Occurrence and Control of Microbial Contaminants of Emerging Concern through the Urban Water Cycle: Molecular Profiling of Opportunistic and Antibiotic Resistance

Emily Dawn Garner

Dissertation submitted to the faculty of the Virginia Polytechnic Institute and State University in partial fulfillment of the requirements for the degree of

Doctor of Philosophy In Civil Engineering

Amy Pruden, Chair Marc A. Edwards Leigh-Anne H. Krometis Brian D. Badgley

February 22, 2018 Blacksburg, VA

Keywords: opportunistic pathogens, antibiotic resistance, stormwater, drinking water, distribution system, wastewater reclamation, direct potable reuse

Copyright 2018 Occurrence and Control of Microbial Contaminants of Emerging Concern through the Urban Water Cycle: Molecular Profiling of Opportunistic Pathogens and Antibiotic Resistance

Emily Dawn Garner

ABSTRACT

In an era of pervasive water stress caused by population growth, urbanization, drought, and climate change, limiting the dissemination of microbial contaminants of emerging concern (MCECs) is of the utmost importance for the protection of public health. In this dissertation, two important subsets of MCECs, opportunistic pathogens (OP) and antibiotic resistant genes (ARG), are studied across several compartments of the urban water cycle, including surface water, stormwater, wastewater, recycled water, and potable water. Collectively, this dissertation advances knowledge about the occurrence of OPs and ARGs across these water systems and highlights trends that may be of value in developing management strategies for limiting their regrowth and transmission.

Field studies of two surface water catchments impacted by stormwater runoff demonstrated the prevalence of ARGs in urban stormwater compared to pristine, unimpacted sites, or to days when no precipitation was recorded. The role of wastewater reuse in transmitting OPs and ARGs was also investigated. Traditional tertiary wastewater treatment plants producing water for non- potable use were found to be largely ineffective at removing ARGs, but plants using advanced oxidation processes or ozonation paired with biofiltration to produce direct potable reuse water were highly effective at removing ARGs. Non-potable reclaimed water consistently had greater quantities of sul1, a sulfonamide ARG, and and Mycobacterium, two OPs of significant public health concern, present than corresponding potable systems. Limited regrowth of OPs and ARGs did occur in simulated premise (i.e., building) plumbing systems operated with direct potable reuse waters, but regrowth was comparable to that observed in systems fed with potable water derived from surface or groundwater. Advancements were also made in understanding the role of several hypothesized driving forces shaping the antibiotic resistome in natural and engineered water systems: selection by antimicrobials and other compounds, horizontal gene transfer, and microbial community composition. Finally, whole-genome and metagenomic characterization were applied together towards profiling L. pneumophila in clinical and water samples collected from Flint, Michigan, where an economically-motivated switch to an alternative water source created conditions favorable for growth of this organism and likely triggered one of the largest Legionnaires’ Disease outbreaks in U.S. history.

Occurrence and Control of Microbial Contaminants of Emerging Concern through the Urban Water Cycle: Molecular Profiling of Opportunistic Pathogens and Antibiotic Resistance

Emily Dawn Garner

GENERAL AUDIENCE ABSTRACT

Population growth, urbanization, drought, and climate change have all driven many U.S. municipalities to utilize alternative water sources, such as recycled wastewater, to offset demand on traditional potable water sources. Many water providers have moved towards a modern paradigm of utilizing multiple available water sources, recognizing the interconnectedness of various components of the urban water cycle, leading to opportunities to improve sustainability, optimize infrastructure use, stimulate economic growth, increase coordination among water agencies, and identify new water resources from which to meet consumer needs. Though advancements in treatment technologies throughout the twentieth century have largely succeeded in eliminating waterborne disease outbreaks associated with contamination of municipal water supplies by fecal pathogens in developed countries, several microbial contaminants of emerging concern (MCECs) have garnered attention.

Two major groups of MCECs are considered in this dissertation: antibiotic resistance, including antibiotic resistant (ARB) and antibiotic resistance genes (ARG), and opportunistic pathogens (OP), such as , the causative agent of Legionnaires’ Disease. ARB are a rising cause of disease around the world and are a major challenge to modern medicine because they make antibiotics used for treatment ineffective. OPs, the leading cause of waterborne disease in the U.S. and other developed countries, have become prevalent in engineered water systems where low nutrient concentrations, warm water temperatures, and long stagnation times can facilitate their growth. Immunocompromised people, including smokers and the elderly, are especially vulnerable to infection with OPs. The role of the urban water cycle in facilitating the spread of these MCECs is not well understood. Here they were studied across several compartments of the urban water cycle, including surface water, stormwater, wastewater, recycled water (spanning a variety of intended uses, from non-potable to direct potable reuse), and potable water.

Field studies were conducted of two watersheds impacted by stormwater runoff, one in the arid Colorado Front Range under conditions of a rare, 1-in-1,000 year rainfall event, and one in the humid climate of southwest Virginia, during three summer storms. Both studies demonstrated the prevalence of ARGs in urban stormwater compared to pristine, unimpacted sites, or to days when no precipitation was recorded.

The role of wastewater reuse in transmitting OPs and ARGs was also investigated. Wastewater treatment plants producing water for non-potable use (i.e. applications such as irrigation, but not for human consumption) were found to be largely inefficient at removing ARGs, and this reclaimed water consistently had greater quantities of the sul1 ARG present than in corresponding potable systems. In these systems, genes associated with the OPs Legionella and

Mycobacterium as well as total bacteria were more abundant in reclaimed water than in corresponding potable systems. In more advanced treatment plants utilizing advanced oxidation processes or ozonation paired with biofiltration to produce direct potable reuse water (i.e. water fit for human consumption), ARGs were very effectively removed by treatment, with abundances often found to be higher in corresponding potable waters derived from surface or groundwater. Limited regrowth of ARGs as well as OPs did occur in simulated home plumbing systems operated with these waters, but regrowth was comparable to that observed in systems fed with potable water derived from surface or groundwater.

Finally, a study of L. pneumophila in the Flint, Michigan drinking water system during use of an alternative water source that has been identified as a likely cause of two Legionnaires’ Disease outbreaks revealed presence of multiple strains of the OP in the system. Genomic comparisons revealed that strains isolated from hospital and residential water samples were highly similar to clinical strains associated with the outbreaks.

Advancements were also made in understanding the role of several hypothesized driving forces in shaping the antibiotic resistome in natural and engineered water systems: selection by antimicrobials and other compounds, horizontal gene transfer, and microbial community composition. Together, these chapters describe an advancement in knowledge regarding the occurrence of OPs and ARGs in a variety of water systems, and highlight trends that may be of value in developing management strategies for limiting regrowth or transmission of these bacteria in various compartments of the urban water cycle.

iv

ACKOWLEDGEMENTS

I would like to express sincere gratitude to my advisor, Dr. Amy Pruden, for her mentorship, guidance, and support. I would also like to thank Dr. Marc Edwards for his encouragement and support of my growth both as an engineer and researcher and as a person. You have both set a tremendous example of compassionate and dedicated researchers and I have been so privileged to work with you on projects that make improve people’s lives.

I would also like to thank Dr. Leigh-Anne Krometis and Dr. Brian Badgley for their support and valuable feedback.

I would like to acknowledge all of the financial support that made this dissertation possible, provided by the National Science Foundation, The Alfred P. Sloan Foundation Microbiology of the Built Environment Program, the Water Environment Research Foundation, the Virginia Water Resources Research Center, the Virginia Tech Institute for Critical Technology and Applied Science Center for Science and Engineering of the Exposome, and the Virginia Tech College of Agriculture and Life Sciences Integrated Grants Program. Thank you as well to the Charles E. Via family, the American Water Works Association Abel Wolman Fellowship, and the National Science Foundation Graduate Research Fellowship for supporting my work.

To the current and former Pruden and Edwards groups members, thank you for teaching me so much and for allowing me to be a member of an incredible team. Thank you all for your encouragement and friendship.

Finally, I would like to thank my friends and family. Thank you to everyone who has made Blacksburg feel like home. Thank you to my family, Mom, Dad, and Lindsay, for teaching me to love nature and to always seek to help others, the convictions that led me to become an environmental engineer. Finally thank you to my husband, Aaron, for your unending patience and for always encouraging me to pursue my dreams. Thank you all for your love and support.

v

TABLE OF CONTENTS

ABSTRACT…...... ii GENERAL AUDIENCE ABSTRACT...... iii ACKNOWLEDGEMENTS…...... v TABLE OF CONTENTS…...... vi LIST OF FIGURES…...... x LIST OF TABLES…...... xi CHAPTER 1 : INTRODUCTION ...... 1 OVERVIEW AND RESEARCH MOTIVATION ...... 1 MICROBIAL CONTAMINANTS OF EMERGING CONCERN ...... 2 Antibiotic Resistance Genes ...... 2 Opportunistic Pathogens ...... 2 RESEARCH OBJECTIVES ...... 3 ANNOTATED DISSERTATION OUTLINE AND ATTRIBUTIONS ...... 3 REFERENCES ...... 7 CHAPTER 2 : A HUMAN EXPOSOME FRAMEWORK FOR GUIDING RISK MANAGEMENT AND HOLISTIC ASSESSMENT OF RECYCLED WATER QUALITY .... 10 ABSTRACT ...... 10 INTRODUCTION ...... 10 UNIQUE ASPECTS OF RWDS DESIGN, OPERATION, AND WATER USE ...... 12 Routes of Exposure ...... 14 Physical and Operational Issues ...... 14 IMPORTANT CHEMICAL DIFFERENCES ANTICIPATED BETWEEN RECYCLED AND POTABLE WATER DISTRIBUTION SYSTEMS ...... 16 Organic Matter ...... 16 Redox zones and degradation of water quality ...... 20 Disinfectant residual ...... 20 CHRONIC CONTAMINANTS...... 21 ARGS, OPS, AND OTHER EMERGING MICROBIAL CONCERNS ...... 23 Opportunistic Pathogens ...... 24 Antibiotic Resistance Genes ...... 24 Viruses ...... 26 Amoebae ...... 27 Algae ...... 28 CONCLUSION ...... 28 ACKNOWLEDGEMENTS ...... 30 REFERENCES ...... 30 CHAPTER 3 : STORMWATER LOADINGS OF ANTIBIOTIC RESISTANCE GENES IN AN URBAN STREAM ...... 39 ABSTRACT ...... 39 INTRODUCTION ...... 39 MATERIALS AND METHODS ...... 41 Site and storm descriptions ...... 41 Sample collection and DNA extraction ...... 41 Molecular analysis and high throughput sequencing ...... 41 Data analysis ...... 42

vi

RESULTS AND DISCUSSION ...... 43 Selection of ARG Targets for Characterizing Storm Loadings ...... 43 Gene loading rates and intra-storm variability...... 43 Event loading rates ...... 45 Association with fecal indicator bacteria and environmental variables ...... 47 Diversity and richness of the resistome ...... 48 CONCLUSIONS...... 51 ACKNOWLEDGEMENTS ...... 51 REFERENCES ...... 51 CHAPTER 4 : METAGENOMIC PROFILING OF HISTORIC COLORADO FRONT RANGE FLOOD IMPACT ON DISTRIBUTION OF RIVERINE ANTIBIOTIC RESISTANCE GENES ...... 58 ABSTRACT ...... 58 INTRODUCTION ...... 58 MATERIALS AND METHODS ...... 59 Sample Collection and Preservation ...... 59 Quantification of ARGs ...... 61 Quantification of Antibiotics and Metals ...... 61 16S rRNA Gene Amplicon Sequencing and Metagenomic Analysis ...... 61 Statistical Analyses ...... 62 RESULTS AND DISCUSSION ...... 62 Metagenomic analysis reveals shift in ARG profile following extreme flooding event ...... 62 Potential for selection pressure indicated by co-occurrence of ARGs and antibiotics ...... 64 Potential for co-selective pressures exerted by heavy metals ...... 67 Metagenomic scaffold associations reveals probable ARGs susceptible to co-resistance ... 67 Role of horizontal gene transfer in shaping the resistome ...... 68 Role of phylogeny in shaping the resistome ...... 69 CONCLUSIONS...... 70 ACKNOWLEDGEMENTS ...... 70 REFERENCES ...... 70 SUPPLEMENTARY INFORMATION FOR CHAPTER 4 ...... 75 CHAPTER 5 : METAGENOMIC CHARACTERIZATION OF ANTIBIOTIC RESISTANCE GENES IN FULL-SCALE RECLAIMED WATER DISTRIBUTION SYSTEMS AND CORRESPONDING POTABLE SYSTEMS ...... 84 ABSTRACT ...... 84 INTRODUCTION ...... 84 METHODS ...... 86 Site description, sample collection, and preservation ...... 86 Water chemistry ...... 87 Quantification of antibiotic resistance genes ...... 88 Shotgun metagenomics and 16S rRNA amplicon sequencing ...... 88 Statistical Analysis ...... 89 RESULTS AND DISCUSSION ...... 89 Metagenomic characterization of the resistome in reclaimed versus potable water ...... 89 Abundance of target ARGs in water and biofilms ...... 91 Associations between ARG abundance and microbial ecological factors ...... 93

vii

Potential for horizontal gene transfer ...... 95 Associations between water chemistry and ARGs ...... 97 Implications for ARG dissemination via reclaimed water ...... 99 ACKNOWLEDGEMENTS ...... 100 REFERENCES ...... 100 SUPPLEMENTARY MATERIAL FOR CHAPTER 5 ...... 108 Antibiotic Analysis ...... 108 Quantification of antibiotic resistance genes ...... 108 CHAPTER 6 : MICROBIAL ECOLOGY AND WATER CHEMISTRY IMPACT REGROWTH OF OPPORTUNISTIC PATHOGENS IN FULL-SCALE RECLAIMED WATER DISTRIBUTION SYSTEMS ...... 124 ABSTRACT ...... 124 INTRODUCTION ...... 124 METHODS ...... 126 Site description, sample collection, and preservation ...... 126 Water Chemistry ...... 126 Quantification of OPs ...... 127 16S rRNA gene amplicon sequencing ...... 128 Shotgun metagenomic sequencing ...... 128 Statistical Analyses ...... 128 RESULTS AND DISCUSSION ...... 128 Overview of surveyed distribution systems ...... 128 Physicochemical water characteristics ...... 129 Occurrence of OP Gene Markers ...... 129 Occurrence of OP Gene Markers in Biofilms ...... 130 Exploration of other potential OPs using Shotgun Metagenomics ...... 132 Relationship between abundance of OPs, water age, and related factors ...... 133 Relationship between water chemistry measurements and abundance of OPs ...... 135 Microbial ecology – OP associations ...... 136 Corrosion-Associated Microbial Activity Assays ...... 137 Implications for OP control in reclaimed distribution systems ...... 138 ACKNOWLEDGEMENTS ...... 139 REFERENCES ...... 139 SUPPLEMENTARY INFORMATION FOR CHAPTER 6 ...... 145 CHAPTER 7 : IMPACT OF BLENDING FOR DIRECT POTABLE REUSE ON PREMISE PLUMBING MICROBIAL ECOLOGY AND REGROWTH OF OPPORTUNISTIC PATHOGENS AND ANTIBIOTIC RESISTANT BACTERIA ...... 156 ABSTRACT ...... 156 INTRODUCTION ...... 156 MATERIALS AND METHODS ...... 158 Rig design and operation ...... 158 Water chemistry ...... 159 Culturing ...... 159 Quantitative polymerase chain reaction ...... 159 16S rRNA gene amplicon sequencing and shotgun metagenomics...... 161 Statistical Analysis ...... 161

viii

RESULTS AND DISCUSSION ...... 161 Comparison of regrowth in simulated premise plumbing rigs...... 161 Microbial community composition of regrowth ...... 163 Regrowth of OPPPs ...... 166 Occurrence of ARGs ...... 167 Regrowth of HPC bacteria capable of growth on antibiotic-supplemented media ...... 168 Microbially-influenced corrosion ...... 170 Water chemistry ...... 171 CONCLUSIONS...... 172 ACKNOWLEDGEMENTS ...... 174 REFERENCES ...... 174 SUPPLEMENTARY INFORMATION FOR CHAPTER 7 ...... 178 Pipe rig pre-testing ...... 178 Simulated treatment ...... 178 Quantitative polymerase chain reaction ...... 178 Data analysis for 16S rRNA gene amplicon sequencing and shotgun metagenomics ...... 179 References ...... 179 CHAPTER 8 : WHOLE GENOME SEQUENCE COMPARISON OF CLINICAL AND DRINKING WATER LEGIONELLA PNEUMOPHILA ISOLATES ASSOCIATED WITH THE FLINT WATER CRISIS...... 185 ABSTRACT ...... 185 INTRODUCTION ...... 186 MATERIALS AND METHODS ...... 187 Study Site Description ...... 187 Sample Collection and Preservation ...... 187 Whole genome sequencing of L. pneumophila isolates ...... 188 Shotgun metagenomic sequencing ...... 189 RESULTS ...... 189 Legionella Isolate characterization ...... 190 Annotation of Shotgun Metagenomic Sequences for Identification of Other Putative Pathogens ...... 192 DISCUSSION ...... 193 ACKNOWLEDGEMENTS ...... 197 REFERENCES ...... 198 SUPPLEMENTARY INFORMATION FOR CHAPTER 8 ...... 201 CHAPTER 9 : CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE WORK ...... 209

ix

LIST OF FIGURES

Figure 2-1: Key aspects of the exposome paradigm for managing RWDS ...... 12 Figure 2-2: Overview of typical normalized composition and potential magnitude of dissolved organic matter (DOM) in drinking water sources compared to recycled water sources .. 18 Figure 2-3: Processes by which antibiotic resistant bacteria and opportunistic pathogens (OPs) can re-grow in RWDSs and relevant exposure routes ...... 25 Figure 3-1: ARG abundance with respect to Stroubles Creek discharge...... 43 Figure 3-2: ARG relative abundances during storm phases ...... 44 Figure 3-3: Cumulative ARG storm loading distributions ...... 45 Figure 3-4: Average ARG storm event loading and corresponding equivalent background period loading...... 46 Figure 3-5: Distribution of ARGs by class in baseline (composite n=3) and peak (n=1) runoff storm samples determined by shotgun metagenomic sequencing ...... 50 Figure 4-1: Poudre River sampling sites ...... 60 Figure 4-2: Metagenomic characterizations of ARGs in Poudre River samples ...... 63 Figure 4-3: Beta Diversity plots of microbial community phylogenetic composition ...... 64 Figure 4-4: Spearman’s Rank Correlation Coefficient between abundance of ARGs and antibiotics or metals ...... 66 Figure 4-5: Co-occurrence of ARGs, MRGs, and genetic markers linked to mobile genetic elements on assembled scaffolds ...... 68 Figure 5-1: Metagenomic characterization of ARGs by antibiotic class ...... 91 Figure 5-2: Network analysis depicting co-occurrence of ARGs among each other as well as with plasmid gene markers on assembled scaffolds ...... 97 Figure 6-1: Average relative abundance of DNA fragments matching additional OPs of interest identified via shotgun metagenomic sequencing ...... 133 Figure 6-2: Temperature, free chlorine, and abundances of 16S rRNA genes, Legionella spp., and Mycobacterium spp. in select distribution systems...... 135 Figure 6-3: Microbial community composition of potable vs. reclaimed distribution system samples ...... 137 Figure 7-1: qPCR abundances of 16S rRNA genes, OPs, and ARGs...... 162 Figure 7-2: Microbial community profiles for simulated premise plumbing pipe rigs ...... 165 Figure 7-3: Shotgun metagenomic abundances of ARGs by antibiotic class ...... 169 Figure 7-4: Organic carbon measurements in water prior to incubation in simulated premise plumbing pipe rigs ...... 173 Figure 8-1: Single nucleotide polymorphism (SNP) analysis of Legionella pneumophila isolates ...... 192 Figure 8-2: Comparison of shotgun metagenomic DNA sequence reads obtained from a cross section of Flint tap water samples ...... 194

x

LIST OF TABLES

Table 2-1: Water quality as a function of treatment process ...... 13 Table 2-2: Overview of non-traditional routes of exposure for recycled water and putative risk of infection or exposure...... 15 Table 2-3: Comparing water quality of typical drinking water vs different recycled water applications ...... 17 Table 2-4: Proposed threshold values to achieve biostability in drinking water distribution systems ...... 19 Table 2-5: Case studies of existing application of advanced treatment processes for intended reuse purposes ...... 22 Table 3-1: Spearman’s rank correlation coefficients between ARGs, fecal indicator bacteria, and physicochemical water quality parameters ...... 48 Table 5-1: Overview of surveyed potable and reclaimed systems ...... 87 Table 5-2: Frequency of qPCR detection and abundance of ARGs ...... 94 Table 6-1: Overview of surveyed potable and reclaimed systems ...... 127 Table 6-2: Frequency of qPCR detectiona for 16S rRNA and opportunistic genes ... 131 Table 6-3: Spearman’s rank correlation coefficients for correlations between 16S rRNA or opportunistic pathogen gene markers and physicochemical water quality parameters .. 134 Table 7-1: Blending scenarios, blending water source, treatment, disinfectants, and blending location tested for each utility ...... 160 Table 7-2: Abundnaces of microorganisms associated with microbially-influenced corrosion 171 Table 8-1: Summary of isolates by sequence type (ST), serogroup (SG), and sample origin .... 191

xi

CHAPTER 1 : INTRODUCTION

OVERVIEW AND RESEARCH MOTIVATION

Engineering of water infrastructure has facilitated many of the greatest advancements of modern society with respect to protecting public health and providing convenient and reliable access to water resources. A stark decrease of approximately 40% in mortality rates in the early twentieth century has been largely attributed to the application of water treatment technologies, such as chlorination and filtration, for removing the microorganisms responsible for typhoid, , and in water.1 As water treatment became commonplace in the U.S., advancements in the engineering of distribution system infrastructure have facilitated the delivery of safe water to consumers’ homes. While these engineering advancements have been critical in addressing the most imminent threats to public health associated with drinking water, most being pathogens of fecal origin, new challenges have arisen regarding the emergence of new contaminants and society’s ability to procure sustainable water resources.

Population growth, urbanization, drought, and climate change have all driven many U.S. municipalities to utilize alternative water sources, such as recycled wastewater, to offset demand on traditional potable water sources.2,3 Of the 32 billion gallons of wastewater produced in the U.S. each day, only approximately 7-8% is reused.2 In addition, de facto reuse, or the use of a potable source water that is impacted by upstream wastewater discharges, has become increasingly widespread. A study of 25 U.S. utilities demonstrated that under-low flow conditions, potable water supplies consisted of between 7 and 100% of flow resulting from upstream wastewater discharges.4 Given the prevalence of de facto reuse as well as the emergence of advanced water treatment technologies, the strict division of water resources into categories such as surface water, groundwater, stormwater, wastewater, recycled water, and drinking water is becoming antiquated and fails to provide an appropriately nuanced characterization of most water resources. A more holistic, integrated system of considering these resources that better accounts for the complexities of water quality as well as the opportunities associated with modern water resources is needed.

The Water Research Foundation has proposed the “One Water” paradigm for describing water resources, which is an “integrated planning and implementation approach to managing finite resources for long-term resilience and reliability, meeting both community and ecosystem needs”.5 This approach highlights the interconnectedness of various components of the urban water cycle, leading to opportunities to improve sustainability, optimize infrastructure use, stimulate economic growth, increase coordination among agencies, and identify new water resources from which to meet consumer needs.5 Accordingly, this dissertation applies the “One Water” framework for understanding and addressing challenges associated with microbial contaminants of emerging concern (MCECs). Two major sub-groups of MCECs are addressed in this dissertation: indicators of antibiotic resistance, including antibiotic resistant bacteria (ARB) and antibiotic resistance genes (ARG) and opportunistic pathogens (OP). Here these MCECs are examined across many aspects of the interconnected “One Water” cycle, including surface water, stormwater, wastewater, recycled water, and potable water. In particular, recycled water is emphasized and a spectrum of recycled water practices are considered ranging from non-potable reuse (i.e., use of treated wastewater to meet non-potable demand such as for irrigation) to direct potable reuse (DPR; i.e. highly treated wastewater intended for direct human consumption).

1

MICROBIAL CONTAMINANTS OF EMERGING CONCERN

Antibiotic Resistance Genes

While the application of filtration and disinfection in modern water treatment has largely addressed the challenges associated with traditional waterborne pathogens associated with fecal contamination of water resources, new challenges for controlling the spread of microbial diseases have arisen. In particular, growing attention has been focused on the potential for the urban water cycle to disseminate ARB and their associated ARGs.6–8 Excretion of ARB and ARGs that pass through the human and animal gut, along with residual unmetabolized antibiotics, into municipal wastewater and to agricultural waste streams creates opportunities for dissemination of antibiotic resistance to downstream users. Antibiotic resistance is a pressing public health concern, responsible for at least two million infections and 23,000 deaths in the U.S. annually.9 Thus, it is critical to understand the role of the urban water cycle in disseminating ARGs as well as to identify approaches to limit such propagation.

While limiting transmission of resistant human pathogens is of the utmost importance, autochthonous bacteria carrying ARGs should also be considered, as these could constitute reservoirs of ARGs in the environment that could subsequently be transferred to pathogens via horizontal gene transfer.10–12 The potential for horizontal gene transfer of ARGs between live cells (i.e., conjugation), via bacteriophage infection (i.e., transduction), or via assimilation of extracellular DNA (i.e., natural transformation) presents a unique challenge for water treatment, as traditional treatment goals, such as the inactivation of , may not be sufficient to limit dissemination of ARGs.13,14 In addition, numerous compounds relevant to water can exhibit selective properties in favor of ARB. Residual antibiotics, heavy metals, herbicides, nanoparticles, and disinfectants are all likely to be present at some stage of the urban water cycle and have all been shown to select for or correlate with ARB or ARG in the environment.15–22 Even at sub-lethal concentrations of these compounds likely to occur in some water environments, selection of ARB as well as stimulation of horizontal gene transfer can occur.22–28 The unique challenges associated with the fate and transport of ARGs in water environments warrant research into the predominant mechanisms governing behavior of these contaminants as well as into strategies to limit their dissemination.

Opportunistic Pathogens

Another class of microorganisms has emerged over recent decades as a key contributor to waterborne disease. OPs are a class of microorganisms that are native aquatic bacteria and, unlike traditional waterborne pathogens, are not associated with fecal contamination.29 OPs are thought to be the primary source of waterborne outbreaks in developed countries, with Legionella pneumophila alone responsible for more drinking water-associated outbreaks than any other pathogen in the U.S. since its surveillance began in 2001.30,31 Other common waterborne OPs include Acanthamoeba polyphaga, Naegleria fowleri, Acinetobacter baumanni, Mycobacterium avium, Burkholderia pseudomallei, Stenotrophomonas maltophilia, , and Aspergillus fumigatus. OPs tend to grow in engineered water systems and therefore cannot be controlled through water treatment alone, but their control relies on factors such as distribution system operation, maintenance of a secondary disinfectant residual, and premise plumbing

2

characteristics and usage.32 Engineered water distribution systems conveying either potable or recycled water have several characteristics that facilitate the growth of OPs as water travels from the treatment plant to the point of use. OPs are oligotrophic organisms, capable of growing in low nutrient concentrations typical of these water systems.29,33 Many OPs are resistant to disinfection and they tend to grow well in biofilms where they are protected from unamenable conditions.29,34 Other water chemistry parameters can also contribute to the growth of OPs in distribution systems; for example, elevated iron has been linked with regrowth of L. pneumophila.35–37 In addition, many OPs have complex ecological relationships with other water microorganisms. Among these relationships are competition, antagonism, and obligate parasite-host interactions. Thus, understanding the broader microbial community present in water systems and its influence on the presence of OPs is of interest.34

Another notable characteristic of OPs is that they tend to infect via exposure routes other than ingestion. L. pneumophila, A. baumanni, M. avium, B. pseudomallei, S. maltophilia, and A. fumigatus can infect hosts' lungs via inhalation of aerosols;38–40 P. aeruginosa can infect via the bloodstream, eyes, ears, skin, or lungs;41 and A. polyphaga can cause infection of the eyes or central nervous system following inhalation or penetration of skin lacerations.42 These non- ingestion exposure routes of OPs are particularly important given that non-potable recycled water is often used for irrigation, cooling, and other applications that may result in aerosolization, thus creating opportunities for exposure via inhalation. In addition, recycled water can be used for snowmaking, irrigation of athletic and recreational facilities, and other applications that can result in dermal contact, creating further opportunities for infection or colonization of human hosts by OPs.43

RESEARCH OBJECTIVES

The aim of the research described herein was to characterize the specific routes of dissemination and factors contributing to the propagation of MCECs (i.e. OPs and ARGs) through an integrated, “One Water” perspective of water management. The specific objectives pursued were to:

1. Investigate the role of stormwater in transporting ARGs in surface water catchments, 2. Assess the role of wastewater reuse in disseminating ARGs, 3. Characterize the capacity of recycled wastewater to support growth of opportunistic pathogens in distribution systems and premise plumbing, and 4. Assess the growth of opportunistic pathogens in a compromised potable water system.

ANNOTATED DISSERTATION OUTLINE AND ATTRIBUTIONS

Chapter 1: Introduction

This chapter details the motivation for the research described herein and provides context for the specific research objectives addressed in this dissertation.

Chapter 2: A human exposome framework for guiding risk management and holistic assessment of recycled water quality

3

This manuscript is a critical review examining the need for a more comprehensive framework for use in assessing the public health risks associated with recycled water use. This review explores important distinctions between traditional potable water and recycled water in terms of chemical composition and its ability to support regrowth of microorganisms in distribution systems and premise plumbing. This manuscript emphasizes the need to monitor water quality at the point of use and to consider non-ingestion routes of exposure. It also outlines the characteristics of ARB and ARGs that make them well suited for growth in recycled water, but these same characteristics are also relevant to surface water, stormwater, and potable water, as well.

This manuscript has been published: Garner, E., Zhu, N., Strom, L., Edwards, M., Pruden, A. (2016). A human exposome framework for guiding risk management and holistic assessment of recycled water quality. Environ. Sci.: Water Res Technol. 2:580-598.

Attributions: Chapter 2 was co-first authored with Ni Zhu. While the entire manuscript was written collaboratively, Zhu led the authorship of the section of the manuscript titled “Important chemical differences anticipated between recycled and potable water distribution systems,” while I led “ARGs, OPs, and other emerging microbial concerns.” Co-author Laurel Strom contributed to the discussion of free-living amoebae. Marc Edwards and Amy Pruden contributed guidance on formulation of the critical commentary and assistance in manuscript preparation and review.

Chapter 3: Stormwater loading of antibiotic resistance genes in an urban stream

Chapter 3 addresses objective (1) by systematically exploring the loading of ARGs associated with stormwater in Stroubles Creek, located in Blacksburg, VA. Five ARGs (two sulfonamide: sul1 and sul2; two tetracycline: tet(O) and tet(W); and one macrolide: erm(F)) were monitored in Stroubles Creek during three storm events and compared to baseline concentrations to assess the extent to which stormwater runoff contributes to ARG dissemination in surface water. Physicochemical and hydrometeorological factors were also measured to identify factors contributing to ARG dissemination. Shotgun metagenomic sequencing was applied to a subset of samples to investigate the breadth of the resistome (i.e. the full complement of known resistance genes), transcending the limitations traditionally associated with molecular monitoring of ARGs.

This manuscript has been published: Garner, E., Benitez, R., von Wagoner, E., Sawyer, R., Shaberg, E., Hession, W. C., Krometis, L. A. H., Badgley, B. D., Pruden, A. (2017). Stormwater loading of antibiotic resistance genes in an urban stream. Water Research. 123:144-152.

Attributions: I conducted all analysis of samples, analyzed data, and led writing of the manuscript for this chapter. Romina Benitez, Emily von Wagoner, Richard Sawyer, and Erin Schaberg collected samples. W. Cully Hession, Leigh Anne Krometis, and Brian Badgley contributed to the experimental design, supervised field work, and assisted with manuscript preparation and review. Amy Pruden contributed guidance on molecular applications and assisted with manuscript preparation and review.

4

Chapter 4: Metagenomic profiling of historic Colorado Front Range flood impact on distribution of riverine antibiotic resistance genes

Chapter 4 further addresses objective (1) by seizing the opportunity to monitor ARGs in the Cache la Poudre River in Northern Colorado before and after historic flooding, as well as after 10 months of recovery post-flood. In addition, antibiotics and metals were also monitored to investigate the role of such compounds in potential selection for ARB and ARGs. Horizontal gene transfer was also explored as a potential mechanism contributing to dissemination of ARGs. Shotgun metagenomic sequencing was again used to surpass the limitations of traditional molecular analysis and allow identification of all known ARGs in collected samples.

This manuscript has been published: Garner, E., Wallace, J. S., Argoty, G. A., Wilkinson, C., Fahrenfeld, N., Heath, L., Zhang, L., Arabi, M., Aga, D. S., Pruden, A. (2016). Metagenomic profiling of historic Colorado Front Range flood impact on distribution of riverine antibiotic resistance genes. Scientific Reports. 6:38432.

Attributions: I coordinated collection of samples, conducted all molecular analyses, analyzed data, and led the writing of this chapter. Joshua Wallace and Diana Aga conducted analysis of antibiotics and metals. Gustavo Argoty, Lenwood Heath, and Liqing Zhang assisted with shotgun metagenomic data analysis. Caitlin Wilkinson and Nicole Fahrenfeld contributed to sample collection and analysis. Mazdak Arabi supervised sample collection. Amy Pruden provided guidance on experimental design and data interpretation, and assisted in manuscript preparation and review.

Chapter 5: Metagenomic characterization of antibiotic resistance genes in full-scale reclaimed water distribution systems and corresponding potable systems

Chapter 5 addresses research objective (2) and describes a survey of four full-scale non-potable reclaimed water distribution systems. In addition to monitoring ARGs both at the treatment plant and at five points of use in each system, potential for selection by antibiotics and metals was explored. Horizontal gene transfer was also considered as a mechanism for propagation of ARGs in reclaimed water distribution systems. This manuscript is currently being reviewed for publication in Environmental Science & Technology.

Attributions: I managed coordination among utilities for this project, planned and facilitated sample collections conducted by utilities, conducted all molecular analysis of samples, analyzed data, and led the writing of this chapter. Co-authors for this manuscript are Chaoqi Chen, Kang Xia, Jolene Bowers, David Engalthaler, Jean McLain, Marc Edwards, and Amy Pruden. Chen and Xia conducted analysis of antibiotics. Bowers, Engalthaler, McLain, Edwards, and Pruden contributed to the experimental design and data interpretation, as well as manuscript preparation and review.

Chapter 6: Microbial ecology and water chemistry impact regrowth of opportunistic pathogens in full-scale reclaimed water distribution systems

5

This manuscript describes an investigation focused on objective (3). In this study, the samples collected in the study described in chapter 5 were further analyzed for the presence of OP gene markers. The role of the microbial interactions between OPs and the rest of the microbial community was investigated, as was the role of water chemistry in contributing to regrowth of OPs during distribution. This manuscript is currently being prepared for submission to Environmental Science & Technology.

Attributions: I managed coordination among utilities for this project, planned and facilitated sample collections conducted by utilities, conducted all molecular analysis of samples, analyzed data, and led the writing of this chapter. Co-authors for this manuscript are Jean McLain, Jolene Bowers, David Engalthaler, Marc Edwards, and Amy Pruden. McLain, Bowers, Engalthaler, Edwards, and Pruden contributed to the experimental design and data interpretation, as well as manuscript preparation and review.

Chapter 7: Impact of blending for direct potable reuse on premise plumbing microbial ecology and regrowth of opportunistic pathogens and antibiotic resistant bacteria

This manuscript further explores objectives (2) and (3) by investigating OPS and ARGs in simulated premise plumbing for direct potable reuse systems. This chapter outlines a study of the abundance of ARGs and OPs after simulated use of premise plumbing with water derived from direct potable reuse. Four utilities exploring potential application of DPR provided treated wastewater for bench- or pilot-scale treatment simulated direct potable reuse. DPR waters were blended with each utility’s traditional potable water (surface or groundwater) prior to simulated premise plumbing use. The role of microbial ecology and water chemistry in contributing to regrowth of OPs and ARB/ARGs were also considered. This manuscript is currently being prepared for submission to Water Research.

Attributions: I coordinated collection of samples, conducted all chemical, culture-based and molecular-based analyses, analyzed data, and led the writing of this chapter. Co-authors for this chapter are Mandu Inyang, Elisa Garvey, Jeffrey Parks, Eric Dickerson, Justin Sutherland, Andrew Salveson, Marc Edwards, and Amy Pruden. Inyang and Dickerson operated premise plumbing rigs and collected on-site data. Parks constructed the rigs. Garvey, Sutherland, and Salveson contributed to the experimental design and coordinated management of the project. Edwards and Pruden contributed to the experimental design, data interpretation, and preparation of the manuscript.

Chapter 8: Whole genome sequence comparison of clinical and drinking water Legionella pneumophila isolates associated with the Flint Water Crisis

This manuscript addresses objective (4) by studying a full-scale potable water distribution system experiencing microbial upset leading to propagation of an OP, Legionella pneumophila. This chapter details a genomic characterization of clinical and water Legionella isolates obtained from the city of Flint, Michigan following the Flint Water Crisis, in which use of an alternative water source likely created conditions favorable for growth of Legionella. This manuscript is currently being prepared for submission to Environmental Health Perspectives.

6

Attributions: I coordinated all sequencing, conducted data analysis, and led the writing of this chapter. Co-authors for this chapter are Connor Brown, David Otto Schwake, William J. Rhoads, Gustavo Arango-Argoty, Liqing Zhang, Guillaume Jospin, David Coil, Jonathan Eisen, Marc Edwards, and Amy Pruden. Brown and Schwake contributed to sample collection and isolation of L. pneumophila. Arango-Argoty, Zhang, Jospin, Coil, and Eisen contributed to the bioinformatics analysis. Edwards and Pruden contributed to the experimental design and data interpretation, as well as manuscript preparation and review.

Chapter 9: Conclusions and Recommendations for Future Work

This chapter synthesizes findings and summarizes the contribution of this research to the field of environmental engineering. Recommendations for future research are also presented.

REFERENCES

(1) Cutler, D.; Miller, G. The Role of Public Health Improvements in Health Advances: The Twentieth-Century United States Demography 2018, 42 (1), 1–22. (2) USEPA, Guidelines for Water Reuse, United States Environmental Protection Agency, 2012. (3) Gosling, S. N.; Arnell, N. W. A global assessment of the impact of climate change on water scarcity. Clim. Change 2016, 134 (3), 371–385. (4) Rice, J.; Wutich, A.; Westerhoff, P. Assessment of de facto wastewater reuse across the U.S.: Trends between 1980 and 2008. Environ. Sci. Technol. 2013, 47, 11099–11105. (5) Water Research Foundation. Blueprint for One Water; 2017. (6) Pruden, A.; Pei, R. T.; Storteboom, H.; Carlson, K. H. Antibiotic resistance genes as emerging contaminants: Studies in northern Colorado. Environ. Sci. Technol. 2006, 40 (23), 7445–7450. (7) Wellington, E. M. H.; Boxall, A. B.; Cross, P.; Feil, E. J.; Gaze, W. H.; Hawkey, P. M.; Johnson-Rollings, A. S.; Jones, D. L.; Lee, N. M.; Otten, W.; et al. The role of the natural environment in the emergence of antibiotic resistance in gram-negative bacteria. Lancet Infect. Dis. 2013, 13 (2), 155–165. (8) Ashbolt, N. J.; Amezquita, A.; Backhaus, T.; Borriello, P.; Brandt, K. K.; Collignon, P.; Coors, A.; Finley, R.; Gaze, W. H.; Heberer, T.; et al. Human Health Risk Assessment (HHRA) for Environmental Development and Transfer of Antibiotic Resistance. Environ. Health Perspect. 2013, 121 (9), 993–1001. (9) US CDC, Antibiotic Resistance Threats in the United States, 2013. (10) Graham, D. W.; Olivares-Rieumont, S.; Knapp, C. W.; Lima, L.; Werner, D.; Bowen, E. Antibiotic Resistance Gene Abundances Associated with Waste Discharges to the Almendares River near Havana, Cuba. Environ. Sci. Technol. 2011, 45 (2), 418–424. (11) Lupo, A.; Coyne, S.; Berendonk, T. U. Origin and evolution of antibiotic resistance: the common mechanisms of emergence and spread in water bodies. Front. Microbiol. 2012, 3, 1–13. (12) Kristiansson, E.; Fick, J.; Janzon, A.; Grabic, R.; Rutgersson, C.; Weijdegard, B.; Soderstrom, H.; Larsson, D. G. J. Pyrosequencing of Antibiotic-Contaminated River Sediments Reveals High Levels of Resistance and Gene Transfer Elements. PLoS One 2011, 6 (2), e17038.

7

(13) Von Wintersdorff, C. J. H.; Penders, J.; Van Niekerk, J. M.; Mills, N. D.; Majumder, S.; Van Alphen, L. B.; Savelkoul, P. H. M.; Wolffs, P. F. G. Dissemination of antimicrobial resistance in microbial ecosystems through horizontal gene transfer. Frontiers in Microbiology. 2016, 7 (173). (14) Dodd, M. C. Potential impacts of disinfection processes on elimination and deactivation of antibiotic resistance genes during water and wastewater treatment. J. Environ. Monit. 2012, 14 (7), 1754–1771. (15) Sörme, L.; Lagerkvist, R. Sources of heavy metals in urban wastewater in Stockholm. Sci. Total Environ. 2002, 298, 131–145. (16) Brar, S. K.; Verma, M.; Tyagi, R. D.; Surampalli, R. Y. Engineered nanoparticles in wastewater and wastewater sludge – Evidence and impacts. Waste Manag. 2010, 30 (3), 504–520. (17) Karumathil, D. P.; Yin, H.-B.; Kollanoor-Johny, A.; Venkitanarayanan, K. Effect of chlorine exposure on the survival and antibiotic gene expression of multidrug resistant in water. Int. J. Environ. Res. Public Health 2014, 11 (2), 1844– 1854. (18) Shi, P.; Jia, S.; Zhang, X. X.; Zhang, T.; Cheng, S.; Li, A. Metagenomic insights into chlorination effects on microbial antibiotic resistance in drinking water. Water Res. 2013, 47 (1), 111–120. (19) Huang, J.-J.; Hu, H.-Y.; Wu, Y.-H.; Wei, B.; Lu, Y. Effect of chlorination and ultraviolet disinfection on tetA-mediated tetracycline resistance of . Chemosphere 2013, 90 (8), 2247–2253. (20) Calomiris, J. J.; Armstrong, J. L.; Seidler, R. J. Association of Metal Tolerance with Multiple Antibiotic Resistance of Bacteria Isolated from Drinking Water. Appl. Environ. Microbiol. 1984, 47 (6), 1238–1242. (21) Armstrong, J. L.; Calomiris, J. J.; Seidler, R. J. Selection of antibiotic-resistant standard plate count bacteria during water treatment. Appl. Environ. Microbiol. 1982, 44 (2), 308– 316. (22) Kurenbach, B.; Marjoshi, D.; Amábile-cuevas, C. F.; Ferguson, G. C.; Godsoe, W.; Gibson, P.; Heinemann, J. a. Sublethal Exposure to Commercial Formulations of the Herbicides Changes in Antibiotic Susceptibility in Escherichia coli and serovar Typhimurium. MBio 2015, 6 (2), e0009-15. (23) Zhang, P. Y.; Xu, P. P.; Xia, Z. J.; Wang, J.; Xiong, J.; Li, Y. Z. Combined treatment with the antibiotics kanamycin and streptomycin promotes the conjugation of Escherichia coli. FEMS Microbiol. Lett. 2013, 348 (2), 149–156. (24) Xia, Z. J.; Wang, J.; Hu, W.; Liu, H.; Gao, X. Z.; Wu, Z. H.; Zhang, P. Y.; Li, Y. Z. Improving conjugation efficacy of Sorangium cellulosum by the addition of dual selection antibiotics. J. Ind. Microbiol. Biotechnol. 2008, 35 (10), 1157–1163. (25) Song, B.; Wang, G. R.; Shoemaker, N. B.; Salyers, A. A. An Unexpected Effect of Tetracycline Concentration: Growth Phase-Associated Excision of the Bacteroides Mobilizable Transposon NBU1. J. Bacteriol. 2009, 191 (3), 1078–1082. (26) Prudhomme, M.; Attaiech, L.; Sanchez, G.; Martin, B.; Claverys, J. P. Antibiotic stress induces genetic transformability in the human pathogen Streptococcus pneumoniae. Science. 2006, 313 (5783), 89–92. (27) Úbeda, C.; Maiques, E.; Knecht, E.; Lasa, Í.; Novick, R. P.; Penadés, J. R. Antibiotic- induced SOS response promotes horizontal dissemination of pathogenicity island-encoded

8

virulence factors in staphylococci. Mol. Microbiol. 2005, 56, 836–844. (28) Beaber, J. W.; Hochhut, B.; Waldor, M. K. SOS response promotes horizontal dissemination of antibiotic resistance genes. Nature 2004, 427, 72–74. (29) Falkinham, J. Common Features of Opportunistic Premise Plumbing Pathogens. Int. J. Environ. Res. Public Health 2015, 12 (5), 4533–4545. (30) Yoder, J., Roberts, V., Craun, G.F., Hill, V., Hicks, L., Alexander, N.T., V.R., Calderon, R.L., Beach, M.J., Carolina, N., 2008. Surveillance for Waterborne Disease and Outbreaks Associated with Drinking Water and Water not Intended for Drinking --- United States, 2005--2006. Morb. Mortal. Wkly. Rep. 57, 39–62. (31) Brunkard, J. M.; Ailes, E.; Roberts, V. A.; Hill, V.; Hilborn, E. D.; Craun, G. F.; Rajasingham, A.; Kahler, A.; Garrison, L.; Hicks, L.; Carpenter, J.; Wade, T. J.; Beach, M. J.; Yoder, M. Surveillance for waterborne disease outbreaks associated with drinking water– United States, 2007–2008. Morb. Mortal. Wkly. Rep. 2008, 57 (SS-9), 1–38. (32) Ashbolt, N. Environmental (Saprozoic) Pathogens of Engineered Water Systems: Understanding Their Ecology for Risk Assessment and Management. Pathogens 2015, 4 (2), 390–405. (33) Williams, K.; Pruden, A.; Falkinham, J.; Edwards, M. Relationship between Organic Carbon and Opportunistic Pathogens in Simulated Glass Water Heaters. Pathogens 2015, 4 (2), 355–372. (34) Wang, H.; Edwards, M. A.; Falkinham 3rd, J. O.; Pruden, A.; Falkinham, J. O.; Pruden, A. Probiotic Approach to Pathogen Control in Premise Plumbing Systems? A Review. Env. Sci. Technol. 2013, 47 (18), 10117–10128. (35) Rhoads, W. J.; Garner, E.; Ji, P.; Zhu, N.; Schwake, D. O.; Pruden, A.; Edwards, M. A. Distribution System Operational Deficiencies Coincide with Reported Legionnaires’ Disease Clusters in Flint, Michigan. Env. Sci. Technol. 2017. 51 (20), 11986-11995. (36) Habicht, W.; Müller, H. E. Occurrence and parameters of frequency of Legionella in warm water systems of hospitals and hotels in Lower Saxony. Zentralblatt fur Bakteriol. Mikrobiol. und Hyg. 1988, 186 (1), 79−88. (37) Bargellini, A.; Marchesi, I.; Righi, E.; Ferrari, A.; Cencetti, S.; Borella, P.; Rovesti, S. Parameters predictive of Legionella contamination in hot water systems: Association with trace elements and heterotrophic plate counts. Water Res. 2011, 45 (6), 2315–2321. (38) Fraser, D. W.; Tsai, T. R.; Orenstein, W.; Parkin, W. E. Legionnaires’ Disease: Description of an Epidemic of . N. Engl. J. Med. 1989, 297, 1189–1197. (39) Peleg, A. Y.; Seifert, H.; Paterson, D. L. Acinetobacter baumannii: Emergence of a Successful Pathogen. Clin. Microbiol. Rev. 2008, 21 (3), 538–582. (40) Horsburgh, C. R. Mycobacterium avium complex infection in the acquired immunodeficiency syndrome. N. Engl. J. Med. 1991, 324 (19), 1332–1338. (41) Bodey, G. P.; Bolivar, R.; Fainstein, V.; Jadeja, L. Infections Caused by Pseudomonas aeruginosa. Rev. Infect. Dis. 1983, 5 (2), 279–313. (42) Marciano-Cabral, F.; Cabral, G. Acanthamoeba spp. as agents of disease in humans. Clin. Microbiol. Rev. 2003, 16 (2), 273–307. (43) Garner, E.; Zhu, N.; Strom, L.; Edwards, M.; Pruden, A. A human exposome framework for guiding risk management and holistic assessment of recycled water quality. Environ. Sci. Water Res. Technol. 2016, 2, 580-598.

9

CHAPTER 2 : A HUMAN EXPOSOME FRAMEWORK FOR GUIDING RISK MANAGEMENT AND HOLISTIC ASSESSMENT OF RECYCLED WATER QUALITY

Emily Garner, Ni Zhu, Laurel Strom, Marc Edwards, Amy Pruden

ABSTRACT

Challenges associated with water scarcity and increasing water demand are leading many cities around the globe to consider water reuse as a step towards water sustainability. Recycled water may be used in a spectrum of applications, from irrigation or industrial use to direct potable reuse, and thus presents a challenge to regulators as not all applications require the same level of treatment. We propose that traditional drinking water standards identifying “safe” water quality are insufficient for recycled water and that using the “human exposome” as a framework to guide development of a risk management strategy offers a holistic means by which to base decisions impacting water quality. A successful and comprehensive plan for water reuse must consider 1) health impacts associated with both acute and chronic exposures, 2) all routes of exposure by which individuals may encounter recycled water, and 3) water quality at the true point of use after storage and transport through pipe networks, rather than at the point of treatment. Based on these principles we explore key chemical differences between recycled and traditional potable water, implications for distribution systems with respect to design and operation, occurrence of chronic contaminants, and the presence of emerging and often underappreciated microbial contaminants. The unique nature of recycled water has the potential to provide rapid regrowth conditions for certain microbial contaminants in these systems, which must be considered to achieve safe water quality at the point of use.

INTRODUCTION

Water reuse is essential for satisfying domestic and industrial water demand worldwide and achieving water sustainability.1 Domestic wastewater can be treated to the necessary level of quality and reused to reduce loss of treated effluent via discharge, relieve pressures on depleting groundwater aquifers, and minimize extraction of water from fragile environments such as drought-stricken surface waters. Treatment of wastewater for reuse is also cost effective compared to alternative approaches, such as obtaining freshwater from desalination.2 Particularly when wastewater is treated for direct or indirect potable reuse, a “multi-barrier” treatment framework is typically used to ensure that multiple means of removing pathogens or harmful chemicals will protect public health in the case of a process failure or other unexpected event that could compromise water quality.3 However, while this approach is logical for controlling acute health threats associated with water as it leaves the treatment facility, it does not address concerns with respect to low-level chronic exposures or changes in water quality during distribution to the point of use.4

A major concern for water reuse in general is the lack of federal regulations.5 Further, the few nascent recycled water quality regulations and guidelines available, typically at the state and local level, have been narrowly focused on fecal indicator bacteria (i.e., total and fecal coliforms).6– 8 This approach addresses traditional concerns regarding fecal-associated pathogens, but does not necessarily provide insight into safeguarding microbial water quality during distribution.9,10 In

10

particular, microbial contaminants that are of concern due to their ability to grow within distribution systems, such as opportunistic pathogens (OP), antibiotic resistant bacteria and their associated antibiotic resistance genes (ARG), and free-living amoebae (FLA), have little or no relationship with fecal indicator bacteria based standards. Alternative frameworks for assessing and managing recycled water quality more holistically are emerging. For example, adaptations of the Hazard Analysis and Critical Control Point (HACCP)11 paradigm, which originated in the food safety industry, and the World Health Organization’s (WHO) Water Safety Plan (WSP)2,12 have been proposed for use as risk management frameworks for recycled water. Application of these adaptable frameworks could have the advantage of drawing attention to the entire treatment process, rather than focusing solely on absence of indicator organisms as a proxy for safe water. Still, we identify three key elements that should be taken into account for a truly comprehensive consideration of public health concerns: 1) evaluation of health impacts associated with both acute and chronic exposures; 2) accounting for all routes of exposure by which individuals may encounter contaminants in recycled water; and 3) consideration of water quality at the true point of use after storage and transport through pipe networks, rather than at the point of treatment.

A more holistic approach to characterizing the physical, chemical, and microbial characteristics of recycled water, as well as the routes by which humans are exposed, can be derived from the emerging concept of the “human exposome.” The exposome has been defined as “the cumulative measure of environmental influences and associated biological responses throughout the lifespan, including exposures from the environment, diet, behavior, and endogenous processes.” 13 The exposome includes general (e.g., climate, urban environment), specific (e.g., water, food, air), and internal (e.g., metabolism, gut/lung microbes) factors and their role in disease.14 Clearly, water, and its corresponding chemical and microbial properties is a fundamental component of the exposome. Water is fundamental to human health, survival, and hygiene and is an integral part of daily life, including direct contact (i.e., drinking, cooking, and showering) and indirect exposures via bioaerosols (i.e., cooling water, flush toilets, or lawn irrigation). We propose that adopting the exposome paradigm as a model for recycled water quality assessment can be used to guide development of an HACCP, WSP, or other comprehensive risk management strategy that more accurately reflects the true risks and exposures associated with water reuse and can strengthen implementation of existing risk management strategies.

In terms of safeguarding recycled water from the point of treatment to the point of use, we note that there are important distinctions between potable and recycled water that should not be ignored, particularly with respect to design, operation, and maintenance of recycled water distribution systems (RWDS). In this critical review, we note key chemical differences between recycled and potable waters, with a particular emphasis on organic matter, and explore implications for RWDSs with respect to design, operation, and intended application. We also discuss how these key differences impact the presence of chronic contaminants and emerging concerns about the presence of OPs, ARGs, and other microbial contaminants (Figure 2-1). Our goal is to proactively address plausible public health risks associated with practical realities of recycled water use.

11

Figure 2-1: Key aspects of the exposome paradigm for managing RWDS. Aspects emphasize holistic consideration of potential exposures to recycled water, including A) chemical distinctions of recycled vs traditional potable water, such as enriched organic matter/nutrients, disinfectant decay, critical reactive zones and chronic contaminants; B) emerging concerns about ARGs, OPs and other microbial contaminants; C) nontraditional routes of exposure, including inhalation, dermal contact.

UNIQUE ASPECTS OF RWDS DESIGN, OPERATION, AND WATER USE

There is a broad continuum of applications for water reuse, ranging from unintended de facto reuse to direct potable reuse (DPR) produced by advanced treatment processes. De facto reuse refers to a situation where reuse of treated wastewater occurs but is not planned, for example, when a drinking water treatment plant intake is located downstream from a wastewater discharge.6 In the U.S., de facto reuse is widespread and becoming increasingly common in recent decades. Rice et al. found that of the top 25 drinking water treatment plants most impacted by upstream wastewater treatment plant (WWTP) discharges in the United States, the fraction of their water source comprised of WWTP discharges increased from between 2 to 16% in 1980 to an average of 68% under typical streamflow conditions in 2008.15 Some treatment plants received as much as 100% WWTP effluent under low flow conditions. Indirect potable reuse refers to the use of treated wastewater to augment other potable source waters following retention in an environmental buffer.6 Common environmental buffers include groundwater aquifer recharge and subsequent withdrawal prior to drinking water treatment or intentional discharge of wastewater effluent upstream or into a reservoir from which water is withdrawn for drinking water treatment. DPR consists of treating wastewater for direct use as a source water for drinking water treatment. DPR is currently limited in full scale application, with Windhoek, Namibia16 and Big Spring, Texas,

12

USA17 serving as prime examples, but there is growing interest in expanding DPR infrastructure in the U.S.

Wastewater may also be treated for non-potable reuse when it can offset water demand associated with landscape or recreation area irrigation, agricultural and food crop irrigation, snowmaking, industrial use such as in cooling towers or in natural gas production, or to augment environmental waters such as in groundwater aquifer recharge, river or stream flow augmentation, or in wetlands.6 Though treatment requirements may reasonably be lower in these cases, all of these non-potable reuse scenarios have relevant human exposures that should be taken into consideration.

Table 2-1 illustrates that there is a wide range of observed recycled water quality characteristics as a function of increasing levels of treatment. Unlike drinking water, where consistent regulations are applied, recycled water is used for a wide spectrum of applications with required treatment varying based on intended use. For example, the most stringent requirements are applied to DPR, whereas residual nutrients can be viewed as a beneficial fertilizer in non- potable reuse scenarios. Hence, efficient reuse treatments ideally match the intended purpose of the recycled water. A number of U.S. states pioneering water reuse have recognized this concept of “fit-for-purpose,” tailoring recycled water treatment regulatory guidelines based on end uses. Determining the ideal configuration of treatment processes for different reuse scenarios would greatly benefit from research integrating water treatment outcomes with the exposome paradigm for more comprehensively considering chemical and microbial risks.

Table 2-1: Water quality as a function of treatment process Treatment processes CAS CAS CAS with MBR MBR CAS CAS Filtration Water quality parameter (units) with with filtration with with effluen with MBR with filtratio BNR and and chlorin UV/ t BNR chlorine n filtration chlorine e ozone Turbidity (NTU) 2-15 0.5-4 2-8 0.3-2 ≤1 1.5-8.7 3.6-6.3 1.7-4.3 0.2-0.5 Total suspended solids (mg/L) 5-25 2-8 5-20 1-4 <2 TOC (mg-C/L) 10-40 8-30 8-20 1-5 0.5-5 12-16 6-8 3-5 2-3 BDOC 5-7 1-2 1-2 <1 (mg-C/L) AOC (mg-C/L) 0.2-1.4 1-2 1-2 1-2 <1 Total nitrogen (mg-N/L) 15-35 15-35 3-8 2-5 <10 5-10 <1 1-3 5-6 Total phosphorus (mg-P/L) 4-10 4-8 1-2 ≤2 <0.3-5 3-5 1-2 5-9 <1 Volatile organic compounds (µg/L) 10-40 10-40 10-20 10-20 10-20 Total coliforms (CFU/100 mL) 104-105 103-105 104-105 104-105 <100 <1 1-10 1-10 1-10 Protozoa and cysts 10-102 0-10 0-10 0-1 0-1 43110 10-100 <1 ≤1 (CFU/100 mL) Viruses (CFU/100 mL) 10-103 10-103 101-103 10-103 1-103 present present present negative 29,192 Source ; CAS = conventional activated sludge; BNR = biological nutrient removal; MBR = membrane bioreactor

13

Routes of Exposure

Stemming from the continuum of recycled water uses described above, there is also a range of relevant exposure scenarios (Table 2-2). Use of a traditional water quality paradigm based on monitoring of fecal indicator bacteria as a sole benchmark for establishing safe water quality neglects risk associated with non-fecal pathogens and routes of exposure other than ingestion. Though ingestion and aspiration are potential modes of exposure associated with DPR, inhalation and dermal contact are important, yet often overlooked, exposure routes for potable water and even more so for recycled water. Recycled water is commonly used for purposes that generate inhalable aerosols, including use in cooling towers,18 spray irrigation,19,20 fire-fighting,21 toilet flushing,22 as well as for aesthetic purposes, such as in decorative fountains.23–25 Importantly, relevant exposure zones may be vast, with Legionnaire’s disease infection associated with inhalation of aerosols from cooling towers located more than a mile away.18 There is also a strong likelihood of dermal contact when recycled water is used for irrigation in public recreation areas. This may be particularly relevant in irrigated parks, athletic fields, and snowmaking for recreational purposes. Dermal abrasions or other lacerations that may be pre-existing or occur during use of these facilities create an additional route of exposure for infection. When recycled water is used for food crop irrigation, chemical and microbial constituents may also be transmitted to humans on or within crops via ingestion.26 When recycled water is used for DPR, a number of often overlooked routes of exposure should also be considered, for example, including use in humidifiers, ice machines, and decorative water features. A more holistic characterization of human exposures to recycled water constituents via such non-conventional routes is important for accurate assessment of health risk associated with use of recycled water.

Physical and Operational Issues

Water and wastewater infrastructure degradation has become one of the leading threats to public health and water security.27 Aging and poorly managed pipes can lead to a drastic decrease in water quality of the transported water, arising from complex interactions among chemical, physical, and microbial constituents. New design considerations are needed to ensure sustainable management of recycled water infrastructure, considerate of their distinct physical and operational characteristics relative to potable water systems.

In a recent survey of 71 recycled water systems in the US and , Jjemba et al. identified infrastructure issues as the most prevalent problem associated with managing and maintaining water quality in RWDSs. Over 20% of recycled water facilities listed infrastructure integrity as a water quality concern. The extensive list of infrastructure challenges revealed by the survey includes infrastructure deterioration from high chlorine residual, maintenance of desired pressure and flow during low and inconsistent usage, lack of redundant design and storage, complicated branched distribution systems designed to supply multiple recycled applications from a centralized treatment plant, high corrosivity of water damaging metal pipes, and effective monitoring of the chemical and microbial quality.4 For example, to control microbial activity resulting from nutrient-rich recycled water, up to 40 mg/L chlorine has been used in some systems, which can potentially result in widespread damage to water infrastructure.28 Use of reservoirs as a way to satisfy on-demand recycled water applications has also been observed to be a challenge, with impaired water quality resulting from long stagnation time and proliferation of algae and

14

Table 2-2: Overview of non-traditional routes of exposure for recycled water and putative risk of infection or exposure Recycled Water Documented Documented Concentrations in Application (in Concentrations in Route of Recycled Water addition to Putative Hazard Infectious Dose Drinking Water Exposure other potable [Percent Samples positive uses) (positive concentration range)] Opportunistic pathogen infection of wounds 107 193 194 Snowmaking, Staphylococcus aureus 17% 6.25% Irrigation of 108-109 colony forming 88% (1±1 - 9±10 CFU/100 5.6% (up to 700.3±158.7 Pseudomonas aeruginosa 108 Dermal athletic and units (CFU)a; 195 mL)b; 29 gene copies / mL)111 Contact recreation Acanthamoeba spp.169 104 trophozoitesc; 196 Not detected111 facilities, Toilet 181 197 Cyanobacteria toxicity (up to ~140 µg/mL chlorophyll) 22 flushing 8% for MRSA of 17 % for Antibiotic resistant infections198 susceptible S. aureus d; 193 Opportunistic pathogen infection of lungs 81% (0.4x103±0.2x103 - 5.6% (up to 219.4±23.8 Legionella pneumophila e; 20 103 - 106 CFU 200 3.5x103± 16x103CFU/100 gene copies / mL)111 98.1% (2.1±104 - 4.2±103 gene Mycobacterium spp.201 104 - 107 CFUag; 195 Cooling towers, copies / mL)111 Spray irrigation, 5.6% (up to 700.3±158.7 Pseudomonas aeruginosa 202 108 - 109 CFUa; 195 88% (1±1 - 9±10 CFU/100 mL)b; 29 Inhalation Toilet flushing22, gene copies / mL)111 and Fire suppress- 19 193 194 Aspiration Staphylococcus aureus 17% 6.25% 21 ion , Car Naegleria fowleri 168,169 103 -105 trophozoites168,203 8-27%168 199 washing Cyanobacteria toxicity181 (up to ~140 µg/mL chlorophyll)197 8% for MRSA of 17 % for Antibiotic resistant infections19 susceptible S. aureus d; 193 Fecal pathogens Disinfection byproducts79 (9.70 - 399.37 µg TTHMh /L79) Recreation, Opportunistic pathogen infection Nasal Potable reuse 168 168 Aspiration Naegleria fowleri 8-27% (sinus irrigation) Recreation, Opportunistic pathogen infection Eye and Direct potable Acanthamoeba keratitis 169 104 trophozoi-tes196 Ear reuse, Indirect 5.6% (up to 700.3±158.7 Contact Pseudomonas aeruginosa 108 108 - 109 CFUa; 195 88% (1±1 - 9±10 CFU/100 mL)b; 29 potable reuse gene copies/ mL)111 8% for MRSA of 17 % for 119–121 Coloniz- Antibiotic resistant infections d; 193 susceptible S. aureus ation and Various Opportunistic pathogen infection Delayed 108 8 9 a; 195 b; 29 Infection Pseudomonas aeruginosa 10 - 10 CFU 88% (1±1 - 9±10 CFU/100 mL) Staphylococcus aureus 107 17%193 6.25%194 a Oral route of infection; bBased on Pseudomonas spp.; cBased on Acanthamoeba keratitis ; dMRSA = Methicillin resistant Staphylococcus aureus ; ePutative hazards consider both Legionella pneumophila and other pathogenic Legionella ; fBased on detection of Legionella spp.; gBased on Mycobacterium avium ; hTTHM = Total trihalomethanes aquatic vegetation.29 Experience from potable water systems has also shown that interaction of iron pipes with water containing high organic content and oxygen tends to promote iron release, producing unacceptable discolored water following stagnation.30 It is important to bear in mind that a shift in water chemistry can have disastrous unintended consequences for corrosion,31–33 and given the unique chemistry of recycled water (Table 2-3), it will be especially critical to bear this in mind. For example, previous studies have documented cases where switching potable water pipes to DPR pipes resulted in destabilization of the of existing corrosion scales and biofilms and an undesirable degradation of water quality at the point of use.34 Despite the unresolved challenges associated with transporting recycled water, this alternative type of water also presents a creative opportunity for solving the challenging issue of leaking pipes. Tang et al. have successfully demonstrated the autogenous repair phenomenon in copper and iron pipes in drinking water distribution systems (DWDS) via beneficial corrosion deposition.35 Optimistically, with a more

15

diverse water chemistry profile, recycled water may be an even better candidate for protecting aging pipes.

Temperature, as an overarching parameter, is another critical factor that could have profound implications in designing and monitoring water reuse systems.36 Not only is temperature directly related to microbial activity, disinfectant residual decay, corrosion rate, and dissolved oxygen levels, it is also indirectly linked to consumption patterns, flow patterns and velocity, and bulk water and biofilm interactions. Elevated recycled water temperatures may stem from extended stagnation times, particularly during the day in cases where irrigation is conducted at night to limit evaporation, as well as from use of above-ground pipelines, which facilitate transport of recycled water over long distances.

For on-demand non-potable water reuse applications, such as agricultural irrigation, landscaping, and toilet flushing, many studies have observed distinct consumption variations in daily and seasonal demand patterns.37,38 For example, on a daily scale, the generation of wastewater effluent usually peaks in the daytime when people are active, but the demand of irrigation water usually occurs at night with an offset time of approximately 12 hours. Discrepancies in user patterns makes water stagnation and storage, along with associated water quality deterioration, a prominent concern in design and maintenance of recycled systems. Multiple studies have also documented water quality deterioration during winter or high rainfall periods in systems largely used for irrigation, due to low user frequency.39

True water age may differ substantially from the designed hydraulic residence time of the recycled water systems based on the actual end-use applications. Emerging work in premise (i.e., building) plumbing systems, i.e. the water pipe networks within homes and buildings, has highlighted unique systematic features in terms of longer stagnation time, elevated temperature, and loss of disinfectant residual, which serve to stimulate microbial proliferation precisely at the point of use, thus amplifying any potential exposure risk to end users.40,41 Similar investigations are needed to quantify the risk of exposure associated with user-driven demand patterns in non- potable reuse systems.

IMPORTANT CHEMICAL DIFFERENCES ANTICIPATED BETWEEN RECYCLED AND POTABLE WATER DISTRIBUTION SYSTEMS

Organic Matter

One of the most distinctive characteristics of recycled water is the nature of dissolved organic matter (DOM) and its occurrence at elevated levels. The organic matter in typical potable waters consists of natural organic matter (NOM) derived mostly from oil, planktonic and vegetative matter, and decay by-products in natural water sources. However, it is important to note that DOM present in recycled waters may be quite distinct from that of potable water due to different sources and treatment processes. In a recent review comparing organic matter data published in the last 15 years for drinking water and recycled water systems, Hu et al. identified four distinct classes of NOM in recycled water: recalcitrant DOM, soluble microbial products from

16

Table 2-3: Comparing water quality of typical drinking water vs different recycled water applications Recycled water applications Direct Drinking Private, Indirect Industrial Parameter Environ- Implications for Distribution Water urban and potable applica- mental irrigation recharge tions reuse pH 6.5-8.5 6-9 6-9 7-9 7-8.5 Total -Provides the most limiting nutrient dissolved 500 source for bacterial regrowth in solids (mg/L) distribution systems; Carbon Chemical source -Consumption of carbons in the oxygen 100 70-100 70-100 70 distribution system is observed to relate demand with increased bacterial activity 29,204 (mg/L) Biochemical oxygen 43393 43393 10 demand (mg/L) Near DO (mg/L) >0.5 >3 >8 >3 saturation Total suspended solids N/A 10 10 10 (mg/L) -Control bacterial growth in the distribution system; Chlorine residual -Excess chlorine can cause carbon <4 0.2-1.0 0.05 0.05 (mg/L) fragmentation and DBPs formation -Chlorine may exacerbate antibiotic resistance 147,148 Total Kjeldahl N -Concerns for nitrification and <10 15-20 10-20 10 (mg/L) denitrification Ammonia-N (mg/L) < 0.2 a 2-20 1.5 0.2 1.5 -Concern for nitrification Total phosphorus -Eutrophication and degradation of water 2-5 0.2 0.2 (mg/L) quality -Caused “red water” -Promote growth of corrosion bacteria Iron (mg/L) 0.3 2 2 and damage pipe integrity -May select for antibiotic resistant bacteria205 -promote growth of certain corrosion bacteria -toxic to certain bacterial and aquatic Copper (mg/L) 1 0.2-1.0 0.2-1.0 at elevated levels -May select for antibiotic resistant bacteria143,206 -May select for antibiotic resistant Zinc (mg/L) 5 0.5-2 0.5-2 bacteria143,206 -May select for antibiotic resistant Pesticide (mg/L) 0.05 0.05 bacteria207 Fecal coliforms Indicator bacteria for pathogenicity of Zero Zero < 200b Zeroc <200a (CFU/100 mL) water (Source: 6,208,209) a WHO guidelines for drinking water quality b Based on 7-day median with none > 800 per 100 mL c Based on 7-day median with non > 14 per 100 mL

17

biological wastewater treatment units, transformation products from advanced treatment, and emerging contaminants associated with anthropogenic activities.42 It was concluded that DOM composition differed significantly between recycled water and drinking water evaluated against five critical chemical indicators: dissolved organic carbon (DOC), dissolved organic nitrogen, assimilable organic carbon (AOC), estrogenic activity, and disinfection byproduct (DBP) formation potential (Figure 2-2). DOC in drinking water ranged from 1.5-11.2 mg/L with a median of 3.9 mg/L while that in recycled water ranged from 3.6-14.6 mg/L with a median of 7.5 mg/L, indicating recycled water as a much more nutrient rich environment for microbial regrowth and byproduct formation. The heightened levels of biotoxicity, in terms of estrogenic levels, is also widely reported in studies examining effluent organic matter compositions, suggesting a potential health risk when used for recycling applications.43

Figure 2-2: Overview of typical normalized composition and potential magnitude of dissolved organic matter (DOM) in drinking water sources compared to recycled water sources. Presented in terms of dissolved organic carbon (DOC), dissolved organic nitrogen (DON), assimilable organic carbon (AOC), estrogenic activity, total haloacetic acid formation potential (THAAFP), and total trihalomethane formation potential (TTHMFP). (Reprinted from Science of The Total Environment, 551-552, Hong-Ying Hu, Ye Du, Qian-Yuan Wu, Xin Zhao, Xin Tang, Zhuo Chen, Differences in dissolved organic matter between reclaimed water source and drinking water source, page 133-142, 2016, with permission from Elsevier.)

