface water >1.8km downstream.

The sewer environment as a reservoir of ARGs

 Previous investigations into the sewer microbiome as a reservoir for ARGs

had not been conducted. The relative abundance of sul1 and tet(G) in

wastewater and sediment were found to be similar with both sample types

exhibited higher ARG concentrations in winter, potentially from higher

excreted resistance rates in municipal wastewater in winter.

 Sewer type (combined or separate) was not found to influence ARG

concentrations. Metals correlated with ARG and the silt fraction in sediment

and drove principle component analysis.

The conclusions of this dissertation contribute to the knowledge base about ARG and other microbial contaminants in the urban water environment. CSO was observed as a source of ARG with intra- and inter-storm variability in the release of microbial contaminants during wet weather flows. The results also provide critical insight into the performance of hydrodynamic separation and rapid disinfection with PAA for different microbial agents. Novel application of a viability-based PCR confirmed the mechanism of PAA disinfection and was used to determine disinfection kinetics for a broader range of microbial targets and to quantify the relative pool of non-viable ARG in wastewater effluent in comparison to receiving waters. This knowledge is critical for understanding which environments have the greatest potential to serve as hotspots for horizontal gene transfer via transformation. Finally, the role of sewer sediments in attenuating microbial 156 agents including ARG was demonstrated. Season, sewer and metal concentrations were found to be important factor affecting observed ARG concentrations. Overall this work provides new insight into the role of sewers, sewage, and wet weather flows as sources of microbial agents and the potential for end-of-pipe treatment to reduce the risk of microbial contaminants in the water environment. These results may be useful for modeling microbial fate and assessing the risk posed by emerging microbial contaminants in the urban environment.

Suggested future work stemming from this research includes investigating the microbial contribution of the stormwater component of CSO for the design of full-scale end-of-pipe treatment. Pilot tests with actual CSO could be conducted using the results obtained in the dissertation as guidelines for treatment. For full-scale end-of-pipe treatment with PAA, any potentially negative effects from the release of larger volumes of PAA to the estuarine environment should be investigated. Additionally, treatment conditions should be optimized to maximize removals during actual wet weather events.

To further investigate the fraction of viable-cell ARG versus total concentrations of

ARGs after wastewater treatment, it would be advisable to sample the water column closer to the WWTP outfalls to look for evidence of significant fractions of eDNA and nonviable cells with ARGs that would be available for transformation. Future research could also focus on bed sediments as reservoirs of extracellular ARGs downstream of wastewater treatment plant discharges. The research conducted exploring the sewer environment can help inform mitigation strategies for sewer overflows and preventative sewer maintenance and builds a foundation for further characterization of the sewer microbiome. 157

References Ahmed, W., Goonetilleke, A., Powell, D., and Gardner, T. (2009). "Evaluation of multiple sewage-associated Bacteroides PCR markers for sewage pollution tracking." Water Res., 43(19), 4872-4877. Alkhaddar, R. M., Higgins, P. R., Phipps, D. A., and Andoh, R. Y. G. (2001). "Residence time distribution of a model hydrodynamic vortex separator." Urban Water, 3(1– 2), 17-24. Aminov, R., Garrigues-Jeanjean, N., and Mackie, R. (2001). "Molecular ecology of tetracycline resistance: development and validation of primers for detection of tetracycline resistance genes encoding ribosomal protection proteins." Appl. Environ. Microbiol., 67(1), 22-32. Aminov, R. I., Chee-Sanford, J. C., Garrigues, N., Teferedegne, B., Krapac, I. J., White, B. A., and Mackie, R. I. (2002). "Development, Validation, and Application of PCR Primers for Detection of Tetracycline Efflux Genes of Gram-Negative Bacteria." Appl Environ Microbiol, 68(4), 1786-1793. Anderson, I. C., Rhodes, M., and Kator, H. (1979). "Sublethal stress in Escherichia coli: a function of salinity." Applied and Environmental Microbiology, 38(6), 1147- 1152. Andoh, R. Y. G., and Saul, A. J. (2003). "The use of hydrodynamic vortex separators and screening systems to improve water quality." Water Sci. Technol., 47(4), 175-183. Antonelli, M., Rossi, S., Mezzanotte, V., and Nurizzo, C. (2006). "Secondary effluent disinfection: PAA long term efficiency." Environmental Science & Technology, 40(15), 4771-4775. Arnone, R. D., and Walling, J. P. (2007). "Waterborne pathogens in urban watersheds." Journal of Water and Health, 5(1), 149-162. Ashbolt, N. J., Amézquita, A., Backhaus, T., Borriello, P., Brandt, K. K., Collignon, P., Coors, A., Finley, R., Gaze, W. H., and Heberer, T. (2013). "Human health risk assessment (HHRA) for environmental development and transfer of antibiotic resistance, ." Environmental Health Perspectives, 121(9), 993. Ashbolt, N. J., Heberer, T., van den Eede, C., Schönfeld, J., Krone, S., Backhaus, T., Li, X.-Y., Borriello, P., Amezquita, A., Larsson, D. G. J., Brandt, K., Smalla, K., Silley, P., Lawrence, J., and Ryan, J. (2013). "Human Health Risk Assessment (HHRA) for Environmental Development and Transfer of Antibiotic Resistance." Environ. Health Perspect. Ashbolt, N. J., Heberer, T., van den Eede, C., Schönfeld, J., Krone, S., Backhaus, T., Li, X.-Y., Borriello, P., Amezquita, A., Larsson, D. G. J., Brandt, K., Smalla, K., Silley, P., Lawrence, J., and Ryan, J. (2013). "Human Health Risk Assessment (HHRA) for Environmental Development and Transfer of Antibiotic Resistance." Environ. Health Perspect., 121(9), 993-1001. Ashley, R., and Crabtree, R. (1992). "Sediment origins, deposition and build-up in combined sewer systems." Water Science and Technology, 34(8), 1-12. Ashley, R., and Verbanck, M. (1996). "Mechanics of sewer sediment erosion and transport." Journal of Hydraulic Reseach 34(6), 753-770. Ashley, R. M., and Hvitved-Jacobsen, T. (2003). Management of sewer sediments. In Wet-weather flow in the urban watershed: Technology and management, Lewis Publishers: Boca Raton, FL. 158

ASTM International (2005). "Standard test methods for laboratory determination of water (moisture) content of soil and rock by mass." ASTM. ASTM International (2016). "Standard practices for determining microbial colony counts from waters analyzed by plating methods."Conshohocken, PA. Bae, S., and Wuertz, S. (2009). "Discrimination of viable and dead fecal Bacteroidales bacteria by quantitative PCR with propidium monoazide." Appl Environ Microbiol, 75(9), 2940-2944. Banihashemi, A., Dyke, M., and Huck, P. (2012). "Long‐amplicon propidium monoazide‐PCR enumeration assay to detect viable Campylobacter and Salmonella." J. Appl. Microbiol., 113(4), 863-873. Banihashemi, A., Van Dyke, M. I., and Huck, P. M. (2012). "Long-amplicon propidium monoazide-PCR enumeration assay to detect viable Campylobacter and Salmonella." J Appl Microbiol, 113(4), 863-873. Bayonne Municipal Utilities Authority (2017). "Wet weather flow treatment and disinfection demonstration project." Berlamont, J. E., and Torfs, H. M. (1996). "Modeling (partly) cohesive sediment transport in sewer systems." Water Science and technology, 33(9), 171-178. Biswal, B. K., Khairallah, R., Bibi, K., Mazza, A., Gehr, R., Masson, L., and Frigon, D. (2014). "Impact of UV and peracetic acid disinfection on the prevalence of virulence and antimicrobial resistance genes in uropathogenic Escherichia coli in wastewater effluents." Applied and Environmental Microbiology, 80(12), 3656- 3666. Block, P. (2016). "The decomposition kinetics of peracetic acid and hydrogen peroxide in municipal wastewaters." WEFTECNew Orleans, LA, 555-563. Boehm, A. B., Van De Werfhorst, L. C., Griffith, J. F., Holden, P. A., Jay, J. A., Shanks, O. C., Wang, D., and Weisberg, S. B. (2013). "Performance of forty-one microbial source tracking methods: A twenty-seven lab evaluation study." Water Res., 47(18), 6812-6828. Breazeal, M. V. R., Novak, J. T., Vikesland, P. J., and Pruden, A. (2013). "Effect of wastewater colloids on membrane removal of antibiotic resistance genes." Water research, 47(1), 130-140. Casey, J. A., Curriero, F. C., Cosgrove, S. E., Nachman, K. E., and Schwartz, B. S. (2013). "High-density livestock operations, crop field application of manure, and risk of community-associated methicillin-resistant Staphylococcus aureus infection in Pennsylvania." JAMA Intern Med, 173(21), 1980-1990. Centers for Disease Control and Prevention (2013). "Antibiotic resistance threats in the United States, 2013." Department of Health and Human Services, ed. Chang, P. H., Juhrend, B., Olson, T. M., Marrs, C. F., and Wigginton, K. R. (2017). "Degradation of extracellular antibiotic resistance genes with UV254 treatment." Environ Sci Technol, 51(11), 6185-6192. Characklis, G. W., Dilts, M. J., Simmons Iii, O. D., Likirdopulos, C. A., Krometis, L.-A. H., and Sobsey, M. D. (2005). "Microbial partitioning to settleable particles in stormwater." Water Res., 39(9), 1773-1782. Cheng, N. S. (1997). "Simplified Settling Velocity Formula for Sediment Particle." Journal of Hydraulic Engineering, 123(2). 159

Chhetri, R. K., Bonnerup, A., and Andersen, H. R. (2016). "Combined Sewer Overflow pretreatment with chemical coagulation and a particle settler for improved peracetic acid disinfection." Journal of Industrial and Engineering Chemistry, 37, 372-379. Chhetri, R. K., Flagstad, R., Munch, E. S., Hørning, C., Berner, J., Kolte-Olsen, A., Thornberg, D., and Andersen, H. R. (2015). "Full scale evaluation of combined sewer overflows disinfection using performic acid in a sea-outfall pipe." Chemical Engineering Journal, 270, 133-139. Chhetri, R. K., Thornberg, D., Berner, J., Gramstad, R., Ojstedt, U., Sharma, A. K., and Andersen, H. R. (2014). "Chemical disinfection of combined sewer overflow waters using performic acid or peracetic acids." Science of the Total Environment, 490, 1065-1072. Chopra, I., and Roberts, M. (2001). "Tetracycline antibiotics: mode of action, applications, molecular biology, and epidemiology of bacterial resistance." Microbiol Mol Biol Rev, 65(2), 232-260 ; second page, table of contents. Clark, N., Cooksey, R., Hill, B., Swenson, J., and Tenover, F. (1993). "Characterization of glycopeptide-resistant enterococci from US hospitals." Antimicrob. Agents Chemother., 37(11), 2311-2317. Comi, M., and Mickley, M. (2017). "Citizen Science Monitoring for Pathogen Indicators Along the NJ Raritan Bayshore: A Final Report for the NY-NJ Harbor and Estuary Program." Conter, A., Gangneux, C., Suzanne, M., and Gutierrez, C. (2001). "Survival of Escherichia coli during long-term starvation: effects of aeration, NaCl, and the rpoS and osmC gene products." Research in Microbiology, 152(1), 17-26. Contreras, P. J., Urrutia, H., Sossa, K., and Nocker, A. (2011). "Effect of PCR amplicon length on suppressing signals from membrane-compromised cells by propidium monoazide treatment." J Microbiol Methods, 87(1), 89-95. Contreras, P. J., Urrutia, H., Sossa, K., and Nocker, A. (2011). "Effect of PCR amplicon length on suppressing signals from membrane-compromised cells by propidium monoazide treatment." J. Microbiol. Methods, 87(1), 89-95. Copeland, C. (2014). "Green Infrastructure and issues in managing urban stormwater." Congressional Research Service, Washington, DC. Corinaldesi, C., Danovaro, R., and Dell'Anno, A. (2005). "Simultaneous recovery of extracellular and intracellular DNA suitable for molecular studies from marine sediments." Appl Environ Microbiol, 71(1), 46-50. Coutu, S., Rossi, L., Barry, D. A., Rudaz, S., and Vernaz, N. (2013). "Temporal variability of antibiotics fluxes in wastewater and contribution from hospitals." PLoS One, 8(1), e53592. Crabtree, R. (1989). "Sediments in Sewers." Water and Environment Journal, 3(6), 569- 578. De Luca, G., Sacchetti, R., Zanetti, F., and Leoni, E. (2008). "Comparative study on the efficiency of peracetic acid and chlorine dioxide at low doses in the disinfection of urban wastewaters." Annals of Agricultural and Environmental Medicine, 15(2), 217-224. Devarajan, N., Laffite, A., Graham, N. D., Meijer, M., Prabakar, K., Mubedi, J. I., Elongo, V., Mpiana, P. T., Ibelings, B. W., and Wildi, W. (2015). "Accumulation 160

