Dissertation (2.087Mb) (2024)

ABSTRACT

Advancing an Understanding of Ecological Risk Assessment Approaches for Ionizable Contaminants in Aquatic Systems.

Theodore W. Valenti Jr., Ph.D.

Committee Chairperson: Bryan W. Brooks, Ph.D.

Freshwater is increasingly becoming a finite resource in many regions of the world. Gaps between estimated water supply and demand continue to narrow and the prospects of acquiring additional sources of freshwater remain limited. Furthermore, economically efficient water resource management practices are perplexed by increasing urbanization and changing land-use in semi-arid regions. Although repeated use of water is a practical and effective means for easing strain on water supplies, there is concern that unnecessary contamination may diminish future value of this important resource. Some surface waters in semi-arid regions of the U.S. are effluent-dominated as flow is comprised of >90% treated wastewater. Ionizable compounds are chemicals often associated with urban development and examples include pharmaceuticals, agrochemicals, natural toxins, and other common contaminants (e.g. ammonia). Because continued population growth and urbanization are likely to increase contaminant release and alter dilution capacity of receiving systems, it is important that best management approaches are developed at the watershed scale to limit water quality degradation

associated with ionizable compounds. Current methods for prospective and retrospective ecological risk assessments of ionizable compounds seldom consider site-specific conditions during the analysis of effects of phase. Ionization state is largely controlled by the acid/base dissociation constant (pKa) and pH of the solution where a compound resides. Stream water quality can therefore influence ionization state, which is important because the unionized forms a more lipophilic and have a greater propensity to cross cellular membranes. Consequently, the unionized forms are hypothetically more toxic. I completed toxicity tests in the laboratory using various contaminants as model ionizable compounds over a gradient of environmentally-relevant surface water pH and then related measured toxicological endpoints to observed pH of surface waters using both discrete and probabilistic ecological risk assessment approaches. The result of my studies clearly demonstrated that site-specific pH may influence the toxicity of ionizable contaminants. Potential modifications to conceptual frameworks of ecological risk assessment for ionizable contaminants are suggested so that uncertainty can be reduced.

Advancing an Understanding of Ecological Risk Assessment Approaches for Ionnizable Contaminants in Aquatic Systems

by

Theodore W. Valenti Jr., B.S., M.S.

A Dissertation

Approved by the Institute of Ecological, Earth, and Environmental Science

______Joseph W. White, Ph.D., Chairperson

Submitted to the Graduate Faculty of Baylor University in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

Approved by the Dissertation Committee

______Bryan W. Brooks, Ph.D., Chairperson

______C. Kevin Chambliss, Ph.D.

______Robert D. Doyle, Ph.D.

______Ryan S. King, Ph.D.

______Joseph C. Yelderman Jr., Ph.D.

Accepted by the Graduate School August 2010

______J. Larry Lyon, Ph.D., Dean

Page bearing signatures is kept on file in the Graduate School.

Copyright © 2010 by Theodore W. Valenti Jr.

All rights reserved

TABLE OF CONTENTS

LIST OF FIGURES ...... viii LIST OF TABLES ...... x ACKNOWLEDGMENTS ...... xii DEDICATIONS ...... xiii CHAPTER ONE ...... 1 Introduction ...... 1 General Overview ...... 1 Why is Site-Specific pH Important to Ecological Risk Assessment? ...... 2 Factors that Influence Site-Specific pH ...... 5 Scope of Dissertation ...... 7 CHAPTER TWO ...... 8 Aquatic Toxicity of Sertraline to Pimephales promelas at Environmentally Relevant Surface Water pH ...... 8 Introducti ...... 8 Materials and Methods ...... 11 Experimental Conditions ...... 11 Pimephales promelas Bioassays ...... 12 Statistical Analyses ...... 15 Analytical Methods ...... 15 Continuous Water Quality Monitoring Data...... 17 Predictions of Ecotoxicological Responses ...... 18 Results ...... 19 Analytical Results ...... 19 Acute Studies ...... 19 Short-term Chronic Experiments ...... 21 Continuous Water Quality Monitoring Data...... 22 Predictions of Ecotoxicological Responses ...... 22 v

Discussion ...... 24 Behavioral Responses To SSRIs ...... 31 Importance Of Site-Specific pH...... 34 CHAPTER THREE ...... 36 Sublethal Effects of the Selective Serotonin Reuptake inhibitor (SSRI) Sertraline on Fathead Minnow Under Potential Worst-case Environmental Exposure Scenarios. .... 36 Introduction ...... 36 Materials and Methods ...... 42 Test Organisms ...... 42 Experimental Design ...... 42 Behavioral Trials ...... 43 Analytical Quantification of Sertraline ...... 48 Comparing Measured Versus Predicted Fish Plasma Concentrations ...... 49 Statistical Analysis ...... 50 Results ...... 51 Analytical Quantification of Sertraline ...... 51 SERT Binding Study...... 52 Behavioral Trials ...... 53 Discussion ...... 56 CHAPTER FOUR ...... 61 A Mechanistic Explanation for pH-Dependent Ambient Aquatic Toxicity of Prymnesium parvum Carter ...... 61 Introduction ...... 61 Material and Methods ...... 64 Bioassays with Samples Obtained from Reservoirs Experiencing Blooms ...... 64 Laboratory Culture Preparation ...... 66 Bioassays with Samples Obtained from Laboratory Cultures ...... 67 Statistical Analysis ...... 68 Estimation of Prymnesin-1 and -2 Physicochemical Properties ...... 68 Results ...... 70 pH Dependent Toxicity in Field Studies: Lakes Granbury and Whitney ...... 70

vi

pH Dependent Toxicity in Laboratory Cultures ...... 74 Prymnesin-1 and -2 Physicochemical Properties ...... 75 Discussion ...... 75 CHAPTER FIVE ...... 83 Interannual Hydrological and Nutrient Influences on Diel pH in Wadeable Streams: Implications for Ecological Risk Assessment of Ionizable Contaminants...... 83 Introduction ...... 83 Material and methods ...... 88 Study Sites ...... 88 Diel Water Quality Monitoring ...... 89 Nutrient Measurements ...... 91 pH Influences on Aquatic Toxicity ...... 91 Daily oscillation risk ratio ...... 92 Relationship Between High TP and Elevated pH ...... 92

Estimates of BCF and Dlip-water at Stream Sites ...... 92 Statistical Analysis ...... 94 Results ...... 95 Continuous Water Quality Monitoring and Nutrients ...... 95 pH Influences on Aquatic Toxicity ...... 98 Daily Oscillation Risk Ratios...... 101

Predicted BCF and Dlip-water ...... 103 Discussion ...... 104 APPENDIX A: ...... 117 Influence of pH on amine toxicology and implications for harmful algal bloom ecology ...... 117 APPENDIX B ...... 135 Licensing Agreement for Appendix A ...... 135 Licensing Agreement for Chapter Two ...... 136 Licensing Agreement for Chapter Four ...... 137 REFERENCES ...... 139

vii

LIST OF FIGURES

Figure 1. The figure depicts the change in ionization state for hypothetic compounds with different pKa values between environmentally relevant surface pH gradients spanning between pH 6 – 9 and pH 5 - 10.

Figure 2. 48-h median lethal concentration (LC50) values for Pimephales promelas exposed to sertraline at three pH treatment levels (6.5, 7.5, and 8.5).

Figure 3. Time-to-death for juvenile Pimephales promelas exposed to 500 μg sertraline L-1 in reconstituted hard water adjusted to three different pHs.

Figure 4. Seasonal frequency distributions of pH by the number of observations at two continuous monitoring stations in the Brazos River Basis, Texas, USA.

Figure 5. The 48-h mean (±standard deviation) mean lethal concentration (LC50) values for Pimephales promelas presented as total and percent unionized sertraline concentrations.

Figure 6. A photo of the dive tank used to assess differences in the behavior of adult male Pimephales promelas unexposed and exposed to sertraline based on work by Levin et al. 2007.

Figure 7. A photo of the plus maze used to assess differences in behavior for sertraline exposed and control adult male Pimephales promelas.

Figure 8. A drawing of the NOLDUS apparatus used to assess differences in the behavior of adult male Pimephales promelas unexposed and exposed to sertraline.

Figure 9. The measured versus predicted fish plasma concentration based on the Huggett et al. model (2003). Closed dots are based on Log D, open dots are based on Dlipwater.

Figure 10. The mean amount of time that Pimephales promelas exposed to sertraline spent in white areas of plus maze.

Figure 11. The mean number of times that Pimephales promelas exposed to sertraline crossed into white areas of plus maze.

Figure 12. The mean number of times that Pimephales promelas exposed to sertraline crossed different areas in the dive tank.

viii

Figure 13. The mean amount of times that Pimephales promelas exposed to sertraline spent in the bottom area of the dive tank.

Figure. 14. Average survivorship (±SD, n=4) of Pimephales promelas exposed to dilutions of Lake Whitney water collected during a bloom of Prymnesium parvum in 2007.

Figure 15. Mean neonate production for Daphnia magna (±SD, n=5) exposed to diluted samples of Lake Whitney water collected during a Prymnesium parvum bloom in 2007.

Figure. 16. LC50 values for Pimephales promelas exposed to cultures of Prymnesium parvum grown in the laboratory using two different nutrient conditions (high nutrients – f/2 medium; low nutrients – f/8 medium).

Figure 17. The percent survivorship of Pimephales promelas exposed to samples of Prymnesium parvum grown under different nutrient conditions with cells (whole culture) and cells removed (filtrate).

Fig.ure 18. The structures of prymnesin-1 and prymnesin-2 with the hydrophobic and hydrophilic portions of each compound differentiated.

Figure 19. Daily patterns in measured temperature and dissolved oxygen on the left axis and pH (Dashed lines) on the right axis sampled under low (2006) and high (2007) conditions.

Figure 20. Water column concentrations of total nitrogen (TN), total phosphorus (TP), daily change in dissolved oxygen, and pH at 23 stream sites in the Brazos watershed under low flow (2006) and high flow (2007) conditions

Figure 21. The results of nonparametric changepoint analysis using surface water TP as the predictor variable and % of time that pH >8.5 as the response variable.

Figure 22. The allowable water column concentrations of TN at sites based on pH- dependent relationships reported in ambient water quality criteria for ammonia (US EPA 2009).

Figure 23. Predicted LC50 value for sertraline based on pH-dependent toxicological reported by Valenti et al. (2009).

Figure 24. The percent of the day that predicted bioconcentration factors will be greater than 1000 for several pharmaceuticals that are weak bases at stream sites during low (2006) and high (2007) hydrology.

ix

LIST OF TABLES

Table 1. Analytical verification of nominal sertraline treatment levels for acute and short-term chronic experiments with Pimephales promelas.

Table 2. Results of short-term chronic experiments with Pimephales promelas exposed to sertraline at pH 6.5, 7.5, and 8.5. The average survivorship, growth, and feeding rate for the various sertraline concentrations at different pH.

Table 3. The pH seasonal averages at two Texas Commission of Environmental Quality continuous water quality monitoring stations in the Brazos Watershed (TX, USA).

Table 4. The likelihood that the predicted sertraline median lethal concentration (LC50) and growth and feeding 10% effective concentration (EC10) values for Pimephales promelas will be below the comparative value at two sites based on aquatic toxicological models derived during laboratory bioassays and distributions of surface water pH.

Table 5. Measured sertraline concentration in treatment water for the 28 d chronic experiments with Pimephales promelas.

Table 6. The results of the SERT bound by radiolabeled citalopram experiments.

Table 7. The results of NOLDUS experimental trials including the mean (± standard error) amount of time adult Pimephales promelas spent in the shelter, the mean total distances (± standard error) traveled, and mean (± standard error) velocity

Table 8. The percent survivorship in undiluted samples and 48-h LC50 values in terms of percent reservoir water for Pimephales promelas exposed to Lake Granbury samples from three stations during a Prymnesium parvum bloom in March 2007.

Table 9. The 48- and 96-hr LC50 values in terms of percent reservoir water for Daphnia magna exposed to a composite sample obtained from Lake Granbury during a Prymnesium parvum bloom in 2007.

Table 10. The LC50 value and respective 95% confidence intervals for experiments completed with Pimephales promelas and cultures of Prymnesium parvum grown in f/2 and f/8 media that were either unfiltered or filtered to remove cells.

x

Table 11. The predicted physiochemical properties of prymnesin-1 and -2 based on computer modeling and hand computation.

Table 12. The location, physical descriptors, dams, wastewater outfall, and land-use breakdown for the 24 sites sampled during 2006 and 2007.

Table 13. The dissociation constant (pKa value), octanol-water partitioning coefficient (Log D), bioconcentration factors, and Dlipwater for several weak base pharmaceuticals.

Table 14. Potential nutrient inputs, average dissolved oxygen (mg/L) + standard deviation (SD), average pH+ SD, and temperature + SD at stream sites in the Brazos River Watershed, TX during low flow (2006) and high flow (2007) conditions.

Table 15. The daily oscillation risk ratio (DORR) for the weak bases ammonia and sertraline at stream sites in the Brazos River Watershed, TX during low flow (2006) and high flow (2007) conditions.

Table 16. The weighted mean Klipw for seven weak base pharmaceuticals at 23 stream sites the Brazos River Watershed, TX during low flow (2006) and high flow (2007) conditions.

xi

ACKNOWLEDGMENTS

This research was partially funded by a U.S. Geological Survey/Texas Water

Research Institute grant to TW Valenti and Dr. BW Brooks, the Glassco*ck Fund for

Excellence in Environmental Science to TW Valenti, and a U.S. Protection Agency grant

(U.S. EPA grant EM96638001) to B.W. Brooks, and a Texas Parks and Wildlife

Department/U.S. Fish and Wildlife Service grant to B.W. Brooks, Dr. J.P. Grover, and

Dr. DL Roelke.. Travel support to conferences came from the Baylor University

Graduate School, SETAC travel award, and southwest regional SETAC chapter.

I thank my committee (Dr. BW Brooks, Dr. C.K. Chambliss, Dr. RS King, Dr.

RD Doyle, and Dr. J Yeldermen) as well as Dr. GG Gould and Dr. R Brain for sharing

their insight and providing me guidance in the laboratory. Also, G Gable and L

Schwierzke from TAMU for assistance with field sampling as well as my colleagues and

peers at Baylor University, including Dr. JK Stanley, J Back, J Taylor, F Urena-Boeck,

ML Lahousse, SV James, JP Berninger, KA Connors, and KN Prosser for all their

assistance and support.

Foremost, my parents, fiancée, and family and their unconditional love and

support throughout my education. Mom, Dad, Dominick, Jamie, and Mary, you taught

me valuable lessons in life that cannot be found in any book and instilled in me a strong

sense of self-value. My determination and drive are fueled by your caring and you play a

pivotal role in my accomplishments. Sheena L. Shipley, your love and encouragement

allow me to enjoy life and I am happy that we shall share the rest of our lives together.

xii

DEDICATION

To my family

xiii

CHAPTER ONE

Introduction

General Overview

Freshwater is increasingly becoming a finite resource in Texas and at the global

scale (Gleick 2003a). The gap between estimated water supply and demand in the state is

narrowing, and the prospect of acquiring additional sources of freshwater are limited (Oki

and Kanae 2006). Furthermore, rapid urbanization in select regions (Murdock et al.

1997) perplexes economically efficient water resource management practices. Population

growth may continue to proportionally increase demand for municipal water use and

further strain the state’s already tight water budget. To account for some of these

shortages, it is imperative that Texas implements policies targeting conservation and water reuse (Gleick 2003b). Although repeated use of water is a practical and effective

means for easing strain on the water supply, there is concern that unnecessary

contamination may diminish future value of this important resource (Toze 2006 a + b) .

Already some surace waters in the state are classified as perennially effluent-dominant

and Brooks et al. (2006) described instances when base flow of some rivers in Texas are

comprised of >90% treated wastewater. Ionizable compounds are chemicals often

associated with urban development and examples include pharmaceutical and personal

care products (PPCP), pesticides, fertilizers and ammonia. Because continued population

growth and urbanization will likely increase the release of these contaminants into

1

waterways, it is important that best management approaches are developed at the watershed scale to decrease water quality degradation by ionizable compounds.

Current ethods for prospective and retrospective ecological risk assessments of

ionizable compounds seldom consider site-specific conditions during the analysis of

effects of phase (Cleuvers 2003, Bound and Voulvoulis 2004, Sanderson et al. 2004,

Suter et al. 2007, Suter et al. 2000). This oversight may needlessly increase uncertainty

associated with water quality management decisions. Ionization state is largely controlled

by the acid/base dissociation constant (pKa) and pH of the solution where a compound resides. Consequentially, instream water quality parameters will influence the proportion

of ionized and unionized forms of a compound (Van Wezel 1998). This may have profound implications on aquatic risk assessment as unionized forms are often more toxic because of their greater lipophilicity The U.S. EPA states that site-specific ambient water quality criteria hould be developed if differences in physical and chemical characteristics of water influence the biological availability and/or hazard of a given contaminant of concern (US EPA 1996, 1999). The overlying goal of my dissertation work is to emphasize the need for site-specific pH considerations during ecological risk assessments for both traditional and emerging ionizable contaminants of concern.

Why is Site-Specific pH Important to Ecological Risk Assessment?

Contaminants, such as some pesticides, fertilizers, PPCPs, are designed to be ionizable to maximize efficacy for their intended purpose. In the case of pesticides this may be exemplified by manufacturer’s suggestions to adjust the pH of spray water before applying products to crops. The importance of pH for agrichemical use and associated implications for phytotoxicty has long be recognized (Blackman and Robsertson-

2

Cuninghame 1952) and is exemplified by more recent work by Green and Hale (2005).

They described that the efficacy of the weak acid herbicide nicosulfurnon can be

optimized by first increasing the pH of spray water to make the compound more soluble,

and then reducing the pH of spray water before application to increases its lipophilicity so

that it will have a greater propensity for uptake by target plants. Another example of how

ionization state may influence biological activity is a pharmaceutical designed to works

on the central nervous system. Most drugs are administered orally and therefore must

first pass through the digestive tract and then be transported via the blood to specific

targets within the body. Various regions of the body have different pH (stomach: pH 1.5,

blood: pH 7, CNS: pH 8), which can alter the bioavailability of drugs by influencing their

physiochemical properties (Hernandez and Rathinavelu 2006, Kwon 2001). These

interactions are summarized in the pH-partition theory of drug absorption, which takes

into account three factors that influence the partitioning process of a drug between water

and lipid at different pH: 1) the dissociation constant (pKa), 2) lipophilicity of the compound, and 3) pH of the absorption site (Jollow and Brodie 1972, Kwon 2001).