Biological stability, i.e., the ability of drinking water to suppress microbial growth in the absence of disinfectants,44 is especially of concern for safe transport and storage of treated water. Ideally, low nutrient water will limit growth in the distribution system, a strategy applied successfully in some European countries for eliminating the need for secondary disinfectant in DWDSs.45 The proportion of DOC that facilitates bacterial regrowth is typically measured by

18

either biodegradable dissolved organic carbon (BDOC) or AOC assays. An array of methods have been used to best evaluate the bacterial growth potential of various types of water samples, with established approaches generally being to measure the decrease in measured DOC over time or an increase in indicator bacteria counts as a proxy for biologically available DOC. To date, there is no widely accepted standardized method to quantify biostability. Reported threshold BDOC and AOC values to achieve biostability in drinking water systems using different methods are documented in Table 2-4. Existing surveys of recycled water systems have indicated orders of magnitude higher levels of organic carbon than in typical U.S. drinking water systems.46,47 In particular, biodegradable organic matter has been observed to be four or five times higher than that of drinking water,29 while AOC can range from 505 to 918 µg/L in moderately treated recycled water,48 compared to 18 to 189 µg/L in drinking water.49

Table 2-4: Proposed threshold values to achieve biostability in drinking water distribution systems Carbon source Threshold values Criteria Reference ≤ 0.15 mg-C/L 210,211,212 ≤ 0.25 mg-C/L Stable BDOC values 213 BDOC ≤ 0.30 mg-C/L at 15°C 210 ≤ 0.15 mg-C/L No coliform growth 214 10 µg-C/L No heterotrophic plate count growth 215 50 µg-C /L No coliform growth 216 AOC 50 µg-C /L No V. cholerae growth 217 100 µg-C /L No E. Coli growth 49

The abundance and type of biodegradable carbon in recycled water calls into question the extent to which the science of potable water delivery is directly translatable to recycled water distribution. In the only available study of its kind, Jjemba et al. examined four RWDSs in the U.S. and observed a trend of AOC and BDOC consumption with increasing residence time, with an average reduction of 475 µg/L AOC and 370 µg/L BDOC from the distribution system point of entry to the point of use.29 They concluded that the change in AOC and BDOC was due to enhanced microbial activity, indicating significant changes in both the quantity and the quality of the available carbon in the RWDSs. In parallel simulated RWDS loop studies, high organic carbon was also observed to result in rapid consumption of disinfectant residuals in the distribution systems.29 Up to 6 mg/L of chlorine was completely consumed within minutes in all systems, leaving the remainder of the distribution system vulnerable to bacterial growth.29

Given the unique nature of organic matter and microbial composition of recycled water, existing assays such as those for AOC or BDOC, may not be suitable. Only one study could be found specifically aimed at adapting the AOC assay to recycled water.50 By including test strains that are more ecologically representative of the sample waters, Zhao et al. concluded that the standard P17 and NOX strains applied in the AOC assay largely underestimate levels in recycled water.50 Khan et al. have similarly highlighted the need to optimize the BDOC method for recycled water with their modified protocol improving repeatability and precision of results as verified by independent biochemical oxygen demand and chemical oxygen demand measurements.51

19

Another negative consequence of NOM in distributed water is that it can accelerate biocorrosion of pipes, which in turn can further stimulate AOC generation.39,52,53 BDOC is also believed to play an important role in microbiologically induced corrosion.54,55 Recycled water, as an abundant source of sulfate and nitrogen species, is likely to provide a nutrient-rich environment for iron-oxidizing/reducing bacteria28,56,57 and sulfate-reducing bacteria34 to thrive in the DS, further raising concerns about the potential for recycled water to accelerate damage to pipe networks.

Redox zones and degradation of water quality

The distribution system can be thought of as a complex reactor with interrelated chemical and biological reactions occurring spatially and temporally as the water passes through the pipes.58 The chemistry of treated potable water changes significantly during transport, with deteriorating DWDS water quality documented since the early 1920s.59–64 Masters et al. illustrated the water distribution system reaction phenomenon by demonstrating the formation of sequential redox zones as a function of water age in simulated DWDS.65 Given greater physiochemical and microbial complexity in RWDS, we speculate that they would foster development of even more dramatic reactive zones, as a function of key physical and hydraulic design parameters such as residence time, flow pattern, hydraulic surface to volume ratio, and pipe layout. Consistent with this expectation, studies in lab and field-scale RWDS recently demonstrated elevated microbial activity as indicated by rapid AOC/BDOC consumption, even at the earliest water age, and attenuated organic carbon at higher water ages.29 Similarly in a 15 month monitoring study of RWDSs, the general pattern observed was an initial reactive zone where rapid microbial regrowth and chemical reactions occurred followed by relatively constant microbial and chemical reactivity along the length of the pipes.66 Recognizing the reality of reactive zones in distribution systems and more deliberately monitoring them may provide valuable insight into predicting and preempting potential problems resulting from issues related to water chemistry.

Disinfectant residual

The intricate relationship between chlorine-based disinfectants and microbial and chemical stability has been intensely studied in DWDSs. Due to its strong oxidizing power, chlorination is generally the disinfectant of choice for microbial control in drinking water treatment. Chlorination can greatly reduce general bacterial counts and help satisfy drinking water microbial regulations. However, as a strong oxidizing agent, it is also known to interact with reductive species, metals, organic matter and pipe materials and, as a result, significantly impact the downstream water chemistry.67 The most widely noted issue with disinfectants is the fragmentation of complex carbon compounds, thus increasing the fraction of biologically available carbon when high concentrations of chlorine are used.29,46,47 Given the tendency to use fecal indicators as a benchmark for assessing recycled water quality, it can be tempting for utilities to dose high concentrations of chlorine. Due to a higher chlorine demand typical of recycled water, disinfectant residual may be rapidly lost, leaving the rest of the RWDS vulnerable to microbial instability.66 Also important to note is that there is growing concern regarding the efficacy of chlorine-based disinfectants against emerging resistant pathogens, which might be more abundant in recycled water than traditional potable water.68–71 The potential for indiscriminate use of disinfectants to inadvertently select disinfectant-resistant bacteria in the RWDS is worthy of exploring in future

20

research. Further, the ability of bacteria to repair and recover in the distribution system following the shock of ultraviolet irradiation (UV), chlorine, or other disinfectant should be considered, as exemplified by recovery of viable but non-culturable bacteria.72–74

Another issue worthy of consideration is the potential for enhanced DBP formation in recycled waters.75,76 In a study comparing DBP formation between wastewater effluent and surface water, Sirivedhin and Gray found that effluent-derived organic matter stimulated formation of higher proportions of brominated DBPs.77 Nurizzo et al. evaluated the DBP formation potential with various disinfection agents and concluded that hypochlorite yielded the greatest total trihalomethanes, exceeding the Italian regulation for agricultural reuse, even when starting with high quality recycled water.78 While DBPs tend to be ignored in recycled waters, particularly for non-potable applications, it is important to recognize that inhalation is also a relevant exposure route to consider, with one model characterizing the inhalation exposure to trihalomethanes of irrigation workers using recycled water suggesting that there was a 13% risk of exceeding acceptable exposure levels for cancer risk.79 The DBP issue illustrates that there can be tradeoffs between microbial control and chemical risks and that clearer guidance and alternative approaches are needed for recycled water to avoid negative consequences of blindly over-chlorinating.

CHRONIC CONTAMINANTS

The exposome highlights the importance of considering exposures over the course of one’s lifetime, and thus, chronic contaminants are an important hazard worthy of consideration during risk assessment of recycled water. WWTPs are generally not designed with the intention of removing micro-constituents, such as pharmaceuticals and personal care products, recalcitrant organic compounds, heavy metals, nanomaterials, and industrial agricultural additives.80–84 Jelic et al. was able to detect 29 pharmaceutical products in the final effluent of one WWTP, versus 32 in the influent.83 Even when discharged at micro-concentrations, up to hundreds of nanograms per liter of targeted micropollutants can still be consistently detected in receiving water bodies and levels can accumulate.80 In a study that monitored 15 different WWTPs generating recycled water for groundwater recharge, detectable levels of all 20 most commonly used antibiotics were still found at elevated concentrations of 212-4,035 ng/L in recycled water and 19-1,270 ng/L in groundwater.85 Several studies have also observed seasonal patterns of higher discharge of pharmaceuticals and personal care products in wastewater effluent during low flow and less during high flow periods.86,87

While increasing research attention and regulatory efforts have been devoted to understanding prevalence of non-conventional chemical constituents in WWTPs and in receiving environments,88 studies specifically focusing on characterization and risk assessment of emerging chemical constituents of concern in the context of recycled water applications are limited.89–91 Advanced oxidation processes are particularly promising for removal of these pharmaceuticals and other organic compounds (Table 2-5).92,93 Negative ecological effects of chemical constituents on the aquatic environment have received much attention.94,95 Although a multi-barrier approach consisting of sequential treatment processes has promise, questions remain regarding the ideal treatment for various contaminant types and reasonable end point concentrations that are protective of human and ecological health. Given the diverse applications of recycled water, relevant, accurate and comprehensive risk models are needed considerate of the various environmental

21

spheres of influence. Wastewater effluent discharged to surface water has resulted in detection of emerging pharmaceutical products in 80% of surface water samples.96 Thus, the science and practice of distributing recycled water should proceed with a comprehensive approach to understanding of the fate and impacts of these emerging contaminants in relevant environments.

Table 2-5: Case studies of existing application of advanced treatment processes for intended reuse purposes. Treatment trains rely on use of biological activated carbon (BAC), reverse osmosis (RO), ultrafiltration, and UV. Advanced Intended treatment Key Results Reference reuse processes · H2O2/ozone process demonstrated higher than 90% average removal rate in 21 of 31 targeted trace organic contaminants and hormonal products · BAC unit achieved higher than 95% removal for all targeted contaminant except Piloted Ozone/H2O2 + benzophenone indirect 218 BAC · High degree microbial inactivation potable reuse · Raised concerns on elevated AOCs and microbial regrowth potential after H2O2/ozone treatment and · Fluorescence excitation-emission matrix showed distinctively transformed organic matter footprints after treatment · Diminishing DOC removal rate after DOC and breakthrough is reached Standalone BAC nitrogen 219 · More than 50% of total nitrogen removal removal rate · Ozone and ozone/peroxide showed similar trace organic contaminant removal performance, likely due to inherently high Ozone/peroxide + General reuse hydroxyl radicals in wastewater effluent. 220 RO applications · Formation of up to 48ng/L NDMA is observed in wastewater effluent ozone systems, raising concern for future reuse applications · 13 out of 291 targeted compounds are Ultrafiltration+ Groundwa-ter detected in post-UV and post-RO water 89 RO+UV recharge · Calculated risk quotient for detected chemicals indicates safe reuse

22

ARGS, OPS, AND OTHER EMERGING MICROBIAL CONCERNS

It is important to recognize the complexity of RWDSs as an ecological habitat and that microbial concerns reach beyond traditional indicator organism paradigms. Here we consider these emerging microbial concerns within a comprehensive microbiome/exposome framework. Several recent studies have utilized DNA sequencing to provide new insight into the composition of the drinking water microbiome, but few have attempted to characterize the recycled water microbiome. Recycled water, and even potable water, both represent surprisingly complex microbial niches, housing a vast array of microbial species about which little is known. Normal fecal indicator bacterial monitoring fails to provide information about the broader microbiome, particularly with respect to oligotrophic organisms residing in distribution systems. Thus, a more holistic approach for characterizing water quality is needed to accurately describe the water quality at the point of use. Here we elaborate on microbial aspects of the exposome that are generally unrecognized in the regulatory landscape and are particularly relevant to RWDSs. While occurrence of fecal-associated pathogens is also of importance in recycled water systems, we have limited the scope of this review to emerging microbial concerns.

Epidemiological studies examining associations between recycled water exposure and disease have been limited and are crucial to identifying potential for disease transmission, determining suitability for public use, and informing effective risk mitigation strategies. For example, Durand and Schwebach did not find an association of gastrointestinal illness when irrigating public parks with non-potable recycled water versus potable water (6% versus 7% of park users reporting symptoms associated with recycled wastewater irrigation versus potable water irrigation, respectively), though wet grass conditions during park usage were associated with an increased rate of illness.97 A study of food crop irrigation with recycled water over a five year period found no undesirable consequences to the quality of vegetables or soil, thus exposure restrictions for farm workers were not deemed necessary.98 In one study conducted in the U.S., even irrigation using trickling filter effluent wastewater was not associated with an increased rate of infection of rotavirus for residents of surrounding areas.99 A study that examined occurrence of methicillin resistant Staphylococcus aureus (MRSA) in spray irrigation workers using recycled water did not find the presence of the resistant organism in nasal swabs from any workers tested, though the odds of carrying a non-resistant strain of the organism were slightly higher among spray irrigation workers than among office workers.19 While isolated reports of disease stemming from exposure to recycled water are helpful, rigorous, long-term epidemiological studies are needed to more precisely determine sources of disease and accurately characterize risk and to address emerging concerns.

Of rising interest is the influence of the distinct physiochemical nature of recycled water on the regrowth or attenuation of emerging pathogens and contaminants, particularly considering exposures relevant to non-conventional water reuse applications. Especially when organic carbon is no longer a growth-limiting resource, conventional fecal bacterial indicators are likely to be even less relevant to shifts in microbial ecology during distribution and the associated health risk. Efforts are underway to recognize the importance of microbial ecological interactions in distribution systems and the potential to harness them to foster a distribution system that favors the growth of non-pathogenic bacteria over pathogenic ones.63 For example, Egli has identified the survival and growth strategies of various microbes in low-nutrient and stressed environments and

23

competition between pathogens and the indigenous microbiota.100 A strategy of capitalizing upon specific ecological interactions, such as nutritional competition, antagonist growth, and symbiotic relationships for improved water quality and human health has been previously proposed for drinking water.101 This presents a potentially transformative and highly relevant approach for guiding RWDS management.

Opportunistic Pathogens

RWDSs offer several unique characteristics that make them particularly well-suited for supporting regrowth of OPs. OPs in DWDSs are thought to be the primary source of waterborne disease in developed countries, including the U.S.102,103 Unlike most fecal pathogens, OPs do not typically impact the gastrointestinal system but rather they infect via alternative routes. To name a few, Legionella pneumophila, Acinetobacter baumanni, Mycobacterium avium, Burkholderia pseudomallei, Stenotrophomonas maltophilia, and Aspergillus fumigatus can infect hosts’ lungs via inhalation;104–106 S. aureus infects via broken skin or mucus membranes;107 Pseudomonas aeruginosa can infect hosts via the bloodstream, eyes, ears, skin, or lungs;108 and Acanthamoeba spp. can cause infection of the eyes or central nervous system when inhaled or upon penetration of skin lesions.109 These alternative routes of infection make OPs of particular interest for recycled water, where exposure routes other than simple ingestion are more relevant. Inhalation of aerosols from cooling towers or spray irrigation and dermal contact with irrigated surfaces, are important routes of exposure that should be accounted for when considering risks associated with OPs in recycled water.

OPs possess several distinct properties that make them particularly well suited for growth in RWDSs (Figure 2-3). OPs tend to be resistant to disinfection, ranging from 21-658 times as resistant to chlorine as Escherichia coli, as in the cases of P. aeruginosa and A. baumanii, respectively.110 Many OPs are also resistant to phagocytosis by amoebae, becoming enclosed within an amoebic cyst, where they can be further protected from disinfectants and other harsh environmental conditions. Biofilms, where OPs tend to reside, offer protection from similar environmental assaults, in addition to acidic and alkaline conditions and shear force from high flow velocities.101,110 OPs also tend to grow at low organic carbon concentrations, which is pertinent to both DWDSs and RWDSs.110 Stagnation is a notorious risk factor for OP outbreak, and is common in RWDSs due to intermittent demand and seasonal shutdown.111

Antibiotic Resistance Genes

Antibiotic resistance among human pathogens is a major public health concern. In the U.S., the Centers for Disease Control has estimated that antibiotic resistant bacteria cause at least two million infections and 23,000 deaths each year.112 ARGs are now well-known to be elevated in human-contaminated surface waters,113–116 however with respect to human pathogens, specifically, there is reasonable evidence that they can gain ARGs from environmental bacteria via horizontal gene transfer (Figure 2-3).117,118 Therefore, all members of the microbiome carrying ARGs are potentially of concern, particularly those that are common in human pathogens. In addition to the possibility of infection by antibiotic resistant bacteria upon exposure, human hosts may also become colonized and infected later.119–121 Similarly, it is possible that horizontal gene transfer may occur from colonized non-pathogenic bacteria to pathogenic ones, leading to antibiotic

24

resistant infection. Thus, infection by antibiotic resistant bacteria may occur at a time and place separate from that of the initial exposure, which complicates traditional dose-response risk assumptions. ARGs or bacteria expressing ARGs corresponding to resistance to aminoglycoside, beta-lactam, chloramphenicol, fluoroquinolone, lincomycide, linezolid, lipopeptide, macrolide, sulfonamide, tetracyclines, and vancomycin antibiotics have been previously identified in recycled water or environments directly impacted by irrigation, infiltration, or groundwater recharge using recycled water.122–128 Since antibiotic resistance is a natural phenomenon inherent among many bacteria, studies that compare these abundances to relevant control environments, such as corresponding potable water or environments unimpacted by recycled water are of particular value. While the nature of reusing human wastewater means that prior to treatment, human pathogens or other bacteria carrying ARGs will be enriched compared to other source waters, multiple studies have demonstrated that ARGs are often not removed during treatment, and in some cases, are even amplified.129–132 Additionally, a study by Fahrenfeld et al. found that ARGs may also increase during distribution of recycled water as a broader range of monitored ARGs were present in point of use samples than in samples leaving the treatment plant.122

Figure 2-3: Processes by which antibiotic resistant bacteria and opportunistic pathogens (OPs) can re-grow in RWDSs and relevant exposure routes.

Various features of recycled water potentially make it a prime medium for the growth of antibiotic resistant bacteria and propagation of their associated ARGs during distribution. In particular, residual antibiotics that escape removal during treatment can exert selective pressure and encourage persistence of ARG-carrying bacteria. Though antibiotics will likely be found in recycled water at sub-lethal concentrations, this low level exposure has actually been shown to encourage the persistence of bacteria that carry ARGs via several mechanisms.133–135 Gullberg et al. found that bacteria maintained plasmids carrying beta-lactam resistance genes even at concentrations of antibiotics and heavy metals nearly 140 times below the compound’s minimum inhibitory concentration.136 Other studies have also demonstrated that sublethal antibiotics can

25

stimulate propagation of ARGs by activating horizontal gene transfer.137–142 Prudhomme et al. demonstrated that intermediate concentrations of streptomycin induced genetic transformation in Streptococcus pneumoniae.140 Beaber et al. demonstrated that ‘SOS response’ among induced by the presence of ciprofloxacin enhances transfer of resistance genes via conjugation.142 Low levels of antibiotics or other selective agents also act to encourage adaptive evolution including development of resistance mutants.135

Antibiotics are not the only antimicrobials with potential to select for ARGs in recycled water systems. Heavy metals, such as copper and iron (which are commonly used in distribution systems), have long been suspected to select for ARGs in a variety of environments.143 Metal- driven selection of ARGs is also of concern due to the presence of various heavy metals capable of ARG selection common in many wastewaters, such as copper, zinc, nickel, mercury, and even nanosilver.144,145 Disinfectants have also been known to select for ARGs.146–148 Following chlorination, E. coli carrying the tetA tetracycline resistance gene were found to be even more tolerant to tetracycline than non-chlorinated E. coli.148 Chlorination has also been reported to concentrate a variety of ARGs in potable water.147

In addition to increasing ARGs via mutations, natural selection, and horizontal gene transfer, presence of residual antibiotics can enhance biofilm formation.149 Studies of Staphylococcus aureus, E. coli, and P. aeruginosa have all indicated that sub-inhibitory concentrations of various antibiotics induce biofilm formation.150–152 Extensive biofilm formation provides a fertile environment for the transfer of ARGs via horizontal gene transfer. Dense microbial communities existing in biofilms with extensive cell to cell contact facilitate transfer via conjugation.153,154 Notably, a key component of biofilms, extracellular polymeric substances, is partially comprised of DNA expelled from cells.155 This may provide a reservoir of free DNA- based ARGs, which have been shown to be available for uptake into cells via transformation.156,157 Biofilms themselves offer protection from antibiotics or other antimicrobial agents via the principle of collective resistance, where cells are physically shielded from exposure to the antimicrobial.158

While transmission of antibiotic resistant bacteria is an acute public health threat, the possibility that water reuse may exacerbate the overall spread of antibiotic resistance has been suggested.159 The water cycle as a whole has recently been subject to scrutiny as a potentially important, yet understudied, route for the spread of antibiotic resistance.160–162 Given the gravity of the antibiotic resistance problem and several lines of reasoning that water reuse can contribute to its spread, additional research is urgently needed to determine whether consideration of antibiotic resistance should be of central concern to comprehensive long-term risk management strategies.

Viruses

Though removal of viruses in recycled water is of great importance, the presence of viruses in recycled water is rarely monitored. Treatment goals and regulations regarding virus removal are typically presented in the form of expected log-removal achieved through treatment such as disinfection, largely due to the analytical difficulty of direct virus detection.6 Low recovery rates, complex and time-consuming laboratory culture procedures, slow turn-around time for culture results, and inability of molecular techniques to differentiate viable from non-viable viruses are 26

major challenges. Problems with the indirect monitoring paradigm may arise, however, because viruses may be resistant to some modes of disinfection. For example, adenoviruses are known to resist ultraviolet irradiation.6

A recent study of the viral metagenome (i.e., total DNA extracted from viral component) revealed approximately 108 virus-like particles (VLP) per mL in non-potable recycled water, 1,000 times more than that measured in potable water.163 Further, genetic markers corresponding to viruses targeting eukaryotes were non-detectable in potable water, while two percent of the viruses in recycled water corresponded to eukaryotic hosts. This is logical, indicating that recycled water is more susceptible to carrying viruses associated with humans than traditional potable water.

Bacteriophages, which represent the vast majority of viruses in both potable and recycled waters, have their own relevance to human health as they act as agents of transfer for ARGs among bacteria via transduction. Bacteriophages have been largely neglected as constituents in potable and recycled water, though they have been found to be highly abundant in both raw wastewater (108 VLP/mL) and in potable water (105-106 VLP/mL).163 Though transduction is generally considered a rare transfer event, occurring only once in every 107-109 phage infections,164 the shear abundance of bacteriophages documented in wastewater and potable water suggests that it is likely a significant phenomenon in recycled water.

Amoebae

FLA are of growing concern in drinking water plumbing. Many FLA, such as Acanthamoeba spp. and Vermameoba spp., graze on bacterial biofilms, and in doing so, can serve as an important vector for amplifying and disseminating OPs.165 For example, Legionella, Mycobacterium, and Pseudomonas spp. can amplify within FLA when grazed upon, which enhances their dissemination and virulence.165

FLA themselves can sometimes be pathogenic, as is the case with keratitis or primary amoebic meningoencephalitis (PAM).165 Similar to other OPs, non-ingestion routes of exposure are important for pathogenic FLA. PAM is contracted when N. fowleri is forced into the nasal cavity and migrates to the brain, while keratitis occurs when pathogenic Acanthamoeba spp. infect the eye.166 Such exposures have been documented both in recreational and drinking water.109,167,168 However, relevant to recycled water, inhalation is under investigation as a primary transmission method.168,169

The design of RWDSs can instigate the growth of biofilm, providing a reservoir for FLA and an environment to promote interactions with amoeba-resisting bacteria. Recycled waters have complex microbial communities and high availability of nutrients and other organic matter, creating optimal conditions for biofilm establishment.170 Recent studies have also shown increased chlorine resistance of FLA in the presence of naturally established biofilm171,172 and even non- biofilm Vermamoeba spp. have been observed to resist chlorination.173

The relationship between L. pneumophila and FLA has been the most closely studied. Resistance to amoebae can provide protection from disinfection, competition, environmental stress and predation for L. pneumophila.174–176 Additionally, different FLA have been shown to survive

27

a wide range of temperatures, from 10-45°C, with some cases indicating survival near 0°C, potentially allowing for protection of Legionella spp. and other amoeba-resisting bacteria during winter, while amoeba are encysted.165,177–179 Little is still known about the diversity and abundance of amoebae and their interactions with amoeba-resisting bacteria in drinking water, let alone recycled water. Gaining a better understanding of the interactions between these microorganisms and the ways in which this may aid the growth of pathogenic bacteria is essential for better understanding the exposome associated with recycled water.

Algae

Algae are a common nuisance in recycled water systems. Though algal growth frequently occurs in systems that use open storage rather than in distribution system pipes, algal cells can persist throughout distribution systems where they have been found to correlate with AOC and BDOC.29 Decaying algal cells can even be a source of BDOC, contributing to the regrowth of other microbial constituents. Increased regrowth resulting from organic carbon made available from decaying algal cells has also been linked with a loss of oxygen and dissipation of chlorine residual.29 Elevated concentrations of algae may carry potential for the production of harmful algal toxins. Cyanobacteria toxins have been linked to liver damage, neurotoxicity, gastroenteritis, pneumonia, and even death.180 Though these symptoms primarily arise from ingestion of the toxins, skin irritations and allergic reactions have been noted following dermal contact with cyanobacteria toxins and respiratory disease has been documented following suspected inhalation of the toxins.181 In cases where non-potable reuse occurs, these problems may be particularly challenging to identify as taste and odor complaints from consumers will be unlikely.

CONCLUSION

Given the increasing trend of water reuse across the globe, it is important that all aspects, including end-users, treatment plant management and operation, regulation, and public health protection are taken into consideration in the planning and implementation of water reuse risk management strategies. In this paper, we summarized the inherently different biochemistry of recycled water in the distribution system as a function of various usage and operational factors. We also discussed acute and long-term risks from the chemical and microbial contaminants that may result from multi-dimensional usage routes and the associated exposure risks associated with various end use of the recycled waters.

Increased awareness of traditionally underappreciated routes of exposure is key to the safe use of recycled water. The history of drinking water epidemiology provides numerous examples of infection via atypical routes of exposure. For example, in 2015 an outbreak of Legionnaire’s disease that killed 12 people in New York City was linked to infection via aerosolized bacteria from cooling towers sourced from potable drinking water.182 N. fowleri infection from drinking water has occurred from use for nasal irrigation and in children via bathing or playing on an outdoor water slide.168 Prevention of infection as a result of unintended exposures with recycled water requires proactive action when planning for treatment and distribution of recycled water. One key “critical control point” that must be considered in this planning is the distribution system to avoid degradation of water quality during distribution. Treating recycled water to remove nutrients and achieve biostability is one promising approach to help ensure safe water at the point

28

of use, but additional treatment at the point of use may also be necessary in some cases, as both the physiochemical and microbial water quality change significantly during distribution and in premise plumbing systems. A key research gap exists regarding the most effective approaches for achieving biostability of recycled water during distribution. Specific and cost-effective engineering controls for nutrient recycling and limiting regrowth during distribution must be identified for respective intended uses. Identification of emerging chronic contaminants and microbial contaminants is also important in minimizing potentially harmful exposures. Rigorous studies that examine the health implications of non-traditional routes of exposure are quite limited and are challenging to design given the lack of available knowledge about infectious doses (particularly based on non-oral routes of exposure), magnitude of exposure via non-traditional routes, and concentrations of emerging contaminants that are typical of recycled water. In addition, virulence and individual susceptibility varies widely for many of the microbial constituents discussed, making it important to consider exposure of immunocompromised populations when assessing risk. In addition to these research gaps, development of quantitative microbiological risk assessments (QMRA) would be extremely valuable for assessing the risk associated with the presence OPs, ARGs, FLA, and viruses in recycled water. Epidemiological studies are also critical for linking actual human illness and associated microbial sources with recycled water. Finally, research is needed to tailor treatment processes to serve specific intended end uses (e.g. Table 2- 1), along with addressing emerging concerns identified here, while also developing best management practices for distribution systems and premise plumbing for preventing re-growth and deterioration of water quality in RWDSs.

While overarching regulations that consider the comprehensive implications and scope of water recycling are currently lacking in many places, there are also practical lessons we can learn from international leaders on adopting a comprehensive risk management approach towards water reuse, notably the Australian Guidelines for Water Recycling, the WHO’s framework of WSP, and the HACCP paradigm. Complementing these strategies with a holistic approach focused on the human exposome creates a framework in which consideration of user exposures drives establishment of water quality standards. Though a regulatory framework that addresses the exposure risks and potential for regrowth associated with use of recycled water is an ideal long term goal, an interim approach of more basic best management practices, as suggested by Jjemba et al., are a reasonable starting place to enable municipalities utilizing recycled water to proactively act to limit bacterial regrowth and preserve water quality at the point of use.29 Best management practices should continually be revised as knowledge gaps are addressed, to ensure that the most meaningful water quality indicators are targeted.

Adoption of the human exposome paradigm aims to ensure comprehensive understanding of the risks and uncertainties regarding alternative recycled water sources. Enhanced knowledge could provide critical guidance on safe management and inform much-needed regulations as the use of recycled water expands. However, while this exposome approach highlights the multi- dimensional risks and uncertainties regarding use of recycled water, it also must be recognized that water reuse plays an integral role in addressing the grand challenge of water scarcity. It is estimated that a third of the world’s population is currently living with moderate to high levels of water stress183 and approximately 50% will suffer water shortages by 2025.184 Implementing water reuse projects is imperative to meet water needs in drought-stricken areas, despite potential risks and concerns. As estimated by Brown, current groundwater sources, serving more than half of the

29

world’s population, are largely overdrafted.185 Lack of new alternative water supplies, compounded with increasing water demand, would further intensify water scarcity stress. Schoreder et al. has estimated that the potential benefits of reuse could offset water supply for a community of 1 million people by 75 million gallons per day.186 In cases where water scarcity lends to the likelihood of de facto reuse, then it is better to have intentional reuse guided by best management practices to minimize risks.

Equally important to the exposure risk from recycled water is lack of access to traditional potable water sources and poor water quality due to degraded source water. Globally, there are over five million deaths associated with poor water quality every year.187 Achieving an environmentally sustainable and socially beneficially water demand management plan requires proactive evaluation of the highest priority needs and identification of the key drivers and barriers to the implementation of water recycling projects. Positive associations between information availability and the acceptance of water reuse have been noted.188–190 As end users become more educated about this alternative water source, their willingness to use recycled water increases.191 Nonetheless, comprehensively addressing all possible public health concerns will be an essential pillar to advancing water sustainability.

ACKNOWLEDGEMENTS

This work was supported by The Alfred P. Sloan Foundation Microbiology of the Built Environment Program, the National Science Foundation Award # 1438328, the Water Environment Research Foundation Paul L. Busch Award, and Graduate Research Fellowship Program Grant # DGE 0822220. We would like to thank Owen Strom for assistance creating figures.

REFERENCES

1 J. Anderson, Water Sci. Technol. Water Supply, 2003, 3, 1–10. 2 D. Goodwin, M. Raffin, P. Jeffrey and H. M. Smith, Environ. Sci. Water Res. Technol., 2015, 1, 709–722. 3 D. Gerrity, B. Pecson, R. S. Trussell and R. R. Trussell, J. Water Supply Res. T., 2013, 62, 321-338. 4 P. Jjemba, W. Johnson, Z. Bukhari and M. LeChevallier, J. Water Reuse Desalin., 2014, 4, 209. 5 F. Digiano and M. Aitkin, Workshop on Water Reuse and Dual Systems for Potable and Non‐Potable Water Supply. U.S. Environmental Protection Agency, Final Report for Cooperative Agreement: 83422201‐0, 2010, 51 pp. 6 USEPA, Guidelines for Water Reuse, United States Environmental Protection Agency, 2012. 7 NRMMC EPHC & AHMC, Australian Guidelines for Water Recycling: Managing Health and Environmental Risks (Phase 1), Natural Resource Ministerial Management Council, Environment Protection and Heritage Council and Australian Health Ministers, 2006. 8 J. Lahnsteiner and G. Lempert, Water Sci. Technol., 2007, 55, 441-448. 9 M. Agulló-Barceló, F. Oliva and F. Lucena, Environ. Sci. Pollut. Res., 2013, 20, 4448– 4454.

30

10 V. J. Harwood, A. D. Levine, T. M. Scott, V. Chivukula, J. Lukasik, S. R. Farrah and J. B. Rose, Appl. Environ. Microbiol., 2005, 71, 3163–3170. 11 J. Swierc, D. Page, J. Van Leeuwen and P. J. Dillon, CSIRO L. Water Tech. Rep., 2005, 20. 12 J. Bartram, L. Corrales, A, Davison, D. Deere, D. Drury, B. Gordon, G. Howard, A. Rinehold and M. Stevens, Water safety plan manual: step-by-step risk management for drinking-water suppliers, World Health Organization, Geneva, 2009. 13 G. W. Miller and D. P. Jones, Toxicol. Sci., 2014, 137, 1–2. 14 C. P. Wild, Cancer Epidemiol. Biomarkers Prev., 2005, 14, 1847–1850. 15 J. Rice, A. Wutich and P. Westerhoff, Environ. Sci. Technol., 2013, 47, 11099–11105. 16 M. Lafforgue and V. Lenouvel, Environ. Sci. Water Res. Technol., 2015, 1, 622–631. 17 J. Sutherland, E. Steinle-Darling, A. Salveson and E. Mancha, presented in part at Water Quality Technology Conference, Fort Worth, July, 2015. 18 D. G. Addiss, J. P. Davis, M. LaVenture, P. J. Wand, M. A. Hutchinson and R. M. McKinney, Am. J. Epidemiol., 1989, 130, 557–568. 19 R. E. Rosenberg Goldstein, S. a Micallef, S. G. Gibbs, X. He, A. George, A. A. R. Sapkota, S. W. Joseph and A. A. R. Sapkota, Int. J. Environ. Res. Public Health, 2014, 11, 4340– 4355. 20 F. Brissaud, E. Blin, S. Hemous and L. Garrelly, Water Sci. Technol., 2008, 57, 781–787. 21 K. E. Borboudaki, N. V. Paranychianakis and K. P. Tsagarakis, Environ. Manage., 2005, 36, 610–623. 22 C. P. Gerba, C. Wallis and J. L. Melnick, Appl. Microbiol., 1975, 30, 229–237. 23 W. G. Hlady, R. C. Mullen, C. S. Mintz, B. G. Shelton, R. S. Hopkins and G. L. Daikos, Am. J. Epidemiol., 1993, 138, 555–562. 24 T. E. Haupt, R. T. Heffernan, J. J. Kazmierczak, H. Nehls-Lowe, B. Rheineck, C. Powell, K. K. Leonhardt, A. S. Chitnis and J. P. Davis, Infect. Control Hosp. Epidemiol., 2012, 33, 185–191. 25 T. N. Palmore, F. Stock, M. White, M. Bordner, A. Michelin, J. E. Bennett, P. R. Murray and D. K. Henderson, Infect Control Hosp Epidemiol, 2009, 30, 764–768. 26 J. B. Rose and C. S. Clark, Crit. Rev. Environ. Control, 1986, 16, 231–256. 27 ASCE 2013 Report Card for America's Infrastructure, http://infrastructurereportcard.org/a/#p/home, (accessed December 2015). 28 H. Wang, C. Hu, X. Hu, M. Yang and J. Qu, Water Res., 2012, 46, 1070–1078. 29 P. K. Jjemba, L. a. Weinrich, W. Cheng, E. Giraldo and M. W. LeChevallier, Appl. Environ. Microbiol., 2010, 76, 4169–4178. 30 P. Sarin, V. L. Snoeyink, J. Bebee, K. K. Jim, M. A. Beckett, W. M. Kriven and J. A. Clement, Water Res., 2004, 38, 1259–1269. 31 USEPA, http://archive.epa.gov/region03/dclead/web/html/index-3.html (accesed January 2016). 32 Mlive, http://www.mlive.com/news/flint/index.ssf/2015/09/new_testing_shows_flint_water.html (accessed January 2016) 33 MATCOR, http://www.matcor.com/water-line-break-corrosion-a-headache-for-tucson- water/ (accessed January 2016) 34 F. Yang, B. Shi, Y. Bai, H. Sun, D. a. Lytle and D. Wang, Water Res., 2014, 59, 46–57. 35 M. Tang, S. Triantafyllidou and M. A. Edwards, Environ. Sci. Technol., 2015, 49, 8697- 8703.

31

36 M. V. Story and C. E. Kaucner, Water Quality Research Australia: Research Report 79, 2009, 122 pp. 37 H. Rufenacht and H. Guibentif, J. Water Supply Res. Technol., 1997, 46, 196–201. 38 R. M. Willis, R. A. Stewart, P. R. Williams, C. H. Hacker, S. C. Emmonds and G. Capati, Desalination, 2011, 272, 201–211. 39 J. Nawrocki, U. Raczyk-Stanisławiak, J. Świetlik, A. Olejnik and M. J. Sroka, Water Res., 2010, 44, 1863–1872. 40 P. Ji, J. Parks, M. A. Edwards and A. Pruden, PLoS One, 2015, 10, e0141087. 41 W. J. Rhoads, A. Pruden and M. A. Edwards, Environ. Sci. Water Res. Technol., 2016, 2, 164–173. 42 C. Z. Hu H-Y, Du Y, Wu Q-Y, Zhao X, and Tang X, Sci Total Env., 2016, 551-552, 133– 142. 43 H. K. Shon, S. Vigneswaran and S. A. Snyder, Crit. Rev. Environ. Sci. Technol., 2006, 36, 327–374. 44 B. Rittman and V. Snoeyink, J. Am. Water Works Assoc., 1984, 76, 106–114. 45 P. W. J. J. van der Wielen and D. van der Kooij, Water Res. , 2010, 44 , 4860–4867. 46 P. Thayanukul, F. Kurisu, I. Kasuga and H. Furumai, Water Res., 2013, 47, 225–232. 47 L. A. Weinrich, P. K. Jjemba, E. Giraldo and M. W. LeChevallier, Water Res., 2010, 44, 5367–75. 48 M. Karim and M. W. LeChevallier, Microbiological quality of reuse water; Internal Report, Voorhees, NJ, 2005. 49 M. W. LeChevallier, N. J. Welsh and D. B. Smith, Appl. Envir. Microbiol., 1996, 62, 2201– 2221. 50 X. Zhao, H. Hu, S. Liu, F. Jiang, X. Shi, M. Li and X. Xu, Front. Environ. Sci. Eng., 2013, 7, 483–491. 51 E. Khan, R. W. Babcock Jr., I. H. Suffet and M. K. Stenstrom, Water Environ. Res., 1998, 70, 1025–1032. 52 L. S. McNeill and M. Edwards, J. Am. Water Works Assoc., 2014, 93, 88–100. 53 M. Edwards and N. Sprague, Corros. Sci., 2001, 43, 1–18. 54 A. K. Camper, Int. J. Food Microbiol., 2004, 92, 355–64. 55 M. W. LeChevallier, C. D. Lowry, R. G. Lee and D. L. Gibbon, J. Am. Water Works Assoc., 1993, 85, 111–123. 56 H. Wang, C. Hu, L. Zhang, X. Li, Y. Zhang and M. Yang, Water Res., 2014, 65, 362–370. 57 Y. Zhu, H. Wang, X. Li, C. Hu, M. Yang and J. Qu, Water Res., 2014, 60, 174–181. 58 A. K. Camper and P. Dircks, Distribution system as reactor, 1996, https://www.biofilm.montana.edu/resources/images/industrial-systems- processes/distribution-system-reactor.html, (accessed January 2016). 59 S. T. Powell, J. Am. Water Works Assoc., 1921, 8, 522–529. 60 A. M. Buswell, J. Am. Water Works Assoc., 1938, 30, 1651–1654. 61 N. J. Howard, J. Am. Water Works Assoc., 1940, 32, 1501–1506. 62 J. Hall, A. D. Zaffiro, R. B. Marx, P. C. Kefauver, E. R. Krishnan, R.C. Haught and J. G. Herrmann, J. Am. Water Works Assoc., 2007, 99, 66–77. 63 D. Berry, C. Xi and L. Raskin, Curr. Opin. Biotechnol., 2006, 17, 297–302. 64 P. Piriou, B. E. Rittmann and J. E. Woolschlager, Urban water J., 2005, 2, 69–79. 65 S. Masters, H. Wang, A. Pruden and M. A. Edwards, Water Res., 2015, 68, 140–149. 66 O. M. Ajibode, C. Rock, K. Bright, J. E. T. Mclain, C. P. Gerba and I. L. Pepper, J. Water

32

Reuse Desalin., 2013, 3, 185–196. 67 M. K. Ramseier, A. Peter, J. Traber and U. von Gunten, Water Res., 2011, 45, 2002–2010. 68 K. A. Steed and J. O. Falkinham, Appl. Environ. Microbiol., 2006, 72, 4007–4011. 69 R. H. Taylor, J. O. Falkinham, C. D. Norton and M. W. LeChevallier, Appl. Environ. Microbiol., 2000, 66, 1702–1705. 70 I. Marchesi, P. Marchegiano, A. Bargellini, S. Cencetti, G. Frezza, M. Miselli and P. Borella, J. Hosp. Infect., 2011, 77, 47–51. 71 M. Dupuy, S. Mazoua, F. Berne, C. Bodet, N. Garrec, P. Herbelin, F. Ménard-Szczebara, S. Oberti, M.-H. Rodier, S. Soreau, F. Wallet and Y. Héchard, Water Res., 2011, 45, 1087– 94. 72 J.-J. Huang, H.-Y. Hu, F. Tang, Y. Li, S.-Q. Lu and Y. Lu, Water Res., 2011, 45, 2775– 2781. 73 M. Guo, J. Huang, H. Hu and W. Liu, Biomed. Environ. Sci., 2011, 24, 400–407. 74 J. Süss, S. Volz, U. Obst and T. Schwartz, Water Res., 2009, 43, 3705–16. 75 R. M. Clark and M. Sivaganesan, J. Environ. Eng., 1998, 124, 1203–1210. 76 S. W. Krasner, Philos. Trans. A. Math. Phys. Eng. Sci., 2009, 367, 4077–4095. 77 T. Sirivedhin and K. A. Gray, Water Res., 2005, 39, 1025–1036. 78 C. Nurizzo, M. Antonelli, M. Profaizer and L. Romele, Desalination, 2005, 176, 241–253. 79 C. C. Wang, Z. G. Niu and Y. Zhang, J. Hazard. Mater., 2013, 262, 179–188. 80 C. M. Coetsier, S. Spinelli, L. Lin, B. Roig and E. Touraud, Environ. Int., 2009, 35, 787– 792. 81 S. Mompelat, B. Le Bot and O. Thomas, Environ. Int., 2009, 35, 803–814. 82 M. Gros, M. Petrović, A. Ginebreda and D. Barceló, Environ. Int., 2010, 36, 15–26. 83 A. Jelic, M. Gros, A. Ginebreda, R. Cespedes-Sánchez, F. Ventura, M. Petrovic and D. Barcelo, Water Res., 2011, 45, 1165–1176. 84 M. Huerta-Fontela, M. T. Galceran and F. Ventura, Water Res., 2011, 45, 1432–1442. 85 Y. Ma, M. Li, M. Wu, Z. Li and X. Liu, Sci. Total Environ., 2015, 518-519, 498–506. 86 D. W. Kolpin, M. Skopec, M. T. Meyer, E. T. Furlong and S. D. Zaugg, Sci. Total Environ., 2004, 328, 119–30. 87 G. A. Loraine and M. E. Pettigrove, Environ. Sci. Technol., 2006, 40, 687–695. 88 J. D. Peterson, R. R. Murphy, Y. Jin, L. Wang, M. B. Nessl and K. Ikehata, Water Environ. Res., 2011, 83, 1853–1875. 89 F. Busetti, M. Ruff and K. L. Linge, Environ. Sci. Water Res. Technol., 2015, 1, 659–667. 90 S. Weber, S. Khan and J. Hollender, Desalination, 2006, 187, 53–64. 91 S. Khan, in Sustainable Water, ed. R. Hester and R. Harrison,The Royal Society of Chemistry, London, 2011, ch. 6, pp. 114–137. 92 M. M. Huber, S. Canonica, G. Y. Park and U. Von Gunten, Environ. Sci. Technol., 2003, 37, 1016–1024. 93 I. H. Kim, N. Yamashita, Y. Kato and H. Tanaka, Water Sci. Technol., 2009, 59, 945–955. 94 L. H. M. L. M. Santos, A. N. Araújo, A. Fachini, A. Pena, C. Delerue-Matos and M. C. B. S. M. Montenegro, J. Hazard. Mater., 2010, 175, 45–95. 95 C. R. Tyler, S. Jobling and J. P. Sumpter, Crit. Rev. Toxicol., 2008, 28, 319–361. 96 D. W. Kolpin, E. T. Furlong, M. T. Meyer, E. M. Thurman, S. D. Zaugg, L. B. Barber and H. T. Buxton, Environ. Sci. Technol., 2002, 36, 1202–1211. 97 R. Durand and G. Schwebach, Am. J. Public Health, 1989, 79, 1659–1660. 98 B. Sheikh, R. P. Cort, W. R. Kirkpatrick, R. S. Jaques and T. Asano, 1990, 62, 216–226.

33

99 R. L. Ward, D. R. Knowlton, J. Stober, W. Jakubowski, T. Mills, P. Graham and D. E. Camann, Water Res., 1989, 23, 1503–1509. 100 T. Egli, Water Res., 2010, 44, 4826–37. 101 H. Wang, M. A. Edwards, J. O. Falkinham 3rd and A. Pruden, Env. Sci Technol, 2013, 47, 10117–10128. 102 A. Pruden, M. A. Edwards, J. O. Falkinham III, M. Arduino, J. Bird, R. Birdnow, E. Bedard, A. Camper, J. Clancy, E. Hilborn, V. Hill, A. Martin, S. Masters, N. R. Pace, M. Prevost, A. Rosenblatt, W. Rhoads, J. E. Stout and Y. Zhang, State of the Science and Research Needs for Opportunistics Pathogens in Premise Plumbing: Methodology, microbial ecology, and epidemiology, Water Research Foundation, Denver, CO, 2013. 103 J. M. Brunkard, E. Ailes, V. A. Roberts, V. Hill, E. D. Hilborn, G. F. Craun, A. Rajasingham, A. Kahler, L. Garrison, L. Hicks, J. Carpenter, T. J. Wade, M. J. Beach and J. S. Yoder, Morb. Mortal. Wkly. Rep., 2008, 57, 1–38. 104 D. W. Fraser, T. R. Tsai, W. Orenstein and W. E. Parkin, N. Engl. J. Med., 1989, 297. 105 A. Y. Peleg, H. Seifert and D. L. Paterson, Clin. Microbiol. Rev., 2008, 21, 538–582. 106 C. R. Horsburgh, N. Engl. J. Med., 1991, 324, 1332–1338. 107 F. D. Lowy, N. Engl. J. Med.,1998, 339, 520-532. 108 G. P. Bodey, R. Bolivar, V. Fainstein and L. Jadeja, Rev. Infect. Dis., 1983, 5, 279–313. 109 F. Marciano-Cabral and G. Cabral, Clin. Microbiol. Rev., 2003, 16, 273–307. 110 J. Falkinham, Int. J. Environ. Res. Public Health, 2015, 12, 4533–4545. 111 H. Wang, M. Edwards, J. O. Falkinham, and A. Pruden, Appl Environ Microbiol., 2012, 78, 6285-6294. 112 US CDC, Antibiotic Resistance Threats in the United States, 2013. 113 A. Pruden, M. Arabi and H. N. Storteboom, Environ. Sci. Technol., 2012, 46, 11541–11549. 114 D. W. Graham, S. Olivares-Rieumont, C. W. Knapp, L. Lima, D. Werner and E. Bowen, Environ. Sci. Technol., 2011, 45, 418–424. 115 G. C. A. Amos, P. M. Hawkey, W. H. Gaze and E. M. Wellington, J. Antimicrob. Chemother., 2014, 69, 1785–1791. 116 E. Marti, J. Jofre and J. L. Balcazar, PLoS One, 2013, 8, e78906. 117 R. Benveniste and J. Davies, Proc. Natl. Acad. Sci. U. S. A., 1973, 70, 2276–2280. 118 K. J. Forsberg, A. Reyes, B. Wang, E. M. Selleck, M. O. A. Sommer and G. Dantas, Science, 2012, 337, 1107–1111. 119 K. A. Davis, J. J. Stewart, H. K. Crouch, C. E. Florez and D. R. Hospenthal, Clin. Infect. Dis., 2004, 39, 776–782. 120 C. von Eiff, K. Becker, K. Machka, H. Stammer and G. Peters, N. Engl. J. Med., 2001, 344, 11–16. 121 D. Bogaert, R. De Groot and P. W. M. Hermans, Lancet Infect. Dis., 2004, 4, 144–154. 122 N. L. Fahrenfeld, Y. J. Ma, M. O’Brien and A. Pruden, Front. Microbiol., 2013, 4, 1-10. 123 F. Barker-Reid, E. M. Fox and R. Faggian, J. Water Health, 2010, 8, 521–531. 124 J. McLain and C. Williams, Sustainability, 2014, 6, 1313–1327. 125 R. E. R. Goldstein, S. A. Micallef, S. G. Gibbs, A. George, E. Claye, A. Sapkota, S. W. Joseph and A. R. Sapkota, Sci. Total Environ., 2014, 466, 404–411. 126 F. H. Wang, M. Qiao, Z. E. Lv, G. X. Guo, Y. Jia, Y. H. Su and Y. G. Zhu, Environ. Pollut., 2014, 184, 247–253. 127 F. Wang, M. Qiao, J. Su, Z. Chen, X. Zhou and Y. Zhu, Environ. Sci. Technol., 2014, 48, 9079–9085.

34

128 U. Bockelmann, H. H. Dorries, M. N. Ayuso-Gabella, M. S. de Marcay, V. Tandoi, C. Levantesi, C. Masciopinto, E. Van Houtte, U. Szewzyk, T. Wintgens and E. Grohmann, Appl. Environ. Microbiol., 2009, 75, 154–163. 129 L. Rizzo, C. Manaia, C. Merlin, T. Schwartz, C. Dagot, M. C. Ploy, I. Michael and D. Fatta- Kassinos, Sci. Total Environ., 2013, 447, 345–360. 130 Y. L. Zhang, C. F. Marrs, C. Simon and C. W. Xi, Sci. Total Environ., 2009, 407, 3702– 3706. 131 A. Schlüter, R. Szczepanowski, A. Pühler and E. M. Top, FEMS Microbiol. Rev., 2007, 31, 449–477. 132 M. Munir, K. Wong and I. Xagoraraki, Water Res., 2011, 45, 681–693. 133 A. Liu, A. Fong, E. Becket, J. Yuan, C. Tamae, L. Medrano, M. Maiz, C. Wahba, C. Lee, K. Lee, K. P. Tran, H. Yang, R. M. Hoffman, A. Salih and J. H. Miller, Antimicrob. Agents Chemother., 2011, 55, 1204–1210. 134 G. McVicker, T. K. Prajsnar, A. Williams, N. L. Wagner, M. Boots, S. a Renshaw and S. J. Foster, PLoS Pathog., 2014, 10, e1003959. 135 D. I. Andersson and D. Hughes, Nat. Rev. Microbiol., 2014, 12, 465–478. 136 E. Gullberg, L. M. Albrecht, C. Karlsson, L. Sandegren and D. I. Andersson, MBio, 2014, 5, 19–23. 137 P. Y. Zhang, P. P. Xu, Z. J. Xia, J. Wang, J. Xiong and Y. Z. Li, Fems Microbiol. Lett., 2013, 348, 149–156. 138 Z. J. Xia, J. Wang, W. Hu, H. Liu, X. Z. Gao, Z. H. Wu, P. Y. Zhang and Y. Z. Li, J. Ind. Microbiol. Biotechnol., 2008, 35, 1157–1163. 139 B. Song, G. R. Wang, N. B. Shoemaker and A. A. Salyers, J. Bacteriol., 2009, 191, 1078– 1082. 140 M. Prudhomme, L. Attaiech, G. Sanchez, B. Martin and J. P. Claverys, Science, 2006, 313, 89–92. 141 C. Úbeda, E. Maiques, E. Knecht, Í. Lasa, R. P. Novick and J. R. Penadés, Mol. Microbiol., 2005, 56, 836–844. 142 J. W. Beaber, B. Hochhut and M. K. Waldor, Nature, 2004, 427, 72–74. 143 C. Baker-Austin, M. S. Wright, R. Stepanauskas and J. V. McArthur, Trends Microbiol., 2006, 14, 176–182. 144 L. Sörme and R. Lagerkvist, Sci. Total Environ., 2002, 298, 131–145. 145 S. K. Brar, M. Verma, R. D. Tyagi and R. Y. Surampalli, Waste Manag., 2010, 30, 504– 520. 146 D. P. Karumathil, H.-B. Yin, A. Kollanoor-Johny and K. Venkitanarayanan, Int. J. Environ. Res. Public Health, 2014, 11, 1844–1854. 147 P. Shi, S. Jia, X. X. Zhang, T. Zhang, S. Cheng and A. Li, Water Res., 2013, 47, 111–120. 148 J. J. Huang, H.-Y. Hu, Y.-H. Wu, B. Wei and Y. Lu, Chemosphere, 2013, 90, 2247–2253. 149 A. Boehm, S. Steiner, F. Zaehringer, A. Casanova, F. Hamburger, D. Ritz, W. Keck, M. Ackermann, T. Schirmer and U. Jenal, Mol. Microbiol., 2009, 72, 1500–1516. 150 M. Blickwede, R. Goethe, C. Wolz, P. Valentin-Weigand and S. Schwarz, J. Antimicrob. Chemother., 2005, 56, 315–323. 151 D. Li, A. Renzoni, T. Estoppey, C. Bisognano, P. Francois, W. L. Kelley, D. P. Lew, J. Schrenzel and P. Vaudaux, Antimicrob. Agents Chemother., 2005, 49, 916–924. 152 L. R. Hoffman, D. A. D’Argenio, M. A. MacCoss, Z. Zhang, R. A. Jones and S. I. Miller, Nature, 2005, 436, 1171–1175.

35

153 M. Hausner and S. Wuertz, Appl. Environ. Microbiol., 1999, 65, 3710–3713. 154 S. Molin and T. Tolker-Nielsen, Curr. Opin. Biotechnol., 2003, 14, 255–261. 155 T. Das, S. Sehar and M. Manefield, Environ. Microbiol. Rep., 2013, 5, 778–786. 156 Y. Li, P. C. Y. Lau, J. H. Lee, R. P. Ellen and D. G. Cvitkovitch, Society, 2001, 183, 897– 908. 157 S. Hannan, D. Ready, A. S. Jasni, M. Rogers, J. Pratten and A. P. Roberts, FEMS Immunol. Med. Microbiol., 2010, 59, 345–349. 158 N. M. Vega and J. Gore, Curr. Opin. Microbiol., 2014, 21, 28–34. 159 A. Pruden, Environ. Sci. Technol., 2014, 48, 5–14. 160 N. J. Ashbolt, A. Amezquita, T. Backhaus, P. Borriello, K. K. Brandt, P. Collignon, A. Coors, R. Finley, W. H. Gaze, T. Heberer, J. R. Lawrence, D. G. J. Larsson, S. A. McEwen, J. J. Ryan, J. Schonfeld, P. Silley, J. R. Snape, C. Van den Eede and E. Topp, Environ. Health Perspect., 2013, 121, 993–1001. 161 E. M. H. Wellington, A. B. Boxall, P. Cross, E. J. Feil, W. H. Gaze, P. M. Hawkey, A. S. Johnson-Rollings, D. L. Jones, N. M. Lee, W. Otten, C. M. Thomas and a P. Williams, Lancet Infect. Dis., 2013, 13, 155–165. 162 H. K. Allen, J. Donato, H. H. Wang, K. a Cloud-Hansen, J. Davies and J. Handelsman, Nat. Rev. Microbiol., 2010, 8, 251–259. 163 K. Rosario, C. Nilsson, Y. W. Lim, Y. Ruan and M. Breitbart, Environ. Microbiol., 2009, 11, 2806–2820. 164 M. Muniesa, M. Colomer-lluch and J. Jofre, Future Microbiol., 2013, 8, 739–752. 165 J. M. Thomas, N. J. Ashbolt, N. Exposure, U. States, E. P. Agency, J. M. Thomas and N. J. Ashbolt, Environ. Sci. Technol., 2011, 45, 860–869. 166 US CDC, Antibiotic Resistance Threats in the United States, 2013, 114 pp. 167 B. Blair, P. Sarkar, K. R. Bright, F. Marciano-Cabral and C. P. Gerba, Emerg. Infect. Dis., 2008, 14, 1498–1499. 168 T. A. Bartrand, J. J. Causey and J. L. Clancy, J. Am. Water Works Assoc., 2014, 106, E418– E432. 169 A. J. Martinez and G. S. Visvesvara, Brain Pathol., 1997, 7, 583–598. 170 M. C. Liang, Z. G. Ning and Y. K. Li, Environ. Earth Sci., 2015, 73, 4325–4338. 171 V. Thomas, J. Loret, M. Jousset and G. Greub, Environ. Microbiol., 2008, 10, 2728–2745. 172 H. C. Miller, J. Wylie, G. Dejean, A. H. Kaksonen, D. Sutton, K. Braun and G. J. Puzon, Environ. Sci. Technol., 2015, 49, 11125–11131. 173 J. M. Thomas and N. J. Ashbolt, Environ. Sci. Technol., 2011, 45, 860–869. 174 J. Lu, H. Y. Buse, V. Gomez-Alvarez, I. Struewing, J. Santo Domingo and N. J. Ashbolt, J. Appl. Microbiol., 2014, 117, 905–918. 175 M. T. García, S. Jones, C. Pelaz, R. D. Millar and Y. Abu Kwaik, Environ. Microbiol., 2007, 9, 1267–1277. 176 P. K. Jjemba, W. Johnson, Z. Bukhari, and M. W. LeChevallier. Pathogens, 2015, 4, 470- 502. 177 J. L. Griffin, Science, 1972, 178, 869–870. 178 H. Y. Buse and N. J. Ashbolt, Lett. Appl. Microbiol., 2011, 53, 217–224. 179 M. A. Söderberg, J. Dao, S. R. Starkenburg and N. P. Cianciotto, Appl. Environ. Microbiol., 2008, 74, 5583–5588. 180 B. C. Hitzfeld, S. J. Höger and D. R. Dietrich, Environ. Health Perspect., 2000, 108, 113– 122.

36

181 G. Codd, S. Bell, K. Kaya, C. Ward, K. Beattie and J. Metcalf, Eur. J. Phycol., 1999, 34, 405–415. 182 Mueller B. New York Times. http://www.nytimes.com/2015/10/02/nyregion/legionnaires- bacteria-regrew-in-bronx-cooling-towers-that-were-disinfected.html?_r=0, (accessed January 2016). 183 B. E. Hatt, A. Deletic and T. D. Fletcher, J. Environ. Manage., 2006, 79, 102–113. 184 Austrailian Government National Health and Medical Research Council, 2004. 185 L. R. Brown, Full Planet, Empty Plates: The New Geopolitics of Food Scarcity, 2012. 186 E. Schroeder, G. Tchobanoglous, H. L. Leverenz and T. Asano, Natl. Water Resesearch Inst. White Pap., 2012, 1–20. 187 D. S. Adapa S, Bhullar N, J Clean Prod, 2016. 188 D. Baumann, Appl. Geogr., 1983, 3, 79–84. 189 J. G. Lohman, L.C., Milliken, U.S. Dep. Inter. Denver, 1985. 190 S. Dolnicar, A. Hurliman, and L. D. Nghiem, J. Environ. Manage., 2010, 91, 1288–1293. 191 K. P. Tsagarakis, A. N. Menegaki, K. Siarapi and F. Zacharopoulou, J. Risk Res., 2013, 16, 133–144. 192 T. Asano, F. L. Burton, H. L. Leverenz, R. Tsuchihashi, and G. Tchobanoglous, Water Reuse: Issues, Technologies, and Applications, McGraw-Hill, New York City, 2016. 193 R. E. Rosenberg Goldstein, S. A. Micallef, S. G. Gibbs, J. A. Davis, X. He, A. George, L. M. Kleinfelter, N. A. Schreiber, S. Mukherjee, A. Sapkota, S. W. Joseph and A. R. Sapkota, Environ. Health Perspect., 2012, 120, 1551–1558. 194 M. W. LeChevallier and R. J. Seidler, Appl. Environ. Microbiol., 1980, 30, 739–742. 195 P. Rusin, J. Rose, C. Haas and C. Gerba, Rev Env. Contam Toxicol, 1997, 152, 57–83. 196 F. Xianmin, Z. Wenyu, W. Yuehua, Z. Donghai, J. Xiaoming and L. Shijie, Anat. Rec., 2015, 298, 1509–1517. 197 P. K. Jjemba, L. Weinrich, W. Cheng, E. Giraldo and M. W. LeChevallier, Guidance Document on the Microbiological Quality and Biostability of Reclaimed Water Following Storage and Distribution, 2010. 198 L. Macmillian, New York Times Green Blog, http://green.blogs.nytimes.com/2012/10/10/do-not-publish-on-the-slopes-gliding-on- wastewater/?_r=0, (accessed January 2016). 199 Q. Y. Wu, Y. Y. Li and H. Y. Hu, Acta Sci. Circumstantiae, 2013, 33, 844–849. 200 T. W. Armstrong and C. N. Haas, Risk Anal., 2007, 27, 1581–1596. 201 T. P. Primm, C. A. Lucero and J. O. Falkinham, Clin. Microbiol. Rev., 2004, 17, 98–106. 202 B. J. Williams, J. Dehnbostel and T. S. Blackwell, Respirology, 2010, 15, 1037–1056. 203 Y. J. Lee, J. H. Kim, H. J. Sohn, J. Lee, S. Y. Jung, Y. J. Chwae, K. Kim, S. Park and H. J. Shin, Parasite Immunol., 2011, 33, 382–389. 204 H. Ryu, A. Alum and M. Abbaszadegan, Environ. Sci. Technol., 2005, 39, 8600–8605. 205 C. W. Knapp, S. M. McCluskey, B. K. Singh, C. D. Campbell, G. Hudson and D. W. Graham, PLoS One, 2011, 6, e27300. 206 M. J. Mohammad Rusan, S. Hinnawi and L. Rousan, Desalination, 2007, 215, 143–152. 207 B. Kurenbach, D. Marjoshi, C. F. Amábile-cuevas, G. C. Ferguson, W. Godsoe, P. Gibson and J. A. Heinemann, mBio, 2015, 6, 1–9. 208 M. Salgot, E. Huertas, S. Weber, W. Dott and J. Hollender, Desalination, 2006, 187, 29– 40. 209 J. K. Edzwald, Water Quality & Treatment: A Handbook on Drinking Water, Sixth Edition,

37

McGraw-Hill, New York City, 2011. 210 C. Volk, C. Renner, C. Robert and J. C. Joret, Environ. Technol., 1994, 15, 545–556. 211 P. Servais, P. Laurent and D. Gatel, presented in part at Proc. of the Annual AWWA Water Qual. Technol. Conf., New Orleans, 1995. 212 P. Laurent, M. Prévost, J. Cigana, P. Niquette and P. Servais, Water Res., 1999, 33, 1387– 1398. 213 P. Niquette, P. Servais and R. Savoir, Water Res., 2001, 35, 675–682. 214 C. J. Volk and M. W. LeChevallier, J. Am. Water Works Assoc., 2000, 92, 64–76. 215 D. van der Kooij, A. Visser and W. A. M. Hijnen, J. Am. Water Works Assoc., 1982, 74, 540–545. 216 M. W. LeChevallier, W. Schulz and R. G. Lee, Appl. Envir. Microbiol., 1991, 57, 857–862. 217 M. Vital, H. P. Fuchslin, F. Hammes and T. Egli, Microbiology, 2007, 153, 1993–2001. 218 D. Gerrity, S. Gamage, J. C. Hola dy, D. B. Mawhinney, O. Quiñones, R. A. Trenholm and S. A. Snyder, Water Res., 2011, 45, 2155–65. 219 Ç. Kalkan, K. Yapsakli, B. Mertoglu, D. Tufan and A. Saatci, Desalination, 2011, 265, 266– 273. 220 A. N. Pisarenko, B. D. Stanford, D. Yan, D. Gerrity and S. A. Snyder, Water Res., 2012, 46, 316–26.

38

CHAPTER 3 : STORMWATER LOADINGS OF ANTIBIOTIC RESISTANCE GENES IN AN URBAN STREAM

Emily Garner, Romina Benitez, Emily von Wagoner, Richard Sawyer, Erin Schaberg, W. Cully Hession, Leigh-Anne H. Krometis, Brian D. Badgley, and Amy Pruden

ABSTRACT

Antibiotic resistance presents a critical public health challenge and the transmission of antibiotic resistance via environmental pathways continues to gain attention. Factors driving the spread of antibiotic resistance genes (ARGs) in surface water and sources of ARGs in urban stormwater have not been well-characterized. In this study, five ARGs (sul1, sul2, tet(O), tet(W), and erm(F)) were quantified throughout the duration of three storm runoff events in an urban inland stream. Storm loads of all five ARGs were significantly greater than during equivalent background periods. Neither fecal indicator bacteria measured (E. coli or enterococci) was significantly correlated with sul1, sul2, or erm(F), regardless of whether ARG concentration was absolute or normalized to 16S rRNA levels. Both E. coli and enterococci were correlated with the tetracycline resistance genes, tet(O) and tet(W). Next-generation shotgun metagenomic sequencing was conducted to more thoroughly characterize the resistome (i.e., full complement of ARGs) and profile the occurrence of all ARGs described in current databases in storm runoff in order to inform future watershed monitoring and management. Between 37-121 different ARGs were detected in each stream sample, though the ARG profiles differed among storms. This study establishes that storm-driven transport of ARGs comprises a considerable fraction of overall downstream loadings and broadly characterizes the urban stormwater resistome to identify potential marker ARGs indicative of impact.

INTRODUCTION

The World Health Organization has deemed the emergence and spread of antibiotic resistance a crisis that “threatens the very core of modern medicine” (World Health Organization, 2015). Though concerns first centered on nosocomial patterns of resistance, the potential role of environmental pathways in facilitating the spread of antibiotic resistance among bacteria has gained considerable attention. Multiple studies have documented the contamination of surface waters with antibiotic resistance genes (ARGs) originating from wastewater treatment plants (Garcia-Armisen et al., 2011; Graham et al., 2011; Munir et al., 2011), agricultural runoff (Chee- Sanford et al., 2009; Fahrenfeld et al., 2014; Joy et al., 2013), and urban stormwater (McLellan et al., 2007; Zhang et al., 2016). Though numerous studies have documented increased loadings of pathogens and fecal indicator bacteria to surface water following rainfall (Hathaway and Hunt, 2010; Liao et al., 2015; McCarthy et al., 2012; Sidhu et al., 2012; Surbeck et al., 2006), potentially associated increases in loadings of antibiotic resistant bacteria and their associated ARGs have not been considered.

Given that soil bacteria represent a natural reservoir of ARGs, simple detection in environmental matrices is not necessarily of concern. However, point and non-point source pollution can serve as anthropogenic sources of ARGs to the environment (Pruden et al., 2012), thus the potential for dissemination of ARGs to waterborne and/or opportunistic environmental

39

pathogens via horizontal gene transfer calls for consideration. The ability of bacteria to acquire ARGs horizontally between live cells (conjugation), via bacteriophage infection (transduction), or via assimilation from the extracellular environment (natural transformation) necessitates consideration of the total abundance of ARGs in an environmental sample (i.e., the resistome) (von Wintersdorff et al., 2016). The potential risk of transfer of extracellular ARGs and ARGs carried by non-pathogenic bacteria to pathogens in aquatic environments is largely uncharacterized at this point. For example, recent work demonstrated that the plasmid-mediated colistin resistance gene MCR-1 can easily pass between strains of Escherichia coli, , and Pseudomonas aeruginosa (Liu et al., 2016), which are common intestinal/environmental species. Consequently, tracking resistance risks requires inclusion not only of known pathogens or microorganisms currently expressing resistance, but also resistance encoding genetic material that may be incorporated by pathogens.

Surface water has been identified as a reservoir of diverse and even novel ARGs (Amos et al., 2014a; Bengtsson-Palme et al., 2014; Garner et al., 2016; Kristiansson et al., 2011; Port et al., 2012). Stormwater in particular possesses many characteristics that may lead to the selection and amplification of these genes. Stormwater can be contaminated from an array of point and non- point sources, including land-applied manure, septic tanks, combined sewer overflows and leaky sewers (Kelsey et al., 2004; Parker et al., 2010; Sauer et al., 2011). It also frequently contains substantial quantities of heavy metals (Sansalone and Buchberger, 1997) and antibiotics (Davis et al., 2006; Joy et al., 2013; Xu et al., 2013), which are both well-known to select bacteria that possess ARGs, even at sub-inhibitory concentrations (Andersson and Hughes, 2014; Gullberg et al., 2014; Liu et al., 2011; McVicker et al., 2014). Sub-inhibitory concentrations are of great interest given that they are generally more environmentally-relevant, but also because “inhibitory” concentrations are only defined in limited strain, and/or media-specific contexts. At low levels, heavy metals and antibiotics have also been observed to stimulate horizontal gene transfer (Beaber et al., 2004; Klümper et al., 2016; Prudhomme et al., 2006; Song et al., 2009; Úbeda et al., 2005; Xia et al., 2008; Zhang et al., 2013), increasing the potential for resistant native aquatic bacteria to transfer ARGs to pathogenic bacteria introduced by stormwater.

While urban stormwater and associated runoff have been thoroughly documented as a source of pathogens (Cizek et al., 2008; Qureshi, 1979; Selvakumar and Borst, 2006; Sidhu et al., 2012), the role of storms in propagating ARGs has received little attention. Although patterns of incidence have been examined in several watersheds to date (Amos et al., 2014a; Chen et al., 2013a; Graham et al., 2011; Luo et al., 2010; Marti et al., 2013), the sources and mechanisms contributing to these observations, and their connections to various anthropogenic inputs are poorly understood. Thus, the identification of likely sources and transport processes of ARGs during storms represents an important knowledge gap.

The objectives of this study were to: (1) characterize the abundance of five ARGs (two sulfonamide: sul1 and sul2; two tetracycline: tet(O) and tet(W); and one macrolide: erm(F)) in an inland urban stream throughout the duration of three rainfall events and during baseline conditions; (2) identify physicochemical and hydrometeorological factors related to the occurrence or abundance of ARGs in stormwater runoff; and (3) investigate the breadth of the resistome detectable during storms relative to baseline levels using next-generation high-throughput DNA

40

sequencing. Understanding the incidence and movement of ARGs within urban streams will better inform watershed management strategies to mitigate downstream risks.

MATERIALS AND METHODS

Site and storm descriptions

Samples were collected from Stroubles Creek in Blacksburg, Virginia, USA at the Stream Research, Education, and Management Laboratory (StREAM Lab; http://www.bse.vt.edu/site/streamlab/). The Stroubles Creek watershed has been extensively described in previous studies (Liao et al., 2015, 2014; VADEQ, 2006); the 14.4 km2 drainage area above the study’s sampling point is 84% urban/residential land use (served by municipal sanitary sewers), 13% agricultural land use (primarily pasture and cropland), and 3% forested land. On-site instrumentation stations record a suite of physicochemical variables (temperature, specific conductivity, pH, turbidity, dissolved oxygen) via multiparameter water quality sondes (YSI Inc.) as well as streamflow (stage) via a gauge (Campbell Scientific, Inc.) (Liao et al., 2014).

Samples were collected during three summer storms occurring on June 27, July 2, and July 10 2013 and are herein referred to as storms 1, 2, and 3, respectively. Rainfall depths of 6, 17, and 12 mm were recorded for the three storms, respectively, and total event runoff volumes were calculated to be 8,100 m3, 37,000 m3, and 70,000 m3 (Liao et al., 2015). Additional hydrometeorological characteristics of the studied storms have been previously published (Liao et al., 2014, 2015).

Sample collection and DNA extraction

Water samples were collected automatically in sterile 750 mL bottles every 15 (storm 1) or 30 (storms 2 and 3) minutes using a 6712 ISCO sampler (Teledyne, Lincoln, NE) over each storm’s duration. Three baseline samples were also collected from the sample site during dry weather periods (e.g., no appreciable precipitation or change in stream stage for the previous 24 hours). Samples were transported on ice and stored at 4ºC prior to processing. Within 24 hours of collection, samples were thoroughly shaken to mix and 50 mL aliquots were filter-concentrated onto 0.4 µm pore size polycarbonate membrane filters (Millipore, Billerica, MA). Filters were transferred to 2 mL sterile tubes and stored at -80ºC. Filters were cut into approximately 1 cm2 fragments using a flame-sterilized blade and transferred to DNA extraction tubes. DNA was extracted from the filters using a PowerSoil DNA Isolation Kit (MO BIO Laboratories, Inc., Carlsbad, CA) according to manufacturer instructions.

Molecular analysis and high throughput sequencing

ARGs were quantified in triplicate reactions from DNA extracts using qPCR with previously published protocols for 16S rRNA genes (Suzuki et al., 2000) and five ARGs: sul1, sul2 (Pei et al., 2006), tet(O), tet(W) (Aminov et al., 2001), and erm(F) (Chen et al., 2007). A subset of samples was initially analyzed at dilutions of 1:10, 1:20, 1:50, and 1:100 to determine the minimum dilution required to minimize inhibition (results not shown); ultimately, a dilution factor of 1:10 was selected and applied to all extracts. Triplicate standard curves of ten-fold serial diluted standards

41

of each target gene ranging from 102 to 108 gene copies/µl for 16S rRNA and 101 to 107 gene copies/µl for ARGs were included on each 96-well plate, along with a triplicate negative control. The minimum acceptable qPCR standard curve R2 was 0.978. The limit of quantification was established as the lowest standard that amplified in triplicate in each run, and was equivalent to 104 gene copies per L of sampled bulk water for sul1, sul2, tet(O), tet(W), and 16S rRNA genes, and 105 gene copies per L of sampled bulk water for erm(F).