of clinically relevant antibiotic-resistance genes, bacterial load, and metals in freshwater lake sediments in Central Europe." Environmental science & technology, 49(11), 6528-6537. Di Cesare, A., Fontaneto, D., Doppelbauer, J., and Corno, G. (2016). "Fitness and Recovery of Bacterial Communities and Antibiotic Resistance Genes in Urban Wastewaters Exposed to Classical Disinfection Treatments." Environmental Science & Technology, 50(18), 10153-10161. Diene, S. M., Bruder, N., Raoult, D., and Rolain, J.-M. (2011). "Real-time PCR assay allows detection of the New Delhi metallo-β-lactamase (NDM-1)-encoding gene in France." Int. J. Antimicrob. Agents, 37(6), 544-546. Dodd, M. C. (2012). "Potential impacts of disinfection processes on elimination and deactivation of antibiotic resistance genes during water and wastewater treatment." Journal of Environmental Monitoring, 14(7), 1754-1771. Dodd, M. C. (2012). "Potential impacts of disinfection processes on elimination and deactivation of antibiotic resistance genes during water and wastewater treatment." Journal of Environmental Monitoring, 14(7), 1754-1771. Domínguez Henao, L., Delli Compagni, R., Turolla, A., and Antonelli, M. (2018). "Influence of inorganic and organic compounds on the decay of peracetic acid in wastewater disinfection." Chemical Engineering Journal, 337, 133-142. Donovan, E., Staskal, D., Unice, K., Roberts, J., Haws, L., Finley, B., and Harris, M. (2008). "Risk of gastrointestinal disease associated with exposure to pathogens in the sediments of the Lower Passaic River." Applied and Environmental Microbiology, 74(4), 1004-1018. Donovan, E., Unice, K., Roberts, J. D., Harris, M., and Finley, B. (2008). "Risk of gastrointestinal disease associated with exposure to pathogens in the water of the lower Passaic River." Appl. Environ. Microbiol., 74(4), 994-1003. Eaton, T., O'Mullan, G. D., and Rouff, A. A. (2013). "Assessing continuous contamination discharge from a combined sewer outfall (CSO) into a tidal wetland creek: bacteriological and heavy metals indicators." Annals of Environmental Science, 7, 79-92. El-Rehaili, A. M. (1995). "Response of BOD, COD and TOC of secondary effluents to chlorination." Water Res., 29(6), 1571-1577. Eramo, A., Delos Reyes, H., and Fahrenfeld, N. ((in review)). "Partitioning of antibiotic resistance genes and fecal indicators varies intra and inter-storm during combined sewer overflows." Eramo, A., Delos Reyes, H., and Fahrenfeld, N. L. (2017a). "Partitioning of Antibiotic Resistance Genes and Fecal Indicators Varies Intra and Inter-Storm during Combined Sewer Overflows." Front Microbiol, 8, 1-13. Eramo, A., Medina, W. M., and Fahrenfeld, N. L. (2017b). "Peracetic acid disinfection kinetics for combined sewer overflows: indicator organisms, antibiotic resistance genes, and microbial community." Environ Sci Water Res Technol, 3(6), 1061- 1072. Fahrenfeld, N., and Bisceglia, K. J. (2016). "Emerging investigators series: sewer surveillance for monitoring antibiotic use and prevalence of antibiotic resistance: urban sewer epidemiology." Environ. Sci.: Water Res. Technol., 2(5), 788-799. 161

Fahrenfeld, N., and Bisceglia, K. J. (2016). "Emerging investigators series: sewer surveillance for monitoring antibiotic use and prevalence of antibiotic resistance: urban sewer epidemiology." Environmental Science: Water Research & Technology, 2(5), 788-799. Fahrenfeld, N., Knowlton, K., Krometis, L. A., Hession, W. C., Xia, K., Lipscomb, E., Libuit, K., Green, B. L., and Pruden, A. (2014). "Effect of manure application on abundance of antibiotic resistance genes and their attenuation rates in soil: Field- scale mass balance approach." Environ. Sci. Technol., 48(5), 2643-2650. Fahrenfeld, N. L., Ma, Y., O'Brien, M., and Pruden, A. (2013). "Reclaimed water as a reservoir of antibiotic resistance genes: distribution system and irrigation implications." Frontiers Microbiol., 4. Flint, K. (1987). "The long‐term survival of Escherichia coli in river water." Journal of Applied Microbiology, 63(3), 261-270. Forsberg, K. J., Reyes, A., Wang, B., Selleck, E. M., Sommer, M. O. A., and Dantas, G. (2012). "The shared antibiotic resistome of soil bacteria and human pathogens." Science, 337(6098), 1107-1111. Garner, E., Wallace, J. S., Argoty, G. A., Wilkinson, C., Fahrenfeld, N., Heath, L. S., Zhang, L., Arabi, M., Aga, D. S., and Pruden, A. (2016). "Metagenomic profiling of historic Colorado Front Range flood impact on distribution of riverine antibiotic resistance genes." Sci Rep, 6, 38432. Gasperi, J., Garnaud, S., Rocher, V., and Moilleron, R. (2008). "Priority pollutants in wastewater and combined sewer overflow." Science of the Total Environment, 407(1), 263-272. Gehr, R., Wagner, M., Veerasubramanian, P., and Payment, P. (2003). "Disinfection efficiency of peracetic acid, UV and ozone after enhanced primary treatment of municipal wastewater." Water Research, 37(19), 4573-4586. Gervin, L., and Brix, H. (2001). "Removal of nutrients from combined sewer overflows and lake water in a vertical-flow contsructed wetland system." Water Science and Technology, 44(11-12), 171-176. González, A., Gehr, R., Vaca, M., and López, R. (2012). "Disinfection of an Advanced Primary Effluent with Peracetic Acid and Ultraviolet Combined Treatment: A Continuous-Flow Pilot Plant Study." Water Environment Research, 84(3), 247- 253. Graham, M. S., Lomax, G., and Skipp, G. "Peracetic Acid (PAA) versus Chlorination/De-chlorination – A Disinfection Comparison." Proc., WEFTEC, Water Environment Federation, 4672-4684. Guardabassi, L., Schwarz, S., and Lloyd, D. H. (2004). "Pet animals as reservoirs of antimicrobial-resistant bacteria: Review." J. Antimicrob. Chemother., 54(2), 321- 332. Gutierrez-Cacciabue, D., Cid, A. G., and Rajal, V. B. (2016). "How long can culturable bacteria and total DNA persist in environmental waters? The role of sunlight and solid particles." Sci Total Environ, 539, 494-502. Hannouche, A., Chebbo, G., and Joannis, C. (2014). "Assessment of the contribution of sewer deposits to suspended solids loads in combined sewer systems during rain events." Environmental Science and Pollution Research, 21(8), 5311-5317. 162

Harwood, V. J., Levine, A. D., Scott, T. M., Chivukula, V., Lukasik, J., Farrah, S. R., and Rose, J. B. (2005). "Validity of the indicator organism paradigm for pathogen reduction in reclaimed water and public health protection." Appl. Environ. Microbiol., 71(6), 3163-3170. Hovi, T., Stenvik, M., Partanen, H., and Kangas, A. (2001). "Poliovirus surveillance by examining sewage specimens. Quantitative recovery of virus after introduction into sewerage at remote upstream location." Epidemiology & Infection, 127(1), 101-106. Hu, Q., Zhang, X. X., Jia, S., Huang, K., Tang, J., Shi, P., Ye, L., and Ren, H. (2016). "Metagenomic insights into ultraviolet disinfection effects on antibiotic resistome in biologically treated wastewater." Water Res, 101, 309-317. Huang, J. J., Hu, H. Y., Lu, S. Q., Li, Y., Tang, F., Lu, Y., and Wei, B. (2012). "Monitoring and evaluation of antibiotic-resistant bacteria at a municipal wastewater treatment plant in China." Environ Int, 42, 31-36. Huang, J. J., Hu, H. Y., Tang, F., Li, Y., Lu, S. Q., and Lu, Y. (2011). "Inactivation and reactivation of antibiotic-resistant bacteria by chlorination in secondary effluents of a municipal wastewater treatment plant." Water Res, 45(9), 2775-2781. Hvitved-Jacobsen, T., Vollertsen, J., and Nielsen, A. H. (2013). Sewer processes: microbial and chemical process engineering of sewer networks, CRC press. Jarnheimer, P.-Å., Ottoson, J., Lindberg, R., Stenström, T.-A., Johansson, M., Tysklind, M., Winner, M.-M., and Olsen, B. (2004). "Fluoroquinolone antibiotics in a hospital sewage line; occurrence, distribution and impact on bacterial resistance." Scandinavian journal of infectious diseases, 36(10), 752-755. Karvelas, M., Katsoyiannis, A., and Samara, C. (2003). "Occurrence and fate of heavy metals in the wastewater treatment process." Chemosphere, 53(10), 1201-1210. Khan, S., Knapp, C. W., and Beattie, T. K. (2016). "Antibiotic Resistant Bacteria Found in Municipal Drinking Water." Environmental Processes, 3(3), 541-552. Kildare, B. J., Leutenegger, C. M., McSwain, B. S., Bambic, D. G., Rajal, V. B., and Wuertz, S. (2007). "16S rRNA-based assays for quantitative detection of universal, human-, cow-, and dog-specific fecal Bacteroidales: a Bayesian approach." Water Res, 41(16), 3701-3715. Kildare, B. J., Leutenegger, C. M., McSwain, B. S., Bambic, D. G., Rajal, V. B., and Wuertz, S. (2007). "16S rRNA-based assays for quantitative detection of universal, human-, cow-, and dog-specific fecal Bacteroidales: A Bayesian approach." Water Res., 41(16), 3701-3715. Kim, M., Gutierrez-Cacciabue, D., Schriewer, A., Rajal, V. B., and Wuertz, S. (2014). "Evaluation of detachment methods for the enumeration of Bacteroides fragilis in sediments via propidium monoazide quantitative PCR, in comparison with Enterococcus faecalis and Escherichia coli." J Appl Microbiol, 117(5), 1513- 1522. Kitis, M. (2004). "Disinfection of wastewater with peracetic acid: a review." Environment International, 30(1), 47-55. Kozich, J. J., Westcott, S. L., Baxter, N. T., Highlander, S. K., and Schloss, P. D. (2013). "Development of a dual-index sequencing strategy and curation pipeline for analyzing amplicon sequence data on the MiSeq Illumina sequencing platform." Appl. Environ. Microbiol. 163