Similar to these pharmaco*kinetic principles, toxico*kinetic principles may be applied for aquatic ecosystems and help scientists more accurately infer environmental hazard.

The environmental availability and physicochemical properties of ionizable contaminants could potentially change appreciably over the range of pH for most minimally impacted freshwater ways in the United States (pH 6-9) depending on their dissociation (Van Wezel 1998). Several factors influence ionization state, including temperature, aqueous pH, and the structural configuration of the compound, which actually determines the pKa value. The pKa is merely a reflection of when the

3

compounds will exist equally (50:50%) as ionized and unionized moieties. Weak bases will increasing become unionized as the aqueous pH approaches and surpasses the compounds pKa value; the opposite is true for weak acids as they are more likely to exist as the unionized form at lower pH. Ultimately, contaminants that have pKa values within the range of surface water pH are more likely to vary in their ionization state at or among sites (Figure 1).

100 pH 5 ‐ 10 state 80 pH 6 ‐ 9 60

ionization 40

in 20

0 change

% 1234567891011121314 pKa value

Figure 1. The figure depicts the change in ionization state for hypothetic compounds with different pKa values between environmentally relevant surface pH gradients spanning between pH 6 – 9 and pH 5 - 10.

Ionization state may alter ambient toxicity to aquatic life and the environmental relevance of such is emphasized by the integration of site-specific pH adjustment factors into national ambient water quality criteria in the United States for compounds such as such as ammonia (NH3) and pentachlorophenol (PCP) (US EPA 1996, 1999). Both the weak base NH3 (pKa = 9.3) and the weak acid PCP (pKa = 4.7) are more toxic to aquatic

4

life when they occur in the environment predominately as the unionized forms (US EPA

1996, 1999). To account for these differences, pH adjustment factors are derived by

relating site-specific pH to laboratory derived toxicological data from experiments completed with the compound of interest at various ambient pH levels. Site-specific pH consideration can result in markedly different acceptable loads in receiving systems; varying by 13-fold for NH3 and 60-fold difference for PCP between sites with contrasting

surface water pH values of 6 and 9 (US EPA 1996, 1999).

Factors that Influence Site-Specific pH

There can be substantial spatiotemporal variability in surface water pH at sites

within and among watersheds due to differences in geomorphology, hydrology, and

climate (Allan 1995). The surface water pH at a given site culminates from time-

dependent interactions that influence the discharge and chemical composition of base

flow, storm flow, and groundwater. Base flow and storm flow are likely more variable in

the headwaters and may be influenced substantially by precipitation and/or winter

melting, especially in flashy systems. Storm events may cause pulses or dilution of

various constituents from surrounding land areas; which may collectively cause

temporary shifts in the typically pH at a site (Allan 1995). Groundwater interactions with bedrock are also very important and are controlled by residency time and the type of parent material, soil depth, slope, and water table. The size and shape of a water body may also have connotations for atmospheric interactions that influence site-specific pH.

Carbon dioxide (CO2) exchange between the atmosphere and surface water affects pH as

CO2 in water may be converted to carbonic acid, thus causing a decrease in pH (Maberly

1996). Atmospheric exchange is more likely to have an effect in shallow water bodies

5

that have large surface area: volume ratio. The impacts of atmospheric exchange are

perhaps best evidenced by current problems of acid rain; however, the global increase of

CO2 causes broader concerns about implications associated with global climate change

(Fung et al. 2005, Doney et al. 2007). Inevitably, interactions between surface water and

either bed rock or the atmosphere are influenced by climate, especially temperature, precipitation, and partial pressure. Consequently, seasonal shifts in pH are often experienced at sites and may be caused by the aforementioned abiotic interactions, as well as biotic interactions that fluctuate throughout the day. In general, CO2 is the

preferred inorganic source of carbon for photosynthesis for aquatic species of plants

(Sand-Jensen et al. 1992, Raven 1970, King 1970), and it may be rapidly sequestered

from the water column during daylight hours (Talling 1976). High rates of

photosynthetic activity may culminate in a rise of pH during the day for freshwater

systems if the rate of CO2 removal exceeds replacement by respiration or diffusion from

the atmosphere and sediment (Pearl 1988). The relevance of carbon depletion is

evidenced in impaired water bodies by elevated pH conditions (> 9.0) as well as

increased diel oscillation of pH (Allan 1995, Maberly 1996, Pearl 1988).

Human alterations to the environment may influence the aforementioned

interactions through pollution, eutrophication, altered hydrology, changing land use

patterns, and the construction of impoundments. These changes may directly influence

pH by altering the biogeochemistry of a site or indirectly by affecting ecosystem

processes (e.g., primary production and respiration dynamics). Increased atmospheric

carbon dioxide has led to the acidification of some water bodies, and even has

connotations for altering the chemistry of the ocean (Doney et al. 2007). Global climate

6

may also influence the functionality of terrestrial plants, thus altering the availability and transport of anion and cations, as well as macro and micronutrients in the environment

(Paerl 1997, Doney et al. 2007, Bowling et al. 2008, Smith et al. 1999).

Scope of Dissertation

The focus of my dissertation is to further expand on an understanding of the importance of site-specific pH as a critical variable during ecological risk assessments of ionizable compounds. The first chapter provides background information that places the importance of pH in the context of ecological risk assessment and briefly summarizes the factors that contribute to spatiotemporal variability in surface water pH. The second chapter details the pH-dependent toxicity of a model pharmaceutical sertraline and then relates how seasonal differences in site-specific pH may alter risk characterization. The third chapter delves more deeply into potential alternative endpoints in fish associated with exposure to sertraline using concepts traditionally used to infer behavioral changes in mammals during drug development as well as automated digital tracking technologies.

The fourth chapter details a potential mechanistic explanation for the pH-dependent toxicity of toxins released by Prymnesium parvum associated with differences in ionization state. The fifth chapter explores diel oscillation of pH for stream sites that represent a nutrient gradient and then relates how these differences may influence national ambient water quality criteria for both traditional and emerging contaminants of concern.

7

CHAPTER TWO

Aquatic Toxicity of Sertraline to Pimephales promelas at Environmentally Relevant Surface Water pH

Introduction

NOTE: Chapter two is published in Environmental Toxicology and Chemistry (2009) 28: 2685-2694. Please refer to Appendix B for the licensing agreement.

For over 50 years researchers have recognized that ionization state may alter the

biological activity of xenobiotics (Simon and Beevers 1951, Simon and Beevers 1952 a +

b, Blackman and Robertson-Cunningham 1953). This observation is attributed to the unionized form crossing cellular membranes more readily than the dissociated moiety

due to its lower polarity; hence, it is if often regarded as being more bioavailable for

uptake. It is important to consider ionization state during both prospective and retrospective risk assessment because ambient toxicity of some contaminants to aquatic life may vary depending on site-specific pH. Spatiotemporal variability in surface water pH may be a reflection of differences in geomorphology, hydrology, climate, and/or other human disturbances. The relevance of these differences to risk assessment is emphasized by the integration of site-specific pH adjustment factors into national ambient water quality criteria in the United States for ammonia (NH3) (US EPA 1985, 1999) and

pentachlorophenol (PCP) (US EPA 1986, 1996). Both the weak base NH3 (pKa = 9.3) and the weak acid PCP (pKa = 4.7) are more toxic to aquatic life when they occur in the

environment predominately as the unionized forms. To account for these differences, pH

adjustment factors have been created by relating site-specific pH to laboratory derived

8

toxicological data from experiments completed with the compound of interest at various

pH levels. Site-specific pH consideration can result in markedly different acceptable

loads in receiving systems, varying by 13-fold for NH3 and 60-fold difference for PCP

between sites with contrasting surface water pH values of 6 and 9. Although it is likely

that similar mechanisms may influence the behavior of select pharmaceuticals in aquatic systems, ecological risk assessments rarely consider ionization state during exposure and

effects analysis.

Many pharmaceuticals are implicitly designed as ionizable compounds to ensure

that active components of administered doses reach specific target locations within the

body. As well as being detected in surface waters, pharmaceuticals have been

demonstrated to bioaccumulate in fish (Brown et al. 2007, Brooks et al. 2005, Ramirez et

al. 2009). The pharmaco*kinetic properties and partitioning behavior of pharmaceuticals

are often well understood due to requirements for drug development. Although this

information may be extrapolated or modeled to predict environmental fate, it is far more

challenging to infer biological effect because other pertinent data is often lacking. For

example, uptake rates of some pharmaceuticals and personal care products (PPCPs) will

inevitably be influenced by their ionization state as well as simultaneously by other

factors, such as the presence of cations and anions, temperature, and species-specific

characteristics. The partitioning behavior of drugs is important for ecological risk

assessment because exchange between the water column and gills of aquatic organisms

will ultimately influence steady-state plasma concentrations, which has direct

connotations to the potential culmination of adverse effects associated with exposure.

9

An increasing body of literature suggests that pharmaceuticals may exert sublethal effects in wildlife due to the evolutionary conservation of human drug targets

(Gunnarsson et al. 2008, Kreke and Dietrich 2008). Despite identifying analogous drug targets in humans and other species, it is uncertain whether exposure to organisms will result in the same magnitude or type of effects as those experienced by humans because little is known about potential interspecific differences in uptake, disposition, binding affinity, and functional responses. As pharmaceuticals are being tailored with greater specificity for certain biological targets to improve efficacy at relatively lower concentrations, there is increased concern that environmental exposure may cause adverse effects to non-target species that have similar drug receptors (Kreke and Dietrich

2008). These concerns are further heightened for aquatic organisms because xenobiotics absorbed across the gills directly enter the circulatory system and are therefore not susceptible to first-pass metabolism.

Sertraline was selected as a model compound to investigate how differences in exposure pH may influence ecotoxicological endpoints. Sertraline is a selective serotonin reuptake inhibitor (SSRI) that is being prescribed more prevalently to treat depression and other diseases (Minagh et al. 2009). There is a robust international market for the drug as evidenced by yearly human consumption rates of 122, 157, and 76 g/1000 inhabitants in Denmark, Norway, and Finland, respectively (Christensen et al. 2007).

Sertraline was a useful model for the present study because it has a pKa value of 9.47.

Therefore, its ionization state will change markedly between pH 6 and 9, which is a representative range of environmentally relevant surface water pH. As well as being identified in sewage discharge and effluent dominated streams (Schultz and Furlong

10

2008), Brooks et al. (2005) detected sertraline and its primary metabolite

desmethylsertraline in the muscle, liver, and brain of several fish species. In addition,

Ramirez et al. (2010) reported sertraline as the pharmaceutical consistently displaying the

greatest maximum concentration in both fish tissue fillets and livers during a U.S.

national pilot study of PPCPs in fish tissue.

Such observations are important because sertraline elicits its effects on the

serotonergic system, which is a highly conserved in the animal kingdom (Gunnarsson et

al. 2008, Gould et al. 2007, Smith 1999). In addition to laboratory experiments assessing

the toxicity of sertraline at various pH values, continuously monitored quality assured

instream pH data for two stream sites in the Brazos River Basin, Texas, USA was

obtained for a three year period and then evaluated relative to laboratory derived pH-

dependent toxicity relationship for Pimephales promelas using both discrete and

probabilistic approaches.

Materials and Methods

Experimental Conditions

All experiments were conducted at 25 ± 1º C and had photoperiods of 16:8 h

light: dark. Reconstituted hard water (RHW), as described in standard methods (APHA,

AWWA, WEF 1998), was used as the dilution water and control. Dissolved oxygen and

conductivity were measured using an YSI Model 55 (Yellow Springs, OH, USA)

handheld dissolved oxygen meter and salinity, conductivity, and temperature on an YSI

Model 30 handheld system. An Orion 720A plus pH/ISE meter (Beverly, MA, USA)

-1 was used to determine pH. Alkalinity (mg L as CaCO3) and hardness were measured

11

amperiometrically and by colorimetric titration, respectively, according to standard methods (APHA, AWWA, WEF 1998). Mean (± standard deviation (SD)) water quality parameters for RHW at test initiation were within acceptable ranges described by guidelines (APHA, AWWA, WEF 1998, US EPA 2002) for all parameters, except pH, as dissolved oxygen = 7.7 (± 1.3) mg L-1, specific conductance = 583 (± 8.6) μS cm-1,

-1 -1 hardness = 180 (± 5.2) mg L as CaCO3, and alkalinity = 116 (± 4.2) mg L as CaCO3.

Analytical grade nitric acid and sodium hydroxide (VWR Scientific, West Chester, PA,

USA) were added to adjust the pH of RHW to desired levels. The pH of exposure water at test initiation were within 0.05 units of the desired level, and drifted by less than 0.5 during any experiment. Sertraline hydrochloride (Ranbaxy Laboratories, New Delhi,

India) was used to prepare exposure concentrations, which were based on data from preliminary range finding toxicity experiments.

Pimephales promelas Bioassays

Acute studies Experiments with P. promelas were completed concurrently at pH

6.5, 7.5, and 8.5 on three separate occasions according to slightly modified U.S.

Environmental Protection Agency (U.S. EPA) protocol (2002a). Each experimental unit consisted of 10 organisms in a 600-ml glass beaker filled with 500 ml of exposure water.

To minimize pH drift, experimental units contained a volume of exposure water greater than recommended and were covered with Parafilm (Pechiney Plastic Packaging,

Chicago, IL, USA). Four replicates were prepared for the control and six sertraline concentrations, including 30, 60, 120, 250, 500, and 1000 μg sertraline L-1. Test solution for each concentration was first prepared, then divided into three aliquots, and each

12

aliquot was adjusted to the correct pH. This approach was taken to ensure that sertraline

concentrations were the same across the three pH treatments. All P. promelas used in experiments were <48 h old. Individuals were fed newly hatched brine shrimp (Artemia sp.) 2 h prior to the exposure, but were not fed during experiments. Survival was assessed at 24 and 48 h. A time-to-death study, separate from the aforementioned experiments, was also completed. Individuals were exposed to either control (RHW) or

500 μg sertraline L-1 treatments adjusted to pH 6.5, 7.5, and 8.5. Four replicates of 10 individuals were prepared for each treatment. Survivorship was assessed after 2, 4, 6, 8,

12, 16, 20, 24, 36, and 48 h. Water quality parameters were collected at test initiation and completion for all experiments.

Short-term chronic studies A 7-d short-term chronic experiment was conducted according to slightly modified U.S. EPA protocol (2002b). Five sertraline concentrations, plus a RHW control, were prepared for each pH level. After observing markedly different acute endpoints for the three pH treatments, we choose to use lower concentrations of sertraline for the higher pH treatments. At pH 6.5 concentrations were

60, 120, 250, 500, and 1000 μg sertraline L-1 , at pH 7.5 concentrations were 30, 60, 120,

250, and 500 μg sertraline L-1, and at pH 8.5 concentrations were 15, 30, 60, 120, and

250 μg sertraline L-1. Several concentrations overlapped between the different pH treatments so that comparisons could easily be made; however, we hoped that these concentration series would enable us to capture both effect and no-effect concentrations at pH 6.5, 7.5, and 8.5. Each treatment level had four replicates of 10 individuals and experiments were completed in 600-ml glass beaker filled with 500 ml of exposure water.

Beakers were covered with Parafilm, and exposure water was renewed daily.

13

Survivorship was monitored daily and individuals were fed Artemia sp. twice daily

(AM/PM). After 7 d of exposure, survivorship was assessed and individuals from a

replicate were pooled minus three individuals used for the feeding behavior trials (see below). The individuals were transferred to previously weighted drying dishes after being euthanized according to Baylor Animal Care Policy and then placed in an oven set

at 80º C for 48 h. After drying the fish the weights of the drying dishes were again

determined on a Mettler Toledo Model MX5 microbalance (Columbus, OH, USA).

Average growth was determined by subtracting the initial drying dish weight from the

final drying dish weight containing fish divided by the number of fish.

Feeding behavior. Feeding behavior trials were completed with individuals from

treatments that had a sufficient number of survivors after the 7-d exposure. These

included the control, 60, 120, and 250 μg sertraline L-1 at pH 6.5, control, 30, 60, and 120

μg sertraline L-1 at pH 7.5, and control, 15, and 30 μg sertraline L-1 at pH 8.5.

Experiments were completed based on an approach by Stanley et al. (2007) with slight

modifications. Three fish from each replicate were randomly selected and isolated

individually in 100 ml glass beakers filled with 100 ml of RHW; thus, a total of twelve

fish per treatment level were examined. Food was withheld from fish for 24 h prior to

feeding trials. Experiments were initiated by introducing 40 Artemia sp. nauplii to an individual fish. Each fish was allowed 15-min to feed, after which time it was removed and the total number of remaining nauplii was enumerated to determine rates of feeding

(Artemia minute-1). Consumption rates of three individuals for a replicate were

determined and then pooled to calculate an overall replicate average, which was used for

14

subsequent statistical analyses (n = 4 per treatment). Individuals from a replicate were pooled to avoid concerns of pseudo-replication.

Statistical Analyses

Statistical significance of response variables was assigned at α = 0.05 for all tests.

The 50% lethal concentrations (LC50) values were calculated using Toxstat. The probit

method was used if data met assumptions; otherwise, the Trimmed Spearman-Karber

method was applied (2002a). Average LC50 values for the three pH ranges were

compared with analysis of variance using the computer program Jumpin version 5.0

(SAS institute, Cary, NC, USA). The 50% time-to-death (LT50) values were calculated

based on equations derived from best-fit line models.

No-observable-adverse effect concentration (NOAEC) and lowest-observable-

adverse effect concentrations (LOAEC) were calculated based on the statistical approach

described by standard protocols (2002b). Effect concentration values (EC10, EC25,

EC50) for chronic experiments were obtained using a reparameterized logistic three

parameter model in Sigma Plot (Systat Software, San Jose, CA, USA) according to Brain

et al. (2005) based on Stephenson et al. (2000).

