A chemical prioritization process: Applications to contaminants of emerging concern in freshwater ecosystems (Phase I)

Abstract

Contaminants of emerging concern (CECs), such as pharmaceuticals, personal care products, and hormones, are frequently found in aquatic ecosystems around the world. Information on sublethal effects from exposure to commonly detected concentrations of CECs is lacking and the limited availability of toxicity data makes it difficult to interpret the biological significance of occurrence data. However, the ability to evaluate the effects of CECs on aquatic ecosystems is growing in importance, as detection frequency increases. The goal of this study was to prioritize the chemical hazards of 117 CECs detected in subsistence species and freshwater ecosystems on the Grand Portage Indian Reservation and adjacent 1854 Ceded Territory in Minnesota, USA. To prioritize CECs for management actions, we adapted Minnesota Pollution Control Agency’s Aquatic Toxicity Profiles framework, a tool for the rapid assessment of contaminants to cause adverse effects on aquatic life by incorporating chemical-specific information. This study aimed to 1) perform a rapid-screening assessment and prioritization of detected CECs based on their potential environmental hazard; 2) identify waterbodies in the study region that contain high priority CECs; and 3) inform future monitoring, assessment, and potential remediation in the study region. In water samples alone, 50 CECs were deemed high priority. Twenty-one CECs were high priority among sediment samples and seven CECs were high priority in fish samples. Azithromycin, DEET, diphenhydramine, fluoxetine, miconazole, and verapamil were high priority in all three media. Due to the presence of high priority CECs throughout the study region, we recommend future monitoring of particular CECs based on the prioritization method used here. We present an application of a chemical hazard prioritization process and identify areas where the framework may be adapted to meet the objectives of other management-related assessments.

Graphical Abstract

Fish are a primary subsistence food used by the Anishinaabeg (people) of Grand Portage Band of Lake Superior Chippewa historically and presently and thus sets the context for this paper exploring potential impacts of contaminants on this culturally important resource. The Grand Portage Band is a federally recognized Indian tribe in extreme northeastern Minnesota and proudly exercises its rights to food sovereignty through subsistence hunting and fishing.

1. Introduction

Contaminants of emerging concern (CECs) are a diverse group of chemicals – often defined as chemicals that were previously unknown, unrecognized, or unregulated – that are generally poorly understood with respect to transport, fate, and toxicity in the environment (Nilsen et al., 2019). Many of these chemicals, including pharmaceuticals, personal care products, and hormones, are widely detected in aquatic environments and have been found at toxicologically relevant concentrations in wastewater and other areas with high human disturbance (Blair et al., 2013; Fairbairn et al., 2018; Kiesling et al., 2019) as well as in remote, less developed regions (Deere et al., 2020; Elliott & VanderMeulen, 2017; Ferrey et al., 2015; 2020). Typical wastewater treatment plant (WWTP) technologies are not designed to remove CECs; thus, most CECs remain in wastewater effluent after treatment (Rizzo et al., 2019). While WWTPs are often considered the main anthropogenic point source of CECs to surface waters, the occurrence of CECs in remote areas has only recently been investigated. Atmospheric transport of these CECs and their deposition through precipitation likely play a role in the appearance of CECs in remote locations (Ferrey et al., 2018). With growing evidence that nonpoint sources of pollution contribute to CEC presence in rural environments, assessing the effects they might have on aquatic systems is important.

While data regarding acute lethality from high concentrations of some CECs in aquatic organisms does exist (Fent et al., 2006; Santos et al., 2010), information on sublethal effects, such as neuroendocrine or immune effects, due to exposure to concentrations that are commonly detected in the environment is generally lacking (Nilsen et al., 2019). In addition to effects from the acute or chronic exposure to single CECs, cumulative effects from exposure to mixtures are also of concern, especially when physiological effects become demographically and ecologically significant (Adeel et al., 2017; Thelusmond et al., 2018). The ability to evaluate the individual and interactive effects of CECs on fish, aquatic ecosystems, and ultimately all forms of life is critical, particularly as the frequency of their occurrence continues to rise in natural waterbodies around the world (Baker & Kasprzyk-Hordern, 2013; Battaglin et al., 2018). However, the limited availability of toxicity data leading to the lack of regulatory or screening values for most CECs makes it difficult to interpret the biological significance of occurrence data.