Krometis, L.-A. H., Characklis, G. W., Simmons Iii, O. D., Dilts, M. J., Likirdopulos, C. A., and Sobsey, M. D. (2007). "Intra-storm variability in microbial partitioning and microbial loading rates." Water Res., 41(2), 506-516. Krometis, L. A., Characklis, G. W., Simmons, O. D., 3rd, Dilts, M. J., Likirdopulos, C. A., and Sobsey, M. D. (2007). "Intra-storm variability in microbial partitioning and microbial loading rates." Water Research, 41(2), 506-516. LaPara, T. M., Madson, M., Borchardt, S., Lang, K. S., and Johnson, T. J. (2015). "Multiple Discharges of Treated Municipal Wastewater Have a Small Effect on the Quantities of Numerous Antibiotic Resistance Determinants in the Upper Mississippi River." Environ Sci Technol, 49(19), 11509-11515. Lazarova, V., Janex, M. L., Fiksdal, L., Oberg, C., Barcina, I., and Pommepuy, M. (1998). "Advanced wastewater disinfection technologies: short and long term efficiency." Water Science and Technology, 38(12), 109-117. Lefevre, F., Audic, J., and Ferrand, F. (1992). "Peracetic acid disinfection of secondary effluents discharged off coastal seawater." Water Science and Technology, 25(12), 155-164. Li, D., Tong, T., Zeng, S., Lin, Y., Wu, S., and He, M. (2014). "Quantification of viable bacteria in wastewater treatment plants by using propidium monoazide combined with quantitative PCR (PMA-qPCR)." J Environ Sci, 26(2), 299-306. Li, J., Cheng, W., Xu, L., Jiao, Y., Ali Baig, S., and Chen, H. (2015). "Occurrence and removal of antibiotics and the corresponding resistance genes in wastewater treatment plants: effluents’ influence to downstream water environment." Enviorn Sci Pollut Res, 10. Liu, S.-S., Qu, H.-M., Yang, D., Hu, H., Liu, W.-L., Qiu, Z.-G., Hou, A.-M., Guo, J., Li, J.-W., and Shen, Z.-Q. (2018). "Chlorine disinfection increases both intracellular and extracellular antibiotic resistance genes in a full-scale wastewater treatment plant." Water Res, 136, 131-136. Luo, J. F., Lin, W. T., and Guo, Y. (2010). "Method to detect only viable cells in microbial ecology." Appl Microbiol Biotechnol, 86(1), 377-384. Luo, Y., Yang, F., Mathieu, J., Mao, D., Wang, Q., and Alvarez, P. J. J. (2013). "Proliferation of multidrug-resistant new delhi metallo-beta-lactamase genes in municipal wastewater treatment plants in northern China." Environmental Science and Technology Letter, 1(1), 26-30. Luprano, M. L., De Sanctis, M., Del Moro, G., Di Iaconi, C., Lopez, A., and Levantesi, C. (2016). "Antibiotic resistance genes fate and removal by a technological treatment solution for water reuse in agriculture." Sci Total Environ, 571, 809- 818. Mailhot, A., Talbot, G., and Lavallée, B. (2015). "Relationships between rainfall and Combined Sewer Overflow (CSO) occurrences." Journal of Hydrology, 523, 602- 609. Mantilla-Calderon, D., Hong, P., (2017). "Fate and Persistence of a Pathogenic NDM-1- Positive Escherichia coli Strain in Anaerobic and Aerobic Sludge Microcosms." Appl and Environ Microbiol, 83(13), 1-13. Mao, D., Luo, Y., Mathieu, J., Wang, Q., Feng, L., Mu, Q., Feng, C., and Alvarez, P. J. (2014). "Persistence of extracellular DNA in river sediment facilitates antibiotic resistance gene propagation." Environ Sci Technol, 48(1), 71-78. 164

Mao, D., Luo, Y., Mathieu, J., Wang, Q., Feng, L., Mu, Q., Feng, C., and Alvarez, P. J. J. (2013). "Persistence of extracellular DNA in river sediment facilitates antibiotic resistance gene propagation." Environ. Sci. Technol., 48(1), 71-78. Martínez, J. L., Coque, T. M., and Baquero, F. (2015). "What is a resistance gene? Ranking risk in resistomes." Nature Reviews Microbiology, 13(2), 116-123. McFadden, M., Loconsole, J., Schockling, A., Nerenberg, R., and Pavissich, J. (2017). "Comparing peracetic acid and hypochlorite for disinfection of combined sewer overflows: Effects of suspended-solids and pH." Sci Total Environ, 599, 533-539. McGann, P., Snesrud, E., Maybank, R., Corey, B., Ong, A. C., Clifford, R., Hinkle, M., Whitman, T., Lesho, E., and Schaecher, K. E. (2016). "Escherichia coli Harboring mcr-1 and blaCTX-M on a Novel IncF Plasmid: First Report of mcr-1 in the United States." Antimicrob Agents Chemother, 60(7), 4420-4421. McKinney, C. W., and Pruden, A. (2012). "Ultraviolet disinfection of antibiotic resistant bacteria and their antibiotic resistance genes in water and wastewater." Environ. Sci. Technol., 46(24), 13393-13400. McKinney, C. W., and Pruden, A. (2012). "Ultraviolet disinfection of antibiotic resistant bacteria and their antibiotic resistance genes in water and wastewater." Environ Sci Technol, 46(24), 13393-13400. McLellan, S. L., Fisher, J. C., and Newton, R. J. (2015). "The microbiome of urban waters." Int Microbiol, 18(3), 141-149. McLellan, S. L., Hollis, E. J., Depas, M. M., Van Dyke, M., Harris, J., and Scopel, C. O. (2007). "Distribution and Fate of Escherichia coli in Lake Michigan Following Contamination with Urban Stormwater and Combined Sewer Overflows." Journal of Great Lakes Research, 33(3), 566-580. McLellan, S. L., Hollis, E. J., Depas, M. M., Van Dyke, M., Harris, J., and Scopel, C. O. (2007). "Distribution and Fate of Escherichia coli in Lake Michigan Following Contamination with Urban Stormwater and Combined Sewer Overflows." Journal of Great Lakes Research, 33(3), 566. Meyer, D., Molle, P., Esser, D., Troesch, S., Masi, F., and Dittmer, U. (2012). "Constructed wetlands for combined sewer overflow treatment—comparison of German, French and Italian approaches." Water, 5(1), 1-12. Mezzanotte, V., Antonelli, M., Azzellino, A., Citterio, S., and Nurizzo, C. (2003). "Secondary effluent disinfection by peracetic acid (PAA): microrganism inactivation and regrowth, preliminary results." Water Science and Technology: Water Supply, 3(4), 269-275. Michelbach, S. (1995). "Origin, resuspension and settling characteristics of solids transported in combined sewage." Water Science and Technology, 7(1), 61-77. Muyzer, G., De Waal, E. C., and Uitterlinden, A. G. (1993). "Profiling of complex microbial populations by denaturing gradient gel electrophoresis analysis of polymerase chain reaction-amplified genes coding for 16S rRNA." Appl. Environ. Microbiol., 59(3), 695-700. Muyzer, G., De Waal, E. C., and Uitterlinden, A. G. (1993). "Profiling of complex microbial populations by denaturing gradient gel electrophoresis analysis or polymerase chain reaction-amplified genes coding for 16S rRNA." Appl Environ Microbiol 59(3), 695-700. 165

Narciso-da-Rocha, C., Varela, A. R., Schwartz, T., Nunes, O. C., and Manaia, C. M. (2014). "blaTEM and vanA as indicator genes of antibiotic resistance contamination in a hospital-urban wastewater treatment plant system." J Glob Antimicrob Resist, 2(4), 309-315. New Jersey Future "New Jersey Combined Sewer System Fact Sheet." . (January 16, 2017). New York City Department of Environmental Protection (2014). "Combined Sewer Overflow Long Term Control Plan for Hutchinson River." New York City Department of Environmetal Protection. Newton, R. J., Bootsma, M. J., Morrison, H. G., Sogin, M. L., and McLellan, S. L. (2013). "A microbial signature approach to identify fecal pollution in the waters off an urbanized coast of Lake Michigan." Microb. Ecol., 65(4), 1011-1023. Newton, R. J., Bootsma, M. J., Morrison, H. G., Sogin, M. L., and McLellan, S. L. (2013). "A Microbial Signature Approach to Identify Fecal Pollution in the Waters Off an Urbanized Coast of Lake Michigan." Microbial Ecology, 65, 13. Newton, R. J., McLellan, S. L., Dila, D. K., Vineis, J. H., Morrison, H. G., Eren, A. M., and Sogin, M. L. (2015). "Sewage reflects the microbiomes of human populations." mBio, 6(2). Nielsen, K. M., Johnsen, P. J., Bensasson, D., and Daffonchio, D. (2007). "Release and persistence of extracellular DNA in the environment." Environ Biosafety Res, 6(1- 2), 37-53. Nocker, A., Cheung, C. Y., and Camper, A. K. (2006). "Comparison of propidium monoazide with ethidium monoazide for differentiation of live vs. dead bacteria by selective removal of DNA from dead cells." J Microbiol Methods, 67(2), 310- 320. Nocker, A., Richter-Heitmann, T., Montijn, R., Schuren, F., and Kort, R. (2010). "Discrimination between live and dead cells in bacterial communities from environmental water samples analyzed by 454 pyrosequencing." Int. Microbiol., 13(2), 59-65. Nocker, A., Richter-Heitmann, T., Montijn, R., Schuren, F., and Kort, R. (2010). "Discrimination between live and dead cellsin bacterial communities from environmental water samples analyzed by 454 pyrosequencing." Int Microbiol, 13(2), 59-65. Nocker, A., Sossa-Fernandez, P., Burr, M. D., and Camper, A. K. (2007). "Use of propidium monoazide for live/dead distinction in microbial ecology." Appl. Environ. Microbiol., 73(16), 5111-5117. Nocker, A., Sossa-Fernandez, P., Burr, M. D., and Camper, A. K. (2007a). "Use of propidium monoazide for live/dead distinction in microbial ecology." Appl Environ Microbiol, 73(16), 5111-5117. Nocker, A., Sossa, K. E., and Camper, A. K. (2007b). "Molecular monitoring of disinfection efficacy using propidium monoazide in combination with quantitative PCR." J Microbiol Methods, 70(2), 252-260. NY State Department of Health (2014). "Drinking Water Advisories Following Hurricane Sandy." 166

. (January 23, 2014, 2014). Passerat, J., Ouattara, N. K., Mouchel, J.-M., Vincent, R., and Servais, P. (2011). "Impact of an intense combined sewer overflow event on the microbiological water quality of the Seine River." Water Res., 45(2), 893-903. Passerat, J., Ouattara, N. K., Mouchel, J. M., Rocher, V., and Servais, P. (2011). "Impact of an intense combined sewer overflow event on the microbiological water quality of the Seine River." Water Research, 45(2), 893-903. Patoczka, J., Dening, J., and Rolak, J. (2016). "Bayonne, NJ combined sewer overflow treatment demonstration project: preliminary disinfection results." WEFTECNew Orleans, LA, 4017-4037. Patoczka, J., Dening, J., and Rolak, J. (2016). "Bayonne, NJ Combined Sewer Overflow Treatment Demonstration Project: Preliminary Disinfection Results." 89th Annual Water Environment Federation Technical Exhibition and Conference (WEFTEC)New Orleans, LA, 4017-4037. Pei, R., Kim, S.-C., Carlson, K. H., and Pruden, A. (2006). "Effect of river landscape on the sediment concentrations of antibiotics and corresponding antibiotic resistance genes (ARG)." Water Res., 40(12), 2427-2435. Pei, R., Kim, S. C., Carlson, K. H., and Pruden, A. (2006). "Effect of river landscape on the sediment concentrations of antibiotics and corresponding antibiotic resistance genes (ARG)." Water Res, 40(12), 2427-2435. Pepper, I. L., Brooks, J. P., and Gerba, C. P. (2018). "Antibiotic resistant bacteria in municipal wastes: Is there reason for concern?" Environmental science & technology, 52(7), 3949-3959. Phillips, P. J., Chalmers, A. T., Gray, J. L., Kolpin, D. W., Foreman, W. T., and Wall, G. R. (2012). "Combined sewer overflows: an environmental source of hormones and wastewater micropollutants." Environ Sci Technol, 46(10), 5336-5343. Pruden, A., Arabi, M., and Storteboom, H. (2012). "Correlation between upstream human activities and riverine antibiotic resistance genes." Environ Sci Tech, 46(21), 11541-11549. Pruden, A., Larsson, D., Amézquita, A., Collignon, P., Brandt, K., Graham, D., Lazorchak, J., Suzuki, S., Silley, P., and Snape, J. (2013). "Management options for reducing the release of antibiotics and antibiotic resistance genes to the environment." Environmental health perspectives, 121(8), 878. Puerta, J., Suarez, J., and Diaz-Ferros, V. (2002). "Contaminant loads of CSOs at the wastewater treatment plant of a city in NW Spain." Urban Water, 4(3), 291-299. R Core Team. 2013. R: A language and environment for statistical computingR Foundation for Statistical Computing, Vienna, Austria. Riverkeeper (2018). "Combined Sewage Overflows (CSOs)." . (5/31/18, 2018). Rizzo, L., Manaia, C., Merlin, C., Schwartz, T., Dagot, C., Ploy, M. C., Michael, I., and Fatta-Kassinos, D. (2013). "Urban wastewater treatment plants as hotspots for antibiotic resistant bacteria and genes spread into the environment: a review." Sci Total Environ, 447, 345-360. 167