Shotgun metagenomics were conducted on the sample representing the maximum stream stage (i.e., peak flow) during each storm, as well a composite of the three baseline samples, combined by equal DNA mass. Samples were prepared using the Nextera XT library preparation (Illumina, San Diego, CA) and sequenced on an Illumina HiSeq 2500 using a 100-cycle paired- end protocol at the Biocomplexity Institute of Virginia Tech Genomics Research Lab. Paired end reads were merged using FLASH (Magoč and Salzberg, 2011). Quality filtering was conducted using Trimmomatic (Bolger et al., 2014) according to default parameters. 16S rRNA genes were annotated using BLASTN (Altschul et al., 1997) against the GreenGenes ribosomal RNA database (DeSantis et al., 2006). ARGs were annotated using the DIAMOND protein aligner (Buchfink et al., 2014) against the subset download of the Comprehensive Antibiotic Resistance Database (McArthur et al., 2013) that excludes genes that confer resistance via specific mutations (accessed August 2015). A minimum amino acid identity of 90% and a minimum e-value of 10-5 were required. Metagenomes were uploaded to the metagenomics RAST server (MG-RAST) (Meyer et al., 2008) and are publicly available under the accession numbers 4628882.3 – 4628885.3.

Data analysis

All statistical comparisons were conducted using R (v. 3.2.1) with a significance cutoff of α=0.05 unless otherwise noted. Normality of ARG datasets was assessed using a Shapiro-Wilk test. All ARG datasets failed to meet normality requirements, with the exception of sul1, thus non- parametric statistical analyses were applied for all ARG comparisons. ARG abundances during various storm phases were compared using a Kruskal-Wallis rank sum test, followed by a pairwise Wilcoxon rank sum test. Spearman’s rank correlation coefficients were calculated to assess correlations between ARG abundances, fecal indicator bacteria, water quality parameters, and hydrometeorological parameters. Storm ARG event loads (EL; gene copies / event) and equivalent background period loads (EBP; gene copies / EBP) were calculated as previously described for fecal indicator bacteria (Liao et al., 2015):

푁 퐸퐿 = ∑푖=1 푄푖 퐶푖∆푡 (1)

퐸퐿 퐸퐵푃 = ⁄퐷퐿 (2) where Qi = ith discrete discharge (L/s); Ci = ith discrete ARG concentration (gene copies/L); ∆t = sampling interval (s); N = number of discrete samples collected; and DL = mean dry-weather loads in the duration of the storm event (gene copies). Event ARG loads were compared to equivalent background period loads using a Wilcoxon rank sum test.

42

RESULTS AND DISCUSSION

Selection of ARG Targets for Characterizing Storm Loadings

Five ARGs representing three classes of resistance (two sulfonamide: sul1 and sul2; two tetracycline: tet(O) and tet(W); and one macrolide: erm(F)) were selected for quantification in all samples via qPCR to characterize how loads of ARGs change throughout the course of each storm (Figure 3-1). The five target ARGs were selected due to their documented prevalence in watersheds impacted by anthropogenic and agricultural runoff (Fahrenfeld et al., 2014; Pruden et al., 2006; Storteboom et al., 2010a). Macrolides and tetracyclines are among the most widely used antibiotics globally for both human and agricultural applications (Van Boeckel et al., 2014). Sulfonamides were the first synthetically produced antibiotic for widespread use beginning in the 1930s and extensive resistance has since emerged among clinical isolates (Sköld, 2000). sul1 and sul2 have been widely detected in municipal wastewater (Laht et al., 2014; Munir et al., 2011; Pruden et al., 2012) and sul1 is of particular interest as it has been identified as a strong indicator of anthropogenic influence in surface water (Pruden et al., 2012). tet(O) and tet(W) have been commonly found in wastewater, but are also common targets found in agricultural waste streams (Auerbach et al., 2007; Koike et al., 2007; McKinney et al., 2010). Additionally, all five of these genes were among those identified as candidate indicators of the extent of the impact of antibiotic resistance in the environment by the European Cooperation in Science and Technology Action (Berendonk et al., 2015).

Figure 3-1: ARG abundance with respect to Stroubles Creek discharge. Absolute abundances of ARGs and 16S rRNA genes (gene copies / L; left axis) and stream discharge (m3/s; right axis) for (A) storm 1, (B) storm 2, and (C) storm 3.

Gene loading rates and intra-storm variability

Absolute concentrations of sulfonamide ARGs responded most consistently during storm events, with sul1 elevated above baseline concentrations during storms 1 and 3 (p=0.0070; 0.0060), and sul2 was significantly elevated above baseline concentrations during all three storms (p=0.0141; 0.0122; 0.0077). In contrast, tetracycline and macrolide ARGs were not consistently elevated, with tet(W) only significantly greater than the baseline during storm 2 (p=0.1412) and erm(F) during storm 3 (p=0.0493). Absolute concentrations of tet(O) were not significantly different from baseline concentrations during any storms. Total 16S rRNA genes were significantly elevated across all storms (p=0.0070; 0.0110; 0.0060).

43

ARG concentrations were also compared among defined portions of each storm: rising limb, peak flow, falling limb, and established baseline (Figure 3-2). While absolute abundances are important for determining total loading of ARGs, changes in the abundance of total bacteria can obscure patterns of ARG enrichment with respect to the overall microbial community. Therefore, both absolute (gene copies/L) and relative (gene copies/16S rRNA gene copies) abundances are considered. During storm 1, the absolute concentration (gene copies per L) of sul1, sul2, tet(W), and 16S rRNA genes all trended higher during the peak and rising limb of the storm compared to the baseline, but were only significantly greater during the rising limb (p=0.0119; 0.0138; 0.0138; 0.0119). In terms of relative abundance, only tet(W) and erm(F) were greater during the rising limb (p=0.0138; 0.0138), while sul2 fell below the baseline (p=0.0138) (Figure 3-2). In contrast, during storm 2, only absolute concentrations of sul2 and 16S rRNA genes were significantly elevated relative to the baseline during the falling limb of the storm (p=0.0053; 0.0030). Notably, absolute concentrations of sul2 trended above the baseline by greater than 1-log during all phases of the storm. During storm 3, the absolute concentration of all five ARGs and 16S rRNA genes were elevated above the baseline during the falling limb (p≤0.0140), which is a remarkable contrast to storm 1, where concentrations were elevated during the rising limb. sul2 again trended above the baseline on average during each phase of the storm, by greater than 1.5- log. 16S rRNA genes were elevated above the baseline during each phase of the storm by greater than 1-log.

Figure 3-2: ARG relative abundances during storm phases. Relative abundances of ARGs (ARG copies / 16S rRNA gene copies) during baseline sample collection, storm rising limb, storm crest, and storm falling limb across all storms.

The unique profile of the five target ARGs that occurred during various phases of the storm suggest point and non-point sources of ARGs may vary both over the duration of each storm and among individual storm events. The non-uniform concentration of ARGs throughout each storm suggests that certain urban and agricultural sources influenced the sampling point at different times throughout the sampling scheme. This is similar to the observation that fecal indicator bacteria in general can exhibit large variations during intra-storm sampling, with greater than 0.5-log variation in E. coli and enterococci concentrations documented during a single storm (Liao et al., 2015;

44

Stumpf et al., 2010). The apparent ubiquity of sul1 in Stroubles Creek suggests that it is not introduced to the watershed exclusively during times of rainfall, which may be attributed to a number of unique characteristic the gene possesses. sul1 is widely associated with plasmids and transposons, making it prone to horizontal gene transfer and common in both pathogens and environmental bacteria (Sköld, 2000). Additionally, sul1 tends to reside adjacent to the class 1 integron and a variable number of additional antibiotic ARGs, enabling selection not only by sulfonamide antibiotics, but other antibiotics as well (Huovinen et al., 1995; Mazel, 2006).

If ARG sources leading to the greatest watershed loading of ARGs can be identified, specific watershed management strategies could be identified to limit the long-term propagation of ARGs in watersheds. To explore this possibility, pollutographs were constructed presenting the cumulative gene copy loading versus the cumulative runoff volume. Of particular interest was whether ARGs follow a “first flush” pattern, commonly defined as the transport of 80% of a pollutant loading within the first 30% of a storm’s discharge volume (Bertrand-Krajewski et al., 1998) (Figure 3-3). Across all three storms, none of the five ARGs exhibited a “first flush” pattern. Rather, all ARGs tend to fall below the 1:1 bi-sector of the pollutographs, indicating that the bulk of ARG loading tended to occur in the latter half of each storm discharge volume. This trend was particularly pronounced for sul2 during storm 1, tet(W) during all three storms, and erm(F) during storms 1 and 3. Previous work indicates that fecal indicator bacteria do not always follow a traditional first flush pattern (Krometis et al., 2007; Stumpf et al., 2010), however, the observed “lag” in ARG transport reported in this study is unique.

Figure 3-3: Cumulative ARG storm loading distributions. The cumulative fraction of ARGs for (A) storm 1, (B) storm 2, and (C) storm 3 with respect to the cumulative runoff volume at the collection point. The dashed 1:1 reference bi-sector indicates a constant absolute (genes per L) concentration of ARGs.

Event loading rates

Total event loads for ARGs and 16S rRNA genes were calculated for each storm and compared to the equivalent background period loading that would occur at baseline concentrations for the equivalent duration of a storm (Figure 3-4). Loading of 16S rRNA genes during storm events averaged almost 2-log greater than during the equivalent background period. Similarly, the storm event loads were significantly greater than the equivalent background period loading for all

45

genes, except tet(O) (Wilcoxon Rank Sum test; α=0.10; p=0.1000; 0.0765; 0.0765; 0.0765; and 0.1000 for sul1, sul2, tet(W), erm(F) and 16S genes, respectively). On average, the total load of each ARG across a storm event was greater than 1-log above the equivalent background load. This increased loading was most dramatic in the cases of sul2 and tet(W), which each increased at least 2-log above the equivalent background loading for each storm. Total bacterial DNA markers also increased during this time by greater than 1-log in all cases. This increased event loading is critical because, despite the relatively short storm duration (i.e., a few hours of precipitation) there is real potential for lasting surface water quality impacts. For example, once ARGs enter the watershed environment, they are subject to a number of complex fate and transport mechanisms by which they may persist or propagate throughout the aquatic environment via bulk water or by partitioning to sediments (Pruden et al., 2012). ARGs may be transferred to or taken up by native aquatic bacteria, augmenting reservoirs of resistance that have the potential to be subsequently transferred to pathogenic bacteria (Forsberg, et al., 2012; Wright, 2010). Residual antibiotics and metals can create selective pressure for ARGs. Studies have also suggested that the presence of pesticides and herbicides can select for bacteria possessing ARGs (Bordas et al., 1997; Kurenbach et al., 2015).

Figure 3-4: Average ARG storm event loading and corresponding equivalent background period loading. Error bars represent standard deviation of the gene load of the storms (n=3) or the equivalent background periods (n=3).

Although no significant correlations were identified between total ARG load and hydrometeorological characteristics of the storms (event rainfall depth, event duration, time to peak flow, and event runoff volume), a few patterns are worth noting. sul1 was present at the greatest absolute abundances during storm 1, the storm with the shortest duration (7 hours) and

46

least runoff volume (8,100 m3). sul1 was present at markedly consistent relative concentrations throughout all storm and baseline samples (Figure 3-2), suggesting that sul1 is present in Stroubles Creek during various meteorological conditions and subject to dilution under intense storm conditions. By contrast, tet(O) and erm(F) reached highest absolute concentrations during storm 3. Storm 3 was characterized by the greatest duration (23 hours), a relatively short time to peak flow (3 hours), and the highest event runoff volume (70,000 m3), suggesting that tet(O) and erm(F) are mobilized from contaminant sources under periods of high runoff volume.

Association with fecal indicator bacteria and environmental variables

Monitoring of fecal indicator bacteria, such as E. coli and enterococci, is widely used in regulatory monitoring as a proxy for probable fecal pathogen contamination and associated human health risk. Culturable E. coli and enterococci concentrations have previously been published for the storms of interest (Liao et al., 2015). Weak correlations existed between several of the monitored ARGs and fecal indicator bacteria (Table 3-1). E. coli concentrations correlated significantly with absolute (Spearman’s ρ=0.3627; p=0.0026) and relative concentrations of tet(O) (ρ=0.3411; p=0.0047). E. coli also correlated with absolute abundances of tet(W) (ρ=0.3301; p=0.0064). Enterococci correlated weakly, but significantly, with absolute abundances of sul2 (ρ=0.2436; p=0.0488), tet(W) (ρ=0.3208; p=0.0086), and erm(F) (ρ=0.3144; p=0.0101). Enterococci exhibited strong significant correlations with absolute tet(O) (ρ=0.5218; p<0.0001) concentrations and relative tet(O) (ρ=0.4753; p<0.0001). Enterococci also significantly correlated with concentrations of 16S rRNA genes (ρ=0.2505; p=0.0425). These results suggest that fecal indicator bacteria are not consistently an accurate proxy for ARGs resulting from stormwater runoff. E. coli, in particular, was not well suited as an indicator for sources of the monitored sulfonamide or macrolide ARGs. Enterococci appear to be a more accurate indicator of contamination by all of the monitored genes, with the exception of sul1. Interestingly, tet(O) and tet(W) were the only ARGs significantly correlated with E. coli and the genes with the strongest correlation to enterococci, suggesting that they are likely to be associated with fecal contamination. This is consistent with previous studies that have found tet(O) and tet(W) to be abundant in environments impacted by swine waste streams, dairy manure-treated agricultural soils, and beef, swine, and dairy waste lagoons (Fahrenfeld et al., 2014; Koike et al., 2007; McKinney et al., 2010). tet(O) and tet(W) are known to be carried by a relatively diverse range of bacterial hosts, having been previously identified in at least 20 and 25 genera, respectively, including both Gram-positive and -negative bacteria as well as both pathogens and environmental bacteria (Roberts, 2011). Both genes may be carried chromosomally or within conjugative plasmids, and have been associated with mobile elements, such as transposons (Chopra and Roberts, 2001; Roberts, 2012). Enterococci have been known to carry tet(O) while neither E. coli nor enterococci typically carry tet(W) (Chopra and Roberts, 2001), suggesting that while tet(W) may be associated with other fecal-associated bacteria, the correlations are likely not due to direct carriage by E. coli or enterococci.

47

Table 3-1: Spearman’s rank correlation coefficients between ARGs, fecal indicator bacteria, and physicochemical water quality parameters. Statistically significant correlations (α=0.05) indicated in bold and with an asterisk.

E. coli enterococci temperature turbidity dissolved oxygen conductivity pH sul1 0.123 0.167 0.279* 0.392* -0.188 -0.206 -0.285* sul2 0.165 0.244* 0.327* 0.467* -0.221 -0.165 -0.251* tet(O) 0.363* 0.522* 0.446* 0.753* -0.397* 0.316* -0.063 tet(W) 0.330* 0.321* 0.312* 0.542* -0.266* -0.138 -0.274* erm(F) 0.150 0.314* 0.324* 0.619* -0.508* 0.271* -0.339*

Stream water temperature also correlated significantly with absolute abundances of all ARGs, as well as total 16S rRNA genes (R=0.2790; 0.3272; 0.4460; 0.3116; 0.3242; 0.2795; p=0.0222; 0.0069; 0.0002; 0.0103; 0.0074; 0.0220 for sul1, sul2, tet(O), tet(W), erm(F), and 16S genes respectively). Elevated temperatures in urban stormwater are often associated with runoff from impervious surfaces (Jones et al., 2012) and fecal sources of contamination (Paule-Mercado et al., 2016). Turbidity was also significantly correlated with absolute concentrations of all ARGs and 16S rRNA genes (R=0.3915; 0.4671; 0.7525; 0.5417; 0.6187; 0.4904; p=0.0011; <0.0001; <0.0001; <0.0001; <0.0001; <0.0001 for sul1, sul2, tet(O), tet(W), erm(F), and 16S genes respectively). Though elevated turbidity may be associated with fecal contamination in freshwater streams, stream bed sediment disturbance as well as particulate matter in runoff from impervious surfaces can also contribute significantly to elevated turbidity. Dissolved oxygen was negatively correlated with absolute concentrations of tet(O), tet(W), and erm(F) (R=-0.3967; -0.2663; - 0.5084; p=0.0009; 0.0294; <0.0001). Deficient dissolved oxygen concentrations are widely associated with urban storm runoff (Keefer et al., 1980), suggesting that it is a source of input to Stroubles Creek for these genes.

Diversity and richness of the resistome

While sul1, sul2, tet(O), tet(W), and erm(F) are all well-documented as frequently detected in surface waters impacted by agricultural runoff and treated wastewater, application of next- generation sequencing technologies to samples collected from similar environments have revealed the presence of diverse ARGs beyond these key genes of interest. Therefore, we applied shotgun metagenomic sequencing to a subset of samples to gain insight into the broader resistome present during peak storm conditions as compared to baseline conditions in order to inform ARG selection for future surface water monitoring efforts. Shotgun metagenomic high-throughput DNA sequencing produced 13.6–18.4 million paired 100-bp reads per sample. Between 409–1157 reads per sample (0.003–0.009% of reads) were identified as probable ARG sequences via annotation against the Comprehensive Antibiotic Resistance Database (McArthur et al., 2013). Abundances of ARGs are presented normalized to abundance of 16S rRNA genes, as well as target gene length and 16S rRNA gene length as described previously (Li et al., 2015). Normalized abundance of total ARGs ranged from 0.17 to 0.30 ARGs per 16S rRNA gene. A total of 162 different ARG were annotated across the dataset, with 57, 37, 100, and 121 ARGs annotated in the baseline

48

sample and storms 1, 2, and 3, respectively. Across the dataset, trimethoprim was the most abundant class of antibiotic resistance (35.8% of total ARGs), followed by multidrug (33.8%), beta-lactam (6.8%), polymyxin (6.7%), aminoglycoside (5.6%), and glycopeptide resistance (3.1%) (Figure 3-5). As many as 155 ARGs have been detected in a single sample in previous studies, as well as ARGs capable of conferring resistance to all major classes of antibiotics (Amos et al., 2014b; Bengtsson-Palme et al., 2014; Chen et al., 2016, 2013b; Garner et al., 2016). The relative prevalence of multidrug resistance among detected ARGs is comparable to the findings of other metagenomic studies characterizing ARGs in surface water and associated sediments (Chen et al., 2013a; Garner et al., 2016; Li et al., 2015) and is likely due to the prevalence of multidrug efflux pumps among environmental bacteria (Martinez, 2009). Elevated trimethoprim resistance is less common among comparable metagenomic studies, but trimethoprim ARGs have been detected in environments heavily impacted by aquaculture and agriculture (Byrne-Bailey et al., 2009; Muziasari et al., 2014), making the source of abundant trimethoprim ARGs unclear in this urban watershed.

Notably, multidrug, beta-lactam, peptide, and tetracycline resistance (0.14, 0.030, 0.0083, and 0.0022 ARGs per 16S rRNA genes, respectively) were more prevalent during storm 3 compared to other storms and baseline concentrations. In the baseline sample, however, rifampin, aminocoumarin, fluoroquinolone, and glycopeptide resistance (0.0070, 0.00029, 0.0011, and 0.013 ARGs per 16S rRNA genes, respectively) were more abundant than levels observed during the storms.

Only 14 ARGs were consistently present during the baseline as well as all three storms: one trimethoprim resistance gene (dfrE), two polymyxin ARGs (PmrE, rosB), one nalidixic acid ARG (emrB), and ten genes that are components of multidrug efflux pumps or involved in the modulation of multidrug efflux (acrF, ceoB, mdtB, mdtC, mexB, mexC, mexD, phoP, smeR, smeB). Each storm contained a unique profile of ARGs, with 8, 25, and 38 ARGs annotated uniquely to storms 1, 2, and 3, respectively. There was not a conserved ARG profile across the storms; however, ten ARGs were detected in all storms but were absent in baseline samples: two aminoglycoside resistance genes (aadA, ANT(2”)-Ia), one beta-lactam (OXA-12), one peptide (bacA), one polymyxin (arnA), and five genes related to multidrug efflux pumps (baeS, mdtD, mdtF, mdtL, phoQ). These ten ARGs could be considered as targets for future storm monitoring efforts, though ARGs unique to storm events may vary among different watersheds.

49

Figure 3-5: Distribution of ARGs by class in baseline (composite n=3) and peak (n=1) runoff storm samples determined by shotgun metagenomic sequencing. Length of bars around the plot circumference indicate ARG copies normalized to 16S rRNA genes. Figure produced using the circlize package in R (v. 3.2.1).

While the association of ARGs with common fecal indicator bacteria and physicochemical parameters offers insight into the possible sources of ARG contamination in Stroubles Creek, the tendency of certain ARG patterns to be conserved based upon runoff source could provide the basis for ARG source tracking. Several studies have demonstrated the feasibility of profiling the antibiotic resistance of fecal streptococci to identify likely sources of fecal pollution in surface water and groundwater (Hagedorn, et al., 1999; Wiggins, 1996; Wiggins et al., 1999). Though widely used for over a decade, current source-tracking strategies generally focus on the detection of source-specific genetic markers (i.e. library-independent strategies), given the labor-intensive nature of antibiotic resistance profiling. Patterns of occurrence of tetracycline ARGs have been used to identify urban and agricultural sources of ARGs in surface water (Chen et al., 2013a; Storteboom et al., 2010a). Storteboom et al. (2010a) demonstrated that certain tetracycline ARGs were more frequently associated with agriculture runoff (tet(H), tet (Q), tet (S), and tet (T)), while others were more frequently associated with wastewater treatment plant (WWTP) effluent (tet(C), tet (E), tet (O)). Phylogenetic variations in tet(W) have been used to track sources of ARG contamination in groundwater and surface water (Koike et al., 2007; Storteboom et al., 2010b). Storteboom et al. (2010b) utilized restriction fragment length polymorphism analysis to

50

demonstrate that certain tet(W) phylotypes were associated with environments impacted by agricultural runoff, while different tet(W) phylotypes were indicative of WWTP influence. In future work, such library-independent microbial source tracking methods combined with storm profiling of ARGs could be used to identify waste streams that contribute to the highest watershed loadings of ARGs. Next-generation sequencing offers a powerful tool to be used for examining genetic variation in ARGs and can facilitate the identification of genetic phylotypes associated with specific ARG sources. Management of these ARG sources can help to limit watershed-scale ARG dissemination and potential downstream uptake by pathogenic bacteria.

CONCLUSIONS

Identification of strategies to limit inputs of clinically-relevant ARGs, along with other initiatives to improve storm water quality, can help alleviate the risk of antibiotic resistance spread. This study tracked the effects of storm events on the loadings of ARGs in an affected stream and provided insight into mechanisms involved in transport as well as the behavior of various indicators of antibiotic resistance. Specific conclusions include the following:  Storm-driven transport of ARGs contributed significant loadings to surface waters. Loadings of certain ARGs (sul2 and tet(W)) were more than two orders of magnitude greater during storm conditions than during equivalent background periods.  Key differences were noted in the behavior of different ARGs during storm runoff, yielding new insight into the processes governing the fate and transport of ARGs in watersheds. For example, the tetracycline resistance genes, tet(O) and tet(W) were correlated with the fecal indicator bacteria, E. coli and enterococci, but sul1, sul2, and erm(F) were not.  Further research is needed to understand the seasonal and geographic variation in behavior among ARGs in stormwater runoff as well as to identify key “indicator” ARGs or other genetic elements that are associated with risk of downstream transfer to pathogens and antibiotic resistance.

ACKNOWLEDGEMENTS

This work is supported by the National Science Foundation (NSF) Graduate Research Fellowship Program Grant (DGE 0822220) and RAPID grant (1402651), Virginia Water Resources Research Center Student Grant, Virginia Tech Institute for Critical Technology and Applied Science Center for Science and Engineering of the Exposome, the Virginia Tech College of Agriculture and Life Sciences Integrated Grants Program, the American Water Works Association Abel Wolman Doctoral Fellowship, and NSF Partnership for International Research and Education Award 1545756.

REFERENCES

Altschul, S., Madden, T., Schaffer, A., Zhang, J., Zhang, Z., Miller, W., Dj, L., 1997. Gapped BLAST and PSI- BLAST: a new generation of protein database search programs. Nucleic acids Res. 25(17), 3389–3402. Aminov, R.I., Garrigues-Jeanjean, N., Mackie, R.I., 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.

51

Amos, G.C.A., Zhang, L., Hawkey, P.M., Gaze, W.H., Wellington, E.M., 2014a. Functional metagenomic analysis reveals rivers are a reservoir for diverse antibiotic resistance genes. Vet. Microbiol. 171, 441–7. Amos, G.C.A., Hawkey, P.M., Gaze, W.H., Wellington, E.M., 2014b. Waste water effluent contributes to the dissemination of CTX-M-15 in the natural environment. J. Antimicrob. Chemother. 69, 1785–1791. Andersson, D.I., Hughes, D., 2014. Microbiological effects of sublethal levels of antibiotics. Nat. Rev. Microbiol. 12, 465–478. Auerbach, E.A., Seyfried, E.E., McMahon, K.D., 2007. Tetracycline resistance genes in activated sludge wastewater treatment plants. Water Research 41, 1143–1151. Beaber, J.W., Hochhut, B., Waldor, M.K., 2004. SOS response promotes horizontal dissemination of antibiotic resistance genes. Nature 427, 72–74. Bengtsson-Palme, J., Boulund, F., Fick, J., Kristiansson, E., Larsson, D.G.J., 2014. Shotgun metagenomics reveals a wide array of antibiotic resistance genes and mobile elements in a polluted lake in India. Front. Microbiol. 5, 1–14. Berendonk, T.U., Manaia, C.M., Merlin, C., Fatta-Kassinos, D., Cytryn, E., Walsh, F., Bürgmann, H., Sørum, H., Norström, M., Pons, M.-N., Kreuzinger, N., Huovinen, P., Stefani, S., Schwartz, T., Kisand, V., Baquero, F., Martinez, J.L., 2015. Tackling antibiotic resistance: the environmental framework. Nat. Rev. Microbiol. 13, 310–317. Bertrand-Krajewski, J.-L.L., Chebbo, G., Saget, A., 1998. Distribution of pollutant mass vs volume in stormwater discharges and the first flush phenomenon. Water Research 32, 2341– 2356. Bolger, A.M., Lohse, M., Usadel, B., 2014. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120. Bordas, A.C., Brady, M.S., Siewierski, M., Katz, S.E., 1997. In vitro enhancement of antibiotic resistance development - Interaction of residue levels of pesticides and antibiotics. J. Food Prot. 60, 531–536. Buchfink, B., Xie, C., Huson, D.H., 2014. Fast and sensitive protein alignment using DIAMOND. Nat. Methods 12, 59–60. Byrne-Bailey, K.G., Gaze, W.H., Kay, P., Boxall, A.B.A., Hawkey, P.M., Wellington, E.M.H., 2009. Prevalence of sulfonamide resistance genes in bacterial isolates from manured agricultural soils and pig slurry in the United Kingdom. Antimicrob. Agents Chemother. 53, 696–702. Chee-Sanford, J.C., Mackie, R.I., Koike, S., Krapac, I.G., Lin, Y.F., Yannarell, A.C., Maxwell, S., Aminov, R.I., 2009. Fate and Transport of Antibiotic Residues and Antibiotic Resistance Genes following Land Application of Manure Waste. J. Environ. Qual. 38, 1086–1108. Chen, B., Liang, X., Huang, X., Zhang, T., Li, X., 2013a. Differentiating anthropogenic impacts on ARGs in the Pearl River Estuary by using suitable gene indicators. Water Research 47, 2811–2820. Chen, B., Yang, Y., Liang, X., Yu, K., Zhang, T., Li, X., 2013b. Metagenomic Profiles of Antibiotic Resistance Genes (ARGs) between Human Impacted Estuary and Deep Ocean Sediments. Environ. Sci. Technol. 47, 12753–12760. Chen, B., Yuan, K., Chen, X., Yang, Y., Zhang, T., Wang, Y., Luan, T., Zou, S., Li, X., 2016. Metagenomic Analysis Revealing Antibiotic Resistance Genes (ARGs) and Their Genetic Compartments in the Tibetan Environment. Environ. Sci. Technol. 50, 6670-9. Chen, J., Yu, Z.T., Michel, F.C., Wittum, T., Morrison, M., 2007. Development and application

52

of real-time PCR assays for quantification of erm genes conferring resistance to macrolides- lincosamides-streptogramin B in livestock manure and manure management systems. Appl. Environ. Microbiol. 73, 4407–4416. Chopra, I., Roberts, M., 2001. Tetracycline Antibiotics : Mode of Action , Applications , Molecular Biology , and Epidemiology of Bacterial Resistance Tetracycline Antibiotics : Mode of Action , Applications , Molecular Biology , and Epidemiology of Bacterial Resistance. Microbiol. Mol. Biol. Rev. 65, 232–260. Cizek, A.R., Characklis, G.W., Krometis, L.A., Hayes, J.A., Simmons, O.D., Di Lonardo, S., Alderisio, K.A., Sobsey, M.D., 2008. Comparing the partitioning behavior of Giardia and Cryptosporidium with that of indicator organisms in stormwater runoff. Water Research 42, 4421–4438. Davis, J.G., Truman, C.C., Kim, S.C., Ascough, J.C., Carlson, K., 2006. Antibiotic Transport via Runoff and Soil Loss. J. Environ. Qual. 35, 2250–60. DeSantis, T.Z., Hugenholtz, P., Larsen, N., Rojas, M., Brodie, E.L., Keller, K., Huber, T., Dalevi, D., Hu, P., Andersen, G.L., 2006. Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARB. Appl. Environ. Microbiol. 72, 5069–5072. Fahrenfeld, N., Knowlton, K., Krometis, L.A., Hession, W.C., Xia, K., Lipscomb, E., Libuit, K., Green, B.L., 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, 2643–2650. 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. Garcia-Armisen, T., Vercammen, K., Passerat, J., Triest, D., Servais, P., Cornelis, P., 2011. Antimicrobial resistance of heterotrophic bacteria in sewage-contaminated rivers. Water Research 45, 788–796. Garner, E., Wallace, J.S., Argoty, G.A., Wilkinson, C., Fahrenfeld, N., Heath, L.S., Zhang, L., Arabi, M., Aga, D.S., Pruden, A., 2016. Metagenomic profiling of historic Colorado Front Range flood impact on distribution of riverine antibiotic resistance genes. Sci. Rep. 6, 38432. Graham, D.W., Olivares-Rieumont, S., Knapp, C.W., Lima, L., Werner, D., Bowen, E., 2011. Antibiotic Resistance Gene Abundances Associated with Waste Discharges to the Almendares River near Havana, Cuba. Environ. Sci. Technol. 45, 418–424. Gullberg, E., Albrecht, L.M., Karlsson, C., Sandegren, L., Andersson, D.I., 2014. Selection of a Multidrug Resistance Plasmid by Sublethal Levels of Antibiotics and Heavy Metals. MBio 5, 19–23. Hagedorn, C., Robinson, S.L., Filtz, J.L., Grubbs, S.M., Angier, T.A., Reneau, R.B., 1999. Determining Sources of Fecal Pollution in a Rural Virginia Watershed with Antibiotic Resistance Patterns in Fecal Streptococci. Appl. Environ. Microbiol. 65(12):5522-531. Hathaway, J.M., Hunt, W.F., 2010. Evaluation of First Flush for Indicator Bacteria and Total Suspended Solids in Urban Stormwater Runoff. Water, Air, Soil Pollut. 217, 135–147. Huovinen, P., Sundstrom, L., Swedberg, G., Skold, O., 1995. Trimethoprim and sulfonamide resistance. Antimicrob. Agents Chemother. 39, 279–289. Jones, M.P., Hunt, W.F., Winston, R.J., 2012. Effect of Urban Catchment Composition on Runoff Temperature. J. Environ. Eng. 138, 1231–1236. Joy, S.R., Bartelt-Hunt, S.L., Snow, D.D., Gilley, J.E., Woodbury, B.L., Parker, D.B., Marx, D.B., Li, X., 2013. Fate and Transport of Antimicrobials and Antimicrobial Resistance Genes in Soil and Runoff Following Land Application of Swine Manure Slurry. 47, 12081-8.

53

Keefer, T.N., Simons, R.K., McQuivey, R.S., 1980. Dissolved Oxygen Impact from Urban Storm Runoff, in: Urban Stormwater and Combined Sewer Overflow Impact on Recieving Water Bodies. pp. 223–244. Kelsey, H., Porter, D.E., Scott, G., Neet, M., White, D., 2004. Using geographic information systems and regression analysis to evaluate relationships between land use and fecal coliform bacterial pollution. J. Exp. Mar. Bio. Ecol. 298, 197–209. Klümper, U., Dechesne, A., Riber, L., Brandt, K.K., Gülay, A., Sørensen, S., Smets, B.F., 2016. Metal stressors consistently modulate bacterial conugal plasmid uptake potential in a phylogenetically conserved manner. ISME J. 1-14. Koike, S., Krapac, I.G., Oliver, H.D., Yannarell, A.C., Chee-Sanford, J.C., Aminov, R.I., Mackie, R.I., 2007. Monitoring and source tracking of tetracycline resistance genes in lagoons and groundwater adjacent to swine production facilities over a 3-year period. Appl. Environ. Microbiol. 73, 4813–4823. Kristiansson, E., Fick, J., Janzon, A., Grabic, R., Rutgersson, C., Weijdegard, B., Soderstrom, H., Larsson, D.G.J., 2011. Pyrosequencing of Antibiotic-Contaminated River Sediments Reveals High Levels of Resistance and Gene Transfer Elements. PLoS One 6, 7. Krometis, L.-A.H., Characklis, G.W., Simmons, O.D., Dilts, M.J., Likirdopulos, C.A, Sobsey, M.D., 2007. Intra-storm variability in microbial partitioning and microbial loading rates. Water Research 41, 506–16. Kurenbach, B., Marjoshi, D., Amábile-cuevas, C.F., Ferguson, G.C., Godsoe, W., Gibson, P., Heinemann, J.A, 2015. Sublethal Exposure to Commercial Formulations of the Herbicides Changes in Antibiotic Susceptibility in Escherichia coli and Salmonella enterica serovar Typhimurium 6, 1–9. Laht, M., Karkman, A., Voolaid, V., Ritz, C., Tenson, T., Virta, M., Kisand, V., 2014. Abundances of tetracycline, sulphonamide and beta-lactam antibiotic resistance genes in conventional wastewater treatment plants (WWTPs) with different waste load. PLoS One 9, e103705. Li, B., Yang, Y., Ma, L., Ju, F., Guo, F., Tiedje, J.M., Zhang, T., 2015. Metagenomic and network analysis reveal wide distribution and co-occurrence of environmental antibiotic resistance genes. ISME J. 9, 1–13. Liao, H., Krometis, L.A.H., Cully Hession, W., Benitez, R., Sawyer, R., Schaberg, E., von Wagoner, E., Badgley, B.D., 2015. Storm loads of culturable and molecular fecal indicators in an inland urban stream. Sci. Total Environ. 530–531, 347–356. Liao, H., Krometis, L.-A.H., Hession, W.C., House, L.L., Kline, K., Badgley, B.D., 2014. Hydrometeorological and Physicochemical Drivers of Fecal Indicator Bacteria in Urban Stream Bottom Sediments. J. Environ. Qual. 43, 2034. Liu, A., Fong, A., Becket, E., Yuan, J., Tamae, C., Medrano, L., Maiz, M., Wahba, C., Lee, C., Lee, K., Tran, K.P., Yang, H., Hoffman, R.M., Salih, A., Miller, J.H., 2011. Selective advantage of resistant strains at trace levels of antibiotics: a simple and ultrasensitive color test for detection of antibiotics and genotoxic agents. Antimicrob. Agents Chemother. 55, 1204–10. Liu, Y.Y., Wang, Y., Walsh, T.R., Yi, L.X., Zhang, R., Spencer, J., Doi, Y., Tian, G., Dong, B., Huang, X., Yu, L.F., Gu, D., Ren, H., Chen, X., Lv, L., He, D., Zhou, H., Liang, Z., Liu, J.H., Shen, J., 2016. Emergence of plasmid-mediated colistin resistance mechanism MCR-1 in animals and human beings in China: A microbiological and molecular biological study. Lancet Infect. Dis. 16, 161-8. Luo, Y., Mao, D., Rysz, M., 2010. Trends in Antibiotic Resistance Genes Occurrence in the Haihe

54

River, China. Environ. Sci. Technol. 44, 7220–7225. Magoč, T., Salzberg, S.L., 2011. FLASH: fast length adjustment of short reads to improve genome assemblies. Bioinformatics 27, 2957–2963. Marti, E., Jofre, J., Balcazar, J.L., 2013. Prevalence of Antibiotic Resistance Genes and Bacterial Community Composition in a River Influenced by a Wastewater Treatment Plant. PLoS One 8, e78906. Martinez, J.L., 2009. The role of natural environments in the evolution of resistance traits in pathogenic bacteria. Proc. Biol. Sci. 276, 2521–2530. Mazel, D., 2006. Integrons: agents of bacterial evolution. Nat. Rev. Microbiol. 4, 608–620. McArthur, A.G., Waglechner, N., Nizam, F., Yan, A., Azad, M. a., Baylay, A.J., Bhullar, K., Canova, M.J., De Pascale, G., Ejim, L., Kalan, L., King, A.M., Koteva, K., Morar, M., Mulvey, M.R., O’Brien, J.S., Pawlowski, A.C., Piddock, L.J. V, Spanogiannopoulos, P., Sutherland, A.D., Tang, I., Taylor, P.L., Thaker, M., Wang, W., Yan, M., Yu, T., Wright, G.D., 2013. The comprehensive antibiotic resistance database. Antimicrob. Agents Chemother. 57, 3348–3357. McCarthy, D.T., Hathaway, J.M., Hunt, W.F., Deletic, A., 2012. Intra-event variability of Escherichia coli and total suspended solids in urban stormwater runoff. Water Research 46, 6661–70. McKinney, C.W., Loftin, K.A., Meyer, M.T., Davis, J.G., Pruden, A., 2010. tet and sul Antibiotic Resistance Genes in Livestock Lagoons of Various Operation Type, Configuration, and Antibiotic Occurrence. Environ. Sci. Technol. 44, 6102–6109. McLellan, S.L., Hollis, E.J., Depas, M.M., Dyke, M. Van, Harris, J., Scopel, C.O., 2007. Distribution and Fate of Escherichia coli in Lake Michigan Following Contamination with Urban Stormwater and Combined Sewer Overflows. J. Great Lakes Res. 33, 566–580. McVicker, G., Prajsnar, T.K., Williams, A., Wagner, N.L., Boots, M., Renshaw, S.A., Foster, S.J., 2014. Clonal expansion during Staphylococcus aureus infection dynamics reveals the effect of antibiotic intervention. PLoS Pathog. 10, e1003959. Meyer, F., Paarmann, D., D’Souza, M., Olson, R., Glass, E.M., Kubal, M., Paczian, T., Rodriguez, A., Stevens, R., Wilke, A., Wilkening, J., Edwards, R.A., 2008. The metagenomics RAST server - a public resource for the automatic phylogenetic and functional analysis of metagenomes. BMC Bioinformatics 9. Munir, M., Wong, K., Xagoraraki, I., 2011. Release of antibiotic resistant bacteria and genes in the effluent and biosolids of five wastewater utilities in Michigan. Water Research 45, 681– 693. Muziasari, W.I., Managaki, S., Pärnänen, K., Karkman, A., Lyra, C., Tamminen, M., Suzuki, S., Virta, M., 2014. Sulphonamide and trimethoprim resistance genes persist in sediments at Baltic Sea aquaculture farms but are not detected in the surrounding environment. PLoS One 9, 1–7. Parker, J.K., McIntyre, D., Noble, R.T., 2010. Characterizing fecal contamination in stormwater runoff in coastal North Carolina, USA. Water Research 44, 4186–94. Paule-Mercado, M.A., Ventura, J.S., Memon, S.A., Jahng, D., Kang, J.H., Lee, C.H., 2016. Monitoring and predicting the fecal indicator bacteria concentrations from agricultural, mixed land use and urban stormwater runoff. Sci. Total Environ. 550, 1171–1181. Pei, R.T., Kim, S.C., Carlson, K.H., Pruden, A., 2006. Effect of River Landscape on the sediment concentrations of antibiotics and corresponding antibiotic resistance genes (ARG). Water Research 40, 2427–2435.

55

Port, J.A., Wallace, J.C., Griffith, W.C., Faustman, E.M., 2012. Metagenomic Profiling of Microbial Composition and Antibiotic Resistance Determinants in Puget Sound. PLoS One 7, 13. Pruden, A., Arabi, M., Storteboom, H.N., 2012. Correlation Between Upstream Human Activities and Riverine Antibiotic Resistance Genes. Environ. Sci. Technol. 46, 11541–11549. Pruden, A., Pei, R.T., Storteboom, H., Carlson, K.H., 2006. Antibiotic resistance genes as emerging contaminants: Studies in northern Colorado. Environ. Sci. Technol. 40, 7445–7450. Prudhomme, M., Attaiech, L., Sanchez, G., Martin, B., Claverys, J.P., 2006. Antibiotic stress induces genetic transformability in the human pathogen Streptococcus pneumoniae. Science 313, 89–92. Qureshi, A., 1979. Microbiological studies on the quality of urban stormwater runoff in Southern Ontario, Canada. Water Research 13, 977–985. Roberts, M.C., 2012. Acquired Tetracycline Resistance Genes, Antibiotic Discovery and Development, Vols 1 and 2. Springer, New York. Roberts, M.C., 2011. Environmental macrolide-lincosamide-streptogramin and tetracycline resistant bacteria. Front. Microbiol. 2, 8. Sansalone, J.J., Buchberger, S.G., 1997. Partitioning and First Flush of Metals in Urban Roadway Storm Water. J. Environ. Eng. 123, 134–143. Sauer, E.P., VandeWalle, J.L., Bootsma, M.J., McLellan, S.L., 2011. Detection of the human specific Bacteroides genetic marker provides evidence of widespread sewage contamination of stormwater in the urban environment. Water Research 45, 4081–4091. Selvakumar, A., Borst, M., 2006. Variation of microorganism concentration in urban stormwater runoff with land use and seasons. J. Water Health 4, 109–124. Sidhu, J.P.S., Hodgers, L., Ahmed, W., Chong, M.N., Toze, S., 2012. Prevalence of human pathogens and indicators in stormwater runoff in Brisbane, Australia. Water Research 46, 6652–6660. Sköld, O., 2000. Sulfonamide resistance: mechanisms and trends. Drug Resist. Updat. 3, 155–160. Song, B., Wang, G.R., Shoemaker, N.B., Salyers, A.A., 2009. An Unexpected Effect of Tetracycline Concentration: Growth Phase-Associated Excision of the Bacteroides Mobilizable Transposon NBU1. J. Bacteriol. 191, 1078–1082. Storteboom, H., Arabi, M., Davis, J.G., Crimi, B., Pruden, A., 2010a. Identification of Antibiotic- Resistance-Gene Molecular Signatures Suitable as Tracers of Pristine River, Urban, and Agricultural Sources. Environ. Sci. Technol. 44, 1947–1953. Storteboom, H., Arabi, M., Davis, J.G., Crimi, B., Pruden, A., 2010b. Tracking antibiotic resistance genes in the south platte river basin using molecular signatures of urban, agricultural, and pristine sources. Environ. Sci. Technol. 44, 7397–7404. Stumpf, C.H., Piehler, M.F., Thompson, S., Noble, R.T., 2010. Loading of fecal indicator bacteria in North Carolina tidal creek headwaters: Hydrographic patterns and terrestrial runoff relationships. Water Research 44, 4704–4715. Surbeck, C.Q., Jiang, S.C., Ahn, J.H., Grant, S.B., 2006. Flow Fingerprinting Fecal Pollution and Suspended Solids in Stormwater Runoff from an Urban Coastal Watershed. Environ. Sci. Technol. 40, 4435–4441. Suzuki, M.T., Taylor, L.T., DeLong, E.F., 2000. Quantitative analysis of small-subunit rRNA genes in mixed microbial populations via 5 ’-nuclease assays. Appl. Environ. Microbiol. 66, 4605–4614. Úbeda, C., Maiques, E., Knecht, E., Lasa, Í., Novick, R.P., Penadés, J.R., 2005. Antibiotic-induced

56

SOS response promotes horizontal dissemination of pathogenicity island-encoded virulence factors in staphylococci. Mol. Microbiol. 56, 836–844. VADEQ, VADCR, 2006. Upper Stroubles Creek Watershed TMDL Implementation Plan, Montgomery County, Virginia. http://www.deq.virginia.gov/Portals/0/DEQ/Water/ TMDL/ImplementationPlans/stroubip.pdf (accessed 9 May 2015). Van Boeckel, T.P., Gandra, S., Ashok, A., Caudron, Q., Grenfell, B.T., Levin, S.A., Laxminarayan, R., 2014. Global antibiotic consumption 2000 to 2010: An analysis of national pharmaceutical sales data. Lancet Infect. Dis. 14, 742–750. von Wintersdorff, C.J.H., Penders, J., van Niekerk, J.M., Mills, N.D., Majumder, S., van Alphen, L.B., Savelkoul, P.H.M., Wolffs, P.F.G., 2016. Dissemination of Antimicrobial Resistance in Microbial Ecosystems through Horizontal Gene Transfer. Front. Microbiol. 7, 1–10. Wiggins, B.A., 1996. Discriminant analysis of antibiotic resistance patterns in fecal streptococci, a method to differentiate human and animal sources of fecal pollution in natural waters. Appl. Environ. Microbiol. 62, 3997–4002. Wiggins, B.A., Andrews, R.W., Conway, R.A., Corr, C.L., Dobratz, J., Dougherty, D.P., Eppard, J.R., Knupp, S.R., Mettenburg, J.M., Rinehardt, J.M., Sonsino, J., Zimmerman, M.E., Dobratz, E.J., Limjoco, M.C., Torrijos, R.L., 1999. Use of antibiotic resistance analysis to identify nonpoint sources of fecal pollution. Appl Env. Microbiol 65, 3483–6. World Health Organization, 2015. Global action plan on antimicrobial resistance. Geneva: World Health Organization. Wright, G.D., 2010. Antibiotic resistance in the environment: a link to the clinic? Curr. Opin. Biotechnol., 23(5), 589-94. Xia, Z.J., Wang, J., Hu, W., Liu, H., Gao, X.Z., Wu, Z.H., Zhang, P.Y., Li, Y.Z., 2008. Improving conjugation efficacy of Sorangium cellulosum by the addition of dual selection antibiotics. J. Ind. Microbiol. Biotechnol. 35, 1157–1163. Xu, W., Yan, W., Li, X., Zou, Y., Chen, X., Huang, W., Miao, L., Zhang, R., Zhang, G., Zou, S., 2013. Antibiotics in riverine runoff of the Pearl River Delta and Pearl River Estuary, China: concentrations, mass loading and ecological risks. Environ. Pollut. 182, 402–7. Zhang, P.Y., Xu, P.P., Xia, Z.J., Wang, J., Xiong, J., Li, Y.Z., 2013. Combined treatment with the antibiotics kanamycin and streptomycin promotes the conjugation of Escherichia coli. Fems Microbiol. Lett. 348, 149–156. Zhang, S., Pang, S., Wang, P., Wang, C., Han, N., Liu, B., Han, B., Li, Y., Anim-Larbi, K., 2016. Antibiotic concentration and antibiotic-resistant bacteria in two shallow urban lakes after stormwater event. Environ. Sci. Pollut. Res. 23, 9984–9992.

57

CHAPTER 4 : METAGENOMIC PROFILING OF HISTORIC COLORADO FRONT RANGE FLOOD IMPACT ON DISTRIBUTION OF RIVERINE ANTIBIOTIC RESISTANCE GENES

Emily Garner, Joshua S. Wallace, Gustavo Arango Argoty, Caitlin Wilkinson, Nicole Fahrenfeld, Lenwood S. Heath, Liqing Zhang, Mazdak Arabi, Diana S. Aga, and Amy Pruden

ABSTRACT

Record-breaking floods in September 2013 caused massive damage to homes and infrastructure across the Colorado Front Range and heavily impacted the Cache La Poudre River watershed. Given the unique nature of this watershed as a test-bed for tracking environmental pathways of antibiotic resistance gene (ARG) dissemination, we sought to determine the impact of extreme flooding on ARG reservoirs in river water and sediment. We utilized high-throughput DNA sequencing to obtain metagenomic profiles of ARGs before and after flooding, and investigated 23 antibiotics and 14 metals as putative selective agents during post-flood recovery. With 277 ARG subtypes identified across samples, total bulk water ARGs decreased following the flood but recovered to near pre-flood abundances by ten months post-flood at both a pristine site and at a site historically heavily influenced by wastewater treatment plants and animal feeding operations. Network analysis of de novo assembled sequencing reads into 52,556 scaffolds identified ARGs likely located on mobile genetic elements, with up to 11 ARGs per plasmid-associated scaffold. Bulk water bacterial phylogeny correlated with ARG profiles while sediment phylogeny varied along the river’s anthropogenic gradient. This rare flood afforded the opportunity to gain deeper insight into factors influencing the spread of ARGs in watersheds.

INTRODUCTION

In September 2013, historic flooding impacted the Colorado Front Range, with some locations experiencing a rare 1-in-1,000 year rainfall event1. A record-setting flood peak of 3.3 m was recorded for the Cache La Poudre (Poudre) River in Fort Collins, resulting in a major transformation of the watershed landscape and massive transport of sediment throughout the basin2. Since 2002, the Poudre River has served as a field observatory for characterizing the impact of urban and agricultural activities on antibiotics and antibiotic resistance genes (ARGs)3–7. The distinct gradient of anthropogenic influence as the Poudre River flows from its pristine origin in the Rocky Mountains to areas heavily impacted by animal feedings operations (AFOs) and wastewater treatment plants (WWTPs) has previously served to demonstrate that human activities significantly alter ARG occurrence in river bed sediment and bulk water7. In particular, the upstream capacity of AFOs and WWTPs, weighted to account for the inverse distance of these facilities from riverine sampling sites, strongly correlated (R2=0.92) with sul1, marking it as a key indicator of human influence.

Antibiotic resistance presents a critical challenge to public health. While antibiotic resistance is a natural capability among many bacteria, with diverse ARGs profiled even in remote and ancient soils and caves8–12, widespread use of antibiotics both in livestock and humans is linked with increased frequency of resistant infections reported in clinical settings and increased ARG abundance in aquatic and terrestrial environments9,13–17. Through comparison of DNA sequence

58

similarity, instances have been identified in which pathogens likely obtained ARGs from environmental reservoirs18,19. Though gene transfer events of ARGs from environmental bacteria to human pathogens are thought to be rare20, the consequences can be devastating. For example, the recent emergence and spread of blaNDM-1, which is frequently found on a genetic element carrying several genes conferring resistance to multiple antibiotics, is thought to have originated via horizontal transfer from a plant pathogen to a human pathogen21.

Surface water is now well-documented as a receiving environment for anthropogenic sources of ARGs and also represents a critical linkage back to humans both as a recreational and drinking water resource7,22–25. The factors governing the dissemination of ARGs in watersheds are complex and not well understood. In particular, transport of resistant bacteria and ARGs from human sources, such as WWTPs and AFOs, selection of allochthonous and authochthonous resistant bacteria by antibiotics and other agents, and horizontal gene transfer have been cited as key mechanisms governing the proliferation of ARGs in watersheds22,26,27. Contamination with antibiotics is of particular interest as they could exert direct or co-selective pressures on ARGs of different classes and also stimulate horizontal gene transfer28–32. Likewise, various metals can also stimulate the latter two processes14,33–36, though few studies have elucidated the relationship between occurrence of ARGs and antibiotics or metals in surface water. Understanding the relative roles of antibiotics and metals in proliferating antibiotic resistance in the environment is important for developing effective management guidelines for antimicrobial use and management of urban and agricultural waste streams.

In the wake of unprecedented rainfall in the Poudre River basin, we sought to characterize the impact of flooding and subsequent recovery on the occurrence of ARGs and examine the influence of antibiotics and metals. We annotated shotgun metagenomic reads against existing ARG and heavy metal resistance gene (MRG) databases to profile the resistome of pristine and heavily impacted sites before and after the flood. Correlations of select ARGs, quantified by quantitative polymerase chain reaction (qPCR), with antibiotics and heavy metals were examined. Amplicon sequencing of 16S rRNA genes enabled comparison of the resistome with the microbial phylogenetic composition as an indicator of the relative importance of vertical gene transfer and physical transport of bacteria. To explore the role of horizontal gene transfer in shaping the resistome, metagenomic reads were annotated against a mobile genetic element database. Network analysis of de novo assembled metagenomic scaffolds revealed ARGs exhibiting physical genetic linkages to mobile genetic elements, MRGs, and other ARGs. Comprehensive profiling of ARGs and factors hypothesized to contribute to their selection, proliferation, and spread in the environment before and after an extreme flooding event provided unique insight into the mechanisms governing the dissemination of ARGs in the water environment. This knowledge will be particularly important in upcoming decades when the frequency and severity of storms is expected to increase as a consequence of climate change37.

MATERIALS AND METHODS

Sample Collection and Preservation

Bulk water, including the suspended sediment therein, and bed sediment samples were collected from five previously described river sites, representing a gradient of anthropogenic

59

influence, as well as from a WWTP that discharges into the river5. Briefly, site 1 is a pristine location near the river’s origin in the Rocky Mountains, site 2 is upstream of Fort Collins and receives light agricultural runoff, site 3 is within Fort Collins and receives agricultural and urban stormwater runoff, site 4 is downstream of two WWTPs, (combined average effluent: 42,000 m3/d), and site 5 is downstream of Fort Collins and Greeley and is heavily impacted by adjacent agricultural and urban land use and a 28,000 m3/d WWTP (Figure 4-1). Bulk water was collected from the center of the flow channel in sterile 1-L polypropylene containers for molecular analysis and in 1-L amber glass bottles pretreated as described by Tso et al.38 for antimicrobial analysis. Duplicate bulk water samples for metal analysis were collected in 50-mL metal-free polypropylene centrifuge tubes. Triplicate sediment samples (~30 g) were collected from the top 5 cm of bed sediment using a sterile spade for molecular analysis. Water quality information was collected using a Hydrolab MS5 multiparameter sonde (OTT Hydromet, Loveland, CO). Samples were transported to the lab on ice and preserved within 24 hours of collection.

Figure 4-1: Poudre River sampling sites. Contributing wastewater treatment plants (WWTPs) and animal feeding operations (AFO) are indicated with their respective capacities in million gallons per day (MGD) and animal counts. The figure was created by co-author Mazdak Arabi using the ArcGIS software by ESRI, Release 10.1 (ESRI, Redlands, CA) (http://www.esri.com/software/arcgis).

Bulk water samples for molecular analysis were concentrated onto 0.22 µm mixed cellulose esters membrane filters (Millipore, Billerica, MA). Filters were folded into quarters and cut into 1 cm2 pieces using a sterile blade and transferred to extraction tubes. Sediment was homogenized and 0.5 g was transferred to extraction tubes. DNA was extracted using a FastDNA SPIN Kit for Soil (MP Biomedicals, Solon, OH). A filter blank and DNA extraction blank were also extracted.

Samples for antimicrobial analysis were preserved according to Tso et al38. The method was 13 13 modified to use a surrogate solution containing d4-sulfamethoxazole, C6-sulfamethazine, C- erythromycin, and demeclocycline (500 ng/mL). Samples for metal analysis were acidified to 2% (v/v) with fuming nitric acid and filtered through a 0.45 μm polypropylene syringe filter. A 10-mL

60

aliquot was transferred to two 15-mL metal-free polypropylene centrifuge tubes. One aliquot was spiked with 50 uL of 1000 ng/mL spiking solution made from certified metal standards (BDH Aristar® PLUS 82026-108, 82026-100, respectively, VWR, Inc. Radnor, PA, USA). An equal volume of 2% nitric acid in water (v/v) was added to the remaining aliquot for quantification by single-point standard addition.

Quantification of ARGs

Gene markers were quantified in triplicate reactions from DNA extracts using qPCR, with previously published protocols for 16S rRNA genes39 and five ARGs: sul16, sul26, tet(O)40, tet(W)40, and ermF41. Extracts were diluted between 1:10-1:50 to minimize inhibition. Triplicate standard curves of ten-fold serial diluted standards of each target gene ranging from 108 to 102 gene copies/µl for 16S rRNA and 107 to 101 gene copies/µl for ARGs were included for each run, along with a triplicate negative control. The limit of quantification was established as the lowest standard that amplified in triplicate in each run, ranging from 0.7 to 3.3 log gene copies/ml for bulk water and 3.6 to 6.0 log gene copies/g for sediment, depending on the gene assay and the measured volume or mass of sample.

Quantification of Antibiotics and Metals

Antibiotics were quantified by liquid chromatography-tandem mass spectrometry (LC- MS/MS) as previously described for sulfonamides and tetracyclines38. A separate LC-MS/MS method was adapted from Wallace and Aga42 for macrolide antibiotics to enhance sensitivity. All analytes were normalized to the internal standard d10-carbamazepine (sulfonamides and macrolides) or minocycline (tetracyclines). Metals were quantified by inductively coupled plasma mass spectrometry (ICP-MS) on an X-Series 2 instrument (Thermo Scientific, Waltham, MA) using collision cell technology to reduce polyatomic interferences. Analytes were quantified using single-point standard addition and confirmed by external calibration curve (0.5 to 1000 ng/mL). The concentrations of Cr, Mn, Cu, As, Sr, Ag, and Cd were quantified using the most abundant isotope, normalized to the internal standard 115In. Barium, Ce, Gd, Pt, Pb, Th, and U were quantified analogously via a separate injection to maximize scan time for accurate quantification, normalized to internal standard, 159Tb.

16S rRNA Gene Amplicon Sequencing and Metagenomic Analysis

To explore the composition of the bacterial communities in bulk water and bed sediment, gene amplicon sequencing was conducted using barcoded primers (515f/806r) designed to target the V4 region of the 16S rRNA gene43,44. Triplicate PCR products were composited, and 240 ng of each composite was combined and purified using a QIAquick PCR Purification Kit (Qiagen, Valencia, CA). Sequencing was conducted at the Virginia Bioinformatics Institute (VBI) Genomics Research Laboratory (Blacksburg, VA) on an Illumina MiSeq using a 250-cycle paired- end protocol. Processing of reads was conducted using the QIIME pipeline45 and annotation against the Greengenes database46 (May 2013 release). After quality filtering, between 37,150 - 444,433 reads were obtained per sample and all samples were rarefied to 37,150 randomly selected reads.

61

Shotgun metagenomics were conducted on water samples collected at sites 1 and 5, 12 months pre-flood and 3 and 10 months post-flood. Samples were prepared using the Nextera XT library prep (Illumina, San Diego, CA) and sequenced on an Illumina HiSeq 2500 using a 100- cycle paired-end protocol at VBI. Paired ends reads were merged using FLASH47. Quality filtering was conducted using Trimmomatic48 according to default parameters. Relative abundances were calculated by normalizing gene counts to abundance of 16S rRNA genes, as well as target gene and 16S rRNA gene length as proposed by Li et al49. Absolute abundances were calculated by multiplying relative abundance of ARGs by total abundance of 16S rRNA genes, quantified by qPCR, as noted above. 16S rRNA genes were annotated using BLASTN50 against the Silva ribosomal RNA database51 (version 123). ARGs were annotated against the subset download of the Comprehensive Antibiotic Resistance Database52, which excludes genes that confer resistance via specific mutations (accessed August 2015). MRGs were annotated from the BacMet antibacterial biocide and metal resistance genes database53 (version 1.1), and proteins specific to the mobile genetic elements plasmids and prophages were annotated from the ACLAME database54 (version 0.4). Annotations made against the ACLAME database were manually screened to ensure known ARGs were not included. Functional gene annotation was performed using the DIAMOND protein aligner55 with a best hit approach using an amino acid identity cutoff of 90%, minimum alignment length of 25 amino acids, and 1e-5 e-value cutoff. Sequences were assembled prior to network analysis using the IDBA-UD de novo assembler56 and annotated using DIAMOND with a 1e-5 e-value cutoff. Unassembled sequences were uploaded to MG-RAST57 and are publicly available under accession numbers 462880.3-4628878.3 (Table S1).

Statistical Analyses

Spearman’s Rank Correlation Coefficients were calculated in JMP to assess correlations between ARGs and metals, antibiotics, and water quality parameters using a significance cutoff of α=0.05. UniFrac distances generated in QIIME were imported into PRIMER-E (version 6.1.13) for one-way analysis of similarities (ANOSIM). Metagenomic ARG relative abundances were imported into PRIMER-E and Bray-Curtis distances were used to generate multidimensional scaling plots. This distance matrix was compared with weighted UniFrac similarities for 16S rRNA gene amplicon sequencing using 2STAGE in PRIMER-E. Network analysis visualization was conducted using Gephi (version 0.8.2).

RESULTS AND DISCUSSION

Metagenomic analysis reveals shift in ARG profile following extreme flooding event

Annotation of shotgun metagenomic reads from bulk water samples against the Comprehensive Antibiotic Resistance Database52 indicated that total ARGs per mL bulk water decreased from pre- to post-flood and then increased to near pre-flood abundances by ten months post-flood at both sites 1 and 5 (Figure 4-2A). This decrease and subsequent increase suggests that the flood acted to “dilute” ARGs at both the pristine and impacted sites. Ten months of recovery, however, allowed sufficient time for ARG abundances to return to approximately pre-flood abundances.

62

A total of 277 ARG subtypes were identified across all samples, ranging from 77 to 155 subtypes per site (Figure 4-2B,C). On average, trimethoprim resistance was the most common resistance type (39%), followed by multidrug (30%), polymyxin (11%), aminocoumarin (4%), peptide (4%), and tetracycline resistance (3%). The most common mechanism of resistance was efflux (46%), followed by antibiotic target replacement (39%), cell wall charge alteration (8%), antibiotic inactivation (5%), and molecular bypass (2%; Figure S1).

Figure 4-2: Metagenomic characterizations of ARGs in Poudre River samples. (A) Absolute abundance of ARGs by class per mL bulk water identified from metagenomic sequencing reads annotated against the Comprehensive Antibiotic Resistance Database. ARGs conferring resistance to two or more of the classes macrolide, lincosamide, and streptogramin are denoted as MLS. Venn diagrams represent number of ARGs unique and shared amongst sample dates indicated in number of months relative to flood at (B) site 1 and (C) site 5. (D) Nonmetric multidimensional scaling (NMDS) plot generated from Bray-Curtis similarity matrix of metagenomic ARG composition by date at site 1 (historically pristine) and site 5 (historically impacted). Months indicated are time scale relative to the flooding event.

The profile of individual ARG subtypes varied at both sites, indicating shifts in response to flooding and recovery, as illustrated by nonmetric multidimensional scaling (NMDS) analysis generated from a Bray-Curtis (BC) similarity matrix (Figure 4-2D). Remarkably, NMDS analysis of ARGs indicated bulk water at both sites 1 and 5 shifted three months post-flood (site 1 BC = 58.3, site 5 BC = 58.9) but continued to shift to a unique profile by ten months post-flood (site 1 BC = 60.9, site 5 BC = 61.7). Interestingly, the shift observed at site 5 three months post-flood indicated similarity with site 1 pre-flood, though not statistically significantly (BC = 65.56), suggesting that the flood acted to “dilute” ARGs from the impacted site such that it resembled the pristine site, as is consistent with the decrease in total abundance of ARGs in post-flood samples (Figure 4-2A). Surprisingly, while ARGs returned to pre-flood abundances, the ARG profile did not return to a pre-flood state, suggesting that the flooding may have disseminated new sources of ARGs that persisted at each site. While seasonal variation may also have contributed to the observed fluctuations, the overall stability of the bacterial community at each site across sample

63

dates, relative to notable community variation across sites (Figure 4-3), suggests that seasonal impact on biological variation was minimal.

Figure 4-3: Beta Diversity plots of microbial community phylogenetic composition. Based on 16S rRNA gene amplicon sequencing of Poudre River bulk water (n=1) and bed sediment (n=3) samples coded by sample site and collection date based on distance matrixes generated using a jackknifed unweighted UniFrac metric.

Potential for selection pressure indicated by co-occurrence of ARGs and antibiotics

The potential role of antibiotics as selective agents influencing the re-establishment of ARGs during post-flood recovery was investigated by examining correlations between sulfonamide (sul1, sul2), tetracycline (tet(O), tet(W)), and macrolide (ermF) ARGs in bed sediment and bulk water, quantified using qPCR (Figure S3), and 23 antibiotics (Table S2) in bulk water at all sites (Figure 4-4). Due to the tendency of some antibiotics to lose antibacterial activity if they become sorbed to sediments or form complexes with substances such as humic acids58-60, analysis of antibiotics was limited to the bulk water. Correlations between antibiotics and ARGs in bulk water are hypothesized to be indicative of potential selective pressure while correlations between antibiotics in the bulk water and ARGs in sediment are likely to be indicative of deposition of bacteria that may have been subject to selection in the bulk water.

All ARGs demonstrated significant Spearman’s rank correlations with at least one antibiotic against which they conferred resistance, suggesting direct selection may be a key pressure shaping the resistome (Figure 4-4). Bed sediment sul1 exhibited moderate correlations with sulfamethoxazole and sulfadiazine (Spearman ρ=0.4972, 0.4575; p=0.0028, 0.0065), while bulk water sul2 was moderately correlated with sulfamethoxazole (ρ=0.401.; p=0.023). Bulk water tet(O) exhibited a moderate correlation with anhydrotetracycline and a strong correlation with

64

tetracycline (ρ=0.4133, 0.5764; p=0.0187, 0.0006), while bulk water tet(W) was moderately correlated with tetracycline (ρ=0.4657; p=0.0072). ermF in bed sediment correlated weakly with azithromycin and moderately with tylosin (ρ=0.3528, 0.4764; p=0.0407, 0.0044) and in bulk water weakly with clarithromycin and moderately with erythromycin (ρ=0.3999, 0.4172; p=0.0234, 0.0175).

All ARGs identified were also found to significantly correlate with certain antibiotics against which they do not confer resistance (Figure 4-4; p-values in Table S3), indicating potential for co-selection, which results from co-location of ARGs on the same genetic element, such as a plasmid, transposon, or integron; cross-resistance, which occurs when a single cellular response is capable of combatting multiple chemicals, such is the case with multidrug resistance pumps; or co-regulation, which occurs when two resistance regulation systems are transcriptionally linked34. Notably, numerous correlations were observed between antibiotics and sul1 and tet(O) ARGs. Bed sediment sul1 exhibited a strong correlation with azithromycin, moderate correlation with clarithromycin, and weak correlation with erythromycin. Bulk water tet(O) exhibited a strong correlation with clarithromycin and erythromycin, moderate correlation with sulfamethoxazole, and sulfamethazine, and a weak correlation with azithromycin, while bed sediment tet(O) exhibited a strong correlation with azithromycin and erythromycin, and a moderate correlation with clarithromycin, sulfamethoxazole, and tylosin.

It is challenging to determine whether observed correlations are truly indicative of selective pressure or simply co-transport of antibiotics and ARGs from the same source. Covariation among antibiotics may also obscure true causative relationships of selection between genes and antimicrobial agents. Based on metagenomic data, positive correlations were observed between MLS, rifampin, and fosfomycin ARGs and the antibiotics sulfamethazine (ρ=0.8452, 0.8452, 0.8262, p=0.0341, 0.0341, 0.0427) and clarithromycin (ρ=0.8452, 0.8452, 0.8262, p=0.0341, 0.0341, 0.0427).

While the concentrations of antibiotics observed in the Poudre River samples appear to be below minimum inhibitory concentrations, previous work has indicated that sublethal concentrations may aid in the dissemination of ARGs, via selection and other mechanisms. Gullberg et al.61 found that bacteria carrying plasmids with beta-lactam resistance genes were selected at concentrations of antibiotics and heavy metals nearly 140 times below reported minimum inhibitory concentration. Other studies have indicated that sublethal antibiotics may promote the dissemination of ARGs by stimulating horizontal gene transfer29,30.

65

Figure 4-4: Spearman’s Rank Correlation Coefficient between abundance of ARGs and antibiotics or metals. Correlations in bed sediment (sed) and bulk water (wat) normalized to 16S rRNA genes, as determined by qPCR, and antibiotics or heavy metals. Statistically significant (p<0.05) correlations indicated in bold. Blue shading indicates negative correlation and red shading indicates positive correlation. Antibiotics detected and included in the analysis were: anhydrotetracycline (ATC), azithromycin (AZI), clarithromycin (CLA), chlorotetracycline (CTC), doxycycline (DOX), erythromycin (ERY), 4-epitetracycline (ETC), oxytetracycline (OTC), sulfadimethoxine (SDM), sulfamethoxazole (SMX), sulfamethazine (SMZ), sulfadiazine (SPD), tetracycline (TC), and tylosin (TYL).

66

Potential for co-selective pressures exerted by heavy metals

All three mechanisms of co-selection described above also pertain to heavy metals. A previous study demonstrated that input of tetracycline resistant bacteria to the Poudre River and selection by tetracycline antibiotics was insufficient to explain the level of resistant bacteria present in the river, and identified co-selection by heavy metals as a likely source of resistant bacteria62. Therefore, the possibility of co-selection by heavy metals was investigated by examining correlations between 14 heavy metals (Table S4) and ARGs quantified by qPCR. sul1 exhibited strong Spearman’s rank correlations with several heavy metals. In bulk water and bed sediment, sul1 was positively correlated with silver (Spearman ρ=0.6435, 0.4671; p=0.0004, 0.0140) and negatively with both barium (ρ=-0.6315, -0.4804; p=0.0005, 0.0112) and copper (ρ=- 0.4806, -0.4229; p=0.0.130, 0.0280). Strontium was positively correlated with bulk water and sediment sul2 (ρ=0.5890, 0.4347; p=0.0015, 0.0234) and bulk water tet(O), tet(W), and ermF (ρ=0.4756, 0.4870, 0.4812; p=0.0141, 0.0116, 0.0128). Uranium also exhibited positive correlations with sul2, tet(O), tet(W), and ermF in bulk water (ρ=0.5123, 0.5092, 0.4695, 0.4825; p=0.0075, 0.0079, 0.0155, 0.0125). Such robust correlations indicate a potential for co-selection by heavy metals, namely, by silver for sul1 and by strontium and uranium for sul2, tet(O), tet(W), and ermF. The unique behavior of sul1 compared to the other genes may result from the tendency for sul1 to be located on mobile genetic elements, such as class 1 integrons63–65. This characteristic may enable sul1 to become associated with various other ARGs and MRGs, making sul1 a prime candidate for co-selection. Copper has been previously identified as a metal that is likely to select for certain ARGs14,66, therefore its strong negative correlation with sul1 was unexpected and may be indicative that copper selects for ARGs through mechanisms highly specific to certain conditions. Such negative correlations are not unprecedented, however, as a significant negative correlation was also observed previously between copper and sulfonamide ARGs in livestock lagoon water67. Though no significant positive correlations existed between copper and the five ARGs quantified by qPCR, copper was correlated with total resistance genes derived from the metagenomic data set for peptide (ρ=0.8, p=0.2), tetracycline (ρ=0.8, p=0.2), and sulfonamide (ρ=0.6, p=0.4) classes, though trends were not significant.

Metagenomic scaffold associations reveals probable ARGs susceptible to co-resistance

Network analysis was conducted to explore de novo assembled scaffolds for ARGs and MRGs physically co-located on DNA strands in order to identify genes that are likely candidates for co-resistance as a mechanism of co-selection (Figure 4-5). Of a total of 52,556 scaffolds generated from all samples, 2,707 (5.2%) scaffolds contained more than one ARG and 347 (0.7%) scaffolds contained both ARGs and MRGs. Assembled scaffolds averaged 794 base pairs (bp) and reached a maximum length of 215,852 bp, ranging from 66,797 to 131,397 scaffolds per sample (Table S1). The most abundant ARG class associated with other ARGs revealed by the network analysis corresponded to efflux pumps (26.2%), multidrug resistance (12.3%), and macrolide/lincosamide/streptogramin (10.8%) resistance, while the most abundant co-located MRGs corresponded to copper and arsenic. The most frequent associations observed between genes were macB and bcrA (0.16% of scaffolds), otrC and bcrA (0.06%), PmrA and PmrB (0.05%), and sav1866 and bcrA (0.04%) (Figure 4-5). The ARGs that were subject to qPCR analysis, sul1, sul2, tet(O), tet(W), and ermF, were not found on any scaffolds with other ARGs

67

or MRGs, which may stem from their relatively low abundance in the pool of metagenomic reads, which reduces likelihood of assembly (20% total reads were assembled).