A study initiated by the Grand Portage Band of Lake Superior Chippewa to explore potential contaminant threats to their natural resources and subsistence species found 117 CECs in aquatic environments on the Grand Portage Indian Reservation (GPIR) and adjacent 1854 Ceded Territory in Minnesota, USA (Deere et al., 2020). Its findings raised additional questions about the sources of these chemicals and about their potential hazards on the biological systems of subsistence fish species on which the Tribe depends and to which the Tribe’s culture is inextricably linked. The goal of this study is to prioritize the 117 detected chemicals by the potential hazards they pose to subsistence species and aquatic ecosystems and, thus, to the Ojibwe culture and way of life.

Minnesota Pollution Control Agency’s Aquatic Toxicity Profiles (ATPs) (Streets & Dobbins, 2017) is a rapid assessment tool that incorporates chemical-specific information, including acute toxicity, endocrine activity, physicochemical properties, and frequency of occurrence data, in the evaluation of the potential for environmental contaminants to cause adverse effects on aquatic life. Given the conservative thresholds employed, this non-regulatory screening tool conservatively estimates the potential for a chemical to be hazardous in a way that guards against type II errors (i.e., to falsely conclude no potential adverse effects). The process is primarily anchored in adverse effects at the level of the organism, which is different from other recent CEC prioritization processes based on bioactivity at the molecular level (Corsi et al., 2019). We adapted the ATP framework to 1) perform a rapid-screening assessment and prioritization of detected CECs based on their potential environmental hazard; 2) identify waterbodies in the study region that contain high priority CECs; and 3) inform future monitoring, assessment, and potential remediation in the study region. Here, priority is a relative term in which detected contaminants are ranked against one another based on available information. We hypothesized that, through this framework of prioritization, a subset of higher priority chemicals would be identified. In so doing, research and policy decisions may be more tenable. We present a research-focused application of a chemical hazard prioritization process and identify areas where the framework may be adapted to meet the objectives of other management-related assessments.

2. Methods

2.1. Study design

Study region, site selection, sample collection, and analytical procedures were described previously (Deere et al., 2020). Briefly, we surveyed the presence of CECs in two regions in northeastern Minnesota, the GPIR and the 1854 Ceded Territory. Across three years (2016–2018), we sampled water, sediment, and fish at sites categorized by anthropogenic pressure: 1) wastewater effluent-impacted, which included sites that received discharge from a wastewater treatment plant; 2) developed, which included sites with any level of shoreline development, including human residences and business; and 3) undeveloped, which included sites with no shoreline development. Candidate sites were selected based on land use, proximity to potential point sources, presence of subsistence fish species, and importance of the location for fish harvest by tribal members. Sites were further selected based on multi-criteria decision analysis (Convertino & Valverde, 2013; Deere et al., 2020). We sampled 14 waterbodies in 2016, 19 in 2017, and 2 in 2018, leading to a total of 28 unique aquatic sampling locations across all years. We collected a total of 33 water and sediment samples each and 51 fish tissue samples. Sites included inland lakes as well as locations along the Lake Superior northwestern shore. Samples were collected using Polar Organic Chemical Integrative Samplers (POCIS) for water and grab samples for sediment. POCIS are passive samplers that represent the respiratory and dermal exposure of aquatic organisms to dissolved chemicals over a given time period. We utilized a combination of methods to collect fish samples, including boat-operated electrofishing and gill nets, targeting important subsistence and recreational fish species of different trophic levels. Complete site selection and sampling details are described in Deere et al. (2020).

2.2. Chemical prioritization

Complete ATP methods are described in Minnesota’s Aquatic Toxicity Profiles: Methods and application (Streets & Dobbins, 2017). Briefly, the ATP process applies the assembly of data from a combination of publicly available databases, modeling tools, monitoring data, and limited literature searches to characterize chemicals on the basis of production volume, persistence and prevalence in the environment, potential for accumulation, and biological effects such as lethality and endocrine disruption. Each of these parameters is captured categorically in a yes/no question format, where a “yes” answer receives a score of 1 and “no” answer receives a score of 0. The scores are then summed by chemical to provide an overall priority level. The ATP questions include: 1) Is the contaminant persistent in the environment?; 2) Does the contaminant have the potential to accumulate?; 3) Is the chemical toxic?; 4) Do detected concentrations exceed toxicity?; 5) Is there evidence of endocrine disruption?; and 6) Is this a high production volume chemical?