Analytical Methods

Nominal concentrations of sertraline were verified for one acute experiment and

for select concentrations on days 1, 3, and 6 for the 7-d sub-chronic experiment. A 500

ml sample of the appropriate test solutions was adjusted to pH 4 using high-performance

liquid chromatography grade nitric acid and then run through a 90 cc C-18 solid phase

extraction column (SPE) (Waters, Milford, MA, USA). Before adding the sample, the

15

column was activated by first drawing through 10 ml of nanopure water that was adjusted

to pH 4, followed by 10 ml of high-performance liquid chromatography grade methanol,

and then a final nanopure rinse. The SPE were stored at -20° C until analyses were

completed. Just prior to analyses the SPE cartridges were eluted with 10 ml of methanol

in 10 ml disposable borosilicate glass culture tubes (VWR Scientific, West Chester, PA,

USA). The eluted samples were diluted 10-fold in methanol to reduce the sertraline

concentrations to levels quantifiable by instrumentation. Depending on the expected

concentration in the samples, different aliquots were taken from this last dilution and

were brought into a final volume of 1000µl. Prior to this, a 50 µl of sertraline-d3 5.5 ppm

(internal standard [IS]) was added; resulting in a final concentration of 275 ng L-1.

A LC-MS/MS consisting of a Varian ProStar model 210 binary pump equipped with a model 410 autosampler (Varian, Palo Alto, CA, USA) was used in the present study. Sertraline was analyzed on a 15 cm x 2.1 mm (5µm, 80 Å) Extend-C18 column. A binary isocratic elution consisting of 30% of formic acid 0.1% (v/v) in water and 70% of methanol was employed to achieve analyses in 5 min. Additional chromatographic parameters included an injection volume of 10 µl and a flow rate of 350 µl/min. Eluted analyte and isotope were monitored by MS/MS using a Varian model 1200L triple- quadrupole mass analyzer equipped with an electrospray interface (ESI). To determine the best ionization mode (ESI + or -) and optimal MS/MS transitions for the target analyte and isotope, each compound was infused individually into the mass spectrometer at a concentration of 1 µg/ml in aqueous 0.1% (v/v) formic acid at a flow rate of 10

µl/min. Compounds were initially tested using both positive and negative ionization modes while the first quadrupole was scanned from m/z 50 to [M + 100]. This enabled

16

identification of the optimal source polarity and most intense precursor ion for each

compound. Once these parameters were defined, the energy at the collision cell was

varied, while the third quadrupole was scanned to identify and optimize the intensity of

product ions for each compound. Additional instrumental parameters held constant for all

analytes were as follows: nebulizing gas, N2 at 60 psi; drying gas, N2 at 19 psi;

temperature, 300 oC; needle voltage, 5000 V ESI+, 4500 V ESI-; declustering potential,

40 V; collision gas, argon at 2.0 mTorr.

A solution of IS was added to each calibration point at a concentration of 275 ng

L-1 to generate a relative response ratio. Analyte concentrations determined using an

internal standard calibration procedure. The response factor was calculated by dividing

the peak area for sertraline by the peak area for the IS, and a calibration curve was

prepared by plotting a linear regression (r 2 ≥ 0.999) of the response factor versus analyte

concentration for all calibrators analyzed. Instrument calibration was monitored through

the use of continuing calibration verification samples with an acceptability criterion of

±20%. In a given run, one blank and one continuing calibration verification sample were

interspersed between every five samples for quality assurance purposes. Recoveries of

the IS were compared with the relative response ratio and a concentration for the

unlabeled analyte was calculated.

Continuous Water Quality Monitoring Data

Daily pH data from 2003 to 2005 was obtained for two wadeable streams, CAMS

703 (Gatesville Water Site) and CAMS 704 (Resley Water Site), from the Texas

Commission of Environmental Quality (TCEQ) database of surface water quality with

quality assurance, quality control review (TCEQ 2008);

17

www.tceq.state.tx.us/compliance/monitoring/water/quality/data/wqm/swqm_realtime_sw

f.html.). For continuous water quality data to be considered valid by TCEQ, pre- calibration and post-calibration information must be provided for multiparameter probes.

Nutrient enrichment is a concern at both sites as surrounding watersheds are largely

agricultural. Resley Creek is influenced by dairy confined animal feeding operations,

occurs at higher elevation, and drains into the Leon River, which is the system monitored

at Gatesville, Texas, USA. The Gatesville site is located in close proximity to urban development and subsequently receives municipal wastewater discharge and storm water runoff. These sites were selected due to their close spatial proximity to examine how factors other than geomorphology and climate may influence surface water pH. Data flagged by the TCEQ as being invalid during a quality review were removed prior to analysis. Daily as well as seasonal means for the three year period were determined.

Cumulative frequency plots of daily instream pH measurements were created so that the likelihood of observing a given pH value at each site could be quantified for both overall as well as seasonal data.

Predictions of Ecotoxicological Responses

The LC50 values for acute experiments completed at different pH were plotted with log LC50 values as the response y-axis and pH as the predictor x-axis. A linear model was fit to the data so that LC50 values could be predicted for any pH. Relating these toxicological relationships with cumulative frequency distributions of pH by season allowed us to examine the likelihood of observing a toxicological response. The lack of analytical data for sertraline precluded specific risk characterization of sertraline at these

sites. Alternatively, risk was assessed by examining how variability in site-specific pH

18

may change predicted toxicological responses. For comparative purposes, a range of potential sertraline concentrations at these sites was examined so that the likelihood of observing adverse effects could be set on a relative scale. The concentrations selected for comparative purposes included 125, 150, 175, and 200 μg sertraline L-1 for acute

bioassays, while those for growth and feeding were 50, 75, 100, 125, and 150 μg

sertraline L-1 and 20, 30, and 40 μg sertraline L-1, respectively. These values were

selected because they spanned the lower ranges of exposure concentrations and could be

used to examine differences in potential effects among the three pH treatments.

Additional safety factors were not considered during the analysis. The present study

solely focused on how site-specific pH may influence biological responses, and thus was

concerned primarily with analysis of effects. The scarcity of reported sertraline

concentrations in the environment prevented further investigation of how site-specific pH

may influence environmental exposure scenarios.

Results

Analytical Results

Nominal concentrations of sertraline in exposure water were analytically

confirmed and values were 96 to 118% of their target concentration (Table 1).

Consequently all subsequent data analyses in this study were performed using nominal

treatment concentrations because of the similarities with analytically verified values.

Acute Studies

Survivorship was >95% in control treatments, which consisted of RHW adjusted

to pH 6.5, 7.5, and 8.5, respectively. Dose dependent responses were apparent as

19

survivorship was lower at higher sertraline concentrations. Mean LC50 values for

experiments completed at pH 6.5, 7.5, and 8.5 were significantly different based on

analysis of variance (p<0.05) (Figure. 2); and mortality was generally greater at lower

sertraline concentrations when individuals were exposed to test water with higher pH.

Table 1. Analytical verification of nominal sertraline treatment levels for acute and short-term chronic experiments with Pimephales promelas. n Nominal concentration Measured ± SD (µg/L)a % Recovery ± SD 4 control (0) 0.02 NAb 4 15 14 + 6 96 + 38 4 30 30 + 4 100 + 13 4 60 61 + 5 101 + 8 4 125 141 + 24 118 + 20 4 250 256 + 26 102 + 10 4 500 579 + 44 116 + 9 4 1000 1120 + 69 112 + 7 1 2000 2111 106 a SD = standard deviation. b Not applicable

3.0 g/L) g/L) μ 2.8

2.6

LC50 value ( value LC50 2.4

2.2

2.0

1.8 Pimephales promelas Pimephales

log 1.6 6789 pH treatment levels

Figure 2. 48-h median lethal concentration (LC50) values for Pimephales promelas exposed to sertraline at three pH treatment levels (6.5, 7.5, 8.5). Each data point represents a LC50 value from an independent test. A linear regression model relating LC50 values to pH is also presented.

20

Results from the time-to-death study showed similar trends as the onset of mortality

occurred more rapidly at higher pH (Figure. 3). The LT50 values for individuals exposed

to 500 µg sertraline L-1 at pH 6.5, 7.5, and 8.5 were >48, 32, and 5 h, respectively. No

mortalities were observed in the time-to-death study in control treatments adjusted to

either pH 6.5, 7.5, and 8.5.

100

80

Survival (%) Survival 60

40 pH 6.5 pH 7.5 20 pH 8.5 Pimephales promelas Pimephales

0 0 10203040 TIme (h) Figure 3. Time-to-death for juvenile Pimephales promelas exposed to 500 μg sertraline L-1 in reconstituted hard water adjusted to three different pHs. No mortality was observed in control treatments at pH 6.5, 7.5, and 8.5

Short-term Chronic Experiments

Control survivorship was ≥ 90% and dose-dependent responses were apparent in all

experiments (Table 2). Similar to the relationship between pH levels and LC50 values,

growth and feeding EC10 values were markedly lower at higher pH (Table 2). Feeding rate was generally more sensitive than growth and survivorship. Survivorship, growth,

and feeding endpoints were consistently lower for individuals exposed at pH 8.5, relative

21

to those exposed to lower pH treatment levels (Table 2). At pH 8.5 a NOAEC could not be determined because the lowest test concentration caused a significant reduction in feeding rate relative to the control. The NOAEC and EC10 for feeding rate at pH 8.5 were also substantially lower than similar values for growth (Table 2).

Continuous Water Quality Monitoring Data

The respective overall site pH average (± SD) for Gatesville and Resley from the period of 2003 through 2005 were 7.97 (±0.2, n=944) and 7.66 (±0.28, n=1033). The frequencies of daily averages pH for each site by season are depicted in Figure 4.

Distinctions between seasonal distributions and means are apparent. The respective fall, winter, spring, summer daily means (±SD) for the various sites are presented in Table 3.

Also presented in Table 3 are attributes describing the pH distributions of the two sites by season. Daily mean pH values were generally highest during winter and lowest in the summer. Variability in pH was greater for the Resley site than Gatesville site based on differences in standard deviation between and within seasons (Table 3).

Predictions of Ecotoxicological Responses

The LC50 predictions based on discrete overall site pH means were 146 and 204

μg L-1 for Gatesville and Resley, respectively. Seasonal differences were evident for both sites and predictions of toxicity to juvenile fathead minnows were greatest during winter months and the least during summer as respective values for Gatesville were 132 and 156

μg L-1, while those for Resley were 172 and 240 μg L-1. Overall, contrasting the temporal change by calculating the absolute percent difference in the mean predicted LC50 values

22

for Resley pH distributions resulted in a total of 33.5%; however, total change at

Gatesville was only 16%.

Table 3. The average pH by season for two Texas Commission of Environmental Quality continuous water quality monitoring stations (Gatesville, Resley) in the Brazos Watershed (TX, USA). The attributes of the line used to infer the likelihood of a specific pH for each site are also summarized. SD = standard deviation

Attributes of line fit to pH distributions Average pH Site Season (SD) Intercept Slope r ² Gatesville Fall 7.98 (0.22) 88.95 -80.00 0.994 Winter 8.06 (0.15) 144.77 -130.78 0.994 Spring 7.94 (0.16) 96.70 -86.69 0.942 Summer 7.91 (0.23) 86.24 -77.16 0.991 All 7.97 (0.2) 93.39 -83.81 0.991 Resley Fall 7.57 (0.2) 78.38 -68.62 0.981 Winter 7.82 (0.26) 58.63 -52.00 0.917 Spring 7.73 (0.28) 66.91 -59.26 0.990 Summer 7.51 (0.27) 71.62 -62.53 0.988 All 7.66 (0.28) 65.40 -57.58 0.999

The pH distributions allowed a probabilistic approach that enabled better

estimations of seasonal variability and differences between sites. The span of sertraline

concentrations used corresponded to respective pH of 8.09, 7.92, 7.78, and 7.66 based on

our LC50 pH-dependent toxicity relationship. The respective pH values based on our

EC10 model were 8.13, 7.84, 7.63, 7.47, and 7.33. Predictions based on daily average pH distributions suggested trends similar to those described from discrete predictions.

Table 4 depicts the likelihood that a LC50 value is predicted to be less than the respective

comparative value. For all seasons combined, LC50 values at Gatesville are predicted to

be <125 μg L-1 10% of the time, <150 μg L-1 46% of the time, <175 μg L-1 82% of the time, and <200 μg L-1 97% of the time. Analogous comparisons for the same

23

comparative values at Resley are less indicative of ambient toxicity as respective values

are 2, 14, 40, and 68% (Table 4). Seasonal differences were also apparent. For example,

an LC50 value is predicted to be less than 150 μg L-1 69% of the time during the winter at

Gatesville, but only 37% of the time during the spring. Another example would be that

an LC50 value is predicted to be less than 150 μg L-1 23% of the time during winter for

Resley, but only 3% during spring months (Table 4). Similar trends were also apparent for comparisons based on growth and feeding EC10 values (Table 4).

Discussion

In this study the sensitivity of P. promelas to sertraline changed markedly over a gradient of environmentally relevant surface water pH. Adverse effects were observed at lower concentrations for individuals exposed at pH 8.5 than for those exposed to the lower pH treatments. Furthermore, individuals in control treatments with water adjusted to either pH 6.5, 7.5, or 8.5 had average survivorship ≥90% in both the acute and short- term chronic bioassays. Neither growth or feeding rate were significantly different among pH levels for the sub-chronic bioassay as respective values ranged from 0.36 to

0.41 mg per surviving individual and 1.7 to 1.9 Artemia min-1, respectively (Table 2).

High survivorship and similar sub-lethal measures for controls in the three pH treatments suggests that differences in sensitivities between treatments containing sertraline was not directly due to physiological stress experienced by organisms as a result of pH adjustment approaches, which followed established U.S. EPA methods. Rather, the observed variability in sensitivities at the different pH levels likely resulted from changes in the physiochemical properties of sertraline as a consequence of different pH environments.

24

Table 2. Results of short-term chronic experiments with Pimephales promelas exposed to sertraline at pH 6.5, 7.5, and 8.5. The average survivorship, growth, and feeding rate for the various sertraline concentrations at different pH are reported and accompanying standard deviations are shown in parentheses. (*) treatments that are statistically significantly different from the control (p=0.05). The no-observable-adverse effect concentration (NOAEC) and lowest-observable-adverse effect concentrations (LOAEC) are shown respectively by A and B for each pH treatment.

pH Treatment Survivorship Growth EC10 EC25 EC50 Feeding rate EC10 EC25 EC50 -1 (μg/L) (%) (mg) (Artemia min ) 6.5 Control 90 (8) 0.37 (0.05) 469 496.4 544.4 1.9 (0.2) 69.6 106.9 199.7 60 88 (13) 0.36 (0.05) 1.7 (0.2) 120 A 98 (5) 0.34 (0.07) 1.3 (0.2) 250 B 83 (17) 0.26 (0.04)* 0.8 (03) * 500 50 (14) * 0 (0) * NA

25 1000 0 (0) * 0 (0) * NA 7.5 Control 93 (10) 0.41 (0.04) 118.7 124.6 131.4 1.7 (0.2) 65.6 99 149.5 30 90 (8) 0.4 (0.05) 1.7 (0.3) 60 A 95 (6) 0.42 (0.02) 1.6 (0.1) 120 B 88 (13) 0.3 (0.06) * 1.1 (0.1) * 250 0 (0) * 0 (0) * NA 500 0 (0) * 0 (0) * NA 8.5 Control 90 (8) 0.36 (0.06) 30.3 39 50 1.9 (0.1) 8.7 20.7 80.3 15 B 90 (8) 0.37 (0.04) 1.5 (0.1) * 30 75 (13) * 0.33 (0.06) * 1.3 (0.2) * 60 20 (18) * 0.11 (0.11) * NA 120 0 (0) * 0 (0) * NA 250 0 (0) * 0 (0) * NA EC = effective concentration; NA = not applicable.

a) b) 100

50

6.8 6.9 7 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.88 7.96.86.97 8 8.1 8.2 7.17.27.37.47.57.67.77.87.98 8.3 8.4 8.18.28.38.4

pH pH

26 c) d) 100

No. of observations of No. 50

6.8 6.9 7 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.88 7.96.86.97 8 8.1 8.27.17.27.37.47.57.67.77.87.98 8.3 8.4 8.18.28.38.4 pH

Figure 4. Seasonal frequency distributions of pH by the number of observations at two continuous monitoring stations, Gatesville („) and Resley ( ), in the Brazos River Basis, Texas, USA. Data from 2003 to 2006 was obtained from the Texas Commission on Environmental Quality Surface Water Quality Monitoring Information System.

Table 4. The likelihood that the predicted sertraline median lethal concentration (LC50) and growth and feeding 10% effective concentration (EC10) values for Pimephales promelas will be below the comparative value at two sites based on aquatic toxicological models derived during laboratory bioassays and distributions of surface water pH. -1 Site Endpoint Comparative sertraline values (μg) L Fall Winter Spring Summer All LC50 125 14 11 7 7 9 150 62 86 52 46 55 175 94 100 92 86 93 200 100 100 100 98 100 Gatesville Growth EC10 50 17 17 9 9 12 75 67 91 58 51 61 100 93 100 91 85 92 125 99 100 99 97 99 150 100 100 100 99 100 Feeding EC10 20 35 49 24 22 28 30 94 100 92 86 93

27 40 100 100 100 99 100 LC50 125 0 7 4 0 2 150 5 30 25 5 16 175 30 62 60 26 46 200 67 84 86 60 76 Resley Growth EC10 50 0 9 5 0 3 75 7 3428718 100 29 60 58 25 45 125 56 79 80 50 68 150 79 90 92 72 84 Feeding EC10 20 1 171216 30 30 62 60 26 46 40 78 17 91 70 83

Pimephales promelas was specifically selected as a model organism for these experiments because of its ability to tolerate a wide range of environmental conditions.

Researchers have previously utilized P. promelas in studies over pH gradients similar to that used in our experiments; examples include development of national ambient water quality criteria for NH3 and PCP (US EPA 1985, 1986, 1995, 1999), as well as U.S. EPA toxicity identification evaluation approaches that employ pH adjustments.