Figure 4-5: Co-occurrence of ARGs, MRGs, and genetic markers linked to mobile genetic elements on assembled scaffolds. Figure provides insight into which ARGs are candidates for co-selection or horizontal gene transfer. Network analysis indicating genetic proximity of ARGs, MRGs, plasmid sequences, and prophage sequences, based on co-occurrence of genes on scaffolds constructed using de novo assembly of shotgun metagenomic sequencing reads. Proximity of nodes and width of lines indicate frequency of associations between genes. Genetic markers with 20 or fewer instances of co-occurrence with other genes of interest were excluded from the network analysis rendering.

Role of horizontal gene transfer in shaping the resistome

Metagenomic data were searched for two families of mobile genetic elements, plasmids and prophages, as a proxy for potential for conjugation and transduction, respectively. Genes belonging to 32 different known plasmids were identified, along with genes corresponding to 65 different prophage genomes. A total of 3,912 (7.4%) scaffolds contained both ARGs and plasmid gene markers. Multiple ARGs were frequently found associated with plasmid markers on a single scaffold, with up to 11 ARGs found together on a single plasmid-associated scaffold. Four hundred and ninety-seven (0.9%) scaffolds contained one or more ARGs and prophage genetic markers. Network analysis revealed that the ARGs most frequently found on plasmid scaffolds were macB (16.4% of plasmid scaffolds), sav1866 (5.2%), mdtC (3.1%), otrC (2.9%), novA (2.5%), arnA (1.9%), and mexS (1.8%), while genes associated with copper (7.4%) and arsenic (2.3%) were the most common MRGs. macB (20.5% of prophage scaffolds), dfrE (12.7%), and arnA (5.2%), were the ARGs most frequently found on prophage-associated scaffolds. The five ARGs examined by qPCR were not identified on any scaffolds associated with plasmids or prophages. 68

Although horizontal gene transfer is known to be an important mechanism in the spread of antibiotic resistance and provides an opportunity for pathogenic bacteria to acquire resistance from environmental bacteria19, it has been reported that it is a relatively rare event among soil bacteria and may be a relatively minor influence in shaping the resistome compared to phylogeny20,68. However, it has also been noted that plasmids carrying ARGs are significantly more likely to be conjugative than non-ARG carrying plasmids69 and broad host range plasmids were found to be capable of uptake by a highly diverse portion of the microbiome in a soil bacterial community study70. Although we could not precisely quantify the extent to which horizontal gene transfer shaped the resistome based on the present study, the numerous associations of plasmids and prophages with ARGs were striking, suggesting that it is a significant phenomenon in the riverine environment.

Role of phylogeny in shaping the resistome

Based on jackknifed unweighted UniFrac distance, the microbial community composition observed in the bulk water of each site was distinct from that of the bed sediment (ANOSIM, R=0.868, p=0.001). Beta diversity plots, in which distance between samples is inversely proportional to similarity in phylogenetic composition, revealed that a clear shift in microbial community structure occurs along the anthropogenic gradient of the Poudre River. Sites clustered distinctly from each other, in bulk water (Figure 4-3, ANOSIM, R=0.488, p=0.001) and more strongly in sediment (ANOSIM, R=0.607, p=0.001), but did not exhibit a discernible pattern when plotted based on sampling date for water (ANOSIM, R=0.159, p=0.033) or sediment (ANOSIM, R=0.166, p=0.001). The strong grouping by sample site indicates that anthropogenic influence on phylogeny is likely a more dominant controlling variable than seasonal variation, as well as for ARGs, as documented in previous studies of the Poudre River7, or even variation observed as a result of the flooding. This strong trend of microbial community variation along the anthropogenic gradient of the Poudre River suggests that adjacent land use is a key driver of sediment and bulk water microbial community. Another study also highlighted that watershed land use also plays a role in shaping the sediment microbial community of the Tongue River in Montana, USA71. The resilience of the microbial community in quickly rebounding to pre-flood conditions is consistent with another study that observed that following a whole-ecosystem mixing disturbance of a freshwater lake, the microbial community returned to pre-mixing composition and diversity in only 11 days72. Site 1 community composition was highly distinct from site 5 (ANOSIM, R=0.929, p=0.001) and WWTP effluent was dissimilar to all river sites (ANOSIM, R≥0.947, p=0.001). Site based similarity was less pronounced using weighted UniFrac distance (ANOSIM, R=0.39, p=0.001), which takes into account not only number of unique operational taxonomic units (OTUs) present, as with unweighted UniFrac, but abundance of each OTU. This weaker correlation indicates that rare species were particularly important in defining observed distinctions in microbial community among sites. , Bacteroidetes, and Cyanobacteria were the most abundant phyla in the bulk water, with Actinobacteria, and Verrucomicrobia also contributing to more than 1% of phyla, on average (Figure S4). Similarly, Proteobacteria, Bacteroidetes, Cyanobacteria, and Verrucomicrobia were the most dominant phyla in the sediment, with Acidobacteria, Actinobacteria, Plantomycetes, Chloroflexi, Firmicutes, Nitrospirae, and Gemmatimonadetes all contributing to greater than 1% of phyla (Figure S5). Interestingly, the overall bulk water phylogeny was not correlated with ARG profiles (2STAGE, weighted UniFrac: Spearman’s ρ=-0.1) indicating that phylogeny alone may not be the most

69

important factor controlling the profile of ARGs. This finding conflicts with previous studies that highlight host phylogeny as a key factor influencing antibiotic resistance in soil, sewage sludge, or agricultural environments20,73,74.

CONCLUSIONS

This study uniquely characterized the impact of an extreme rainfall and flooding event on a riverine resistome using next-generation DNA sequencing. Following the flood, total bulk water ARGs decreased following the flood but recovered to near pre-flood abundances by ten months post- flood at both the pristine and impacted sites. Bulk water phylogeny did not correlate with ARG profiles, but sediment phylogeny varied according to the river’s anthropogenic gradient. Quantitative monitoring of ARGs and two classes of selective agents, antibiotics and heavy metals, was suggestive of selective pressure in the reestablishment of the resistome following the flood. Additionally, we identified ARGs found on assembled metagenomic scaffolds associated with other ARGs, MRGs, and mobile genetic element genes as likely candidates for co-selection or horizontal gene transfer. The results of this study help elucidate the mechanisms contributing to proliferation of ARGs in surface water and inform management strategies limiting anthropogenic contributions of ARGs to the environment.

ACKNOWLEDGEMENTS

This work was supported by the National Science Foundation under RAPID Grant No. 1402651 and Graduate Research Fellowship Program Grant DGE0822220. Additional support was provided by a Virginia Water Resources Research Center Student Grant and the Alfred P. Sloan Foundation Microbiology of the Built Environment program. We acknowledge the NSF Major Research Instrumentation Program CHE0959565 for the ICP-MS instrument. We would like to thank Tyler Dell and Douglas Gossett for assistance collecting samples.

REFERENCES

1. National Weather Service, National Oceanic and Atmospheric Administration. Exceedance Probability Analysis for the Colorado Flood Event. (2013). 2. CIRES Western Water Assessment at the University of Colorado, NOAA ESRL Physical Science Division, and the CSU Colorado Climate Center. Severe Flooding on the Colorado Front Range. (2013). http://wwa.colorado.edu/resources/front-range- floods/assessment.pdf. 3. Yang, S. & Carlson, K. Evolution of antibiotic occurrence in a river through pristine, urban and agricultural landscapes. Water Res 37: 4645–4656. (2003). 4. Storteboom, H., Arabi, M., Davis, J.G., Crimi, B. & Pruden, A. Tracking antibiotic resistance genes in the South Platte River basin using molecular signatures of urban, agricultural, and pristine sources. Environ Sci Technol 44: 7397–7404. (2010). 5. Pruden, A., Pei, R.T., Storteboom, H. & Carlson, K.H. Antibiotic resistance genes as emerging contaminants: Studies in northern Colorado. Environ Sci Technol 40: 7445– 7450. (2006). 6. Pei, R.T., Kim, S.C., Carlson, K.H. & Pruden, A. Effect of River Landscape on the sediment concentrations of antibiotics and corresponding antibiotic resistance genes

70

(ARG). Water Res 40: 2427–2435. (2006). 7. Pruden, A., Arabi, M. & Storteboom, H.N. Correlation Between Upstream Human Activities and Riverine Antibiotic Resistance Genes. Environ Sci Technol 46: 11541– 11549. (2012). 8. Perron, G.G., et al. Functional Characterization of Bacteria Isolated from Ancient Arctic Soil Exposes Diverse Resistance Mechanisms to Modern Antibiotics. PLoS One 10: e0069533. (2015). 9. D’Costa, V.M., et al. Antibiotic resistance is ancient. Nature 477: 457–461. (2011). 10. Mindlin, S.Z., Soina, V.S., Petrova, M.A. & Gorlenko, Z.M. Isolation of antibiotic resistance bacterial strains from Eastern Siberia permafrost sediments. Russ J Genet 44: 27–34. (2008). 11. Bhullar, K., et al. Antibiotic resistance is prevalent in an isolated cave microbiome. PLoS One 7: e34953. (2012). 12. Allen, H.K., Moe, L.A., Rodbumrer, J., Gaarder, A. & Handelsman, J. Functional metagenomics reveals diverse beta-lactamases in a remote Alaskan soil. ISME J 3: 243– 251. (2009). 13. Zhu, Y.G., et al. Diverse and abundant antibiotic resistance genes in Chinese swine farms. Proc Natl Acad Sci USA 110: 3435–3440. (2013). 14. Knapp, C.W., et al. Antibiotic Resistance Gene Abundances Correlate with Metal and Geochemical Conditions in Archived Scottish Soils. PLoS One 6: e27300. (2011). 15. Udikovic-kolic, N., Wichmann, F., Broderick, N.A. & Handelsman, J. Bloom of resident antibiotic-resistant bacteria in soil following manure fertilization. Proc Natl Acad Sci USA 111: 15202–15207. (2014). 16. Allen, H.K., et al. Call of the wild: antibiotic resistance genes in natural environments. Nat Rev Microbiol 8: 251–259. (2010). 17. McArthur, J.V., Fletcher, D.E., Tuckfield, R.C. & Baker-Austin, C. Patterns of Multi- Antibiotic-Resistant Escherichia Coli from Streams with No History of Antimicrobial Inputs. Microb Ecol. e-pub ahead of print, doi: 10.1007/s00248-015-0678-4. (2015). 18. Benveniste, R. & Davies, J. Aminoglycoside antibiotic-inactivating enzymes in actinomycetes similar to those present in clinical isolates of antibiotic-resistant bacteria. Proc Natl Acad Sci USA 70: 2276–2280. (1973). 19. Forsberg, K.J., et al. The Shared Antibiotic Resistome of Soil Bacteria and Human Pathogens. Science 337: 1107–1111. (2012). 20. Forsberg, K.J., et al. Bacterial phylogeny structures soil resistomes across habitats. Nature 509: 612–616. (2014). 21. Sekizuka, T., et al. Complete sequencing of the blaNDM-1-positive IncA/C plasmid from Escherichia coli ST38 isolate suggests a possible origin from plant pathogens. PLoS One 6: e25334. (2011). 22. Graham, D.W., et al. Antibiotic Resistance Gene Abundances Associated with Waste Discharges to the Almendares River near Havana, Cuba. Environ Sci Technol 45: 418– 424. (2011). 23. Amos, G.C., Hawkey, P.M., Gaze, W.H. & Wellington, E.M. Waste water effluent contributes to the dissemination of CTX-M-15 in the natural environment. J Antimicrob Chemother 69: 1785–1791. (2014). 24. Czekalski, N., Gascón Díez, E. & Bürgmann, H. Wastewater as a point source of antibiotic-resistance genes in the sediment of a freshwater lake. ISME J 8:1381–1390.

71

(2014). 25. Amos, G.C., et al. Validated predictive modelling of the environmental resistome. ISME J 9: 1467–1476. (2015). 26. Lupo, A., Coyne, S. & Berendonk, T.U. Origin and evolution of antibiotic resistance: the common mechanisms of emergence and spread in water bodies. Front Microbiol 3: 1-13. (2012). 27. Kristiansson, E., et al. Pyrosequencing of Antibiotic-Contaminated River Sediments Reveals High Levels of Resistance and Gene Transfer Elements. PLoS One 6: e17038. (2011). 28. Andersson, D,I. & Hughes, D. Microbiological effects of sublethal levels of antibiotics. Nat Rev Microbiol 12: 465–478. (2014). 29. Zhang, P.Y., et al. Combined treatment with the antibiotics kanamycin and streptomycin promotes the conjugation of Escherichia coli. Fems Microbiol Lett 348: 149–156. (2013). 30. Prudhomme, M., Attaiech, L., Sanchez, G., Martin, B. & Claverys, J.P. Antibiotic stress induces genetic transformability in the human pathogen Streptococcus pneumoniae. Science 313: 89–92. (2006). 31. Beaber, J.W., Hochhut, B. & Waldor, M.K. SOS response promotes horizontal dissemination of antibiotic resistance genes. Nature 427: 72–74. (2004). 32. Xia, Z.J., et al. Improving conjugation efficacy of Sorangium cellulosum by the addition of dual selection antibiotics. J Ind Microbiol Biotechnol 35: 1157–1163. (2008). 33. Stepanauskas, R., et al. Coselection for microbial resistance to metals and antibiotics in freshwater microcosms. Environ Microbiol 8: 1510–1514. (2006). 34. Baker-Austin, C., Wright, M.S., Stepanauskas, R. & McArthur, J.V. Co-selection of antibiotic and metal resistance. Trends Microbiol 14: 176–182. (2006). 35. Gullberg, E., et al. Selection of resistant bacteria at very low antibiotic concentrations. PLoS Pathog 7: e1002158. (2011). 36. Seiler, C. & Berendonk, T.U. Heavy metal driven co-selection of antibiotic resistance in soil and water bodies impacted by agriculture and aquaculture. Front Microbiol 3: 1–10. (2012). 37. Trapp, R.J., et al. Changes in severe thunderstorm environment frequency during the 21st century caused by anthropogenically enhanced global radiative forcing. Proc Natl Acad Sci USA 104: 19719–19723. (2007). 38. Tso, J., Dutta, S., Inamdar, S. & Aga, D.S. Simultaneous analysis of free and conjugated estrogens, sulfonamides, and tetracyclines in runoff water and soils using solid-phase extraction and liquid chromatography-tandem mass spectrometry. J Agric Food Chem 59:2213–2222. (2011). 39. Suzuki, M.T., Taylor, L.T. & DeLong, E.F. Quantitative analysis of small-subunit rRNA genes in mixed microbial populations via 5 ’-nuclease assays. Appl Environ Microbiol 66: 4605–4614. (2000). 40. Aminov, R.I., Garrigues-Jeanjean, N. & Mackie, R.I. Molecular ecology of tetracycline resistance: Development and validation of primers for detection of tetracycline resistance genes encoding ribosomal protection proteins. Appl Environ Microbiol 67: 22–32. (2001). 41. Chen, J., Yu, Z.T., Michel, F.C., Wittum, T. & Morrison, M. Development and application of real-time PCR assays for quantification of erm genes conferring resistance to macrolides-lincosamides-streptogramin B in livestock manure and manure

72

management systems. Appl Environ Microbiol 73: 4407–4416. (2007). 42. Wallace, J.S. & Aga, D.S. Enhancing Extraction and Detection of Veterinary Antibiotics in Solid and Liquid Fractions of Manure. J Environ Qual 45: 471-479. (2016). 43. Caporaso, J.G., et al. Global patterns of 16S rRNA diversity at a depth of millions of sequences per sample. Proc Natl Acad Sci USA 108: 4516–4522. (2011). 44. Caporaso, J.G., et al. Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J 6: 1621–1624. (2012). 45. Caporaso, J.G., et al. QIIME allows analysis of high-throughput community sequencing data. Nat Methods 7: 335–336. (2010). 46. DeSantis, T.Z., et al. Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARB. Appl Environ Microbiol 72: 5069–5072. (2006). 47. Magoč, T. & Salzberg, S.L. FLASH: fast length adjustment of short reads to improve genome assemblies. Bioinformatics 27: 2957–2963. (2011). 48. Bolger, A.M., Lohse, M., & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30: 2114–2120. (2014). 49. Li, B., et al. Metagenomic and network analysis reveal wide distribution and co- occurrence of environmental antibiotic resistance genes. ISME J 1–13. (2015). 50. Altschul, S., et al. Gapped BLAST and PSI- BLAST: a new generation of protein database search programs. Nucleic acids Res 25: 3389–3402. (1997). 51. Quast, C., et al. The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Res 41: D590–D596. (2013). 52. McArthur, A.G., et al. The comprehensive antibiotic resistance database. Antimicrob Agents Chemother 57: 3348–3357. (2013). 53. Pal, C., Bengtsson-Palme, J., Rensing, C., Kristiansson, E. & Larsson, D.G.J. BacMet: antibacterial biocide and metal resistance genes database. Nucleic Acids Res 42: D737– D743. (2014). 54. Leplae, R., Lima-Mendez, G. & Toussaint, A. ACLAME: a CLAssification of Mobile genetic Elements, update 2010. Nucleic Acids Res 38: D57–D61. (2010). 55. Buchfink, B., Xie, C. & Huson, D.H. Fast and sensitive protein alignment using DIAMOND. Nat Methods 12: 59–60. (2014). 56. Peng, Y., Leung, H.C.M., Yiu, S.M. & Chin, F.Y.L. IDBA-UD: a de novo assembler for single-cell and metagenomic sequencing data with highly uneven depth. Bioinformatics 28: 1420–1428. (2012). 57. Meyer, F., et al. The metagenomics RAST server - a public resource for the automatic phylogenetic and functional analysis of metagenomes. BMC Bioinformatics 9. (2008). 58. Gu, C. & Karthikeyan, K.G. Sorption of the antimicrobial ciprofloxacin to aluminum and iron hydrous oxides. Environ. Sci. Technol. 39: 9166–9173. (2005). 59. Gu, C. & Karthikeyan, K.G. Sorption of the antibiotic tetracycline to humic mineral complexes. J. Environ. Qual. 37: 704–711. (2008). 60. Lunestad, B.T. & Goksøyr, J. Reduction in the antibacterial effect of oxytetracycline in sea water by complex formation with magnesium and calcium. Dis. Aquat. Organ. 9: 67– 72. (1990). 61. Gullberg, E., Albrecht, L.M., Karlsson, C., Sandegren, L. & Andersson, D.I. Selection of a Multidrug Resistance Plasmid by Sublethal Levels of Antibiotics and Heavy Metals. MBio 5: 19–23. (2014). 62. Hellweger, F.L. Simple Model of Tetracycline Antibiotic Resistance in Aquatic

73

Environment : Accounting for Metal Coselection. J Environ Eng 139: 913–921. (2013). 63. Gillings, M.R., et al. Using the class 1 integron-integrase gene as a proxy for anthropogenic pollution. ISME J 1–11. (2014). 64. Engelstädter, J., Harms, K. & Johnsen, P.J. The evolutionary dynamics of integrons in changing environments. ISME J 1–12. (2016). 65. Gaze, W.H., et al. Impacts of anthropogenic activity on the ecology of class 1 integrons and integron-associated genes in the environment. ISME J 5: 1253–1261. (2011). 66. Berg, J., et al. Cu exposure under field conditions coselects for antibiotic resistance as determined by a novel cultivation-independent bacterial community tolerance assay. Environ Sci Technol 44: 8724–8728. (2010). 67. McKinney, C.W., Loftin, K.A., Meyer, M.T., Davis, J.G. & Pruden, A. tet and sul Antibiotic Resistance Genes in Livestock Lagoons of Various Operation Type, Configuration, and Antibiotic Occurrence. Environ Sci Technol 44: 6102–6109. (2010). 68. Gibson, M.K., Forsberg, K.J. & Dantas, G. Improved annotation of antibiotic resistance determinants reveals microbial resistomes cluster by ecology. ISME J 1–10. (2014). 69. Pal, C., Bengtsson-palme, J., Kristiansson, E. & Larsson, D.G.J. Co-occurrence of resistance genes to antibiotics, biocides and metals reveals novel insights into their co- selection potential. BMC Genomics 16. (2015). 70. Klümper, U., et al. Broad host range plasmids can invade an unexpectedly diverse fraction of a soil bacterial community. ISME J 9: 934–945. (2014). 71. Gibbons, S.M., et al. Human and Environmental Impacts on River Sediment Microbial Communities. PLoS One 9: e97435. (2014). 72. Shade, A., et al. Lake microbial communities are resilient after a whole-ecosystem disturbance. ISME J 6: 2153–2167. (2012). 73. Durso, L.M., Miller, D.N. & Wienhold, B.J. Distribution and Quantification of Antibiotic Resistant Genes and Bacteria across Agricultural and Non-Agricultural Metagenomes. PLoS One 7: 1–12. (2012). 74. Ju, F., et al. Antibiotic Resistance Genes and Human Bacterial Pathogens: Co- occurrence, Removal, and Enrichment in Municipal Sewage Sludge Digesters. Water Res 91: 1–10. (2015)

74

SUPPLEMENTARY INFORMATION FOR CHAPTER 4

Figure S1: Relative abundance of ARGs as determined by metagenomic analysis of site 1 and site 5 Poudre River bulk water samples. (A) Mechanisms of antibiotic resistance and (B) classes of ARGs determined by metagenomic analysis and annotation against the Comprehensive Antibiotic Resistance Database.

Figure S2: Relative abundance (ARG copies / 16S rRNA gene copies) of ARGs as determined by metagenomic analysis of site 1 and site 5 Poudre River bulk water samples

75

Figure S3: Quantification of select ARGs by quantitative polymerase chain reaction (qPCR) in Poudre River sediment (bars) and bulk water (points), normalized to 16S rRNA genes at 12 months (-12) before the flooding occurred and at five time points following the flooding. X-axis indicates sites and months relative to the flooding event. (*) indicates gene detected below quantification limit in sediment and (+) in water. Error bars represent standard deviation of triplicate qPCR measurements in water and standard deviation of triplicate samples in sediment.

76

Figure S4: Phyla accounting for greater than 1% of the total OTUs in bulk water, determined by 16S rRNA gene sequencing.

77

Figure S5: Phyla accounting for greater than 1% of the total OTUs in bulk water, determined by 16S rRNA gene sequencing. Triplicate sediment samples were sequenced separately and results averaged.

Figure S6: Spearman’s Rank Correlation Coefficient between abundance of ARGs normalized to 16S rRNA genes, as determined by qPCR, and water quality paramters. Statistically significant (p<0.05) correlations are indicated in bold.

78

Figure S7: Rarefaction curves for metagenomic samples.

Figure S8: ARG copies determined by qPCR in WWTP effluent, normalized to 16S rRNA gene copies.

79

Table S1: Characteristics of metagenomic data. All sequences have been deposited in MG-RAST under project name “Fate and Transport of Antibiotics and Antibiotic Resistance Genes during Historic Colorado Flood.” Average Maximum MG-RAST ID # Reads # scaffold scaffold Sample Name (unassembled) (unassembled) scaffolds length length (bp) (bp)

Poudre_WSite1_Sept2012 4628878.3 1,397,640 131,397 1005 137,041

Poudre_WSite1_Dec2013 4628876.3 1,182,555 81,537 644 35,373

Poudre_WSite1_July2014 4628877.3 1,460,502 66,797 794 78,321

Poudre_WSite5_Sept2012 4628881.3 1,320,893 86,492 625 153,801

Poudre_WSite5_Dec2013 4628879.3 2,232,495 96,292 751 91,128

Poudre_WSite5_July2014 4628880.3 1,860,238 97,413 829 215,852

80

Table S2: Antibiotic concentrations in Poudre River bulk water (ng/L). Standard deviation of replicate samples denoted in parentheses and months indicate months post-flood. “W” denotes Wastewater Treatment Plant samples. Antibiotics abbreviations are denoted as follows: anhydrotetracycline (ATC), azithromycin (AZI), clarithromycin (CLA), chlorotetracycline (CTC), doxycycline (DOX), erythromycin (ERY), 4-epitetracycline (ETC), oxytetracycline (OTC), sulfamerazine (SMR), sulfamethoxazole (SMX), sulfamethazine (SMZ), sulfadiazine (SPD), tetracycline (TC), and tylosin (TYL). Sulfameter, sulfamethiazole, sulfamerazine, sulfachloropyridazine, sulfathiazole, roxithromycin, spiramycin, 4-epichlorotetracycline, anhydrochlorotetracycline, demeclocycline (surrogate), minocycline (internal standard), phenyl-13C6-sulfamethazine (13C6-SMZ), d4-sulfamethoxazole (d4-SMX), N-methyl 13C-erythromycin, and d10- carbamazepine (internal standard) were not detected in any samples. Site ATC AZI CLA CTC DOX ERY ETC OTC SMR SMX SMZ SPD TC TYL

1 ND ND ND ND ND 6.7(1.5) ND ND ND 1.1(0.0) ND ND ND ND 2 2.1(0.1) ND ND 3.6(2.0) ND 3.3(0.9) ND ND ND 17(0.3) 0.7(0.0) ND 4.6(1.2) ND 3 5.7(0.0) ND ND 6.7(0.0) ND 7.7(0.0) ND ND 0.9(0.0) 29 (0.0) 0.5(0.0) ND ND ND 4 ND 3.7(0.1) 3.2(1.3) ND ND 11.(3.8) ND ND ND 154(9.5) 3.5(0.2) ND 6.4(1.0) ND 3 months 5 ND 15.(0.6) 11.(4.0) ND ND 16.(0.7) ND ND ND 223(21.) 9.1(1.0) ND ND ND

1 ND ND ND ND ND 1.8(0.0) ND ND ND 0.6(0.1) 0.2(0.0) ND ND ND 2 2.6(0.7) ND ND ND ND 1.6(0.0) 3.2(0.2) ND ND 0.5(0.0) 0.4(0.0) ND 59.(2.0) ND 3 ND 1.9(0.2) 1.6(0.1) ND ND 3.0(0.2) ND ND ND 21 (4.9) ND ND ND ND 4 ND 2.5(0.3) 2.8(1.1) ND ND 10.(4.2) ND ND ND 93(10) ND 5.8(0.9) ND ND 6 months 5 ND 9.3(2.4) 15.(1.0) ND ND 20.(0.0) ND ND ND 55.(1.0) 2.0(0.5) 1.8(0.1) ND ND 1 ND ND ND ND ND 1.5(0.1) ND ND ND ND ND ND ND ND 2 ND ND ND ND ND 1.5(0.0) ND 6.6(0.8) ND ND ND ND ND ND 3 ND ND ND ND ND 1.6(0.0) ND 5.1(0.2) ND 19.(1.6) ND ND ND ND 4 ND ND 0.6(0.0) ND ND 3.5(0.1) ND ND ND 127(6.3) ND 6.5(0.4) ND ND

8 months 5 ND ND 1.0(0.0) ND ND 5.0(0.2) ND ND ND 113(7.9) 2.4(0.1) 1.7(0.2) ND ND W ND 510(55) 18.(1.1) ND 72(6.5) 5.1(0.4) ND ND ND 68.(3.8) ND ND ND 2.1(0.0)

1 ND ND ND ND ND 2.3(0.1) ND ND ND ND ND ND ND ND 3 ND ND 1.0(0.0) ND ND 3.6(0.3) ND ND ND 20.(0.2) ND ND ND ND 4 ND ND 0.6(0.1) ND ND 9.8(0.2) ND ND ND 44.(0.1) ND ND ND ND 5 ND ND 1.9(0.0) ND ND 12.(0.0) ND ND ND 36.(2.9) 1.1(0.0) ND ND ND 10 months W ND 651(4.8) 67.(1.7) ND ND 58.(0.4) ND ND ND 770(41.) ND ND 9.0(1.5) 6.0(0.3) 1 ND ND ND ND ND 1.7(0.1) ND ND ND ND ND ND ND ND 2 ND ND ND ND ND 2.8(0.0) ND ND ND 0.7(0.2) ND ND ND ND 3 ND 16.(0.5) 3.5(0.6) ND ND 4.1(0.1) ND ND ND 12.(0.2) ND ND ND ND

months 4 ND 17.(0.2) 12.(0.1) ND ND 18.(1.0) ND ND ND 231(1.7) 1.1(0.0) 3.5(0.6) ND ND

18 5 ND 60.(2.2) 52.(0.2) ND ND 53.(1.1) ND ND ND 58.(3.5) 2.1(0.1) 3.5(0.1) ND 4.3(0.6) W 4.4(0.2) 250(100) 620(20) ND ND 41.(1.7) ND 18.(0.2) ND 137(0.3) ND 2.2(0.2) 59.(3.2) 7.9(0.1)

81

Table S3: p-values for Kruskal-Wallis rank sum tests for correlations between ARGs and antibiotics of metals. Significant (p<0.05) values indicated in bold. sul1 sul2 tet(O) tet(W) ermF sed wat sed wat sed wat sed Wat sed wat Ag 0.014 0.0004 0.6488 0.2661 0.7732 0.7897 0.2197 0.4207 0.2595 0.2864 As 0.6341 0.5165 0.9109 0.0875 0.9925 0.0004 0.156 0.0775 0.8686 0.0428 ATC 0.1385 0.8748 0.3638 0.8283 0.5209 0.0187 0.1604 0.0787 0.8122 0.7852 AZI 0.0028 0.6716 0.084 0.0996 0.0021 0.0328 0.0194 0.048 0.0407 0.1402 Ba 0.0112 0.0005 0.5486 0.3738 0.6882 0.0048 0.0608 0.1501 0.9313 0.1457 Ce 0.3905 0.5377 0.4453 0.0968 0.0681 0.473 0.1171 0.3863 0.5664 0.9147 CLA 0.0155 0.588 0.1651 0.0297 0.0079 0.0037 0.0555 0.0078 0.064 0.0234 Cr 0.6204 0.9468 0.4608 0.4307 0.755 0.5655 0.6454 0.8458 0.3712 0.8841 CTC 0.4515 0.7914 0.6364 0.7108 0.8723 0.0898 0.2462 0.558 0.7566 0.3645 Cu 0.028 0.013 0.7977 0.4164 0.2552 0.2459 0.4342 0.7183 0.1263 0.8276 DOX 0.1705 0.1176 0.113 0.1173 0.0574 0.1578 0.0813 0.1444 0.0853 0.1305 ERY 0.0229 0.5509 0.7548 0.0437 0.0022 0.0005 0.0049 0.0088 0.1496 0.0175 ETC 0.9601 0.1176 0.6514 0.5595 0.9109 0.3604 0.7836 0.4882 0.4126 0.3099 Gd 0.316 0.4792 0.2073 0.1049 0.342 0.4617 0.1058 0.5577 0.4561 0.8529 Mn 0.3192 0.8286 0.3449 0.6604 0.1302 0.5455 0.2354 0.5222 0.7028 0.9192 OTC 0.1641 0.0787 0.5761 0.5217 0.4081 0.9698 0.1389 0.9953 0.2185 0.5687 Pb 0.5159 0.8011 0.1292 0.7052 0.805 0.8663 0.6834 0.8715 0.371 0.5374 Pt 0.7491 0.2577 0.3006 0.2252 0.7462 0.7887 0.4012 0.8433 0.0203 0.7661 SDM 0.1138 0.6336 0.7255 0.425 0.9109 0.9574 0.7836 0.9577 0.4126 0.9577 SMX 0.0028 0.6148 0.1742 0.023 0.0097 0.0046 0.0227 0.0095 0.0307 0.0008 SMZ 0.11 0.5533 0.8296 0.4429 0.4746 0.0081 0.9784 0.0424 0.6661 0.0557 SPD 0.0065 0.2857 0.3429 0.1065 0.3866 0.4074 0.6908 0.102 0.8154 0.1641 Sr 0.1884 0.905 0.0234 0.0015 0.4179 0.0141 0.6493 0.0116 0.1652 0.0128 TC 0.5532 0.7814 0.5392 0.2495 0.1464 0.0006 0.3646 0.0072 0.4571 0.1517 Th 0.0878 0.0038 0.8713 0.5892 0.7505 0.4899 0.3679 0.8272 0.082 0.1872 TYL 0.0884 0.6962 0.0207 0.185 0.0054 0.3075 0.0017 0.1436 0.0044 0.7039 U 0.4071 0.9181 0.0858 0.0075 0.8281 0.0079 0.2449 0.0155 0.3516 0.0125

82

Table S4: Metal concentrations in Poudre River bulk water (µg/L). Standard deviation of replicate samples denoted in parentheses.

Site Ag As Ba Cd Ce Cr Cu Gd Mn Pt Pb Sr Th U 1 ND 0.6(0.0) 41.(0.4) ND 0.7(0.0) ND 0.5(0.0) ND 22.(0.2) ND ND 89.(0.6) ND 1.6(0.0) 2 ND 0.7(0.0) 46.(0.9) ND 1.0(0.0) 0.3(0.1) 0.6(0.1) ND 39.(0.2) ND 0.3(0.0) 136(0.7) ND 1.9(0.0) 3 ND 0.8(0.2) 47.(0.9) ND 1.1(0.0) 0.4(0.0) 0.9(0.0) ND 47.(0.4) ND 0.3(0.0) 224(2) ND 2.1(0.0)

3 months 4 ND 1.1(0.2) 44.(0.6) ND 0.7(0.0) ND 0.7(0.0) ND 42.(0.0) ND ND 395(0.6) ND 3.2(0.0) 5 ND 2.3(0.2) 42(1) ND ND ND 0.8(0.0) ND ND ND ND 730(14) ND 6.3(0.3) 1 ND 0.5(0.0) 39.(0.5) ND ND ND 0.4(0.0) ND 15.(0.0) ND ND 94.(0.8) ND 1.6(0.0) 2 ND 0.4(0.0) 61(1) ND 4.3(0.0) 2.3(0.0) 2.0(0.0) 0.3(0.0) 100(0.6) ND 1.6(0.0) 116(0.6) ND 1.6(0.0) 3 ND 1(0.1) 50.(0.7) ND 0.3(0.0) ND 1.1(0.0) ND 53.(0.5) ND ND 626(8) ND 3.9(0.1)

6 months 4 ND 2(0.3) 51.(0.8) ND ND ND 0.9(0.0) ND 67.(0.6) ND ND 123(3) ND 8.1(0.0) 5 ND 4.1(0.7) 57(1) ND 0.6(0.0) ND 1.8(0.0) ND 176(0.4) ND 0.3(0.0) 147(6) ND 18.(0.2) 1 ND ND 43(2) ND 5.4(0.3) 2.5(0.0) 2.0(0.0) 0.3(0.0) 76.(0.3) ND 1.7(0.0) 44(0.4) ND 0.7(0.0)

2 ND ND 35.(0.4) ND 3.0(0.0) 1.3(0.0) 1.4(0.0) ND 46(0.4) ND 0.8(0.0) 55.(0.3) ND 0.7(0.0) 3 ND ND 28(1) ND 2.6(0.1) 1.1(0.0) 1.1(0.0) ND 42.(0.1) ND 0.9(0.0) 48.(0.1) ND 0.3(0.0) 4 ND ND 42(2) ND 1.1(0.0) 0.4(0.0) 0.5(0.0) ND 38.(0.9) ND 0.4(0.0) 235(0.5) ND 1.8(0.1) 8 months 5 ND ND 45(1) ND 2.2(0.1) 0.6(0.0) 0.8(0.0) ND 149(0.3) ND 0.9(0.0) 410(20) ND 6(0.3) W ND ND 30.(0.9) ND ND ND 1.6(0.0) ND 19.(0.1) ND ND 318(0.5) ND 1.1(0.0) 1 ND ND 51(2) ND ND ND 3.6(0.0) ND 5.9(0.1) ND ND 33.(0.4) ND 0.3(0.0)

3 ND ND 49(1) ND ND ND 3.2(0.0) ND 19.(0.1) ND ND 292(3) ND 1.2(0.0) 4 ND 1.9(0.1) 54(0.5) ND ND ND 2.7(0.0) ND 11.(0.0) ND ND 723(2) ND 4.3(0.1) 5 ND 3.7(0.2) 70(2) ND ND ND 2.3(0.0) ND 57.(0.8) ND ND 105(10) ND 12.(0.6)

10 months W ND 0.6(0.2) 57(2) ND ND ND 5.2(0.1) ND 12.(0.0) ND ND 846(3) ND 3.5(0.1) 1 0.7(0.1) 0.2(0.0) 31.(0.8) ND ND ND 1.2(0.0) ND 7(0.1) 0.5(0.1) ND 84(3) 0.7(0.0) 0.6(0.0)

2 0.7(0.1) 0.3(0.0) 34(1) ND ND ND 0.7(0.0) ND 14.(0.5) ND ND 150(30) 0.3(0.0) 0.6(0.0) 3 0.5(0.0) 0.3(0.1) 32.(0.5) ND ND ND 0.6(0.1) ND 26(0.5) ND ND 350(50) ND 2.2(0.2) 4 0.8(0.1) 0.6(0.0) 44(1) ND ND ND 0.7(0.0) ND 66(5) ND ND 160(300) 0.2(0.0) 7.7(0.0)

18 months 5 0.5(0.0) 0.7(0.9) 38.(0.2) ND ND ND 1.4(0.0) ND 64(2) ND ND 120(300) ND 14.(0.0) W 0.5(0.0) 0.2(0.0) 33.(0.1) ND ND 0.3(0.0) 1.9(0.0) ND 30(1) ND 0.3(0.0) 388(30) ND 2.4(0.0)

83

CHAPTER 5 : METAGENOMIC CHARACTERIZATION OF ANTIBIOTIC RESISTANCE GENES IN FULL-SCALE RECLAIMED WATER DISTRIBUTION SYSTEMS AND CORRESPONDING POTABLE SYSTEMS

Emily Garner, Chaoqi Chen, Kang Xia, Jolene Bowers, David M. Engelthaler, Jean McLain, Marc A. Edwards, Amy Pruden

ABSTRACT

Water reclamation provides a valuable resource for meeting non-potable water demands. However, little is known about the potential for wastewater reuse to disseminate antibiotic resistance genes (ARGs). Here, samples were collected seasonally in 2014-2015 from four U.S. utilities’ reclaimed and potable water distribution systems before treatment, after treatment, and at five points of use (POU). Shotgun metagenomic sequencing was used to profile the resistome (i.e., full contingent of ARGs) of a subset (n=38) of samples. Four ARGs (qnrA, blaTEM, vanA, sul1) were quantified by quantitative polymerase chain reaction. Bacterial community composition (via 16S rRNA gene amplicon sequencing), horizontal gene transfer (via quantification of intI1 integrase and plasmid genes), and selection pressure (via detection of metals and antibiotics) were investigated as potential factors governing the presence of ARGs. Certain ARGs were elevated in all (sul1; p≤0.0011) or some (blaTEM, qnrA; p≤0.0145) reclaimed POU samples compared to corresponding potable samples. Bacterial community composition was weakly correlated with ARGs (Adonis, R2=0.1424-0.1734) and associations were noted between 193 ARGs and plasmid-associated genes. This study establishes that reclaimed water conveys greater abundances of certain ARGs than potable waters and provides observations regarding factors that likely control ARG occurrence in reclaimed water systems.

INTRODUCTION

Reclamation and reuse of municipal wastewater effluent is increasingly relied on to offset demand on traditional potable water sources. Water reuse can help address challenges such as water shortages, groundwater depletion, surface water contamination, increasing demand due to population growth, and exacerbated water stress due to climate change.1 However, even as its application expands worldwide, there are technical challenges and public health concerns that must be assessed, such as trace contaminants, including antibiotics and personal care products,1 and microbial constituents, such as viruses2,3 and antibiotic resistance elements.4

The potential of water reuse to contribute to the spread of antibiotic resistance has drawn attention.4,5 Antibiotic resistance is a critical public health challenge, with over two million antibiotic resistant bacterial infections documented in the U.S. each year6 and even more globally.7 Strategies to mitigate the spread of antibiotic resistance have primarily focused on optimizing clinical use, limiting application in agriculture, and improving hygiene in hospitals.7–10 While such efforts are vital, they are limited in effectiveness due to difficulty of implementation and because they do not take into consideration broader sources and routes of resistance dissemination associated with natural and built environments.8 Correspondingly, there is now growing movement towards identifying comprehensive mitigation strategies to prevent the evolution and spread of antibiotic resistance as an environmental “contaminant”.8,11,12 In this context, water reuse and the

84

water cycle in general have the potential to contribute to the proliferation of antibiotic resistant bacteria (ARB) and antibiotic resistance genes (ARGs). Numerous studies have now documented the abundance of ARB and ARGs in wastewater, which like many microorganisms, are not always removed completely by traditional wastewater treatment.13–18 Previous studies have indicated that a diverse range of ARB and ARGs are present in reclaimed (i.e., non-potable) water,19–21 but, given that antibiotic resistance is a natural phenomenon occurring in many aquatic and soil bacteria, it is important to move towards advancing understanding of which ARB and ARGs actually pose risk to human health. For example, samples collected from ancient permafrost and unexplored caves contain a surprisingly diverse array of ARGs.22–24 To address the presence of ARGs in even pristine environments, benchmarking the presence of ARGs in water reuse treatment and distribution systems to that of corresponding potable water systems can help provide a frame of reference for assessing potential risks associated with water reuse compared to water derived from surface and groundwater. Discriminating amongst various classes of ARGs and their location in the genome is also important, with ARGs that encode resistance to clinically-important antibiotics and that are capable of disseminating resistance via horizontal gene transfer being of greatest concern.25,26

Municipal sewage represents a composite reservoir of excreted bacteria and associated DNA, where its collection and treatment will likely select for certain strains and, in some cases, could create conditions ideal for horizontal transfer of clinically-important ARGs.27,28 Wastewater treatment plants (WWTP) have been proposed as potential “hot spots” for ARB proliferation.15 Correspondingly, poor efficiency of ARB and ARG removal has been noted with some conventional WWTPs.5,15,29 In particular, antibiotics, and other potential selective agents; such as heavy metals, herbicides, and disinfectants, have been associated with the loading of ARGs in water and soil environments30–33 and their presence in wastewater is expected to play a similar role. Further, shotgun metagenomic approaches and tracking of plasmids and other gene transfer elements have revealed evidence of high rates of horizontal transfer of ARGs among densely populated activated sludge bacteria core to the WWTP biological treatment process.34–37

While considerable research has been devoted to understanding these phenomena in WWTPs and receiving environments, few studies have explored the potential for dissemination of ARB and ARGs via subsequent water reuse. While some studies have examined the implications of irrigation38–40 or groundwater recharge41,42 with reclaimed water for ARG dissemination, only recently has the potential for bacterial regrowth of ARG-carrying bacteria during reclaimed water distribution been reported in the peer-reviewed literature.19 Given that indicator organisms43 and opportunistic pathogens44 have both exhibited patterns of regrowth during distribution in reclaimed water systems, the potential for regrowth of ARG-carrying bacteria warrants consideration. Distribution system biofilms are also worthy of investigation given that biofilms have been identified as reservoirs for pathogenic bacteria in drinking water systems45 and supportive environments for horizontal gene transfer.46

Here we used quantitative polymerase chain reaction (qPCR) to survey four ARGs; blaTEM, qnrA, vanA, and sul1, and the intI1 integrase gene, known to facilitate horizontal gene transfer, in full-scale reclaimed water distribution systems located in four U.S. cities that implement non- potable reuse. Shotgun metagenomic sequencing was applied towards profiling the broader “resistome” (i.e., full contingent of ARGs)47 in a cross-section of samples and the bacterial community composition was tracked using 16S rRNA amplicon sequencing to explore potential

85

microbial ecological drivers of ARG occurrence. The specific objectives of this work were to: 1) characterize the resistome of reclaimed water distribution systems relative to corresponding potable systems; 2) quantify abundances of ARGs inhabiting the water versus the biofilm; 3) measure removal of ARGs during treatment and any subsequent increases during transport of reclaimed water to the point of use (POU); 4) explore associations between ARG occurrence and the bacterial community composition; 5) investigate potential for ARG dissemination via horizontal gene transfer; and 6) examine associations between water chemistry, particularly potential selective agents, and the abundance of ARGs. Realization of the objectives of this work will provide context for understanding the potential for water reuse to disseminate ARB and ARGs and inform development of management strategies for limiting dissemination of ARB and ARGs via distribution system operation.

METHODS

Site description, sample collection, and preservation

Four water utilities utilizing tertiary wastewater treatment to produce reclaimed water participated in sampling and are described in Table 5-1. Details about the four seasonal collection dates conducted for each utility are provided in Table S1. For each reclaimed system, samples were collected of raw wastewater influent, following treatment at the point of entry (POE) to the distribution system, and at five POUs. For each potable system, samples were collected of source water, at the POE, and at five POUs. After flushing for 30 seconds, water samples for molecular analysis were collected via distribution system sampling ports in sterile one liter polypropylene containers. Samples for organic carbon analysis were collected in acid-washed, baked 250 milliliter amber glass bottles. All bottles were prepared with 48 milligrams sodium thiosulfate per liter to quench chlorine, and bottles for molecular analysis were also prepared with 292 milligrams ethylenediaminetetraacetic acid per liter to chelate metals. Water was collected in acid-washed 250 milliliter bottles for chemical analyses. For Utilities A and B, after collecting water samples, biofilm samples were collected at each POU by inserting a sterile cotton-tipped applicator (Fisher Scientific, Waltham, MA) into the distribution system pipe and swabbing the upper 180º of the circumference of the pipe in a single pass. The sample end of the swab was transferred directly to a sterile DNA extraction lysing tube.

Samples were shipped overnight on ice and processed immediately upon arrival, within approximately 24 hours of sample collection. Samples for molecular analysis were concentrated onto 0.22 micron mixed cellulose esters membrane filters (Millipore, Billerica, MA) until the entire one liter sample passed or until the filter became clogged. Filters were folded into quarters, torn into 1 cm2 pieces using sterile forceps, transferred to lysing tubes, and preserved at -20ºC. DNA was extracted from filters and biofilm swabs using a FastDNA SPIN Kit (MP Biomedicals, Solon, OH).

86

Table 5-1: Overview of surveyed potable and reclaimed systems U.S. Region Potable Reclaimed Utility (Climate)a Disinfectant Source Summary of Treatment Disinfectant Southeast Plant #1b – 5-stage c (Humid Surface and Bardenpho Cl2 A NH2Cl b e Subtropical; Groundwater Plant #2 - Activated (NH2Cl)

Cfa) sludge, dentirification Southwest Plant #1d – 4-stage Cl Cl ; 2 (Mediterran; 2 Surface and Bardenphoc (NH Cl)e B occasional 2 Csb) Groundwater ClO UV 2 Plant #2d - Biofiltration Southwest (Mid- Surface and Dual membrane filters or C Latitude Cl NH Cl 2 Groundwater membrane bioreactors 2 Steppe and Desert; Bsh) West Surface and Cl D (Mediterran; Cl Dual media filters 2 2 Groundwater (NH Cl)e Csb) 2 aKöppen climate classification: Cfa = mild temperate, fully humid, hot summer; Csb = mild temperate, dry summer, warm summer; Bsh = dry, dry summer, hot arid bUtility A plants feed into two isolated distribution systems (A1 and A2) cBardenpho refers to activated sludge process modified to optimize biological nutrient removal dUtility B plants feed into 1 combined distribution system eFree chlorine was dosed but water chemistry data indicates that total chlorine >> free chlorine, thus free chlorine reacted with ambient ammonia and resulted in NH2Cl as the primary form of disinfectant residual (Table S5)

Water chemistry

Free chlorine (4500Cl G), total chlorine (4500Cl G), temperature (2550 B), dissolved oxygen (4500-O G), pH (4500-H+ B), turbidity (2130 B), and electrical conductivity (2510 B) were measured on-site according to standard methods.48 Upon return to the lab, one 30 milliliter aliquot was taken for total organic carbon (TOC) and a second aliquot was filtered through pre- rinsed 0.22 micron pore size mixed cellulose esters membrane filters (Millipore, Billerica, MA) for dissolved organic carbon (DOC). Biodegradable dissolved organic carbon (BDOC) was measured as previously described by Servais et al.49 with an incubation time extended to 45 days. Samples were analyzed on a Sievers 5310C portable TOC analyzer (GE, Boulder, CO) according to Standard Method 5310C.48 A host of 28 metals, including sodium, magnesium, phosphorus, chloride, calcium, iron, copper, zinc, and lead were measured using an Electron X-Series inductively coupled plasma mass spectrometer (ThermoFisher, Waltham, MA) according to Standard Method 3125B.48 Nitrate, nitrite, phosphate, and sulfate were quantified via a Dionex- 500 ion chromatography system (ThermoFisher, Waltham, MA) according to Standard Method 4110B.48 Antibiotics were extracted from samples using solid phase extraction according to Sui et al.50 with minor modifications described in the supporting information. The following antibiotics were analyzed using an ultra-performance liquid chromatography-tandem mass spectrometer

87

(UPLC-MS/MS; Agilent 1290 UPLC/Agilent 6490 Triple Quad tandem MS, Agilent Technologies Inc., Santa Clara, CA): cefotaxime, chlorotetracycline, erythromycin, flumequine, nalidixic acid, ormetoprim, ornidazole (anti-protozoan), oxolinic acid, sulfamethazine, sulfamethoxazole, tetracycline, and tylosin.

Quantification of antibiotic resistance genes

Gene copies were quantified on a CFX96 Real Time System (BioRad, Hercules, CA) from DNA extracts in triplicate reactions using qPCR with previously published protocols for 16S 51 52 53 54 55 56 rRNA, blaTEM, qnrA, vanA, sul1, and intI1 genes. These genes were selected based on relevance to human health and environmental transmission of ARGs. The genes represent a spectrum of typical documented prevalence in the environment. For example, sul1 has been widely documented in wastewater impacted environments19,57 and the gene corresponds to a class of antibiotics (sulfonamides) for which resistance of human pathogens has become commonplace.58 In contrast, vanA encodes resistance to a “last resort” drug and is less common in wastewater impacted environments.4,19 A subset of samples was initially analyzed at 5, 10, 20, and 50 fold dilutions to determine the optimal dilution effective for minimizing inhibition. A ten-fold dilution was selected and applied to all extracts. A triplicate negative control and triplicate standard curves of ten-fold serially diluted standards, constructed as described in the supporting information, of each target gene ranging from 101 to 107 gene copies/µl were included on each 96-well plate. The limit of quantification was 10 gene copies per milliliter of sampled water and 103 gene copies per biofilm swab.

Shotgun metagenomics and 16S rRNA amplicon sequencing

To profile the resistome, shotgun metagenomic sequencing was conducted on DNA extracted from the POE and greatest water age POU samples from each reclaimed system on each collection date as previously described,59 with sequencing conducted on an Illumina HiSeq with 2x100-cycle paired end reads at the Biocomplexity Institute of Virginia Tech Genomics Sequencing Center (BI; Blacksburg, VA). One source water sample, the POE, and the greatest water age POU sample from the potable system of each utility (all collected during the summer collection from each utility) were also submitted for sequencing. All potable samples from Utilities C and D and a subset of samples from Utility A yielded insufficient DNA to conduct metagenomic sequencing.

Functional genes were annotated via the MetaStorm platform60 according to default parameters (amino acid identity≥80%; e-value cutoff=1e-10; minimum alignment length=25 amino acids) using annotation to the Comprehensive Antibiotic Resistance Database (CARD, version 1.0.6) for ARGs,61 the Silva ribosomal RNA database for 16S rRNA genes (version 123),62 the BacMet database (version 1.1) for metal resistance genes,63 and the ACLAME database (version 0.4) for plasmids.64 Functional genes were normalized to 16S rRNA genes as previously described by Li et al.65 Absolute abundances were calculated by multiplying relative abundance of target functional genes by total abundance of 16S rRNA genes quantified by qPCR (Figure S1). All metagenomes generated in this study are publicly available via MG-RAST66 under project number 12943 (see Table S2 for sample IDs and read and assembly statistics). Reads were

88

assembled de novo in MetaStorm according to default parameters and scaffolds were annotated as described above for reads.

Bacterial communities were profiled using gene amplicon sequencing targeting the V4 region of the 16S rRNA gene with barcoded primers (515F/806R).67 Some archaea are also detected by these primers. Triplicate PCR products were combined and 240 ng of each composite was pooled and purified using a QIAquick PCR Purification Kit (Qiagen, Valencia, CA). Sequencing was conducted at BI on an Illumina MiSeq using a 2x250-cycle paired-end protocol. Processing of reads was conducted using the QIIME pipeline68 with phylogenetic inference based on alignment against the Greengenes database (May 2013 release).69 Samples were rarefied to 10,000 randomly selected reads (Table S3). Field, filtration, DNA extraction blanks, and a least one PCR blank per lane were included in the analysis.

Statistical Analysis

A Wilcoxon rank sum test for multiple comparisons was applied in JMP (version 13, SAS, Cary, NC) to determine differences between abundances of ARGs across groups of samples. Spearman’s rank sum correlation coefficients were calculated in JMP to assess correlations between ARGs, phyla, Gram-stain type (identified from the literature for each phyla to the extent possible), and water quality parameters. Canonical correspondence analysis was conducted in R (version 3.4.1) using the Adonis function from the vegan package70 to identify ARGs that are significantly correlated with the bacterial community structure. Abundance of ARGs based on metagenomic data were imported into PRIMER-E (version 6.1.13) for Bray-Curtis resemblance matrix construction and one-way analysis of similarities (ANOSIM) to compare profile differences across groups of samples. A significance cutoff of α=0.05 was used for all analyses. Co- occurrences of annotated genes on scaffolds were characterized via network analysis visualization using Gephi (version 0.8.2).

RESULTS AND DISCUSSION

Metagenomic characterization of the resistome in reclaimed versus potable water

Shotgun metagenomic sequencing was used to investigate the abundance of known ARGs by annotating reads against the CARD database. Seventeen classes of antibiotic resistance, as well as multidrug resistance, were identified across all samples (n=38, Figure 5-1). Across the dataset, 590 different ARGs were annotated, with between 16 and 372 different ARGs detected in each sample. The top 25 most abundant ARGs overall included four aminoglycoside (aadA23, ant(2”)- la, aadA, aadA17), one sulfonamide (sul1), one trimethoprim (dfrE), one polymyxin (pmrE), one rifampin (rbpA), two beta-lactam (blaOXA-256, blaOXA-129), two tetracycline (tetC, tetQ), one fluoroquinolone (qnrS6), one macrolide (ereA2), and eight multidrug (msrE, mtrA, mexF, mexK, CRP, adeG, mexW, acrB) ARGs. The average number of sequencing reads aligning to different ARGs per sample was 26.8±9.2 per ten million reads in the potable water and 69.0±42.7 per ten million reads in the reclaimed water, though differences were not significant (Wilcoxon; p=0.0891).

89

Among the potable samples that were successfully sequenced, multidrug ARGs were common (8.4-33.4% of total ARGs) and the most abundant classes of ARGs were trimethoprim (10.6-49.7%), aminocoumarin (0-18.5%), beta lactam (1.3-38.5%), polymyxin (0.4-11.1%), and aminoglycoside (0.9-9.8%) resistance. The abundance of total ARGs in potable water ranged from 3.93–6.83 log gene copies per milliliter and 4.33–5.32 log gene copies per swab in the biofilm.

Reclaimed water ARG profiles were distinct from that of potable samples (ANOSIM; R=0.705, p=0.001; Figure 5-1; Figure S2). In the reclaimed water, Utility A’s ARG profile stood out from that of other utilities (R=0.695-0.932, p=0.001), dominated by aminoglycoside (34.5- 66.8% of total ARGs) and sulfonamide (29.2-51.9%) ARGs, whereas Utilities B, C, and D were generally dominated by multidrug (10.0-40.8%), trimethoprim (8.4-50.5%), sulfonamide (0-38.3), tetracycline (0.97-13.2%), and beta-lactam (1.07-16.9%) ARGs. The biofilms of Utilities A and B exhibited patterns that were more similar to the respective water samples of those utilities than to each other, dominated by aminoglycoside (44.3%), sulfonamide (22.3%), and trimethoprim ARGs (13.8%) for Utility A and multidrug (44.6%), rifampin (24.0%), and trimethoprim (17.9%) for Utility B. The abundance of total ARGs in the reclaimed water ranged from 5.24–6.53 log gene copies per milliliter and 4.82–6.14 log gene copies per biofilm swab.

The resistance profile of the treated water at the POE was markedly different from that of the influent wastewater for each utility (ANOSIM; R=0.971, p=0.002), consistent with the expectation that the treatment process shifts the types and numbers of ARGs relative to raw sewage. As expected, the greatest abundances of total ARGs were generally found in the raw wastewater influent samples (6.81–7.88 log gene copies per milliliter). Influent wastewater tended to be dominated by multidrug (16.1-35.4% of total ARGs), aminoglycoside (12.7-30.8%), beta- lactam (3.1-21%), tetracycline (6.3-16.7%), macrolide (2.4-10.4%), and sulfonamide resistance (1.2-19.5%), though there was variation across utilities (Figure 5-1).

While metagenomic analysis provides tremendous potential for characterizing the resistome and allowing broad detection of all known ARGs, the approach is cost-prohibitive, limiting the number of samples that can be analyzed, as well as being only semi-quantitative. In addition, the lack of sufficient DNA from many samples in this study was a critical limitation to being able to fully compare reclaimed water samples to potable samples, which were much lower in biomass. This consequentially limited sample size for many of the categories (Figure 5-1) and correspondingly limited the ability to draw statistically significant conclusions about the data. The metagenomic analysis completed herein should be viewed as an exploratory characterization, particularly with respect to the potable water resistome.

90

10 9 multidrug other 8 trimethoprim tetracycline 7 sulfonamide streptothricin 6 streptogramin 5 rifampin polymyxin 4 peptide macrolide

log (gene log copies / mL) 3 lincosamide glycopeptide 2 fosfomycin 1 fluoroquinolone chloramphenicol beta-lactam aminoglycoside aminocoumarin

plasmids

Influent (n=1) Influent (n=1) Influent (n=1) Influent

Influent (n=1) Influent total

POE: water(n=1) POE: water(n=4) POE: water(n=4) POE: water(n=2) POE: water(n=3) POE:

POU: water (n=1) water POU: (n=1) water POU: (n=4) water POU: (n=4) water POU: (n=2) water POU: (n=4) water POU:

POU: biofilm (n=1) POU: biofilm (n=1) POU: biofilm (n=1) POU: biofilm (n=1) POU: Utility Utility B Utility A Utility B Utility C Utility D A Potable Reclaimed

Figure 5-1: Metagenomic characterization of ARGs by antibiotic class. Abundance of ARGs by antibiotic class (stacked bars) and plasmids (diamonds) per mL of water sampled or per biofilm swab as determined by annotation of reads from shotgun metagenomic sequencing of samples collected at the point-of-entry (POE) and point-of-use (POU) of four potable water utilities (A, B, C, D). Reads were annotated against the Comprehensive Antibiotic Resistance Database for ARGs and the ACLAME database for plasmids. Sufficient DNA for analysis of potable water samples was only possible at Utilities A and B, the remaining samples were from reclaimed water distribution systems. Error bars indicate standard deviation of total abundance of ARGs or plasmids when statistical power was sufficient (n value indicated for each sample category). This data is also presented as relative abundances normalized to 16S rRNA genes (Figure S6).

Abundance of target ARGs in water and biofilms

To precisely quantify a selection of ARGs corresponding to a range of critically and highly important classes, as defined by the World Health Organization,71 across the full dataset, qPCR was utilized. Important to note is that qPCR, as applied in this study, does not directly distinguish live and dead organisms or intracellular versus extracellular DNA. However, tracking numbers of ARGs through the water systems represents an indicator of net amplification and decay of the target genes, via horizontal transfer and/or growth or death and degradation, respectively.57 91

Several significant differences were noted in target gene numbers when comparing the paired reclaimed and potable distribution systems (Table 5-2). A consistent difference noted across all four utilities was that both 16S rRNA (a proxy for total bacterial cells) and sul1 gene copies per milliliter were more abundant in the reclaimed than potable distribution system samples (Wilcoxon; p≤0.0011). Further, blaTEM was more abundant in the reclaimed than the potable distribution system of Utility D (p<0.0001), and qnrA was greater in the reclaimed water of Utilities B, C, and D (p≤0.0145). vanA was not significantly different in any of the reclaimed versus potable distribution system waters (p≥0.0952). Where it was possible to collect biofilm samples, at Utilities A and B, 16S rRNA genes and sul1 gene copies per swab were all greater in the reclaimed compared to the potable biofilm at Utility A (p≤0.0017), while only sul1 was significantly elevated in the reclaimed biofilm for Utility B (p=0.0018).

sul1 was generally the most abundant ARG in reclaimed water at the POE, with 50–100% of samples positive, averaging from 4.0–5.3 log gene copies per milliliter. blaTEM was present in 14–75% of reclaimed POE samples, averaging from 1.3–1.9 log gene copies per milliliter. qnrA was only detected at the reclaimed POE of Utility D (75% positive; 1.3±0.1 log gene copies per milliliter, on average) while vanA was only detected at the reclaimed POE of Utility A (14% positive; 2.7±0.0 log gene copies per milliliter).

Although sul1 was nearly ubiquitous in reclaimed water, it was only detected at the POE in the potable water of Utility B, with 25% of samples positive at an average of 2.3±0.0 log gene copies per milliliter. qnrA was only detected at the potable POE of Utility D (25% positive; 1.1±0.0 log gene copies per milliliter, on average) and vanA was only detected at the potable POE of Utility B (50% positive; 2.0±0.6 log gene copies per milliliter, on average). blaTEM was detected at the potable POE of Utilities A (33% positive; 1.0±0.0 log gene copies per milliliter) and C (25% positive; 1.4±0.0 log gene copies per milliliter). Similar trends were observed when ARG abundances were normalized to 16S rRNA gene abundances (Table S4).

Interestingly, there were no significant differences between the abundance of ARGS (i.e., gene copies per milliliter) in the potable source water (e.g., reservoirs, groundwaters) and in the corresponding treated water at the POE (Wilcoxon; p≥0.1859). This suggests that ARGs essentially “pass-through” the drinking water treatment process, with the finished water representative of background natural waters. When comparing the raw wastewater influent to the reclaimed POE, reductions in some target genes were observed. 16S rRNA genes were decreased at Utilities A and B (p≤0.0298), and qnrA decreased at Utility B (p=0.0404). With these exceptions, treatment of reclaimed water did not consistently result in a significant reduction in ARGs on a unit volume basis, even when the water was collected post-disinfection. It is surprising to find that 16S rRNA genes were significantly reduced during treatment only at two of the utilities (A and B). Previous studies by Czekalski et al.72 and Mao et al.73 reported similar findings, with minuscule differences between 16S rRNA gene abundances in WWTP raw influent and final effluent. Czekalski reported that larger reductions in viable cell counts were observed using methods such as heterotrophic plate count,72 thus it is possible that 16S rRNA genes are present as extracellular DNA or associated with inactivated cells in final effluent. The lack of removal of ARGs during wastewater treatment adds to the body of evidence that typical biological treatment processes, and even media filtration, do not result in consistent broad-scale removal of ARGs.74– 76 In contrast, other studies have reported successful reductions in ARG abundances during

92

conventional wastewater treatment,14,74 so further research into the efficacy of various treatment processes for removing ARGs is needed. Advanced oxidative processes (AOPs), which employ hydroxyl radicals to more aggressively break down residual organic matter, are of interest and have shown promise for removal of both ARB and ARGs from wastewater.77,78 However, it is important to recognize that not all studies have indicated that AOPs achieve substantial removal of ARGs. In particular, effectiveness appears to vary widely based on the target microorganism and ARG studied,78,79 calling for further research to effectively optimize AOPs for the purpose of minimizing antibiotic resistance risk in reclaimed water.

With the exception of vanA, which decreased from the POE to the POU in Utility A’s reclaimed water (Wilcoxon; p=0.0338), abundances did not change significantly between the POE and the POU for any other ARGs measured by qPCR in potable or reclaimed water (p≥0.2021). Metagenomic analysis revealed that in the potable water, samples averaged 20.1 different ARGs per ten million reads at the POE and 30.2 different ARGs per ten million reads at the POU. In the reclaimed water, POE samples averaged 81.4 ARGs per ten million reads, but only 49.1 ARGs per ten million reads at the POU. In the reclaimed water, this represents a significant decrease in ARG diversity from the POE to the POU (p=0.0272). A previous study found that pipeline transport significantly increased beta-lactam ARGs in potable water at the tap.80 In a study of a reclaimed water distribution system, Fahrenfeld et al. found that the ARGs sul1 and tet(A) were more abundant at the POU than at the POE, while no difference was observed for vanA, ermF, sul2, and tet(O), suggesting that regrowth can occur from the POE to the POU for some ARGS in some reclaimed water distribution systems.19 While the results of the current study did not suggest significant regrowth of ARG-carrying bacteria in the surveyed potable and reclaimed distribution systems, ARGs did generally persist following distribution system transport in both potable and reclaimed systems.

Associations between ARG abundance and microbial ecological factors

Canonical correspondence analysis (Figure S3) indicated that all five target genes had significant, though weak, associations with the overall bacterial community composition (Adonis, R2=0.1424-0.1734; p≤0.001). The target ARGs were each significantly correlated with several phyla and Proteobacteria classes. The strongest correlations between ARGs and various bacterial phyla (or Proteobacteria classes) were between blaTEM and Fibrobacteres, WWE1, Synergistetes, SR1, OD1, and Euryarchaeota (Spearman; ρ=0.2595-0.2767; p<0.0001) qnrA and SR1, TM7, Lentisphaerae, WWE1, Synergistetes, and Fibrobacteres (ρ=0.3797-0.5044; p<0.0001); sul1 and NKB19, WPS-2, TA18 Proteobacteria, TM7, TM6, and (ρ=0.2917-0.4890; p<0.0001); and between vanA and Actinobacteria, Cyanobacteria, Proteobacteria (unclassified at the class level), GN04, Firmicutes, and AncK6 (ρ=0.1226-0.1797; p≤0.0415). The presence of statistically significant correlations between the bacterial community composition and ARGs indicates that at least some of the shifts in ARG abundances noted within the distribution systems may actually be indicative of shifts in numbers and types of bacteria carrying them. In other words, vertical transfer of ARGs via relative selection pressures on existing bacteria carrying these genes is possibly a key contributor to the patterns observed, as reported by others.81,82 Previous studies have demonstrated that ARG abundances are associated with community phylogenetic composition in many environments, including soil,82,83 the human gut,83 and sewage sludge.84

93

Table 5-2: Frequency of qPCR detection and abundance of ARGs. 16S rRNA and ARGs in potable and reclaimed water (log10 gene copies per milliliter) and biofilm (log10 gene copies per swab) samples in the influent or source water, at the point of entry (POE), and point of use (POU); average (standard deviation) of values above the limit of quantification. Values for each sampling location include four sampling events. Biofilm swab samples were collected by the same individual at each utility and in the same manner, so comparisons within one utility may be made, but biofilm abundances should be compared across utilities with caution. Relative abundances (normalized to 16S rRNA genes) are provided in Table S4.

16S rRNA blaTEM intI1 qnrA vanA sul1 Influent / Source Potable A (n=6) 100%; 3.6(0.8) 17%; 1.8(0.0) ND ND 33%; 1.5(0.4) ND B (n=4) 100%; 3.8(0.5) 25%; 2.2(0.0) ND ND 25%; 1.2(0.0) 25%; 1.2(0.0) C (n=9) 100%; 3.7(1.5) 11%; 1.2(0.0) 11%; 4.2(0.0) ND ND 22%; 1.7(0.8) D (n=3) 100%; 4.2(2.1) 33%; 2.8(0.0) 33%; 6.2(0.0) 67%; 1.9(0.5) ND 33%; 5.8(0.0) Reclaimed A (n=6) 100%; 7.3(0.3) 50%; 2.4(0.9) 67%; 5.7(1.2) 33%; 3.0(0.2) ND 50%; 4.9(1.4) B (n=7) 100%; 7.5(1.5) 71%; 3.2(0.4) 71%; 6.7(0.4) 71%; 2.6(0.5) ND 71%; 6.5(0.2) C (n=4) 100%; 6.9(0.4) 75%; 1.6(0.1) 50%; 6.8(2.5) 50%; 1.4(0.1) ND 75%; 5.8(1.2) D (n=3) 100%; 7.2(0.2) 100%; 2.9(0.3) 100%; 6(0.5) 100%; 2.9(0.5) ND 100%; 5.8(0.4) POE - water Potable A (n=3) 100%; 4.6(0.5) 33%; 1.0(0.0) 33%; 5.6(0.0) ND ND ND B (n=4) 100%; 4.6(2.2) ND 25%; 4.0(0.0) ND 50%; 2.0(0.6) 25%; 2.3(0.0) C (n=4) 100%; 2.6(0.7) 25%; 1.4(0.0) 25%; 4.3(0.0) ND ND ND D (n=4) 100%; 2.8(0.8) ND ND 25%; 1.1(0.0) ND ND Reclaimed A (n=7) 100%; 5.7(0.7) 14%; 1.9(0.0) 57%; 5.6(1.2) ND 14%; 2.7(0.0) 100%; 5.0(1.0) B (n=7) 100%; 5.5(1.0) 43%; 1.5(0.5) 29%; 5.0(2.1) ND ND 100%; 4.0(1.0) C (n=4) 100%; 5.4(2.1) 50%; 1.3(0.1) 50%; 6.6(0.2) ND ND 50%; 5.3(0.1) D (n=4) 100%; 5.2(1.8) 75%; 1.5(0.3) 25%; 5.2(0.0) 75%; 1.3(0.1) ND 100%; 4.1(1.4) POU - water Potable A (n=44) 100%; 4.2(0.9) 23%; 1.8(0.8) 25%; 4.7(0.9) 2%; 1.2(0.0) 11%; 1.4(0.3) 18%; 2.7(2.0) B (n=21) 100%; 3.3(1.0) 19%; 1.7(0.7) 19%; 4.9(0.5) ND 14%; 1.9(1.0) 19%; 1.7(0.7) C (n=15) 95%; 2.3(0.6) ND ND ND ND 47%; 1.6(0.5) D (n=20) 85%; 2.4(0.6) ND ND ND ND 5%; 1.2(0.0) Reclaimed A (n=38) 100%; 6.4(0.4) 18%; 1.8(0.8) 71%; 5.7(0.8) ND ND 92%; 5.5(0.8) B (n=17) 100%; 5.8(1.3) 12%; 1.6(0.4) 53%; 5.8(0.6) ND 6%; 1.5(0.0) 88%; 4.8(1.2) C (n=17) 100%; 5.7(2.0) 18%; 1.9(1.1) 53%; 5.1(1.5) 6%; 1.5(0.0) ND 59%; 5.2(2.0) D (n=20) 100%; 5.7(0.8) 80%; 1.5(0.4) 55%; 4.8(0.3) 45%; 1.2(0.2) ND 100%; 4.6(0.9) POU – biofilm Potable A (n=40) 100%; 4.2(0.7) 18%; 1.5(0.4) 20%; 4.2(0.9) ND 5%; 1.5(0.5) 15%; 3.2(1.5) B (n=21) 100%; 3.5(0.8) 10%; 1.6(0.7) ND 5%; 2.1(0.0) 10%; 1.3(0.1) 24%; 2.1(1.6) Reclaimed A (n=33) 100%; 6.0(0.4) 6%; 1.8(1.0) 55%; 4.5(0.8) 3%; 1.6(0.0) ND 85%; 4.6(1.0) B (n=15) 100%; 4.2(0.5) 13%; 1.4(0.1) 20%; 3.5(2.7) 7%; 1.2(0.0) 20%; 1.4(0.1) 80%; 2.7(1.0)

In the absence of ARGs, most sulfonamides, beta-lactams, and fluoroquinolones are effective against both Gram-positive and -negative bacteria, while vancomycin is primarily only effective against Gram-positive bacteria. To the extent that information was available, phyla were sorted by Gram stain type and correlations with ARGs were examined. blaTEM correlated negatively with the abundance of Gram-negative bacteria (Spearman; ρ=-0.1562, p=0.0092). qnrA correlated positively with Gram-positive bacteria (ρ=0.2450, p<0.0001) and negatively with 94

Gram-negative bacteria (ρ=-0.2340, p<0.0001). sul1 correlated negatively with Gram-positive bacteria (ρ=-0.1919, p=0.0013) and positively with Gram-negative bacteria (ρ=0.1747, p=0.0035). vanA correlated positively with Gram-positive bacteria (ρ=0.1807, p=0.0025) and negatively with Gram-negative bacteria (ρ=-0.1353, p=0.0243), which is consistent with the activity of vancomycin against Gram-positive bacteria. While vertical transfer of ARGs appears to be an important factor, the identification of only weak correlations (Adonis R2≤0.1734; Spearman |ρ|≤0.5044) between ARGs and overall bacterial community composition, occurrence of individual phyla, and Gram-stain type indicates that vertical transfer explains only a portion of the variation in ARGs documented in this study. The presence of weak correlations indicates that it is possible that environmental bacteria, such as those inhabiting reclaimed and potable distribution systems, may serve as potential reservoirs of resistance that can become mobilized to pathogenic bacteria. This possible mobilization of ARGs from innocuous to pathogenic bacteria has the potential to occur at several points: 1) in the distribution system, 2) following irrigation or other industrial application, or 3) following human exposure and potential colonization by these commensal bacteria. Further research is needed to confirm that such mobilization occurs in each of these compartments and determine corresponding rates.