We adapted these questions to meet the objectives of the current study. ATPs look at a multitude of factors, some of which are not toxicity; therefore, we focused on utilizing ATPs for hazard identification. We created chemical profiles for CECs detected in each medium (i.e., water, sediment, and fish separately), as well as profiles for CECs detected in water, sediment, or fish (“any media” category). Some questions were adjusted based on availability of data. For example, question 4 (Do detected concentrations exceed toxicity?) could only be answered with fish tissue data and not water data. We did not have POCIS sampling rates, which are necessary to calculate water concentrations from POCIS measurements (Godlewska et al., 2020).

The priority level is based on six questions for water and fish and five questions for sediment and “any media” (Table 1). Therefore, CECs detected in water and fish can have a priority level of up to six and those detected in sediment or “any media” can have a priority level of up to 5. The priority levels range from high (receiving a score of 4–6), intermediate (a score of 3), or low (scores of 0–2). Available data among prioritization questions were inconsistent; therefore, a low priority level might reflect lack of knowledge about a particular chemical rather than lack of concern. This interpretation reflects the focus of prioritizing CECs for advising management decisions rather than on identifying research needs.

Table 1.

Questions answered to prioritize contaminants (adapted from the original ATP framework). If the answer to the question is “yes” it receives a score of 1; if “no” it receives a score of 0. The higher the score, the higher the priority level. T_1/2 = chemical half-life. pKa = acid dissociation constant. TSC = toxic tissue screening concentration.

Media	Question	Criteria	Data Source
All	Is the contaminant persistent in the environment?	T1/2 water > 2 months	EPI Suite
		T1/2 sediment > 6 months
		T1/2 soil > 6 months
		T1/2 air > 2 day
All	Does the contaminant have the potential to accumulate?	log10 KOW ≥ 4	EPI Suite
		log10 KOC ≥ 3
All	Is the chemical’s aqueous toxicity high?a	Acute toxicity value (μg/L) < 10,000	ECOSARc
		Chronic toxicity value (μg/L) < 100
All	Is there evidence of potential endocrine disruption?	Any active estrogen, androgen, or thyroid assays	EDSP21
All	Is the contaminant detected in more than 20% of samples?	Detected in at least	Current study; Deere et al. 2020
		6/28 sites (any media)
		7/33 samples (water and sediment)
		11/51 samples (fish)
Water	Is the chemical neutral at the pH of the water > 10% of the time?	Ionization ratio < 1	CompTox Chemistry Dashboard
Fish	Do fish tissue concentrations exceed toxicity thresholds?b	Maximum concentration > TSC	ECOSAR and CompTox Chemistry Dashboard

^aChemical toxicity expressed as aqueous toxicity in order to provide a standardized ranking system irrespective of the media in which it was detected.

^bThis question is typically asked of all media, but due to availability of data we answered this question for fish only, based on an adapted method.

^cECOSAR toxicity is based on narcosis through aqueous exposure; thus, does not represent all forms of toxicity

2.2.1. Is the contaminant persistent in the environment?

To determine whether each contaminant is predicted to rapidly biodegrade or if it is persistent in the environment, we used the U.S. Environmental Protection Agency (EPA) Estimation Programs Interface (EPI) Suite (U.S. EPA, 2020a), a freely available tool containing physicochemical property and environmental fate models for organic chemicals. EPI Suite has undergone thorough review by a panel of EPA’s independent Science Advisory Board (Morgan & McFarland, 2007). Persistence, in the context of this prioritization method, is described as the half-life of the chemical in water, sediment, soil, or air (Webster et al., 1998). To interpret degradation potential, we used the following half-life (t_1/2) thresholds in water, sediment, soil, and air: t_1/2 in water > 2 months, t_1/2 in sediment > 6 months, t_1/2 in soil > 6 months, and t_1/2 in air > 2 days, respectively (Streets & Dobbins, 2017). If there was evidence of half-lives exceeding any of these thresholds or if the chemical was not predicted to be readily biodegradable in EPI Suite, the contaminant was considered persistent in the environment, thus receiving a score of 1. Half-lives were predicted using the LEV3EPI™ program that contains a level III multimedia fugacity model (Parnis & Mackay, 2021).