Rodenburg, L. A., Du, S., Fennell, D. E., and Cavallo, G. J. (2010). "Evidence for widespread dechlorination of polychlorinated biphenyls in groundwater, landfills, and wastewater collection systems." Environmental science & technology, 44(19), 7534-7540. Roeder, R. S., Lenz, J., Tarne, P., Gebel, J., Exner, M., and Szewzyk, U. (2010). "Long- term effects of disinfectants on the community composition of drinking water biofilms." Int J Hyg Environ Health, 213(3), 183-189. Rossi, S., Antonelli, M., Mezzanotte, V., and Nurizzo, C. (2007). "Peracetic acid disinfection: a feasible alternative to wastewater chlorination." Water Environment Research, 79(4), 341-350. Sanchez-Ruiz, C., Martinez-Royano, S., and Tejero-Monzon, I. (1995). "An evaluation of the efficiency and impact of raw wastewater disinfection with peracetic acid prior to ocean discharge." Water Science and Technology, 32(7), 159-166. Schloss, P. D., Westcott, S. L., Ryabin, T., Hall, J. R., Hartmann, M., Hollister, E. B., Lesniewski, R. A., Oakley, B. B., Parks, D. H., and Robinson, C. J. (2009). "Introducing mothur: open-source, platform-independent, community-supported software for describing and comparing microbial communities." Appl. Environ. Microbiol., 75(23), 7537-7541. Segata, N., Izard, J., Waldron, L., Gevers, D., Miropolsky, L., Garrett, W. S., and Huttenhower, C. (2011). "Metagenomic biomarker discovery and explanation." Genome Biology, 12(6), R60. Shibata, T., Solo-Gabriele, H. M., Fleming, L. E., and Elmir, S. (2004). "Monitoring marine recreational water quality using multiple microbial indicators in an urban tropical environment." Water Res, 38(13), 3119-3131. Skold, O. (2000). "Sulfonamide resistance: mechanisms and trends." Drug Resistance Updates, 3(3), 155-160. Sondergaard, A., San Millan, A., Santos-Lopez, A., Nielsen, S. M., Gonzalez-Zorn, B., and Norskov-Lauritsen, N. (2012). "Molecular organization of small plasmids bearing blaTEM-1 and conferring resistance to beta-lactams in Haemophilus influenzae." Antimicrob Agents Chemother, 56(9), 4958-4960. Spellberg, B. (2010). "Antibiotic Resistance: Promoting Critically Needed Antibiotic Research and Development and Appropriate Use(“Stewardship”) of these Precious Drugs." Testimony of the Infectious Diseases Society of America. Sterk, A., Schets, F. M., Husman, R., Maria, A., Nijs, T., and Schijven, J. F. (2015). "Effect of climate change on the concentration and associated risks of Vibrio Spp. in Dutch recreational waters." Risk Analysis, 35(9), 1717-1729. Stewart, P. S. (2002). "Mechanisms of antibiotic resistance in bacterial biofilms." International Journal of Medical Microbiology, 292(2), 107-113. Storteboom, H., Arabi, M., Davis, J. G., Crimi, B., and Pruden, A. (2010). "Tracking antibiotic resistance genes in the South Platte River basin using molecular signatures of urban, agricultural, and pristine sources." Environ. Sci. Technol., 44(19), 7397-7404. Sun, J., Deng, Z., and Yan, A. (2014). "Bacterial multidrug efflux pumps: mechanisms, physiology and pharmacological exploitations." Biochem Biophys Res Commun, 453(2), 254-267. 168

Sun, L., Klein, E. Y., and Laxminarayan, R. (2012). "Seasonality and temporal correlation between community antibiotic use and resistance in the United States." Clin Infect Dis, 55(5), 687-694. Suter, E., Juhl, A. R., and O’Mullan, G. D. (2011). "Particle association of Enterococcus and total bacteria in the lower Hudson River estuary, USA." Journal of Water Resource and Protection, 3, 715. Suter, E., Juhl, A. R., and O’Mullan, G. D. (2011). "Particle association of enterococcus and total bacteria in the lower Hudson River estuary, USA." Journal of Water Resource and Protection, 3(10), 715. Suzuki, M. T., Taylor, L. T., and DeLong, E. F. (2000). "Quantitative analysis of small- subunit rRNA genes in mixed microbial populations via 5′-nuclease assays." Appl. Environ. Microbiol., 66(11), 4605-4614. ten Veldhuis, J. A. E., Clemens, F. H. L. R., Sterk, G., and Berends, B. R. (2010). "Microbial risks associated with exposure to pathogens in contaminated urban flood water." Water Res, 44(9), 2910-2918. The Weather Company LLC "Weather Underground Historical Weather." . (5 January, 2017). Tian, S., Tian, Z., Yang, H., Yang, M., and Zhang, Y. (2017). "Detection of Viable Bacteria during Sludge Ozonation by the Combination of ATP Assay with PMA- Miseq Sequencing." Water, 9(3), 166. Tondera, K., Klaer, K., Gebhardt, J., Wingender, J., Koch, C., Horstkott, M., Strathmann, M., Jurzik, L., Hamza, I. A., and Pinnekamp, J. (2015). "Reducing pathogens in combined sewer overflows using ozonation or UV irradiation." Int J Hyg Environ Health, 218(8), 731-741. Tremblay, J., Singh, K., Fern, A., Kirton, E. S., He, S., Woyke, T., Lee, J., Chen, F., Dangl, J. L., and Tringe, S. G. (2015). "Primer and platform effects on 16S rRNA tag sequencing." Frontiers Microbiol., 6, 771. Turolla, A., Cattaneo, M., Marazzi, F., Mezzanotte, V., and Antonelli, M. (2018). "Antibiotic resistant bacteria in urban sewage: Role of full-scale wastewater treatment plants on environmental spreading." Chemosphere, 191, 761-769. Tutumi, M., Imamura, K., Hatano, S., and Watanade, T. (1973). "Antimicrobial action of peracetic acid." Food Hygiene and Safety Science (Shokuhin Eiseigaku Zasshi), 14(5), 443-447. U.S. Food and Drug Administration (2018). "Bacteriological Analytical Manual Media M76: Lauryl Tryptose Broth." . (5/5/18, 2018). United States Department Of Agriculture (1954). "Diagnosis and improvement of saline and alkali soils." United States Environmental Protection Agency (1978). "Microbiological Methods for Monitoring the Environment Water and Wastes."Environmental Monitoring and Support Laboratory, Cincinnati. United States Environmental Protection Agency (2002). "Method 1604: Total Coliforms and Escherichia coli in Water by Membrane Filtration Using a Simultaneous Detection Technique " Office of Water. 169

United States Environmental Protection Agency (2012). "Alternative Disinfection Methods Fact Sheet: Peracetic Acid." O. o. W. Management, ed. United States Environmental Protection Agency (2016). "Clean Watersheds Needs Survey 2012." United States Environmental Protection Agency, ed. United States Environmental Protection Agency (2016). "Clean Watersheds Needs Survey 2012 Report to Congress." US Environmental Protection Agency (2004). "EPA Test Method 9056D: Soil and waste pH." US Environmental Protection Agency (2011). "Keeping Raw Sewage and Contaminated Stormwater Out of the Public's Water."New York, New York. US Environmental Protection Agency (2018). "Impaired Waters and TMDLs: Statute and Regulations." (5/31/18, 2018). USGS (2018). "Geologic History of Raritan Bay." Geology of National Parks, 3D and Photographic Tours, . (March 15, 2018). Van Abs, D. J. (2014). "Water Infrastructure in New Jersey's CSO Cities: Elevating the Importance of Upgrading New Jersey's Urban Water Systems." Rutgers The State University of New Jersey, School of Environmental and Biological Sciences, 159. Van Abs, D. J. (2014). "Water Infrastructure in New Jersey’s CSO Cities: Elevating the Importance of Upgrading New Jersey’s Urban Water Systems." Rutgers University. Vandewalle, J. L., Goetz, G. W., Huse, S. M., Morrison, H. G., Sogin, M. L., Hoffmann, R. G., Yan, K., and McLellan, S. L. (2012). "Acinetobacter, Aeromonas and Trichococcus populations dominate the microbial community within urban sewer infrastructure." Environ Microbiol, 14(9), 2538-2552. Varma, M., Field, R., Stinson, M., Rukovets, B., Wymer, L., and Haugland, R. (2009). "Quantitative real-time PCR analysis of total and propidium monoazide-resistant fecal indicator bacteria in wastewater." Water Res, 43(19), 4790-4801. Vikesland, P. J., Pruden, A., Alvarez, P. J. J., Aga, D., Burgmann, H., Li, X. D., Manaia, C. M., Nambi, I., Wigginton, K., Zhang, T., and Zhu, Y. G. (2017). "Toward a Comprehensive Strategy to Mitigate Dissemination of Environmental Sources of Antibiotic Resistance." Environ Sci Technol, 51(22), 13061-13069. Wade, T. J., Pai, N., Eisenberg, J. N. S., and Colford, J. M. (2003). "Do U.S. Environmental Protection Agency Water Quality Guidelines for Recreational Waters Prevent Gastrointestinal Illness? A Systematic Review and Meta- analysis." Environmental Health Perspectives, 111(8), 1102-1109. Walsh, T. R., Weeks, J., Livermore, D. M., and Toleman, M. A. (2011). "Dissemination of NDM-1 positive bacteria in the New Delhi environment and its implications for human health: an environmental point prevalence study." The Lancet Infectious Diseases, 11(5), 355-362. Watkins, R. R., Smith, T. C., and Bonomo, R. A. (2016). "On the path to untreatable infections: colistin use in agriculture and the end of 'last resort' antibiotics." Expert Rev Anti Infect Ther, 14(9), 785-788. Welch, A., Awah, C. U., Jing, S., van Veen, H. W., and Venter, H. (2010). "Promiscuous partnering and independent activity of MexB, the multidrug transporter protein from Pseudomonas aeruginosa." Biochem J, 430(2), 355-364. 170