Potential for horizontal gene transfer

As was the case in this study, it is commonly observed that the ARG profile correlates with phylogenetic composition. However, this does not necessarily explain the full extent of factors influencing abundance of ARGs.82–84 Horizontal gene transfer has been established as an important mechanism for the dissemination of ARGs between species and particularly in the mobilization of ARGs from non-pathogenic environmental flora to pathogens.85–88

The class 1 integron-integrase gene, intI1, is important for gene acquisition as it allows capture of exogenous genes into a cell’s genome and subsequent expression. The gene also has a tendency to acquire a wide range of gene cassettes, including several ARGs, and is well suited for horizontal gene transfer among a variety of environmental and pathogenic organisms.89 intI1 was more abundant in the reclaimed than potable water of Utilities A, C, and D (Wilcoxon; p<0.0001) as well as in the reclaimed compared to the potable biofilm at Utility A (p=0.0017). The ubiquity of intI1 in reclaimed water compared to potable water (Table 5-2) suggests that reclaimed water may be a particularly rich environment for horizontal gene transfer and incorporation of transferred genes into a cell’s genome. The abundance of intI1 in reclaimed biofilms further indicates that their high cellular density may be well suited for horizontal gene transfer. Correlations between sul1 and intI1 were examined, given that sul1 is commonly found on intI1 gene cassettes.89 In potable water, these two gene were correlated in biofilm (Spearman; ρ=0.3061, p=0.0164), but not in water (ρ=0.0502, p=0.6165). In reclaimed water, the genes were correlated in both the water (ρ=0.6422, p<0.0001) and biofilm (ρ=0.4991, p=0.0003). intI1 was also correlated with blaTEM in potable water (ρ=0.5038, p<0.0001) and biofilm (ρ=0.3485, p=0.0059), and reclaimed water (ρ=0.3597, p<0.0001) and biofilm (ρ=0.3935, p=0.0051). intI1 was only correlated with qnrA in reclaimed water (ρ=0.3018, p=0.0012).

Potential for conjugation (i.e., transfer of plasmid DNA via cell-to-cell contact) was explored by identifying genes associated with plasmids via annotation of metagenomic sequence reads against the ACLAME plasmid database. Owing to the unique ecosystem created by the

95

biofilm environment and a high level of contact between microorganisms, biofilms have been documented to facilitate relatively high rates of gene transfer.90 While biofilm samples tended to have relatively high abundances of plasmid-associated genes in the present study (Figure 5-1), the abundances were not significantly different from those of the water for the potable (Wilcoxon; p=0.7728) or the reclaimed (p=0.1412) water systems. There were no significant differences in abundance of metagenomic reads annotated as plasmid associated genes between potable and reclaimed water (p=0.1144), or among utilities for reclaimed water (p≥0.2301). For reclaimed water samples, the abundance of plasmid-associated genes detected at the POU was greater than at the POE (p=0.0004), suggesting that conjugation or selection for plasmid-carrying bacteria may be occurring within the distribution system.

Network analysis of de novo assembled scaffolds derived from metagenomics sequence data was conducted to identify ARGs associated with plasmid gene markers as well as to examine associations between ARGs co-located on the same DNA strand (Figure 5-2, Figure S4). In particular, ARGs present on the same mobile element or DNA strand may be subject to co- selection. Across all samples analyzed, 193 different ARGs were found to be associated with plasmid markers. Those with at least three instances of co-location with plasmid gene markers are highlighted in Figure 5-2. The ARGs most frequently associated with plasmid scaffolds included several multidrug ARGs; acrE, acrA, adeF, mexB, mexK, sav1866, and ceoA (684, 156, 146, 135, 128, 109, and 100 scaffolds, respectively), the trimethoprim ARG dfrE (339 scaffolds), the aminoglycoside ARG amrA (222 scaffolds), and the cationic antibiotic ARG rosA (198 scaffolds). Other notable ARGs that mapped to plasmid scaffolds were sul1 (34 scaffolds), NDM-13 (18 scaffolds), and several extended spectrum beta-lactamases: CTX-M-9 (1 scaffold), PER-7 (1 scaffold), GES-5 (2 scaffolds), 2 SHV gene variants (2 scaffolds), 4 TEM gene variants (7 scaffolds), and 12 OXA gene variants (18 scaffolds). Detection of sul1 and blaTEM occurring together with plasmid-associated genes and correlated with intI1, as noted above, suggests that these genes might be particularly noteworthy candidates as indicators of horizontal gene transfer potential. Detection of the New Delhi metallo-beta-lactamase, NDM-13, on several plasmid scaffolds is also particularly concerning as it is an exemplar of an ARG that is both mobile and encodes resistance to several critically-important antibiotics, including nearly all beta-lactam antibiotics.91

Although horizontal gene transfer rates in the environment are thought to be low,37,92 even rare transfer incidence of ARGs that are of concern to public health, such as NDM-13, merit further consideration in terms of beginning to better understand and predict how occurrence of ARGs may translate to actual risk. Little is known about rates of horizontal gene transfer in wastewater or aquatic environments, and to the authors’ knowledge, no study has quantified these rates specifically in the reclaimed water environment.

96

Figure 5-2: Network analysis depicting co-occurrence of ARGs among each other as well as with plasmid gene markers on assembled scaffolds. Constructed using de novo assembly of shotgun metagenomic sequencing reads from all samples. This analysis highlights ARGs that are the probable candidates for potential horizontal gene transfer (co-occurrences with plasmid associated genes) or co-selection (co-occurrences with ARGs of different classes). Proximity of nodes and width of lines is proportional to numbers of association between genes. Node diameter is proportional to the number of co-occurrences for that gene. A minimum of three co-occurrences were required for inclusion in the graphic. All ARG nodes depicted share a connection with the plasmid node, with the exception of EreA2, indicated by an asterisk. MLS indicates resistance to macrolide, lincosamide, and streptogramin antibiotics.This analysis has also been provided for reclaimed and potable samples individually (Figure S4).

Associations between water chemistry and ARGs

This study comparing ARG occurrence in reclaimed and potable water systems presents an opportunity to examine how ARG occurrence may be shaped by water chemistry. Here we observe that water chemistry can influence the resistome via several mechanisms. For example, nutrients, pH, dissolved oxygen, temperature, disinfectant, and other physicochemical parameters

97

may select the type of bacterial community inhabiting water systems, including taxonomic structure and potential for horizontal gene transfer. In addition, certain agents, such as antimicrobials, disinfectants, and heavy metals, may be present that directly select for bacteria carrying ARGs. Many of these agents have even been shown to select for ARGs or stimulate horizontal gene transfer at sub-inhibitory concentrations, thus even the presence of low abundances of these selective agents merits consideration of their potential to influence the resistome.93–103

Water chemistry data is presented in Table S5. Total chlorine was negatively correlated with sul1 (Spearman; ρ=-0.3117, p=0.0003) and intI1 (ρ=-0.3030, p=0.0007) in reclaimed water, but positively correlated with blaTEM (ρ=0.2799, p=0.0165) and qnrA (ρ=0.2410, p=0.0400) in the potable water. This suggests that the relatively higher levels of chlorine maintained in the potable water may have actually selected for bacteria carrying blaTEM and qnrA. Previous studies have demonstrated that chlorine can sometimes select for ARG-carrying bacteria. Huang et al. found that inactivation of E. coli carrying the tetracycline ARG tetA using chlorine was significantly lower than tetracycline-sensitive E. coli.31 Karumathil et al. found that exposure of Acinetobacter baumannii to chlorine resulted in up-regulation of ARGs.30 Alternatively, maintaining a total chlorine residual above 1.3 mg/L appears to have aided in the control of sul1 and intI1 in the reclaimed water. Several heavy metals that could potentially co-select ARGs located on the same genetic element or be subject to cross-resistance (i.e., same gene and corresponding cellular function combats multiple chemicals, such as multidrug efflux pumps) or co-regulation (i.e., two resistance regulation systems that are transcriptionally linked) were correlated with ARGs in reclaimed water in the distribution system. Iron, copper, zinc, manganese, silver, and cobalt were all positively correlated with qnrA (ρ=0.2546-0.4449; p≤0.0023). Cobalt, silver, and manganese were correlated with blaTEM (ρ=0.3392-0.4184; p≤0.0018), while cobalt, arsenic, silver, and manganese were correlated with sul1 (ρ=0.2489-0.3429; p≤0.0208). Forty-seven different ARGs were identified from the data set that were co-located on de novo assembled scaffolds together with metal resistance genes (Figure S5) and therefore identified to be candidates for co-selection. The most frequent associations were between acrE and zinc resistance genes (421 scaffolds), acrA and mexB with genes conferring resistance to copper and zinc (122, 99 scaffolds, respectively), amrA and gold resistance genes (119 scaffolds), and AAC(6’)-Iaa, adeF, and ceoA with gold resistance genes (90, 67, and 49 scaffolds, respectively).

A subset of potable source water, raw wastewater influent, and treated potable and reclaimed POE samples were also screened for the presence of 14 antimicrobials using UPLC- MS/MS. While 12 of the antibiotics were not detectable in any samples, vancomycin and trimethoprim were both detectable in the wastewater influent of all utilities, with the exception of Utility C, where only trimethoprim was detectable (Table S6). Vancomycin was widely removed during treatment, with the antibiotic only detectable in the treated effluent of one of the Utility B plants. Trimethoprim was detectable in 50% of reclaimed plant effluents. Surprisingly, both antibiotics were detectable in the treated potable water of Utility B and trimethoprim was detectable in that of Utility C. In samples where antibiotic concentrations were quantifiable, trimethoprim was correlated with abundances of sul1, blaTEM, and qnrA (ρ=0.5991-0.7922, p≤0.0121) and vancomycin was correlated with sul1, intI1, qnrA, and blaTEM (ρ=0.5894-8241, p≤0.0128). These results suggest that there is potential for co-selection, cross-selection, or co- regulation to select for the presence of ARGs by antibiotics of different classes in the studied systems, though further research is needed to understand this phenomenon further.

98

Little is known about how other “non-antimicrobial” water chemistry parameters might affect ARGs in water systems. In particular, organic carbon has been identified as a key parameter for controlling the re-growth of bacteria in distribution systems.104,105 BDOC was positively correlated with both blaTEM (ρ=0.3075, p=0.0023) and qnrA (ρ=0.3874, p<0.0001) in reclaimed water, but was not correlated with any ARGs in potable water. This suggests that limiting BDOC may be an important factor in controlling certain ARGs in distribution systems. Similarly, another key nutrient for bacterial growth, phosphorus, was correlated with both blaTEM (ρ=0.1989-0.2977, p≤0.0150) and qnrA (ρ=0.2423-0.2948, p≤0.0035) in both potable and reclaimed water. Together these results suggest that limiting key growth nutrients may be an important factor for limiting propagation of some clinically-relevant ARGs. Since not all ARGs were similarly affected by nutrient availability, it is likely that blaTEM and qnrA are more frequently associated with bacteria that lack the oligotrophic advantage typical of bacteria that are known to prosper in potable and highly treated reclaimed water systems.

Implications for ARG dissemination via reclaimed water

This study provides a head-to-head comparison of ARGs in full-scale reclaimed water distribution systems with corresponding potable water systems operated in the same communities in the United States. Presently there is no means to directly translate ARG numbers to human health risks, such as increased likelihood of acquiring an antibiotic resistant infection. Traditional pathogen risk models do not factor in the dimension of microbes sharing ARGs or consider antibiotic resistant infections as a treatment outcome.25 A significant challenge is that ARGs exist naturally in the background of any aquatic system, with further research needed to identify ARG targets and levels that truly pose a health risk. Profiling the occurrence of ARGs in reclaimed water systems and benchmarking them relative to corresponding potable water systems helps to address a key question: Do reclaimed water systems pose any greater risk than traditional potable water systems in terms of their potential to spread antibiotic resistance? If the resistome of both reclaimed and potable systems are comparable, then this shifts attention to broader ARG management strategies that also encompass potable water, such as the “One Water” concept.106

sul1 was consistently elevated in reclaimed water compared to potable water, while blaTEM, qnrA, and intI1 were each elevated in the reclaimed water of some utilities, compared to the corresponding potable water samples. While wastewater treatment at Utility B reduced qnrA abundances, no other ARGs were significantly removed at any treatment plants, consistent with the notion that WWTPs are not generally designed to minimize antibiotic resistance levels and that this has implications for water reuse. Further, association of 193 different ARGs mapped with plasmid-associated genes suggests that there is potential for proliferation of ARGs via horizontal gene transfer. Correlations between several ARGs and antibiotics, metals, and disinfectant concentrations in distribution systems suggest that selection pressures are worthy of consideration in reclaimed water. However, in contrast with a prior study of a single reclaimed distribution system,19 this more comprehensive analysis of multiple distribution systems over several time points did not indicate a significant trend towards increasing abundance from POE to POU of any of the ARGs analyzed. Overall, while there were indications that horizontal gene transfer and selection of resistant bacteria could occur in the distribution system, this did not seem to result in a net increase in total ARGs (Figure 5-1) or the ARGs specifically targeted by qPCR (Table 5-2) at the POU of the surveyed systems. Further, while a general pattern in reduction of diversity of

99

ARGs during distribution was noted, total abundances did not change significantly. Further research is needed via controlled laboratory conditions to determine quantitatively the relative contribution of selection by antibiotics and other compounds, horizontal gene transfer, microbial community structure, and water chemistry in shaping the resistome of reclaimed water systems.

Overall, this study demonstrates that reclaimed water is distinct in its ARG content, or “resistome,” relative to corresponding potable waters. When statistical differences were noted for various target genes, levels were higher in samples from the reclaimed water systems. Indicators of gene transfer, including plasmids and integrons, were also more prevalent in the reclaimed systems. Altogether, the results indicate that further research is warranted to determine to what extent, if any, the distinct resistome of reclaimed water may translate to human health risk. Exposure routes relevant to non-potable water worthy of consideration include dermal contact resulting from uses such as irrigation of athletic and recreation facilities and snowmaking, and inhalation resulting from use of reclaimed water in cooling towers, spray irrigation, toilet flushing, and fire suppression.107 Additional research is needed to characterize the extent to which human exposures to reclaimed water are associated with the transmission of resistant commensal or pathogenic microorganisms. It is also important to acknowledge that this study was molecular- based and it cannot be discerned precisely which ARGs were present in viable microbial hosts. Given the potential for naked DNA to be taken up downstream via transformation, the precautionary principle advises to continue to seek economical water treatment and management options that minimize the levels of ARGs, particularly clinically-relevant ARGs subject to horizontal gene transfer.

ACKNOWLEDGEMENTS

We thank the participating utilities for conducting sampling and on-site data collection. This work is supported by the National Science Foundation (NSF) Graduate Research Fellowship Program and NSF Collaborative Research grant (CBET 1438328) and Partnership in International Research and Education (OISE 1545756), The Alfred P. Sloan Foundation Microbiology of the Built Environment program, the Water Environment & Research Foundation Paul L. Busch award, the Virginia Tech Institute for Critical Technology and Applied Science Center for Science and Engineering of the Exposome, and the American Water Works Association Abel Wolman Doctoral Fellowship.

REFERENCES

(1) Guidelines for Water Reuse; United States Environmental Protection Agency: 2012. (2) Jjemba, P. K.; Weinrich, L. A.; Cheng, W.; Giraldo, E.; LeChevallier, M. W. Guidance document on the microbiological quality and biostability of reclaimed water and distribution systems (WRF‐05‐002), WateReuse Research Foundation: Alexandria, VA, 2010. (3) Rosario, K.; Nilsson, C.; Lim, Y. W.; Ruan, Y.; Breitbart, M. Metagenomic analysis of viruses in reclaimed water. Environ. Microbiol. 2009, 11 (11), 2806–2820. (4) Pruden, A. Balancing water sustainability and public health goals in the face of growing concerns about antibiotic resistance. Environ. Sci. Technol. 2014, 48 (1), 5–14. (5) Hong, P.-Y.; Al-Jassim, N.; Ansari, M.; Mackie, R. Environmental and Public Health

100

Implications of Water Reuse: Antibiotics, Antibiotic Resistant Bacteria, and Antibiotic Resistance Genes. Antibiotics 2013, 2 (3), 367–399. (6) Antibiotic Resistance Threats in the United States; U.S. Department of Health and Human Services Centers for Disease Control and Prevention: Atlanta, GA, 2013. (7) Global action plan on antimicrobial resistance; World Health Organization: Geneva, Switzerland, 2015. (8) A European One Health Action Plan against Antimicrobial Resistance; European Commission, 2017. (9) National Action Plan for Combating Antibiotic-Resistant Bacteria; U.S. White House: Washington, DC, 2015. (10) National Action Plan on Antimicrobial Resistance; Government of India, 2017. (11) Zhu, B. Y.; Gillings, M.; Simonet, P.; Stekel, D.; Banwart, S.; Penuelas, J. Microbial mass movements. Science 2017, 357 (6356), 1099–1100. (12) Allen, H. K.; Donato, J.; Wang, H. H.; Cloud-Hansen, K. A; Davies, J.; Handelsman, J. Call of the wild: antibiotic resistance genes in natural environments. Nat. Rev. Microbiol. 2010, 8 (4), 251–259. (13) Munir, M.; Wong, K.; Xagoraraki, I. Release of antibiotic resistant bacteria and genes in the effluent and biosolids of five wastewater utilities in Michigan. Water Res. 2011, 45 (2), 681–693. (14) Laht, M.; Karkman, A.; Voolaid, V.; Ritz, C.; Tenson, T.; Virta, M.; Kisand, V. Abundances of tetracycline, sulphonamide and beta-lactam antibiotic resistance genes in conventional wastewater treatment plants (WWTPs) with different waste load. PLoS One 2014, 9 (8), e103705. (15) Rizzo, L.; Manaia, C.; Merlin, C.; Schwartz, T.; Dagot, C.; Ploy, M. C.; Michael, I.; Fatta- Kassinos, D. Urban wastewater treatment plants as hotspots for antibiotic resistant bacteria and genes spread into the environment: a review. Sci. Total Environ. 2013, 447, 345–360. (16) Auerbach, E. A.; Seyfried, E. E.; McMahon, K. D. Tetracycline resistance genes in activated sludge wastewater treatment plants. Water Res. 2007, 41 (5), 1143–1151. (17) LaPara, T. M.; Burch, T. R.; McNamara, P. J.; Tan, D. T.; Yan, M.; Eichmiller, J. J. Tertiary- treated municipal wastewater is a significant point source of antibiotic resistance genes into Duluth-Superior Harbor. Environ. Sci. Technol. 2011, 45 (22), 9543–9549. (18) Szczepanowski, R.; Linke, B.; Krahn, I.; Gartemann, K. H.; Gutzkow, T.; Eichler, W.; Puhler, A.; Schluter, A. Detection of 140 clinically relevant antibiotic-resistance genes in the plasmid metagenome of wastewater treatment plant bacteria showing reduced susceptibility to selected antibiotics. Microbiology 2009, 155, 2306–2319. (19) Fahrenfeld, N. L.; Ma, Y. J.; O’Brien, M.; Pruden, A. Reclaimed water as a reservoir of antibiotic resistance genes: distribution system and irrigation implications. Front. Microbiol. 2013, 4 (130). (20) Barker-Reid, F.; Fox, E. M.; Faggian, R. Occurrence of antibiotic resistance genes in reclaimed water and river water in the Werribee Basin, Australia. J. Water Health 2010, 8 (3), 521–531. (21) Goldstein, R. E. R.; Micallef, S. A.; Gibbs, S. G.; George, A.; Claye, E.; Sapkota, A.; Joseph, S. W.; Sapkota, A. R. Detection of vancomycin-resistant enterococci (VRE) at four US wastewater treatment plants that provide effluent for reuse. Sci. Total Environ. 2014, 466–467, 404–411. (22) Perron, G. G.; Whyte, L.; Turnbaugh, P. J.; Goordial, J.; Hanage, W. P.; Dantas, G.; Desai,

101

M. M. Functional Characterization of Bacteria Isolated from Ancient Arctic Soil Exposes Diverse Resistance Mechanisms to Modern Antibiotics. PLoS One 2015, 10 (3), e0069533. (23) Mindlin, S. Z.; Soina, V. S.; Petrova, M. A.; Gorlenko, Z. M. Isolation of antibiotic resistance bacterial strains from Eastern Siberia permafrost sediments. Russ. J. Genet. 2008, 44 (1), 27–34. (24) Bhullar, K.; Waglechner, N.; Pawlowski, A.; Koteva, K.; Banks, E. D.; Johnston, M. D.; Barton, H. a; Wright, G. D. Antibiotic resistance is prevalent in an isolated cave microbiome. PLoS One 2012, 7 (4), e34953. (25) Ashbolt, N. J.; Amezquita, A.; Backhaus, T.; Borriello, P.; Brandt, K. K.; Collignon, P.; Coors, A.; Finley, R.; Gaze, W. H.; Heberer, T.; et al. Human Health Risk Assessment (HHRA) for Environmental Development and Transfer of Antibiotic Resistance. Environ. Health Perspect. 2013, 121 (9), 993–1001. (26) Martínez, J. L.; Coque, T. M.; Baquero, F. What is a resistance gene ? Ranking risk in resistomes. Nat. Rev. Microbiol. 2014, 13 (2), 116–123. (27) Zhang, Y. L.; Marrs, C. F.; Simon, C.; Xi, C. W. Wastewater treatment contributes to selective increase of antibiotic resistance among Acinetobacter spp. Sci. Total Environ. 2009, 407 (12), 3702–3706. (28) Schlüter, A.; Szczepanowski, R.; Pühler, A.; Top, E. M. Genomics of IncP-1 antibiotic resistance plasmids isolated from wastewater treatment plants provides evidence for a widely accessible drug resistance gene pool. FEMS Microbiol. Rev. 2007, 31 (4), 449–477. (29) Czekalski, N.; Gascón Díez, E.; Bürgmann, H. Wastewater as a point source of antibiotic- resistance genes in the sediment of a freshwater lake. ISME J. 2014, 8, 1381–1390. (30) Karumathil, D. P.; Yin, H.-B.; Kollanoor-Johny, A.; Venkitanarayanan, K. Effect of chlorine exposure on the survival and antibiotic gene expression of multidrug resistant Acinetobacter baumannii in water. Int. J. Environ. Res. Public Health 2014, 11 (2), 1844– 1854. (31) Huang, J.-J.; Hu, H.-Y.; Wu, Y.-H.; Wei, B.; Lu, Y. Effect of chlorination and ultraviolet disinfection on tetA-mediated tetracycline resistance of Escherichia coli. Chemosphere 2013, 90 (8), 2247–2253. (32) Kurenbach, B.; Marjoshi, D.; Amábile-cuevas, C. F.; Ferguson, G. C.; Godsoe, W.; Gibson, P.; Heinemann, J. a. Sublethal Exposure to Commercial Formulations of the Herbicides Changes in Antibiotic Susceptibility in Escherichia coli and Salmonella enterica serovar Typhimurium. mBio 2015, 6 (2), e00009-15. (33) Baker-Austin, C.; Wright, M. S.; Stepanauskas, R.; McArthur, J. V. Co-selection of antibiotic and metal resistance. Trends Microbiol. 2006, 14 (4), 176–182. (34) Yang, Y.; Li, B.; Zou, S.; Fang, H. H. P.; Zhang, T. Fate of antibiotic resistance genes in sewage treatment plant revealed by metagenomic approach. Water Res. 2014, 62, 97–106. (35) Yang, Y.; Li, B.; Ju, F.; Zhang, T. Exploring Variation of Antibiotic Resistance Genes in Activated Sludge over a Four-Year Period through a Metagenomic Approach. Environ. Sci. Technol. 2013, 47 (18), 10197–10205. (36) Zhang, T.; Zhang, X. X.; Ye, L. Plasmid Metagenome Reveals High Levels of Antibiotic Resistance Genes and Mobile Genetic Elements in Activated Sludge. PLoS One 2011, 6 (10), e26041. (37) Kim, S.; Yun, Z.; Ha, U.-H.; Lee, S.; Park, H.; Kwon, E. E.; Cho, Y.; Choung, S.; Oh, J.; Medriano, C. A.; et al. Transfer of antibiotic resistance plasmids in pure and activated sludge cultures in the presence of environmentally representative micro-contaminant

102

concentrations. Sci. Total Environ. 2014, 468–469, 813–820. (38) Wang, F. H.; Qiao, M.; Lv, Z. E.; Guo, G. X.; Jia, Y.; Su, Y. H.; Zhu, Y. G. Impact of reclaimed water irrigation on antibiotic resistance in public parks, Beijing, China. Environ. Pollut. 2014, 184, 247–253. (39) Wang, F.; Qiao, M.; Su, J.; Chen, Z.; Zhou, X.; Zhu, Y. High Throughput Profiling of Antibiotic Resistance Genes in Urban Park Soils with Reclaimed Water Irrigation. Environ. Sci. Technol. 2014, 48, 9079-9085. (40) Han, X. M.; Hu, H. W.; Shi, X. Z.; Wang, J. T.; Han, L. L.; Chen, D.; He, J. Z. Impacts of reclaimed water irrigation on soil antibiotic resistome in urban parks of Victoria, Australia. Environ. Pollut. 2016, 211, 48–57. (41) McLain, J. E.; Williams, C. F. Sustainability of Water Reclamation: Long-Term Recharge with Reclaimed Wastewater Does Not Enhance Antibiotic Resistance in Sediment Bacteria. Sustainability 2014, 6 (3), 1313–1327. (42) Bockelmann, U.; Dorries, H. H.; Ayuso-Gabella, M. N.; de Marcay, M. S.; Tandoi, V.; Levantesi, C.; Masciopinto, C.; Van Houtte, E.; Szewzyk, U.; Wintgens, T.; et al. Quantitative PCR Monitoring of Antibiotic Resistance Genes and Bacterial Pathogens in Three European Artificial Groundwater Recharge Systems. Appl. Environ. Microbiol. 2009, 75 (1), 154–163. (43) Ryu, H.; Alum, A.; Abbaszadegan, M. Microbial Characterization and Population Changes in Nonpotable Reclaimed Water Distribution Systems. Environ. Sci. Technol. 2005, 39 (22), 8600–8605. (44) Jjemba, P. K.; Weinrich, L. a.; Cheng, W.; Giraldo, E.; LeChevallier, M. W. Regrowth of potential opportunistic pathogens and algae in reclaimed-water distribution systems. Appl. Environ. Microbiol. 2010, 76 (13), 4169–4178. (45) Wingender, J.; Flemming, H.-C. Biofilms in drinking water and their role as reservoir for pathogens. Int. J. Hyg. Environ. Health 2011, 214 (6), 417–423. (46) Hausner, M.; Wuertz, S. High rates of conjugation in bacterial biofilms as determined by quantitative in situ analysis. Appl. Environ. Microbiol. 1999, 65 (8), 3710–3713. (47) Costa, V. M. D.; Mcgrann, K. M.; Hughes, D. W.; Gerard, D.; Science, S.; Series, N.; Jan, N. Sampling the Antibiotic Resistome. Science 2014, 311, 374–377. (48) American Public Health Association (APHA). Standard Methods for the Examination of Water and Wastewater; American Water Works Association, Water Environment Federation, 2005. (49) Servais, P.; A. Anzil; Ventresque, C. Simple method for determination of biodegradable dissolved organic carbon in water. Appl. Environ. Microbiol. 1989, 55 (10), 2732–2734. (50) Sui, Q.; Huang, J.; Deng, S.; Chen, W.; Yu, G. Seasonal Variation in the Occurrence and Removal of Pharmaceuticals and Personal Care Products in Different Biological Wastewater Treatment Processes. Environ. Sci. Technol. 2011, 45, 3341–3348. (51) Suzuki, M. T.; Taylor, L. T.; DeLong, E. F. Quantitative analysis of small-subunit rRNA genes in mixed microbial populations via 5’-nuclease assays. Appl. Environ. Microbiol. 2000, 66 (11), 4605–4614. (52) Bibbal, D.; Dupouy, V.; Ferre, J. P.; Toutain, P. L.; Fayet, O.; Prere, M. F.; Bousquet- Melou, A. Impact of three ampicillin dosage regimens on selection of ampicillin resistance in and excretion of bla(TEM) genes in swine feces. Appl. Environ. Microbiol. 2007, 73 (15), 4785–4790. (53) Colomer-Lluch, M.; Jofre, J.; Muniesa, M. Quinolone resistance genes (qnrA and qnrS) in

103

bacteriophage particles from wastewater samples and the effect of inducing agents on packaged antibiotic resistance genes. J. Antimicrob. Chemother. 2014, 69 (5), 1265–1274. (54) Dutka-Malen, S.; Evers, S.; Courvalin, P. Detection of glycopeptide resistance genotypes and identification to the species level of clinically relevant enterococci by PCR. J. Clin. Microbiol. 1995, 33 (1), 24–27. (55) Pei, R. T.; Kim, S. C.; Carlson, K. H.; Pruden, A. Effect of River Landscape on the sediment concentrations of antibiotics and corresponding antibiotic resistance genes (ARG). Water Res. 2006, 40 (12), 2427–2435. (56) Hardwick, S. A.; Stokes, H. W.; Findlay, S.; Taylor, M.; Gillings, M. R. Quantification of class 1 integron abundance in natural environments using real-time quantitative PCR. FEMS Microbiol. Lett. 2008, 278 (2), 207–212. (57) Pruden, A.; Arabi, M.; Storteboom, H. N. Correlation Between Upstream Human Activities and Riverine Antibiotic Resistance Genes. Environ. Sci. Technol. 2012, 46 (21), 11541– 11549. (58) Huovinen, P.; Sundstrom, L.; Swedberg, G.; Skold, O. Trimethoprim and sulfonamide resistance. Antimicrob. Agents Chemother. 1995, 39 (2), 279–289. (59) Garner, E.; Benitez, R.; Wagoner, E. Von; Sawyer, R.; Schaberg, E.; Hession, W. C.; Krometis, L. H.; Badgley, B. D.; Pruden, A. Stormwater loadings of antibiotic resistance genes in an urban stream. Water Res. 2017, 123, 144–152. (60) Arango-Argoty, G.; Singh, G.; Heath, L. S.; Pruden, A.; Xiao, W.; Zhang, L. MetaStorm: A public resource for customizable metagenomics annotation. PLoS One 2016, 11 (9), 1– 13. (61) McArthur, A. G.; Waglechner, N.; Nizam, F.; Yan, A.; Azad, M. a.; Baylay, A. J.; Bhullar, K.; Canova, M. J.; De Pascale, G.; Ejim, L.; et al. The comprehensive antibiotic resistance database. Antimicrob. Agents Chemother. 2013, 57 (7), 3348–3357. (62) Quast, C.; Pruesse, E.; Yilmaz, P.; Gerken, J.; Schweer, T.; Yarza, P.; Peplies, J.; Glockner, F. O. The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Res. 2013, 41, D590–D596. (63) Pal, C.; Bengtsson-Palme, J.; Rensing, C.; Kristiansson, E.; Larsson, D. G. J. BacMet: antibacterial biocide and metal resistance genes database. Nucleic Acids Res. 2014, 42, D737–D743. (64) Leplae, R.; Hebrant, A.; Wodak, S. J.; Toussaint, A. ACLAME: a CLAssification of Mobile genetic Elements. Nucleic Acids Res. 2004, 32, D45-D49. (65) Li, B.; Yang, Y.; Ma, L.; Ju, F.; Guo, F.; Tiedje, J. M.; Zhang, T. Metagenomic and network analysis reveal wide distribution and co-occurrence of environmental antibiotic resistance genes. ISME J. 2015, 1–13. (66) Meyer, F.; Paarmann, D.; D’Souza, M.; Olson, R.; Glass, E. M.; Kubal, M.; Paczian, T.; Rodriguez, A.; Stevens, R.; Wilke, A.; et al. The metagenomics RAST server - a public resource for the automatic phylogenetic and functional analysis of metagenomes. BMC Bioinformatics 2008, 9 (386). (67) Caporaso, J. G.; Lauber, C. L.; Walters, W. A.; Berg-Lyons, D.; Huntley, J.; Fierer, N.; Owens, S. M.; Betley, J.; Fraser, L.; Bauer, M.; et al. Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J. 2012, 6 (8), 1621–1624. (68) Caporaso, J. G.; Kuczynski, J.; Stombaugh, J.; Bittinger, K.; Bushman, F. D.; Costello, E. K.; Fierer, N.; Pena, A. G.; Goodrich, J. K.; Gordon, J. I.; et al. QIIME allows analysis of

104

high-throughput community sequencing data. Nat. Methods 2010, 7 (5), 335–336. (69) DeSantis, T. Z.; Hugenholtz, P.; Larsen, N.; Rojas, M.; Brodie, E. L.; Keller, K.; Huber, T.; Dalevi, D.; Hu, P.; Andersen, G. L. Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARB. Appl. Environ. Microbiol. 2006, 72 (7), 5069–5072. (70) Kindt, R.; Legendre, P.; Hara, B. O.; Simpson, G. L.; Stevens, M. H. H.; Wagner, H. The vegan Package. 2008. (71) Critically Important Antimicrobials for Human Medicine; World Health Organization: Geneva, Switzerland, 2017. (72) Czekalski, N.; Berthold, T.; Caucci, S.; Egli, A.; Bürgmann, H. Increased levels of multiresistant bacteria and resistance genes after wastewater treatment and their dissemination into Lake Geneva, Switzerland. Front. Microbiol. 2012, 3 (106). (73) Mao, D.; Yu, S.; Rysz, M.; Luo, Y.; Yang, F.; Li, F.; Hou, J.; Mu, Q.; Alvarez, P. J. J. Prevalence and proliferation of antibiotic resistance genes in two municipal wastewater treatment plants. Water Res. 2015, 85, 458–466. (74) Yuan, Q.; Guo, M.; Wei, W.; Yang, J. Reductions of bacterial antibiotic resistance through five biological treatment processes treated municipal wastewater. Environ. Sci. Pollut. Res. 2016, 19495–19503. (75) Lee, J.; Hun, J.; Shin, J.; Min, H.; Kim, S.; Seok, M.; Mo, Y. Quantitative and qualitative changes in antibiotic resistance genes after passing through treatment processes in municipal wastewater treatment plants. Sci. Total Environ. 2017, 605–606, 906–914. (76) Xu, J.; Xu, Y.; Wang, H.; Guo, C.; Qiu, H.; He, Y.; Zhang, Y.; Li, X.; Meng, W. Occurrence of antibiotics and antibiotic resistance genes in a sewage treatment plant and its effluent- recieving river. Chemosphere 2015, 119, 1379–1385. (77) Yoon, Y.; Jung, H.; Yoong, D.; Di, W.; Dodd, M. C.; Hur, H.; Lee, Y. Inactivation efficiency of plasmid-encoded antibiotic resistance genes during water treatment with chlorine, UV, and UV/H2O2. Water Res. 2017, 123, 783-793. (78) Alexander, J.; Knopp, G.; Dötsch, A.; Wieland, A.; Schwartz, T. Ozone treatment of conditioned wastewater selects antibiotic resistance genes, opportunistic bacteria , and induce strong population shifts. Sci. Total Environ. 2016, 559, 103–112. (79) Ferro, G.; Guarino, F.; Castiglione, S.; Rizzo, L. Antibiotic resistance spread potential in urban wastewater effluents disinfected by UV/H2O2 process. Sci. Total Environ. 2016, 560– 561, 29–35. (80) Xu, L.; Ouyang, W.; Qian, Y.; Su, C.; Su, J.; Chen, H. High-throughput profiling of antibiotic resistance genes in drinking water treatment plants and distribution systems. Environ. Pollut. 2016, 213, 119–126. (81) Durso, L. M.; Miller, D. N.; Wienhold, B. J. Distribution and Quantification of Antibiotic Resistant Genes and Bacteria across Agricultural and Non-Agricultural Metagenomes. PLoS One 2012, 7 (11), e48325. (82) Forsberg, K. J.; Patel, S.; Gibson, M. K.; Lauber, C. L.; Knight, R.; Fierer, N.; Dantas, G. Bacterial phylogeny structures soil resistomes across habitats. Nature 2014, 509, 612–616. (83) Gibson, M. K.; Forsberg, K. J.; Dantas, G. Improved annotation of antibiotic resistance determinants reveals microbial resistomes cluster by ecology. ISME J. 2014, 1–10. (84) Su, J.; Ouyang, W.; Wei, B.; Huang, F.; Zhao, Y.; Xu, H.; Zhu, Y.-G. Antibiotic resistome and its association with bacterial communities during sewage sludge composting. Environ. Sci. Technol. 2015, 7356-7363.

105

(85) Aminov, R. I. Horizontal gene exchange in environmental microbiota. Front. Microbiol. 2011, 2, 1-19. (86) Sekizuka, T.; Matsui, M.; Yamane, K.; Takeuchi, F.; Ohnishi, M.; Hishinuma, A.; Arakawa, Y.; Kuroda, M. Complete sequencing of the blaNDM-1-positive IncA/C plasmid from Escherichia coli ST38 isolate suggests a possible origin from plant pathogens. PLoS One 2011, 6 (9), e25334. (87) Wright, G. D. Antibiotic resistance in the environment: a link to the clinic? Curr. Opin. Microbiol. 2010, 13 (5), 589–594. (88) Forsberg, K. J.; Reyes, A.; Wang, B.; Selleck, E. M.; Sommer, M. O. A.; Dantas, G. The Shared Antibiotic Resistome of Soil Bacteria and Human Pathogens. Science 2012, 337, 1107–1111. (89) Gillings, M. R.; Gaze, W. H.; Pruden, A.; Smalla, K.; Tiedje, J. M.; Zhu, Y.-G. Using the class 1 integron-integrase gene as a proxy for anthropogenic pollution. ISME J. 2014, 1–11. (90) Madsen, J. S.; Burmolle, M.; Hansen, L. H.; Sorensen, S. J. The interconnection between biofilm formation and horizontal gene transfer. FEMS Immunol. Med. Microbiol. 2012, 65 (2), 183–195. (91) Shrestha, B.; Tada, T.; Miyoshi-Akiyama, T.; Shimada, K.; Ohara, H.; Kirikae, T.; Pokhrel, B. M. Identification of a novel NDM variant, NDM-13, from a multidrug-resistant Escherichia coli clinical isolate in Nepal. Antimicrob. Agents Chemother. 2015, 59 (9), 5847–5850. (92) Shakibaie, M. R.; Jalilzadeh, K. A.; Yamakanamardi, S. M. Horizontal transfer of antibiotic resistance genes among gram negative bacteria in sewage and lake water and influence of some physico-chemical parameters of water on conjugation process. J. Environ. Biol. 2009, 30 (1), 45–49. (93) Liu, A.; Fong, A.; Becket, E.; Yuan, J.; Tamae, C.; Medrano, L.; Maiz, M.; Wahba, C.; Lee, C.; Lee, K.; et al. Selective advantage of resistant strains at trace levels of antibiotics: a simple and ultrasensitive color test for detection of antibiotics and genotoxic agents. Antimicrob. Agents Chemother. 2011, 55 (3), 1204–1210. (94) McVicker, G.; Prajsnar, T. K.; Williams, A.; Wagner, N. L.; Boots, M.; Renshaw, S. a; Foster, S. J. Clonal expansion during Staphylococcus aureus infection dynamics reveals the effect of antibiotic intervention. PLoS Pathog. 2014, 10 (2), e1003959. (95) Andersson, D. I.; Hughes, D. Evolution of antibiotic resistance at non-lethal drug concentrations. Drug Resist. Updat. 2012, 15 (3), 162–172. (96) Gullberg, E.; Cao, S.; Berg, O. G.; Ilbäck, C.; Sandegren, L.; Hughes, D.; Andersson, D. I. Selection of resistant bacteria at very low antibiotic concentrations. PLoS Pathog. 2011, 7 (7), e1002158. (97) Gullberg, E.; Albrecht, L. M.; Karlsson, C.; Sandegren, L.; Andersson, D. I. Selection of a Multidrug Resistance Plasmid by Sublethal Levels of Antibiotics and Heavy Metals. mBio 2014, 5 (5), 19–23. (98) Zhang, P. Y.; Xu, P. P.; Xia, Z. J.; Wang, J.; Xiong, J.; Li, Y. Z. Combined treatment with the antibiotics kanamycin and streptomycin promotes the conjugation of Escherichia coli. FEMS Microbiol. Lett. 2013, 348 (2), 149–156. (99) Xia, Z. J.; Wang, J.; Hu, W.; Liu, H.; Gao, X. Z.; Wu, Z. H.; Zhang, P. Y.; Li, Y. Z. Improving conjugation efficacy of Sorangium cellulosum by the addition of dual selection antibiotics. J. Ind. Microbiol. Biotechnol. 2008, 35 (10), 1157–1163. (100) Song, B.; Wang, G. R.; Shoemaker, N. B.; Salyers, A. A. An Unexpected Effect of

106

Tetracycline Concentration: Growth Phase-Associated Excision of the Bacteroides Mobilizable Transposon NBU1. J. Bacteriol. 2009, 191 (3), 1078–1082. (101) Úbeda, C.; Maiques, E.; Knecht, E.; Lasa, Í.; Novick, R. P.; Penadés, J. R. Antibiotic- induced SOS response promotes horizontal dissemination of pathogenicity island-encoded virulence factors in staphylococci. Mol. Microbiol. 2005, 56 (3), 836–844. (102) Prudhomme, M.; Attaiech, L.; Sanchez, G.; Martin, B.; Claverys, J. P. Antibiotic stress induces genetic transformability in the human pathogen Streptococcus pneumoniae. Science 2006, 313 (5783), 89–92. (103) Beaber, J. W.; Hochhut, B.; Waldor, M. K. SOS response promotes horizontal dissemination of antibiotic resistance genes. Nature 2004, 427, 72–74. (104) Kooij, D. Van Der. Assimilable Organic Carbon as an Indicator of Bacterial Regrowth. J. Am. Water Work. Assoc. 1992, 57–65. (105) Williams, K.; Pruden, A.; Falkinham, J.; Edwards, M. Relationship between Organic Carbon and Opportunistic Pathogens in Simulated Glass Water Heaters. Pathogens 2015, 4 (2), 355–372. (106) Blueprint for One Water; Water Research Foundation; 2017. (107) Garner, E.; Zhu, N.; Strom, L.; Edwards, M.; Pruden, A. A human exposome framework for guiding risk management and holistic assessment of recycled water quality. Environ. Sci. Water Res. Technol. 2016, 2, 580-598.

107

SUPPLEMENTARY MATERIAL FOR CHAPTER 5

Antibiotic Analysis

Antibiotics were extracted from water samples using solid phase extraction according to Sui et al.,1 with minor modifications. A volume of 500 mL water sample filtered through glass microfiber filter paper (Whatman grade GF/F, GE Healthcare Bio-Sciences, PA) was first adjusted to pH=7 and then introduced to a solid phase extraction cartridge (Oasis HLB Cartridges, 60 mg sorbent; Waters, Milford, MA), pre-conditioned with 3 mL of methanol and 3 mL of MilliQ water. The water was passed through a cartridge at an approximate rate of 5 mL min-1. After loading a water sample, the cartridge was first washed with MilliQ water and then vacuum dried for 10 minutes. The target analytes were eluted off each cartridge with 3 mL methanol at a flow rate of 5 -1 mL min . Each eluent was dried at 40°C under N2 gas (RapidVap N2 Evaporation Systems, Labconco Corp., MO), reconstituted with 1 mL 90% acetonitrile, filtered through a 0.2 µm PTFE syringe filter (Fisher, Pittsburgh, PA) into a 2 mL amber glass HPLC vial, and then analyzed using an ultra-performance liquid chromatography-tandem mass spectrometer (UPLC/MS/MS) (Agilent 1290 UPLC/Agilent 6490 Triple Quad tandem mass spectrometry, Agilent Technologies Inc., Santa Clara, CA).

Each antibiotic was identified and its peak area quantified under multiple reaction monitoring mode (MRM), which is achieved by specifying the precursor and two product ions formed after fragmentation with particular collision energies.2 A Zorbax Extend C18 analytical column (4.6 × 50 mm, 5 µm particle size) coupled with a Zorbax Extend C18 guard column (4.6 × 12 mm, 5 µm particle size) was used for chromatographic separation. The temperature of the column oven was kept at 40 °C. Separation was achieved with a gradient elution consisting of two mobile phases (A: 0.1% formic acid in water; B: 0.1% formic acid in 95% acetonitrile) at a flow rate of 0.5 mL min-1. The mobile phase gradient was programed as: 0-4.5 min (Isocratic elution): 90% of A and 10% of B; 4.6-10.0 min (Gradient elution): change from 90% of A and 10% of B to 0% A and 100% of B; 10.1-14.0 min (Purging): 0% of A and 100% of B, and 14.1-18 min (Re- equilibration): 90% of A and 10% of B.

The sample injection volume was 10 µL. Each sample was analyzed three times. Mill-Q water was extracted, as a laboratory system blank control, using the same procedure and no antibiotics were detected in these blanks. All water samples were extracted, cleaned up, and analyzed on the UPLC/MS/MS within the same batch, making it possible to compare the relative quantity of each antibiotic across different samples using the peak area of each analyte. This approach enables fast screening of a large set of antibiotics and eliminates the need of costly absolute quantification of each analyte using expensive standards. The use of UPLC-MS/MS for screening was reported in a semi-quantitative screening method for the 14 antibiotics in the collected wastewater. Any measurements that fell below a signal-to-noise ratio of 3 was considered non-detect.

Quantification of antibiotic resistance genes

To construct standards for each gene, wastewater samples were amplified with each primer set via PCR and analyzed via gel electrophoresis. The sequence of target gene amplicons of the

108

appropriate length were confirmed via Sanger Sequencing conducted at the Biocomplexity Institute of Virginia Tech Genomics Sequencing Center (Blacksburg, VA). Confirmed amplicons were cloned into plasmids using a TOPO TA cloning kit (Thermo Fisher Scientific, Waltham, MA). Standards were then generated and quantified as described previously.3 The specificity of all assays in wastewater or fecal samples has been previously documented3–8 and, when appropriate, melt curves were closely monitored to ensure that melt temperature of amplicons corresponded with standards. Standard curve R2 values and efficiencies are presented for each gene assay in Table S7.

References

(1) Sui, Q.; Huang, J.; Deng, S.; Chen, W.; Yu, G. Seasonal Variation in the Occurrence and Removal of Pharmaceuticals and Personal Care Products in Different Biological Wastewater Treatment Processes. Env. Sci. Technol. 2011, 45, 3341–3348. (2) Naegele, E. Detection of Trace Level Pharmaceuticals in Drinking Water by Online SPE Enrichment with the Agilent 1200 Infinity Series Online-SPE Solution. 2013. (3) Pei, R. T.; Kim, S. C.; Carlson, K. H.; Pruden, A. Effect of River Landscape on the sediment concentrations of antibiotics and corresponding antibiotic resistance genes (ARG). Water Res. 2006, 40 (12), 2427–2435. (4) Suzuki, M. T.; Taylor, L. T.; DeLong, E. F. Quantitative analysis of small-subunit rRNA genes in mixed microbial populations via 5 ’-nuclease assays. Appl. Environ. Microbiol. 2000, 66 (11), 4605–4614. (5) Bibbal, D.; Dupouy, V.; Ferre, J. P.; Toutain, P. L.; Fayet, O.; Prere, M. F.; Bousquet- Melou, A. Impact of three ampicillin dosage regimens on selection of ampicillin resistance in Enterobacteriaceae and excretion of bla(TEM) genes in swine feces. Appl. Environ. Microbiol. 2007, 73 (15), 4785–4790. (6) Dutka-Malen, S.; Evers, S.; Courvalin, P. Detection of glycopeptide resistance genotypes and identification to the species level of clinically relevant enterococci by PCR. J. Clin. Microbiol. 1995, 33 (1), 24–27. (7) Colomer-Lluch, M.; Jofre, J.; Muniesa, M. Quinolone resistance genes (qnrA and qnrS) in bacteriophage particles from wastewater samples and the effect of inducing agents on packaged antibiotic resistance genes. J. Antimicrob. Chemother. 2014, 69 (5), 1265–1274. (8) Hardwick, S. A.; Stokes, H. W.; Findlay, S.; Taylor, M.; Gillings, M. R. Quantification of class 1 integron abundance in natural environments using real-time quantitative PCR. FEMS Microbiol. Lett. 2008, 278 (2), 207–212.

109

Table S1: Seasonal sample collection dates for each utility Utility A B C D Spring May 11-21, June 2-3, 2015 June 4-11, May 18-19, 2015 2015 2015 Summer August 11-20, July 30-August September 30- August 10-11, 2014 11, 2014 October 1, 2015 2015 Fall October 27- September 29- October 7-9, October 20, November 5, 30, 2015 2014 2015 2014 Winter February 2-11, October 20-22, February 12- February 23- 2015 2014* 19, 2015 24, 2015 *Utility B’s reclaimed system is shut down for the winter, so a winter collection was not possible. Samples were collected just before the system was shut down for the winter

Figure S1: Comparison of absolute abundances (normalized to milliltiers of water or biofilm swab) of the sul1 gene determined by two methods: qPCR versus shotgun metagenomics modified by multiplication of 16S rRNA gene abundances (determined by qPCR). This analysis demonstrates that the modified metagenomic method results in sul1 gene abundances that are strongly correlated with the highly quantitative values produced by qPCR Spearman’s ρ = 0.6411, p<0.001). Sample labels represent Utility - System type (D=potable; R=reclaimed) - Sample (POE=point of entry; POU=point of use; INF=raw wastewater influent) - Season (Sp=Spring; Su=Summer; F=Fall; W=Winter) - Matrix (W=water; B=biofilm)

110

Table S2: List of shotgun metagenomic sequenced samples, MG-RAST sample IDs, and assembly information MG- Average System Sample Paired- Number Matrix Utility Season Sample Name RAST N50 scaffold Type Type end reads scaffolds Sample ID length 5,412,528 Drinking biofilm A POU Su Drinkingwater_biofilm_ UtilityA_S5_L1S10 mgs295574 4,760 94,299 1,465 5,887,968 Drinking bulk water A POU Su Drinkingwater_bulkwater_ UtilityA_S5_L1S6 mgs295565 7,474 121,193 1,439 38,668,577 Drinking bulk water B POE Su Drinkingwater_bulkwater_ UtilityB_S0_1 mgs458886 22,674 43,273 1,360 32,868,754 Drinking bulk water B POE Su Drinkingwater_bulkwater_ UtilityB_S0_1_duplicate mgs458889 4,585 54,477 1,262 7,132 Reclaimed bulk water A Influent Su Reclaimedwater_bulkwater_UtilityA_-1_1 mgs458922 927 351,988 748 51,423,687 Reclaimed bulk water A POE F Reclaimedwater_bulkwater_UtilityA_S0_2 mgs458874 2,687 63,023 971 12,036,240 Reclaimed bulk water A POE W Reclaimedwater_bulkwater_UtilityA_S0_3 mgs458865 1,530 56,797 982 4,274,193 Reclaimed bulk water A POE Sp Reclaimedwater_bulkwater_UtilityA_S0_4 mgs458868 32,255 43,020 2,244 742,276 Reclaimed bulk water A POE Su Reclaimedwater_bulkwater_UtilityA_S0_L1S2 mgs295556 2,453 18,883 1,159 32,082,202 Reclaimed biofilm A POU Su Reclaimedwater_biofilm_ UtilityA_S5_L1S9 mgs466972 1,324 162,693 929 40,816,121 Reclaimed bulk water A POU F Reclaimedwater_bulkwater_UtilityA_S5_2 mgs458880 2,376 83,117 1,118 43,183,835 Reclaimed bulk water A POU W Reclaimedwater_bulkwater_UtilityA_S5_3 mgs458907 2,276 143,562 1,116 43,580,278 Reclaimed bulk water A POU Sp Reclaimedwater_bulkwater_UtilityA_S5_4 mgs458913 2,322 213,956 1,209 3,827,661 Reclaimed bulk water A POU Su Reclaimedwater_bulkwater_UtilityA_S5_L1S5 mgs295562 2,260 170,551 1,179 39,243,688 Reclaimed bulk water B Influent Su Reclaimedwater_bulkwater_UtilityB_-1_1 mgs458937 1,024 206,090 775 24,153,588 Reclaimed bulk water B POE W Reclaimedwater_bulkwater_UtilityB_S0_2 mgs458940 993 101,493 770 20,416,825 Reclaimed bulk water B POE Sp Reclaimedwater_bulkwater_UtilityB_S0_3 mgs458943 1,073 68,192 814 48,463,674 Reclaimed bulk water B POE F Reclaimedwater_bulkwater_UtilityB_S0_4 mgs458931 1,334 187,500 894 4,726,224 Reclaimed bulk water B POE Su Reclaimedwater_bulkwater_UtilityB_S0_L1S1 mgs295553 1,000 99,628 794 191,192 Reclaimed biofilm B POU Su Reclaimedwater_biofilm_ UtilityB_S5_L1S7 mgs295568 837 6,656 700 28,668,559 Reclaimed bulk water B POU W Reclaimedwater_bulkwater_UtilityB_S5_2 mgs458925 9,297 56,227 1,952 39,378,131 Reclaimed bulk water B POU Sp Reclaimedwater_bulkwater_UtilityB_S5_3 mgs458871 1,383 9,782 890 18,232,136 Reclaimed bulk water B POU F Reclaimedwater_bulkwater_UtilityB_S5_4 mgs458877 2,887 83,141 1,124 4,530,935 Reclaimed bulk water B POU Su Reclaimedwater_bulkwater_UtilityB_S5_L1S3 mgs295559 3,170 104,347 1,331 20,146,190 Reclaimed bulk water C Influent Sp Reclaimedwater_bulkwater_UtilityC_-1_3 mgs458934 958 66,992 755

111

33,224 Reclaimed bulk water C POE Sp Reclaimedwater_bulkwater_UtilityC_S0_3 mgs458892 1,314 241,370 864 32,079,119 Reclaimed bulk water C POE Su Reclaimedwater_bulkwater_UtilityC_S0_4 mgs458883 1,691 123,451 892 5,300,338 Reclaimed bulk water C POE F Reclaimedwater_bulkwater_UtilityC_S0_L1S11 mgs295577 5,659 12,092 946 40,009,882 Reclaimed bulk water C POU Sp Reclaimedwater_bulkwater_UtilityC_S5_3 mgs458904 1,293 240,845 908 33,917,149 Reclaimed bulk water C POU Su Reclaimedwater_bulkwater_UtilityC_S5_4 mgs458895 1,455 152,566 906 36,097,736 Reclaimed bulk water D Influent Su Reclaimedwater_bulkwater_UtilityD_-1_3 mgs458916 877 192,239 728 40,363,707 Reclaimed bulk water D POE W Reclaimedwater_bulkwater_UtilityD_S0_1 mgs458919 1,046 216,558 795 35,768,857 Reclaimed bulk water D POE Sp Reclaimedwater_bulkwater_UtilityD_S0_2 mgs458910 1,207 188,909 831 37,816,330 Reclaimed bulk water D POE Su Reclaimedwater_bulkwater_UtilityD_S0_3 mgs458898 1,133 227,064 807 44,493,192 Reclaimed bulk water D POU W Reclaimedwater_bulkwater_UtilityD_S5_1 mgs458901 2,169 190,094 1,116 35,693,403 Reclaimed bulk water D POU Sp Reclaimedwater_bulkwater_UtilityD_S5_2 mgs458946 1,142 174,840 810 21,616,266 Reclaimed bulk water D POU Sp Reclaimedwater_bulkwater_UtilityD_S5_2_duplicate mgs458949 983 119,681 769 5,181 Reclaimed bulk water D POU Su Reclaimedwater_bulkwater_UtilityD_S5_3 mgs458862 4,275 66,607 1,316 24,947,026 Reclaimed bulk water D POU F Reclaimedwater_bulkwater_UtilityD_S5_4 mgs458928 990 150,711 791

112

Table S3: List of samples analyzed via 16S rRNA sequencing and included in the analysis of correlations between ARGs and the overall microbial community or individual phyla. All samples collected in this study were prepared for sequencing. Some samples were excluded from sequencing if insufficient DNA could be obtained or if other QA/QC requirements (i.e. excessive negative control amplification) were not met after three amplification attempts. Samples produced an average of 63,864±43,264 reads. Utility Type Season Site Matrix Number samples sequenced POE W 1 W 10 Su POU B 10 S W 2 POE W 1 Potable W W 7 POU B 8 POE W 1 W POU W 10 POE W 1 Sp POU W 9 A INW W 2 POE W 2 Su W 10 POU B 10 POE W 2 W W 10 Reclaimed POU B 2 POE W 2 W W 9 POU B 10 POE W 2 Sp POU W 10 POE W 1 W 6 Su POU B 5 S W 2 Potable POE W 1 W W 5 POU B 4 B Sp POU W 2 W POU W 3 INW W 2 POE W 2 Su W 6 Reclaimed POU B 4 POE W 2 W POU W 5

113

B 4 POE W 2 So POU W 5 POU B 4 POE W 2 W W 4 POU B 3 POE W 1 W POU W 5 Potable S W 3 Sp POE W 1 SU POU W 3 INW W 1 C W POE W 1 POU W 7 W POU W 1 Reclaimed POE W 1 Sp POU W 5 POE W 1 Su POU W 5 Sp POU W 4 POE W 1 Su Potable POU W 4 POE W 1 W POU W 5 POE W 1 D W POU W 5 POE W 1 Sp Reclaimed POU W 5 POE W 1 Su POU W 5 W POU W 5 Field Blank 9 DNA Extraction Blank 2 PCR Blank 5 Abbreviations: Season (Sp=Spring; Su=Summer; F=Fall; W=Winter), Sample (POE=point of entry; POU=point of use; INF=raw wastewater influent; S=source water), Matrix (W=water; B=biofilm)

114

Figure S2: Nonmetric multidimensional scaling (NMDS) plot generated from Bray-Curtis similarity matrix of all metagenomic ARG abundances by utility and system type.

115

Table S4: Frequency of qPCR detection and relative abundance of ARGs (normalized to 16S rRNA genes; log(ARG copies/16S rRNA gene copies)) in potable and reclaimed water and biofilm samples at the point of entry (POE) and point of use (POU); average (standard deviation) of values above the limit of quantification. Values for each sampling location include four sampling events. blaTEM intI1 qnrA vanA sul1 POE - water Potable A (n=3) 33%; -3.0(0.0) 33%; 1.5(0.0) ND ND ND B (n=4) ND 25%; -2.2(0.0) ND 50%; -4.6(-4.9) 25%; -4.0(0.0) C (n=4) 25%; -1.7(0.0) 25%; 1.2(0.0) ND ND ND D (n=4) ND ND 25%; -1.6(0.0) ND ND Reclaimed A (n=7) 14%; -3.9(0.0) 57%; 0.8(1.1) ND 14%; -3.0(0.0) 100%; 0.2(0.4) B (n=7) 43%; -4.0(-3.9) 29%; -0.2(-0.1) ND ND 100%; -1.6(-1.6) C (n=4) 50%; -4.3(-4.9) 50%; 1.0(0.6) ND ND 50%; -0.4(-0.8) D (n=4) 75%; -4.4(-4.4) 25%; -0.8(0.0) 75%; -4.8(-5.3) ND 100%; -1.0(-1.0) POU - water Potable A (n=44) 23%; -0.6(-0.2) 25%; 1.6(1.9) 2%; -2.8(0.0) 11%; -2.1(-2.1) 18%; 1.5(1.9) B (n=21) 19%; -1.9(-1.7) 19%; 1.0(1.0) ND 14%; -1.4(-1.2) 19%; -0.9(-0.7) C (n=15) ND ND ND ND 47%; -0.4(-0.5) D (n=20) ND ND ND ND 5%; -0.8(0.0) Reclaimed A (n=38) 18%; -3.4(-3.0) 71%; -0.1(0.1) ND ND 92%; -0.1(0.2) B (n=17) 12%; -2.2(0.0) 53%; -0.7(-0.6) ND 6%; -2.6(0.0) 88%; 4.8(5.3) C (n=17) 18%; -4.7(-4.6) 53%; -0.2(-0.2) 6%; -6.0(0.0) ND 59%; -0.2(-0.3) D (n=20) 80%; -4.1(-4.1) 55%; -0.8(-0.8) 45%; -4.7(-4.7) ND 100%; -0.7(-0.4) POU - biofilm Potable A (n=40) 18%; -2.3(-2.4) 20%; 1(1.3) ND 5%; -3.1(-3.2) 15%; 0.6(0.9) B (n=21) 10%; -2.4(-2.3) ND 5%; -2.3(0.0) 10%; -1.9(-2.2) 24%; -0.2(0.1) Reclaimed A (n=33) 6%; -2.8(-2.7) 55%; -0.4(0.1) 3%; -3.5(0.0) ND 85%; -0.7(-0.4) B (n=15) 13%; -2.9(0.0) 20%; 1.4(1.6) 7%; -3.0(0.0) 20%; -2.4(-2.4) 80%; 0.0(0.5)

116

Figure S3: Canonical correspondence analysis comparing ARG and microbial community profiles. Points represent gene profiles based on the abundance of sul1, qnrA, vanA, blaTEM, and intI1 determined by qPCR. Triangles represent operational taxonomic units as determined by 16S rRNA gene amplicon sequencing.

117

Figure S4: Network analysis depicting co-occurrence of ARGs among each other as well as with plasmid gene markers on assembled scaffolds constructed using de novo assembly of shotgun metagenomic sequencing reads from (A) reclaimed samples and (B) potable samples. Proximity of nodes and width of lines is proportional to numbers of association between genes. Node diameter is proportional to the number of co-occurences for that gene. A minimum of three co- occurrences were required for inclusion in the graphic. All ARG nodes depicted share a connection with the plasmid node, with the exception of EreA2, indicated by an asterisk. MLS indicates resistance to macrolide, lincosamide, and streptogramin antibiotics.

118

Table S5: Water chemistry data for potable and reclaimed distribution system samples biodegradable dissolved dissolved dissolved temperature total Cl free Cl turbidity conductivity oxygen total organic organic organic (oC) (mg/L) (mg/L) pH (NTU) (S/m) (mg/L) carbon (µg/L) carbon (µg/L) carbon (µg/L)

A (n=44) 25.5 ± 3.3 3.5 ± 1.1 -- 7.9 ± 0.2 1.5 ± 2.7 505.8 ± 119.5 5.6 ± 1 2470 ± 762 2748 ± 1002 465 ± 758 B (n=16) 18.1 ± 3.1 0.7 ± 0.3 0.7 ± 0.3 7.8 ± 0.2 0.3 ± 0.2 361.9 ± 97.2 7.7 ± 0.6 1120 ± 1422 1439 ± 1133 548 ± 564 C (n=20) 28.4 ± 4.4 -- 0.9 ± 0.1 7.9 ± 0.2 0.2 ± 0.2 727.4 ± 49.4 7.2 ± 0.5 188 ± 62 1252 ± 1980 1522 ± 1683

Potable D (n=40) 19.4 ± 1.9 -- 0.2 ± 0.1 7.7 ± 0.1 1.3 ± 0.8 957.5 ± 416.5 5.5 ± 1.2 BDa BDa BDa

A (n=20) 26.7 ± 4.3 0.4 ± 0.4 0.2 ± 0.2 7.7 ± 0.3 5.4 ± 10.8 1354.4 ± 215.4 6.5 ± 1.3 5714 ± 2564 6351 ± 2761 2137 ± 2321 B (n=19) 19.6 ± 2.9 2.3 ± 3.1 1.1 ± 2.1 7.3 ± 0.2 2.5 ± 2.6 736.9 ± 80.3 4 ± 1.4 10123 ± 6173 11087 ± 7014 6094 ± 8777 C (n=20) 26.2 ± 3.7 0.3 ± 0.1 0.4 ± 0.4 7.3 ± 0.1 0.6 ± 0.3 1195.8 ± 17.3 4.3 ± 2.8 2791 ± 1810 2944 ± 1646 2191 ± 2244 Reclaimed D (n=20) 20 ± 1.8 2.7 ± 2.3 0.2 ± 0.2 7.2 ± 0.1 2 ± 1 1542.9 ± 283.1 6.8 ± 2.4 3961 ± 2120 4333 ± 2069 2621 ± 1238 BD = below limit of detection a19/20 samples below limit of detection (4 µg/L)

119

Figure S5: Network analysis depicting co-occurrence of ARGs and metal resistance genes on assembled scaffolds constructed using de novo assembly of shotgun metagenomic sequencing reads in (A) reclaimed samples, (B) potable samples, and (C) all samples. Proximity of nodes and width of lines is proportional to numbers of association between genes. Node diameter is proportional to the number of co-occurences for that gene. A minimum of three co-occurrences were required for inclusion in the graphic. MLS indicates resistance to macrolide, lincosamide, and streptogramin antibiotics.

120

Table S6: Antibiotics detectable in a subset of potable and reclaimed water samples. Values refer to peak area. Ornidazole (anti-protozoan), nalidixic acid, sulfamethoxazole (class: sulfonamide), oxolinic acid (quinolone), flumequine (fluoroqinolone), ormetoprim, sulfamethazine (sulfonamide), tetracycline, cefotaxime (beta-lactam), chlorotetracycline (tetracycline), erythromycin (macrolide), and tylosin (macrolide) were screened but not detectable in any samples. N.D. indicates compound not detectable. Vancomycin Trimethoprim Influent Plant 1 22072 3628247 Influent Plant 2 41236 1537909 Utility A Reclaimed POE Plant 1 N.D. N.D. POE Plant 2 N.D. N.D. Source N.D. N.D. Potable POE 41236 1537909 Influent Plant 1 6359 2351518 Utility B Influent Plant 2 13582 1211878 Reclaimed POE Plant 1 N.D. N.D. POE Plant 2 89515 1294057 Source N.D. N.D. Potable POE N.D. 2410 Utility C Influent N.D. 719542 Reclaimed POE N.D. 75082 Source N.D. N.D. Potable POE N.D. N.D. Utility D Influent 3636 698209 Reclaimed POE N.D. 813197

121

Figure S6: Relative abundance of ARGs (normalized to 16S rRNA gene copies) by antibiotic class (stacked bars) and plasmids (diamonds) per mL of water sampled or per biofilm swab as determined by annotation of reads from shotgun metagenomic sequencing of samples collected at the point-of-entry (POE) and point-of-use (POU) of four potable water utilities (A, B, C, D). Reads were annotated against the Comprehensive Antibiotic Resistance Database for ARGs and the ACLAME database for plasmids. Sufficient DNA for analysis of potable water samples was only possible at Utilities A and B, the remaining samples were from reclaimed water distribution systems. Error bars indicate standard deviation of total abundance of ARGs or plasmids when statistical power was sufficient (n value indicated for each sample category).

122

Table S7: qPCR standard curve R2 and efficiency values for each gene assay (average ± standard deviation) R2 efficiency 16S rRNA 0.989±0.010 101.982±14.255 sul1 0.995±0.004 100.227±8.986 intI1 0.980±0.006 95.46±12.74 vanA 0.988±0.007 99.75±7.278 qnrA 0.993±0.006 100.45±7.85 blaTEM 0.998±0.002 95.48±5.445

123

CHAPTER 6 : MICROBIAL ECOLOGY AND WATER CHEMISTRY IMPACT REGROWTH OF OPPORTUNISTIC PATHOGENS IN FULL-SCALE RECLAIMED WATER DISTRIBUTION SYSTEMS

Emily Garner, Jean McLain, Jolene Bowers, David M. Engelthaler, Marc A. Edwards, Amy Pruden ABSTRACT

Need for global water security has spurred growing interest in wastewater reuse to offset demand for municipal water. While reclaimed (i.e., non-potable) microbial water quality regulations target fecal indicator bacteria, opportunistic pathogens (OPs), which are subject to regrowth in distribution systems and spread via aerosol inhalation and other non-ingestion routes, may be more relevant. This study compares the occurrences of five OP gene markers (Acanthamoeba spp., Legionella spp., Mycobacterium spp., Naegleria fowleri, Pseudomonas aeruginosa) in reclaimed versus potable water distribution systems and characterizes factors potentially contributing to their regrowth. Samples were collected over four sampling events at the point of compliance for water exiting treatment plants and at five points of use at four U.S. utilities bearing both reclaimed and potable water distribution systems. Reclaimed water systems harbored unique water chemistry (e.g. elevated nutrients), microbial community composition, and OP occurrence patterns compared to potable systems examined here and reported in the literature. Legionella spp. genes, Mycobacterium spp. genes, and total bacteria, represented by 16S rRNA genes, were more abundant in reclaimed than potable water distribution system samples (p≤0.0001). This work suggests that further consideration should be given to managing reclaimed water distribution systems with respect to non-potable exposures to OPs.

INTRODUCTION

Growing need for sustainable water sources has spurred interest in direct and indirect potable reuse to supplement traditional surface and groundwater supplies. Approximately 1.6 billion people globally live in watersheds impacted by water scarcity and, by 2050, it is projected that due to climate change and population increase, the number of people affected will roughly double.1 In these areas, wastewater reuse is particularly attractive to meet both potable and non- potable water demand. Non-potable reuse is already common in the U.S. for irrigation of agricultural and urban areas, groundwater recharge, and industrial reuse.2 While advanced treatments enable production of high-quality water, maintaining microbial water quality as water is transported to the point of use may present a greater challenge than that recognized for potable water and premise (i.e., building) plumbing, due to the unique qualities of reclaimed water, including high levels of growth-promoting nutrients, rapid decay of disinfectant residual, stagnation, and elevated distribution system retention times.3

Where regulations exist, typically at the state level in the U.S., microbial water quality in reclaimed systems is typically characterized via monitoring of E. coli, Enterococci, or fecal or total coliforms.2 While these parameters track contamination from fecal bacteria, they are not good surrogates for opportunistic pathogens (OPs), which are non-fecal, such as Legionella pneumophila, Mycobacterium avium, Pseudomonas aeruginosa, Acanthamoeba spp., and Naegleria fowleri.4 Although waterborne disease related to fecal pathogens has nearly been

124

eradicated in most developed countries, OPs are now among the primary sources of tap water- related outbreak in the U.S. and elsewhere with developed water systems.5,6 OPs can infect humans via inhalation of aerosols or dermal, eye, or ear contact,7–10 which are more relevant than ingestion for non-potable reuse applications. L. pneumophila and M. avium are the causative agents of severe lung infections characterized by Legionnaires’ disease and M. avium complex, respectively.11,12 P. aeruginosa can infect hosts via the bloodstream, eyes, ears, skin, or lungs,8 while Acanthamoeba spp. can cause infection of the eyes or central nervous system via inhalation or penetration of skin lesions.13 N. fowleri can infect the brain following entrance of water into the nasal cavity, with infections having been linked to nasal irrigation with neti pots and other hygienic or recreational activities where water can “get up the nose”.14 Exposure via aerosol inhalation could result from use of reclaimed water in cooling towers, spray irrigation, toilet flushing, fire suppression, and car washing.15–17 Further, dermal or eye and ear contact is feasible from use of reclaimed water for irrigation of athletic and recreational facilities, snowmaking, and toilet flushing.4,17 Presently, little is known about the occurrence of OPs in reclaimed water distribution systems, with one field survey having documented their occurrence at the point-of-use.18

While OPs are expected to be present at relatively low concentrations following treatment of recycled water, they are known to thrive in pipe biofilms and are generally tolerant of chlorine and other disinfectants, especially when residing in amoebae.19,20 OPs are also capable of growth under the extremely low organic carbon and nutrient concentrations characteristic of potable water.19 Stagnant conditions, which are common in reclaimed water systems due to seasonal shutdowns and intermittent demand, are also thought to trigger OP regrowth.21

In addition to improved documentation of occurrence patterns of OPs in reclaimed distribution systems, fundamental understanding of how various physicochemical conditions relate to their regrowth potential during transport to the point of use is needed. The role of biostability (i.e., bioavailable nutrient content) of the water and other factors potentially stimulating regrowth of OPs in reclaimed water is of particular interest. Here we surveyed gene markers for Legionella spp., Mycobacterium spp., P. aeruginosa, Acanthamoeba spp., and N. fowleri in the distribution system point of entry (POE) and at five points of use (POU) at four U.S. utilities distributing reclaimed water for non-potable reuse and compared occurrences to corresponding municipal potable water systems over four sampling events. Quantitative polymerase chain reaction (qPCR) was employed to quantify specific OP gene markers of interest, while 16S rRNA amplicon sequencing and shotgun metagenomic sequencing provided broader context of microbial community structure and a means to explore other potential microbes of concern. The specific objectives were to 1) quantify regrowth in distribution systems by comparing OP gene copy numbers at the POE versus various POUs, 2) examine partitioning of OPs between bulk water and biofilms, 3) identify associations between water chemistry, water age and regrowth of OPs, and 4) characterize the relationship between the occurrence of OPs and the microbial community composition of the distribution system.