2.2.2. Does the contaminant have the potential to accumulate?

Bioaccumulation and sediment accumulation potential were determined using octanol-water and organic carbon-water partition coefficients (K_OW and K_OC, respectively). Partition coefficient values were obtained from EPI Suite. A chemical was considered to have the potential to accumulate in biota and/or sediment if it had measured or predicted partitioning properties that exceeded the following guidelines: log10 K_OW ≥ 4 or log10 K_OC ≥ 3 (United Nations, 2003; OECD, 2001). If there was evidence of measured or predicted partitioning properties that exceeded either of these thresholds, the chemical received a score of 1.

2.2.3. Is the chemical’s aqueous toxicity high?

As this method is intended to be rapid and because measured toxicity data (from peer-reviewed literature, government documents, or EPA’s ECOTOX (https://cfpub.epa.gov/ecotox/)) were not available for many of the detected chemicals in this study, we used modeled toxicity values from EPA’s ecological structure activity relationships (ECOSAR) to assess toxicity of each contaminant. ECOSAR is a QSAR model that uses lipid solubility (i.e., K_ow) to predict a chemical’s acute and chronic toxicity to fish and other aquatic organisms based primarily on a narcotic mode of action (Mayo-Bean et al., 2012). Narcosis (i.e., baseline toxicity) is a common mode of action in approximately 40% of organic chemicals (Kienzler et al., 2019). ECOSAR training sets include chemicals with log K_ow values in the range of −3 to 8 and molecular weights less than 1000; therefore, the chemicals in this study are included in ECOSAR’s domain applicability (Tables S1–S4) (Mayo-Bean et al., 2012). Further, ECOSAR models for other toxicities (in addition to narcosis), including specifically acting organic chemicals causing “excess toxicity” and surface-active compounds.

Although ECOSAR is recognized as a robust quantitative structure-activity relationship (QSAR) for aquatic toxicity, we intended for this process to be used for screening purposes; thus, we took steps to check the likelihood that our screening method was conservative (i.e., reduced the chance for a misclassification error by placing a chemical with high toxicity into a lower toxicity category). We therefore included a supplementary assessment with the toxicity estimation software tool (TEST) (https://www.epa.gov/chemicalresearch/toxicity-estimation-software-tool-test). Like ECOSAR, TEST is a QSAR, but it incorporates additional molecular descriptor methods for toxicity estimation including hierarchical method, nearest neighbor method, and more (Martin, 2016). We used the predicted toxicity values from the consensus method, which is an average of predicted toxicities from multiple QSAR methods.

The ECOSAR-based toxicity assessment was used for chemicals detected in any media in order to standardize the relative toxicity ranking system. Acute values were classified as “very toxic” if ≤ 1,000 μg/L or “toxic” if > 1,000 to ≤ 10,000 μg/L. Chronic values were classified as “very toxic” if ≤ 10 μg/L or “toxic” if > 10 to ≤ 100 μg/L (Streets & Dobbins, 2017). Therefore, if the acute toxicity value was ≤ 10,000 μg/L and/or the chronic toxicity value was ≤ 100 μg/L, then the chemical was classified as “toxic.” Acute effects were obtained from the lowest values from fish (96-hour) or daphnid (48-hour) lethal concentration 50% (LC50) values. Chronic effects were obtained from the lowest values from fish or daphnid Chronic Values (ChV). Values used and their ECOSAR class are provided in Table S1.

The acute toxicity thresholds were obtained from internationally harmonized classification systems (United Nations, 2003; OECD, 2001). Harmonized classification systems for chronic effects are not available, so the categories used for chronic values were derived from the acute categories. Chronic effects typically occur at concentrations lower than acute effects (U.S. EPA OW/ORD Emerging Contaminants Workgroup, 2008), and many studies have evaluated the ratio between acute and chronic effects. The acute-to-chronic ratio varies depending on the species and chemical tested and can encompass a wide range of values. An acute-to-chronic extrapolation of 100 has been demonstrated to be protective for greater than 90% of evaluated chemicals, while an acute-to-chronic extrapolation of 10 may only be protective for approximately 50% of chemicals (May et al., 2016). Similar results have been reported with a 90th percentile of acute-to-chronic ratios close to 100 (73–80) (Lange et al., 1998; Raimondo et al., 2007). An acute-to-chronic conversion of 100 was used in this preliminary screening.