Various studies report that ionization state may influence the toxicity of some xenobiotics to organisms. The difference in toxicity between the ionized and unionized moiety can be attributed to variable rates of diffusion into cells, which are determined by the external concentration and the pH of the exposure water (Simon and Beevers 1951,

Simon and Beeves 1952 a + b, Blackman and Robertson-Cunningham 1953, US EPA

1985, Sarrikoski et al. 1986, Sarrikoski and Vilukesela 1981, Whitley 1968). Results from the time-to-death experiment support the hypothesis that differences in ionization state of sertraline may explain aquatic toxicological responses across pH gradients. The unionized moiety is often assumed to be absorbed and delivered to target sites at a higher rate than the ionized from. Therefore, although aquatic organisms may be exposed to the same concentrations, adverse effects should be observed more rapidly in treatments with higher proportions of the unionized moiety; such a relationship was observed during the time-to-death study with fathead minnows (Figure 3).

Nakamura et al. (2008) reported similar pH-dependent toxicological relationships during experiments in which Japanese medaka (Oryzias latipes) were exposed to the

SSRI fluoxetine, a weak base with a pKa of 10.1, at pH 7, 8, and 9 as respective LC50 values were 5500, 1300, and 200 μg L-l. These researchers were able to ascertain that the

28

heightened acute toxicological response at the higher pH was likely due to a greater

bioconcentration factor at increased pH as rates varied markedly over the pH gradient.

Fisher et al. (1999) exposed zebra mussels (Dreissena polymorpha) to PCP and observed

significantly different uptake rates at various pH and temperatures that ranged from 0.33

to 2133 L kg-l d-l, and noted that lethal body resides varied at different pH and

temperature levels, as respective values for pH 6.5, 7.5, and 8.5 were 69, 245, and 782

µmol kg-l at 25 °C and 863, 1351, and 6253 µmol kg-l at 17 °C. Again, these results

suggest that the neutral, unionized moiety is more lipophilic, and that uptake by

organisms is favored when this form is more prevalent.

Because the pH and total sertraline concentrations at which the exposures

occurred were known, we used the Henderson-Hasselbalch equation to determine

ecotoxicological endpoints in terms of percent unionized sertraline. At pH 6.5, 7.5, and

8.5 approximately 5, 12, and 28% of the total sertraline should be present in the

unionized form. Examining LC50 values in terms of total sertraline by percent unionized

in the exposure water produces respective values of 32, 25, and 19 μg L-l unionized

sertraline for pH 6.5, 7.5, and 8.5 (Figure 5).

Similarly, growth EC10 values for % unionized were estimated to be 23, 14, and

8.5 μg L-l for pH 6.5, 7.5, and 8.5, respectively. The magnitude of difference in

endpoints is substantially less when comparing responses across pH treatment levels in terms of percent unionized concentrations compared to total sertraline concentrations.

For example, there is only a 1.5-fold difference in endpoints based on unionized concentrations compared to nearly a 9-fold difference in values based on total concentrations for acute experiments completed at pH 6.5 and 8.5.

29

Figure 5. The 48-h mean (±standard deviation) median lethal concentration (LC50) values for Pimephales promelas presented as total and percent unionized sertraline concentrations. Experiments were completed in triplicate for each pH treatment. The % percent unionized ( ) was determined by integrating total sertraline („) concentrations and exposure pH into the Henderson-Hasselbalch equation.

Further, there was a 15-fold difference between growth EC10 at pH 6.5 and 8.5

for total sertraline, but only a 3-fold difference when comparisons are made in terms of

unionized concentrations. Nakamura et al. (2008) observed approximately a 28-fold

difference in acute endpoints based on total fluoxetine between pH 7 and 9, which

equates to less than a 5-fold difference if comparisons are made based on unionized

levels of fluoxetine. In experiments examining the toxicity of the ionizable weak acid

triclosan to Ceriodaphnia dubia over different pH ranges, Orvos et al. (2002) calculated

acute endpoints based on total triclosan concentrations that were 4-fold lower at pH 6.8 to

7.0 compared to pH 8.2 to 8.5; however, when calculations were based on unionized

concentrations values were 110 and 140 μg L-1, respectively.

30

Although predicted LC50 values at different pH in terms of unionized concentrations results in less deviation than similar comparison in terms of total concentrations, toxicity may not solely be explained by exposure to the unionized moiety.

Saarikoski et al. (1986) concluded that although the rate of absorption is exponentially related to pH in experiments with phenolic and carboxylic acid, the overall rate over a pH gradient is not indicative of what would be predicted based solely on the unionized form in water; suggesting additional mechanism besides internal-external equilibrium of the neutral entity. Erickson et al. (2006) postulated explanations for this discrepancy in fish models, including that the 1) release of excretory products from gills may alter the pH of the area where cells come in immediate contact with contaminant, 2) ionized molecules may be absorbed to support chemical flux and amiable diffusion gradients, and 3) ionized entities are not completely impermeable to uptake across membrane barriers. The latter two suggestions are recognized by site-specific national ambient water quality criteria values for NH3 in the U.S., which are determined by concentrations of both the ionized and unionized moiety.

Behavioral Responses to SSRIs

Gerhardt (2007) suggested that behavior is among the most sensitive measures of impairment by stressors and thus an important consideration in ecotoxicology. Brooks et al. (2003) described the merit of augmenting traditional standardized ecotoxicological responses with alternative endpoints, particularly for SSRIs because they may result in behavioral modifications. Kreke and Dietrich (2008) further identified that standardized endpoints may be insufficient for ecological risk assessments of SSRIs. In the present study, feeding behavior was more sensitive than standardized mortality and growth

31

responses and the magnitude of survival, growth and behavioral toxicity was influenced

by pH gradients. Stanley et al. (2007) also observed feeding behavior in the fathead

minnow model to represent a more sensitive and ecologically relevant response than standardized endpoints (e.g., survival, growth) to the SSRI fluoxetine. Previous studies have reported that behavioral responses can be 10 to 100 times more sensitive than standard toxicological benchmarks, such as survivorship (Gernhardt 2007).

Several factors make feeding behavior a reasonable choice as an alternative endpoint for assessing SSRI exposure; the foremost being the connection between observed effects and potential ecological consequences. Maltby (1999) noted that

lowered energy input associated with reduced feeding as a consequence of environmental exposure may be integrated into energy budget models to predict effects at the population level. Furthermore, a plausible cause and effect relationship between reducing feeding behavior and exposure to sertraline can be concluded based on our current understanding

of its mode-of-action in mammalian models. The desired human therapeutic effect of

sertraline is to inhibit the reuptake of serotonin into presynaptic terminals of the central

nervous system (MacQueen et al. 2001). Sertraline blocks the reuptake carrier that

normally removes serotonin released from the synapse. Consequently, extracellular

levels of serotonin become elevated in mammals when administered sertraline. In

addition, sertraline also blocks the reuptake of the carrier on the cell body, which

subsequently leads to activations of serotonin autoreceptors that negatively modulate

firing rates (Sprouse et al. 1996).

Previous studies have reported the detection of sertraline and its primary

metabolite desmethylsertraline, in muscle, liver, and brain tissues in fish residing in an

32

effluent dominated stream (Brooks et al. 2005); which likely represent worst case scenarios for aquatic exposure to SSRIs in the developed world (Brooks et al. 2006). In order to understand fish responses to SSRIs in these systems comparative pharmacology and read-across approaches can be very useful (Brooks et al. 2009, Ankley et al. 2007).

For example, Benmansour et al. (1999) demonstrated down regulation of the serotonin

transporter (SERT) in brains of rats treated for 21 d with sertraline. This observation is

critical because SERT downregulation is important in the antidepressant action of SSRIs,

resulting in swimming activity responses and other behavioral effects in rat models

(Holmes et al. 2003). More recently, Gould et al. (2007) reported similar SERT

downregulation in zebrafish brains treated for 21 d with sertraline. Such observations

and feeding inhibition of juvenile fathead minnows exposed to the SSRIs sertraline in the

present study and fluoxetine (Stanley et al. 2007) appear to provide multiple lines of

evidence that support a cause and effect relationship between exposure to sertraline and

target-mediated reduced feeding behavior in fish.

Although one may anticipate growth responses to be representative of feeding

rates, standard test designs may preclude this assumption in the case of SSRIs. Reducing

feeding rates may be the function of fewer strikes or lower prey-capture efficiency, and

may not be apparent in growth data because of the longevity and abundance at which

food is available under typical static, renewal test conditions. The current traditional

experimental design described by U.S. EPA (2002b) is advantageous because it may

improve statistical power in subsequent analyses by reducing variability among and

between replicates; however, it may be limiting as it may not allow for detection of sub-

lethal effects that are more subtle, such as strike rate and prey-capture efficiency. Thus,

33

feeding responses in the fathead minnow model presented here and by Stanley et al.

(2007) appear to provide an ecologically relevant measure of effect for fish exposed to

SSRIs. Alternative endpoints, such as feeding response, do not preclude the importance

of traditional endpoints described in standard protocol, but rather may provide additional

data for reducing uncertainty when characterizing environmental effects during risk assessment.

Importance of Site-Specific pH

Using distributions rather than discrete measurements of instream pH, such as overall site averages, affords additional insight and flexibility during both prospective

and retrospective risk assessment. Seasonal and overall site mean values are useful for

identifying broad distinctions at and among sites during risk assessment; however, such

measures often lack the resolution to reflect exposure scenarios to organisms during more

finite scales. Consequently, there is a growing impetus that pH should be approached as

a dynamic variable that is not accurately quantified with periodic discrete sampling

events. Several researchers describe pronounced diel fluctuations in pH and CO2 concentration that are attributable to photosynthetic and respiratory processes (Howland et al. 2000, Guasch et al. 1998, Rebsdorf et al. 1991) as well as changes in gaseous saturation potential for waters due to increases in temperature during the day. The magnitude of pH change at a site is dependent on abiotic and biotic factors; however, previous researchers have documented daily oscillations that exceed 0.5 pH units

(Guasch et al. 1998, Rebsdorf et al. 1991, Allan 1995).

In the present study variability in pH between sites and temporal pH differences at sites suggests potential differences in site-specific ecotoxicological responses to ionizable

34

contaminants, even in receiving systems located within the same subwatershed. It is reasonable to project that the magnitude of pH differences among receiving systems at

broader spatial scales will be more pronounced. Based on the present study findings in

the laboratory these changes may influence the dissociation of sertraline and potentially

other ionizable compounds in the environment. Thus, it appears relevant that site-

specific pH be considered in prospective ecological risk assessments of pharmaceuticals

because ambient pH can influence exposure and toxicity of sertraline and potentially other pharmaceuticals to aquatic life residing in receiving systems. Current ecological risk assessment approaches for pharmaceuticals and other contaminants may over- or under-predict instream effects for ionizable compounds if the pH of a receiving system is not considered.

35

CHAPTER THREE

Sublethal Effects of the Selective Serotonin Reuptake Inhibitor (SSRI) Sertraline on Fathead Minnow Under Potential Worst-Case Environmental Exposure Scenarios.

Introduction

Human pharmaceuticals and personal care productions (PPCPs) have been measured in the discharges of wastewater treatment plants (WWTP) and receiving systems (Koplin et al. 2002, Herberer 2002, Kummerer 2004, Vieno et al. 2005, Brooks et al. 2005, Nikolaou et al. 2007, Williams and Cook 2007). Concerns about the potential environment effects are further heightened by the quantification of drugs in the tissues of aquatic organisms (Larsson et al. 1999, Brooks et al. 2005, Brown et al. 2007, Ramirez et al. 2009, Fick et al. 2010). Selective serotonin reuptake inhibitors (SSRI) are one of the most commonly detected classes of drugs in effluents, and several studies have measured concentrations in the plasma (Citations) as well as the brain tissue (Brooks et al. 2005,

Ramirez et al. 2009) of fish collected below wastewater discharges. The SSRIs are medications widely prescribed to treat depression, anxiety, and other mental ailments

(Citations), and elicit therapeutic effects by preventing serotonin reuptake from synapses by pre-synaptic serotonin transporter proteins, which triggers increased serotonergic neurotransmission (Frazer 2001). The serotonergic receptors are well described in humans and rodents (Owens et al. 1997) and are largely conserved among vertebrates

(Dietl and Palacios 1988). Researchers have demonstrated occupancy of the serotonin transporter by SSRI during in vivo studies with mice models and emphasized that competition studies with radiolabelled ligands are useful for identifying binding affinity

36

of drugs (Scheffel et al. 1994). During in vitro studies, Gould et al. (2007) used hom*ogenate binding with [3H] labeled citalopram to compare the pharmacological profiles of central SERT binding sites in several fish species to those of rats and found similar KD and Bmax values between the genera. Furthermore, their study also demonstrated in vivo down regulation of the SERT transporter in zebrafish following dietary exposure to the SSRI sertraline. The evolutionary conservatism of the SERT transporter site among vertebrates (Dietl and Palacios 1988) suggests that exposure to

SSRIs may therefore cause a physiological affect in fish.

Many PPCPs provide ecological risk assessors with a unique set of challenges and researchers have emphasized concerns that traditional ecological risk assessment approaches may be inadequate for hazard characterization (Lange and Dietrich 2002,

Huggett et al. 2003, Brooks et al 2009). Environmental hazard assessment frameworks for PPCPs in aquatic ecosystems originally relied heavily on lethality endpoints from short-term bioassays as a starting point because data was generally more available (FDA

1998, EMA 2001). Acute to chronic ratios and safety factors could then be applied to establish environmental concentrations likely to minimal adversely affects on organisms associated with chronic exposure (Huggett et al. 2003). The practical constraint of using this ecological risk assessment framework for PPCPs is that most pose little risk of lethality following acute exposure (Webb 2001), yet many are quite potent and can elicit biological change at low concentration because of their high selectively and specificity.

Scott et al. (2004) emphasized that standardized acute lethality tests likely ignore ecological death because exposure to sub-lethal doses may cause organisms to be unable to function in an ecological context. Ultimately, the major dilemma risk assessors face is

37

that traditional endpoints from toxicity tests (e.g., survivorship, growth, reproduction)

only provide a coarse resolution of potential biological effects, thereby perplexing the

accurate characterization of risk for some PPCPs to aquatic organisms.

Gerhardt (2007) noted that biochemical, physiological, and behavioral responses

can provide greater resolution following short-term exposures because they may become

discernable within minutes, whereas life history response and morphological responses

may take months or years to manifest. While biomarkers are extremely practical as tools

for inferring exposure, they have limited use for identifying biological effects as it is

often difficult to elucidate specific causal relationships between biochemical changes and

specific ecological consequences. Alternatively, changes in behavior can represent the

initial response of an organism to a chemical stressor and may help explain observed

reductions in survival, growth, or reproduction (Fernández-Casalderry et al. 1994).

Behavior itself is defined as the visual response of an organism to a culmination of biotic

and abiotic stimuli. The response of an organism to these cues is dependent upon

physiological (internal) signals, as well as environmental or social (external) factors

(Gerhardt 2007). Contaminant-induced alterations in behaviors may therefore reveal

connections between the biochemical, individual, and population levels of biological

organization (Weis et al. 2001). For these reasons, Peakall (1996) emphasized that

behavioral endpoints may be more robust and comprehensive for ecological risk

assessment than either physiological or biochemical parameters.

Researchers have long recognized the potential applicability of animal behavior as

indicators of sublethal stress during laboratory bioassays with fish. For example, Warner et al. (1966) contrasted the behavior of control fish to those exposed to gradients of

38

various contaminants and showed that exposure may adversely affect organisms at concentrations far below those that cause death or immobility. Since then several other studies have investigated how anthropogenic pollutants affect fish behavior (Marcucella and Abramson 1978; Little et al. 1985; Rand 1985; Atchison et al. 1987; Beitinger 1990;

Little and Finger 1990; Døving 1991; Blaxter and Hallers-Tjabbes 1992; Scherer 1992), and Scott and Sloman (2004) emphasized that most of the previous studies focus solely on direct response measures, such as avoidance, coughs, or body tremors. There has been a transition from these pursuits and more recent research has attempted to identify changes in behaviors that are more ecologically relevant, such as foraging (Sandheinrich and Atchison 1990, Kasumyan, 2001, Hahn and Schulz 2007), feeding rates (Grippo and

Health 2003, Valenti et al. 2010, Stanley et al. 2007) predator –prey interactions (Weis et al. 2001, Weis et al. 2003), reproduction (Martinovic et al. 2007), and social hierarchies

(Perreault et al. 2003). Furthermore, tests with fish to assess anxiety, stress, and fear have been developed as tools to study neuropsychopharmacology, neuropathology, and psychopathology (Baraban et al. 2005, Bass et al. 2008, Blaser et al. 2010).

These efforts have been promoted by advances in digital tracking technologies and statistical approaches for interpreting data, which have absolved criticism concerning the potential subjective nature of behavioral endpoints. Kane et al. (2005) described progress in behavioral sciences and provides a detailed review, as well as future perspectives, for the use of automated techniques during experiments with fish models.

One such technology is NOLDUS Ethovision XT, which is a software package that allows automated tracking and analysis of animal movement and activity (NOLDUS website). The system processes video images and has been widely applied to study

39

behavior in mammalian models, which is often associated with drug development.

Tracking swim patterns of individual has proven an effective means to assess sublethal changes in exposed fish (Little and Brewer 2001), and Ethovision XT may prove an effect way to studying sublethal toxic effects of fish. Several studies have used this program to access behavioral responses in larval (Baraban et al. 2005,) and adult zebrafish (Gerlai et al. 2000)

A modified plus maze with both black and white arms is another promising approach for assessing behavioral changes in fish (Sackerman et al. 2010). Several researchers have observed that teleosts have natural preferences to dark environments

(Serra et al. 1999, Maximino et al. 2010), and exposure induced change of this behavior could therefore be suggestive of anxiolytic effects of drugs in fish models. Under this premise, fish exposed to agents that reduce anxiety would be more likely to explore and spend time in white arms relative to unexposed individuals. The modified plus maze for fish is based on the elevated plus-maze (EPM) for rats, which is one of the most popular in vivo animal tests as over 2500 experiments have been published (Carobrez and

Bertoglio 2005). Montgomery (1955) first used the concept and observed that rats were less likely to exhibit exploratory behaviors in ‘open’ arms compared to ‘closed’ arms in a

Y-shaped test apparatus, which he attributed to an avoidance of open alleys due to the fact that they engender higher levels of fear. A modern version of the EPM has been designed with four arms and has been used to assess how drugs may influence anxiety in rats. For example, Handley and McBlanch (1993) noted that the anti-anxiety drug diazepam increased the ratio of open: total arm entries, while the pro-anxiety drug picrotoxin diminished this ratio.