125

METHODS

Site description, sample collection, and preservation

Four U.S. utilities participated in this study (Table 6-1), with both the reclaimed and potable water distribution systems sampled in each city. Utilities were selected based on similar intended reclaimed water use (i.e. all utilities produced non-potable water for use primarily as irrigation water). For each potable or reclaimed system, samples were collected of freshly treated water at the point of compliance/POE to the distribution system and at five locations representing a range of water ages throughout the distribution system at the POU. Flushed bulk water samples were collected from POUs via distribution system sampling ports in sterile 1-L polypropylene containers prepared with 292 mg ethylenediaminetetraacetic acid (EDTA) and 48 mg sodium thiosulfate per liter sampled, to chelate metals and quench chlorine, which could kill cells, damage DNA, or otherwise inhibit or interfere with downstream molecular analyses. Samples for organic carbon analysis were collected in 250 mL amber glass bottles that were acid-washed and baked for five hours at 550ºC. Additional water was collected in separate acid-washed 250 mL bottles for other chemical analyses. For Utilities A and B, after collecting bulk water samples, biofilm samples were collected by inserting a sterile cotton-tipped applicator into the distribution system pipe (Fisher Scientific, Waltham, MA), pressing it to the pipe’s surface, and in a single pass, swabbing the upper 180º of the circumference of the pipe. The swab was transferred directly to a sterile DNA extraction lysing tube and the stem snapped and severed to preserve only the sample end of the swab.

Samples were shipped overnight on ice and processed within approximately 24 hours of sample collection. Samples for molecular analysis were concentrated onto 0.22 µm mixed cellulose esters membrane filters (Millipore, Billerica, MA). Filters were folded into quarters, torn into 1 cm2 pieces using sterile forceps, transferred to lysing tubes, and stored at 20ºC for later analysis. DNA was subsequently extracted using a FastDNA SPIN Kit (MP Biomedicals, Solon, OH). Biological activity reaction tests (BART; Hach, Loveland, CO) were used to examine the presence of active nitrifying, denitrifying, and sulfate-reducing bacteria.

Water Chemistry

Free chlorine, total chlorine, temperature, dissolved oxygen, pH, turbidity, and electrical conductivity were measured on-site using in-house resources used routinely by each participating utility. Upon receipt in the lab, 30 mL was subject to total organic carbon (TOC) analysis and 30 mL was filtered through pre-rinsed 0.22 µm pore size mixed cellulose esters membrane filters (Millipore, Billerica, MA) for dissolved organic carbon (DOC) analysis. Biodegradable dissolved organic carbon (BDOC) was measured as previously described by Servais et al.22 but with the incubation time extended to 45 days. Samples were analyzed on a Sievers 5310C TOC analyzer (GE, Boulder, CO) according to Standard Method 5310C.23 Metals were measured using an Electron X-Series inductively coupled plasma mass spectrometer (Thermo Fisher, Waltham, MA) according to Standard Method 3125B.23 Nitrate, nitrite, phosphate, and sulfate were quantified via a Dionex DX-500 ion chromatographer (Thermo Fisher, Waltham, MA) according to Standard Method 4110B.23

126

Table 6-1: Overview of surveyed potable and reclaimed systems U.S. Region Potable Reclaimed Utility (Climate)a Disinfectant Source Summary of Treatment Disinfectant Southeast Plant #1b - Bardenpho (Humid Surface and Cl2 A NH2Cl b d Subtropical; Groundwater Plant #2 - Activated (NH2Cl) Cfa) sludge, dentirification Southwest Cl Cl ; 2 (Mediterran; 2 Surface and Plant #1c - Bardenpho (NH Cl)d B occasional 2 Csb) Groundwater ClO UV 2 Plant #2c - Biofiltration Southwest Secondary treatment (Mid- Surface and followed by dual C Latitude Cl NH Cl 2 Groundwater membrane filters or 2 Steppe and membrane bioreactors Desert; Bsh) West Secondary treatment Surface and Cl D (Mediterran; Cl followed by dual media 2 2 Groundwater (NH Cl)d Csb) filters 2 aKöppen climate classification: Cfa = mild temperate, fully humid, hot summer; Csb = mild temperate, dry summer, warm summer; Bsh = dry, dry summer, hot arid bUtility A reclaimed water treatment plants feed into two isolated distribution systems (A1 and A2), while the entire municipality is serviced by a single potable water distribution system cUtility B plants feed into 1 combined distribution system dFree chlorine was dosed but water chemistry data indicates that total chlorine >> free chlorine, thus free chlorine reacted with ambient ammonia and resulted in NH2Cl as the primary form of disinfectant residual (Table S2)

Quantification of OPs

OP gene copy numbers were quantified in triplicate reactions from DNA extracts using qPCR with published protocols for 16S rRNA genes,24 Legionella spp. (23S rRNA),25 Mycobacterium spp. (16S rRNA),26 P. aeruginosa (ecfX and gyrB),27 Acanthamoeba spp. (18S rRNA),28 and N. fowleri (internal transcribed spacer region).29 With the exception of N. fowleri, all protocols were validated for specificity in environmental matrices in a prior study.21 The specificity of the N. fowleri assay was confirmed by cloning and sequencing of qPCR products from a cross-section of positive samples (Table S1). In order to identify an optimized dilution for consistently minimizing the effect of PCR inhibition, a subset of DNA extracts was initially analyzed at dilutions of 1:5, 1:10, 1:20, and 1:50, with a dilution factor of 1:10 found to yield optimum quantitation across extracts and qPCR assays. A triplicate negative control and triplicate standard curves of ten-fold serial diluted standards of each target gene ranging from 101 to 107 gene copies/µl were included on each 96-well plate. Plates that yielded quantifiable data for negative wells were reanalyzed to exclude any results possibly impacted by contamination. The limit of quantification was established as the lowest standard that amplified in triplicate in each run, and was equivalent to 10 gene copies per milliliter of bulk water and 103 gene copies per biofilm swab.

127

16S rRNA gene amplicon sequencing

Bacterial community compositions were profiled using gene amplicon sequencing with barcoded primers (515F/806R) targeting the V4 region of the 16S rRNA gene.30,31 Triplicate PCR products were composited and 240 ng of each composite was combined and purified using a QIAquick PCR Purification Kit (Qiagen, Valencia, CA). Sequencing was conducted at the Genomics Research Laboratory at the Biocomplexity Institute of Virginia Tech (BI; Blacksburg, VA) on an Illumina MiSeq using a 250-cycle paired-end protocol. Reads were processed using the QIIME pipeline32 and annotated against the Greengenes database33 (May 2013 release). Samples were rarefied to 10,000 randomly selected reads. Field, filtration, DNA extraction blanks, and a least one PCR blank per lane were included in the analysis.

Shotgun metagenomic sequencing

Shotgun metagenomic sequencing was conducted on the POE and greatest water age POU samples from each system on each collection date. Select potable samples were also sequenced. Libraries were prepared using Nextera XT (Illumina, San Diego, CA) and sequenced on an Illumina HiSeq 2500 using a 100-cycle paired-end protocol at BI. Samples were uploaded to the metagenomics RAST server (MG-RAST) and annotated against the RefSeq database using default parameters.34 Metagenomes are publicly accessible under the sample IDs listed in Table S2.

Statistical Analyses

Spearman’s rank sum correlation coefficients were calculated in JMP (SAS, Cary, NC) to assess correlations between OPs, water quality parameters, phyla, and corrosion bacteria using a significance cutoff of α=0.05. A Wilcoxon rank sum test for multiple comparisons was applied in JMP to determine differences between abundances of OPs across groups of samples. Unweighted UniFrac distances generated in QIIME were imported into PRIMER-E (version 6.1.13) for one- way analysis of similarities (ANOSIM) to determine taxonomic differences between groups of samples.

RESULTS AND DISCUSSION

Overview of surveyed distribution systems

The four reclaimed water distribution systems represented a range of U.S. geographic regions, climate zones, treatment schemes, and disinfectant types (Table 6-1). All utilities are located in climate zones that are warm seasonally or year-round and thus were candidates for potential regrowth of OPs, which generally prefer warmer water.19 Utility A used monochloramine as disinfectant residual, while all other utilities primarily used free chlorine. All potable water was derived from a combination of surface and groundwater sources. All utilities utilized advanced wastewater treatment to produce a relatively high quality finished product for distribution for the purposes of non-potable reuse.

128

Physicochemical water characteristics

The physicochemical water quality characteristics of distribution system samples (Table S3) suggested that, with the exception of Utility C, water in the reclaimed systems was warmer than the corresponding potable system, but only Utilities A and B were significantly warmer (p≤0.0321). TOC, DOC, and BDOC were consistently greater in reclaimed water than potable water (p≤0.0038). Average BDOC concentrations ranged from 2,137 to 6,094 ppb in reclaimed water and 15 to 1,522 ppb in potable water. The BDOC concentrations in reclaimed water were comparable to those reported in previous surveys of reclaimed water distribution systems, which ranged from 400 to 6,300 ppb BDOC.18,35

Turbidity and conductivity were also elevated in reclaimed systems (p≤0.0002). Dissolved oxygen ranged from 5.5 to 7.7 mg/L on average in potable systems and 4.0 to 6.8 mg/L in reclaimed systems. Average total chlorine ranged from 0.7 to 3.5 mg/L in potable systems and 0.3 to 2.7 mg/L in reclaimed systems. In reclaimed systems, where free chlorine was typically dosed for the purpose of serving as a secondary disinfectant residual, in reality it was susceptible to conversion to ambient chloramine residual because of reaction with elevated ammonia in the water. Total chlorine was significantly lower at POU sites than at the POE for all systems except Utility B (p≤0.0380), indicating decay of disinfectant residual. Distance from the POE to the POU, temperature, and TOC have all been identified as important factors contributing to enhanced decay of disinfectant residual in reclaimed systems.36

Occurrence of OP Gene Markers

The copy numbers of gene markers corresponding to five target OPs that are commonly problematic in potable water distribution systems;37–40 Legionella spp., Mycobacterium spp., P. aeruginosa, Acanthamoeba spp., and N. fowleri and 16S rRNA genes were determined via qPCR (Table 2). Given that qPCR provides an upper bound estimate of actual viable OPs, qPCR measurements are hereafter referred to in terms of abundance of their corresponding marker genes (i.e., gene copy numbers). Legionella spp., Mycobacterium spp., and 16S rRNA genes were more abundant in reclaimed than potable water distribution systems (p≤0.0001). In particular, Legionella spp. genes were widely detected in reclaimed water, ranging from 76-89% of samples from each utility being positive at an average of 3.4-4.4 log gene copies per milliliter. Legionella spp. genes were also widespread in Utility A’s potable water distribution system, with 80% of samples positive, though the average abundance was only 1.7 log gene copies per milliliter. Mycobacterium spp. genes were abundant in reclaimed water, with 59-79% of samples positive and average levels ranging from 2.5-3.7 log gene copies per milliliter. P. aeruginosa genes were more abundant in potable systems (p=0.0003), with up to 15% of samples positive from Utility B, but no more than 5% of samples positive from any reclaimed systems. N. fowleri genes were also notably widespread in Utility A’s potable (41% positive) and reclaimed (45% positive) distribution system samples, as well as Utility D’s potable samples (45% positive), though at relatively low abundances (2.1, 1.8, 1.3 log gene copies per milliliter on average, respectively). Although N. fowleri has been previously isolated from tap water, information is not available about the numbers of N. fowleri present in municipal water systems.41,42 It is notable that the frequency of detection of Legionella spp., Mycobacterium spp. and N. fowleri genes was generally highest in Utility A’s potable system, which was the sole utility employing monochloramine as the secondary

129

disinfectant residual, whereas the others all utilized free chlorine. Maintaining a free chlorine residual of at least 0.2 mg/L has been proposed as a key strategy for control of N. fowleri.14 Disinfectant residual type may be an important factor influencing regrowth of these OPs.

In a culture-based survey of four reclaimed distribution systems, Jjemba et al. found average log colony forming units per milliliter ranging from 0.6-1.9 for Legionella spp., 0.16-3.21 for Mycobacterium spp., and 0.001-0.009 for Pseudomonas spp.18 Though Jjemba et al. also found Legionella and Mycobacterium to be widespread in reclaimed water systems, concentrations were notably lower than those observed in the present study. However, it is to be expected that molecular tools provide an upper end estimate of pathogens, since they do not directly differentiate viable versus non-viable cells, while culture-based methods provide a lower end estimate, given that they do not capture viable but non-culturable (VBNC) cells. Legionella spp. commonly enter a VBNC state in water systems, which may relate to their characteristic oligotrophic status, given that VBNC is commonly induced by nutrient starvation.43 Previous studies have demonstrated that Legionella spp., Mycobacterium spp., and P. aeruginosa are all capable of entering a VBNC state, while culturable Legionella spp. CFU can be as much as two orders of magnitude less than corresponding viable cell estimates.44–46

While there were no significant correlations among the different OPs in potable bulk water, Legionella spp. and Mycobacterium spp. genes were positively correlated with each other in reclaimed bulk water (ρ=0.4581, p<0.0001), as were N. fowleri and Acanthamoeba spp. genes (ρ=0.3357, p=0.0011). Legionella spp. were negatively correlated with both N. fowleri and Acanthamoeba spp. genes in reclaimed bulk water (ρ=-0.3771, -0.2517; p≤0.0155).

Occurrence of OP Gene Markers in Biofilms

Assuming a pipe diameter of 4 in, within each 1 ft length of pipe, the potable systems harbored on average 7.89–8.59 log 16S rRNA genes in the biofilm and 5.69–7.59 log 16S rRNA genes in the bulk water. The same pipe segment in the reclaimed systems on average harbored 8.49–10.39 log 16S rRNA genes in the biofilm and 8.99–9.79 log 16S rRNA genes copies in the bulk water. Therefore, biofilms harbored the majority of the microbial community compared to bulk water in both potable and reclaimed systems. Based on qPCR, Legionella spp. and 16S rRNA genes were more abundant in reclaimed than potable biofilms (p<0.0003). Legionella spp. genes were nearly ubiquitous in Utility A’s reclaimed system biofilm, with 97% of samples positive at an average of 3.6 log gene copies per swab. Legionella spp. genes were also prevalent in Utility B’s reclaimed system, with 83% of swabs positive. Mycobacterium spp. genes were frequently detected in the reclaimed biofilms with 50-55% of samples positive, compared to 25-53% in potable systems. N. fowleri-specific genes were present in 6-33% of samples in reclaimed systems and 20-28% of potable samples. Acanthamoeba spp. genes were present in 12-22% reclaimed biofilm samples and 5-23% of potable samples.

130

Table 6-2: Frequency of qPCR detectiona for 16S rRNA and opportunistic pathogen genes in potable and reclaimed bulk water and biofilm samples at the point of entry (POE) and point of use (POU); average (standard deviation) of values above the limit of quantification

It is generally thought that biofilms are critical to proliferation of OPs in potable distribution and domestic plumbing systems.37,50,51 OPs are rarely a concern in water exiting the treatment plant, but are believed to grow and accumulate in biofilms. Biofilms provide protection from physical and chemical disruption, such as chlorine disinfection, and can facilitate ecological interactions, such as predator-prey-parasitic relationships between bacteria and amoebae.19 Though biofilms can be protective to its inhabitants, monochloramine has been found in some studies to be more effective at penetrating iron pipe biofilms and scales than free chlorine, thus the unintentional formation of ambient chloramine observed in this study could actually have positive consequences for biofilm control.36,52,53 However, chloramine can also sometimes undergo more rapid decay than chlorine because of nitrification, which can be especially problematic in warmer climates.54

The higher nutrient content of the reclaimed waters studied herein is a key difference relative to potable water systems and could hypothetically increase OPs regrowth. Consistent with these higher nutrient levels, Legionella spp. genes were more widespread in the biofilms of reclaimed systems compared to corresponding potable systems. On the other hand, Mycobacterium spp. gene markers were frequently, but similarly detected both in Utility A’s potable (53%) and reclaimed (55%) biofilms, though they were more common in Utility B’s reclaimed (50%) biofilm than

131

potable (25%) biofilm, suggesting that other factors could be at play or that a very low nutrient threshold applies to mycobacteria. Gene markers for P. aeruginosa, which is often thought of as a “model” biofilm organism, were relatively rare in both the potable and reclaimed biofilms. Little is known about the role of biofilms in supporting N. fowleri growth in pipe systems, but it has been hypothesized that the amoeba would be well suited for growth in biofilms due to the abundance of bacteria and other particulate matter, which offer both a food source and protection from disinfection.37 Neither N. fowleri nor Acanthamoeba spp. gene copy numbers were significantly different in the reclaimed relative to the other potable systems, suggesting that these amoebae were not particularly sensitive to the differing nutrient levels of these two water types.

In potable water biofilms, Mycobacterium spp. genes were positively correlated with both Legionella spp. and Acanthamoeba spp. genes (ρ=-0.2944, -0.4098; p≤0.0213). In reclaimed biofilms, there were no significant positive correlations among the OPs, but Legionella spp. genes were negatively correlated with Acanthamoeba spp. and P. aeruginosa genes (ρ=-0.4719, -0.3010; p≤0.0356).

The negative correlations between Legionella spp. and amoebae genes in both reclaimed bulk water and biofilms were surprising, since it is widely understood that infection of free-living amoeba hosts is essential for L. pneumophila replication in potable water systems.47 The digestive vacuole environment inside the amoebae is rich in amino acids, such as L-cysteine, which is a critical carbon source for L. pneumophila.48,49 Free-living amoebae are further thought to enhance virulence of L. pneumophila and protect it from disinfection, though they can survive for extended periods of time in absence of amoebae.47 It is possible that L. pneumophila is less dependent on amoebae hosts for replication under the higher nutrient environment of reclaimed water. Legionella spp. and Acanthamoeba spp. genes were also not correlated in the potable water distribution system biofilms (ρ=0.2173, p=0.0925). Overall, these results are suggestive of complex relationships among Legionella and free-living amoebic hosts, such as predator-prey cycling, deposition of persistent Legionella spp. from the bulk water, or alternate free-living amoebic hosts that were not monitored in these systems.

Exploration of other potential OPs using Shotgun Metagenomics

While five OPs were targeted for quantification across all samples, shotgun metagenomics was applied to a subset of samples to explore the occurrence of other potential OPs of concern (Figure 6-1). Genes annotated as Acanthamoeba castellanii, Acinetobacter baumannii, , Burkholderia pseudomallei, L. pneumophila, M. avium, Staphylococcus aureus, and Stenotrophomonas maltophilia were detected in all potable and reclaimed systems from which DNA sequence data was available, while Aspergillus fumigatus, Acanthamoeba polyphaga, and N. fowleri were not annotated in any samples. The majority of Legionella spp. and Mycobacterium spp. infections in the clinic are identified as L. pneumophila and M. avium,55,56 respectively, species which were detected via metagenomics in all samples. The only species of Acanthamoeba spp. detected in the metagenomic analysis was A. castellanii, a species of the amoeba capable of causing an eye infection known as amebic keratitis.13 The absence of detection of N. fowleri in the metagenomics data may indicate that abundances were too low for detection, that the DNA extraction method was not optimal for these amoebae, or that the reference databases are poorly suited for characterization of amoebae. DNA fragments identified as matching with S. aureus were

132

consistently present at low relative abundances in both potable and reclaimed water, ranging from -4.35 – -3.22 log hits per total reads. A. baumannii (-3.41 – -1.75 log hits/reads), A. hydrophilia (- 3.65 – -1.65 log hits/reads), B. pseudomallei (-2.99 – -2.23 log hits/reads), and S. maltophilia (- 3.31 – -2.24 log hits/reads) gene fragments were also consistently detected. Where comparisons were available, there was generally strong agreement in trends between qPCR and metagenomic data. Significant correlations were noted when comparing the two methods for abundances of both Legionella spp. (휌=0.7409, p<0.0001) and Mycobacterium spp. gene markers (휌=0.4809, p=0.0062) (Figure S1).

Figure 6-1: Average relative abundance of DNA fragments matching additional OPs of interest identified via shotgun metagenomic sequencing (hits / total reads). All samples are from the POE and the POU with the greatest water age from each sampling event were submitted for sequencing. Many samples, potable in particular, did not pass the library preparation step due to low DNA yield. Those that passed library preparation and were successfully sequenced are included here.

Relationship between abundance of OPs, water age, and related factors

Because distribution system models were not consistently available for the reclaimed distribution systems, ranked water age estimated based on location in the distribution systems and estimated flow paths was used as a proxy for water age. 16S rRNA genes were found to be more abundant at the POU than the POE in reclaimed systems (p=0.0268), supporting the hypothesis of general bacterial regrowth in these distribution systems. POE versus POU differences in 16S rRNA gene abundances were not noted in the potable systems (p=0.8323). Significant regrowth from the POE to the POU was observed for Legionella spp. genes in Utility A’s reclaimed system (p=0.0089), but regrowth from the POE to the POU was not observed for any of the other OPs (p≥0.0798) in any potable or reclaimed systems. Surprisingly, 16S rRNA genes, Acanthamoeba

133

Table 6-3: Spearman’s rank correlation coefficients for correlations between 16S rRNA or opportunistic pathogen gene markers and physicochemical water quality parameters. Correlations are for all potable and reclaimed water samples. P-values indicated in parentheses. Significant correlations indicated in bold.

spp., and Mycobacterium spp. genes decreased with ranked reclaimed water age (p≤0.0325) (Table 6-3).

While none of the OPs monitored consistently correlated with water age across the four utilities, there were several notable instances where they did correlate in reclaimed systems (Figure 6-2A-2C). For example, samples collected from Utility A’s two reclaimed distribution systems displayed regrowth of total bacteria, indicated by 16S rRNA genes, as well as Legionella spp. and Mycobacterium spp. genes (Figures 6-2A-2B). 16S rRNA genes were consistently at least 1-log higher in abundance at POU than at POE sites in both systems. In the A1 system, Legionella spp. genes increased by at least 0.5-log at POU sites compared to the POE. In the A2 distribution system, Legionella spp. genes were at least 1-log greater at all POU sites than at the POE and Mycobacterium spp. genes also increased at all POU sites relative to the POE. Total bacterial regrowth was observed for Utility C (Figure 6-2C) at some sites, particularly when total chlorine dropped. OP regrowth was not observed from the POE to the POUs; however there was an apparent inverse relationship between chlorine residual and Legionella spp. and Mycobacterium spp. gene abundances at the fourth and fifth sites. While qPCR captures DNA from dead or lysed cells together with that from live cells, the elevated gene markers from the POE to the POUs in these examples are strongly indicative of regrowth. Regrowth was not observed for Utility D (Figure 6- 2D); however, chlorine residual decayed rapidly from 4.1 mg/L at the POE to 0.42 mg/L or less at all POUs. This low disinfectant residual likely permitted both 16S rRNA and Legionella spp. genes to remain abundant throughout the Utility D distribution system. While temperature was not directly correlated with OP regrowth when considering the sample pool as a whole, there were

134

several spikes in temperature that also coincided with an increase in Legionella spp. and Mycobacterium spp. gene abundance (Figure 6-2A-2B).

Figure 6-2: Temperature, free chlorine, and abundances of 16S rRNA genes, Legionella spp., and Mycobacterium spp. in select distribution systems. Samples collected at the treatment plant and throughout the reclaimed distribution systems for key examples of scenarios where Legionella spp. or Mycobacterium spp. were observed. Examples are from (A) Utility A1 sampled August 2014, (B) Utility A2 sampled August 2014, (C) Utility B sampled August 2014, and (D) Utility C sampled October 2014. When modeling data was available from utilities, water age was estimated. In the absence of water age data, pipe distances from the treatment plant were used as a proxy for relative water age. When neither metric was available, water utility staff provided ranked water age data approximating relative water age between site

Relationship between water chemistry measurements and abundance of OPs

Correlations between OPs and physicochemical water quality parameters are presented in Tables 3 and S4. Temperature appeared to be an important factor facilitating regrowth in reclaimed water, as genes associated with 16S rRNA, Acanthamoeba spp., Mycobacterium spp., and N. fowleri were all positively correlated with temperature in bulk water samples (p≤0.0130). In reclaimed biofilms, genes associated with 16S rRNA, Legionella spp., and N. fowleri all correlated

135

with temperature (p≤0.0032). Total chlorine appeared to have a controlling effect on total bacteria in reclaimed water, as indicated by a negative correlation between total chlorine and 16S rRNA genes (p≤0.0185) in both the bulk water and biofilm. With the exception of Legionella spp. and Mycobacterium spp. genes in Utility B’s reclaimed water (Table S4), this correlation did not consistently apply to any OPs, however, suggesting that the tendency of OPs to resist chlorine disinfection is an important factor in providing a selective advantage over the general bacterial community. In reclaimed systems, BDOC was only significantly correlated with Acanthamoeba spp. genes (p<0.0001) in the bulk water. Previous studies of potable water have indicated that concentrations well below 10 ppb are required in order for organic carbon to be a limiting nutrient and constrain regrowth of bacteria in distribution systems,57,58 which may be an unrealistic goal for reclaimed water distributors. Mycobacterium spp. and Legionella spp. genes in reclaimed systems were both positively correlated with phosphorus, while Legionella spp. and 16S rRNA genes were correlated with ammonia, further suggesting that nutrients other than carbon are worthy of future examination as alternative limiting nutrients.

Microbial ecology – OP associations

Based on jackknifed unweighted UniFrac distances (a measure of beta diversity), the microbial community composition of the potable distribution system was distinct from that of reclaimed systems for both bulk water (ANOSIM, R=0.426, p≤0.001) and biofilm samples (R=0.317, p≤0.001) (Figure 6-3). At the POEs, potable and reclaimed bulk water were also unique (R=0.396, p=0.002). The bulk water versus biofilm communities were distinct for both potable (R=0.167, p≤0.001) and reclaimed systems (R=0.364, p≤0.001). Alphaproteobacteria, Deltaproteobacteria, TM6, Verrucomicrobia, and Chlamydiae were enriched in the reclaimed systems (Wilcoxon multiple comparisons, p≤0.0474), while , , Actinobacteria, Cyanobacteria, Firmicutes, and Nitrospirae were enriched in potable samples (p≤0.0032). Gammaproteobacteria, Planctomycetes, and [Thermi] were enriched in potable bulk water samples (p≤0.0191), while and Chlamydiae were enriched in the biofilm (p≤0.0007). In reclaimed systems, Epsilonproteobacteria, Firmicutes, OD1, TM6, and Fusobacteria were enriched in the bulk water (p≤0.0093), while Deltaproteobacteria, Acidobacteria, Gemmatimonadetes, Nitrospirae, Planctomycetes, and Verrucomicrobia were enriched in the biofilm (p≤0.0129).

OPs, such as Legionella, are known to competitively, antagonistically, and even symbiotically interact with other microbes.59–61 To identify potential microbial interactions of interest, correlations were examined between OPs and the phyla detected (Table S5). Several phyla exhibited strong positive correlations with Legionella spp. gene markers: WPS-2, TM7, NKB19, SR1, Lentisphaerae, Tenericutes, and OP11 (Spearman’s ρ=0.3108-0.4913, p≤0.0001). Cyanobacteria, NC10, Nitrospirae, Caldithrix, and [Parvarchaeota] (ρ=-0.1499-(-0.2158), p≤0.0125) exhibited the strongest negative correlations with Legionella spp. gene markers. Tenericutes, SR1, Lentisphaerae, NKB19, WWE1, WPS-2, Euryarchaeota, OP11, TM6, and OP8 (ρ=0.2304-0.2992, p≤0.0001) exhibited the strongest positive correlations with Mycobacterium spp. genes, while no phyla significantly negatively correlated with the gene. Weak correlations were also noted between P. aeruginosa, Acanthamoeba spp., and N. fowleri genes and various phyla (Table S5). Microbial communities were also characterized in PCR blanks (n=5), field blanks (n=8), extraction blanks (n=2), and a filtration blank (n=1) (Figure S2). Blanks yielded

136

microbial community phylum profiles that were significantly different from reclaimed samples (ANOSIM, R=0.401, p=0.0050), but not potable samples (R=0.064, p=0.1540). PCR blanks (p=0.0009), extraction blanks (p=0.0383), and filter blanks (n=1; statistical analysis not possible) all yielded less reads than samples. While field blanks did not produce significantly less reads than samples (p=0.0886), greater volumes of PCR product were pooled to achieve the minimum 240 ng mass from each sample from field blanks (median 15.1 µl) than from samples (median 12.3 µl). Given that more product was pooled to achieve sufficient DNA for sequencing of negative controls, it is likely that the inherent “noise” associated with this method did not overwhelm the microbial communities profiled for samples, but these data should be interpreted with caution. Additional research is needed to determine the accuracy of trends observed herein.

Figure 6-3: Microbial community composition of potable vs. reclaimed distribution system samples. Phyla (and proteobacteria classes) comprising at least 1% of any sample from potable and reclaimed water distribution system (POU) samples from utilities A, B, C, and D. Samples appear in chronological order from left to right according to sample date.

Corrosion-Associated Microbial Activity Assays

BARTs can provide insight into activity of various functional groups of microbes of interest and are commonly employed to assess microbially influenced corrosion (MIC).62–64 The development of redox gradients in distribution systems can create a range of dissolved oxygen levels, lack of chlorine residual, and varying nitrogen species that can create a range of microbial niches in terms of available electron donors and acceptors and result in undesirable consequences to water quality, including MIC.63 Corrosion of pipes may facilitate the growth of OPs by releasing iron, an important micronutrient for the growth of Legionella,65 by increasing surface area where biofilms may establish,66 and consuming chlorine.64,67 Additionally, corrosion tubercles have been shown to harbor high densities of coliforms, with some indication that the same may be true of OPs.64,68 The approximate abundances of nitrifying bacteria, denitrifying bacteria, and sulfate- reducing bacteria were determined using BARTs (Table S6). Nitrifying bacteria were only 137

detected in potable water of Utility A, but were detected in the reclaimed water of utilities A, C, and D, all systems that tended to contain ambient ammonia. Denitrifying bacteria were detected in both the potable and reclaimed water of all utilities. Sulfate-reducing bacteria were only detected in the potable water of Utility C, but were detected in all reclaimed waters.

Nitrifying bacteria were positively correlated with Legionella spp. genes in both bulk water and biofilm as well as Mycobacterium spp. genes in bulk water (ρ=0.5652, 0.3786, 0.4498; p≤0.0002). Nitrifying bacteria have been widely documented in potable and reclaimed water systems due to the availability of ammonia in systems utilizing chloramines as a disinfectant residual.54,62,69,70 Nitrification has been previously associated with regrowth of OPs due to rapid decay of disinfectant residual, particularly chloramines.71 Denitrifying bacteria were correlated with biofilm N. fowleri genes (ρ=0.3607, p=0.0034). Although denitrification has only rarely been documented in potable distribution systems,72,73 denitrifying bacteria thrive under conditions of low dissolved oxygen and high organic matter,74,75 which were often observed in the reclaimed water systems surveyed in this study. Although N. fowleri typically prefer oxygen-rich environments, other Naegleria species are able to survive under relatively low oxygen concentrations,76 so they may find a potential niche in the low oxygen conditions required for denitrification. One previous study noted a negative correlation between dissolved oxygen and N. fowleri in surface water.77 Sulfate-reducing bacteria were positively correlated with Legionella spp. genes in both bulk water and biofilm, and Mycobacterium spp. genes in bulk water (ρ=0.5307, 0.6991, 0.4152; p≤0.0001). Sulfate-reducing bacteria and mycobacteria have been previously identified together as the dominant microbial groups inhabiting a biofilm in a chloraminated potable water system.78

Implications for OP control in reclaimed distribution systems

While OP-related illnesses are likely largely underreported,50 exposures associated with potable water are a leading cause of waterborne disease in developed countries.5 The potential disease burden associated with OP exposure resulting from reclaimed water use has not been characterized, with this study revealing that OP genes (specifically, Legionella spp. and Mycobacterium spp.) were more abundant in the reclaimed water systems than corresponding potable systems surveyed. This work demonstrates that growth of OPs in reclaimed water is strongly tied to the unique water chemistry and microbial ecology of reclaimed water. However, the vast majority of knowledge about regrowth of OPs in distribution systems and premise plumbing is based on understanding of potable water systems, which, as suggested by this study, are not likely directly translatable. This study clearly demonstrates that reclaimed water generally differs from potable water in many aspects (nutrient concentration, temperature, etc.), with differences in corresponding OP occurrence patterns. Additionally, the microbial community composition in reclaimed distribution systems is unique from that of potable systems, with even greater differences observed as the water is transported through the system. Differences in associations between Legionella spp. genes and amoebic hosts demonstrate that interactions between members of the microbial community are unique in potable versus reclaimed water. Therefore, traditional knowledge about the behavior of OPs in potable distribution systems and premise plumbing is not necessarily applicable to reclaimed water.

138

There were several key limitations of this study. Only four systems were studied with highly variable treatment approaches, so the observed trends may not be representative of all systems. In addition, given the complexities of studying full-scale systems, many factors that could contribute to the occurrence of OPs could not be quantitatively considered, for example, usage patterns, climate impacts, and historical treatment, disinfection, and operation patterns. Finally, the large number of statistical comparisons made in this study warrant critical consideration of the comparisons made. More research is needed to better document and ultimately understand the complex factors governing OP behavior in reclaimed water systems. In potable water, there are no regulatory requirements for monitoring OPs, with the focus primarily being on ingestion of fecal pathogens. Thus, serious consideration is needed regarding regulatory and monitoring requirements for OPs both in reclaimed and potable water systems, given that the most relevant routes of exposure are overlooked in the current regulatory paradigm, or lack thereof.4 Research is needed to further advance hazard identification, exposure assessment, and establish the infectious dose of these pathogens. Comprehensive risk assessment is needed to better understand the potential impact to public health associated with transmission of OPs via use of reclaimed water.

ACKNOWLEDGEMENTS

We thank the participating utilities for conducting sampling and on-site data collection. This work is supported by the National Science Foundation (NSF) Graduate Research Fellowship Program Grant (DGE 0822220) and NSF Collaborative Research grant (1438328), The Alfred P. Sloan Foundation Microbiology of the Built Environment (MoBE) program, the Water Environment & Research Foundation Paul L. Busch award, the Virginia Tech Institute for Critical Technology and Applied Science Center for Science and Engineering of the Exposome, and the American Water Works Association Abel Wolman Doctoral Fellowship.

REFERENCES

(1) Gosling, S. N.; Arnell, N. W. A global assessment of the impact of climate change on water scarcity. Clim. Change 2016, 134 (3), 371–385. (2) Guidelines for Water Reuse; United States Environmental Protection Agency: 2012. (3) Jjemba, P.; Johnson, W.; Bukhari, Z.; LeChevallier, M. Review of the leading challenges in maintaining reclaimed water quality during storage and distribution. J. Water Reuse Desalin. 2014, 4 (4), 209–1237. (4) Garner, E.; Zhu, N.; Strom, L.; Edwards, M.; Pruden, A. A human exposome framework for guiding risk management and holistic assessment of recycled water quality. Environ. Sci. Water Res. Technol. 2016, 2, 580–1598. (5) Brunkard, J. M.; Ailes, E.; Roberts, V. A.; Hill, V.; Hilborn, E. D.; Craun, G. F.; Rajasingham, A.; Kahler, A.; Garrison, L.; Hicks, L.; Carpenter, J.; Wade, T. J.; Beach, M. J.; Yoder, M. Surveillance for waterborne disease outbreaks associated with drinking water – United States, 2007–2008. Morb. Mortal. Wkly. Rep. 2008, 57 (SS-9), 1–38. (6) Craun, G. F.; Brunkard, J. M.; Yoder, J. S.; Roberts, V. A.; Carpenter, J.; Wade, T.; Calderon, R. L.; Roberts, J. M.; Beach, M. J.; Roy, S. L. Causes of outbreaks associated with drinking water in the United States from 1971 to 2006. Clin. Microbiol. Rev. 2010, 23 (3), 507–528.

139

(7) Brissaud, F.; Blin, E.; Hemous, S.; Garrelly, L. Water reuse for urban landscape irrigation: aspersion and health related regulations. Water Sci. Technol. 2008, 57 (5), 781–787. (8) Bodey, G. P.; Bolivar, R.; Fainstein, V.; Jadeja, L. Infections Caused by Pseudomonas aeruginosa. Rev. Infect. Dis. 1983, 5 (2), 279–313. (9) Primm, T. P.; Lucero, C. A.; Falkinham, J. O. Health Impacts of Environmental Mycobacteria. Clin. Microbiol. Rev. 2004, 17 (1), 98–106. (10) Williams, B. J.; Dehnbostel, J.; Blackwell, T. S. Pseudomonas aeruginosa: Host defence in lung diseases. Respirology 2010, 15 (7), 1037–1056. (11) Fraser, D. W.; Tsai, T. R.; Orenstein, W.; Parkin, W. E.; Beecham, H. J.; Sharrar, R. G.; Harris, J.; Mallison, G. F.; Martin, S. M.; McDade, J. E.; Shepard, C. C.; Brachman, P. S. Legionnaires’ Disease: Description of an Epidemic of Pneumonia. N. Engl. J. Med. 1989, 297 (22), 1189–11197. (12) Horsburgh, C. R. Mycobacterium avium complex infection in the acquired immunodeficiency syndrome. N. Engl. J. Med. 1991, 324 (19), 1332–1338. (13) Marciano-Cabral, F.; Cabral, G. Acanthamoeba spp. as agents of disease in humans. Clin. Microbiol. Rev. 2003, 16 (2), 273–307. (14) Bartrand, T. A.; Causey, J. J.; Clancy, J. L. Naegleria fowleri: An emerging drinking water pathogen. J. Am. Water Works Assoc. 2014, 106 (10), E418–E432. (15) Wu, Q. Y.; Li, Y. Y.; Hu, H. Y. Exposure dose of reclaimed water for car wash. Acta Sci. Circumstantiae 2013, 33 (3), 844–849. (16) Borboudaki, K. E.; Paranychianakis, N. V.; Tsagarakis, K. P. Integrated wastewater management reporting at tourist areas for recycling purposes, including the case study of Hersonissos, . Environ. Manage. 2005, 36 (4), 610–623. (17) Gerba, C. P.; Wallis, C.; Melnick, J. L. Microbiological hazards of household toilets: droplet production and the fate of residual organisms. Appl. Microbiol. 1975, 30 (2), 229– 237. (18) Jjemba, P. K.; Weinrich, L. A.; Cheng, W.; Giraldo, E.; LeChevallier, M. W. Regrowth of Potential Opportunistic Pathogens and Algae in Reclaimed-Water Distribution Systems. Appl. Environ. Microbiol. 2010, 76 (13), 4169–4178. (19) Falkinham, J. Common Features of Opportunistic Premise Plumbing Pathogens. Int. J. Environ. Res. Public Health 2015, 12 (5), 4533–4545. (20) Thomas, J. M.; Ashbolt, N. J. Do free-living amoebae in treated drinking water systems present an emerging health risk? Environ. Sci. Technol. 2011, 45 (3), 860–869. (21) Wang, H.; Edwards, M.; Falkinham, J. O.; Pruden, A. Molecular Survey of the Occurrence of Legionella spp., Mycobacterium spp., Pseudomonas aeruginosa, and Amoeba Hosts in Two Chloraminated Drinking Water Distribution Systems. Appl. Environ. Microbiol. 2012, 78 (17), 6285–6294. (22) Servais, P.; Billen, G.; Hascoet, M-C. Determination of the Biodegradable Fraction of Dissolved Organic Matter in Waters. Water Res. 1995, 21 (4), 445 – 450. (23) American Public Health Association (APHA). Standard Methods for the Examination of Water and Wastewater; American Water Works Association, Water Environment Federation, 2005. (24) Suzuki, M. T.; Taylor, L. T.; DeLong, E. F. Quantitative analysis of small-subunit rRNA genes in mixed microbial populations via 5’-nuclease assays. Appl. Environ. Microbiol. 2000, 66 (11), 4605–4614.

140

(25) Nazarian, E. J.; Bopp, D. J.; Saylors, A.; Limberger, R. J.; Musser, K. A. Design and implementation of a protocol for the detection of Legionella in clinical and environmental samples. Diagn. Microbiol. Infect. Dis. 2008, 62 (2), 125–132. (26) Radomski, N.; Lucas, F. S.; Moilleron, R.; Cambau, E.; Haenn, S.; Moulin, L. Development of a real-time qPCR method for detection and enumeration of Mycobacterium spp. in surface water. Appl. Environ. Microbiol. 2010, 76 (21), 7348–7351. (27) Anuj, S. N.; Whiley, D. M.; Kidd, T. J.; Bell, S. C.; Wainwright, C. E.; Nissen, M. D.; Sloots, T. P. Identification of Pseudomonas aeruginosa by a duplex real-time polymerase chain reaction assay targeting the ecfX and the gyrB genes. Diagn. Microbiol. Infect. Dis. 2009, 63 (2), 127–131. (28) Rivière, D.; Szczebara, F. M.; Berjeaud, J.-M.; Frère, J.; Héchard, Y. Development of a real-time PCR assay for quantification of Acanthamoeba trophozoites and cysts. J. Microbiol. Methods 2006, 64 (1), 78–83. (29) Mull, B. J.; Narayanan, J.; Hill, V. R. Improved Method for the Detection and Quantification of Naegleria fowleri in Water and Sediment Using Immunomagnetic Separation and Real-Time PCR. J. Parasitol. Res. 2013, 608367. (30) Caporaso, J. G.; Lauber, C. L.; Walters, W. A.; Berg-Lyons, D.; Lozupone, C. A.; Turnbaugh, P. J.; Fierer, N.; Knight, R. Global patterns of 16S rRNA diversity at a depth of millions of sequences per sample. Proc. Natl. Acad. Sci. U.S.A. 2011, 108, 4516–4522. (31) Caporaso, J. G.; Lauber, C. L.; Walters, W. A.; Berg-Lyons, D.; Huntley, J.; Fierer, N.; Owens, S. M.; Betley, J.; Fraser, L.; Bauer, M.; Gormley, N.; Gilbert, J. A.; Smith, G.; Knight, R. Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J. 2012, 6 (8), 1621–1624. (32) Caporaso, J. G.; Kuczynski, J.; Stombaugh, J.; Bittinger, K.; Bushman, F. D.; Costello, E. K.; Fierer, N.; Pena, A. G.; Goodrich, J. K.; Gordon, J. I.; Huttley, G. A.; Kelley, S. T.; Knights, D.; Koenig, J. E.; Ley, R. E.; Lozupone, C. A.; McDonald, D.; Muegge, B. D.; Pirrung, M.; Reeder, J.; Sevinsky, J. R.; Turnbaugh, P. J.; , Walters, W. A.; Widmann, J.; Yatsunenko, T.; Zaneveld, J.; Knight, R. QIIME allows analysis of high-throughput community sequencing data. Nat. Methods 2010, 7 (5), 335–336. (33) DeSantis, T. Z.; Hugenholtz, P.; Larsen, N.; Rojas, M.; Brodie, E. L.; Keller, K.; Huber, T.; Dalevi, D.; Hu, P.; Andersen, G. L. Greengenes, a Chimera-Checked 16S rRNA Gene Database and Workbench Compatible with ARB. Appl. Environ. Microbiol. 2006, 72 (7), 5069–5072. (34) Meyer, F.; Paarmann, D.; D’Souza, M.; Olson, R.; Glass, E. M.; Kubal, M.; Paczian, T.; Rodriguez, A.; Stevens, R.; Wilke, A.; Wilkening, J.; Edwards, R. A. The metagenomics RAST server - a public resource for the automatic phylogenetic and functional analysis of metagenomes. BMC Bioinformatics 2008, 9 (386). (35) Weinrich, L. A; Jjemba, P. K.; Giraldo, E.; LeChevallier, M. W. Implications of organic carbon in the deterioration of water quality in reclaimed water distribution systems. Water Res. 2010, 44 (18), 5367–5375. (36) Jjemba, P. K.; Johnson, W.; Bukhari, Z.; LeChevallier, M. W. Occurrence and Control of Legionella in Recycled Water Systems. Pathogens 2015, 4 (3), 470–502. (37) Pruden, A.; Edwards, M. A.; Falkinham, J. O.; Arduino, M.; Bird, J.; Birdnow, R.; Bedard, E.; Camper, A.; Clancy, J.; Hilborn, E.; et al. State of the Science and Research Needs for Opportunistics Pathogens in Premise Plumbing: Methodology, microbial ecology, and epidemiology, Final Report; Water Research Foundation: Denver, CO, 2013.

141

(38) Falkinham, J.; Pruden, A.; Edwards, M. Opportunistic Premise Plumbing Pathogens: Increasingly Important Pathogens in Drinking Water. Pathogens 2015, 4 (2), 373–386. (39) Falkinham, J. Common Features of Opportunistic Premise Plumbing Pathogens. Int. J. Environ. Res. Public Health 2015, 12 (5), 4533–4545. (40) Wang, H.; Bedard, E.; Prevost, M.; Camper, A. K.; Hill, V. R.; Pruden, A. Methodological approaches for monitoring opportunistic pathogens in premise plumbing: A review. Water Res. 2017, 117, 68–86. (41) Puzon, G. J.; Lancaster, J. A.; Wylie, J. T.; Plumb, J. J. Rapid Detection of Naegleria Fowleri in Water Distribution Pipeline Biofilms and Drinking Water Samples. Env. Sci. Technol. 2009, 43 (17), 6691–6696. (42) Esterman, A.; Dorsch, M.; Cameron, S.; Roder, D.; Robinson, B.; Lake, J.; Christy, P. The association of Naegleria fowleri with the chemical, microbiological, and physical characteristics of South Australian water supplies. Water Res. 1984, 18 (5), 549–553. (43) Steinert, M.; Emody, L.; Amann, R.; Hacker, J. Resuscitation of Viable but Nonculturable Legionella pneumophila\rPhiladelphia JR32 by Acanthamoeba castellanii. Appl. Env. Microbiol. 1997, 63 (5), 2047–2053. (44) Oliver, J. D. The viable but nonculturable state in bacteria. J. Microbiol. 2005, 43 (S), 93– 100. (45) Oliver, J. D. Recent findings on the viable but nonculturable state in pathogenic bacteria. FEMS Microbiol. Rev. 2010, 34 (4), 415–425. (46) Moritz, M. M.; Flemming, H. C.; Wingender, J. Integration of Pseudomonas aeruginosa and Legionella pneumophila in drinking water biofilms grown on domestic plumbing materials. Int. J. Hyg. Environ. Health 2010, 213 (3), 190–197. (47) Murga, R.; Forster, T. S.; Brown, E.; Pruckler, J. M.; Fields, B. S.; Donlan, R. M. Role of biofilms in the survival of Legionella pneumophila in a model potable-water system. Microbiology 2001, 147 (11), 3121–3126. (48) Manske, C.; Hilbi, H. Metabolism of the vacuolar pathogen Legionella and implications for virulence. Front. Cell. Infect. Microbiol. 2014, 4, 1–10. (49) Ewann, F.; Hoffman, P. S. Cysteine Metabolism in Legionella pneumophila : Characterization of an L-Cystine-Utilizing Mutant. Appl. Environ. Microbiol. 2006, 72 (6), 3993–4000. (50) Ashbolt, N. Environmental (Saprozoic) Pathogens of Engineered Water Systems: Understanding Their Ecology for Risk Assessment and Management. Pathogens 2015, 4 (2), 390–405. (51) Wingender, J.; Flemming, H.-C. Biofilms in drinking water and their role as reservoir for pathogens. Int. J. Hyg. Environ. Health 2011, 214 (6), 417–423. (52) LeChevallier, M. W.; Cawthon, C. D.; Lee, R. G. Inactivation of biofilm bacteria. Appl. Environ. Microbiol. 1988, 54 (10), 2492–2499. (53) Samrakandi, M. M.; Roques, C.; Michel, G. Influence of trophic conditions on exopolysaccharide production: bacterial biofilm susceptibility to chlorine and monochloramine. Can. J. Microbiol. 1997, 43, 751–758. (54) Zhang, Y.; Love, N.; Edwards, M. Nitrification in Drinking Water Systems. Crit. Rev. Environ. Sci. Technol. 2009, 39, 153–208. (55) Marston, B. J.; Lipman, H. B.; Breiman, R. F. Surveillance for Legionnaires’ Disease: Risk Factors for Morbidity and Mortality. Arch Intern Med 1994, 154 (21), 2417–2422. (56) Johnson, M. M.; Odell, J. A. Nontuberculous mycobacterial pulmonary infections. J.

142

Thorac. Dis. 2014, 6 (3), 210-220. (57) Kooij, D. V. D. Assimilable Organic Carbon as an Indicator of Bacterial Regrowth. J. Am. Water Work. Assoc. 1992, 57–65. (58) Williams, K.; Pruden, A.; Falkinham, J.; Edwards, M. Relationship between Organic Carbon and Opportunistic Pathogens in Simulated Glass Water Heaters. Pathogens 2015, 4 (2), 355–372. (59) Declerck, P.; Behets, J.; Delaedt, Y.; Margineanu, A.; Lammertyn, E.; Ollevier, F. Impact of non-Legionella bacteria on the uptake and intracellular replication of Legionella pneumophila in Acanthamoeba castellanii and Naegleria lovaniensis. Microb. Ecol. 2005, 50 (4), 536–549. (60) Guerrieri, E.; Bondi, M.; Sabia, C.; De Niederhäusern, S.; Borella, P.; Messi, P. Effect of bacterial interference on biofilm development by Legionella pneumophila. Curr. Microbiol. 2008, 57 (6), 532–536. (61) Wadowsky, R. M.; Yee, R. B. Satellite growth of Legionella pneumophila with an environmental isolate of Flavobacterium breve. Appl. Environ. Microbiol. 1983, 46 (6), 1447–1449. (62) Jjemba, P. K.; Weinrich, L. A.; Cheng, W.; Giraldo, E.; LeChevallier, M. W. Guidance document on the microbiological quality and biostability of reclaimed water and distribution systems (WRF‐05‐002), WateReuse Research Foundation: Alexandria, VA, 2010. (63) Masters, S.; Wang, H.; Pruden, A.; Edwards, M. A. Redox gradients in distribution systems influence water quality, corrosion, and microbial ecology. Water Res. 2015, 68, 140–149. (64) Emde, K. M. E.; Smith, D. W.; Facey, R. Initial Investigation of Microbially Influenced Corrosion (MIC) in a Low Temperature Water Distribution System. Wat. Res. 1992, 26 (2), 169–175. (65) Ristroph, J. D.; Hedlund, K. W.; Gowda, S. Chemically defined medium for Legionella pneumophila growth. J Clin Microbiol 1981, 13 (1), 115–119. (66) Yang, F.; Shi, B.; Gu, J.; Wang, D.; Yang, M. Morphological and physicochemical characteristics of iron corrosion scales formed under different water source histories in a drinking water distribution system. Water Res. 2012, 46 (16), 5423–5433. (67) Frateur, I.; Deslouis, C.; Kiene, L.; Levi, Y.; Tribollet, B. Free chlorine consumption induced by cast iron corrosion in drinking water distribution systems. Water Res. 1999, 33 (8), 1781–1790. (68) LeChevallier, M. W.; Babcock, T. M.; Lee, R. G. Examination and Characterization of Distribution-System Biofilms. Appl. Environ. Microbiol. 1987, 53 (12), 2714–2724. (69) Zhang, Y.; Edwards, M. Accelerated chloramine decay and microbial growth by nitrification in premise plumbing. J. Am. Water Work. Assoc. 2009, 101 (11), 51–62. (70) Narahimhan, R.; Brereton, J.; Abbaszadegan, M.; Ryu, H.; Butterfield, P.; Thompson, K.; Werth, H. Characterizing Microbial Water Quality in Reclaimed Water Distribution Systems, American Water Works Association Research Foundation and WateReuse Foundation, 2005. (71) Wang, H.; Masters, S.; Hong, Y. J.; Stallings, J.; Falkinham, J. O.; Edwards, M. A.; Pruden, A. Effect of Disinfectant, Water Age, and Pipe Material on Occurrence and Persistence of Legionella, mycobacteria, Pseudomonas aeruginosa, and Two Amoebas. Environ. Sci. Technol. 2012, 46 (21), 11566–11574. (72) Lee, S. H.; Connor, J. T. O.; Banerji, S. K. Biologically mediated corrosion and its effects

143

on water quality in distribution systems. J. Am. Water Work. Assoc. 2015, 72 (11), 636– 645. (73) Nguyen, C.; Elfland, C.; Edwards, M. Impact of advanced water conservation features and new copper pipe on rapid chloramine decay and microbial regrowth. Water Res. 2012, 46 (3), 611–621. (74) Knowles, R. Denitrification. Microbiol. Rev. 1982, 46 (1), 43–70. (75) Zumft, W. G. Cell biology and molecular basis of denitrification. Microbiol. Mol. Biol. Rev. 1997, 61 (4), 533 ff. (76) Opperdoes, F. R.; De Jonckheere, J. F.; Tielens, A. G. M. Naegleria gruberi metabolism. Int. J. Parasitol. 2011, 41 (9), 915–924. (77) Tyndall, R. L.; Ironside, K. S.; Metler, P. L.; Tan, E. L.; Hazen, T. C.; Fliermans, C. B. Effect of thermal additions on the density and distribution of thermophilic amoebae and pathogenic Naegleria fowleri in a newly created cooling lake. Appl. Environ. Microbiol. 1989, 55 (3), 722–732. (78) Gomez-Smith, C. K.; LaPara, T. M.; Hozalski, R. M. Sulfate Reducing Bacteria and Mycobacteria Dominate the Biofilm Communities in a Chloraminated Drinking Water Distribution System. Environ. Sci. Technol. 2015, 49 (14), 8432–8440.

144

SUPPLEMENTARY INFORMATION FOR CHAPTER 6

Figure S1: Spearman’s rank correlation coefficients were determined between gene marker abundances determined by qPCR (x-axes) and shotgun metagenomic sequencing (y-axes) for (A) Legionella spp. and (B) Mycobacterium spp.

145

Figure S2 – A) and B) read counts for negative controls included in 16S rRNA amplicon sequencing, including read counts for samples for comparison

146

Table S1: BLAST matches for cloned qPCR products for specificity confirmation of N. fowleri assay Sample % Highest Highest Similarity BLAST match (description) Identity Similarity BLAST match (Accession No.)

1 95% KT375442.1 Naegleria fowleri 18S ribosomal RNA gene, partial sequence; internal transcribed spacer 1, 5.8S ribosomal RNA gene, and internal transcribed spacer 2, complete sequence; and 28S ribosomal RNA gene, partial sequence

2 98% KT375442.1 Naegleria fowleri 18S ribosomal RNA gene, partial sequence; internal transcribed spacer 1, 5.8S ribosomal RNA gene, and internal transcribed spacer 2, complete sequence; and 28S ribosomal RNA gene, partial sequence

3 98% KT375442.1 Naegleria fowleri 18S ribosomal RNA gene, partial sequence; internal transcribed spacer 1, 5.8S ribosomal RNA gene, and internal transcribed spacer 2, complete sequence; and 28S ribosomal RNA gene, partial sequence

4 97% KT375442.1 Naegleria fowleri 18S ribosomal RNA gene, partial sequence; internal transcribed spacer 1, 5.8S ribosomal RNA gene, and internal transcribed spacer 2, complete sequence; and 28S ribosomal RNA gene, partial sequence

5 98% KT375442.1 Naegleria fowleri 18S ribosomal RNA gene, partial sequence; internal transcribed spacer 1, 5.8S ribosomal RNA gene, and internal transcribed spacer 2, complete sequence; and 28S ribosomal RNA gene, partial sequence

6 95% KT375442.1 Naegleria fowleri 18S ribosomal RNA gene, partial sequence; internal transcribed spacer 1, 5.8S ribosomal RNA gene, and internal transcribed spacer 2, complete sequence; and 28S ribosomal RNA gene, partial sequence For all sequences, no BLAST matches were found for any other tested Naegleria species (species tested: gruberi, americana, RNG, australiensis, lovaniensis, clarki, italica, polaris, pagei, AG-2012)

147

Table S2: Shotgun metagenomic sequenced samples System Sample Paired-end MG-RAST Matrix Utility Sample Name Type Type reads Sample ID

Drinking biofilm A POU Drinkingwater_biofilm_ UtilityA_S5_L1S10 5,412,528 mgs295574

Drinking bulk water A POU Drinkingwater_bulkwater_ UtilityA_S5_L1S6 5,887,968 mgs295565

Drinking bulk water B POE Drinkingwater_bulkwater_ UtilityB_S0_1 38,668,577 mgs458886

Drinking bulk water B POE Drinkingwater_bulkwater_ UtilityB_S0_1_duplicate 32,868,754 mgs458889

Reclaimed bulk water A Influent Reclaimedwater_bulkwater_UtilityA_-1_1 7,132 mgs458922

Reclaimed bulk water A POE Reclaimedwater_bulkwater_UtilityA_S0_2 51,423,687 mgs458874

Reclaimed bulk water A POE Reclaimedwater_bulkwater_UtilityA_S0_3 12,036,240 mgs458865

Reclaimed bulk water A POE Reclaimedwater_bulkwater_UtilityA_S0_4 4,274,193 mgs458868

Reclaimed bulk water A POE Reclaimedwater_bulkwater_UtilityA_S0_L1S2 742,276 mgs295556

Reclaimed biofilm A POU Reclaimedwater_biofilm_ UtilityA_S5_L1S9 32,082,202 mgs466972

Reclaimed bulk water A POU Reclaimedwater_bulkwater_UtilityA_S5_2 40,816,121 mgs458880

Reclaimed bulk water A POU Reclaimedwater_bulkwater_UtilityA_S5_3 43,183,835 mgs458907

Reclaimed bulk water A POU Reclaimedwater_bulkwater_UtilityA_S5_4 43,580,278 mgs458913

Reclaimed bulk water A POU Reclaimedwater_bulkwater_UtilityA_S5_L1S5 3,827,661 mgs295562

Reclaimed bulk water B Influent Reclaimedwater_bulkwater_UtilityB_-1_1 39,243,688 mgs458937

Reclaimed bulk water B POE Reclaimedwater_bulkwater_UtilityB_S0_2 24,153,588 mgs458940

Reclaimed bulk water B POE Reclaimedwater_bulkwater_UtilityB_S0_3 20,416,825 mgs458943

Reclaimed bulk water B POE Reclaimedwater_bulkwater_UtilityB_S0_4 48,463,674 mgs458931

Reclaimed bulk water B POE Reclaimedwater_bulkwater_UtilityB_S0_L1S1 4,726,224 mgs295553

Reclaimed biofilm B POU Reclaimedwater_biofilm_ UtilityB_S5_L1S7 191,192 mgs295568

Reclaimed bulk water B POU Reclaimedwater_bulkwater_UtilityB_S5_2 28,668,559 mgs458925

Reclaimed bulk water B POU Reclaimedwater_bulkwater_UtilityB_S5_3 39,378,131 mgs458871

Reclaimed bulk water B POU Reclaimedwater_bulkwater_UtilityB_S5_4 18,232,136 mgs458877

148

Reclaimed bulk water B POU Reclaimedwater_bulkwater_UtilityB_S5_L1S3 4,530,935 mgs295559

Reclaimed bulk water C Influent Reclaimedwater_bulkwater_UtilityC_-1_3 20,146,190 mgs458934

Reclaimed bulk water C POE Reclaimedwater_bulkwater_UtilityC_S0_3 33,224 mgs458892

Reclaimed bulk water C POE Reclaimedwater_bulkwater_UtilityC_S0_4 32,079,119 mgs458883

Reclaimed bulk water C POE Reclaimedwater_bulkwater_UtilityC_S0_L1S11 5,300,338 mgs295577

Reclaimed bulk water C POU Reclaimedwater_bulkwater_UtilityC_S5_3 40,009,882 mgs458904

Reclaimed bulk water C POU Reclaimedwater_bulkwater_UtilityC_S5_4 33,917,149 mgs458895

Reclaimed bulk water C POU Reclaimedwater_bulkwater_UtilityC_S5_L1S12 5,284,941 mgs295580

Reclaimed bulk water D Influent Reclaimedwater_bulkwater_UtilityD_-1_3 36,097,736 mgs458916

Reclaimed bulk water D POE Reclaimedwater_bulkwater_UtilityD_S0_1 40,363,707 mgs458919

Reclaimed bulk water D POE Reclaimedwater_bulkwater_UtilityD_S0_2 35,768,857 mgs458910

Reclaimed bulk water D POE Reclaimedwater_bulkwater_UtilityD_S0_3 37,816,330 mgs458898

Reclaimed bulk water D POU Reclaimedwater_bulkwater_UtilityD_S5_1 44,493,192 mgs458901

Reclaimed bulk water D POU Reclaimedwater_bulkwater_UtilityD_S5_2 35,693,403 mgs458946

Reclaimed bulk water D POU Reclaimedwater_bulkwater_UtilityD_S5_2_duplicate 21,616,266 mgs458949

Reclaimed bulk water D POU Reclaimedwater_bulkwater_UtilityD_S5_3 5,181 mgs458862

Reclaimed bulk water D POU Reclaimedwater_bulkwater_UtilityD_S5_4 24,947,026 mgs458928

149

Table S3: Water chemistry data for potable and reclaimed distribution system samples biodegradable dissolved dissolved dissolved temperature total Cl free Cl turbidity conductivity oxygen total organic organic organic (oC) (mg/L) (mg/L) pH (NTU) (S/m) (mg/L) carbon (µg/L) carbon (µg/L) carbon (µg/L)

A (n=44) 25.5 ± 3.3 3.5 ± 1.1 -- 7.9 ± 0.2 1.5 ± 2.7 505.8 ± 119.5 5.6 ± 1 2470 ± 762 2748 ± 1002 465 ± 758 B (n=16) 18.1 ± 3.1 0.7 ± 0.3 0.7 ± 0.3 7.8 ± 0.2 0.3 ± 0.2 361.9 ± 97.2 7.7 ± 0.6 1120 ± 1422 1439 ± 1133 548 ± 564 C (n=20) 28.4 ± 4.4 -- 0.9 ± 0.1 7.9 ± 0.2 0.2 ± 0.2 727.4 ± 49.4 7.2 ± 0.5 188 ± 62 1252 ± 1980 1522 ± 1683

Potable D (n=40) 19.4 ± 1.9 -- 0.2 ± 0.1 7.7 ± 0.1 1.3 ± 0.8 957.5 ± 416.5 5.5 ± 1.2 BDa BDa BDa

A (n=20) 26.7 ± 4.3 0.4 ± 0.4 0.2 ± 0.2 7.7 ± 0.3 5.4 ± 10.8 1354.4 ± 215.4 6.5 ± 1.3 5714 ± 2564 6351 ± 2761 2137 ± 2321 B (n=19) 19.6 ± 2.9 2.3 ± 3.1 1.1 ± 2.1 7.3 ± 0.2 2.5 ± 2.6 736.9 ± 80.3 4 ± 1.4 10123 ± 6173 11087 ± 7014 6094 ± 8777 C (n=20) 26.2 ± 3.7 0.3 ± 0.1 0.4 ± 0.4 7.3 ± 0.1 0.6 ± 0.3 1195.8 ± 17.3 4.3 ± 2.8 2791 ± 1810 2944 ± 1646 2191 ± 2244 Reclaimed D (n=20) 20 ± 1.8 2.7 ± 2.3 0.2 ± 0.2 7.2 ± 0.1 2 ± 1 1542.9 ± 283.1 6.8 ± 2.4 3961 ± 2120 4333 ± 2069 2621 ± 1238 BD = below limit of detection a19/20 samples below limit of detection (4 µg/L)

150

Table S4: Spearman’s rank correlation coefficients and p-values for correlations between 16S rRNA or opportunistic pathogen gene markers and physicochemical water quality parameters in potable and reclaimed water samples from each utility. Significant correlations indicated in bold Bulk Water Biodegradable Dissolved Temperature Total chlorine Ranked water age dissolved Phosphorus Ammonia Potable oxygen Utility A organic carbon 16S rRNA genes -0.0779 (0.6153) 0.1651 (0.2843) 0.4094 (0.0058) -0.4688 (0.0320) -0.5101 (0.0004) 0.1565 (0.3102) 0.0040 (0.9823) Acanthamoeba spp. 0.0802 (0.6049) 0.0744 (0.6312) -0.2450 (0.1090) -0.0918 (0.6924) 0.2712 (0.0749) -0.1782 (0.2471) -0.0491 (0.7862) Legionella spp. 0.2487 (0.1035) -0.0092 (0.9527) -0.0583 (0.7069) 0.1453 (0.5297) -0.0899 (0.5617) 0.0230 (0.8824) 0.0937 (0.6040) Mycobacterium spp. 0.3774 (0.0116) -0.0208 (0.8932) 0.0261 (0.8663) -0.4282 (0.0528) -0.2158 (0.1595) 0.1349 (0.3828) -0.1676 (0.3512) N. fowleri 0.4429 (0.0026) 0.1716 (0.2652) 0.2290 (0.1348) 0.3265 (0.1486) 0.2161 (0.1589) -0.3682 (0.0139) -0.0061 (0.9733) P. aeruginosa 0.1984 (0.1966) -0.0149 (0.9233) 0.0088 (0.9550) 0.2344 (0.3064) 0.2090 (0.1734) -0.1372 (0.3745) 0.1825 (0.3095) Utility B 16S rRNA genes 0.2412 (0.3682) -0.1030 (0.7042) -0.0722 (0.7558) -0.200 (0.6059) 0.1147 (0.6723) -0.696 (0.0005) NA Acanthamoeba spp. 0.1444 (0.5938) 0.3831 (0.1431) -0.1433 (0.5353) 0.0685 (0.8611) -0.2644 (0.3224) -0.1784 (0.4392) NA Legionella spp. 0.2851 (0.2845) -0.1031 (0.7040) 0.0711 (0.7594) -0.3105 (0.4160) 0.0731 (0.7878) -0.2645 (0.2465) NA Mycobacterium spp. 0.2590 (0.3327) -0.3560 (0.1759) 0.0414 (0.8584) -0.6881 (0.0405) -0.5243 (0.0371) 0.6149 (0.0030) NA N. fowleri -0.019 (0.9444) 0.5210 (0.0385) -0.1962 (0.3941) 0.3104 (0.4163) 0.1327 (0.6241) -0.4615 (0.0352) NA P. aeruginosa 0.3342 (0.2058) 0.5897 (0.0162) -0.0232 (0.9205) -0.4108 (0.2721) -0.1565 (0.5627) -0.2661 (0.2436) NA Utility C 16S rRNA genes 0.4147 (0.1243) NA -0.0475 (0.8664) -0.4286 (0.3965) -0.5929 (0.0198) 0.0966 (0.7320) -0.3000 (0.6238) Acanthamoeba spp. 0.6299 (0.0067) NA 0.0380 (0.8848) -0.0546 (0.8979) -0.3413 (0.1800) 0.6144 (0.0087) -0.6669 (0.2189) Legionella spp. -0.0319 (0.9034) NA 0.1868 (0.4728) 0.2182 (0.6036) -0.0372 (0.8874) -0.3222 (0.2072) 0.3000 (0.6238) Mycobacterium spp. 0.2208 (0.3945) NA -0.4754 (0.0538) -0.5455 (0.1619) -0.3557 (0.1612) -0.1999 (0.4417) NA N. fowleri -0.7911 (0.0002) NA -0.0807 (0.7583) 0.1690 (0.6891) 0.6588 (0.004) -0.2352 (0.3634) NA P. aeruginosa -0.4085 (0.1035) NA 0.3384 (0.1840) 0.5774 (0.1340) 0.3572 (0.1592) -0.0911 (0.7280) NA Utility D 16S rRNA genes 0.0837 (0.7256) NA -0.5527 (0.0115) NA -0.2515 (0.2848) -0.2949 (0.2068) NA Acanthamoeba spp. 0.3251 (0.1620) NA -0.3281 (0.1579) NA -0.1344 (0.5721) 0.0023 (0.9925) NA Legionella spp. 0.2677 (0.2539) NA -0.0987 (0.6788) NA 0.1082 (0.6499) -0.5835 (0.0069) NA Mycobacterium spp. 0.1672 (0.4810) NA -0.0197 (0.9342) NA -0.3628 (0.1159) 0.4673 (0.0377) NA N. fowleri 0.0290 (0.9035) NA 0.2031 (0.3905) NA 0.0474 (0.8427) -0.4118 (0.0712) NA P. aeruginosa -0.2983 (0.2014) NA 0.0352 (0.8829) NA 0.0734 (0.7586) -0.2836 (0.2257) NA Reclaimed

151

Utility A 16S rRNA genes 0.2805 (0.0880) 0.1004 (0.5486) 0.2479 (0.1334) -0.3981 (0.0397) 0.0708 (0.6772) 0.1841 (0.2685) 0.3313 (0.0851) 0.6373 Acanthamoeba spp. (<0.0001) 0.0086 (0.9593) -0.0054 (0.9742) 0.2179 (0.2749) -0.0693 (0.6836) -0.0829 (0.6207) -0.3276 (0.0888) Legionella spp. -0.0432 (0.7966) 0.0196 (0.907) 0.0608 (0.7171) -0.4332 (0.024) -0.1873 (0.2671) 0.0646 (0.7001) 0.0954 (0.6292) -0.6943 Mycobacterium spp. 0.4968 (0.0015) 0.1771 (0.2874) -0.0897 (0.5924) (<0.0001) 0.2058 (0.2217) 0.1671 (0.3160) 0.2792 (0.1502) N. fowleri 0.4753 (0.0026) -0.0248 (0.8827) 0.0046 (0.978) 0.195 (0.3298) -0.0701 (0.6803) 0.2766 (0.0927) -0.1281 (0.5159) P. aeruginosa -0.3232 (0.0478) -0.1078 (0.5194) 0.0681 (0.6846) 0.2299 (0.2488) -0.0611 (0.7194) -0.2468 (0.1353) NA Utility B 16S rRNA genes -0.3988 (0.1128) -0.7794 (0.0002) -0.329 (0.1972) 0.1958 (0.5419) -0.3007 (0.3423) -0.2426 (0.3480) NA Acanthamoeba spp. 0.3313 (0.1939) 0.2032 (0.4341) 0.0500 (0.8490) 0.4421 (0.1501) 0.2797 (0.3786) -0.3117 (0.2233) NA Legionella spp. -0.3616 (0.1538) -0.6536 (0.0044) -0.1498 (0.5660) 0.2465 (0.4399) -0.5429 (0.0682) -0.0934 (0.7215) NA Mycobacterium spp. -0.1822 (0.4841) -0.7265 (0.0010) -0.0391 (0.8817) 0.5493 (0.0643) -0.0490 (0.8797) -0.4548 (0.0666) NA N. fowleri 0.0983 (0.7074) -0.0304 (0.9077) -0.4598 (0.0633) NA -0.1404 (0.6633) 0.0346 (0.8952) NA P. aeruginosa 0.3066 (0.2314) 0.2041 (0.4320) 0.3138 (0.2200) NA 0.1310 (0.6848) 0.0510 (0.8458) NA Utility C 16S rRNA genes 0.3605 (0.1552) 0.2 (0.7471) -0.4134 (0.0991) 0.0727 (0.8317) -0.5172 (0.0335) 0.0723 (0.7903) 0.200 (0.7471) Acanthamoeba spp. 0.3406 (0.1810) NA -0.2401 (0.3533) 0.3772 (0.2528) 0.2126 (0.4126) -0.1398 (0.6056) NA Legionella spp. 0.2986 (0.2444) 0.4 (0.5046) -0.2099 (0.4189) -0.335 (0.314) -0.3502 (0.1682) -0.26 (0.3307) 0.600 (0.2848) Mycobacterium spp. 0.5159 (0.0340) NA 0.0415 (0.8742) -0.4419 (0.1736) 0.1295 (0.6203) -0.7421 (0.001) NA N. fowleri -0.7414 (0.0007) NA 0.0845 (0.747) -0.1706 (0.6161) 0.741 (0.0007) 0.4752 (0.0629) NA P. aeruginosa NA NA NA NA NA NA NA Utility D 16S rRNA genes -0.318 (0.1718) 0.3633 (0.1154) -0.2698 (0.2500) 0.4762 (0.2329) 0.1368 (0.5651) 0.0286 (0.9048) NA Acanthamoeba spp. 0.0868 (0.7160) 0.3479 (0.1328) -0.4708 (0.0362) -0.2474 (0.5546) -0.0808 (0.7348) 0.1299 (0.5851) NA Legionella spp. -0.0090 (0.9698) 0.1587 (0.5039) -0.0123 (0.9591) 0.7619 (0.0280) 0.3398 (0.1426) -0.3061 (0.1893) NA Mycobacterium spp. -0.2081 (0.3787) -0.1023 (0.6677) -0.1411 (0.5530) -0.3095 (0.4556) 0.2482 (0.2913) -0.0026 (0.9912) NA N. fowleri 0.1212 (0.6107) -0.0088 (0.9705) -0.1868 (0.4304) -0.4566 (0.2554) -0.2636 (0.2614) 0.2649 (0.2590) NA P. aeruginosa NA NA NA NA NA NA NA Biofilm Potable Utility A 16S rRNA genes 0.3259 (0.0401) 0.0527 (0.7466) 0.1271 (0.4345) -0.3281 (0.1703) -0.3448 (0.0293) -0.1363 (0.4016) -0.1102 (0.5621) Acanthamoeba spp. 0.5396 (0.0003) 0.1106 (0.4969) 0.0784 (0.6305) 0.3444 (0.1487) 0.1849 (0.2533) -0.3819 (0.015) -0.1153 (0.544) Legionella spp. 0.2526 (0.1159) 0.0124 (0.9395) 0.2325 (0.1488) 0.2125 (0.3825) -0.1524 (0.3478) -0.1468 (0.3659) -0.0127 (0.9471)

152

Mycobacterium spp. 0.4429 (0.0042) 0.0595 (0.7154) 0.1964 (0.2246) -0.1441 (0.5561) -0.0797 (0.6247) -0.0272 (0.8676) -0.1714 (0.365) N. fowleri 0.0439 (0.7881) -0.2909 (0.0686) -0.0658 (0.6867) 0.5493 (0.0149) -0.1381 (0.3956) -0.0962 (0.5549) -0.1091 (0.5659) P. aeruginosa 0.0615 (0.706) -0.2382 (0.1388) -0.0515 (0.7524) 0.3443 (0.1489) 0.2441 (0.1291) 0.0302 (0.8531) -0.2135 (0.2574) Utility B 16S rRNA genes 0.4588 (0.0738) 0.3179 (0.2302) 0.2908 (0.201) 0.5 (0.1705) 0.3206 (0.226) -0.4779 (0.0284) NA Acanthamoeba spp. 0.023 (0.9326) -0.2597 (0.3314) 0.4034 (0.0697) -0.3195 (0.402) -0.4551 (0.0765) 0.4298 (0.0518) NA Legionella spp. 0.6801 (0.0037) 0.3388 (0.1993) 0.2383 (0.2982) 0.2609 (0.4978) 0.1197 (0.6589) -0.175 (0.4481) NA Mycobacterium spp. 0.3499 (0.184) 0.0589 (0.8284) 0.279 (0.2207) -0.2739 (0.4758) 0.0065 (0.9808) 0.1948 (0.3974) NA N. fowleri 0.2251 (0.4018) 0.2175 (0.4183) -0.1053 (0.6498) 0.7794 (0.0133) 0.1584 (0.558) -0.1213 (0.6005) NA P. aeruginosa -0.4142 (0.1107) -0.095 (0.7264) -0.0285 (0.9023) 0.3651 (0.3339) -0.1812 (0.5018) -0.0615 (0.7911) NA Reclaimed Utility A 16S rRNA genes 0.1469 (0.4145) -0.1169 (0.517) 0.1727 (0.3366) 0.4439 (0.0385) -0.3586 (0.0439) -0.3062 (0.0831) -0.3989 (0.0535) Acanthamoeba spp. 0.2881 (0.104) 0.0204 (0.9104) 0.012 (0.947) -0.1204 (0.5935) 0.3211 (0.0732) 0.0505 (0.7803) 0.0665 (0.7576) Legionella spp. -0.2106 (0.2394) 0.0137 (0.9396) -0.0536 (0.7671) -0.1807 (0.4209) -0.066 (0.7197) 0.1641 (0.3615) 0.223 (0.2948) Mycobacterium spp. 0.1046 (0.5623) -0.0023 (0.9899) 0.087 (0.6301) -0.4051 (0.0614) 0.4483 (0.0101) 0.1733 (0.3349) 0.4293 (0.0363) N. fowleri 0.6008 (0.0002) -0.0608 (0.7369) -0.1304 (0.4696) 0.2197 (0.3259) -0.2785 (0.1227) 0.1647 (0.3598) -0.3607 (0.0833) P. aeruginosa 0.1114 (0.537) 0.0929 (0.6071) 0.1327 (0.4617) -0.1204 (0.5935) 0.0876 (0.6337) 0.2612 (0.142) 0.1977 (0.3546) Utility B 16S rRNA genes 0.5707 (0.0263) 0.3357 (0.2212) 0.2871 (0.281) 0.0636 (0.8525) -0.1841 (0.6354) -0.2964 (0.2834) NA Acanthamoeba spp. 0.392 (0.1485) 0.6898 (0.0044) -0.1031 (0.7041) -0.3642 (0.2708) 0.2119 (0.5841) 0.0625 (0.8248) NA Legionella spp. 0.1083 (0.7008) -0.1555 (0.58) 0.1053 (0.698) 0.5057 (0.1125) -0.4603 (0.2125) -0.4075 (0.1316) NA Mycobacterium spp. 0.209 (0.4547) 0.1691 (0.5469) -0.1169 (0.6662) -0.2897 (0.3876) -0.0253 (0.9485) 0.2878 (0.2983) NA N. fowleri -0.4186 (0.1205) -0.4393 (0.1013) -0.0462 (0.8651) 0.4781 (0.1369) 0.1558 (0.6889) -0.2455 (0.3778) NA P. aeruginosa -0.4565 (0.0872) -0.1404 (0.6177) 0.1477 (0.5851) 0.1561 (0.6467) 0.6875 (0.0407) 0.0945 (0.7377) NA NA indicates correlation not possile due to insufficient data points above detection or data not collected

153

Table S5: Spearman’s rank correlation coefficients and p-values for correlations between 16S rRNA genes or opportunistic pathogens and phyla (or proteobacteria classes) determined by 16S rRNA sequencing. Brackets indicate suggested taxonomies. phyla ρ p-value phyla ρ p-value phyla ρ p-value Positive Mycobacterium spp. Positive Acanthamoeba spp. Positive Legionella spp. Correlations Correlations Correlations WPS-2 0.4198 <.0001 WS5 0.1549 0.0098 TM7 0.4083 <.0001 Tenericutes 0.2992 <.0001 Firmicutes 0.1526 0.011 NKB19 0.3975 <.0001 SR1 0.2714 <.0001 Cyanobacteria 0.1385 0.0211 SR1 0.3718 <.0001 Lentisphaerae 0.2693 <.0001 Betaproteobacteria 0.1384 0.0212 Negative Acanthamoeba spp. Lentisphaerae 0.3429 <.0001 NKB19 0.2654 <.0001 Correlations Tenericutes 0.3207 <.0001 WWE1 0.2634 <.0001 SR1 -0.1455 0.0154 Positive N. fowleri OP11 0.3108 <.0001 WPS-2 0.2575 <.0001 Correlations Euryarchaeota Crenarchaeota Synergistetes 0.2752 <.0001 (Archaea) 0.2571 <.0001 (Archaea) 0.1537 0.0104 WWE1 0.2685 <.0001 OP11 0.2417 <.0001 Firmicutes 0.1188 0.0482 Negative N. fowleri TM6 0.2546 <.0001 TM6 0.2417 <.0001 Correlations BHI80-139 0.2232 0.0002 OP8 0.2304 0.0001 Alphaproteobacteria -0.1242 0.0389 BRC1 0.2082 0.0005 Fibrobacteres 0.2237 0.0002 SR1 -0.1271 0.0345 Bacteroidetes 0.2066 0.0005 TM7 0.2056 0.0006 Nitrospirae -0.1386 0.021 Chlamydiae 0.191 0.0014 Synergistetes 0.201 0.0008 [Thermi] -0.2231 0.0002 Positive P. aeruginosa Thermotogae 0.182 0.0024 OD1 0.1988 0.0009 Correlations Spirochaetes 0.1749 0.0035 GN02 0.1818 0.0024 Cyanobacteria 0.1829 0.0022 Fibrobacteres 0.1645 0.0061 Spirochaetes 0.1697 0.0046 AncK6 0.1546 0.01 OP8 0.1555 0.0095 OP3 0.1578 0.0085 Chloroflexi 0.1532 0.0107 Armatimonadetes 0.1472 0.0142 Chlamydiae 0.1523 0.0112 WS3 0.1291 0.0317 PAUC34f 0.1444 0.0162 Armatimonadetes 0.1447 0.016 Negative P. aeruginosa Correlations Euryarchaeota (Archaea) 0.1382 0.0214 Deferribacteres 0.1356 0.024 SBR1093 -0.1232 0.0405 GN02 0.1368 0.0228 Bacteroidetes 0.1309 0.0294 NKB19 -0.1439 0.0165 Fusobacteria 0.1322 0.0278 BRC1 0.1299 0.0306 TM7 -0.1496 0.0127 WS4 0.1288 0.0321 H-178 0.1237 0.0396 WPS-2 -0.1707 0.0044

154

Deferribacteres 0.1241 0.0391 Thermotogae 0.1194 0.0472 TM6 -0.1876 0.0017 Deltaproteobacteria 0.1193 0.0473 Negative Legionella spp. Correlations Actinobacteria -0.1249 0.0378 Chloroflexi -0.1304 0.0301 FCPU426 -0.1398 0.0199 [Parvarchaeota] (Archaea) -0.1499 0.0125 Caldithrix -0.1683 0.005 Nitrospirae -0.1931 0.0012 NC10 -0.1934 0.0012 Cyanobacteria -0.2158 0.0003

Table S6: Nitrifying bacteria, denitrifying bacteria, sulfate-reducing bacteria and heterotrophic aerobic bacteria in potable and reclaimed bulk water samples determined by Biological Activity Reaction Test (Hach, Loveland, CO). Values indicate frequency of detection (average ± standard deviation).