Wilson, M. A., Mohseni, O., Gulliver, J. S., Hozalski, R. M., and Stefan, H. G. (2009). "Assessment of Hydrodynamic Separators for Storm-Water Treatment." Journal of Hydraulic Engineering, 135(5). Wilson, M. A., Mohseni, O., Gulliver, J. S., Hozalski, R. M., and Stefan, H. G. (2009). "Assessment of hydrodynamic separators for storm-water treatment." Journal of Hydraulic Engineering, 135(5), 383-392. Wisconsin State Lab of Hygiene (1993). "ESS Method 340.2: Total Suspended Solids, Mass Balance (Dried at 103–105 C) Volatile Suspended Solids (Ignited at 550_C)."Madison, WI. Wisconsin State Lab of Hygiene (1993). "ESS Method 340.2: Total Suspended Solids, Mass Balance / Volatile Suspended Solids." Environmental Sciences SectionInorganic Chemistry Unit. Woodman, M. E. (2008). "Direct PCR of intact bacteria (colony PCR)." Current Protocols in Microbiology, A. 3D. 1-A. 3D. 6. Yoneda, K., Chikumi, H., Murata, T., Gotoh, N., Yamamoto, H., Fujiwara, H., Nishino, T., and Shimizu, E. (2005). "Measurement of Pseudomonas aeruginosa multidrug efflux pumps by quantitative real-time polymerase chain reaction." FEMS Microbiol Lett, 243(1), 125-131. Yoneda, K., Chikumi, H., Murata, T., Gotoh, N., Yamamoto, H., Fujiwara, H., Nishino, T., and Shimizu, E. (2005). "Measurement of Pseudomonas aeruginosa multidrug efflux pumps by quantitative real-time polymerase chain reaction." FEMS Microbiol. Lett., 243(1), 125-131. Young, S., Juhl, A., and O'Mullan, G. D. (2013). "Antibiotic-resistant bacteria in the Hudson River Estuary linked to wet weather sewage contamination." J. Water Health, 11(2), 297–310. Young, S., Juhl, A., and O'Mullan, G. D. (2013). "Antibiotic-resistant bacteria in the Hudson River Estuary linked to wet weather sewage contamination." Journal of Water and Health, 11(2), 297-310. Yuan, Q. B., Guo, M. T., and Yang, J. (2015). "Fate of antibiotic resistant bacteria and genes during wastewater chlorination: implication for antibiotic resistance control." PLoS One, 10(3), 1-11. Zhang, X.-X., Zhang, T., and Fang, H. (2009). "Antibiotic resistance genes in water environment." Appl. Microbiol. Biotechnol., 82(3), 397-414. Zhang, Y., Snow, D. D., Parker, D., Zhou, Z., and Li, X. (2013). "Intracellular and extracellular antimicrobial resistance genes in the sludge of livestock waste management structures." Environ Sci Technol, 47(18), 10206-10213. Zhang, Y., Zhuang, Y., Geng, J., Ren, H., Xu, K., and Ding, L. (2016). "Reduction of antibiotic resistance genes in municipal wastewater effluent by advanced oxidation processes." Sci Total Environ, 550, 184-191. 171

Appendix A: Shifts in microbial community structure and function in surface waters impacted by unconventional oil and gas waste water revealed by metagenomics Science of the Total Environment 580 (2017) 1205–1213

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Short Communication Shifts in microbial community structure and function in surface waters impacted by unconventional oil and gas wastewater revealed by metagenomics

N.L. Fahrenfeld a,⁎, Hannah Delos Reyes a, Alessia Eramo a,DeniseM.Akobb, Adam C. Mumford b, Isabelle M. Cozzarelli b a Rutgers, The State University of New Jersey, Civil and Environmental Engineering, 96 Frelinghuysen Rd, Piscataway, NJ 08504, United States b U.S. Geological Survey, National Research Program, 12201 Sunrise Valley Dr., Reston, VA 20192, United States

HIGHLIGHTS GRAPHICAL ABSTRACT

• Microbial communities change down- stream from a wastewater disposal fa- cility. • Deltaproteobacteria and Methanomicrobia are in higher abun- dance in affected sediments. • Increases in genes for dormancy and sporulation were found downstream from the facility. • Select efflux pump genes increase downstream but not total antibiotic re- sistance genes.

article info abstract

Article history: Unconventional oil and gas (UOG) production produces large quantities of wastewater with complex geochemistry Received 30 September 2016 and largely uncharacterized impacts on surface waters. In this study, we assessed shifts in microbial community Received in revised form 12 December 2016 structure and function in sediments and waters upstream and downstream from a UOG wastewater disposal facility. Accepted 12 December 2016 To do this, quantitative PCR for 16S rRNA and antibiotic resistance genes along with metagenomic sequencing were Available online 27 December 2016 performed. Elevated conductivity and markers of UOG wastewater characterized sites sampled downstream from Editor: D. Barcelo the disposal facility compared to background sites. Shifts in overall high level functions and microbial community structure were observed between background sites and downstream sediments. Increases in Deltaproteobacteria Keywords: and Methanomicrobia and decreases in Thaumarchaeota were observed at downstream sites. Genes related to dor- Hydraulic fracturing mancy and sporulation and methanogenic respiration were 18–86 times higher at downstream, impacted sites. Dormancy and sporulation The potential for these sediments to serve as reservoirs of antimicrobial resistance was investigated given frequent Antibiotic resistance genes reports of the use of biocides to control the growth of nuisance bacteria in UOG operations. A shift in resistance pro- Biocides files downstream of the UOG facility was observed including increases in acrBandmexB genes encoding for multi- Sediment drug efflux pumps, but not overall abundance of resistance genes. The observed shifts in microbial community structure and potential function indicate changes in respiration, nutrient cycling, and markers of stress in a stream impacted by UOG waste disposal operations. © 2016 Elsevier B.V. All rights reserved.

⁎ Corresponding author at: Rutgers, The State University of New Jersey, 96 Frelinghuysen Rd., Piscataway, NJ 08854, United States. E-mail address: [email protected] (N.L. Fahrenfeld).

http://dx.doi.org/10.1016/j.scitotenv.2016.12.079 0048-9697/© 2016 Elsevier B.V. All rights reserved. 1206 N.L. Fahrenfeld et al. / Science of the Total Environment 580 (2017) 1205–1213

1. Introduction 2015). Recent investigations associated (1) QAC exposure in river sedi- ments with increases in metagenomic signatures of biodegradation Recent technological advances combining horizontal drilling with genes, efflux pumps, cell envelope modification, chaperones, and oxida- hydraulic fracturing have led to a boom in the development of uncon- tive stress (Oh et al., 2013), and (2) glutaraldehyde exposure of bacterial ventional oil and gas (UOG) resources from low permeability forma- isolates in simulated produced waters with osmotic stress, energy pro- tions (Engle et al., 2014). UOG resources include shale gas, coal bed duction and conservation, membrane integrity, and protein transport in methane, and tight oil. UOG production results in large volumes of the transcriptome (Vikram et al., 2014). wastewater (Gregory et al., 2011), and there are potential environmen- This study aimed to determine (1) if inputs of UOG wastewater into tal risks associated with holding, transporting, and disposing of these a surface water resulted in shifts in microbial community structure and fluids. Improper disposal, treatment (Ferrar et al., 2013; Hladik et al., functional potential in the community metagenome; and (2) whether 2014; Volz et al., 2011) and accidental releases remain a concern. UOG impacts led to elevated levels of antibiotic resistance in surface water wastewater is a complex mixture with high total dissolved solids and sediments. To meet these aims, surface water and bed sediments (TDS) (Fontenot et al., 2013; Gaudlip and Paugh, 2008; Lester et al., from an impacted stream adjacent to a Class II underground injection 2013; Murali Mohan et al., 2013a; Murali Mohan et al., 2013b; control (UIC) facility in West Virginia were analyzed using Olmstead et al., 2013; Rowan et al., 2011b; Vidic et al., 2013; Volz metagenomic sequencing and quantitative PCR (qPCR). Impacts to the et al., 2011; Warner et al., 2011), naturally occurring radioactive mate- stream were identified using chemical analyses and published previous- rials (Rowan et al., 2011a; Rowan et al., 2015), organic (Akob et al., ly (Akob et al., 2016). The Class II UIC well is used to dispose of wastewa- 2015; Orem et al., 2014; Strąpoć et al., 2011; Volz et al., 2011), and ter from UOG production, including wastewater from shale gas and coal heavy metal compounds. In the event wastewater and production fluids bed methane wells; additional details about the disposal facility can be are released to the environment [e.g., (Cozzarelli et al., in press; Lauer found in Akob et al. (2016) and the endocrine disrupting activities of et al., 2016)] the impact of these chemical components, including addi- these waters are described in Kassotis et al. (2016). Results presented tives such as biocides used in hydraulic fracturing fluids (Kahrilas et al., in this study provide insight into the impacts of UOG wastewater on 2015), remains unclear. stream microbial community structure and function. The complex biogeochemistry of wastewater generated during UOG production is underscored by shifts in microbial communities and or- ganic and inorganic chemical composition observed across the period 2. Materials and methods of well development (Cluff et al., 2014; Murali Mohan et al., 2013a; Orem et al., 2014; Rowan et al., 2015). Over time, there is an increase 2.1. Sampling in the relative abundance of facultative anaerobes in UOG wastewater and microorganisms that can tolerate high salt concentrations and bio- In June 2014, water and bed sediment samples were collected from cides (Cluff et al., 2014; Liang et al., 2016; Murali Mohan et al., 2013a). tributaries of Wolf Creek in West Virginia, USA, including an unnamed

Viable H2S-producing, fermenting, and methanogenic microorganisms tributary that runs through the UIC disposal facility and a background have been cultured from UOG wastewater (Akob et al., 2015; Daly site in a separate drainage (Fig. 1A). The disposal facility includes the et al., 2016; Liang et al., 2016). Interestingly, these microbes were culti- disposal well, which injects wastewater to 792.5 m below surface, vable despite the use of biocides to control their growth during UOG brine storage tanks, an access road, and (formerly) two small, lined im- production (Johnson et al., 2008). These shifts in microbial community poundment ponds, which were used for temporary storage of wastewa- have been associated with microbial functional changes. Increases ter to allow for settling of particulates prior to injection. The ponds across time in sporulation, dormancy, and stress response and shifts in operated from 2002 to spring 2014 when they were removed and the metabolism of iron and sulfur have been reported (Murali Mohan area re-contoured. A detailed site description is provided in Akob et al. et al., 2014). (2016) including information on the production wells contributing Releases of UOG wastewater to the environment could present an waste to this facility. No records of pre-treatment activities are available. uncharacterized potential secondary impact of the UOG water cycle by Four sites were sampled along the stream that runs through the dis- promoting microbial community functional changes in impacted envi- posal facility: up-gradient, background Site 4, mid Site 6 (adjacent to the ronments. For example, releases of high salinity UOG wastewater into injection well), and down gradient Sites 3 and 7 (downstream from for- freshwater streams would be expected to alter microbial community mer impoundment ponds) (Fig. 1 and Table 1). A stream in a separate structure and function, thereby affecting nutrient cycling. Increases in drainage (“background drainage”, Site 2) with no known oil and gas salinity due to deicing of roads have been shown to disrupt nitrogen cy- wastewater inputs was also sampled to provide an additional control cling, likely due to alterations of microbial communities in roadside soils (Table 1, Fig. 1). Both water and sediment were collected at each sam- and surface waters (Green and Cresser, 2007; Green et al., 2008). UOG pling site. wastewater also contains microbial communities (Akob et al., 2015; Water samples were collected in sterile, one-liter amber glass bottles Cluff et al., 2014; Daly et al., 2016; Liang et al., 2016; Murali Mohan by submerging bottles, filling completely, and capping with minimal et al., 2013b) which may thrive in impacted environments and are likely headspace. A duplicate biological replicate sample was collected at to be adapted to and suited for biodegradation of the organic com- Site 3 and processed separately as an internal control (Table 1)for pounds in wastewater. both sediment and water samples. Water samples were stored on ice Of interest is the impact of chemicals used to enhance oil and gas in the field, shipped in coolers overnight to Rutgers where they were fil- production on changes in metabolism, stress, and other toxic responses ter concentrated [0.22 μm, mixed cellulose ester (Millipore, Billerica, resulting from environmental releases. Biocides are frequently used in MA)], then stored at −20 °C for further analysis. Sediment samples UOG operations to control corrosion, souring, and biofouling and their (~50 mL) were collected using aseptic technique from the upper 5 cm potential for selecting for antibiotic resistant bacteria has been raised of the streambed and frozen on dry ice in the field and during shipping. as a potential secondary impact (Kahrilas et al., 2015). Antibiotic resis- Chemical analyses on water and sediment samples are described in tance is a pressing public health issue and links between environmental Akob et al. (2016). Water was analyzed for alkalinity, cations, anions, reservoirs of resistance and clinical infections have been observed strontium, oxygen and hydrogen isotopes, nonvolatile dissolved organic (Forsberg et al., 2012). A variety of biocides with different mechanisms carbon (NVDOC), trace inorganic elements, and disinfection byproducts. of action may be used in UOG production including lytic quaternary am- Sediment was analyzed for carbon, nitrogen, and sulfur elemental anal- monium/amine compounds (QAC), electrophilic compounds such as ysis, Fe speciation, and total inorganic elements. Sample collection was glutaraldehyde, and oxidizers such as peracetic acid (Kahrilas et al., limited due to the study area being located on private property. N.L. Fahrenfeld et al. / Science of the Total Environment 580 (2017) 1205–1213 1207

Fig. 1. a. Location of study area and b. sampling map of UOG wastewater facility wastewater injection well and the former impoundment ponds (outlined in yellow). Brine storage tanks are also shown and circled in purple. Sampling site labels include conductivity data.