40

Another novel approach for assessing anxiety specific to fish models that is gaining in popularity is use of a dive tank (Sackerman et al. 2010). This is a very simplistic approach in which a fish is placed in a novel environment and the time it

spends at various depths is monitored. Typically, fish placed in a novel environment will

immediately dive to the bottom of a tank, which is likely an anti-predation mechanism.

This may also be a useful tool to assess anxiety, which is inherently coupled to predator

avoidance, and fish exposed to anti-anxiety drugs would be more likely to spend time

away from the bottom compared to unexposed fish. Several researchers have successful

used this approach to assess anxiety in zebrafish (Levin et al. 2007, Bencan and Levin

2008, Egan et al. 2009).

In our experiment we examined whether water exposures of the SSRI sertraline

caused increases in plasma concentrations that subsequently led to physiological changes

in the SERT transporter of adult male Pimephales promelas (fathead minnow).

Furthermore, we attempted to distinguish whether these changes led to altered behavior

by using a barrage of behavioral trials. These experiments included use of the modified plus maze for aquatic organisms, dive tanks, and digital tracking software. The ultimately goal of the study was to relate how increased plasma concentrations of a model xenobiotic could cause physiological changes that cascaded into aberrant behaviors in

exposed fish.

\

41

Materials and Methods

Test Organisms

All fish were maintained and examined according to an approved Baylor

University animal care protocol. Newly hatched Pimephales promelas (<24 h) were purchased from a commercial supplier (Environmental consulting and testing Inc,

Superior, WI). Individuals were housed in a flow through system supplied with aged, de- chlorinated tap water at a constant temperature of 25±1º C under a 16:8 light: dark photoperiod. For the first 30 d individuals were fed twice daily solely on a diet of newly hatched Artemia spp. and thereafter were fed a mixture of Artemia spp. and certified test- grade flake food. Individual were aged to 120 d before exposures were initiated. Only adult male fish that exhibited pronounced reproductive features were used in bioassays.

Experimental Design

Exposures were completed in 5-g experimental units (aquarium) filled with 20-L of test solution. Each aquarium housed 6 adult male P. promelas and had 3 shelters (3-in

PVC cut in half length wise into sections of 10 cm) to reduce potential altercations. De- chlorinated tap water was used as the control treatment and dilution water for each of the

3 sertraline treatments. A stock of sertraline hydrochloride (Sigma Aldrich, St. Louis,

MO, USA) was prepared and used to create the sertraline concentrations. Treatment levels included 3, 10, and 30 µg/L sertraline and each had 3 replicate experimental units, thus a total of eighteen individual adults per treatment. The exposures for each set of replicates were staggered by one day to accommodate for the substantial time that it

42

would take to complete the behavioral trials. Each set of replicates consisted of four aquariums from each of the treatment levels, including a control.

Exposures were 28-d in duration and fresh exposure water was prepared daily and

75% v/v water renewals were performed. All aquaria were housed in a single walk-in incubator set at 25 ± 1º C with a 16:8 light: dark photoperiod. Temperature in the aquaria

was monitored daily. Water quality parameters, including dissolved oxygen,

conductivity, pH, chlorine, and ammonia were measured 3 times a week.

Behavioral Trials

Novel Dive Tank. The specific methods for behavioral trials in the dive tank are

described in Sackerman et al. (2010) and the overall concept for its use was based on

Levin et al. (2008). The dive tank was a transparent, hexagon 19 L fish tank (Marineland

Eclipse System 5 Hex combo) filled to a depth of 30 cm. Lines dividing the tank into

quarters were drawn on the outside with marker to aid observation (Figure 6). The tank

sat on a black countertop, and its back wall was covered with white vinyl to enhance

contrast for video recording. Fish in the dive tank were observed and digitally recorded

with a SONY Handycam (DCR-SX40) for 300 seconds to determine the amount of time

fish spent in each zone. The camera was position on the side at the center height of the aquarium. The plus maze was repeatedly rinsed three times with de-chlorinated tap water before being refilled for the next trial.

Novel light-dark plus maze. The aquatic plus maze test was performed according to methods described by Sackerman et al. (2010). The height, width, and depth of the plus maze were 71 cm, 51 cm, and 10 cm, respectively (Ezra Scientific, San Antonio,

43

TX). The plus maze module had a 10 cm2 center section surrounded by four arms (three were10 x 10 cm, while the fourth was 10 x 15 cm). The two shorter opposing arms were lined with black polyethylene, while the other arms were lined with white polyethylene cut from folders and secured to the walls with binder clips. The maze was place on a grey commercial grade cart and the center was left open. An image of the plus maze is shown in Figure 7. The maze was filled to a depth of 8 cm.

Fish netted from the dive tank were released into the center section of the plus maze and observed and digitally recorded for 300 seconds with a SONY Handycam

(DCR-SX40). The camera was positioned directly over the center of the plus maze.

During the trial, the number of crosses into white arms, total duration of time spent in white arms, and time fish spent motionless in the center section upon introduction

(initially frozen) were calculated from digital recordings. This scoring was based on the technique used for rats in the elevated plus maze (Lapiz-Bluhm et al. 2008). The plus maze was repeatedly rinsed three times with de-chlorinated tap water before being refilled for the next trial.

NOLDUS experiments. Once the behavior of 1 individual from each of the treatments was assessed using the dive tank and plus-maze, the 4 fish were allowed one hour of acclimation prior to loading into the NOLDUS behavioral apparatus (Figure 8).

The order in which fish were removed from a treatment was determined by using a random number table. The order was also used to determine their position in the

NOLDUS apparatus. These experiments were designed to examine shelter-seeking and nest-

44

Figure 6.. A photo of the dive tank used to assess differences in the behavior of adult male Pimephales promelas unexposed and exposed to sertraline based on work by Levin et al. 2007.

Figure 7. A photo of the plus maze used to assess differences in behavior for sertraline exposed and control adult male Pimephales promelas. The overall design is based on similar apparatuses used in drug development to assess anxiety in rats.

45

guarding behavior under both light and dark conditions. The apparatus essentially consisted of a SONY Handycam (HDR-SR11) that was equipped with a visible light filter (850 nm) positioned over the center of four test areas. The camera was connected to a video port on the back of a computer that had NOLDUS digital tracking software installed using AV cables. Each test area was a white plastic bin (20 L, 35 cm x 40 cm) that housed a shelter and was filled with 15 L of de-chlorinated tap water. The camera and NOLDUS video setting were previously calibrated to optimize tracking with individuals not used in the trials. In addition, the arena was also calibrated by using a ruler placed at several locations in each of the 4 arenas. Two high powered infrared lights (Manufacturer) were positioned below the 4 arenas to allow for indirect illumination with infrared light. These lights remained on throughout the duration of the experiment. Two light fixtures fitted with 60 watt bulbs were position above the areas that were either turned on or off to simulate light and dark.

SIDE VIEW TOP VIEW

Figure 8. A drawing of the NOLDUS apparatus used to assess differences in the behavior of adult male Pimephales promelas unexposed and exposed to sertraline.

46

Fish were transferred to the areas using an aquarium net and the trial initiated

within 10 s once all four fish were loaded. The order that fish were introduced was based

on the random number table. The camera was set to record and a trial was initiated on

the NOLDUS software program menu. At the beginning of the trial the lights were on;

however, after 300 second the lights were then switched off. This processes was repeated

a total of 3 times so that during each trial the behavior of the fish could be assessed 3

different times under both light and dark conditions. The NOLDUS software

automatically analyzed the total distance traveled, swim velocity, and overall movement

for each of the 300 second light and dark time interval for each fish. The time that the

fish spent in the shelter was calculated manually by reviewing digital video files to ensure

that the fish was actually inside of the shelter and not merely swimming above it.

Following a completed NOLDUS trial the fish were removed, their weight and

length measured, and then decapitated. The head was then immediately place inside of a

microcentrifuge tube and stored in a -80º C freeze for future brain analysis. A

heparinized hematocrit tube from StatSpin (Westwood, MA) was used to collect blood.

Once this process was completed for all 4 fish, the blood was spun at 10,000 rpm for 3 min in a microcentrifuge. The resulting plasma was then harvested and placed in a microcentrifuge tube with a micropipette so that a volume could be estimated. The plasma was the immediately diluted 1:10 with 100mM phosphate buffer saline buffer with preservatives pH 7.0 filtered with 0.22 micron (PBS) (Immunalysis Corporation,

Ponoma, CA) and placed under refrigeration at -4º C. Another round of plus maze, dive

tank, and NOLDUS apparatus trials was completed as previous described using

individuals from the same group of tanks.

47

Analytical Quantification of Sertraline

Sertraline direct ELISA kits (Immunalysis Corporation, Pomona, CA) were used to quantify plasma sertraline concentrations in control and exposed fish. The

Immunalysis sertraline direct ELISA kit is based upon the competitive binding to antibody of enzyme labeled antigen and unlabeled antigen in proportion to the concentration of the reaction mixture. The plasma of individuals in the same tank was pooled to ensure adequate volumes for analytical measurements. A calibration curve was developed by collecting plasma from fish not used during the exposure, which was then diluted with PBS. This mixture was then loaded in triplicate in the ELISA kit as the control. Next, the mixture was spiked with sertraline to a final concentration of 800 ng/ml. Then 100 ul of the 800 ng/ml plasma sertraline standard was added with a micropipette to the ELISA kit in triplicate. The 400 ng/ml standard was achieved by adding 50 ul of the 800 ng/ml mixture in triplicate to the ELISA kit. The 200, 100, 50, and 25 ng/ml concentrations were prepared similarly by adding half of the volume of the next higher concentration.

Plasma samples from exposed fish were then added to the wells of the ELISA kit.

To ensure that absorbance readings were within the calibration curve range, three dilution of the plasma were prepared. This was achieved by adding 25, 50 and 100 ul of exposed fish plasma in triplicate to the ELISA kit to establish respective dilutions of 0.25, 0.5, and

1. After incubation the kit was then placed on a calibrated microplate reader and absorbance was measured at 450 nm and 630 nm within 30 min of loading the last sample. A detailed description of the directions for the Sertraline direct ELISA kit can be found at the Immunalysis website. Sertraline direct ELISA kits were also used to

48

confirm nominal concentrations of sertraline in exposure water using methods similar to

those previously described.

Comparing Measured versus Predicted Fish Plasma Concentrations

Measured plasma concentrations of sertraline were compared to predictions from

a slightly modified model developed by Fitzsimmons et al. (2001). For calculations, Log

KOW value (Eq.1 ) was substituted for the predicted Log D value of 3.77 (Scifinder scholar) and the LogDlipwater value (Eq. 2) of 4.06, which was based on the equation

described in Escher and Schwarzenbach (2000). The exposure concentrations of 0, 3, 10,

and 30 µg/L were substituted in for the environmental concentrations (EC) values used in

Eq.3. The human therapeutic plasma concentration (HTPC) was defined as 142 ng/mL

(Thomson 2008) and used to calculated Eq 4.

Log PB :W 0.73 x Log DO : 0.88 (1)

Log PB :W 0.73 x Log DL : 0.88 (2)

FF PC EC x PB : (3)

ER HTPC/FSSPC (4)

Saturation Radioligand Binding to Serotonin Transporters in Whole Brain hom*ogenates

SERT saturation binding to [3H] citalopram in membrane hom*ogenate preparations from fathead minnow whole brains were performed following the methods of D’Amato et al. (1987), with minor modifications. Whole brain from three fish were

49

pooled into each hom*ogenate preparation, and tissue was dispersed at 30,000 rpm for 20

sec into 25 ml of ice-cold 50 mM Tris, 120 mM NaCl, 5 mM KCl, pH 7.4 at 26ºC buffer

using a tissue hom*ogenizer (Polytron 3100, Kinematica, Bohemia, NY). The

hom*ogenate was spun for 10 min at 30,600 x g in a 4°C centrifuge (Avanti A-J,

Beckman-Coulter, Brea, CA). The supernatant was discarded and the pellet re-suspended

in 5 ml buffer on ice using a hand-held tissue hom*ogenizer. Another 20 ml of buffer was

added and the hom*ogenate was centrifuged again for 10 min at 30,600 x g. The final

pellet was re-suspended to obtain a protein concentration near 1 mg/ml, as determined with Bradford reagent (Sigma, St. Louis, MO) and measured on a spectrophotometer

(DU-640, Beckman). The hippocampal hom*ogenate was incubated at 26°C for 1 hour in buffer containing 0.1 – 12 nM of [3H] citalopram (PerkinElmer, Boston, MA). Non- specific binding was defined by 50 µM sertraline (Pfizer, Groton, CT). Incubation was terminated by addition of 4 ml of buffer, pH 7.4 at 4°C. Labeled hom*ogenates were captured by filtration under vacuum with a tissue harvester (Brandel, Gaithersburg, MD) onto Whatman GF/B filter paper strips (Brandel) pre-soaked in 5% polyethyleneimine

(Sigma). Filters were washed twice more with 4 ml of buffer. [3H] citalopram trapped in membrane tissue on the filters was determined using a liquid scintillation counter (LS

6500, Beckman) with 40% efficiency.

Statistical Analysis

Statistical significance of response variables was assigned at α = 0.05 for all tests.

The mean responses of individuals for each replicates were determined (n=6). These means were then used for statistical analysis (n=3). No-observable-adverse effect concentration (NOAEC) and lowest-observable-adverse effect concentrations (LOAEC)

50

for results of behavioral trials were calculated based on the statistical methodology described by standard protocols. Endpoints for the plus maze and dive tank were compared among the different treatments using analysis of variance (ANOVA).

For results of experiments using NOLDUS, ANOVAs were performed between the various sertraline treatments by each time period. The time that fish spent in shelters was tallied by hand by reviewing video data, while distance traveled and velocity were calculated using NOLDUS software by creating time bins of 300s for each trial. For endpoints that had significant differences among the means, Dunnett’s comparisons were then completed between control groups and those exposed to sertraline for each specific time, respectively.

Results

Analytical Quantification of Sertraline

Nominal treatment concentrations for water exposures were very close to those measured (Table 5). The results of ELISA experiments also clearly demonstrated a dose- dependent relationship between concentrations of sertraline in water and plasma concentration in exposed fish. Sertraline was not detected in the plasma of control fish but was quantifiable for the other treatments. The respective mean (±standard deviation) plasma concentrations for the 3, 10, and 30µg/L exposure treatments were 280±110,

720±70, and 1900±50 ng/mL. These measured sertraline plasma concentrations were very similar to those predicted by the fish plasma concentration model based on water column exposure concentrations (Figure 9). Furthermore, the effects ratios (ER) based the ratio measured fish plasma concentrations and human therapeutic plasma

51

concentrations were all <1, suggestive that potential biological effects would be realized in exposed fish. These measured values were closer to predicted values based on

LogDOct:water compared to LogDlip:water as the relationship was nearly one to one based on

slopes of 0.85 and 0.5, respectively.

Table 5. Measured sertraline concentration in treatment water for the 28 d chronic experiments with Pimephales promelas.

Water concentration (µg/L) Nominal Sertraline (µg/L) Week 1 Week 2 Week 3 Mean ± SD 0 <1.5 <1.5 <1.5 <1.5 3 3.5 3.3 3.4 3.4 ± 0.1 10 9.9 12 11 11 ± 1 30 26 29 27 27 ± 1

2500

2000

1500

1000

500

0 0 1000200030004000 Measured plasma concentration (ng/mL) concentration plasma Measured Predicted plasma concentration (ng/mL) Figure 9. The measured versus predicted fish plasma concentration based on the Huggett et al. model (2003). Closed dots are based on Log Doctw, open dots are based on Dlipw.

SERT Binding Study

Physiological changes in the brains of fish were also apparent for fish with

elevated plasma concentrations of sertraline following water borne exposure. The mean

52

numbers of low affinity SERT transport sites in fish exposed to sertraline were significantly lower compared to the control (Table 6). However, while all treatments had lower SERT for the high affinity site, only those in the 3 µg/L treatment were significantly different than the control.

Table 6 The results of the SERT bound by radiolabeled citalopram experiments.

Sertraline High affinity site Low affinity site Treatment (µg/L) mean±SE mean±SE 0 27 ± 2.1 148 ± 8.2 3 16 ± 3.2* 101 ± 4.1* 10 18 ± 1.5 110 ± 9.4* 30 18 ± 2.4 111 ± 9.3*

Behavioral Trials

There was no significant difference in the behavioral endpoints measured for the plus maze among the different sertraline treatments. Although not significant (p=0.78), the mean number of times (± standard error) that fish exposed to 10 or 30 µg/L crossed into the white arms (4.9 ± 1.4 and 4.8 ± 2.3 crosses, respectively) were greater than the number by individuals in control and 3 µg/L treatments (3.3 ± 1.4 and 3.1 ± 1.5 crosses, respectively) (Figure 10). In addition, the mean number of total crosses between any color were also not significantly different (p=0.73); however, the mean (± standard error) number of total crosses increased with increasing sertraline treatment were 9.3 (±2.1),

11.9 (± 3.8), 12.4 (± 1.6) and 14.6 (± 9.6) for the control, 3, 10, and 30 µg/L (Figure 11).

Observed behavioral endpoints from the dive tank experiments were not significantly different between sertraline treatments, although the mean numbers of total crosses were less for the 10 and 30 µg/L treatments (Figure 12). Fish from all treatment levels spent predominantly all time (>95%) in the bottom area of the dive tank (Figure 13). 53

100

80

60

40

20

0 Mean time in white (sec) (n=3) (sec) white in time Mean 0 10203040

Sertraline concentration Figure 10. The mean amount of time that Pimephales promelas exposed to sertraline spent in white areas of plus maze. Fish spent a total of 300s in the plus maze.