Potable Nitrifying Bacteria Denitrifying Bacteria Sulfate-reducing Bacteria A (n=28) 18% (22400 ± 29760) 7% (50000 ± 0) ND C (n=16) ND 75% (708333 ± 433188) 6% (1200 ± 0) D (n=16) ND 6% (10000 ± 0) ND Reclaimed A (n=30) 27% (3125 ± 4249) 27% (143750 ± 66480) 93% (83800 ± 146471) B (n=8) ND 63% (34000 ± 21909) 38% (73833 ± 48353) C (n=15) 73% (26091 ± 23106) 67% (635000 ± 473198) 40% (402 ± 459) D (n=14) 93% (73077 ± 20160) 14% (30000 ± 28284) 86% (61617 ± 138917)

155

CHAPTER 7 : IMPACT OF BLENDING FOR DIRECT POTABLE REUSE ON PREMISE PLUMBING MICROBIAL ECOLOGY AND REGROWTH OF OPPORTUNISTIC PATHOGENS AND ANTIBIOTIC RESISTANT BACTERIA

Emily Garner, Mandu Inyang, Elisa Garvey, Jeffrey Parks, Eric Dickerson, Justin Sutherland, Andrew Salveson, Marc Edwards, Amy Pruden

ABSTRACT

Little is known about how introducing recycled water intended for direct potable reuse (DPR) into distribution systems and premise (i.e., plumbing) will affect water quality at the point of use, particularly with respect to effects on microbial populations and regrowth. The potential to trigger growth of opportunistic pathogens (OPs) and the spread of antibiotic resistance genes (ARGs), each representing serious and growing public health concerns, by introducing DPR water has not previously been evaluated. In this study, the impact of blending DPR water with traditional potable water sources was investigated with respect to treatment techniques, blending location, and blending ratio. Water from four U.S. utilities was treated in bench- and pilot-scale treatment trains to simulate DPR with blending. Water was incubated in simulated premise plumbing rigs made of PVC pipe and brass coupons to measure regrowth of total bacteria (16S rRNA genes, heterotrophic plate counts), OPs (Legionella spp., Mycobacterium spp.), ARGs (qnrA, vanA) and a marker of horizontal gene transfer and multi-drug resistance (intI1). The microbial community composition was profiled and the resistome (i.e., all ARGs present) was characterized in select samples. While regrowth of 16S rRNA genes consistently occurred across tested scenarios (p≤0.0001), total bacteria were not more abundant in the water or biofilm of any DPR scenario than in the corresponding conventional potable condition (p≥0.0748). Regrowth of OPs and ARGs was not significantly greater in water or biofilm for any DPR blends treated with advanced oxidation compared to corresponding potable water (p≥0.1047). These results reveal no evidence that blended DPR water will create unusual problems with either total bacteria, OPs, and ARGs in premise plumbing.

INTRODUCTION

Population growth, urbanization, climate change, drought, and diminishing traditional potable water sources have driven many municipalities to consider using alternative water sources to meet future water demand (Gosling and Arnell, 2016; US EPA, 2012). One option for augmenting traditional potable water sources is to implement direct potable reuse (DPR), or advanced treatment of municipal sewage to achieve high-quality water suitable for potable use. Advanced water treatment technologies typically applied for DPR include membrane filtration (e.g., ultrafiltration, reverse osmosis) and advanced oxidation processes (AOPs; e.g., ozonation, ultraviolet irradiation combined with hypochlorite or hydrogen peroxide) (Gerrity et al., 2013). While these advanced treatment technologies have exhibited strong potential for removing emerging contaminants, such as endocrine disrupting chemicals, pharmaceuticals, and personal care products (Kim et al., 2007; Snyder et al., 2007; Wang et al., 2016; Watkinson et al., 2007), their use also creates unique challenges. Reverse osmosis, for example, might alter corrosivity (Applegate, 2017; Gerrity et al., 2013). To address this challenge, as well as the limitation that wastewater reuse alone is likely insufficient to generate enough water to supply a municipality’s

156

potable water demand, recycled wastewater may be blended with a traditional potable source water (Gerrity et al., 2013). However, little is known about how blending waters with such distinct water chemistries might impact biological stability and microbial water quality in distribution systems and premise plumbing.

While advanced water purification (AWP) is likely to be extremely effective at reducing viable bacteria in DPR water, regrowth of bacteria is still likely to occur in distribution system pipes and premise plumbing if the processed water does not achieve sufficient biological stability (Chowdhury, 2012; Thayanukul et al., 2013). The U.S. Centers for Disease Control has identified opportunistic pathogens (OPs), such as Legionella, to be a leading source of waterborne disease outbreaks in the U.S. and called for awareness of the potential for water quality to degrade in pipe systems (Centers for Disease Control and Prevention, 2011; Yoder et al., 2008). OPs are known to thrive under the conditions typical of premise plumbing, including high water age, depleted disinfectant residual, elevated water temperatures, the presence of plumbing materials that react with disinfectant residuals or leach nutrients, high surface area to volume ratios, and highly variable water flow conditions leading to long periods of stagnation (Falkinham, 2015; Nguyen et al., 2012; Rhoads et al., 2016). Concerns are also emerging regarding the potential to spread antibiotic resistance genes (ARG) via water reuse (Hong et al., 2013; Pruden, 2014), though to the authors’ knowledge, the presence of ARGs in DPR water has never been studied. Reuse of wastewater involves higher initial concentrations of antibiotics, antibiotic resistant bacteria (ARB), and ARGs than are typical of most potable source waters, thus implementing treatment schemes capable of removal of these contaminants is critical. While treatment trains typical of reuse have shown promise for removing such contaminants more effectively than traditional wastewater treatment, AOPs can result in incomplete removal of antibiotics (Watkinson et al., 2007), while other advanced treatment approaches might fail to remove, and in some cases might even enrich, certain ARGs (Alexander et al., 2016; Czekalski et al., 2016; Yoon et al., 2017). Incomplete removal of ARB and ARGs during treatment creates potential for regrowth before water reaches the consumer during distribution and in premise plumbing. Additionally, horizontal gene transfer and uptake of extracellular ARGs (i.e., natural transformation) could facilitate dissemination of antibiotic resistance within pipe systems (Garner et al., 2016b).

Here we evaluated the impact of blending water source type and quality, water treatment methods, and blending ratio on bacterial regrowth potential in premise plumbing. Full-scale wastewater treatment trains from each of four partner water utilities were supplemented with bench- or pilot-scale AWP processes to achieve water quality suitable for DPR. DPR water was blended with each municipality’s potable source water at ratios ranging from 0-50%, consistent with utility projections for future DPR implementation, and incubated in pipe rigs with regular water changes over eight weeks and compared to the corresponding potable water control. Rig influent water (i.e., the simulated point of compliance; POC) and effluent water (i.e., simulated point of use; POU) and biofilm samples were collected and analyzed for gene markers of total bacteria (16S rRNA genes), OPs (Legionella spp., Mycobacterium spp., and P. aeruginosa), two ARGs of clinical significance (vanA, qnrA), and a key horizontal gene transfer element (intI1), along with culturing of common bacterial indicators (heterotrophic plate count, E. coli, and enterococci), comprehensive microbial community profiling via 16S rRNA gene amplicon sequencing, and shotgun metagenomic sequencing of select samples. Various biochemical indicators of microbial regrowth potential, including biodegradable dissolved organic carbon

157

(BDOC) and biological activity reaction tests (BARTs), were also analyzed and compared. This study provides valuable insight into potential microbial responses and public health concerns when introducing blended DPR water to distribution systems.

MATERIALS AND METHODS

Rig design and operation

Simulated premise plumbing pipe rigs were constructed of polyvinylchloride (PVC) pipe with brass inserts. Rigs were 1.5 m long with an inner diameter of 5 cm. Brass inserts were 46 cm long, 1.3 cm in diameter, and comprised of standard yellow brass that is nominally 60% copper, 35% zinc, and 3% lead. Forty pipe rigs were constructed and pre-tested according to National Sanitation Foundation (NSF) Standard 61 to remove outliers (NSF, 2007) as described in the supplementary information. The test scenarios studied are summarized in Table 7-1. DPR water was treated using either pilot- or bench-scale treatment as described in the supplementary information. Treated waters were dosed with chlorine or chloramine residuals consistent with those typically employed by each utility: chloramine at 1.8-2.2 mg/L total chlorine for Utility 1, chloramine at 3.8 mg/L total chlorine for Utility 2, 1.5 mg/L free chlorine for Utility 3, and chloramine at 2.6 mg/L total chlorine for Utility 4. To produce enough treated water for eight weeks of rig operation, water was treated in two batches at weeks 1 and 4. After blending and treatment, to preserve the physicochemical characteristics of treated water for long-term water changes, water for use during the third and fourth weeks after each batch treatment was pasteurized as described by Escobar and Randall (2000). To pasteurize, jugs of treated water were placed in a water bath and heated until the inner temperature reached 72°C for 30 minutes. All treated water was stored in one-gallon amber glass jugs at 4°C until use. After storage, chlorine and chloramine residuals were dosed as needed to once again reach target concentrations. Premise plumbing operation was simulated by replacing 100% of the volume in each rig with fresh stored, treated water three times per week. Duplicate pipe rigs for each test scenario were incubated at room temperature (~20°C) during eight weeks of this simulated operation.

Water was collected from the final two water changes during week eight in sterile one liter polypropylene containers for molecular analysis and in acid-washed, baked 250 milliliter amber glass bottles for carbon analysis. At the conclusion of each eight-week incubation, biofilm samples were collected by swabbing once along the length of the brass insert with a sterile cotton-tipped applicator. The sample end of the swab was transferred directly to a sterile DNA extraction lysing tube. Water samples for molecular analysis were filter-concentrated immediately after collection onto 0.2 µm cellulose nitrate filters in pre-packaged, sterile filter funnels (Nalgene, Rochester, NY). Filters were folded into quarters, torn into 1 cm2 pieces using sterile forceps, transferred to lysing tubes, and frozen at -20ºC. DNA was extracted using a FastDNA SPIN Kit (MP Biomedicals, Solon, OH) according to manufacturer instructions.

158

Water chemistry

A 30 milliliter aliquot was taken for total organic carbon (TOC) analysis and a second aliquot was filtered through pre-rinsed 0.22 µm pore size mixed cellulose esters membrane filters (Millipore, Billerica, MA) for dissolved organic carbon (DOC) analysis. BDOC was measured as previously described by Servais et al. (1989) with incubation time extended to 45 days. Samples were analyzed on a Sievers 5310C portable TOC analyzer (GE, Boulder, CO) according to Standard Method 5310C (APHA, 2005).

Culturing

The fraction of the total HPC capable of growth in the presence of antibiotics was determined by plating POC and POU samples on R2A media (Hardy Diagnostics, Santa Maria, CA) with and without one of nine antibiotics added. Media was supplemented separately with ampicillin (4 µg/mL), ciprofloxacin (0.5 µg/mL), chloramphenicol (4 µg/mL), gentamicin (2 µg/mL), oxacillin (1 µg/mL), rifampin (0.5 µg/mL), sulfamethoxazole (128 µg/mL), tetracycline (2 µg/mL), and vancomycin (0.5 µg/mL) (BD, Franklin Lakes, NJ). To select the antibiotic concentrations used, a trial was conducted using local tertiary treated recycled water after subsequent granular activated carbon filtration. The concentration of each antibiotic that produced a 2-log reduction in plate count compared to the R2A agar without antibiotics was selected. Culturing was performed according to standard method 9215C (APHA, 2005). Briefly, four ten- fold serial dilutions of water sample were prepared and 0.1 milliliter of each was spread onto prepared R2A agar. Plates were incubated for seven days at 37°C and enumerated, with an upper and lower limit of quantification (LOQ) of 20 and 200 colonies per plate. E. coli and enterococci were cultured from week eight simulated POC and POU water samples using Colilert and Enterolert Quantitrays (IDEXX, Westbrook, ME). BART tests (Hach, Loveland, CO) were used to approximate the presence of active nitrifying, denitrifying, sulfate-reducing, acid-producing, slime-producing, and heterotrophic aerobic bacteria.

Quantitative polymerase chain reaction

OP gene markers and ARGs were quantified by quantitative polymerase chain reaction (qPCR) using previously published primers and thermocycler conditions (Table S1). The universial bacterial gene, 16S rRNA, genes associated with three OPs (Legionella spp., Mycobacterium spp., Pseudomonas aeruginosa), two ARGs (a quinolone resistance gene, qnrA, and a vancomycin resistance gene, vanA), along with the class 1 integron integrase gene intI1 were quantified via qPCR. Reaction components are described in detail in the supplementary information. Prior to all analyses, 16S rRNA genes were quantified in a representative subset of samples diluted ten-, 20-, 50-, and 100-fold as well as undiluted to identify the minimum concentration at which inhibition was negligible. A ten-fold dilution was selected and applied to all samples. All qPCR runs included a triplicate negative control and triplicate standard curves consisting of ten-fold serial dilutions ranging from 107-101 gene copies/µl for all genes except 16S rRNA, for which 108-102 gene copies/µl were used. The limit of quantification (LOQ) for all genes was 10 gene copies per reaction.

159

Table 7-1: Blending scenarios, blending water source, treatment, disinfectants, and blending location tested for each utility Source of potable Reuse water treatmentc Treatment Utility Scenarioa Disinfectant water for blending performed prior to blending following blending 100% Surface water -- O3, coagulation, Treated potable water f 90% Surface/10% O3-BAC O3, BAF flocculation, 1 derived from surface NH Cl 90% Surface/10% AWP sedimentation, 2 waterf MF, RO, UV/AOPf 50% Surface/50% AWP filtrationb 100% Groundwater Treated potable water -- 90% Groundwater/10% AWP derived from MF, RO, UV/AOPp 2 50% Groundwater/50% AWP groundwater (treated -- NH2Cl 50% Groundwater/50% with iron and UF, RO, UV/AOP, and AWP-Past manganese removalf) pasteurizationp 100% Surface water -- 95% Surface/5% Tertiary Secondary treatment with Treated potable water Coagulation, nitrification, partial derived from surface flocculation, 3 90% Surface/10% Tertiary denitrification, and biological Cl2 water (treated with f sedimentation, f phosphorus removal b O3 ) filtration 90% Surface/10% O3-BAC f O3, BAF 50% Surface/50% O3-BAC 100% Groundwater Treated potable water -- derived from 90% Groundwater/10% AWP MF, RO, UV/AOPf O , coagulation, groundwaterf 3 flocculation, 4 100% Surface -- NH Cl Treated potable water sedimentation, 2 90% Surface/10% AWP MF, RO, UV/AOPf derived from surface filtrationb 90% Surface/10% Industrial waterf Industrial treatmentf AWP aSurface water and Groundwater refer to treated potable water derived from the designated source cAll reuse water treatment is performed subsequent to secondary treatment unless specified otherwise fFull scale; pPilot scale; bBench scale ; O3 = Ozonation; UV = ultraviolet irradiation, UV/AOP = ultraviolet irradiation with hypochlorite or hydrogen peroxide MF = membrane filtration; RO = reverse osmosis; UF = ultrafiltration; BAF = biologically active filtration

160

16S rRNA gene amplicon sequencing and shotgun metagenomics

Bacterial communities were profiled using gene amplicon sequencing with barcoded primers (515F/806R) targeting the V4 region of the 16S rRNA gene (Caporaso et al., 2012). Triplicate PCR products were composited, and 240 ng of each composite was combined and purified using a QIAquick PCR Purification Kit (Qiagen, Valencia, CA). Sequencing was conducted at the Biocomplexity Institute of Virginia Tech Genomics Sequencing Center (BI; Blacksburg, VA) on an Illumina MiSeq using a 250-cycle paired-end protocol. Field, filtration, DNA extraction blanks, and a least one PCR blank per lane were included in the analysis.

Shotgun metagenomic sequencing was conducted on the POC and week eight POU samples from each system, on week eight POU water and biofilm samples from each potable source water, and each 10% DPR blend scenario. Sequencing was conducted as previously described (Garner et al., 2016a) on an Illumina HiSeq 2500 with 100-cycle paired end reads at BI. Select source and potable samples were also sequenced.

Data analysis of 16S rRNA gene amplicon sequencing and shotgun metagenomic data was conducted as previously described (Garner et al., 2016a), with details also provided in the supplementary information.

Statistical Analysis

Statistical differences between BDOC, 16S rRNA genes, and HPC in samples were tested by Tukey HSD in JMP (v. 13). For OPs and ARGs, which were non-normally distributed, a Kruskal Wallis rank sum test with a posthoc pairwise Wilcoxon test was performed in JMP. Spearman’s rank correlation was used to test for correlations between BDOC and 16S rRNA genes, OPs, or ARGs, while a Pearson product-moment correlation was used to test for correlations between HPC and 16S rRNA genes. Weighted UNIFRAC distance matrices generated in QIIME were imported to PRIMER-E (version 6.1.13) for analysis of similarities (ANOSIM).

RESULTS AND DISCUSSION

Comparison of regrowth in simulated premise plumbing rigs

Treated blends of DPR water and each utility’s traditional potable water were incubated in pipe rigs to simulate water use in premise plumbing to examine regrowth of bacteria. 16S rRNA genes, a proxy for total bacteria, were measured in week eight at the simulated POC and POU samples following eight weeks of simulated use in the pipe rigs to examine regrowth of total bacteria (Figure 7-1). Across conditions, the abundance of 16SrRNA genes at the simulated POU was greater than at the POC (paired Wilcoxon; p≤0.0001). The most regrowth was observed in Utility 1 scenarios utilizing 90-100% Surface water (i.e., treated potable water derived from surface water; 2.7-4.2 log increase), Utility 4’s 100% Groundwater scenario (i.e., treated potable water derived from groundwater; 3.19 log increase), and Utility 3’s 90% Surface/10% Tertiary blend (3.3 log increase). Only the 90% Surface/10% AWP condition from Utility 3 did not result in regrowth of total bacteria (0.4 log decrease), while all other scenarios produced between 0.2 and 2.0 log increase in 16S rRNA genes. Thus, most conditions stimulated re-growth of bacteria, as

161

expected, but surprisingly, it was sometimes the 100% traditional potable water that yielded the most re-growth.

Figure 7-1: qPCR abundances of 16S rRNA genes, OPs, and ARGs. Abundances of 16S rRNA genes, opportunistic pathogen gene markers (Legionella spp., Mycobacterium spp., P. aeruginosa), antibiotic resistance genes (vanA, qnrA), and the class 1 integron integrase gene intI1 from samples collected after eight weeks of pipe rig incubation at the simulated point of compliance (POC), simulated point of use (POU), and in the biofilm (BF). Asterisks indicate the gene was detected below the limit of quantification. Surface water and Groundwater conditions refer to treated potable water derived from each source. AWP = advanced water purification using membrane filtration, reverse osmosis and ultraviolet irradiation with the addition of hypochlorite or hydrogen peroxide; O3-BAC = ozone followed by biologically active carbon filtration; Past = refers to treatment using pasteurization; Industrial refers to wastewater derived from an industrial source rather than municipal wastewater, as in all other cases. Refer to Table 7-1 for additional details regarding treatment schemes. 162

Comparing within Utility 2 and Utility 3 scenarios, there were no differences in total regrowth across the different treatments and blends (i.e., no difference in log increase in 16S rRNA gene abundances), indicating that DPR scenarios did not produce more regrowth than the corresponding potable source waters (Wilcoxon; p≥0.2996). For Utility 1, the DPR scenario with the greatest blend ratio (50% Surface/50% AWP) resulted in more regrowth than the corresponding lesser blend ratio (p=0.0490), but still produced less regrowth than the traditional potable source water alone (p≥0.1632). For Utility 4, the 100% Surface water and 100% Groundwater scenarios also produced more regrowth than some of the DPR blends, specifically 10% AWP DPR condition (p≤0.0404). The 90% Surface/10% Industrial AWP blend produced an equivalent amount of regrowth as the corresponding potable source water (p=0.426).

Biofilms are of interest given that they are thought to represent the primary reservoir for microbial regrowth in distribution systems (Liu et al., 2014), though recent reports highlight that substantial regrowth can occur in the bulk phase too (Proctor et al., 2016). In the present study, there were no apparent effects of DPR blends on the biomass density of the biofilms after the eight- week incubation. In the case of each of the four utilities, the 16S rRNA gene abundances in biofilms were not significantly greater in DPR blend scenarios than the corresponding potable source water scenarios (p≥0.0748).

Enumeration of 16S rRNA genes by qPCR captures genes from both live and dead cells, thus relative differences were compared above to estimate regrowth. To compare the findings with a standard culture-based method, heterotrophic plate counts were also enumerated for a subset of scenarios. 16S rRNA gene copy numbers and HPCs were found to be significantly correlated (Pearson, R2=0.5127, p=0.0053). While substantial regrowth of total bacteria was noted for the majority of conditions, neither HPC nor 16S rRNA gene abundances indicated greater regrowth in any DPR scenarios than in the corresponding potable water condition. These results are congruent with expectations based on available literature, as regrowth of total bacteria (via proxys such as HPC, total coliforms, and 16S rRNA genes) is well-documented in traditional potable water distribution systems and premise plumbing (LeChevallier et al., 1991; Wang et al., 2012). Much less information is available regarding regrowth during distribution of reuse water, though studies of non-potable reclaimed water systems have similarly demonstrated substantial regrowth (Jjemba et al., 2010; Narahimhan et al., 2005).

E. coli and enterocci were also cultured from the simulated POC and POU of each rig, but no positives were detected. This was consistent with the expectation that fecal indicators do not survive well in relatively cool, oligotrophic water systems; which represent a very different environment than the mammalian gut, and therefore are not generally subject to regrowth.

Microbial community composition of regrowth

16S rRNA gene amplicon sequencing was carried out to gain insight into the kinds of bacteria subject to regrowth under the various scenarios. Notably, the composition of the microbial community typically shifted during pipe rig incubation. Within each scenario, the microbial community composition of simulated POU water samples were significantly different from POC samples (ANOSIM, R=0.706, p=0.001) (Figure 7-2). Also, the composition of POU samples varied widely across utilities (R=0.450, p=0.001) and among scenarios within each utility

163

(R=0.430, p=0.001). Within each scenario, water and biofilm communities were not significantly different (R=0.216, p=0.064), suggesting that there was interplay between the bulk water and biofilm under these stagnant pipe rig conditions, leading to a high degree of similarity between the two matrices. This contrasts previous studies of potable premise plumbing, in which bulk water microbial communities were found to be largely unique compared to that of the corresponding biofilm (Ji et al., 2017; Liu et al., 2014).

For Utility 1 water, all simulated POU samples were dominated by Alphaproteobacteria (61.0-99.9%), though POC samples were dominated by a combination of Alphaproteobacteria, Betaproteobacteria, and Gammaproteobacteria. Biofilms were also overwhelmingly dominated by Alphaproteobacteria, with the exception of the 50% Surface/50% AWP scenario, which was dominated by Gammaproteobacteria (37.5-65.9%) and Betaproteobacteria (22.4-32.1%). The vast majority of Alphaproteobacteria detected in POU water and biofilm samples belonged to the Methylobacteriaceace family (57.9-99.9% of total), while the 50% Surface/50% AWP scenario biofilm samples included Sphingomonadaceae as the dominant Alphaproteobacteria, Comamonadaceae as the dominant Betaprotebacteria, and Moraxellaceae as the dominant Gammaproteobacteria. Of Utility 2 samples, the 100% Groundwater POU water and biofilm samples were dominated by Alphaproteobacteria (51.7-98.0%), while the AWP blends were typically dominated by Betaproteobacteria (6.4-30.9%) and Gammaproteobacteria (12.3-81.7%). The 50% Groundwater/50% AWP-Past blend was dominated by Actinobacteria (62.7-75.9%). Utility 3 POU water and biofilm samples were largely dominated by Gammaproteobacteria (16.2- 99.4%) and Alphaproteobacteria (12.8-81.6%) for the 100% Surface scenario and both tertiary blends. The O3-BAC conditions were largely dominated by Actinobacteria, Betaproteobacteria, Alphaproteobacteria and Bacilli. Utility 4 samples were largely dominated by Alphaproteobacteria, Betaproteobacteria, and Actinobacteria across all scenarios.

While 16S rRNA amplicon sequencing does not have the resolution to confirm the presence of pathogens, it is possible to screen for phylogenetic groups known to contain pathogenic strains. Across scenarios, of the Actinobacteria present, 93% belonged to the genus Mycobacterium. Of the Alphaproteobacteria, 41% belonged to the genus Methylobacterium, 23% belonged to the genus Sphingomonas, and 14% belonged to Blastomonas. 21% of Betaproteobacteria belonged to the Comomonadaceae family but could not be classified at the genus level. Ralstonia, Acidovorax, and Limnohabitans were responsible for 15%, 15%, and 10% of Betaproteobacteria, respectively. Gammaproteobacteria primarily belonged to the Nevskia, Acinetobacter, and Pseudomonas genera (44%, 17%, and 10%, respectively) and no known enteric pathogens, which was consistent with fecal indicator monitoring noted above. Thus, based on amplicon sequencing, only a few potentially pathogenic groups were identified: Mycobacteria, Enterobacter, and Pseudomonas, both of which contain several non-pathogenic strains as well.

The microbial community composition shifted markedly from the simulated POC to POU, with samples collected at the POC having a greater alpha diversity (Simpson; 0.916±0.077) than simulated POU (0.541±0.315) or biofilm samples (0.670±0.302) (p≤0.0006). This suggests that the conditions present in the premise plumbing rigs selected for bacteria well-suited for regrowth, rather than indiscriminately enriching all bacteria. Blended waters trended towards having a higher alpha diversity than potable waters at the POC (0.930±0.067 vs. 0.882±0.097) as well as the

164

Figure 7-2: Microbial community profiles for simulated premise plumbing pipe rigs. Samples profiled at the simulated point of compliance (POC), simulated point of use (POU), and in the biofilm (BF) as determined by amplicon sequencing of using universal primers targeting V4 region of Bacteria and Archaeae. (1) and (2) indicate experimental replicate premise plumbing rigs. Surface water and Groundwater conditions refer to treated potable water derived from each source. AWP = advanced water purification using membrane filtration, reverse osmosis and ultraviolet irradiation with the addition of hypochlorite or hydrogen peroxide; O3-BAC = ozone followed by biologically active carbon filtration; Past = refers to treatment using pasteurization; Industrial refers to wastewater derived from an industrial source rather than municipal wastewater, as in all other cases. Refer to Table 7-1 for additional details regarding treatment schemes.

165

simulated POU (0.582±0.320 vs. 0.435±0.291) and biofilm (0.681±0.300 vs. 0.646±0.323), though differences were not significant (p≥0.1526).

Regrowth of OPPPs

The occurrence of OPs at the simulated POU versus POC was explored quantitatively via qPCR targeting genes specific for Legionella spp., Mycobacterium spp., and P. aeruginosa. In Utility 1 scenarios, there were no significant differences among scenarios for OPs at the POC (Wilcoxon, p≥0.4533) or in the conditioned biofilm (p≥0.1859). There were no significant differences at the simulated POU in Legionella spp. levels for Utility 1 scenarios (p≥0.1709) and the gene was only detected at sub-quantifiable levels in the 90% Surface/10% O₃-BAC conditions and 100% Surface conditions. Mycobacterium spp. genes were more abundant in the 90% Surface/10% O₃-BAC scenario than the 90% Surface/10% AWP scenario (p=0.0044), indicating that membrane filtration may offer benefits to limiting regrowth of this genus of bacteria compared to biofiltration. The potable source water scenario resulted in more Mycobacterium spp. genes in the rig POU than either AWP DPR blend condition (p=0.0013), suggesting that blending the potable source water with highly treated DPR water actually offered benefits for limiting regrowth of Mycobacteria spp.

While Legionella spp. genes were detected in all Utility 2 rig POU samples, there were no significant differences in abundance of the gene among scenarios (p≥0.1047). Pasteurized water also supported regrowth of Mycobacterium spp., while the corresponding non-pasteurized scenario did not (p=0.0046). There were no significant differences in abundance of OP gene markers detected in biofilms across scenarios (p≥0.2453). In Utility 3 scenarios, Legionella spp. gene markers were not detected in any POU water or biofilm samples. Greater Mycobacterium spp. regrowth was observed in the 90% Surface/10% O₃-BAC and the 50% Surface/50% O₃-BAC POU water than in the potable source water POU (p≤0.0323). In the biofilm, Mycobacterium spp. was detected in both the 90% Surface/10% Tertiary and 50% Surface/50% O₃-BAC rigs, but concentrations were not significantly greater than scenarios where the gene was not detected (p≥0.2207). There were no OP gene markers detected in any POC, POU, or biofilm samples from Utility 4 scenarios. P. aeruginosa genes were not detected in POC, POU or biofilm samples collected from any scenarios from any utilities. Thus, the Pseudomonadaceae detected by 16S rRNA gene amplicon sequencing were likely non-pathogenic strains.

When compared to traditional potable source waters, DPR treatment schemes from Utilities 1, 2, and 4 all produced reuse waters that were successful at limiting regrowth of OPs. Membrane filtration appears to be a particularly promising treatment approach for limiting regrowth of OPs, as membrane filtered waters from Utility 1 tended to harbor less regrowth than biofiltered waters. When treatment approach alone was not sufficient to limit OP regrowth, selection of an optimal blend ratio appears to be a particularly promising approach for limiting regrowth, as more regrowth was observed in greater blend ratios for Legionella spp. in Utility 2 and Mycobacterium spp. in Utility 3.

Previous studies have indicated that both Legionella spp. and Mycobacterium spp. are widespread in non-potable reclaimed water at the POU (Fahrenfeld et al., 2013; Jjemba et al., 2010; Whiley et al., 2015), highlighting the importance of identifying AWP treatment approaches

166

that can produce finished water that effectively limits regrowth of OPs in distribution systems and premise plumbing. Though regrowth of OPs has not been previously studied in DPR distribution systems or the associated premise plumbing, results of this study indicate that AWP is highly effective at producing biostable waters that do not support regrowth of OPs. The conditions in this study represent a worst-case scenario for premise plumbing with long stagnation periods, however further study examining the relevance of these trends in full-scale systems over longer time periods and with mature biofilms would be valuable.

Occurrence of ARGs

While the presence of diverse ARGs belonging to the antibiotic classes glycopeptide, macrolide, sulfonamide, tetracycline have been previously documented in non-potable reuse water distribution systems (Fahrenfeld et al., 2013), to the authors’ knowledge, this study is the first to examine the presence of ARGs in DPR waters. There are currently no standard methodologies for monitoring antibiotic resistance in recycled water, but it is becoming common practice to quantify a number of target ARGs as conservative indicators for assessing potential for selection and spread of resistance in various aquatic reservoirs (Berendonk et al., 2015). Monitoring ARGs provides a conservative indicator because, while bacteria may be killed, their DNA carrying ARGs still has the potential to be taken up by downstream bacteria. Here, two ARGs, vanA and qnrA, which encode resistance to critically-important antibiotics, vancomycin and quinolones, respectively, were quantified via qPCR. The gene capture element intI1 was also measured as a broad indicator for anthropogenic influence and potential for horizontal transfer of multi-antibiotic resistance (Figure 7-1).

In Utility 1 samples, vanA was widely detected, but at sub-quantifiable concentrations, with the 90% Surface/10% AWP scenario yielding the only quantifiable occurrence (Figure 7-1). There were no significant differences between scenarios for vanA concentration in either the POU water (Wilcoxon, p=0.1709) or the biofilm (p≥0.4142), and qnrA and intI1 were not detected in any biofilm samples. In Utility 2 samples, ARGs and intI1 were detected sporadically at sub- quantifiable levels and there were no differences between scenarios for the POU water (p≥0.0764) or biofilm (p≥0.4795) for any of the genes. For Utility 3 scenarios, intI1 and qnrA were occasionally detected but not quantifiable. There were no significant differences in gene abundances between scenarios for either gene in the POU water (p≥0.3816) or biofilm samples (p≥0.6171). No ARGs were detected in any samples from Utility 4 scenarios. Together, these results are a promising preliminary indication that these highly treated DPR waters do not pose added risk in terms of producing waters enriched in ARGs, relative to traditional potable water, or in proliferating ARGs in the distribution system. However, this issue merits further monitoring, particularly as more standardized tools for ARG monitoring become available and over longer study periods.

A subset of samples representing the traditional potable source water and the 10% DPR blend of each scenario were subject to shotgun metagenomic sequencing with annotation against the Comprehensive Antibiotic Resistance Database (McArthur et al., 2013) to broadly profile ARGs beyond those quantified by qPCR. A large portion of the samples (all samples from Utility 2 and a subset from Utilities 3 and 4) did not yield sufficient DNA to conduct metagenomic sequencing. Of the scenarios that could be sequenced (Figure 7-3), annotation of shotgun

167

metagenomic reads against CARD revealed that total abundances of known ARGs ranged from 1.5 to 6.5 log gene copies per milliliter in bulk water and from 3.7 to 6.0 log gene copies per biofilm swab in samples that were able to be sequenced. From the samples that were sequenced, a total of 212 different ARGs were detected across the dataset, ranging from five to 94 different ARGs per sample. ARG profiles varied strongly according to utility (Figure S1; ANOSIM, R=0.789, p=0.001), but did not vary as strongly by scenario within each utility (R=0.307, p=0.012). Multidrug, trimethoprim, and aminoglycoside resistance were abundant across all utilities, while rifampin, tetracycline, and peptide resistance were notably abundant only in Utility 4 samples. Overall, the ten most abundant genes across the dataset included one trimethoprim resistance gene (dfrE), two rifampin genes (RbpA, arr-1), one aminoglycoside gene (AAC(2')-Ib), and five multidrug resistance genes (mtrA, mdtB, adeG, ceoB, acrB, mexF). Utility 4 samples were largely dominated by dfrE, mtrA, RbpA, AAA(2')-Ib, and arr-1, while Utility 1 and 3 samples were dominated by dfrE, mdtB, adeG, ceoB, and acrB. This strong grouping of ARGs by utility in standardized premise plumbing rigs, suggests that overarching factors such as geography and physicochemical and microbial characteristics of potable source water (e.g. the original source of the vast majority of wastewater) are likely to be critical determinants influencing the resistome of DPR water.

Regrowth of HPC bacteria capable of growth on antibiotic-supplemented media

While molecular methods are extremely powerful for detecting ARGs in viable cells, irrespective of their culturability, molecular methods also capture DNA from dead cells, extracellular DNA, and DNA incorporated into the biofilm in the form of extracellular polymeric substances. To verify the presence of viable of phenotypically resistant heterotrophic bacteria, water samples were plated onto R2A media amended with antibiotics (Table S2). While comparing results across conditions can provide an indication of the potential for blending reuse waters with traditional source waters to increase growth on antibiotic-supplemented media, there are important limitations. Because R2A agar captures a variety of heterotrophic bacteria, it is impossible to identify the minimum inhibitory concentration (MIC) for each present species. Growth may not necessarily be indicative of resistance, but rather intrinsic resistance (for example, some antibiotics are not effective against Gram positive bacteria), intermediate resistance, or sub-inhibitory antibiotic concentrations. With these limitations in mind, we compared HPC growth in the presence of a suite of antibiotics as further evidence to identify treatment schemes that potentially limit resistant strains or community shifts favoring species with inherent resistance or higher MICs. Such an approach has been previously applied for drinking water (Xi et al., 2009).

168

Figure 7-3: Shotgun metagenomic abundances of ARGs by antibiotic class. Abundance of ARGs are reported per milliliter of water sampled or per biofilm swab as determined by shotgun metagenomic sequencing of samples collected after eight weeks of conditioning at the point of compliance (POC), point of use (POU), and in the biofilm (BF). Numbers in parentheses indicate the replicate rig number. Sufficient DNA for analysis of potable water samples was only recoverable for select samples from Utilities 1, 3, and 4. AWP = advanced water purification using membrane filtration, reverse osmosis and ultraviolet irradiation with the addition of hypochlorite or hydrogen peroxide; O3-BAC = ozone followed by biologically active carbon filtration; Past = refers to treatment using pasteurization; Ind refers to wastewater derived from an industrial source rather than municipal wastewater, as in all other cases. Refer to Table 7-1 for additional details regarding treatment schemes.

Of the 100% Surface water and 50% Surface/50% AWP blend tested from Utility 1, the 50% Surface/50% AWP blend consistently produced less HPCs capable of growth on media amended with antibiotics, indicating that blending with highly-treated AWP water might have benefits for reducing microbial water quality problems in general. While neither condition had detectable HPCs at the POC, regrowth of HPCs occurred during simulated use in both sets of premise plumbing rigs. Particularly high levels of regrowth were noted from the 90% Surface/10% AWP blend on media amended with sulfamethoxazole (240, 69% of total HPCs for rigs 1 and 2, respectively) and vancomycin (65, 258%). Utility 2 did not produce any quantifiable HPCs capable of growth on antibiotic-amended HPC agar. Utility 3’s 100% Surface condition did not produce any quantifiable HPCs at the POC or POU, but regrowth of total and resistant HPCs occurred for both tertiary blends. The 90% Surface/10% Tertiary blend rigs produced at least 1.8 log more total HPCs than the 95% Surface/5% Tertiary blend, demonstrating that selection of the appropriate blend ratio can effectively control regrowth of HPCs in premise plumbing. From Utility 4, only the two potable waters (100% Surface and 100% Groundwater) produced quantifiable HPCs capable of growth on antibiotic media. Again, these results suggest that blending traditional

169

potable source waters with highly treated reuse water can actually have benefits for limiting regrowth of total HPCs and HPCs capable of growth on antibiotic media.

Microbially-influenced corrosion

Other nuisance bacteria may regrow during distribution that are of concern for maintaining water infrastructure and aesthetics and can indirectly cause health concerns. BART tests were used to identify bacteria associated with microbially-influenced corrosion following incubation in the pipe rigs (Table 7-2). Heterotrophic aerobic bacteria were rarely detected, and when they were present, they were present at low abundances, below approximately 7,000 colony forming units (CFU) per milliliter. Sulfate-reducing bacteria were detected in all utilities at abundances up to approximately 2,200,000 CFU per milliliter, indicating that anaerobic conditions likely developed within rigs, regardless of utility or scenario. Acid-producing bacteria were only detected at a quantifiable range in the 90% Surface/10% Tertiary blend from Utility 3, indicating that overall, these bacteria can be controlled by AWP and selection of an appropriate blend ratio. Slime- producing bacteria were ubiquitous, but were particularly abundant in in the 100% Surface and tertiary blends of Utility 3. While nitrifying bacteria were not detected in any samples, denitrifying bacteria were notably abundant in the tertiary blends of Utility 3 and all scenarios from Utility 4, further suggesting that anaerobic conditions developed during simulated distribution. The development of low dissolved oxygen conditions in premise plumbing is not uncommon, and has been previously documented after elevated water age is reached in distribution systems (Masters et al., 2015), especially in systems experiencing nitrification (Wilczak et al., 1996). Given that Utilities 1, 2, and 4 all use a chloramine residual, nitrification is a particular concern given the availability of nitrogen via ammonia (Zhang et al., 2009).

While bacteria associated with microbially-influenced corrosion are important because of their ability to contribute to corrosion of distribution system pipes and plumbing materials, they can also be detrimental to water aesthetics. Sulfate-reducing bacteria, slime-producing bacteria, and denitrifying bacteria emerged as the primary concerns in the studies scenarios. Sulfate- reducing bacteria can contribute to corrosion pitting and undesirable taste and odor problems, such as the production of black slimes and rotten-egg odor (Jacobs and Edwards, 2000). The elevated abundance of sulfate-reducing bacteria under nearly all conditions suggests that DPR waters may be particularly susceptible to supporting the growth of these microorganisms. Slime-producing bacteria are associated with the production of excessive biofilms and extracellular polymeric substances that can corrode metal pipes, plug pipes, and cause undesirable taste and odor and water cloudiness (Little et al., 2007). Denitrifying bacteria may be associated with increased pH, corrosion of metal pipes, and undesirable taste and odor (Masters et al., 2015). Given that sulfate- reducing and denitrifying bacteria can grow only in anaerobic conditions, distribution system operation will be critical for limiting growth of these organisms in full-scale systems. Limiting stagnation and minimizing water age at the POU can aid in the prevention of redox conditions favorable to growth of these microorganisms (Masters et al., 2015). Additionally, maintaining a disinfectant residual at the point-of-use can aid in control of these organisms and their undesirable consequences.

170

Table 7-2: Abundnaces of microorganisms associated with microbially-influenced corrosion. Approximate abundances (CFU/mL) of bacteria associated with microbially-influenced corrosion measured at the simulated point of compliance (POC) as determined using Biological Activity Reactivity Tests. AWP = advanced water purification using membrane filtration, reverse osmosis and ultraviolet irradiation with the addition of hypochlorite or hydrogen peroxide; O3-BAC = ozone followed by biologically active carbon filtration; Past = refers to treatment using pasteurization. Refer to Table 7-1 for additional details regarding treatment schemes. HAB SRB APB SLYM N DN a 100% Surface ND Present ND Present ND NA 90% Surface/10% O₃-BAC ND Present ND Present ND NA 90% Surface/10% AWP ND Present ND Present ND NA Utility 1 50% Surface/50% AWP ND Present ND Present ND NA

100% Groundwater ND 2,200,000 ND 500 ND ND 90% Groundwater/10% AWP 6,500 2,200,000 ND 500 ND 3,000 50% Groundwater/50% AWP ND 2,200,000 ND 500 ND 3,000 Utility 2 50% Groundwater/50% Past.-AWP 6,500 2,200,000 ND 500 ND ND 100% Surface ND 2,200,000 ND 1,750,000 ND 1,800,000

95% Surface/5% Tertiary ND 2,200,000 ND 1,750,000 ND 1,800,000 90% Surface/10% Tertiary 6,500 2,200,000 70,000 1,750,000 ND 1,800,000

Utility 3 90% Surface/10% O₃-BAC ND 2,200,000 ND 500 ND 3,000 50% Surface/50% O₃-BAC ND 2,200,000 <100 500 ND ND

100% Groundwater ND 2,200,000 ND 440,000 ND 1,800,000 90% Groundwater/10% AWP ND 2,200,000 ND 440,000 ND 1,800,000 90% Surface/10% AWP ND 2,200,000 ND 440,000 ND 1,800,000 Utility 4 90% Finished/10% Industrial-AWP ND 6,000 ND 350,000 ND 215,000 aDue to a recording error, the approximate concentration cannot be accurately determined; ND = not detected, NA = not tested, HAB = heterotrophic aerobic bacteria, SRB = sulfate-reducing bacteria, APB = acid-producing bacteria, SLYM = slime-producing bacteria, N = nitrifying bacteria, DN = denitrifying bacteria

Water chemistry

Organic carbon is a critical nutrient supporting regrowth of microorganisms in treated water. While TOC and DOC have long been used as correlates to the level of organic carbon in the water during distribution, BDOC has been proposed as an alternative indicator that more accurately reflects the biodegradable fraction and overall biostability of water (Servais et al., 1987). Average TOC for each scenario ranged from 50.7 to 5790 ppb, average DOC ranged from 4.5 to 5710 ppb, and average BDOC ranged from sub-quantifiable to 1850 ppb (Figure 7-4). Previous studies that have found BDOC in reclaimed water to range from 400 to 6200 ppb (Jjemba et al., 2010) and from 20 to 930 ppb in potable water (Charnock and Kjønnø, 2000; Ribas et al., 1991). With the exception of the 90% Surface/10% AWP and 100% O₃-BAC conditions from Utility 1, all other reuse scenarios fell below 930 ppb BDOC, indicating that they are of comparable biostability to potable waters documented in the literature.

With the exception of the 90% Surface/10% O₃-BAC scenario, all Utility 1 treatment schemes yielded significantly higher concentrations of BDOC than the 100% surface condition

171

(Wilcoxon, p≤0.0452), indicating that all DPR treatment scenarios would result in a greater potential for bacterial regrowth than the traditional potable treated water of that utility. Membrane filtration appears particularly promising for reducing BDOC, as the 90% Surface/10% O3-BAC treatment train resulted in less BDOC than the comparable 90% Surface/10% AWP treatment (p=0.0031). Though blend ratio did not significantly affect BDOC concentrations for scenarios utilizing biofiltration treatment, the blend ratio did affect BDOC in membrane-based treatment, with the 90% Surface/10% AWP being greater than the 50% Surface/50% AWP scenario (p=0.0034). Of Utility 2 and 4 scenarios, no DPR treatment schemes produced BDOC concentrations greater than the respective potable scenario (p≥0.0998). Of Utility 3 scenarios, both reuse schemes produced BDOC concentrations that were not significantly different from the finished condition (p≥0.1503), though the 100% surface condition produced by pilot scale treatment was greater than the BDOC concentration of the 100% finished condition produced by the full-scale potable treatment plant (p=0.0241).

There were no significant positive correlations between BDOC and total bacteria, as measured by 16S rRNA genes or HPC, or between BDOC and any OP or ARG gene markers for any of the utilities (Spearman, p≥0.2188). This result suggests that in this highly treated DPR water, organic carbon is not the limiting nutrient determining biostability of finished water. Previous studies have demonstrated that in order for organic carbon to be the limiting nutrient constraining bacterial regrowth, extremely low concentrations of below 10 ppb as AOC are required (Kooij, 1992; Williams et al., 2015).

CONCLUSIONS

Advancements in treatment technologies have facilitated the ability to produce high-quality water from wastewater suitable for potable purposes, but little is known about the biological stability of this DPR water. This study simulated use of premise plumbing, where microbial regrowth potential is anticipated to be the greatest, using blended DPR water from four U.S. water utilities. To our knowledge, this study represents the first such simulation, and here we aimed to be comprehensive in the approach, including five different baseline waters from four water utilities along with comprehensive culture- and molecular-based characterization of the resident microbial communities and factors influencing their regrowth. Results were compared to the regrowth observed for traditional potable water currently in use at the same utilities. Across tested treatment and blending scenarios, total bacteria (i.e., 16S rRNA genes) were more abundant at the POU and the POC, for both DPR blends and traditional potable water. However, regrowth of total bacteria, OPs, and ARGs was not significantly greater for any DPR blends treated with AOPs than in the corresponding potable water. The one scenario in which OP regrowth exceeded the potable condition was a 10% blend of tertiary treated DPR water, but this regrowth was effectively limited by selecting an appropriate blend ratio (5%). The overall microbial community composition was unique at the POU compared to the POC, with greater alpha diversity observed at the POC, suggesting that simulated premise plumbing conditions selected for particular bacteria that were well-suited for regrowth. Measurements of BDOC suggested that DPR water generally possessed greater potential to facilitate regrowth of bacteria than the corresponding potable waters, though the lack of corresponding regrowth under these conditions suggests that organic carbon was not the limiting nutrient in these waters. Regrowth was observed for several microorganisms associated with microbially-influenced corrosion. Sulfate-reducing bacteria, in particular, grew in

172

Figure 7-4: Organic carbon measurements in water prior to incubation in simulated premise plumbing pipe rigs. Total organic carbon (TOC), dissolved organic carbon (DOC), and biodegradable dissolved organic carbon (BDOC) concentrations measured at the point of compliance (POC). When both a 100% Finished and 100% Surface condition are indicated, the 100% Finished condition represents the corresponding surface water treated at the full-scale potable water treatment plant, as opposed to the bench-scale treatment simulated for the 100% Surface condition. AWP = advanced water purification using membrane filtration, reverse osmosis and ultraviolet irradiation with the addition of hypochlorite or hydrogen peroxide; O3-BAC = ozone followed by biologically active carbon filtration; Past = refers to treatment using pasteurization. Refer to Table 7-1 for additional details regarding treatment schemes.

173

all conditions, so control of these bacteria in DPR systems could present a critical challenge. It is important to recognize; however, that this study was carried out over a short time period of only eight weeks, which was aimed to capture key differences of the kinds of microbes expected to colonize, proliferate, and establish in premise plumbing when DPR water is introduced. Given that microbial communities undergo extensive, long-term succession patterns, additional longer- term monitoring is recommended as DPR waters are integrated into potable water distribution systems. Additional research is also needed to determine the applicability of these findings when implementing DPR under less stringent treatment scenarios or when using higher blends of DPR water.

ACKNOWLEDGEMENTS

We thank the utilities that participated in this study for their support. This work is supported by the Water Research Foundation (WRF 4536) and National Science Foundation through the Graduate Research Fellowship Program, CBET Award 1438328, and PIRE Award 1545756. Additional support was provided by the Alfred P. Sloan Foundation Microbiology of the Built Environment program, the Water Environment & Research Foundation Paul L. Busch award, the Virginia Tech Institute for Critical Technology and Applied Science Center for Science and Engineering of the Exposome, and the American Water Works Association Abel Wolman Doctoral Fellowship.

REFERENCES

Alexander, J., Knopp, G., Dötsch, A., Wieland, A., Schwartz, T., 2016. Ozone treatment of conditioned wastewater selects antibiotic resistance genes, opportunistic bacteria, and induce strong population shifts. Sci. Total Environ. 559, 103–112. American Public Health Association (APHA), 2005. Standard Methods for the Examination of Water and Wastewater. American Water Works Association, Water Environment Federation. Applegate, L.E., 2017. Posttreatment of Reverse Osmosis Product Waters. J. Am. Water Work. Assoc. 78, 59–65. Berendonk, T.U., Manaia, C.M., Merlin, C., Fatta-Kassinos, D., Cytryn, E., et al., 2015. Tackling antibiotic resistance: the environmental framework. Nat. Rev. Microbiol. 13, 310–317. Caporaso, J.G., Lauber, C.L., Walters, W.A., Berg-Lyons, D., Huntley, J., et al., 2012. Ultra-high- throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J. 6, 1621–1624. Centers for Disease Control and Prevention, 2011. Surveillance for Waterborne Disease Outbreaks and Other Health Events Associated with Recreational Water — United States, 2007–2008. Morb. Mortal. Wkly. Rep. 60. Charnock, C., Kjønnø, O., 2000. Assimilable organic carbon and biodegradable dissolved organic carbon in Norwegian raw and drinking waters. Water Research 34, 2629–2642. Chowdhury, S., 2012. Heterotrophic bacteria in drinking water distribution system: a review., Environmental monitoring and assessment. Environ. Monit. Assess. 184(10), 6087-6137. Czekalski, N., Imminger, S., Salhi, E., Veljkovic, M., Kle, K., Drissner, D., Hammes, F., Bu, H., Gunten, U.V., 2016. Inactivation of Antibiotic Resistant Bacteria and Resistance Genes by Ozone: From Laboratory Experiments to Full-Scale Wastewater Treatment. Environ. Sci. Technol. 50, 11862-11871.

174

Escobar, I.C., Randall, A.A., 2000. Sample storage impact on the assimilable organic carbon (AOC) bioassay. Water Research 34, 1680–1686. Fahrenfeld, N.L., Ma, Y.J., O’Brien, M., Pruden, A., 2013. Reclaimed water as a reservoir of antibiotic resistance genes: distribution system and irrigation implications. Front. Microbiol. 4, 10. Falkinham, J., 2015. Common Features of Opportunistic Premise Plumbing Pathogens. Int. J. Environ. Res. Public Health 12, 4533–4545. Garner, E., Benitez, R., Wagoner, E. Von, Sawyer, R., Schaberg, E., Hession, W.C., Krometis, L.H., Badgley, B.D., Pruden, A., 2017. Stormwater loadings of antibiotic resistance genes in an urban stream. Water Research 123, 144–152. Garner, E., Zhu, N., Strom, L., Edwards, M., Pruden, A., 2016. A human exposome framework for guiding risk management and holistic assessment of recycled water quality. Environ. Sci. Water Res. Technol. 2, 580-598. Gerrity, D., Pecson, B., Trussell, R.S., Trussell, R.R., 2013. Potable reuse treatment trains throughout the world. J. Water Supply Res. Technol. 62, 321–338. Gosling, S.N., Arnell, N.W., 2016. A global assessment of the impact of climate change on water scarcity. Clim. Change 134, 371–385. Hong, P.-Y., Al-Jassim, N., Ansari, M., Mackie, R., 2013. Environmental and Public Health Implications of Water Reuse: Antibiotics, Antibiotic Resistant Bacteria, and Antibiotic Resistance Genes. Antibiotics 2, 367–399. Jacobs, S., Edwards, M., 2000. Sulfide scale catalysis of copper corrosion. Water Research 34, 2798–2808. Ji, P., Rhoads, W.J., Edwards, M.A., Pruden, A., 2017. Impact of water heater temperature setting and water use frequency on the building plumbing microbiome. ISME J. 11, 1318–1330. Jjemba, P.K., Weinrich, L.A., Cheng, W., Giraldo, E., LeChevallier, M.W., 2010. Regrowth of potential opportunistic pathogens and algae in reclaimed-water distribution systems. Appl. Environ. Microbiol. 76, 4169–4178. Kim, S.D., Cho, J., Kim, I.S., Vanderford, B.J., Snyder, S.A., 2007. Occurrence and removal of pharmaceuticals and endocrine disruptors in South Korean surface, drinking, and waste waters. Water Research 41, 1013–1021. Kooij, D. Van Der, 1992. Assimilable Organic Carbon as an Indicator of Bacterial Regrowth. J. Am. Water Work. Assoc. 57–65. LeChevallier, M.W., Schulz, W., Lee, R.G., 1991. Bacterial nutrients in drinking water. Appl. Envir. Microbiol. 57, 857–862. Little, B.J., Mansfeld, F.B., Arps, P.J., Earthman, J.C., 2007. Microbiologically Influenced Corrosion, in: Corrosion in Special Environments. pp. 662–685. Liu, G., Bakker, G.L., Li, S., Vreeburg, J.H.G., Verberk, J.Q.J.C., Medema, G.J., Liu, W.T., Van Dijk, J.C., 2014. Pyrosequencing reveals bacterial communities in unchlorinated drinking water distribution system: An integral study of bulk water, suspended solids, loose deposits, and pipe wall biofilm. Environ. Sci. Technol. 48, 5467–5476. Masters, S., Wang, H., Pruden, A., Edwards, M.A., 2015. Redox gradients in distribution systems influence water quality, corrosion, and microbial ecology. Water Research 68, 140–149. McArthur, A.G., Waglechner, N., Nizam, F., Yan, A., Azad, M.A., et al., 2013. The comprehensive antibiotic resistance database. Antimicrob. Agents Chemother. 57, 3348– 3357. Narahimhan, R., Brereton, J., Abbaszadegan, M., Ryu, H., Butterfield, P., Thompson, K., Werth,

175

H., 2005. Characterizing Microbial Water Quality in Reclaimed Water Distribution Systems. AwwaRF Report. National Sanitation Foundation, 2007. NSF International Standard/American National Standard 61: Drinking Water System Components-Health Effects. Ann Arbor, MI. Nguyen, C., Elfland, C., Edwards, M., 2012. Impact of advanced water conservation features and new copper pipe on rapid chloramine decay and microbial regrowth. Water Research 46, 611– 21. Proctor, C.R., Gächter, M., Kötzsch, S., Rölli, F., Sigrist, R., Walser, J.-C., Hammes, F., 2016. Biofilms in shower hoses – choice of pipe material influences bacterial growth and communities. Environ. Sci. Water Res. Technol. 8–11. Pruden, A., 2014. Balancing Water Sustainability and Public Health Goals in the Face of Growing Concerns about Antibiotic Resistance. Environ. Sci. Technol. 48, 5–14. Rhoads, W.J., Pruden, A., Edwards, M.A., 2016. Survey of green building water systems reveals elevated water age and water quality concerns. Environ. Sci. Water Res. Technol. 2, 164-173. Ribas, F., Frias, J., Lucena, F., 1991. A new dynamic method for the rapid determination of the biodegradable dissolved organic carbon in drinking water. J. Appl. Bacteriol. 71, 371–378. Servais, P., A. Anzil, Ventresque, C., 1989. Simple method for determination of biodegradable dissolved organic carbon in water. Appl. Environ. Microbiol. 55(10), 2732–2734. Servais, P., Billen, G., Hascoet, M.C., 1987. Determination of the Biodegradable Fraction of Dissolved Organic Matter in Waters. Water Research 21, 445–450. Snyder, S.A., Wert, E.C., Lei, H.D., Westerhoff, P., Yoon, Y., 2007. Removal of EDCs and pharmaceuticals in drinking and reuse treatment processes. AwwaRF Report. Thayanukul, P., Kurisu, F., Kasuga, I., Furumai, H., 2013. Evaluation of microbial regrowth potential by assimilable organic carbon in various reclaimed water and distribution systems. Water Research 47, 225–232. US Environmental Protection Agency, 2012. Guidelines for Water Reuse. Wang, H., Masters, S., Hong, Y.J., Stallings, J., Falkinham, J.O., Edwards, M.A., Pruden, A., 2012. Effect of Disinfectant, Water Age, and Pipe Material on Occurrence and Persistence of Legionella, mycobacteria, Pseudomonas aeruginosa, and Two Amoebas. Environ. Sci. Technol. 46, 11566–11574. Wang, W., Wu, Q., Huang, N., Wang, T., Hu, H., 2016. Synergistic effect between UV and chlorine (UV/chlorine) on the degradation of carbamazepine: Influence factors and radical species. Water Research 98, 190–198. Watkinson, A.J., Murby, E.J., Costanzo, S.D., 2007. Removal of antibiotics in conventional and advanced wastewater treatment: Implications for environmental discharge and wastewater recycling. Water Research 41, 4164–4176. Whiley, H., Keegan, A., Fallowfield, H., Bentham, R., 2015. The presence of opportunistic pathogens, Legionella spp., L. pneumophila and Mycobacterium avium complex, in South Australian reuse water distribution pipelines. J. Water Health 13, 553. Wilczak, A., Jacangelo, J.G., Marcinko, J.P., Odell, L.H., Kirmeyer, G.J., Wolfe, R.L., 1996. Occurrence of nitrification in chloraminated distribution systems. J. Am. Water Work. Assoc. 88, 74–85. Williams, K., Pruden, A., Falkinham, J., Edwards, M., 2015. Relationship between Organic Carbon and Opportunistic Pathogens in Simulated Glass Water Heaters. Pathogens 4, 355– 372. Xi, C., Zhang, Y., Marrs, C.F., Ye, W., Simon, C., Foxman, B., Nriagu, J., 2009. Prevalence of

176

antibiotic resistance in drinking water treatment and distribution systems. Appl. Environ. Microbiol. 75, 5714–5718. Yoder, J., Roberts, V., Craun, G.F., Hill, V., Hicks, L., Alexander, N.T., V.R., Calderon, R.L., Beach, M.J., Carolina, N., 2008. Surveillance for Waterborne Disease and Outbreaks Associated with Drinking Water and Water not Intended for Drinking --- United States, 2005- -2006. Morb. Mortal. Wkly. Rep. 57, 39–62. Yoon, Y., Jung, H., Yoong, D., Di, W., Dodd, M.C., Hur, H., Lee, Y., 2017. Inactivation efficiency of plasmid-encoded antibiotic resistance genes during water treatment with chlorine, UV, and UV/H2O2. Water Research 123, 783–793. Zhang, Y., Love, N., Edwards, M., 2009. Nitrification in Drinking Water Systems, Critical Reviews in Environmental Science and Technology.

177

SUPPLEMENTARY INFORMATION FOR CHAPTER 7

Pipe rig pre-testing

Forty pipe rigs were constructed and pre-tested according to National Sanitation Foundation Standard (NSF) 61 to obtain the most reproducible lead and copper (NSF, 2007). Briefly, rigs were rinsed three times with distilled water, followed by three times with NSF extraction water (synthesized water with the following characteristics: pH = 8.0; alkalinity = 500 mg/L as CaCO3, dissolved inorganic carbon = 122 mg/L, free chlorine = 2 mg/L). Rigs were filled with NSF extraction water and allowed to stagnate for one day. Water was discarded, then rigs were treated with three consecutive 12-hour stagnation periods. The four rigs producing the greatest lead and copper concentration variation from these three periods were excluded from the study.

Simulated treatment

For Utility 1, following blending, all waters were subject to bench-scale treatment consisting of O3 pretreatment (2 mg/L), coagulation with alum (38 mg/L), cationic polymer (1.5 mg/L), and non-ionic polymer (0.18 mg/L), flocculation, sedimentation, filtration (1.5 μm glass fiber filter), and chlorination followed by the addition of aqua ammonia to form chloramines (target of 1.8 to 2.2 mg/L total chlorine after 11 minutes). For Utility 2, recycled water was treated prior to blending with pilot-scale treatment of pasteurization, ultrafiltration, reverse osmosis, and ultraviolet irradiation with the addition of hydrogen peroxide. Following blending, waters were subject to secondary disinfection with chlorine followed by ammonia to achieve a target residual of 3.8 mg/L total chlorine. For Utility 3, for O3-BAC conditions, recycled water was treated with ozone and biofiltration prior to blending. Tertiary waters were treated via secondary treatment with nitrification, partial denitrification, and biological phosphorus removal via full-scale treatment. After blending, water was treated at the bench-scale with coagulation with ferric chloride (0.6 mg/L), flocculation, sedimentation, filtration (0.7 μm glass fiber filter), and chlorination (target dose of 1.5 mg/L after 2 hours). For Utility 4, groundwater and surface water were each blended with recycled water that had been previously treated with microfiltration, reverse osmosis, ultraviolet irradiation with the addition of hydrogen peroxide, and stabilization via pH adjustment and the addition of calcium chloride to achieve a Langelier saturation index between -0.5 and 0.5. Following blending, blends were treated at the bench scale via O3 (0.5 mg/L), coagulation with ferric chloride (1.5 mg/L) and cationic polymer (1.2 mg/L), flocculation, sedimentation, filtration (1.5 μm glass fiber filter), and chlorine followed by aqua ammonia for chloramination (target dose of 2.6 mg/L total chlorine after 16.5 minutes), and the additional of zinc orthophosphate for corrosion inhibition. If disinfectant residuals degraded during storage, chlorine or chloramines were spiked to bring concentrations back up to target concentrations described above before rig incubation.

Quantitative polymerase chain reaction

All qPCR assays were conducted using previously published primers and thermocycler conditions (Table S1). The following genes were targeted for quantification of OPs: a highly specific region of the 23S rRNA gene for Legionella spp. (Nazarian et al., 2008), a highly specific region of the

178

16S rRNA gene for Mycobacterium spp. (Radomski et al., 2010), and the ecfX and gyrB genes for Pseudomonas aeruginosa (Anuj et al., 2009). Two ARGs were also quantified: a quinolone resistance gene, qnrA (Colomer-Lluch et al., 2014), and a vancomycin resistance gene, vanA (Dutka-Malen et al., 1995), along with the class 1 integron integrase gene intI1 (Hardwick et al., 2008). The universal bacterial gene, 16S rRNA, was also quantified (Suzuki et al., 2000). All non- probe assays (16S rRNA genes, vanA, intI1) were performed in triplicate 10 µl reactions that included 5 µl SsoFast EvaGreen SuperMix (Bio-Rad, Hercules, CA), 0.8 µl of forward and reverse primers at 5 µM (Integrated DNA Technologies, Coralville, IA), 2.4 µl molecular grade water, and 1 µl sample. All probe assays (Legionella spp., Mycobacterium spp., P. aeruginosa, qnrA) were performed in triplicate 10 µl reactions that included 5 µl SsoFast Probes SuperMix (Bio-Rad), 0.5 µl of each forward and reverse primer at 5 µM, 0.19 µl of each probe at 10 µM, 1 µl sample, and molecular grade water to reach the total reaction volume.