2.2. Microbial analyses obtain sufficient DNA (~200 ng) (Waksman Genomic Core Facility, Rutgers). Paired-end (150 bp) sequencing was performed and generat- DNA was extracted from filter concentrated water samples ed 96 million reads over three separate runs of the same plate (~500 mL) and homogenized sediment samples (~0.5 g, wet weight) (Table S2). Duplicate biological replicate samples collected from Site 3 using a FastDNA Soil Spin Kit (MP Biomedicals, CA, USA) according to (named Site 3 and Site 3d) were processed separately. Sequences manufacturer instructions. Duplicate to quintuplet extracts were pooled were analyzed using the MG-RAST pipeline (Meyer et al., 2008)and from sediments and submitted for Illumina NextSeq500 sequencing to pooled following analysis. Pipeline options including removal of

Table 1 Field parameters, non-volatile dissolved organic carbon (NVDOC) and major anion and cation concentrations of water samples collected in June 2014 in tributaries of Wolf Creek, including a stream adjacent to a UOG wastewater disposal facility. Site locations are indicated in Fig. 1; Dup. refers to a duplicate field sample. Data courtesy of Akob et al. (2016).

Sample Type, location Dups + blanks pH Conductivity NVDOC Cl− Na+ Ba+ Field Fe II (μS/cm) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L)

Site 2 Background, separate drainage 6.51 109 1.42 1.70 7.02 95 0.1 Site 4 Background, upstream from disposal facility 6.47 74.0 1.13 0.88 6.96 136 0.2 Site 6 Adjacent to the injection well shed 6.85 82.0 2.20 1.76 6.86 110 0.8 Site 7 Downstream from former impoundment ponds Dup. 6.36 416 2.49 115 63.4 653 8 Site 3 Downstream from the disposal facility Dup. + field blank 6.09 379 3.24 119 56.0 594 3.5 1208 N.L. Fahrenfeld et al. / Science of the Total Environment 580 (2017) 1205–1213 artificial replicates produced by sequencing artifacts (Gomez-Alvarez 3. Results et al., 2009), dynamic trimming using a modified Dynamic Trim (Cox et al., 2010) for sequences with 5 bp below a 15 phred score. Gene call- Analyses previously described (Table 1)(Akob et al., 2016)demon- ing was performed using FragGeneScan (Rho et al., 2010). Taxonomic strated geochemical impacts consistent with UOG wastewater to down- classifications of sequences identified in the metagenomes were per- stream Sites 7 and 3 due to activities at the disposal facility. Background formed on sequences classified by the lowest common ancestor Sites 2 (located in a separate drainage) and 4 (located upstream from (Huson et al., 2007) with maximum E-value cutoff of 10−5, minimum disposal facility) were not impacted. Site 6, located adjacent to the injec- percent identity 60%, and minimum alignment length cutoff of 15 tion well, was minimally impacted, e.g., had conductivity and major ion amino acids. After pooling sequences from the three runs, sequences concentrations comparable to background sites, but Sr isotopic analysis were sub-sampled randomly (N = 1,658,437 for Bacteria, N = 20,498 suggested small inputs of coalbed methane wastewater (Akob et al., for Archaea) using a custom bootstrap in R to provide an equal number 2016). Sites 7 and 3 were characterized by increased conductivity of sequences for comparing community structure between samples. (Fig. 1) and elevated levels of ions including chloride (two orders of Proteins were annotated using MG-RAST SEED subsystem hierarchical magnitude above background), sodium (ten times above background), classification (max E-value cutoff 10−5, minimum percent identity and barium (five to six times background levels). Based on these obser- 60%, and minimum alignment length cutoff of 15 amino acids) for vations, Sites 7 and 3 were designated as impacted due to the significant Level 1–3 functions as defined by SEED and pooled over the three runs geochemical alterations observed. Akob et al. (2016) described addi- (Overbeek et al., 2005). SEED provides a curated hierarchy for annota- tional geochemical shifts between the impacted and background sites. tions: Level 1 is the highest classification for subsystems related to a structure (e.g., cell wall) or functional process (e.g., respiration, stress 3.1. UOG wastewater impacts on microbial community and function response), Level 3 is similar to a KEGG pathway, and Level 4 is the actual functional assignment. Relative abundance was determined by normal- Significantly different bacterial community structures were ob- izing to total protein annotations for a given sample and these values served between impacted Sites 7 and 3 and between background Sites were compared to those from background Site 2, which was located in 2 and Sites 4, and minimally impacted Site 6 (p = 0.001–0.005, Fig. 2, a separate drainage. with rarefaction curve included as Fig. S1 and operational taxonomic To further investigate the presence of ARG, genes called as proteins unit table as Table S3). Bacterial community structure at Site 6 was in MG-RAST pooled over the three runs were queried against the Anti- more similar to the control sites, consistent with comparable geochem- biotic Resistance Genes Database (ARDB) (Liu and Pop, 2009)using istry between these sites and the findings of Akob et al. (2016), where BlastX with an E-value cutoff of 10−5 (Yang et al., 2013). The threshold 16S rRNA gene amplicon sequencing (with a different DNA extraction for amino acid identity was ≥90% and sequence alignment set to ≥25 method) was performed on paired samples. Bacterial community struc- amino acids (Kristiansson et al., 2011; Zhang et al., 2011). Resulting se- tures from replicate sediment samples collected from Site 3 (samples 3 quences were normalized to total clean reads (sequences passing qual- and 3d) were 98.6% similar. ity control which included dereplication and trimming described Proteobacteria dominated the communities at all sites, with above) per sample and reported as parts per million (ppm, Table S2). Alphaproteobacteria, Betaproteobacteria, Deltaproteobacteria,and Three sequencing runs were performed for each sample and joined Gammaproteobacteria being the major classes represented. prior to data analysis. Sequences are available under accession numbers Deltaproteobacteria were more abundant at impacted sites (9.8 ± 4606751–4606762, 4614547–4614552, and 4635906–4635911 in the 0.1%) compared to background (unimpacted) sites (6.2 ± 1.2%). Unclas- MG-RAST database. sified bacteria were the next most abundant group (21.9–27.1%). Simi- qPCR was performed for select antibiotic resistance genes (ARG) larly the Archaeal community at Site 7 formed a unique cluster from the (acrB, sul1, sul2, ermF, tet(G), tet(O)) on diluted (1:100) water and sed- other sampling sites (Fig. 3), however Site 3 did not have a significantly iment DNA extracts using a SybrGreen protocol (Supplementary data; different Archaeal community structure compared to the background Table S1). All qPCR standard curves were constructed from 10-fold seri- Site 2 in a separate drainage. Site 7 was dominated by the class al dilutions of cloned genes ranging from 108 to 102 gene copies per μL. Methanomicrobia (59.8%). Methanomicrobia were present at higher Samples were analyzed in triplicate (technical replicate) with a stan- (p = 3.8 × 10−8) and unclassified Thaumarchaeota were present at dard curve and negative control included in each run. Details on qPCR lower (p = 4.5 × 10−5) relative abundance at the downstream sites. primers and thermocycling conditions are presented in Table S1. Unimpacted Sites 2, 4, and minimally impacted Site 6 formed a clus- Results were analyzed with Primer 6 statistical software (PrimerE, ter with 99.8% similarity for relative abundance of Subsystem Level 1 Ivy Bridge, UK) and visualized with ggplot2 (Wickham, 2009) from R functions (Fig. S2 with functional rarefaction curve included as statistical software (R Core Team, 2013), SigmaPlot (Systat Software, Fig. S3). Site 3 and 3d formed a single significantly different cluster Inc., San Jose, CA), and Venny (Oliveros, 2007). qPCR values were com- (p = 0.01), as did Site 7. Spikes in select dormancy and sporulation pared between unimpacted (Sites 2 and 4), minimally impacted Site 6 genes (Subsystem Level 3 functions, similar to a KEGG pathway) were and impacted sites (Sites 3 and 3d, 7) using a Wilcoxon test. Correla- observed at Site 7 (Figs. 4, S4) while the levels of these genes at other tions between log-normalized gene copy numbers from the background sites and impacted Site 3 were comparable to Site 2. The el- metagenomic sequencing and conductivity were tested using a Spear- evated dormancy and sporulation annotations observed at Site 7 in- man rank test. Hierarchical clustering analysis using Bray-Curtis similar- clude genes for dipicolinate synthesis (29 times higher than Site 2), ities was performed on log(X + 1) transformed ARG, bacterial small acid-soluble spore proteins (36.1 times higher), spore germina- community, and Archaeal community abundance data from the tion (21.2 times higher), and sporulation cluster III A (86.3 times metagenomic sequencing and significant multivariate structure was de- higher). Likewise, select genes related to methanogenic respiration termined using a similarity profile (SIMPROF, Primer 6) test. Compari- were elevated at Site 7 including methanophenazine hydrogenase and sons of (a) relative abundance of Archaeal phyla and (b) metagenomic coenzyme F420 hydrogenase at 25.5 and 17.8 times higher than back- ARG annotations normalized by clean reads between unimpacted and ground Site 2, respectively (Figs. 4, S5). Cadmium resistance genes downstream sites was performed by a Kruskal Wallis test followed by were found at double the relative abundance at Site 7 compared to a post-hoc pairwise t-test with a Bonferroni correction for multiple background Site 2 (Fig. S6) despite Cd being below detection in waters comparisons. Suites of MG-RAST annotated functions were compared and sediments. Generally, genes related to stress response (Fig. S7), ar- between impacted and unimpacted sites using ANOSIM (Primer 6). Rar- omatic metabolism (Fig. S8), sulfur metabolism (Fig. S9), and nitrogen efaction curves for 16S rRNA genes and functional annotations were metabolism (Fig. S10) at Site 7 and 3 were less than two times greater generated using mothur (v.1.36.1) (Schloss et al., 2009). than the relative abundance in background metagenomes. Select N.L. Fahrenfeld et al. / Science of the Total Environment 580 (2017) 1205–1213 1209

Fig. 2. Heat map illustrates relative abundance of different bacterial classes in sediment with major phyla labeled. Cluster analysis indicated significant shifts in impacted sediment (Site 7 & 3) bacterial community compared to background and unimpacted Site 6. Clusters with significant differences are connected with black bars in the cluster analysis. Numbers on heat map indicate rounded relative abundances (%). antibiotic resistance functions were also elevated (7.9–9.4 times back- indicator of water quality, and ARG was observed for smeE, a fluoroquin- ground Site 2) at Sites 7 and 3 including bacteriocin-like peptides olone resistance gene that also serves as a multidrug efflux pump (Fig. S11). (Spearman rank p = 0.017). No relationships were observed between other observed ARGs and conductivity as an indicator of water quality 3.2. Antibiotic resistance assigned by ARDB (p = 0.13–1.0). The most frequently detected resistance function was multidrug re- Forty-three different resistance genes were identified in the sistance with an average relative abundance across sites of 3.1 ± metagenomic sequencing data of sediment samples using the ARDB 1.0 ppm, followed by bacitracin with an average relative abundance of (Fig. 5). The most frequently detected genes across the sites were 1.6 ± 0.5 ppm (Fig. 6). Eight of 43 ARGs were above detection at all bacA, mexB, mexF, and acrB. The highest levels of acrB and mexB were sites, and 11 were unique to Site 3 (Fig. S12). observed at downstream sites (p = 0.0003–0.03). No difference be- tween bacAandmexF levels were observed comparing unimpacted 3.3. Quantification of select ARG and downstream sites (both p = 1.0). Cluster analysis performed on re- sistance gene profiles indicated a relationship between resistance pat- qPCR for select ARG on water and sediment samples showed that terns and sample geographic location. Two of the impacted sites (Sites sul1, tet(G), and acrB were above detection for all samples (Fig. S13 7 and 3) formed a cluster with 66.8% similarity. Background (Sites 2 and Table S5). No differences in qPCR gene copy numbers were ob- and 4) and the site located adjacent to the injection well, which was served between the sites (all p = 0.2–0.8). The qPCR analysis detected minimally impacted (Site 6), formed a cluster with 71.1% similarity. Se- ARG that were not observed in the metagenomes, implying qPCR was quences from duplicate samples from Site 3 (3 and 3d) formed a cluster more sensitive than the sequencing at the depth performed. For exam- with 70% similarity. A relationship between conductivity, used as an ple, in the metagenomic results sul1 was above detection in two 1210 N.L. Fahrenfeld et al. / Science of the Total Environment 580 (2017) 1205–1213