8

6

4

2

0 crosses of white (n=3) number Mean 010203040

Sertraline concentration Figure 11. The mean number of times that Pimephales promelas exposed to sertraline crossed into white areas of plus maze.

54

12

) 10 n=3 ( 8

6 ne crosses

li 4

f

# o 2

0 0 10203040

Sertraline concentration

Figure 12 The mean number of times that Pimephales promelas exposed to sertraline crossed different areas in the dive tank. Fish spent a total of 300s in the dive tank.

100

80

60

40

20

Mean percentMean of time on bottom (n=3) 0 10203040

Sertraline concentration

Figure 13. The mean amount of times that Pimephales promelas exposed to sertraline spent in the bottom area of the dive tank.

55

The results of the behavioral trials with NOLDUS clearly demonstrated an

appreciable difference among the behavior of exposed and unexposed fish. Control fish

spent greater amounts of time in the shelter during light periods (1,3,and 5) compared to

the other treatments (Table 7) During light periods 3 and 5, control fish spent more than

twice the amount of time in the shelter (mean ± standard error, 217 ± 31 and 239 ± 36 seconds, respectively) compared to those exposed to any level of sertraline as these fish only spent 43-98 seconds in shelters (Table 7). The amount of time that fish spent inside the shelter during dark conditions (time periods 2,4,and 6) were similar for all treatments and means (± standard errors) ranged between 8 ± 3 and 34 ± 29 seconds. In addition, the mean distance of total travel (cm) was also significantly different between control and fish exposed to sertraline during the 2nd and 3rd light periods (Table 7). Exposed fish

travelled significantly higher distances during these light periods. Notably, the mean

velocities that fish travelled (cm/second) were not significantly different between any

time periods (Table 7). Therefore, differences in the mean total distance travelled during

light periods may have been merely manifested by the fact that the fish were spent less

time in shelters.

Discussion

In our study we demonstrated that water borne exposure to sertraline caused

elevated concentrations in the plasma of P. promelas that affected the SERT-binding

characteristics in the brains of exposed compared to unexposed fish (Table 6). These

results are supported by previous research, which demonstrated that oral doses of the

sertraline in food reduced SERT binding in zebrafish brains (Gould et al. 2007).

Furthermore, we related these physiological changes to potentially ecologically relevant

56

behavioral modification of shelter seeking/ nest guarding (Table 7). In previous studies, the alternative endpoint of feeding rate was observed to be more sensitive than growth or survivorship when juvenile P. promelas were exposed to SSRI in the water column

(Valenti et al. 2010, Stanley et al 2007). In our current study the sublethal response of shelter seeking in adult males (3 µg/L) was actually more sensitive than those for feeding with juveniles (15 µg/L) during exposures to sertraline. These findings suggest that the measurement of SERT occupancy following exposure to SSRIs could potential provide a better understanding of the pharmacodynamics and pharmaco*kinetics properties of these drugs and the potential risk they pose to aquatic organisms.

The measured plasma concentrations of sertraline in fish exposed to all of the water exposure concentrations tested in our study exceeded therapeutic plasma concentrations for humans. The predicted calculated effect ratios for the 3, 10, and 30

µg/L sertraline water exposures at pH 8.5 based on LogDOct:Water values were all < 1 as respective values were 0.65, 0.19, and 0.06. The respective means (±standard deviations) based on measured plasma concentrations were 0.57 (±0.1), 0.29 (±0.03), and 0.09

(±0.01). The modeled and measured ERs are similar and both indicate that potential toxicological effects may occur in fish based on plasma levels, which are further supported by the similarity in SERT transport sites (Gould et al. 2003). However, these relationships are only suggestive and cannot clearly prove that adverse affects of ecological consequence may occur in fish because the functionality of the serotonergic system varies among genera.

57

For use as a tool in ecological risk assessment, it is imperative that observed behavioral modification have ecological relevance and can be subsequently related to fitness or survival of individuals (Scott and Sloman 2004). The decrease in shelter

Table 7. The results of NOLDUS experimental trials including the mean (± standard error) amount of time adult Pimephales promelas spent in the shelter, the mean total distance(± standard error) traveled, and mean(± standard error) velocity. Time period 1,3, and 5 had lights on while time periods 2,4, and 6 were in the dark. Fish spent a total of 300s in the chamber.

Time Sertraline Time in shelter ± SE Distance ± SE Velocity ± SE Period (µg/L) (sec) (cm) (cm/sec) 1 0 258 ± 22 327 ± 143 16 ± 13 1 3 173 ± 44 372 ± 107 3 ± 1 1 10 203 ± 17 658 ± 276 11 ± 7 1 30 202 ± 49 568 ± 313 16 ± 12 2 0 30 ± 16 1640 ± 66 15 ± 5 2 3 23 ± 6 1590 ± 106 19 ± 12 2 10 33 ± 10 1728 ± 204 7 ± 0.6 2 30 21 ± 9 1890 ± 128 7 ± 0.7 3 0 217 ± 31 847 ± 290 10 ± 3 3 3 91 ± 42* 1760 ± 84* 8 ± 0.2 3 10 55 ± 14* 1300 ± 172* 8 ± 0.5 3 30 81 ± 26* 1746 ± 240* 9 ± 0.7 4 0 19 ± 1 1697 ± 152 21 ± 26 4 3 8 ± 3 1792 ± 138 7 ± 1 4 10 13 ± 5 2023 ± 225 8 ± 3 4 30 20 ± 15 2044 ± 74 7 ± 1 5 0 239 ± 36 905 ± 274 8 ± 3 5 3 78 ± 34* 1493 ± 6* 9 ± 2 5 10 98 ± 47* 1918 ± 195* 7 ± 0.3 5 30 43 ± 20* 2054 ± 152* 9 ± 1 6 0 34 ± 29 1574 ± 238 9 ± 4 6 3 9 ± 6 1777 ± 117 7 ± 1 6 10 5 ± 3 1870 ± 90 9 ± 3 6 30 4 ± 2 1982 ± 82 7 ± 0.4

58

seeking behavior we observed by adult male P. promelas exposed to sertraline during light periods has implications for both survival and fitness. Individuals willing to spend more time away from shelters may face greater predatory risk and thus their overall survival rate may be reduced. In terms of reproduction, fathead minnows typically spawn adhesive eggs to the underside of aquatic plants or woody debris (Nelson and Paetz

1992). Shelter seeking may be intrinsically linked to nest guarding behavior by male fathead minnows, which is important for defining territories and attracting mates

(Jamieson 1995). Martinovic et al. (2007) exposed P. promelas to environmental estrogens in the water and noted alterations in nest guarding behavior that ultimately led to complete reproductive failure. Kreke and Dietrich (2008) summarized the effects of serotonin and SSRIs in aquatic vertebrates and observed changes in various reproductively relate pathways often within 30 to 90 minutes following in vivo injections.

Other researchers have demonstrated that serotonin influences reproduction (Khan and

Thomas 1992) and aggression (Adams et al. 1996) in fish, and consequently exposure to

SSRI may spur similar behavioral modifications.

The altered behavioral patterns of patterns during the NOLDUS experiments may have been linked to anxiolytic properties associated with SSRIs and exposed individuals may have less reserved about exploring their environment. However, our results from the plus-maze and dive tank do not support this and therefore these behavior responses may not have been related to anxiety, but rather potentially to other physiological changes associated with exposure to SSRIs. Kreke and Dietrich (2008) suggested that following chronic exposure SSRIs could reach the retina and pineal organ in fish, which could affect serotonin reuptake at these sites. This could alter normal patterns of

59

serotonin and melatonin and consequently lead to changes in photosensitive behaviors,

such as hunting, feeding, or reproduction (Kreke and Dietrich 2008). Therefore, the

observed changes of behavior of P. promelas during light periods may have been caused by differences in the way in which they perceive their environment. Additional research examining the specific receptors in fish that may potential be affected by SSRIs exposure is clearly warranted.

There are several potential advantages associated with incorporating behavioral studies into toxicological assessment of pharmaceuticals to aquatic life beyond the fact that changes in behavior occur more rampantly than gross-levels of toxicity, such as survivorship or life-history characteristics. This is potentially advantageous in that it may

reduce cost and time lag associated with evaluating risk of pharmaceuticals using

biological entities. Furthermore, because observing behavior can be achieved non- invasively, researcher may complete time-dependent studies that entail both exposure and

recovery.

60

CHAPTER FOUR

A Mechanistic Explanation for pH-Dependent Ambient Aquatic Toxicity of Prymnesium parvum Carter

Introduction

NOTE: Chapter two is published in Toxicon (2010) 55:990-998. Please refer to Appendix B for the licensing agreement.

Harmful algal blooms (HABs) may have devastating impacts on aquatic

ecosystems, resulting in severe impacts to fisheries. Increases in the frequency and

severity of HABs on the global scale has triggered scientific inquiry to define factors

causing these trends (Zingone and Enevoldsen 2000, Anderson et al 2002, Hallegraeff

2003); however, among the greatest challenges for managers is the spread of invasive

species. Prymnesium parvum is an example of an invasive HAB species that has

transitioned from marine origins to inland systems. Identified nearly a century ago as a problem in marine environments because of its toxic blooms (Liebert and Deerns 1920),

P. parvum is more recently recognized as an invasive species threatening inland systems

in the arid and semiarid southwestern and south central United States (Baker et al. 2007,

Roelke et al. 2007, Schwierzke et al 2010).

Anthropogenic changes to the hydrologic cycle, eutrophication, and salinization

of waterways are associated with the spread of HABs (Anderson et al. 2002). Such

changes, intertwined with climatological and geological factors, have rendered some

Texas reservoirs to be within the tolerance range of P. parvum (Larsen and Bryant 1998,

Baker et al. 2007, Baker et al. 2009). Specifically, the species’ euryhaline nature has

61

apparently facilitated its transition from coastal marine and estuarine ecosystems to these weakly saline inland impoundments. Since the first harmful blooms of P. parvum in

Texas were documented in the Pecos River (James and De La Cruz 1989), P. parvum has spread to other systems in Texas resulting in toxic blooms and fish kills (Roelke et al

2007, Schwierzke et al. 2010).

Prymnesium parvum is a mixotrophic haptophyte that can gain energy photosynthetically as well as phagotrophically by feeding on other microorganisms

(Skovgaard and Hansen 2003). Exposure to P. parvum toxins can lyse cells (Yariv and

Hestrin 1961, Tillmann 2003), disrupt cell membrane integrity (Yariv and Hestrin 1961,

Padilla 1970, Kim and Padilla 1977, Brooks et al. 2010), and affect gill functions of aquatic organisms (Ulitzur and Shilo 1966). Ecologists have proposed several purposes for the production and release of toxins by P. parvum, including acquisition of prey

(Stoecher et al. 2006), elimination of algal competitors (Fistarol et al. 2003, Graneli and

Hansen 2006, Uronen et al. 2007), or reduced grazing pressure (Rosetta and McManus

2003, Tillmann 2003). A variety of factors, including nutrient limitation, salinity, temperature, and light are known to influence cell growth and the toxicity of laboratory cultures of P. parvum (Shilo and Aschner 1953, Padilla 1970, Dafni et al. 1972, Larsen et al. 1993, Larsen and Bryant 1998, Johansson and Granéli 1999, Granéli and Johansson

2003, Baker et al. 2007, Baker et al. 2009). Few studies have focused on factors governing the behavior of the toxins once they are released, or considered how bloom formation might alter the environment (e.g., light attenuation, nutrient availability, dissolved oxygen, pH) in ways that could influence the bioavailability and potency of P. parvum toxins.

62

Shilo and Aschner (1953) proposed that P. parvum toxins were proteins with

high molecular weights. To date, the only characterized toxins are prymnesin-1 and -2,

large chains of 90 carbon atoms and trans-1,6-dioxadecaline units with conjugated

double/triple bonds at each terminal end; their respective chemical formulas are

C107H154Cl3NO44 and C96H136Cl3NO35 (Igarashi et al. 1999). These compounds are amphiphilic, with uneven distributions of sugars and hydroxyl groups, and three chlorine atoms and one nitrogen atom. Both prymnesins are structurally similar to other HAB toxins such as maitotoxin and ciguatoxin, which are characterized by a network of hydroxylated polycyclic ether units (Murata and Yasumoto 2000). The amine present on the prymnesins suggests that these compounds might be weak ionizable bases with pKa values > 8. Prior studies suggested that some of the toxins released by P. parvum are ionizable, becoming more toxic to fish exposed at higher pH, with toxicity eliminated below pH 7 (Shilo and Ashner 1953, McLaughlin 1958, Ulitzur and Shilo 1964).

However, these experiments were completed under marine conditions, and prior to the development of standardized aquatic bioassays

This study examines whether pH also influences the toxicity of P. parvum toxin in less saline waters representative of Texas reservoirs where blooms have occurred.

Simultaneous bioassays were performed at three pH levels with samples obtained during

P. parvum blooms occurring in 2007 from two reservoirs, and with samples of laboratory cultures and culture filtrates. Further, the chemical structures of prymnesin-1 and -2 were examined to estimate their physiochemical properties. We hypothesized that toxins released by P. parvum are ionizable weak bases.

63

Material and Methods

Bioassays with Samples Obtained from Reservoirs Experiencing Blooms

Lake Whitney. Lake Whitney is a reservoir constructed in 1951 on the Brazos

River, with a capacity of 4.68 x 108 m3, surface area of 95 km2, and shoreline of 362 km

(Bailes and Hudson 1982). Two 4-L samples were collected in NALGENE® I-Chem

Certified Series™ 300 LDPE Cubitainers™ (Fisher Scientific) from Lake Whitney

during a bloom in March 2007, transported to the laboratory on ice, and stored under refrigeration at 4º C. This lake sample contained 61.5 x 103 P. parvum cells mL-1

(enumerated microscopically with a hemocytometer ). Ambient pH at the site in Lake

Whitney when the sample was collected was pH 8.4. Total ammonia in the samples was

<1 mg/L in whole samples, which is below ambient water quality criteria for the temperature and pH at which our experiments were completed. Dilutions in our toxicity experiments at which toxicity was observed further confirmed that dose dependent responses were not due to ammonia. Toxicity tests were initiated within 96 h of sample collection following EPA recommendations for ambient toxicity studies (US EPA 2002).

Acute bioassays with <48 h old Pimephales promelas were conducted in 100-mL glass beakers. Three replicates of seven individuals were prepared for each treatment level. Reconstituted hard water (RHW) prepared according to APHA et al. (1999) was used as the diluent and control (treatment consisting of 100% RHW). Treatment levels were prepared by diluting lake water with RHW to the following percentages of lake water: control (RHW), 0.01, 0.1, 1, 5, 10, and 20%. These treatment levels were selected based upon ambient toxicity data from prior water quality monitoring efforts. A volume

64

of 3-L was prepared for each treatment level, which was then divided into three aliquots

of 1-L that were then adjusted to pH units of 6.5, 7.5, or 8.5 (+ 0.05) prior to dispensing

experimental aliquots. The pH adjustments were achieved by slowly titrating 10%

HPLC-grade nitric acid, which generally followed U.S. Environmental Protection

Agency protocols for pH adjustment in Toxicity Identification Evaluations (US EPA

1991). Test individuals were fed newly hatched brine shrimp (Artemia sp.) 2 h prior to the exposure, but were not fed during experiments (US EPA 2002). Survivorship was assessed at 24 and 48 h, and temperature, dissolved oxygen, and pH were measured at test initiation and completion. Exposures were conducted at 25±1ºC under a 16:8 light: dark photoperiod.

A 10-d Daphnia magna reproductive study was also completed with Lake

Whitney water (US EPA 1994, modified as in Dzialowski et al. 2006). Treatment levels consisted of control (RHW), 12, 25, 50, and 100% lake sample water. Test solutions were prepared and adjusted to desired pH as previously described. Experimental units were 100-mL beakers filled with 80 mL of test solution. Organisms were fed daily with

Pseudokirchneriella subcapitata (formerly Selenastrum capricornutum) augmented with filtrate from a cerophyll suspension (approximately 2 g / L RHW). The final green algae cell concentration was 30 x 106 cells mL-1. One D. magna individual < 24 h old was

introduced in each beaker and transferred every other day to fresh test solution. Five

replicates were prepared for each treatment level. Survivorship and fecundity were

monitored daily. The experiment was completed for 10 d at 25±1ºC under a 16:8 light:

dark photoperiod.

65

Lake Granbury. Lake Granbury is a reservoir constructed in 1969, with a

capacity of 167 x 106 m3, a surface area of 34 km2, and an average depth of ~5 meters.

Samples were collected during a P. parvum bloom that caused fish kills in March 2007 at

three fixed monitoring stations and were handled prior to bioassay initiation as previously

described. Ambient pH at the sites ranged between pH 8.2 and 8.4. Total ammonia

concentrations were <0.25 mg/l in the three samples, which again were lower than levels

associated with ammonia toxicity (US EPA 1999). Cell counts for samples from Sites 1-

3 were 29 x 103, 36 x 103, and 36 x 103 cells mL-1, respectively.

Experiments were conducted with P. promelas similar to those previously

described in order to assess acute toxicity at the three sites. Six treatments, including a

control (RHW), 6, 12, 25, 50, and 100% lake water were prepared using RHW as the

diluent and adjusted to pH 6.5, 7.5 and 8.5. A 96-h acute exposure experiment was

initiated with < 24 h old D. magna using a composite sample from the three stations.

Treatments included a control (RHW), 6, 12, 25, 50, and 100% lake water. Ten replicates were prepared for each treatment level. Exposures were completed in 100-mL glass beakers filled with 80 mL of test solution, and water in experimental units was renewed at 48 h. Organisms were fed daily the same concentration of the mixture described for the 10-d D. magna experiment, at which time survivorship was assessed.

Exposures were conducted at 25±1ºC under a 16:8 light: dark photoperiod.

Laboratory Culture Preparation

The UTEX LL 2797 (University of Texas, Austin, Texas, USA) strain of P. parvum was used to initiate cultures. Cultures were grown in 20-L glass carboys filled with 14-L of an artificial seawater (ASW) prepared according to Berges et al. 2001 and

66

then diluted to a working salinity of 5.8 g L-1 with ultrapure water (18 MΩ cm-1).

Afterwards, nutrients (NaNO3 and NaH2PO4) were added at concentrations of f/2 and f/8

media (Guillard 1975); vitamins and trace metals were the same for both types of media.