Data analysis for 16S rRNA gene amplicon sequencing and shotgun metagenomics

Processing of 16S rRNA gene amplicon sequencing reads was conducted using the QIIME pipeline (Caporaso et al., 2010) with annotation against the Greengenes database (May 2013 release; DeSantis et al., 2006). Samples were rarefied to 10,000 randomly selected reads. Alpha diversity was calculated in QIIME using the Simpson index.

From shotgun metagenomic sequencing data, annotation of ARGs was conducted on the MetaStorm platform (Arango-Argoty et al., 2016) according to default parameters using annotation to the Comprehensive Antibiotic Resistance Database (CARD; version 1.0.6) for ARGs (McArthur et al., 2013), Silva ribosomal RNA database (version 123) for 16S rRNA genes (Quast et al., 2013), BacMet database (version 1.1) for metal resistance genes (Pal et al., 2014), and ACLAME database (version 0.4) for plasmid-associated genes (Leplae et al., 2004). Relative abundances were calculated by normalizing gene counts to abundance of 16S rRNA genes as well as target gene and 16S rRNA gene length as proposed by Li et al. (2015). Absolute abundances were calculated by multiplying relative abundance of ARGs by total abundance of 16S rRNA genes, quantified by qPCR. All metagenomes generated in this study are publicly available via MG-RAST (Meyer et al., 2008) under project number 12943. Reads were assembled de novo in MetaStorm according to default parameters and scaffolds were annotated as described above for reads. Co-occurrences of annotated genes on scaffolds were characterized via network analysis visualization using Gephi (version 0.8.2).

References

Anuj, S.N., Whiley, D.M., Kidd, T.J., Bell, S.C., Wainwright, C.E., Nissen, M.D., Sloots, T.P., 2009. Identification of Pseudomonas aeruginosa by a duplex real-time polymerase chain reaction assay targeting the ecfX and the gyrB genes. Diagn. Microbiol. Infect. Dis. 63, 127– 131. Arango-Argoty, G., Singh, G., Heath, L.S., Pruden, A., Xiao, W., Zhang, L., 2016. MetaStorm: A public resource for customizable metagenomics annotation. PLoS One 11, 1–13. Caporaso, J.G., Kuczynski, J., Stombaugh, J., Bittinger, K., Bushman, F.D., Costello, E.K., Fierer, N., Pena, A.G., Goodrich, J.K., Gordon, J.I., Huttley, G.A., Kelley, S.T., Knights, D., Koenig, J.E., Ley, R.E., Lozupone, C.A., McDonald, D., Muegge, B.D., Pirrung, M., Reeder, J.,

179

Sevinsky, J.R., Tumbaugh, P.J., Walters, W.A., Widmann, J., Yatsunenko, T., Zaneveld, J., Knight, R., 2010. QIIME allows analysis of high-throughput community sequencing data. Nat. Methods 7, 335–336. Colomer-Lluch, M., Jofre, J., Muniesa, M., 2014. Quinolone resistance genes (qnrA and qnrS) in bacteriophage particles from wastewater samples and the effect of inducing agents on packaged antibiotic resistance genes. J. Antimicrob. Chemother. 69, 1265–1274. DeSantis, T.Z., Hugenholtz, P., Larsen, N., Rojas, M., Brodie, E.L., Keller, K., Huber, T., Dalevi, D., Hu, P., Andersen, G.L., 2006. Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARB. Appl. Environ. Microbiol. 72, 5069–5072. Dutka-Malen, S., Evers, S., Courvalin, P., 1995. Detection of glycopeptide resistance genotypes and identification to the species level of clinically relevant enterococci by PCR. J. Clin. Microbiol. 33, 24–27. Hardwick, S.A., Stokes, H.W., Findlay, S., Taylor, M., Gillings, M.R., 2008. Quantification of class 1 integron abundance in natural environments using real-time quantitative PCR. FEMS Microbiol. Lett. 278, 207–212. Leplae, R., Hebrant, A., Wodak, S.J., Toussaint, A., 2004. ACLAME: a CLAssification of Mobile genetic Elements. Nucleic Acids Res. 32, D45-9. Li, B., Yang, Y., Ma, L., Ju, F., Guo, F., Tiedje, J.M., Zhang, T., 2015. Metagenomic and network analysis reveal wide distribution and co-occurrence of environmental antibiotic resistance genes. ISME J. 1–13. McArthur, A.G., Waglechner, N., Nizam, F., Yan, A., Azad, M. a., Baylay, A.J., Bhullar, K., Canova, M.J., De Pascale, G., Ejim, L., Kalan, L., King, A.M., Koteva, K., Morar, M., Mulvey, M.R., O’Brien, J.S., Pawlowski, A.C., Piddock, L.J. V, Spanogiannopoulos, P., Sutherland, A.D., Tang, I., Taylor, P.L., Thaker, M., Wang, W., Yan, M., Yu, T., Wright, G.D., 2013. The comprehensive antibiotic resistance database. Antimicrob. Agents Chemother. 57, 3348–3357. Meyer, F., Paarmann, D., D’Souza, M., Olson, R., Glass, E.M., Kubal, M., Paczian, T., Rodriguez, A., Stevens, R., Wilke, A., Wilkening, J., Edwards, R.A., 2008. The metagenomics RAST server - a public resource for the automatic phylogenetic and functional analysis of metagenomes. BMC Bioinformatics 9(386). National Sanitation Foundation, 2007. NSF International Standard/American National Standard 61: Drinking Water System Components-Health Effects. Ann Arbor, MI. Nazarian, E.J., Bopp, D.J., Saylors, A., Limberger, R.J., Musser, K. a, 2008. Design and implementation of a protocol for the detection of Legionella in clinical and environmental samples. Diagn. Microbiol. Infect. Dis. 62, 125–132. Pal, C., Bengtsson-Palme, J., Rensing, C., Kristiansson, E., Larsson, D.G.J., 2014. BacMet: antibacterial biocide and metal resistance genes database. Nucleic Acids Res. 42, D737– D743. Quast, C., Pruesse, E., Yilmaz, P., Gerken, J., Schweer, T., Yarza, P., Peplies, J., Glockner, F.O., 2013. The SILVA ribosomal RNA gene database project: improved data processing and web- based tools. Nucleic Acids Res. 41, D590–D596. Radomski, N., Lucas, F.S., Moilleron, R., Cambau, E., Haenn, S., Moulin, L., 2010. Development of a real-time qpcr method for detection and enumeration of Mycobacterium spp. in surface water. Appl. Environ. Microbiol. 76, 7348–7351.

180

Suzuki, M.T., Taylor, L.T., DeLong, E.F., 2000. Quantitative analysis of small-subunit rRNA genes in mixed microbial populations via 5’-nuclease assays. Appl. Environ. Microbiol. 66, 4605–4614.

Table S1: Primer and probe sequences and annealing temperatures used for qPCR Annealing Gene Primer/Probe Sequence (5’-3’) Reference temperature 16S rRNA F: CGGTGAATACGTTCYCGG (Suzuki et 55.0 (universal) R: GGWTACCTTGTTACGACTT al., 2000) F: CCCATGAAGCCCGTTGAA 23S rRNA R: ACAATCAGCCAATTAGTACGAGTTAGC (Nazarian 55.0 (Legionella spp.) Probe: HEX- et al., 2008) TCCACACCTCGCCTATCAACGTCGTAGT-BHQ 16S rRNA F: CCTGGGAAACTGGGTCTAAT (Radomski (Mycobacterium R: CGCACGCTCACAGTTA 55.0 et al., 2010) spp.) Probe: FAM-TTTCACGAACAACGCGACAAACT-BHQ ecfX F: CGCATGCCTATCAGGCGTT (Pseudomonas R: GAACTGCCCAGGTGCTTGC aeruginosa) Probe: HEX-ATGGCGAGTTGCTGCGCTTCCT-BHQ (Anuj et al., F: CCTGACCATCCGTCGCCACAAC 60.0 gyrB 2009) R: CGCAGCAGGATGCCGACGCC (Pseudomonas Probe: FAM-CCGTGGTGGTAGACCTGTTCCCAGACC- aeruginosa) BHQ F: AGGATTGCAGTTTCATTGAAAGC (Colomer- qnrA R: TGAACTCTATGCCAAAGCAGTTG 60.0 Lluch et al., Probe: FAM-TATGCCGATCTGCGCGA-BHQ 2014) F: GGGAAAACGACAATTGC (Dutka- vanA R: GTACAATGCGGCCGTTA 54.0 Malen et al., 1995) F: CTGGATTTCGATCACGGCACG (Hardwick intI1 66.0 R: ACATGCGTGTAAATCATCGTCG et al., 2008)

181

Figure S1: Nonmetric multidimensional scaling (NMDS) plot generated from Bray-Curtis similarity matrix of all metagenomic ARG abundances by utility and system type.

182

Table S2: Log-transformed colony-forming units per milliliter sample forming on R2A agar supplemented with antibiotics from samples collected at the simulated point of compliance (POC) and simulated point of use (POU) for each duplicate premise plumbing rig. Conditions tested include R2A amended with no antibiotics (NONE), ampicillin (AMP; 4 µg/mL), ciprofloxacin (CIP; 0.5 µg/mL), chloramphenicol (CHL; 4 µg/mL), gentamicin (GEN; 2 µg/mL), oxacillin (OXA; 1 µg/mL), rifampin (RIF; 0.5 µg/mL), sulfonamide (SUL; 128 µg/mL), tetracycline (TET; 2 µg/mL), and vancomycin (VAN; 0.5 µg/mL). Abundances below the limit of quantification (LOQ; 100 CFU/mL) are shown in gray while measurements above the LOQ are shown in black. ND=no detection. Utility Scenario Sample None AMP CIP CHL GEN OXA RIF SUL TET VAN POC ND ND ND ND ND ND ND ND ND ND 90% Surface/ POU-1 3.9 ND 3.1 3.4 ND ND ND 4.3 ND 4.3 10% AWP POU-2 6.1 6.1 6.0 5.5 ND 6.0 ND 5.9 6.0 5.9 1 POC ND ND ND ND ND ND ND ND ND ND 50% Surface/ POU-1 1.0 0.5 0.8 0.5 50% AWP ND ND ND ND ND ND POU-2 2.4 2.1 1.7 ND ND 1.9 ND 1.8 2.0 1.7 POC ND ND ND ND ND ND ND ND ND ND 100% POU-1 ND ND ND ND ND ND ND 0.5 0.5 ND Groundwater POU-2 ND ND ND ND ND ND ND ND ND ND 90% POC ND ND ND ND ND ND ND ND ND ND Groundwater/ POU-1 0.5 ND ND ND ND ND ND ND ND ND 10% AWP POU-2 0.5 ND ND ND ND ND ND ND ND ND 2 50% POC ND ND ND ND ND ND ND ND ND ND Groundwater/ POU-1 ND ND ND ND ND ND ND ND ND ND 50% AWP POU-2 1.5 ND ND ND ND ND ND ND ND ND 50% POC ND ND ND ND ND ND ND ND ND ND Groundwater/ POU-1 ND ND ND ND ND ND ND ND ND ND 50% AWP- Past POU-2 2.1 ND ND ND ND 1.9 ND 1.1 ND ND POC 0.5 ND ND ND ND 2.0 1.8 ND ND ND 100% POU-1 ND ND ND ND ND 2.0 1.8 ND ND ND Surface POU-2 0.8 ND ND ND ND 2.0 1.8 0.5 ND ND POC 0.5 ND ND ND ND 2.0 1.8 ND ND ND 95% Surface/ 3 POU-1 2.6 1.5 1.2 ND 2.2 2.5 2.1 2.0 ND 1.2 5% Tertiary POU-2 1.9 0.5 ND ND 1.9 1.8 1.6 1.7 ND 0.8 POC 1.1 0.5 ND ND ND 2.0 1.8 ND ND 0.5 90% Surface/ POU-1 4.3 3.9 ND ND 4.3 4.2 4.1 ND 2.5 3.6 10% Tertiary POU-2 4.4 3.2 3.2 2.2 4.3 4.4 4.2 4.1 1.9 4.0 POC ND ND ND ND ND ND ND ND ND ND 100% POU-1 3.0 2.8 ND ND 0.5 2.6 2.6 ND ND ND Groundwater POU-2 3.2 2.2 ND ND ND 3.1 3.2 ND ND ND POC ND ND ND ND ND ND ND ND ND ND 4 100% POU-1 ND ND ND ND ND ND 0.5 ND ND ND Surface POU-2 ND 0.5 ND ND ND ND ND ND ND ND POC ND ND ND ND ND ND ND ND ND ND POU-1 2.8 2.2 ND ND ND 2.5 2.6 ND ND ND 183

90% Groundwater/ 10% AWP POU-2 3.1 2.8 ND ND ND 3.0 3.0 ND ND ND POC ND ND ND ND ND ND ND ND ND ND 90% Surface/ POU-1 ND ND ND ND ND ND ND ND ND ND 10% AWP POU-2 ND ND ND ND ND ND ND ND ND ND 90% Surface/ POC ND ND ND ND ND ND ND ND ND ND 10% POU-1 1.6 1.0 ND ND ND 0.8 1.0 ND ND ND Industrial AWP POU-2 2.8 ND ND ND ND 2.0 1.8 ND ND ND

184

CHAPTER 8 : WHOLE GENOME SEQUENCE COMPARISON OF CLINICAL AND DRINKING WATER LEGIONELLA PNEUMOPHILA ISOLATES ASSOCIATED WITH THE FLINT WATER CRISIS

Emily Garner, Connor Brown, David Otto Schwake, William J. Rhoads, Gustavo Arango- Argoty, Liqing Zhang, Guillaume Jospin, David Coil, Jonathan A. Eisen, Marc A. Edwards, Amy Pruden

ABSTRACT

Background: Two outbreaks of Legionnaires’ Disease (LD) occurred in Genesee County, Michigan during 2014 and 2015. Previous work demonstrated that higher iron, depleted chlorine, and warmer temperatures characteristic of use of Flint River as the potable source water coincided with these outbreaks and were associated with elevated Legionella pneumophila genes in large buildings using Flint tap water.

Objectives: Here we compare whole genome sequences of clinical and water L. pneumophila isolates associated with the Flint LD outbreaks.

Methods: Whole genome sequences were obtained from 103 L. pneumophila isolates collected from Flint area tap water between March and August 2016 and compared to ten clinical isolates associated with the 2015 outbreak.

Results: A diverse range of L. pneumophila strains were documented over a cross-section of Flint tap water samples. Three clinical isolates and four potable water isolates collected from a Flint hospital and a Flint residence had a high degree of genomic similarity (average nucleotide identity=99.16–99.971%), all belonging to L. pneumophila sequence type (ST) 1 and serogroup 1. Serogroup 6 isolates belonging to the previously uncharacterized ST 2518 were widespread in samples collected throughout a Flint hospital in March 2016. Genes associated with Shigella spp., Stenotrophomonas maltophilia, and 22 other putative pathogens were found to be no more relatively abundant in Flint tap water samples than in other U.S. potable water systems.

Conclusions: Though few clinical isolates are available from the LD outbreaks, the high degree of similarity demonstrated between select water and clinical isolates indicates that the Flint potable water system was a probable source of some L. pneumophila infections.

185

INTRODUCTION

In January 2016, the Michigan Department of Health and Human Services (MDHHS) and the Genesee County Health Department (GCHD) publicly announced that two outbreaks of Legionnaires’ Disease (LD) had occurred in Genesee County, MI (MDHHS 2016a, 2016b). LD is a severe form of pneumonia caused by inhalation of certain virulent species of aerosolized bacteria belonging to the genus Legionella. The first outbreak occurred from June 2014 to March 2015 (n=45) and the second from May to October 2015 (n=47), with a known total of 92 cases and 12 deaths (MDHHS 2016a, 2016b). In Flint, MI, which is located in Genesee County, the corrosive Flint River was in use as a new drinking water source from April 2014 to October 2015, a period spanning that of the LD outbreaks, without implementation of federally-mandated corrosion control. This resulted in systemic degradation of water quality, including elevated lead in the tap water over a prolonged period now known as the “Flint Water Crisis” (Pieper et al. 2017). This disruption in water quality also likely stimulated the growth of Legionella pneumophila, the species that is most frequently the causative agent of LD and is responsible for over 90% of reported outbreaks (Marston et al. 1994), in Flint’s distribution and plumbing systems (Rhoads et al. 2017).

Our prior work highlighted coincidence of the LD outbreaks with elevated iron (a natural consequence of corrosion of iron water mains), reduced levels of free chlorine disinfectant residuals, and elevated water temperatures, all factors known to stimulate growth of Legionella (Rhoads et al. 2017). Zahran et al. (2018) similarly reported that the odds of Flint residents being diagnosed with LD during use of the Flint River increased 6.3 fold and noted associations with low chlorine residuals. Further, 23S rRNA genes of Legionella spp. and macrophage infectivity potentiator (mip) genes of L. pneumophila were found to be elevated towards the end of the second outbreak in the tap water of large buildings in Flint, relative to levels reported for other U.S. water systems not experiencing outbreak (Schwake et al. 2016). On the other hand, mip levels were largely below detection in Flint single-family residences during the water crisis (Schwake et al. 2016). Large buildings, such as hospitals, are generally thought to be more susceptible to Legionella regrowth relative to much simpler plumbing systems characteristic of single family homes (Sabria and Yu 2002), and Legionella control often focuses on appropriate management of risks in large buildings (ANSI/ASHRAE 2015).

In addition to the problems noted above occurring during the Flint Water Crisis, rampant corrosion and unusually cold temperatures also compromised the integrity of drinking water mains. The incidence of water main breaks was between 1.34-2.21 times higher during the crisis than during 2010-2013 (Rhoads et al. 2017), creating the potential for increased contamination of potable water by fecal or opportunistic pathogenic bacteria due to detachment of scale or intrusion (Garrison et al. 2016). During this time, the water exceeded standards for fecal coliform bacteria and E. coli, necessitating declaration of multiple boil water advisories (Fonger 2014a, 2014b). Other health concerns drew attention, including increased incidence of rashes among Flint residents throughout the duration of Flint River use (Unified Coordination Group 2016) and an outbreak of in Genesee and Saginaw counties from March to December 2016 (CDC and MDHHS 2016; Unified Coordination Group 2016), but no evidence has yet emerged to link these concerns to transmission of waterborne pathogens (CDC and MDHHS 2016; Unified Coordination Group 2016). Additionally, reportedly in response to concern circulating in the Flint

186

community via social media, the GCHD released a statement informing residents about health conditions associated with infection by the multidrug-resistant pathogen, Stenotrophomonas maltophilia (GCHD 2017; Young 2017).

Here we characterized L. pneumophila isolated from Flint tap water shortly after the the 2015 LD outbreaks subsided, and over the subsequent year as the water quality improved and no LD outbreak was observed, using next-generation sequencing. Gene markers corresponding to 24 fecal and opportunistic pathogens were identified across a cross section of samples representing hospitals and homes before and after switching back to Detroit water, and compared to data available for other U.S. cities using shotgun metagenomic sequencing. Legionella isolates were obtained from the tap water of a hospital and single family residences in Genesee County over several months after switching back to Detroit water and subject to whole genome sequencing. One hundred and three drinking water isolates were compared to ten clinical isolates collected during the second outbreak in terms of sequence type (ST), average nucleotide identity (ANI), and single nucleotide polymorphism (SNP) analysis.

MATERIALS AND METHODS

Study Site Description

Bulk water and biofilm samples were collected during five sampling campaigns: two while the city was using the Flint River as the drinking water source (August 18-19, 2015 and October 15-16, 2015) and three approximately five (March 7-9, 2016), eight (June 21-27, 2016), and ten months (August 2016) after the city resumed purchasing water with corrosion control from the original Detroit Water and Sewer Department (DWSD) supplier. The August 2015 and August 2016 collections targeted samples from hot and cold water taps in single-story homes and businesses to characterize water quality at the point of use, where corrosion impacts and regrowth of bacteria were anticipated to be most problematic. The October 2015 sampling targeted hot and cold water taps from the two largest hospitals in Flint, “Hospital #1” and “Hospital #2” as designated in Schwake et al. (2016), where water quality is vulnerable to extensive plumbing systems and there is potential for immunocompromised populations to be exposed. The March 2016 sampling consisted of repeat sampling from the homes, businesses, and Hospital #1 sampled previously (Hospital #2 denied access the second time), while the June 2016 sampling focused on sampling in homes (both hospitals denied access the third time). Control samples were also collected from nearby Flint Township, which received DWSD water consistently throughout the duration of the study, and from a nearby school using well water.

Sample Collection and Preservation

One or two liter samples were collected from all taps into sterile polypropylene bottles (Nalgene, Rochester, NY) with 24 mg of sodium thiosulfate per liter added as a chlorine quenching agent. Cold water samples were taken by collecting the first flush from the tap. Hot water samples were collected after flushing for 30 seconds. Select biofilm samples were collected, after all bulk water samples were collected from that tap, by removing the tap aerator, inserting a sterile swab (Fisher, Hampton, NH) into the faucet, swabbing one full pass around the circumference of the inner surface, and transferring to a sterile Lysing Matrix A tube (MP Biomedicals, Solon, OH). In

187

addition, in June 2016, homes were extensively sampled as part of a water heater cleaning campaign, with the following samples collected before and after the cleaning protocol: hot and cold stagnant samples from the kitchen tap, a stagnant shower or bathtub sample of blended hot and cold taps, hot water flushed until constant temperature was reached from the kitchen tap, the hot water heater drain valve, and a cold water sample collected after flushing for five minutes from the outside hose bib or nearest tap to the service entry point to collect water before it is exposed to home plumbing. Between 250 and 500 milliliters were aliquoted into sterile containers for subsequent culture analysis. All samples were transported to the lab and processed within approximately 30 hours of sampling. Samples for culture were transported at room temperature while samples for molecular analysis were transported on ice.

Aliquots for culture were filter-concentrated onto a sterile 0.22 μm pore size mixed- cellulose ester membrane (Millipore, Billerica, MA) and resuspended in 5 mL sterile tap water prior to culturing Legionella according to standard methods (CDC 2005). For molecular analyses, the remaining volume was filter-concentrated in the same manner onto a second filter, which was subsequently fragmented using sterile forceps and stored at -20 °C until DNA could be extracted using a FastDNA SPIN Kit (MP Biomedicals, Solon, OH) according to manufacturer instructions. Biofilm samples were extracted in the same manner after transferring swabs directly to extraction tubes. Quantities of Legionella spp. and L. pneumophila gene markers from these samples have been published previously (Rhoads et al. 2017; Schwake et al. 2016). DNA was extracted from Legionella cultures by resuspending colonies in 50 μl of molecular grade water, freezing at -20 °C, and rapidly thawing at 90°C for 10 minutes.

Whole genome sequencing of L. pneumophila isolates

Whole genome sequencing was conducted on DNA extracts from 103 water L. pneumophila isolates and ten clinical isolates originating from patients in Genesee County diagnosed with LD in 2015 (Table S1). Clinical isolates were provided by MDHHS without identifying patient information. Two positive control strains of known identity and two negative controls of non-Legionella isolates (selected from plates prepared according to the L. pneumophila standard method but failing to be confirmed as L. pnuemophila due to irregular morphology) were also sequenced. DNA extracts were quantified via a Qubit 2.0 Fluorometer (Thermo Fisher, Waltham, MA) and analyzed via gel electrophoresis to verify DNA integrity. Sequencing was conducted by MicrobesNG (Birmingham, United Kingdom) on a MiSeq platform (Illumina, San Diego, CA) with 2 x 250 bp paired-end reads. Libraries were constructed using a modified Nextera DNA library preparation kit (Illumina, San Diego, CA). Reads were trimmed using Trimmomatic (Bolger et al. 2014) and de novo assemblies were generated using SPAdes (Bankevich et al. 2012).

Whole genome sequence analysis

16S rRNA gene sequences were extracted from sequence data using the Rapid Annotations Using Subsystem Technology server (Aziz et al. 2008) and Legionella species assignments were determined via BLASTn of the sequence against the NCBI nucleotide database via the web server. Phylogenetic trees were generated using FastTree (Price et al. 2010) based on extracted 16S rRNA gene sequences, and 37 single-copy housekeeping genes in nucleotide space and amino acid space using PhyloSift (Darling et al. 2014). ANI was calculated as previously described (Goris et al.

188

2007) and SNPs were identified using kSNP3.0 (Gardner et al. 2015) with maximum likelihood estimation. Nine known L. pneumophila genomes associated with previous LD outbreaks were included in the analysis as reference strains for comparison (Table S2).

Serogroup analysis

L. pneumophila isolate genomes belonging to serogroup 1 were identified via detection of the wzm gene (Mérault et al. 2011) in whole genomes using BLAST with a minimum nucleotide identity of 98% and e-value of 1e-5. DNA sequence-based classifications were verified and unknown serogroups were determined using direct fluorescent antibody (DFA) staining with FITC-conjugated antibodies (m-TECH, Milton, GA). To address problems with non-specific binding when stained cells were prepared according to manufacturer instructions, the protocol was modified as follows: isolates grown in buffered yeast extract broth (per liter: 10 g yeast extract, 1 g alpha ketoglutaric acid, 10 g 2-(carbamoylmethylamino)ethanesulfonic acid, 0.4 g L-cystine monohydrochloride, 0.25 g ferric pyrophosphate) were centrifuged at 5,000 x g and resuspended in 1X phosphate buffered saline (PBS). To separate 25 µl aliquots of cells suspended in PBS, 5 µl of each FITC-conjugated antibodies were added and the suspension was incubated at 20oC for 30 minutes. Cells were washed with 1X PBS three times, then viewed with an AxioSkop2 plus fluorescence microscope (Carl Zeiss Microscopy, Oberkocken, Germany).

Shotgun metagenomic sequencing

Shotgun metagenomic sequencing was performed on DNA extracts from hot (n=3) and cold (n=4) water samples collected in August and October 2015, while the Flint River water source was online, and hot (n=4) and cold (n=3) water samples collected in March 2016, after the municipality resumed purchasing water from DWSD. Additionally, control samples (n=8) were collected, including one from Flint Township, which consistently received DWSD water, one from a school that had consistently operated using well water, and two from taps in three additional U.S. municipalities located in Virginia, Florida, and Arizona (Garner et al., in review; Ji et al., in preparation). Sequencing was also performed on a biofilm sample (n=1) collected from a Flint residential tap of particular interest, due to positive detection of Legionella based on quantitative polymerase chain reaction (qPCR), and a sample collected from the Flint River (n=1). Samples were multiplexed for shotgun metagenomic sequencing using the barcodes presented in Table S3. Sequencing was conducted at the Biocomplexity Institute of Virginia Genomics Sequencing Center (Blacksburg, VA) on an Illumina HiSeq 2500 platform using a 100-cycle paired-end protocol and Accel-NGS 2S Plus DNA library preparation (Swift Biosciences, Ann Arbor, MI).

Shotgun metagenomic data was uploaded to MG-RAST (Meyer et al. 2008), where merging of paired-end reads and quality filtering was performed according to default parameters. Reads were annotated against the RefSeq database for taxonomic classification using the best hit annotation approach with a stringent minimum amino acid identity cutoff of 90%, minimum alignment length of 15 amino acids, and 1e-5 e-value cutoff to minimize the potential for inaccurate annotations. Samples containing high relative abundances of reads annotated as Shigella spp. and S. maltophilia were further analyzed via endpoint PCR for presence of the ipaH and 23S rRNA genes specific to each species, respectively, using previously described protocols (Hsu et al. 2010; Whitby et al. 2000). Metagenomes are publicly available on the MG-RAST server

189

under project mgp21599. Statistically significant differences between abundances of metagenomic annotations (normalized to number of reads) were tested using a nonparametric Kruskal Wallis rank sum test with a post-hoc pairwise Wilcoxon test performed in R (v3.3.2).

RESULTS

Legionella Isolate characterization

Across the 192 total samples collected across this study from which isolation was attempted, 22 hospital samples were positive for culturable L. pneumophila in March 2016, 3 residence samples were positive in June 2016, and 3 residence samples were positive in August 2016. No isolates were obtained from businesses receiving DWSD water, but 3 cold and 3 hot taps at the school serviced by well water were positive. Isolates were named according to the following system: First letter indicates building type/location (H=hospital; R=residence; W=school using well water; P=large public building), second letter indicates sample collection location (H=hot tap; C=cold tap; D=water heater drain valve; S=shower), followed by a unique numeric identifier. Clinical strains are denoted C1-10.

According to phylogenetic analysis of 16S rRNA genes mined from whole genome sequences, all clinical and water isolates, except for eight of the nine well water isolates, were identified as L. pneumophila. The positive control strain was correctly identified as L. pneumophila, with SNP analysis further classifying it according to its known provenance (130b), while the negative control strain was also confirmed to be non-Legionella (Stenotrophomonas maltophilia). Serotyping via presence of the wzm gene for serogroup 1 and DFA staining for other serogroups indicated that all L. pneumophila isolates belonged to serogroups 1 and 6 (Table 8-1).

L. pneumophila isolates obtained from clinical and water samples were found to belong to several STs (Table 8-1). Of serogroup 1 isolates, all belonged to STs 1, 44, 159, 192, 211, 213, 222 or to a previously uncharacterized ST that we submitted to the EWGLI database and has now been designated as ST 2513. Serogroup 6 isolates all belonged to a previously uncharacterized ST that we submitted to the EWGLI database and has now been designated as ST 2518. The vast majority of hospital isolates belonged to ST 2518, while isolates originating from residential tap water belonged primarily to ST 192. Only ST 1 was represented by both clinical and water isolates, specifically, three clinical isolates, three isolates from hospital tap water, and one isolate from residential tap water.

The seven isolates classified as ST 1 were further found to share a high degree of genomic similarity with each other, with ANI values ranging from 99.164 to 99.971%. In particular, clinical isolate C3 shared the highest degree of similarity with Flint tap water isolates, with ANI values ranging from 99.601 to 99.846%. Clinical isolate C3 was found to have the highest ANI similarity to HH56. Clinical isolate C2 displayed the next highest degree of similarity to the water isolates, with ANI values ranging from 99.175 to 99.218%. The only other clinical isolate exhibiting ANI values >98% when compared to a water isolate was clinical isolate C8 when compared with several of the ST2518 isolates recovered from hot and cold hospital lines, from the cold water line of a large public building, and the hot water of a school using well water (HC01-14, HH01-16, HH18- 24, HH26-27, HH29-55, PC01, WH03).

190

Table 8-1: Summary of isolates by sequence type (ST), serogroup (SG), and sample origin ST SG Isolate origin 1 1 3 hospital water (HH17, HH25, HH56), 1 residence water (RH08), 3 clinical (C2, C3, C7) 44 1 1 clinical (C6) 159 1 1 clinical (C1) 192 1 19 residence water (RC01, RC02, RC03, RC04, RC06, RC07, RD01, RD02, RD03, RD04, RD05, RH02, RH03, RH04, RH05, RH07 RH07, RS01, RS02) 211 1 1 clinical (C8) 213 1 2 clinical (C4, C5) 222 1 1 clinical (C9) 2513a 1 1 clinical (C10) 2518a 6 66 hospital water (HC01, HC02, HC03, HC04, HC05, HC06, HC07, HC08, HC09, HC10, HC11, HC12, HC13, HC14, HH01, HH02, HH03, HH04, HH05, HH06, HH07, HH08, HH09, HH10, HH11, HH12, HH13, HH14, HH15, HH16, HH18, HH19, HH20, HH21, HH22, HH23, HH24, HH26, HH27, HH29, HH30, HH31, HH32, HH33, HH34, HH35, HH36, HH37, HH38, HH39, HH40, HH41, HH42, HH43, HH44, HH45, HH46, HH47, HH48, HH49, HH50, HH51, HH52, HH53, HH54, HH55), 1 public building (PC01), 1 well water (WH03) ND ND HH28, RC05, RH01, RS03, WC01, WC02, WC03, WC04, WH01, WH02, WH04, WH05 aNew STs entered in EWGLI database; ND = could not be determined

When classified based on SNP similarity, calculated via maximum likelihood methods, isolates formed distinct clades that were generally consistent with the ST classification (Figure 8- 1). The ST 1 clade varied by 2-1062 SNPs, with isolates varying from the reference Paris strain by only 371-505 SNPs. Water isolates in this clade varied from clinical isolates by as few as 38 SNPs. Several other distinct clades emerged in which water isolates were grouped primarily by building type. A large clade of primarily ST 2518 isolates included the majority of the hospital samples, one well water isolate, and one large public building isolate. Another clade contained only isolates originating from Flint residence water samples belonging to ST 192. The SNP analysis results were generally consistent with the phylogenetic results (Figure S1, Figure S2, Figure S3), confirming the grouping of ST 1 isolates into one clade, with hospital water isolates and residential water isolates generally forming two separate and larger clades.

The ST of eight isolates derived from well water could not be determined due to the absence of L. pneumophila-specific alleles. We hypothesize that these isolates were mistakenly phenotypically characterized as L. pneumophila based on colony morphology and actually belong to a different species of Legionella. ANI values comparing these isolates with the positive control L. pneumophila strain (130b) ranged from 62.645 to 62.969%. This indicates that these isolates are not L. pneumophila, given that genomes belonging to a single species generally share ANI values >95% (Rodriguez-R and Konstantinidis 2014). The eight well water strains that were not L. pneumophila appear to be most closely related to L. taurinensis, L. rubrilucens, or L. erythra, as the 16S rRNA genes extracted from these genomes shared greater than 99% nucleotide similarity to all three species.

191

Figure 8-1: Single nucleotide polymorphism (SNP) analysis of Legionella pneumophila isolates. Analysis conducted in kSNP3.0 and visualized using FigTree 14.3. Isolates originated from clinical (yellow), hospital water (blue), residence water (red), public building water (purple), or buildings supplied by well water (green). With the exception of buildings supplied by well water, all buildings were serviced by Flint municipal water. Reference strains are detailed in Table S2.

Annotation of Shotgun Metagenomic Sequences for Identification of Other Putative Pathogens

The potential for presence of other waterborne fecal and opportunistic pathogens was screened using shotgun metagenomic sequencing of water samples and annotation of reads corresponding to 24 select pathogens using MG-RAST. A cross section of DNA extracts obtained from Flint tap water during use of Flint River water (August 2015, October 2015) and 5 months after return to DWSD water (March 2016) were analyzed. These included residences, businesses, and hospitals in Flint, with comparison to raw Flint River water, a Flint Township business that remained consistently on DWSD water, a school serviced by well water, and tap water from three U.S. municipalities (AZ, FL, and VA) (Figure 8-2).

No genes annotated as potentially belonging to pathogens were found to be higher in relative abundance in Flint tap water relative to water analyzed from four other municipal systems and one well systems (p≥0.1032, Wilcoxon). Only Cryptosporidium spp. genes were more abundant in non-Flint samples than Flint samples (p=0.0326). When broken down by date,

192

Legionella spp. and L. pneumophila genes were found at a greater relative abundance in October 2015 samples than in the samples collected from other municipalities and the well system (p=0.0225), while Shigella spp. genes were more abundant in March 2016 Flint samples than in the other municipal and well samples (p=0.0184).

Among Flint samples, several genes annotated as highly similar to those pertaining to the 24 selected pathogen genomes were more abundant in small buildings (residences and businesses) than large buildings (hospitals). These included: Aeromonas hydrophilia (p=0.0043), (p=0.0157), (p=0.0481), Mycobacterium spp. (p=0.0338), and Mycobacterium avium (p=0.0338). There were no significant differences between putative pathogen gene abundances in hot versus cold Flint samples (p≥0.0741). Metagenomic reads obtained were annotated as three different species of Legionella in Flint samples, with L. pneumophila generally being the most abundant, followed by L. longbeachae and L. drancourtii.

Samples with the greatest relative abundance of Shigella spp. (Business 3 - cold, Hospital #1 - tap iii - hot, and Hospital #1 - tap iii - cold) were further analyzed for the presence of the Shigella-specific ipaH gene via PCR to confirm the metagenomic results; however, the gene was not detected in any of the three samples (Figure 8-2B). Similarly, samples with the greatest abundance of genes annotated as S. maltophilia were screened for the presence of the 23S rRNA gene specific to this species, but the gene was not detected (Figure 8-2B). In contrast, there was a strong correlation between the relative abundance of Legionella spp. genes determined by shotgun metagenomics and the absolute abundance determined previously by qPCR (Schwake et al. 2016) (Spearman’s rank sum correlation test; ρ=0.6687, p=0.0345).

DISCUSSION

Few clinical sputum isolates appear to have been collected or preserved from the outbreaks in Flint, with only urine-antigen testing having been conducted in the majority of cases. This is unfortunate given that the LD outbreak in Genesee County is among the largest in U.S. history when considered per capita and how vitally important clinical isolates are for learning from past outbreaks and preventing future outbreaks. Although 31 of 92 LD patients’ home residences were serviced by Flint water (MDHHS 2016a, 2016b), no sputum isolates were preserved from patients residing in homes serviced by Flint water (personal communication, MDHHS). Additionally, given that the LD outbreaks were not announced until January 2016, three months following the conclusion of the second outbreak, few environmental specimens were collected during the period that the outbreaks actually occurred. Thus, a more definitive study of environmental sources of the outbreaks is not possible. The present study is the only to our knowledge that includes shotgun metagenomic DNA sequence analysis of Flint tap water collected during the actual Flint Water Crisis, along with comprehensive analysis of Legionella isolates collected throughout the distribution system within the following 6 months to 1 year as the system began to recover.

It has previously been demonstrated that a single strain of L. pneumophila can colonize buildings and persist over multiple years (Perola et al. 2005; Rangel-Frausto et al. 1999; Scaturro et al. 2007). Thus, it is reasonable to assume that water isolates collected in 2016 were likely representative of any strains colonizing water systems over the previous months or even years.

193

Figure 8-2: Comparison of shotgun metagenomic DNA sequence reads obtained from a cross section of Flint tap water samples. Samples collected during use of Flint River water (August and October 2015) and after switch back to DWSD on October 16, 2015 (March 2016). Raw Flint River source water, tap water from a residence in Flint Township that continually received DWSD water (DWSD Residence), well water within Genesee County, and municipal tap water in three other U.S. states (Virginia, Florida, and Arizona). Heatmap is based on annotation of reads conducted in MG-RAST with a best hit approach, 90% identity cutoff, and 1e-5 e-value cutoff.

194

Interestingly, several strains of L. pneumophila were found to inhabit Flint tap water collected in homes, hospitals, and businesses. Of the 10 clinical isolates provided, these also belonged to several strains. However, only three of the clinical strains (C2, C3, C7) displayed high similarity to any water isolates, all belonging to a single clade, as defined by SNP analysis, containing both hospital water and residential water isolates. The remaining water isolates were generally subdivided into two SNP-defined clades, one primarily dominated by residential tap water samples, and the other by hospital tap water samples (Figure 8-1).

Based on this study, there is reasonable evidence that Flint tap water was a likely source of L. pneumophila for LD patients infected with ST 1. High degrees of similarity (2-1,062 SNPs) between three clinical isolates and four water isolates belonging to ST 1 were apparent and consistent with phylogenetic and ANI analysis. Previous studies have documented that while some outbreaks are characterized by clinical strains that differ by as few as <5 SNPs, other outbreaks may differ by as many as 418 core SNPs (Raphael 2016). Thus, the SNP variability between water and clinical strains of ST 1 in this study is comparable to the range of variation documented within other outbreaks. Given that the ST 1 water isolates were collected from both hospital and residential taps, this strain appears to be somewhat widespread in the water distribution system, spanning multiple Flint buildings. However, the presence of several distinct phylogenetic clades of L. pneumophila isolated from Flint water systems further demonstrates that a single strain of L. pneumophila did not dominate the system citywide. We hypothesize that this is likely due to the presence of conditions favorable to Legionella growth, which we previously documented in the Flint system (Rhoads et al. 2017), facilitating the proliferation of multiple strains of L. pneumophila in different buildings and parts of the system. Similarly, the broad distribution of clinical isolates across eight STs supports the hypothesis that any waterborne exposures that resulted in LD could hypothetically have originated from a diverse array of L. pneumophila strains. The variety of STs associated with clinical isolates also suggests that the clinical cases profiled here originated from a variety of different exposure sources. In addition, the markedly elevated relative abundance of L. pneumophila gene markers annotated from the October 2015 metagenomic dataset is also consistent with previous surveys indicating elevated levels of Legionella spp. and L. pneumophila specific gene markers in the Flint system during this period based on qPCR (Schwake et al. 2016). Furthermore, the relatively low abundances of L. pneumophila genes annotated from the metagenomic dataset (Figure 8-2A) in the Flint River sample demonstrates that the potable source water was likely not the primary source of L. pneumophila genes, but rather regrowth in the distribution system is likely the cause of L. pneumophila genes in Flint tap water.

All clinical isolates characterized in this study belonged to L. pneumophila serogroup 1, which is the cause of 57% of reported LD cases in the U.S., though this number is likely underreported given that over 30% of cases are attributed to undetermined L. pneumophila serogroups (Marston et al. 1994). The data are also largely skewed because the widely applied urine antigen test for LD only confirms L. pneumophila serogroup 1. L. pneumophila of ST 1 has been widely implicated in Legionnaires’ Disease outbreaks worldwide, including outbreaks in France (Ginevra et al. 2012), China (Qin et al. 2016), Germany (Borchardt et al. 2007), Canada (Reimer et al. 2010), and the U.S. (Kozak-Muiznieks et al. 2014). In the U.S., ST 1 is thought to be both the most common cause of sporadic cases of LD, as well as the most common waterborne ST found in potable and non-potable water alike (Kozak-Muiznieks et al. 2014).

195

Water isolates belonging to serogroup 6, all classified as ST 2518, were found to be widespread in samples collected from Hospital #1 in March 2016. A study of L. pneumophila isolates collected from Flint tap water in September and October of 2016 also found that serogroup 6 isolates were widespread in residential premise plumbing water samples, though these isolates were all found to belong to STs 367 and 461 (Byrne et al. 2018). Serogroup 6 strains identified by Byrne et al. (2018) were found to be at least as infectious of macrophages as a known virulent laboratory strain, emphasizing the potential for LD to be caused by non-serogroup 1 strains. Improved diagnostic tools are critically needed to address the potential for LD caused by non- serogroup 1 strains, and more research is needed to confirm the relevance of serogroup 6 strains for human infectivity.

It is also interesting to note that the vast majority of L. pneumophila isolates obtained from taps serviced by Flint water in this study originated from hot water taps, 38% of which were positive for culturable L. pneumophila, compared to only 16% of cold tap samples. While L. pneumophila typically multiplies at temperatures between 25°C and 37°C (Wadowsky et al. 1985) and prospers in hot water plumbing systems (Rhoads et al. 2015), it has also been widely documented in cold water taps, with one study finding as many as 47% of surveyed taps positive for genes specific to L. pneumophila serogroup 1 (Donohue et al. 2014).

In addition to the documented LD outbreaks and elevated levels of lead, several other health concerns emerged in Flint, including known contamination of the potable water system with E. coli and coliform bacteria, an outbreak of Shigellosis, and widespread occurrence of rashes (CDC and MDHHS 2016; GCHD 2017; Fonger 2014a, 2014b; Unified Coordination Group 2016; Young 2017). In particular, the cause of the rashes and source of the Shigella have never been confirmed. Given that waterborne bacterial agents are capable of causing such afflictions, metagenomic sequencing was applied to select water samples collected from August 2015 to March 2016 to screen for the occurrence of DNA sequences corresponding to suspect agents. It is important to note that shotgun metagenomics is an emerging methodology, with no standard protocols yet available, and cannot directly confirm the presence or viability of an actual pathogen. In particular, detection of some level of background DNA lingering in extracellular form or within cells killed as a result of water treatment, especially for fecal contaminants that are non-native to drinking water and do not readily flourish in the drinking water environment, is likely and not necessarily representative of live organisms. Also important to note is that shotgun metagenomic sequencing provides relative abundances of target genes (i.e., normalized to total reads) and thus does not measure the total number of putative pathogens. In other words, a sample might be highly enriched in a DNA sequence corresponding to a putative pathogen, but the total biomass may actually be very low, suggesting abundance of the pathogen would also be correspondingly low. Given that few water samples were collected or preserved during the time of the actual Flint Water Crisis, shotgun metagenomics was applied as an exploratory screening tool to gain further insight into the microbial water quality of Flint water during this period.

Shotgun metagenomic sequence data from Flint water were compared data from samples collected from buildings located near Flint, but serviced by DWSD or well water, as well as samples from three different U.S. cities’ potable water systems collected at the point of use. This approach provided various local and national non-Flint samples as points of relative comparison

196

and provided context for understanding what was found in the Flint samples. No clear differences between relative abundance of reads annotated as pathogens in Flint versus non-Flint samples emerged, when comparing the two datasets as a whole. However, the March 2016 levels of Shigella spp. gene markers were relatively more abundant compared to other cities. The Shigellosis outbreak in Genesee County began in March 2016, which is consistent with the possibility of contamination of water with the bacterium during this timeframe (CDC and MDHHS 2016). However, endpoint PCR analysis indicated that Shigella-specific genes were undetectable in the samples from which the species was annotated via metagenomics. Similarly, although S. maltophilia was detected at high relative abundances in a portion of the metagenomic analyses, when these same samples were analyzed via endpoint PCR, the species was not detected. Given that PCR-based methods performed with well-validated assays would be expected to be far more sensitive and specific than shotgun metagenomic sequencing and annotation, together, these results suggest that shotgun metagenomic analysis likely resulted in false-positive annotations for some pathogens. Because shotgun metagenomics captures a random subset of full genomes of a microbial community, incorrect annotations are hypothesized to likely result from the detection of genes belonging to a lineage that shares a common ancestor with the putative pathogens targeted in this study. In contrast, the relative abundance of Legionella spp. genes determined via shotgun metagenomics was well-correlated with abundances of the genus determined by qPCR. Thus, use of such next-generation sequence technologies to broadly profile the microbial community and identify putative pathogens of interest is does provide some value for screening the composition of samples of interest. However, use of these methods needs to be refined and should be employed with extreme caution and may be inappropriate for certain pathogens that are particularly difficult to annotate with certainty.

This study characterizes clinical and water L. pneumophila isolates from Flint, Michigan and the surrounding area and establishes a high degree of similarity between four water isolates originating from Flint tap water and three clinical strains. This study also establishes that a variety of L. pneumophila strains were culturable from the Flint system, demonstrating that multiple strains were potentially transmitted via the drinking water. Still, the remaining seven clinical strains analyzed in this study showed low similarity to the water isolates and remain of unknown provenance. Given the widespread distribution of various strains of L. pneumophila throughout the Flint water distribution system, it is also possible that the serogroup 1 ST 1 strain implicated here as having the greatest degree of similarity between water and clinical isolates, could also occur in other putative sources of transmission, such as cooling towers.

ACKNOWLEDGEMENTS

This study was partially supported by U.S. National Science Foundation RAPID Award (1556258), Graduate Research Fellowship Program Grant (DGE 0822220) and supplementary funding associated with grant 1336650. Additional support was provided by the Alfred P. Sloan Foundation Microbiology of the Built Environment program, the State of Michigan for a study of effects of flushing residential hot water heaters (summer 2016), the American Water Works Association Abel Wolman Fellowship, and the Institute for Critical Technology and Applied Science at Virginia Tech. We also thank the members of the Flint Water Study Team at Virginia Tech, who volunteered their time to collect samples, and the Flint citizens and businesses that supported this study, and we thank Joan Rose for allowing us to utilize her laboratory.

197

REFERENCES

American National Standards Institute / American Society of Heating, Refrigerating and Air- Conditioning Engineers (ANSI/ASHRAE). 2015. Standard 188 Legionellosis: Risk Management for Building Water Systems. Aziz RK, Bartels D, Best AA, DeJongh M, Disz T, Edwards RA, et al. 2008. The RAST Server: Rapid Annotations using Subsystems Technology. BMC Genomics 9:75. Bankevich A, Nurk S, Antipov D, Gurevich AA, Dvorkin M, Kulikov AS, et al. 2012. SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing. J. Comput. Biol. 19:455–477. Bolger AM, Lohse M, Usadel B. 2014. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30:2114–2120. Borchardt J, Helbig JH, Lück PC. 2007. Occurrence and distribution of sequence types among Legionella pneumophila strains isolated from patients in Germany: common features and differences to other regions of the world. Eur. J. Clin. Microbiol. Infect. Dis. 27:29–36. Byrne, BG, McColm, S, McElmurry, SP, Kilgore, PE, Sobeck, J, Sadler, R, Love, NG, Swanson, MS. 2018. Prevalence of Infection-Competent Serogroup 6 Legionella pneumophila within Premise Plumbing in Southeast Michigan. 9:e00016-18. Centers for Disease Control and Prevention (CDC). 2005. Procedures for the Recovery of Legionella from the Environment. Centers for Disease Control and Prevention (CDC), Michigan Department of Health & Human Services (MDHHS). 2016. Interim Findings from the Shigellosis Outbreak Investigation — Genesee and Saginaw Counties, Michigan, 2016. Available: http://www.michigan.gov/documents/mdhhs/MI_Shigellosis_Dec_Update_CommunityPr esentation_545873_7.pdf [accessed 6 December 2017]. Genesee County Health Department (GCHD). 2017. Media Release RE: Stenotrophomonas maltophilia. Available: https://gchd.us/wp-content/uploads/2017/04/Stenotrophomonas-4- 19-17-1.pdf. [accessed 6 December 2017]. Darling AE, Jospin G, Lowe E, Matsen FA IV, Bik HM, Eisen JA. 2014. PhyloSift: phylogenetic analysis of genomes and metagenomes. PeerJ 2:e243. Donohue MJ, O’Connell K, Vesper SJ, Mistry JH, King D, Kostich M, et al. 2014. Widespread molecular detection of Legionella pneumophila Serogroup 1 in cold water taps across the United States. Environ. Sci. Technol. 48: 3145–3152. Fonger R. 2014a. Flint expands boil water advisory area after more positive tests for total coliform bacteria. MLive.com. Available: http://www.mlive.com/news/flint/index.ssf/2014/09/flint_expands_boil_water.html [accessed 6 December 2017]. Fonger R. 2014b. Flint issues boil water advisory for section of the city after positive test for total coliform bacteria. MLive.com. Available: http://www.mlive.com/news/flint/index.ssf/2014/09/flint_issues_boil_water_adviso.html [accessed 6 December 2017]. Gaia V, Fry NK, Afshar B, Luck PC, Meugnier H, Etienne J, et al. 2005. Consensus Sequence- Based Scheme for Epidemiological Typing of Clinical and Environmental Isolates of Legionella pneumophila. J. Clin. Microbiol. 43:2047–2052.

198

Gardner SN, Slezak T, Hall BG. 2015. kSNP3.0: SNP detection and phylogenetic analysis of genomes without genome alignment or reference genome. Bioinformatics 31:2877–2878. Garner E, Chen C, Xia K, Bowers J, Engelthaler DM, McLain J, et al. Metagenomic characterization of antibiotic resistance genes in full-scale reclaimed water distribution systems and corresponding potable systems. Manuscript in review. Garrison LE, Kunz JM, Cooley LA, Moore MR, Lucas C, Schrag S, et al. 2016. Vital Signs: Deficiencies in Environmental Control Identified in Outbreaks of Legionnaires’ Disease - North America, 2000-2014. Morb. Mortal. Wkly. Rep. 65:576–584. Ginevra C, Jacotin N, Diancourt L, Guigon G, Arquilliere R, Meugnier H, et al. 2012. Legionella pneumophila sequence type 1/Paris pulsotype subtyping by spoligotyping. J. Clin. Microbiol. 50:696–701. Gordon M, Yakunin E, Valinsky L, Chalifa-Caspi V, Moran-Gilad J, ESCMID Study Group for Legionella Infections. 2017. A bioinformatics tool for ensuring the backwards compatibility of Legionella pneumophila typing in the genomic era. Clin. Microbiol. Infect. 23:306–310. Goris J, Konstantinidis KT, Klappenbach JA, Coenye T, Vandamme P, Tiedje JM. 2007. DNA- DNA hybridization values and their relationship to whole-genome sequence similarities. Int. J. Syst. Evol. Microbiol. 57:81–91. Hsu B-M, Wu S-F, Huang S-W, Tseng Y-J, Ji D-D, Chen J-S, et al. 2010. Differentiation and identification of Shigella spp. and enteroinvasive Escherichia coli in environmental waters by a molecular method and biochemical test. Water Research 44:949–955. Ji P, Rhoads WJ, Edwards MA, Pruden A. Metagenomic characterization of the hot water plumbing microbiome and relationship with opportunistic pathogen occurrence. Manuscript in preparation. Kozak-Muiznieks NA, Lucas CE, Brown E, Pondo T, Taylor TH, Frace M, et al. 2014. Prevalence of Sequence Types among Clinical and Environmental Isolates of Legionella pneumophila Serogroup 1 in the United States from 1982 to 2012. 52:201-211. Marston BJ, Lipman HB, Breiman RF. 1994. Surveillance for Legionnaires’ disease. Risk factors for morbidity and mortality. Arch. Intern. Med. 154:2417–2422. Mérault N, Rusniok C, Jarraud S, Gomez-Valero L, Cazalet C, Marin M, et al. 2011. Specific real- time PCR for simultaneous detection and identification of Legionella pneumophila serogroup 1 in water and clinical samples. Appl. Environ. Microbiol. 77:1708–1717. Meyer F, Paarmann D, D’Souza M, Olson R, Glass EM, Kubal M, et al. 2008. The metagenomics RAST server – a public resource for the automatic phylogenetic and functional analysis of metagenomes. BMC Bioinformatics 9:386. Michigan Department of Health and Human Services (MDHHS). 2016a. Legionellosis Outbreak- Genesee County, June, 2014 – March, 2015, Full Analysis. Michigan Department of Health and Human Services (MDHHS). 2016b. Legionellosis Outbreak- Genesee County May, 2015 – October, 2015, Full Analysis. Perola O, Kauppinen J, Kusnetsov J, Karkkainen U-M, Christian Luck P, Katila M-L. 2005. Persistent Legionella pneumophila colonization of a hospital water supply: efficacy of control methods and a molecular epidemiological analysis. APMIS 113:45–53. Pieper KJ, Tang M, Edwards MA. 2017. Flint Water Crisis Caused By Interrupted Corrosion Control: Investigating “Ground Zero” Home. Environ. Sci. Technol. 51:2007–2014. Price, M.N., Dehal, P.S., and Arkin, A.P. (2010) FastTree 2 -- Approximately Maximum- Likelihood Trees for Large Alignments. PLoS ONE, 5(3):e9490.

199

Qin T, Zhang W, Liu W, Zhou H, Ren H, Shao Z, et al. 2016. Population structure and minimum core genome typing of Legionella pneumophila. Sci. Rep. 6:21356. Rangel-Frausto MS, Sigfrido Rangel-Frausto M, Rhomberg P, Hollis RJ, Pfaller MA, Wenzel RP, et al. 1999. Persistence of Legionella Pneumophila in a Hospital’s Water System: A 13- Year Survey. Infect. Control Hosp. Epidemiol. 20:793–797. Raphael BH, Baker DJ, Nazarian E, Lapierre P, Bopp D, Kozak-Muiznieks NA et al. 2016. Genomic Resolution of Outbreak-Associated Legionella pneumophila Serogroup 1 Isolates from New York State. Appl. Environ. Microbiol. 28:3582-3590. Reimer AR, Au S, Schindle S, Bernard KA. 2010. Legionella pneumophila monoclonal antibody subgroups and DNA sequence types isolated in Canada between 1981 and 2009: Laboratory Component of National Surveillance. Eur. J. Clin. Microbiol. Infect. Dis. 29:191–205. Rhoads WJ, Garner E, Ji P, Zhu N, Parks J, Schwake DO, et al. 2017. Distribution System Operational Deficiencies Coincide with Reported Legionnaires’ Disease Clusters in Flint, Michigan. Environ. Sci. Technol. 51:11986–11995. Rhoads WJ, Ji P, Pruden A, Edwards MA. 2015. Water heater temperature set point and water use patterns influence Legionella pneumophila and associated microorganisms at the tap. Microbiome 3:67. Rodriguez-R LM, Konstantinidis KT. 2014. Bypassing Cultivation To Identify Bacterial Species. Microbe 9:111–118. Sabria M, Yu VL. 2002. Hospital-acquired legionellosis: solutions for a preventable infection. Lancet Infect. Dis. 2:368–373. Scaturro M, Dell’Eva I, Helfer F, Ricci ML. 2007. Persistence of the Same Strain of Legionella pneumophila in the Water System of an Italian Hospital for 15 Years. Infect. Control Hosp. Epidemiol. 28:1089–1092. Schwake DO, Garner E, Strom OR, Pruden A, Edwards MA. 2016. Legionella DNA Markers in Tap Water Coincident with a Spike in Legionnaires’ Disease in Flint, MI. Environ. Sci. Technol. Lett. 3:311–315. Unified Coordination Group—Flint, Michigan. 2016. Flint Rash Investigation. Wadowsky RM, Wolford R, McNamara AM, Yee RB. 1985. Effect of temperature, pH, and oxygen level on the multiplication of naturally occurring Legionella pneumophila in potable water. Appl. Environ. Microbiol. 49:1197–1205. Whitby PW, Carter KB, Burns JL, Royall JA, LiPuma JJ, Stull TL. 2000. Identification and detection of Stenotrophomonas maltophilia by rRNA-directed PCR. J. Clin. Microbiol. 38:4305–4309. Young M. 2017. Social media scare prompts health department to release info on bacteria. MLive.com. Available: http://www.mlive.com/news/flint/index.ssf/2017/04/social_media_scare_prompts_hea.ht ml [accessed 9 December 2017]. Zahran S, McElmurry SP, Kilgore, PE, Mushinski, D, Press, J, Love, et al. 2018. Assessment of the Legionnaires’ disease outbreak in Flint, Michigan. Proc. Nat. Acad. Sci. Published online ahead of print, Feb 5, 2018.

200

SUPPLEMENTARY INFORMATION FOR CHAPTER 8

Table S1: Summary of environmental and clinical L. pneumophila isolates subject to whole genome sequencing Year Month Isolate Building Sample Tap or Flushed/ Water Source (April SampleIDa ST SG Collected Collected Type Typeb Source Stagnant 2014-October 2015) C1 2015 Clinical 159 1 C2* 2015 Clinical 1 1 C3* 2015 Clinical 1 1 C4 2015 Clinical 213 1 C5 2015 Clinical 213 1 C6 2015 Clinical 44 1 C7 2015 Clinical 1 1 C8 2015 Clinical 211 1 C9* 2015 Clinical 222 1 C10 2015 Clinical 2513 1 HC01 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC02 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC03 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC04 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC05 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC06 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC07 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC08 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC09 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC10 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC11 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC12 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC13 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HC14 2016 March Water Hospital #1 Cold Stagnant Flint 2518 6d HH01 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH02 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH03 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH04 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d

201

HH05 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH06 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH07 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH08 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH09 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH10 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH11 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH12 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH13 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH14 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH15 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH16 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH17 2016 March Water Hospital #1 Hot Stagnant Flint 1 1 HH18 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH19 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH20 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH21 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH22 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH23 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH24 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH25 2016 March Water Hospital #1 Hot Stagnant Flint 1 1 HH26 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH27 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH28 2016 March Water Hospital #1 Hot Stagnant Flint ND ND HH29 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH30 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH31 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH32 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH33 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH34 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH35 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH36 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH37 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d

202

HH38 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH39 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH40 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH41 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH42 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH43 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH44 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH45 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH46 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH47 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH48 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH49 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH50 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH51 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH52 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH53 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH54 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH55 2016 March Water Hospital #1 Hot Stagnant Flint 2518 6d HH56 2016 March Water Hospital #1 Hot Stagnant Flint 1 1 PC01 2016 March Water Public Cold Stagnant Flint 2518 6e building RC01 2016 June Water Residence Cold Stagnant Flint 192 1 RC02 2016 June Water Residence Cold Flushed Flint 192 1 RC03 2016 August Water Residence Cold Stagnant Flint 192 1 RC04 2016 June Water Residence Cold Stagnant Flint 192 1 RC05 2016 June Water Residence Cold Stagnant Flint ND 1 RC06 2016 June Water Residence Cold Stagnant Flint 192 1 RC07 2016 June Water Residence Cold Flushed Flint 192 1 RD01 2016 June Water Residence HWHDV NA Flint 192 1 RD02 2016 June Water Residence HWHDV NA Flint 192 1 RD03 2016 June Water Residence HWHDV NA Flint 192 1 RD04 2016 June Water Residence HWHDV NA Flint 192 1 RD05 2016 June Water Residence HWHDV NA Flint 192 1

203

RH01 2016 June Water Residence Hot Flushed Flint NA 1 RH02 2016 June Water Residence Hot Stagnant Flint 192 1 RH03 2016 June Water Residence Hot Flushed Flint 192 1 RH04 2016 June Water Residence Hot Flushed Flint 192 1 RH05 2016 June Water Residence Hot Stagnant Flint 192 1 RH06 2016 June Water Residence Hot Stagnant Flint 192 1 RH07 2016 June Water Residence Hot Stagnant Flint 192 1 RH08 2016 August Water Residence Hot Stagnant Flint 1 1 RS01 2016 June Water Residence Shower Stagnant Flint 192 1 RS02 2016 August Water Residence Shower Stagnant Flint 192 1 RS03 2016 June Water Residence Shower Stagnant Flint ND ND WC01 2016 March Water Well Water Cold Stagnant well ND ND WC02 2016 March Water Well Water Cold Stagnant well ND ND WC03 2016 March Water Well Water Cold Stagnant well ND ND WC04 2016 March Water Well Water Cold Stagnant well ND ND WH01 2016 March Water Well Water Hot Stagnant well ND ND WH02 2016 March Water Well Water Hot Stagnant well ND ND WH03 2016 March Water Well Water Hot Stagnant well 2518 6d WH04 2016 March Water Well Water Hot Stagnant well ND ND WH05 2016 March Water Well Water Hot Stagnant well ND ND pos_con* + control 42 1 neg_con* - control NA NA aIsolates were named according to the following system: First letter indicates building type/location (H=hospital; R=residence; W=school using well water; P=large public building), second letter indicates sample collection location (hot water tap (H), cold water tap (C),water heater drain valve (D), shower (S)), followed by a unique numeric identifier. Clinical strains are denoted C1-10. bUnless otherwise indicated, all buildings were serviced by Flint municipal water derived from the Flint River during the Flint Water Crisis dPresumed serogroup 6 based on direct fluorescent antibody staining of a phylogenetically diverse subset of isolates belonging to serogroup 2518 eVerified serogroup 6 using direct fluorescent antibody staining *indicates isolate was prepared and sequenced twice with consistent results as an additional control. HWHDV= hot water heater drain valve; ST = sequence type; SG = serogroup; ND=could not be determined due to insufficient genome coverage; NA=not applicable

204

Table S2: Clinical reference strains selected for comparison to water isolates. Sample ID GenBank Accession Origin Serogroup Sequence Type Number LP Philadelphia AE017354.1 USA 1 ST-136 LP ATCC 43290 CP003192.1 USA 12 ST-187 LP Alcoy CP001828.1 Spain 1 ST-578 LP Corby CP000675.2 UK 1 ST-51 LP Lens CR628337.1 France 1 ST-15 LP 130b FR687201.1 USA 1 ST-42 LP Paris CR628336.1 France 1 ST-1 LP Lorraine FQ958210.1 France 1 ST-47 LPHL06041035 FQ958211.1 France 1 ST-734 Legionella clemsonensis NZCP016397.1 NZLN614827.1 NZLN681225.1 NC013861.1 NZCP004006.1

Table S3: MG-RAST identifiers associated with each metagenome and barcodes used to multiplex samples for shotgun metagenomic sequencing Sample MG-RAST ID Barcode

Flint River mgm4735990.3 CGATGT

Residence A (cold) mgm4735991.3 TGACCA

Residence A (biofilm) mgm4735992.3 ACAGTG

DWSD Residence mgm4735993.3 GCCAAT

Business 1 (cold) mgm4735994.3 CAGATC

Hospital #1 - tap i (cold) mgm4735995.3 CTTGTA

Hospital #2 (cold) mgm4736011.3 AGTCAA

Hospital #2 (hot) mgm4736012.3 AGTTCC

Hospital #1 - tap ii (hot) mgm4736013.3 ATGTCA

Hospital #1 - tap iii (hot) mgm4736016.3 CCGTCC

Hospital #1 - tap i (cold) mgm4736014.3 GTCCGC

Business 2 (hot) mgm4736020.3 GTGAAA

Residence B mgm4736017.3 ATCACG

Well water mgm4736022.3 TTAGGC

Business 3 (cold) mgm4736023.3 ACTTGA

Hospital #1 - tap iii (cold) mgm4736029.3 TAGCTT

Hospital #1 - tap iii (hot) mgm4736028.3 GATCAG

Hospital #1 - tap i (cold) mgm4736027.3 GGCTAC

205

Figure S1: Phylogenetic tree generated using FastTree (Price et al. 2010) based on extracted 16S rRNA gene sequences using PhyloSift (Darling et al. 2014).

206

Figure S2: Phylogenetic tree generated using FastTree (Price et al. 2010) based on 37 single-copy housekeeping genes in amino acid space using PhyloSift (Darling et al. 2014)

207

Figure S3: Phylogenetic tree generated using FastTree (Price et al. 2010) based on 37 single-copy housekeeping genes in nucleotide space using PhyloSift (Darling et al. 2014)

208

CHAPTER 9 : CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE WORK

In order to protect public health and advance sustainability of society’s water resources, it is imperative that research is conducted to characterize microbial contaminants throughout the urban water cycle and identify strategies to limit their transmission to downstream populations. While source water protection, filtration, and disinfection have all contributed tremendously to improvements in public health and accessibility to safe drinking water in developed countries, additional microbial challenges have emerged. Regrowth of opportunistic pathogens (OP) in distribution systems and premise plumbing and the transmission of antibiotic resistant bacteria (ARB) and their associated resistance genes (ARG) represent two key challenges for which ongoing research is needed. This dissertation contributed to this need by characterizing the occurrence of OPs and ARGs throughout the urban water cycle and identified potential strategies by which control of these contaminants may be achieved, given further research and development.

Chapter 2 of this dissertation described the mechanisms by which these organisms can be transmitted, particularly by recycled wastewater systems, and identified critical challenges with respect to wastewater reuse and safeguarding public health. We proposed adoption of a “human exposome” framework to guide development of risk management strategies for holistic assessment of recycled water quality. Namely, this would require consideration of both acute and chronic exposures to recycled water, consideration of all routes of exposure to recycled water, including not only ingestion, but also inhalation and dermal contact, and assessment of water quality at the point of use, rather than at the location of treatment.

In the third chapter of this dissertation, monitoring of an urban watershed during storm flow conditions revealed that storm-driven transport of ARGs contributed significantly to surface water loadings of ARGs, with loadings of certain ARGs being as high as two orders of magnitude greater during storm conditions than during equivalent background periods. In addition, some ARGs (e.g., tet(O) and tet(W)) were found to correlate with fecal indicator bacteria, while others were not, suggesting that the processes governing fate and transport are not the same for all ARGs.

The fourth chapter of this dissertation identified a decrease in ARGs in a watershed as a result of extreme rainfall and flooding, presumably due to dilution of ARGs suspended in surface water and transport of sediment ARGs downstream. Within ten months of post-flood recovery, however, the system had returned to pre-flood abundances of ARGs, suggesting that ARG sources persisted after the flood. Bacterial phylogeny was not correlated with ARG in water or sediment samples, but correlations were noted between ARGs and several antibiotics and metals, suggesting they may have exerted selective pressure for ARB in post flood recovery. Identification of ARGs on scaffolds assembled from metagenomics data co-occurring with mobile genetic element- associated genes highlighted the potential for intercellular transmission of ARGs.

In Chapter 5, a survey of four paired full-scale non-potable reclaimed and potable water distribution systems in the United States revealed that the ARG sul1 was consistently elevated in reclaimed water compared to potable water, while other ARGs were elevated only at select utilities. Of critical importance is the finding that ARGs were generally not significantly reduced by tertiary wastewater treatment or drinking water treatment. A multitude of ARGs were found to be co- located with plasmid gene markers on metagenomic scaffolds, again demonstrating potential for

209

intercellular transmission of ARGs to occur. Correlations between several ARGs and potential selective agents (e.g., antibiotics, metals, and disinfectants) in distribution systems suggest that these agents may select for ARGs or shape a microbial community that is pre-disposed to carriage of ARGs. Weak correlations were also noted between ARGs and the overall microbial community, as well as several key phyla.

In the sixth chapter, the survey of reclaimed and potable distribution systems was further examined and genes associated with two genera containing OPs (Legionella spp., Mycobacterium spp.) and total bacteria (16S rRNA genes) were found to be more abundant in reclaimed water systems than corresponding potable systems. This study identified key characteristics of reclaimed water that generally differ from potable water (i.e., nutrient concentration, temperature), that are likely to contribute to the observed differences in OP occurrence. In addition, correlations were observed between different amoebic hosts and Legionella spp. genes, suggesting that different interactions between members of the microbial community may contribute to differences in OP abundances between reclaimed and potable water.

The seventh chapter explored the role of pilot- or bench-scale direct potable reuse (DPR) treatment at four U.S. utilities in producing biologically-stable water during distribution and stagnation in premise plumbing. All utilities and treatment scenarios produced water that resulted in regrowth of total bacteria (16S rRNA genes) during incubation in simulated premise plumbing rigs. However, the regrowth of total bacteria, OPs, and ARGs was not significantly greater for any DPR blends treated with advanced oxidative processes (AOP) than in corresponding traditional potable waters. Biodegradable dissolved organic carbon, a measure of biological stability, was not correlated with total bacteria, any OP gene markers, or any ARGs, suggesting that in this highly treated water, organic carbon is likely not the limiting nutrient controlling regrowth.

In the eighth chapter, samples collected from the compromised Flint, Michigan drinking water distribution system during the Flint Water Crisis revealed that strains of Legionella pneumophila isolated from the hot water of three hospital taps and one residence tap had a high degree of genomic similarity to a subset of outbreak associated strains collected from patients in the area. In addition, a diverse variety of strains were found in the Flint municipal water system, suggesting that the crisis contributed to the growth of multiple strains of L. pneumophila.

Together, these chapters describe an advancement in knowledge regarding the occurrence of OPs and ARGs in a variety of water systems, and highlights trends that may be of value in developing management strategies for limiting regrowth or transmission of these bacteria in various compartments of the urban water cycle. The research described herein raises numerous additional needs for research that are critical to the continued advancement of science in this realm.

In particular, numerous research needs exist regarding the behavior of ARGs in various water systems. This research provides evidence that several key processes relevant to antibiotic resistance (i.e., horizontal gene transfer; selection by antibiotics, metals, and disinfectants; interactions with the microbial community; and influence of water chemistry) are likely to be occurring throughout the urban water cycle, but additional research is needed to confirm the occurrence of these phenomena in situ. It is also important to determine the rates at which these processes occur in situ, as this information could offer valuable insight into which processes could

210

most effectively be targeted as potential strategies for control of the transmission of ARGs. In addition, research is critically needed to develop approaches by which to quantify risk associated with detection of ARGs in water systems. Traditional pathogen risk models are poorly suited to characterize risk associated with ARGs because they fail to account for horizontal gene transfer and selection of ARGs by chemical compounds. Accordingly, information is needed about what the disease burden is associated with antibiotic resistant infections associated with the waterborne transmission of ARB and ARGs

In addition, identification of key “indicator” ARGs or associated genetic elements that can be used for monitoring of risk associated with antibiotic resistance would be a valuable development that would enable utilities, governments, and researchers to characterize ARGs in their systems or watersheds. Given that ARGs are naturally occurring and can be found even in pristine environments, some level of background abundance of ARGs is likely to occur in nearly all water systems. It is important to identify which genes and at what levels are acceptable background concentrations, and which are cause for concern. Finally, strategies for intervention are critically needed. Currently, the most effective treatment approaches and management strategies for limiting spread of ARGs are unclear. For example, while AOPs have shown promise for removing ARGs from wastewater, there are conflicting reports that indicate that removal is dependent on the target microorganism and genes monitored. Further research is needed to optimize the ability of AOPs or other treatment methods to remove ARGs.

Key research needs also exist regarding the occurrence of OPs, particularly in recycled wastewater systems. Firstly, the disease burden associated with OP exposure via recycled water needs to be characterized and research is needed to advance hazard identification, exposure assessment, and even to establish the infectious dose for key OPs. Such advancements could serve to inform decisions about management strategies and identify which routes of exposure can most effectively be limited to minimize OP exposure. Research is needed to advance hazard identification, exposure assessment The results described in this dissertation suggests that very different processes may govern the occurrence and regrowth of OPs in reclaimed versus potable water, so it may not be appropriate to assume that management strategies that are effective in potable systems will also be effective in recycled water system. Additional research is needed to determine the suitability of using traditional potable water control strategies in reclaimed systems, and novel treatment and control strategies for these systems are needed. In addition, research is needed to better discern the relative role of each of the complex factors identified as likely contributors to influencing OP regrowth.

211