2.2.4. Is there evidence of potential endocrine disruption?

The presence of contaminants in aquatic systems may disrupt the endocrine system of organisms at concentrations lower than what may cause toxic effects such as death or decreased growth (Niemuth & Klaper, 2018; Jiaying Wang et al., 2018). Therefore, we utilized the EPA’s Endocrine Disruptor Screening Program (EDSP) to review any potential endocrine effects. EDSP data can be accessed through EPA’s Chemistry Dashboard (U.S. EPA, 2020b); specifically, the EDSP21 section under the “Bioactivity” tab of the Dashboard. If there was any evidence of activity in the assays, it was considered evidence of potential for endocrine disruption and the chemical received a score of 1. The EDSP assesses chemicals for endocrine-related activity. The activity of a chemical in a specific assay does not relate to whole organism toxicity, but rather the potential for the chemical to affect endocrine pathways, which may induce an adverse health outcome.

2.2.5. Is the contaminant detected in more than 20% of samples?

The original ATP framework prioritizes chemicals often in the absence of occurrence data; thus, it relies on the EPA (U.S. EPA, 2020c) and Organization for Economic Co-operation and Development (OECD, 2009) lists of high production volume chemicals as data sources to characterize the likelihood that a chemical will be present in the environment. Both sources identify chemicals that are produced in or imported to the U.S. at a rate of at least 1 million pounds per year. If a chemical is included on either or both of these lists, then it would receive a score of 1. To evaluate our site-specific occurrence data, we modified this question to ask: “Is the contaminant detected in more than 20% of samples?” For the “any media” priority level, we assessed detection frequency by site; if the chemical was detected in at least one medium per site, then this question was given an answer of “yes.” Therefore, if a contaminant was detected in any media at six or more of the 28 sites we examined in our study, it was given a score of 1, indicating that it was detected more frequently than other contaminants. For individual medium (water, sediment, and fish) priority levels, if the chemical was detected in the respective medium samples at a frequency of at least 20%, then this question was given a score of 1 for that chemical.

If we used the original ATP question, “Is this a high production volume chemical?,” some of the results might have changed. For example, caffeine was detected in approximately 15% of water samples, so was given a score of 0 in the modified question. However, caffeine is listed as a high production volume chemical so it would have been given a score of 1 using the original ATP question. Whereas gemfibrozil, which was detected in approximately 30% of water samples and was given a score of 1 for our chemical profile, is not listed as a high production volume chemical so would have been given a score of 0 using the original ATP question.

2.2.6. Is the chemical neutral at the pH of the water > 10% of the time?

The aquatic toxicity of ionizable organic compounds, such as most pharmaceuticals, is dependent on water pH (Escher et al., 2020). Chemicals that are neutral at the pH of the water containing them are more likely to be absorbed by aquatic organisms than chemicals that are not neutral because of their increased tendency to cross cell membranes (Alsop & Wilson, 2019). The acid dissociation constant (pKa) of the chemical indicates whether the chemical will be neutral at a given pH (Babić et al., 2007). To determine whether a chemical was neutral at the pH of the water, which was measured at the time of sampling, we calculated an ionization ratio (or acid/base ratio) using available pH data from the study lakes at time of sampling and pKa estimates obtained from EPA’s Chemistry Dashboard (U.S. EPA, 2020b), as modeled using OPERA (Mansouri et al., 2019):

$$\frac{A^{-}}{AH}={10}^{pH-pKa}$$

where A⁻ is the concentration of the conjugate base and AH is the concentration of the conjugate acid. If the ionization ratio was less than 1, then the chemical was considered neutral. The estimate is conservative, as the ionization state is a continuum rather than a threshold phenomenon. This ATP question was applied only to chemicals detected in water samples. Further, pH data was available for water sampled in 2017 but not 2016 or 2018, which included 19 of 28 sites. While most of the chemicals detected in 2017 were also detected in 2016 and 2018, we were not able to assess the ionization potential for 12 chemicals either because we did not have pH data (6 chemicals) or pKa estimates (6 chemicals). Based on the available data, if a chemical was neutral at least 10% of the time, then it was given a score of 1 for this question.