Three replicates were prepared for each treatment and all carboys were inoculated with

103 cells mL-1 of P. parvum from stock cultures in late exponential phase grown in the

corresponding medium at the stock salinity of 5.8 g L-1. Cultures were maintained in

incubators at 20±1oC for a 12:12 light: dark cycle with an irradiance of ~140 µE m-2 d-1.

Carboys were repositioned and mixed daily by gently swirling.

Bioassays with Samples Obtained from Laboratory Cultures

Several experiments with larval P. promelas were completed using these P. parvum cultures. An initial experiment examined the toxicity of the three replicate carboys of both the f/2 and f/8 cultures. ASW adjusted to a salinity of 5.8 g L-1 used in

the media served as the diluent and controls. Bioassays were completed in 100-ml

beakers filled to capacity with test solution. Treatments included a control (ASW), 1,

2.5, 10, 25, and 100% culture water containing P. parvum cells. Four replicates of five

individuals were prepared for each treatment. Experiments were conducted at 25±1ºC

under a 16:8 light: dark photoperiod.

For subsequent studies, we separately pooled f/2 and f/8 cultures and then filtered

half of each volume through GF/C filters (Whattman GF/C; VWR International, West

Chester, Pennsylvania, USA). Acute toxicity to P. promelas was determined for cultures

(f/2, f/8) that were unfiltered and filtered (cell-free filtrate), then these samples were manipulated to pH 6.5, 7.5 or 8.5 following procedures outlined above. ASW adjusted to a salinity of 5.8 g L-1 was used as the diluent and control. An additional RHW treatment

67

for quality assurance was also prepared. Each acute toxicity study included a control

(ASW), RHW, 0.1, 1, 5, 10, 25, and 100% media treatment. Four replicates of five P. promelas <48 h old were used for each treatment level. Experiments were conducted at

25±1ºC under a 16:8 light: dark photoperiod.

Statistical Analysis

LC50 values for acute toxicity to P. promelas and D. magna were calculated by

Probit analysis if data met assumptions; otherwise, the Trimmed Spearman-Karber

method was applied using TOXSTAT computer software (US EPA 2002). SAS (SAS

Institute, Cary, NC, USA) was used for other statistical analyses. For the 10-d

experiment with D. magna, significant differences in survivorship between the control at

each of the respective pH treatments and reservoir water dilutions were determined using

Fisher’s Exact Test. Significant differences in reproduction were assessed by an

ANOVA comparing the mean neonate production per female for all treatments (α =

0.05), followed by Dunnett’s test comparing the controls to each of the treatments at a

respective pH (α = 0.05). In addition, we compared the mean control responses between

different pH levels for the respective endpoints using ANOVAs for each series of

experiments to confirm health of test organisms.

Estimation of Prymnesin-1 and -2 Physicochemical Properties

Calculation of physicochemical parameters for prymnesin-1 and -2 (Figure 14)

was carried out using ACD/Labs (Advanced Chemistry Development, Inc., Toronto,

Ontario, Canada) ChemSketch, pKa calculator, LogD calculator, and LogP calculator

(Version 9). LogP was calculated for each whole molecule, and percent distribution of

68

species and pKa values were calculated considering the hydrophobic component of the

molecule containing the primary amine group (Murata and Yasumoto 2000). The full molecule could not be handled by the program due to the large number of ionizable sites,

especially on the hydrophilic portion of the molecule. Calculated values for LogP

(octanol: water partitioning coefficient when the compound is primarily unionized),

LogD (coefficient of octanol: water partitioning ratio of ionized to unionized over a pH

range), and pKa (acid dissociation constant) are estimates based on the use of an

extensive database of fragments and predicted inductive effects based on substituents near ionizable sites. LogD was calculated as the sum of logD for the hydrophobic fragment at different pHs with the LogP of the hydrophilic fragment (neutral form), which remains neutral at all relevant pHs (< 12).

OH OH OH Cl OH H O H O Prymnesin-2 O OH H O O OH OH OH Cl H O O OH O OH O O O O O H O OH OH OH Cl

hydrophobic hydrophilic portion portion (common)

H O Cl H O OH H O OH NH2 Prymnesin-1 O O H O OH OH H O H O H O O OH O O O O OH OH H O OH O Cl O OH O O OH O O O OH O O O O O O H O OH OH OH OH

Figure 14. The structures of prymnesin-1 and prymnesin-2 with the hydrophobic and hydrophilic portions of each compound differentiated. The primary amine is highlighted.

69

Results

pH Dependent Toxicity in Field Studies: Lakes Granbury and Whitney

Ambient toxicity to juvenile P. promelas of Lake Whitney samples collected

during a Spring 2007 bloom was reduced when pH was adjusted to < 7.5 (Fig. 15). The

48-h LC50 value (95% confidence intervals) for experiments completed at pH 8.5 was 1.9

x 103 (1.6-2.7 x 103) cells mL-1, whereas comparable values at pH 6.5 and 7.5 were 7.8 x

103 (7.1-8.8 x 103) and 4.1 x 103 (2.6-5.3 x 103) cells mL-1. There was a similar pH-

dependent toxicological relationship during experiments with D. magna, as fecundity was

significantly reduced at lower densities of P. parvum cells when exposure occurred at

higher pH (Fig. 16). No reproduction was observed at any pH in 100% Lake Whitney water. There was no significant difference in reproduction between controls at each pH

or unmodified RHW.

Susceptibility of P. promelas to samples from Lake Granbury during the 2007 bloom

also indicated a pH-dependent toxicological relationship. Ambient toxicity to fish was

ameliorated in lake samples from two stations and substantially reduced in a third by

lowering pH (Table 8). There was insufficient mortality for samples collected from two

sites to calculate LC50 values at pH 6.5 and 7.5; thus, these values are conservatively

reported as > 100%. The sample from Site 2 was the only Lake Granbury sample for

which LC50 values could be determined for pH 6.5, 7.5, and 8.5. There was

approximately a four-fold difference in LC50 values between the pH 6.5 and 8.5

treatments, with higher toxicity at higher pH (Table 8).

70

100

80 7.5 8.5 60

40 survivorship (%) survivorship

71 20 Pimephales promelas 0

Mean control 6 62 615 3075 6150 12300 -1 Prymnesium parvum cells ml

Figure. 15. Average survivorship (±SD, n=4) of Pimephales promelas exposed to dilutions of Lake Whitney water collected during a bloom of Prymnesium parvum in 2007. Cell density is expressed as the % lake water multiplied by the density of P. parvum cells in the undiluted sample. The error bars represent the standard deviation. Missing error bars are due to 100% survivorship in all replicates for a treatment, hence there was no variation to derive a prediction.

72

Figure. 16. Mean neonate production for Daphnia magna (±SD, n=5) exposed to diluted samples of Lake Whitney water collected during a Prymnesium parvum bloom in 2007. Cell density is expressed as the % lake water multiplied by the density of Prymnesium parvum cells in the undiluted sample. The error bars represent standard deviations for each value. * represents treatments that are significantly lower than the respective controls (p<0.05). Controls did not vary significantly among the three pH treatments (p>0.05).

Table 8. The percent survivorship in undiluted samples and 48-h LC50 values in terms of percent reservoir water for Pimephales promelas exposed to Lake Granbury samples from three stations during a Prymnesium parvum bloom in March 2007.

Upper and lower % Survivorship in 48-h LC50 95% confidence 48-h LC50 Site Cells per ml pH undiluted sample (% lake water) intervals (P. parvum cell/ml) 6.5 81 > 100A nc nc 1 2.9 x 103 7.5 71 > 100A nc nc 8.5 24 72 61-85 21 x 103 6.5 38 80 59 - 100 28 x 103 2 3.6 x 103 7.5 5 54 41 -71 19 x 103 8.5 0 22 15 -31 78 x 102 6.5 90 > 100A nc nc 3 3.6 x 103 7.5 67 > 100A nc nc 8.5 0 43 36 -53 16 x 103

Similarly, toxicity to D. magna was reduced in low pH in a 96-h experiment

exposing individuals to a composite sample from all three sites (Table 2). LC50 values

could not be calculated at pH 6.5 due to insufficient mortality, but point estimates for pH

7.5 and 8.5 differed by nearly two-fold, with higher toxicity to D. magna at higher pH

(Table 9).

Table 9. The 48- and 96-hr LC50 values in terms of percent reservoir water for Daphnia magna exposed to a composite sample obtained from Lake Granbury during a Prymnesium parvum bloom in 2007.

LC50 Upper and lower 95% LC50 Time pH (% site water) confidence intervals (P. parvum cell/ml) 6.5 >100A nc 34 x 103 48 7.5 65.5 42 -100 22 x 103 8.5 46.7 31- 70 16 x 103 6.5 >100A nc 34 x 103 96 7.5 57.4 47-70 19 x 103 8.5 30.7 26 -37 10 x 103

A = There was insufficient mortality to generate a point estimate. nc = Not calculable.

73

pH Dependent Toxicity in Laboratory Cultures

Cultures were terminated on day 28 after reaching late stationary phase, and experiments were immediately performed to assess the toxicity of each replicate culture.

Cells were enumerated at this time showing densities in high nutrient (f/2) cultures of 2.0 x 105, 1.5 x 105, and 2.1 x 105 cells mL-1, and densities in low nutrient (f/8) cultures of

1.5 x 105, 1.3 x 105, and 1.5 x 105 cells mL-1. Survival in the ASW and RHW controls was > 90% for all tests at all pH levels. The LC50 values for experiments with P. promelas were consistently lower for the low nutrient (f/8) treatment compared to those for high nutrient (f/2) (Figure 17). Estimated LC50s were more variable between replicates for the f/2 treatment and increased exposure time resulted in greater toxicity, whereas temporal effects were less evident for the f/8 treatment (Figure 17).

2.0x106 )

-1 f/2 media f/8 media

1.5x106 cells ml cells

1.0x106 P. parvum P.

500.0x100.5x1063 value ( 50

LC 0.0 0 1020304050

Time (h)

Figure. 17. LC50 values for Pimephales promelas exposed to cultures of Prymnesium parvum grown in the laboratory using two different nutrient conditions (high nutrients – f/2 medium; low nutrients – f/8 medium). Each data point represents the LC50 value for an individual culture.

74

Samples of f/2 and f/8 whole cultures and cell-free filtrate were consistently more

toxic to P. promelas when exposure occurred at pH 8.5 compared to pH 7.5 or 6.5

(Figure 18). For the f/2 treatment, 50% of exposed individuals died at pH 6.5 in

undiluted whole culture; however, only 15% died in the cell free filtrate. Cell free

filtrates were also less potent than the whole culture at pH 7.5 and 8.5 for the f/2 treatment; however, differences in toxicity between whole cultures and cell-free filtrates

were not as apparent for the f/8 treatment (Figure. 18). The LC50 values were markedly

lower for filtered and unfiltered cultures grown in f/8 media compared to those in f/2

media; however, endpoints were consistently lower at higher pH for all experiments

(Table 10).

Prymnesin-1 and -2 Physicochemical Properties

The structures of prymnesin-1 and -2 (Figure 14) lead to an estimated pKa value of 8.9 for both prymnesin-1 and -2 (Table 11). LogD between pH 5.5 and 8.5 ranged between 2.8 and 5.2 for prymnesin-1, and 2.5 and 4.9 for prymnesin-2, respectively. At

pH 6.5 approximately 16% of the prymnesins are predicted to be ionized, whereas at pH

8.5 only 0.002% are predicted to be ionized (Table 11).

Discussion

Our studies with laboratory cultures and samples from reservoirs experiencing P.

parvum blooms consistently indicate that toxins released by P. parvum are more potent

when exposure occurred at a higher pH of 8.5 than at lower pH. The predicted

physiochemical properties of prymnesins indicate that these toxins are weak bases (pKa =

75

100 100 f/2 Media FIltered f/2 Media Ufiiltered 80 80

60 60 survivorship(%) survivorship (%) survivorship 40 40

pH 6.5 pH 6.5 20 pH 7.5 20 pH 7.5 pH 8.5 pH 8.5

Pimphales promelas 0 Pimphales promelas 0 0 40x10 3 80x10 3 120x10 3 160x10 3 0 40x10 3 80x10 3 120x10 3 160x10 3 Prymnesium parvum cells mL -1 Prymnesium parvum cells mL -1

100 100 f/8 Media Filtered

(%) f/8 Media Unfiltered pH 6.5 pH 6.5 76 p pH 7.5 pH 7.5 hi 80 80 pH 8.5 pH 8.5 vors i 60 60 surv as survivorship (%) l 40 40 es prome l 20 20 a h mp

Pi 0 Pimphales promelas 0 0 40x10 3 80x10 3 120x10 3 160x10 3 0 40x10 3 80x10 3 120x10 3 160x10 3 Prymnesium parvum cells mL -1 Prymnesium parvum cells mL -1

Figure 18. The percent survivorship of Pimephales promelas exposed to samples of Prymnesium parvum grown under different nutrient conditions with cells (whole culture) and cells removed (filtrate).

Table 10. The LC50 value and respective 95% confidence intervals for experiments completed with Pimephales promelas and cultures of Prymnesium parvum grown in f/2 and f/8 media that were either unfiltered or filtered to remove cells.

LC50 value Upper and lower 95% Media Treatment pH (% media) confidence intervals 6.5 >100A nc

f/2 Unfiltered 7.5 35 26-46

8.5 18 12-25

6.5 >100A nc

f/2 Filtered 7.5 51 36-73

8.5 30 23-40

6.5 7 5-10

f/8 Unfiltered 7.5 1.7 1.4-2.4

8.5 0.7 0.4-1.1

6.5 4.3 3-6

f/8 Filtered 7.5 2.6 1.8-3.6

8.5 2 1.1-2.7

A = There was insufficient mortality to generate a point estimate. nc = Not calculable.

77

Table 11. The predicted physiochemical properties of prymnesin-1 and -2 based on computer modeling and hand computation.

Property Prymnesin-1 Prymnesin-2 Log P 6.0 ± 1.5 5.6 ± 1.5 Log P (hydrophobic portion) 7.5 ± 0.8 7.5 ± 0.8 Log P (hydrophilic portion) -1.7 ± 1.5 -2.0 ± 1.5 Log D (pH 6.5) 3.4 ± 1.5 3.1 ± 1.5 Log D (pH 7.5) 4.3 ± 1.5 4.0 ± 1.5 Log D (pH 8.5) 5.2 ± 1.5 4.9 ± 1.5 + % ionized (NH3 ) (pH 6.5) 16 16 + % ionized (NH3 ) (pH 7.5) 0.02 0.02 + % ionized (NH3 ) (pH 8.5) 0.002 0.002 pKa (1º amine) 8.9 ± 0.1 8.9 ± 0.1 pKa (hydroxyl groups) 13-15 13-15

8.9) and, thus, a greater proportion of the prymnesins were likely unionized in higher pH

treatment levels (e.g., 8.5). We propose that a higher proportion of prymnesins in unionized forms at pH 8.5 explains the greater toxicity observed in field and laboratory

studies. This novel explanation for pH-dependent ambient toxicity associated with P.

parvum suggests that variability in pH among and within aquatic systems may be an

important factor governing the occurrence of fish kills.

Unionized forms of contaminants often have greater propensity to cross cellular

membranes due to their lower polarity and thus are more likely to partition into

organisms (Simon and Beevers 1951, Sarrikoski et al. 1986, US EPA 1986, Fisher et al.

1999, US EPA 1999, Nakamura et al. 2008, Valenti et al 2010). The importance of

ionization state for ambient toxicity and environmental management is evidenced by the

integration of site-specific ambient water quality criteria for contaminants such as

pentachlorophenol and ammonia (US EPA 1986,1999). Ammonia, like the prymnesins,

78

has a pKa value of ~9. In addition, ammonia is a weak base and the ionization state of the compound changes appreciably across environmentally relevant surface water pH gradients (US EPA 1999). Consequently, acceptable ammonia loads in stream are 13- fold lower if the receiving system has a pH of 9 compared to a pH 6. Weak bases have a greater propensity to cross cellular membranes if the pH at which the exposure occurs approaches and surpasses the compound’s pKa value (US EPA 1999, US EPA 1986,

Simon and Beevers 1951, Nakamura et al. 2008, Fisher et al. 1999). Greater interaction with target sites (e.g., gill membranes) would increase the likelihood of adverse effects in exposed individuals; hence, prymnesin-1 and -2 would poses greater risk to aquatic life when these toxins exist predominantly as the unionized form.

In laboratory tests examining the effectiveness of ammonium and barley straw extract to control P. parvum, Grover et al. (2007) only observed toxicity in samples with pH > 8. These observations were consistent with Lindholm et al. (1999) who observed fish kills attributed to P. parvum in a brackish-water lake when pH ranged between 8.9 and 9.4. Additional studies with P. parvum conducted under higher salinity conditions have reported similar pH influences during in vivo experiments. Shilo and Ashner (1953) noted that fish were 5-times more sensitive at pH 9 compared to pH 6, and demonstrated that the effects of pH manipulation were reversible during adjustments to and from pH 7 and 6. Ulitzur and Shilo (1964) investigated the toxicity of P. parvum toxins, along with various chemicals identified as cofactors, over a range of pH 7 to 9 and consistently noted markedly greater toxicity at higher pH. McLaughlin (1958) observed that high pH shortened the exposure time associated with onset of mortality. The pH-dependent activity of P. parvum toxins could also reduce internal damage to cells that are producing

79

or storing toxins. Extracts of P. parvum induced “self-toxicity,” reducing growth rates and causing lysis (Olli and Trunov 2007). It is plausible that these ionizable toxins are stored inside cells of P. parvum at lower physiological pH, and are thus more ionized than when released outside the cell, where pH values may be higher.

Some previous results from in vitro hemolytic experiments with P. parvum contradict the results of the in vivo experiments reported here and elsewhere. Blood cells rupture more often when exposures are completed at pH < 6 (Igarashi et al. 1996, 1998

Kim and Padilla 1977). Prymnesin-1 and -2 have multiple ionizable groups so that changes in the protonation state could alter the configuration of the toxins. In turn, the interaction of prymnesins with specific binding sites in blood cells and fish gill membranes could depend on the structural configuration, which may be influenced by pH, and thus may be different among such in vivo and in vitro experiments.