Fig. 3. a. Cluster analysis and b. heat map illustrating the relative abundance (%) of Archaea counts classified at the class level using the lowest common ancestor method in sediment. Numbers overlaid on the heat map are the relative abundance (%). samples (Tables S4) but for qPCR these genes were found at relatively 4. Discussion constant levels in all samples (Fig. S13 and Table S5). Other resistance genes tested (sul2, tet(O), ermF) were below detection or quantification Microbial community composition and potential metabolic activity via qPCR and in the metagenomes. were altered in stream sediments due to activities at a UOG wastewater disposal facility that resulted in geochemical changes to the stream. The alterations in impacted stream sediments indicated potential shifts in nutrient cycling and redox conditions characterized by the loss of nitrate-oxidizing Nitrospira, decreases in ammonia oxidizing Thaumarchaeota, and increases in anaerobic Methanomicrobia.Inaddi- tion, members of Firmicutes were elevated at the impacted Site 7 (5.3%) compared to Site 3 (2.2 ± 0.7%) and unimpacted sites (0.7 ± 0.1%). Organisms within the Firmicutes phylum are abundant in high sa- linity UOG produced waters from the Marcellus Shale and other UOG formations (Akob et al., 2015; Cluff et al., 2014; Liang et al., 2016; Murali Mohan et al., 2013a; Struchtemeyer and Elshahed, 2012). UOG wastewater potentially contains microbial communities representative of both production fluids (chemical additives and water taken from a variety of sources) and formation waters (Cluff et al., 2014; Daly et al., 2016; Murali Mohan et al., 2013a), which are different from surficial stream sediment communities. However, the microbial community shifts in the impacted sediments are likely a result of UOG wastewater selecting for different communities in stream sediments due to chemi- cal changes given the high dilution of the UOG wastewater in the stream Fig. 4. Comparison of relative abundance of Level 3 functions in sediment samples for select dormancy and sporulation and respiration genes to background Site 2. Error bars (0.001 part brine to 0.999 parts fresh water) based on previously report - - on Site 3 represent replicate samples from Site 3. Br and Cl concentrations (Akob et al., 2016). Further, UOG wastewater N.L. Fahrenfeld et al. / Science of the Total Environment 580 (2017) 1205–1213 1211

Fig. 6. a. Relative abundance and b. percentage of antibiotic resistance elements in sediment by antibiotic target annotated using the ARDB. Specific conductance measured in water shown on second y-axis in panel a.

potential function of communities due to impacts from the disposal fa- cility. Again, this is likely due to the changes in geochemistry due to the inputs of UOG wastewater. Increases in spore forming bacteria (Murali Mohan et al., 2013a) and in genes for these functions (Murali Mohan et al., 2014) have been observed over time in produced waters. At Site 7, the Firmicutes phylum contained spore forming Bacillus and Clostridia, which accounted for 10.6% of the microbial community. Increases in nitrosative stress at impacted sites are consistent with observed impacts to nitrogen cycling in soils impacted by road salt application (Green and Cresser, 2007; Green et al., 2008). The observed shifts in the microbial community and these taxonomic and functional populations were asso- ciated with shifts in overall functional annotations. However, genes ele- Fig. 5. a. Cluster analysis of resistance gene profiles for ARDB annotated results and b. heat vated at Site 7 were often not elevated at Site 3, despite these both being map of relative abundance of antibiotic resistance genes, normalized to total clean reads in impacted sites. This could be explained by the geochemical variability sediment. Site locations are shown in Fig. 1. Sites 3d and 3 are replicate samples from Site 3. observed at the two impacted sites, which includes slightly lower con- ductivity and lower concentrations of elements associated with UOG wastewater at Site 3 compared to Site 7. These observations may also after nine days of production was dominated by Gammaproteobacteria be impacted by the small sample size used in this study. However, (Murali Mohan et al., 2014), but this signal is not retained in our surficial Akob et al. present further sediment analysis with site replicates dem- stream samples indicating that selection and not input of bacteria is onstrating that the differences between Sites 3 and 7 are due to geo- likely driving community dynamics. An increase in Deltaproteobacteria chemical impacts and not likely from spatial heterogeneity in at impacted sites is characteristic of polluted environments and shifts sediment based on similar C/N/S profiles at both sites. towards anaerobic conditions. The observation of Methanomicrobia A shift in antibiotic resistance profiles in sediments was observed in dominating the Archaea community at Site 7 is consistent with observa- downstream samples. No biocides identified in a recent review tions of these as the sole Archaea observed in unaerated and biocide (Kahrilas et al., 2015) share the polycyclic and carboxylic chemical amended impoundments for UOG produced waters (Murali Mohan structures characteristic of fluoroquinolones. While UOG wastewater et al., 2013b). Similar to the 16S rRNA gene amplicon sequencing previ- can be rich in organic chemicals including polycyclic aromatic hydrocar- ously reported, shifts in microbial community composition correlated bons and heterocyclic compounds (Khan et al., 2016; Orem et al., 2014), with changes in geochemistry in impacted stream sites. However, our elevated NVDOC was not observed at the impacted sites. The highest study goes beyond this previous work to (1) present the microbial com- relative abundance of multidrug resistance genes was observed at Site munity structure using a sequencing method without the potential PCR 7, and many of these resistance elements encoded for efflux pumps. biases in amplicon sequences, (2) provide the structure of the Archaeal Multidrug efflux pumps transport a range of chemicals including surfac- community, and (3) show that the potential function of these commu- tants, which can be used as an additive for hydraulic fracturing, and nities is further altered due to activities at the site. therefore increases in the presence of efflux pumps may represent a re- Increases in select sporulation, dormancy and methanogenesis sponse to other chemicals in the UOG wastewater matrix. It is also pos- genes were observed at the impacted Site 7, showing a shift in the sible these shifts in resistance profiles are due to shifts in the microbial 1212 N.L. Fahrenfeld et al. / Science of the Total Environment 580 (2017) 1205–1213 community (Martinez et al., 2015). For example, increases in acrBand community indicate a need to understand the microbial ecosystem im- mexB in impacted sediments may be characteristic of community shifts pacts from accidental releases or improper treatment of UOG wastewa- that include increases of E. coli and Pseudomonas aeruginosa (Martinez ter. Functional shifts in UOG wastewater impacted sediments may have et al., 2015). However, Gammaproteobacteria were not elevated above implications for the treatment and beneficial reuse of UOG wastewater. background (3.7 ± 0.8%) at impacted Site 7 (3.7%) or Site 3 (4.7 ± In particular, while elevated levels of select multidrug efflux pumps 0.2%). High relative abundance of efflux pumps has also been observed were observed in UOG impacted sediments the ARG relative abun- in saline sediments, with acrBandmexB accounting for 8 and 4% of total dances observed here were comparable to saline sediments impacted ARG in sediments with anthropogenic impacts (Chen et al., 2013). by anthropogenic activity (Chen et al., 2013), both of which were The impact of biocide use in UOG on environmental antibiotic resis- lower than municipal wastewater, which is considered a hot spot of re- tance and the fate of biocides in the UOG water cycle is poorly character- sistance. Measurements of functional shifts in a range of UOG wastewa- ized (Kahrilas et al., 2015). In this study, specificbiocideswerenot ter from different sources and biocide usages is needed to understand detected in impacted stream waters (Keith Loftin, personal communica- the full impact of UOG wastes on environmental antibiotic resistance tion). However, disclosure of biocides in the FracFocus.org database and ecosystem function. (Groundwater Protection Council and Interstate Oil and Gas)provides an indication of the biocides used in UOG development for shale gas Funding which is a source of wastewater for disposal at this site [site history was previously described (Akob et al., 2016)]. Using compiled lists of This research was supported by a research stipend to HDR from the biocides in use for UOG development, one can generate a list of expected Douglass Project, graduate fellowship to AE from Rutgers University resistances if biocide use was impacting the resistance profile of the School of Engineering, NLF university startup funds, and funding to community. Interestingly, resistance genes directly related to QAC DMA, IMC, and ACM from the U.S. Geological Survey Toxic Substances (qacA, qacB, qacC), which are commonly used by the UOG industry, Hydrology Program. were all below detection in all samples. This could be explained either Any use of trade, firm, or product names is for descriptive purposes by lack of QAC use in the wastewater disposed of at the facility, degrada- only and does not imply endorsement by the U.S. Government. tion or sorptive losses of QAC, or induction of biocide tolerance through other mechanisms. Biodegradation and/or abiotic transformations of Appendix A. Supplementary data QAC have been demonstrated under aerobic, nitrate and fermenting conditions, but not methanogenic conditions (Tezel, 2009). It is also Supplementary data to this article can be found online at http://dx. possible that complex geochemical conditions in UOG wastewater af- doi.org/10.1016/j.scitotenv.2016.12.079. fected bacterial response to any biocides present in the wastewater, as was observed in a study of the biocide glutaraldehyde on model organ- References ism Pseudomonas fluorescens (Vikram et al., 2014). Given the hypothesis that UOG wastewater could serve as a selecting Akob, D.M., Cozzarelli, I.M., Dunlap, D.S., Rowan, E.L., Lorah, M.M., 2015. Organic and inor- ganic composition and microbiology of produced waters from Pennsylvania shale gas environment for environmental antibiotic resistance, it is important to wells. Appl. Geochem. 60, 116–125. compare the ARG profiles and levels observed in Wolf Creek to other re- Akob, D.M., Mumford, A.C., Orem, W.H., Engle, M.A., Klinges, J.G., Kent, D.B., et al., 2016. ported environments, particularly those reported to be hot spots of an- Wastewater disposal from unconventional oil and gas development degrades stream quality at a West Virginia injection facility. Environ. Sci. Technol. 50, 5517–5525. tibiotic resistance and therefore potential risks to public health, like Chen, B., Yang, Y., Liang, X., Yu, K., Zhang, T., Li, X., 2013. Metagenomic profiles of antibi- municipal wastewater. ARG relative abundance for our sites was com- otic resistance genes (ARGs) between human impacted estuary and deep ocean sed- parable to levels observed in the saline, anthropogenic impacted sedi- iments. Environ. Sci. Technol. 47, 12753–12760. – Cluff, M.A., Hartsock, A., MacRae, J.D., Carter, K., Mouser, P.J., 2014. Temporal changes in ments of the Pearl River Estuary (2.8 7 ppm) (Chen et al., 2013). microbial ecology and geochemistry in produced water from hydraulically fractured However, more ARG types were observed at in our sites (43 ARG) Marcellus shale gas wells. Environ. Sci. Technol. 48, 6508–6517. than in the Pearl River Estuary (27 ARG) (Chen et al., 2013), which Cox, M.P., Peterson, D.A., Biggs, P.J., 2010. SolexaQA: at-a-glance quality assessment of – was dominated by sulfonamide resistance (38%). Generally lower ARG Illumina second-generation sequencing data. BMC Bioinf. 11, 1 6. Cozzarelli, I.M., Skalak, K.J., Kent, D.B., Engle, M.A., Benthem, A., Mumford, A.C., et al., 2016. levels on a ppm basis and fewer ARG types were detected in this Environmental Signatures and Effects of a Spill of Untreated Oil and Gas Wastewater study compared to metagenomic studies in municipal wastewater in the Williston Basin, North Dakota. Sci. Total Environ. (in press). using the ARDB (Yang et al., 2013; Yang et al., 2014; Zhang et al., Daly, R.A., Borton, M.A., Wilkins, M.J., Hoyt, D.W., Kountz, D.J., Wolfe, R.A., et al., 2016. Mi- crobial metabolisms in a 2.5-km-deep ecosystem created by hydraulic fracturing in 2011). These differences are likely due to the fact that municipal waste- shales. Nat. Microbiol. 1, 16146. water is a stronger source of ARG and that the waste inputs in the Pearl Engle, M.A., Cozzarelli, I.M., Smith, B.D., 2014. USGS Investigations of Water Produced River Estuary study were chemically different than the UOG wastewa- During Hydrocarbon Reservoir Development (No. 2014-3104). US Geological Survey. Ferrar, K.J., Michanowicz, D.R., Christen, C.L., Mulcahy, N., Malone, S.L., Sharma, R.K., 2013. ter. However, differences in DNA extraction method (Morgan et al., Assessment of effluent contaminants from three facilities discharging Marcellus shale 2010), and bioinformatics pipeline can affect results in –omics studies wastewater to surface waters in Pennsylvania. Environ. Sci. Technol. 47, 3472–3481. and care should be taken when comparing between studies using differ- Fontenot, B.E., Hunt, L.R., Hildenbrand, Z.L., Carlton Jr., D.D., Oka, H., Walton, J.L., et al., 2013. An evaluation of water quality in private drinking water wells near natural ent techniques. gas extraction sites in the Barnett shale formation. Environ. Sci. Technol. 47, 10032–10040. 5. Conclusions Forsberg, K.J., Reyes, A., Wang, B., Selleck, E.M., Sommer, M.O.A., Dantas, G., 2012. The shared antibiotic resistome of soil bacteria and human pathogens. Science 337, 1107–1111. This work demonstrates that UOG wastewater releases are associat- Gaudlip, A.W., Paugh, L.O., 2008. Marcellus Shale Water Management Challenges in ed with a shift in microbial community structure and functions related Pennsylvania (No. SPE 119898). Society of Petroleum Engineers Shale Gas Production to dormancy and sporulation, respiration, and antimicrobial resistance, Conference, Irving, TX. Gomez-Alvarez, V., Teal, T.K., Schmidt, T.M., 2009. Systematic artifacts in metagenomes which is hypothesized to lead to alterations in microbial activity in situ. from complex microbial communities. ISME J. 3, 1314–1317. Therefore, it provides further evidence of microbiological impacts at this Green, S.M., Cresser, M.S., 2007. Nitrogen cycle disruption through the application of de- – UOG disposal site. The potential for further downstream impacts re- icing salts on upland highways. Water Air Soil Pollut. 188, 139 153. Green, S.M., Machin, R., Cresser, M.S., 2008. Effect of long-term changes in soil chemistry mains to be demonstrated given that samples here were collected on induced by road salt applications on N-transformations in roadside soils. Environ. the disposal facility property. The increases in dormancy and shifts in Pollut. 152, 20–31. respiration are characteristic of microbial communities responding to Gregory, K.B., Vidic, R.D., Dzombak, D.A., 2011. Water management challenges associated with the production of shale gas by hydraulic fracturing. Elements 7, 181–186. environmental disturbances, which in this study were characterized Groundwater Protection Council and Interstate Oil and Gas. Frac Focus Chemical Disclo- by shifts in the geochemistry. The observed shifts in the microbial sure Registry. (Online) Available: fracfocus.org. N.L. Fahrenfeld et al. / Science of the Total Environment 580 (2017) 1205–1213 1213