Alternatively, prymnesins might not be the only, or even the most important toxins produced by P. parvum, and hemolytic activity in vitro might have different determinants than lethal activity in vivo (Schug et al. 2010).

Cell density alone has been long recognized as a poor predictor of toxicity for samples containing P. parvum (Reich and Aschner 1947, Baker et al. 2007, Grover et al.

2007), and this generalization remains apparent during monitoring in Texas reservoirs.

Ionization state of the toxins may partially explain some of this variability and reduce uncertainty related to ecological risk assessments and risk management of P. parvum blooms. The observed pH-dependent toxicological relationships and the physiochemical properties predicted by computer modeling suggest that the toxins prymnesin-1 and -2 act as weak bases in aqueous solutions. Because their predicted pKa values are within the

80

range of variation of pH in many surface waters, modest variations in pH could have a large influence on toxicity.

The production of ionizable toxins offers potential advantages to P. parvum and

may be related to biochemical adaptations associated with its marine origins. The results

of our studies and others suggest that the toxins released by P. parvum are more potent to

gill-breathing organisms when exposure occurs at pH levels representative of those

measured in marine systems (e.g. pH > 8). Moreover, blooms of P. parvum and other

HABs can alter the environment and cause pH to increase through depletion of carbon

dioxide during daytime photosynthesis (Pearl 1988). In fish hatchery ponds impacted by

P. parvum, pH measurements vary by more than one unit between the daylight and

evening hours (Shilo and Shilo 1953). Thus, P. parvum not only produces toxins during

bloom formation, but could also make conditions that increase the potency of their toxins.

For example, our research team recently observed high pH levels in Lake Granbury

during a P. parvum bloom that resulted in ambient toxicity to fish, compared to lower pH

levels before and after this bloom (Roelke et al. 2010).

Considering site-specific pH may be especially important for ecological risk

assessments of P. parvum because of the inherent linkage between physiochemical

properties of waters and the organisms that inhabit them. There is far greater

spatiotemporal variability in the pH of inland waters compared to marine systems. Some

of this variability arises from natural variations in geomorphology, geochemistry, and

climate. Anthropogenic activities also influence the pH of inland waters. Inland waters

where P. parvum blooms have occurred are often affected by altered hydrology, land use

changes in the catchment, and increased nutrient loading. In the southwestern and south

81

central U.S., P. parvum blooms and fish kills are often limited to waters where pH is typically high due to an arid climate, limestone bedrock, and sparse vegetation.

Consequently, prospective ecological risk assessment approaches may be possible for predicting the occurrence of harmful blooms of P. parvum by relating watershed land-use and geography to water quality.

82

CHAPTER FIVE

Interannual Hydrological and Nutrient Influences on Diel pH in Wadeable Streams: Implications for Ecological Risk Assessment of Ionizable Contaminants.

Introduction

Climate change may reduce future water availability for semi-arid regions of the southwest United States that already experience substantial variability in natural stream flow and prolonged periods of drought (Sun et al 2008, Hurd et al 1999). Flow may be markedly decreased during periods of low precipitation, which may lead to the pooling or elimination of some lotic systems (Carpenter et al 1992, Smakhtin 2001). Drought events may also limit the assimilation capacity of systems receiving point-source discharges or cause deviations in site-specific water chemistry due to altered transient pathways of water, which could change biological and geochemical processes (Carpenter et al 1992,

Smakhtin 2001, Brooks et al 2006, Boxall et al. 2010). The most extreme scenario may include suppressed hydrologic regimes in which effluent-dominated streams are losing systems with instream flows strongly influenced or completely comprised by return flows from effluent discharges (Brooks et al 2006). The culmination of effects that reduced flow will have on the aquatic risks of contaminants is difficult to predict and presents a burgeoning challenge for environmental assessment and management efforts of aquatic systems in semi-arid regions.

Population growth, urbanization, and the intensification of agricultural operations in some semi-arid regions have led to heightened release of nutrients into surface waters

(Boxall et al 2009, Heathwaite 2010). Increased nutrient availability may cause shifts in

83

community composition and standing biomass (Marti et al 2004, Schindler et al 2006,

Burcher and Benfield 2006), thereby altering ecosystem metabolism (e.g., production and

respiration dynamics) (Connell and Miller 1984, Grimm et al 2000, Walsh et al 2005).

Resulting changes in ecosystem interactions may influence surface water quality. Other

contaminants, such as pharmaceuticals and personal care products (PPCPs),

agrochemicals, and industrial constituents or byproducts, are often associated with the

same sources responsible for eutrophication of surface waters (e.g., waste water treatment plants (WWTP), livestock rearing facilities, agricultural fields; Brooks et al 2008). The

combination of these stressors makes it challenging to predict environmental hazards

because site-specific conditions can alter the physicochemical properties, bioavailability, and toxicity of some contaminants (Farrington 1991, Walsh et al 2005, Van Wezel 1998).

Further, nutrient enrichment can modify toxicological thresholds of anthropogenic contaminants, particularly at low trophic levels (Fulton et al 2009).

More than three-quarters of the essential medicinal drugs described by the World

Health Organization and approximately one-third of modern pesticides have ionizable groups (Manallack 2007, Franco 2010). Ionization state is important for ecological risk assessment because it can influence the environmental fate and biological effects of some contaminants (Van Wezel 1998). There are inherent connotations between ionization state and lipophilicity that are exemplified by pH influence partitioning of drugs (Jollow and Brodie 1972, Kwon 2001). These relationships are further evidenced by differing

Log D and bioconcentration factors (BCF) for ionizable compounds over ranges of environmentally relevant surface water pH (Schwarzenbach et al 1993, Rand 1995,

Boethling and MacKay 2000, Hernandez and Rathinavelu 2006, Valenti et al 2010b).

84

The potential for weak ionizable bases to adversely affect organisms may be

heightened in aquatic systems located in semi-arid regions of the United States that

experience elevated surface water pH due to altered hydrology and increased primary

production. As surface water pH approaches and surpasses the pKa value of a weak base, the compound will increasingly exist in the unionized form, which is often regarded as more toxic because of its greater propensity to cross cellular membranes (Hernaddez and

Tathinovelu 2006). The importance of ionization state for ecological hazard is

emphasized by the integration of site-specific pH adjustment factors into United States

Environmental Protection Agency’s National Ambient Water Quality Criteria (NAWQC) for some contaminants. One example is ammonia, a weak base with a pKa value of 9.2.

To account for differences in pH between sites that can affect ionization state and biological effects, criteria adjustment factors are determined by relating site-specific pH to toxicological data derived from laboratory experiments completed over pH gradients

(USEPA 1986, 2009). Acceptable concentrations of ammonia in the water column,

expressed as total nitrogen (TN), may vary appreciably between sites. For example at 24°

C, the criterion maximum concentration (CMC) for surface waters at a site with no

mussels present and a pH of 6.5 is 31.4 mg TN /L, whereas comparable values for a site

with a pH of 9 is only 0.85 mg/L (US EPA 2009).

The practical constraint of implementing criteria based on pH-dependent toxicity

is defining site-specific conditions. In fact, it is critical to appreciate that the surface

water pH measurement culminates from fluctuating interactions that vary on different

spatial and temporal scales among the atmosphere, hydrology, climate, geomineralogy,

and physical morphology of watersheds over longitudinal gradients (Santschi 1988, Allan

85

1995, Rebsdorf et al 1991, Hill and Neal 1997, Fitzhugh et al 1999). Bedrock

mineralogy, soil compression, till depth, elevation gradient, precipitation patterns, and

vegetation may cause appreciable differences in surface water geochemistry between

sites (Omernik and Griffith 1991, Allan et al 1993, Omernik and Bailey 1997, Fitzhugh et

al 1999). On more finite spatial scales distinctions between surface water pH at sites may

also be realized due to additional factors, such transfers of allochthonous nutrient

subsides (Jefferies 2000), groundwater and tributary inputs (Maberly 1996), proximity to wetlands (Hunt et al 1997, Fitzhugh et al 1999), availability of acid-neutralizing materials

(Driscoll et al 1987, Fitzhugh et al 1999), and both point and non-point pollution

(Heathwaite 2010, Kim et al 2010). Although there is an understanding of how environmental heterogeneity may contribute to spatial variability in surface water pH between sites, less is known about more complex interactions among such factors that may cause temporal variability at a single site.

Interpreting relationships between the spatial dynamics of such variables and surface water pH are challenged by the effects of temporal factors that occur on various time scales. Seasonal and annual variability in precipitation may influence ionic composition of surface waters by controlling inputs of groundwater, subsurface water, and overland flow (Raxcher et al 1987, Carpenter et al 1992, Findlay 1995, Smakhtin

2001). On a more resolute temporal scale, inorganic carbon in surface waters may become uncoupled with concentrations of carbon dioxide in the atmosphere during the day if the rates of biological transformation exceed rates of physiochemical transfer between environmental compartments (Maberly 1996). Diel oscillations in pH will likely have correlates to seasonality as the succession shifts of phytoplankton and macrophyte

86

communities, coupled with changing in standing biomass, can influence the potential

carbon dioxide demand of aquatic systems (Marberly 1996). This disparity may be

especially true for effluent-dominated systems in the southwestern and south central

United States that already experience low hydrologic regimes (Brooks et al 2006). The

reduction or absence of groundwater augmentation in these systems may eliminate

important inputs of carbon dioxide. Furthermore, increased ecosystem biomass may

cause high rates of respiration while subsidies of organic material from WWTPs, pasture

lands, or other terrestrial sources may cause substantial decomposition to occur in surface waters (Moss 2010). These processes may culminate in temporary depression of pH in surface waters as concentrations of inorganic carbon in water exceed rates of degassing

(Maberly 1996). Conversely, assimilation of organic carbon driven by photosynthesis can cause elevated pH during light hours as carbon dioxide is removed from the water column and previous research has demonstrated eutrophication of surface waters may facilitate high rates of primary production that potentiate extreme daily oscillations in pH at a site (Halstead and Tash 1982, Maberly 1996, Guasch et al 1998, Kent et al 2005,

Tank et al 2009).

Researchers have previously explored the implication of daily change in pH on metal and nutrient availability in aquatic systems (Crumpton and Isenhart 1988, Garban et al 1999, Jones et al 2004, Morris et al 2005, Nimick et al 2007). In this study, we explore interannual variability of diel pH oscillation patterns across 23 stream sites in the

Brazos River, Texas, USA watershed during different hydrologic regimes and examine how pH variability throughout the day at these sites may influence risk characterization for select weak bases. Ammonia was be used as a model weak base to determine the

87

variability of allowable site-specific instream concentration of TN throughout the day due

to temporal pH differences. Measured diel changes in pH were also used to develop

aquatic toxicity predictions for the weak base sertraline, which is a pharmaceutical from

the class of selective serotonin reuptake inhibitors that has been previously shown to

exhibit pH-dependent toxicity (Valenti et al 2009). Similar assessments for other

contaminants of emerging concern are often prevented because toxicity data over

gradients of environmentally relevant pH seldom exists. We further examined

relationships between BCF values and site-specific pH to infer the potential bioaccumulation of other model weak bases.

Material and methods

Study Sites

Data were collected from 23 wadeable streams in July-August 2006 and

September -October 2007. Sites were selected in the Brazos River basin, Texas, USA

within the Cross Timbers Level III Ecoregion (King et al 2009). These sites were

purposely selected to span a gradient of phosphorus enrichment, yet still captured a full

range of natural variability in geology, drainage networks, stream size, and other physiographic factors (King et al 2009). The permitted wastewater outfalls, land use for the catchment, and potential nutrients inputs for each stream site are presented in Table

12. Several sites (e.g., Bluf-01, Harr-01, Mbos-01, Sbos-01) had >10% crop land. Some sites were located directly downstream of WWTP discharges (e.g., Leon-02, Nbos-01,

Nbos-05, Nolc-01, Nolr-01), whereas others likely had at least a portion of flow

88

comprised of WWTP effluent (e.g., Cowh-01, Nbos-02,-03,-04, Meri-01, Mbos-01, Sbos-

01, Leon-01).

Diel Water Quality Monitoring

Diel changes in dissolved oxygen (DO), pH, and temperature were determined by deploying YSI 600 XLM and YSI 6600 (YSI instruments, Yellow Springs, Ohio, USA) multiparameter datasonds at each site for 48-h. Data were collected from 23 wadeable streams in 2006 and 2007 between September and October. Data collected during 2006 was during a near-record drought and flow was minimal or absent at nearly all sites, with effluent-dominated streams as an exception. For example, annual departures from normal precipitation in the studied watershed areas in 2006 were between -102 and -406 mm, while comparable values in 2007 were between 102 and 508 mm (National Weather

Service website). Exceptional high stream flows occurred during the summer of 2007

and data collection occurred during a brief period of stream-flow recession (King et al

2009). Instruments were placed in the central channel in areas of discernable flow, if

possible. Values for each parameter were collected at 15-min intervals and data was

transferred to a computer at the completion of deployment. All instruments were calibrated with all reagents at room temperature within 24-h prior to data collection

(TCEQ 2003). For pH calibration, probes were calibrated with buffer solutions at pH 7

and 10. As an additional quality assurance following data collection, post-calibration

were completed and using error limits for pH, DO, and temperature of 0.5 standard units,

± 5% error at saturation, and ±1° C, respectively (TCEQ 2003).

89

Table 12. The location, physical descriptors, dams, wastewater outfall, and land-use breakdown for the 24 sites sampled during 2006 and 2007.

Outfalls Water Dev. Forest Shrub Grass Pasture Crop Wetland Imp. Nutrient input Site (MGD) (%) (%) (%) (%) (%) (%) (%) (%) (%) STEE-01 0 0.8 0.5 14.7 52.4 26.8 2.6 2.1 0.5 0.1 Minimal LAMP-02 0 0.2 0.7 14 53.6 29.2 1 1.1 0.3 0.1 Minimal NEIL-01 0 0.2 0.2 35.6 1.4 57.7 0.7 2.2 1.8 0.2 Minimal PALU-01 0 0.4 1.1 32.9 11 49.8 3.1 0.8 1 0.2 Minimal ROCK-01 0 0.1 1.3 24.3 39.9 34 0.1 0.1 0.3 0.2 Minimal CORY-01 0 0 1.5 26.9 5.3 59.2 1.5 3.5 2 0.2 Minimal DUFF-01 0 0.2 0.4 22.4 13.1 57.1 1.5 4.1 1.1 0.2 Minimal LAMP-01 0 0.3 0.6 10.9 56.6 28.8 1.5 0.9 0.3 0.1 Minimal COWH-01 0.06 0.1 0.7 18.8 43 33.8 1.6 1.5 0.4 0.1 2° WWTP NBOS-02 3.5 0.6 6.5 10.7 20.1 43.9 8.5 7.8 1.9 1.3 2° WWTP

90 NBOS-03 3.5 0.5 4.3 13.3 22.6 44.6 7.1 6 1.6 0.8 2° WWTP NBOS-04 3.75 0.5 3.1 18.4 13.5 55.1 4.1 3.5 1.7 0.6 2° WWTP MERI-01 0.04 0.4 0.3 32.3 0.4 63.7 0.5 1 1.4 0.2 2° WWTP BLUF-01 0 0.1 0.6 8.8 0 69.1 1.4 17.2 2.9 0 Crop HARR-01 0 0.1 12 1.6 0 42.6 3 38.6 2.1 2 Crop. MBOS-01 0.09 0.1 1.6 10.5 0 63.2 2 19.9 2.7 0.1 2° WWTP / Crop. SBOS-01 1.1 0.2 6.7 3.7 0 48.5 4.9 33.9 2.2 0.9 2° WWTP / Crop. LEON-01 3 0.8 3.1 12.2 33.1 33.3 8.4 8.5 0.8 0.3 2° WWTP / Pasture LEON-02 6.08 0.7 2.7 13.5 30.6 37.6 6.8 7.2 0.9 0.3 1° WWTP/ Pasture NBOS-01 3.5 0.6 8 11 17.6 36.8 12.1 12 1.9 1.6 1° WWTP / Pasture NBOS-05 5.28 0.5 2.5 23.8 8.5 56.2 3.4 3.1 2 0.5 1° WWTP NOLR-01 6.73 1.6 12.3 3.2 0.1 65.7 8.8 5.9 2.3 2.5 1° WWTP/Pasture NOLC-01 33.77 0.5 32.4 24 7.9 30.2 1.4 1.3 1.6 11.7 1° WWTP

Nutrient Measurements

Determination of total phosphorus was conducted using the molybdate-blue

method in persulfate digested samples. Total nitrogen in water samples was determined

by analysis of nitrate plus nitrite-nitrogen using the cadmium reduction method in

persulfate-digested samples. A Lachat Quickchem 8500 Flow Injection Autoanalyzer was

use for quantification. 250 ml unfiltered samples were preserved with concentrated

H2SO4 (pH < 2) and stored in the refrigerator for a period no longer than 28 days.

pH Influences on Aquatic Toxicity

The mean, minimum, and maximum daily pH values and the measured values at

0800, 1100, 1400 and 1700 hrs were determined for each site. In addition, cumulative

frequency distributions of daily instream pH measurements were created so that the

likelihood of observing a given pH value at each site could be quantified. The pH data

for discrete measures (mean, minimum, maximum, specific time points) and the values

for each cumulative frequency distribution were then used to examine site-specific

NAWQC for ammonia (Eq. 1; (USEPA 2009).

Dissertation (2.087Mb) (2024)
Top Articles
Latest Posts
Article information

Author: Greg Kuvalis

Last Updated:

Views: 6666

Rating: 4.4 / 5 (55 voted)

Reviews: 86% of readers found this page helpful

Author information

Name: Greg Kuvalis

Birthday: 1996-12-20

Address: 53157 Trantow Inlet, Townemouth, FL 92564-0267

Phone: +68218650356656

Job: IT Representative

Hobby: Knitting, Amateur radio, Skiing, Running, Mountain biking, Slacklining, Electronics

Introduction: My name is Greg Kuvalis, I am a witty, spotless, beautiful, charming, delightful, thankful, beautiful person who loves writing and wants to share my knowledge and understanding with you.