Hladik, M.L., Focazio, M.J., Engle, M., 2014. Discharges of produced waters from oil and gas Orem, W., Tatu, C., Varonka, M., Lerch, H., Bates, A., Engle, M., et al., 2014. Organic sub- extraction via wastewater treatment plants are sources of disinfection by-products to stances in produced and formation water from unconventional natural gas extraction receiving streams. Sci. Total Environ. 466–467, 1085–1093. in coal and shale. Int. J. Coal Geol. 126, 20–31. Huson, D.H., Auch, A.F., Qi, J., Schuster, S.C., 2007. MEGAN analysis of metagenomic data. Overbeek, R., Begley, T., Butler, R.M., Choudhuri, J.V., Chuang, H.-Y., Cohoon, M., et al., Genome Res. 17, 377–386. 2005. The subsystems approach to genome annotation and its use in the project to Johnson, K., French, K., Fichter, J., Oden, R., 2008. Use of microbiocides in Barnett shale gas annotate 1000 genomes. Nucleic Acids Res. 33, 5691–5702. well fracturing fluids to control bacteria related problems. Corrosion (NACE Interna- R Core Team, 2013. R: A Language and Environment for Statistical Computing. R Founda- tional, 2008). tion for Statistical Computing, Vienna, Austria. Kahrilas, G.A., Blotevogel, J., Stewart, P.S., Borch, T., 2015. Biocides in hydraulic fracturing Rho, M., Tang, H., Ye, Y., 2010. FragGeneScan: predicting genes in short and error-prone fluids: a critical review of their usage, mobility, degradation, and toxicity. Environ. Sci. reads. Nucleic Acids Res. 38, e191. Technol. 49, 16–32. Rowan, E.L., Engle, M.A., Kirby, C.S., Kraemer, T.F., 2011a. Radium content of oil- and gas- Kassotis, C.D., Iwanowicz, L.R., Akob, D.M., Cozzarelli, I.M., Mumford, A.C., Orem, W.H., et field produced waters in the northern Appalachian basin (USA): summary and al., 2016. Endocrine disrupting activities of surface water associated with a West discussion of data. U.S. Geological Survey Scientific Investigations Report, Virginia oil and gas industry wastewater disposal site. Sci. Total Environ. 557, pp. 5135–5166. 901–910. Rowan, E.L., Engle, M.A., Kraemer, T.F., 2011b. Formation waters from reservoirs in the Khan, N.A., Engle, M., Dungan, B., Holguin, F.O., Xu, P., Carroll, K.C., 2016. Volatile-organic Appalachian basin, New York, Pennsylvania, and West Virginia: observations on molecular characterization of shale-oil produced water from the Permian basin. major and minor element, and radium geochemistry. GSA Annual Meeting in Chemosphere 148, 126–136. Minneapolis. Kristiansson, E., Fick, J., Janzon, A., Grabic, R., Rutgersson, C., Weijdegård, B., et al., 2011. Rowan, E.L., Engle, M.A., Kraemer, T.F., Schroeder, K.T., Hammack, R.W., Doughten, M.W., Pyrosequencing of antibiotic-contaminated river sediments reveals high levels of re- 2015. Geochemical and isotopic evolution of water produced from Middle Devonian sistance and gene transfer elements. PLoS One 6, e17038. Marcellus shale gas wells, Appalachian basin, Pennsylvania. AAPG Bull. 99, 181–220. Lauer, N.E., Harkness, J.S., Vengosh, A., 2016. Brine spills associated with unconventional Schloss, P.D., Westcott, S.L., Ryabin, T., Hall, J.R., Hartmann, M., Hollister, E.B., et al., 2009. oil development in North Dakota. Environ. Sci. Technol. 50, 5389–5397. Introducing mothur: open-source, platform-independent, community-supported Lester, Y., Yacob, T., Morrissey, I., Linden, K.G., 2013. Can we treat hydraulic fracturing software for describing and comparing microbial communities. Appl. Environ. flowback with a conventional biological process? The case of guar gum. Environ. Microbiol. 75, 7537–7541. Sci. Technol. Lett. 1, 133–136. Strąpoć, D., Mastalerz, M., Dawson, K., Macalady, J., Callaghan, A.V., Wawrik, B., et al., Liang, R., Davidova, I.A., Marks, C.R., Stamps, B.W., Harriman, B.H., Stevenson, B.S., et al., 2011. Biogeochemistry of microbial coal-bed methane. Annu. Rev. Earth Planet. Sci. 2016. Metabolic capability of a predominant Halanaerobium sp. in hydraulically frac- 39, 617–656. tured gas wells and its implication in pipeline corrosion. Front. Microbiol. 7, 988. Struchtemeyer, C.G., Elshahed, M.S., 2012. Bacterial communities associated with hydrau- Liu, B., Pop, M., 2009. ARDB—antibiotic resistance genes database. Nucleic Acids Res. 37, lic fracturing fluids in thermogenic natural gas wells in north Central Texas, USA. D443–D447. FEMS Microbiol. Ecol. 81, 13–25. Martinez, J.L., Coque, T.M., Baquero, F., 2015. What is a resistance gene? Ranking risk in Tezel, U., 2009. Fate and Effect of Quaternary Ammonium Compounds in Biological Sys- resistomes. Nat. Rev. Microbiol. 13, 116–123. tems. ProQuest, Georgia Institute of Technology. Meyer, F., Paarmann, D., D'Souza, M., Olson, R., Glass, E., Kubal, M., et al., 2008. The Vidic, R.D., Brantley, S.L., Vandenbossche, J.M., Yoxtheimer, D., Abad, J.D., 2013. Impact of metagenomics RAST server - a public resource for the automatic phylogenetic and shale gas development on regional water quality. Science 340. functional analysis of metagenomes. BMC Bioinf. 9, 386. Vikram, A., Lipus, D., Bibby, K., 2014. Produced water exposure alters bacterial response to Morgan, J.L., Darling, A.E., Eisen, J.A., 2010. Metagenomic sequencing of an in vitro simu- biocides. Environ. Sci. Technol. 48, 13001–13009. lated microbial community. PLoS One 5, e10209. Volz, C.D., Ferrar, K., Michanowicz, D., Christen, C., Kearney, S., Kelso, M., et al., 2011. Con- Murali Mohan, A., Bibby, K.J., Lipus, D., Hammack, R.W., Gregory, K.B., 2014. The functional taminant characterization of effluent from the Pennsylvania Brine Treatment Inc. potential of microbial communities in hydraulic fracturing source water and pro- Josephine Facility Being Released Into Blacklick Creek, Indiana County, PA. EPA Hy- duced water from natural gas extraction characterized by metagenomic sequencing. draulic Fracturing Study Technical Workshop 3, Fate and Transport, March 28–29. PLoS One 9, e107682. Warner, N., Osborn, S., Jackson, R., Vengosh, A., 2011. Boron and strontium isotopes as Murali Mohan, A., Hartsock, A., Bibby, K.J., Hammack, R.W., Vidic, R.D., Gregory, K.B., sensitive tracers for indicating potential shallow groundwater contamination from 2013a. Microbial community changes in hydraulic fracturing fluids and produced Marcellus formation brines. GSA Annual Meeting in Minneapolis. water from shale gas extraction. Environ. Sci. Technol. 47, 13141–13150. Wickham, H., 2009. ggplot2: Elegant Graphics for Data Analysis. Springer, New York. Murali Mohan, A., Hartsock, A., Hammack, R.W., Vidic, R.D., Gregory, K.B., 2013b. Microbial Yang, Y., Li, B., Ju, F., Zhang, T., 2013. Exploring variation of antibiotic resistance genes in communities in flowback water impoundments from hydraulic fracturing for recov- activated sludge over a four-year period through a metagenomic approach. Environ. ery of shale gas. FEMS Microbiol. Ecol. 86, 567–580. Sci. Technol. 47, 10197–10205. Oh, S., Tandukar, M., Pavlostathis, S.G., Chain, P.S., Konstantinidis, K.T., 2013. Microbial Yang, Y., Li, B., Zou, S., Fang, H.H.P., Zhang, T., 2014. Fate of antibiotic resistance genes in community adaptation to quaternary ammonium biocides as revealed by sewage treatment plant revealed by metagenomic approach. Water Res. 62, 97–106. metagenomics. Environ. Microbiol. 15, 2850–2864. Zhang, T., Zhang, X.X., Ye, L., 2011. Plasmid metagenome reveals high levels of antibiotic Oliveros, J.C., 2007. VENNY. An Interactive Tool for Comparing Lists With Venn Diagrams. resistance genes and mobile genetic elements in activated sludge. PLoS One 6, Olmstead, S.M., Muehlenbachs, L.A., Shih, J.-S., Chu, Z., Krupnick, A.J., 2013. Shale gas de- e26041. velopment impacts on surface water quality in Pennsylvania. Proc. Natl. Acad. Sci. 110, 4962–4967.