2.2.7. Do fish tissue concentrations exceed toxicity thresholds?

Tissue toxicity thresholds for our detected CECs were limited. Therefore, to determine if fish tissue concentrations may have exceeded toxicity thresholds, we estimated tissue screening concentrations (TSC) (Dyer et al., 2000). The TSC (μg/kg) for each chemical is a product of the chronic toxicity value (μg/L) and the bioconcentration factor (BCF) (L/kg):

$$TSC=chronictoxicityvalue\ast BCF$$

BCF values (the ratio of the concentration of the chemical in fish tissue to the concentration in water) were obtained from the Chemistry Dashboard (U.S. EPA, 2020b). Bioaccumulation factor (BAF) estimates could also be used in this context (Costanza et al., 2012); however, as BAF and BCF values did not yield different rankings, we chose to use BCF values in this evaluation. Additionally, as empirical (i.e., not modeled) BCF values are experimentally determined using standard protocols, they are more readily comparable across chemicals than BAF values.

2.3. Statistical Analysis

All data analysis was performed with R Version 4.0.2 (R Core Team, 2020). Following a significant Kruskal-Wallis test, Dunn’s test of multiple comparisons was performed post-hoc to explore differences between priority levels and anthropogenic pressure categories, using the dunn.test package (Dinno, 2017). P-values were adjusted using the Benjamini-Hochberg method (Benjamini & Hochberg, 1995). We assessed correlation among profile questions using Spearman’s rho rank correlation coefficients (“Spearman Rank Correlation Coefficient,” 2008) in the Hmisc package (Harrell, 2020) and created correlation plots in the corrplot package (Wei & Simko, 2017).

As part of the descriptive summary, we assigned all detected contaminants to primary use categories, as previously described (Deere et al., 2020). Briefly, we used the World Health Organization (WHO) Anatomical Therapeutic Chemical classification system (WHO, 1993). Chemicals were first classified based on their anatomical or pharmacological groups and then assigned into the primary use categories we created. For those chemicals not in the WHO database, we classified them according to their classification in published literature.

3. Results

Chemical profiles were assembled for 117 CECs detected in aquatic systems in 28 northeastern Minnesota waterbodies (lakes) (Tables S2 – S4). Across any media (water, sediment, and/or fish), 38 chemicals were deemed high priority (Table S5). In water (POCIS) samples alone, 50 CECs received a high priority level (Table 2). Twenty-one contaminants ranked as high priority among sediment samples (Table 3), and seven chemicals were high priority in fish samples (Table 4).

Table 2.

High priority contaminants detected in water samples collected through POCIS. Maximum possible priority level = 6.

Chemical	Primary use	Priority level
Diphenhydramine	Antihistamine	6
Estrone	Hormone	6
17 alpha-Estradiol	Hormone	5
17 beta-Estradiol	Hormone	5
Atorvastatin	Cardiovascular modulating agent	5
Benztropine	Anticholinergic	5
Bisphenol A	Plastic residue	5
Citalopram	Antidepressant	5
Fluoxetine	Antidepressant	5
Gemfibrozil	Cardiovascular modulating agent	5
Hydrocodone	Opioid analgesic	5
Metoprolol	Cardiovascular modulating agent	5
Paroxetine	Antidepressant	5
Roxithromycin	Antimicrobial	5
Tamoxifen	Antineoplastic	5
Triclosan	Disinfectant	5
Verapamil	Cardiovascular modulating agent	5
17 alpha-Ethinyl-Estradiol	Hormone	4
Albuterol	Bronchodilator	4
Allyl Trenbolone	Hormone	4
Alprazolam	Antianxiety	4
Amitriptyline	Antidepressant	4
Amlodipine	Cardiovascular modulating agent	4
Amphetamine	Stimulant	4
Androstenedione	Hormone	4
Azithromycin	Antimicrobial	4
Carbamazepine	Antiepileptic	4
Clotrimazole	Antimicrobial	4
Cocaine	Stimulant	4
Codeine	Opioid analgesic	4
DEET	Insect repellant	4
Desogestrel	Hormone	4
Diltiazem	Cardiovascular modulating agent	4
Equilenin	Hormone	4
Glyburide	Antidiabetic	4
Mestranol	Hormone	4
Miconazole	Antimicrobial	4
Progesterone	Hormone	4
Promethazine	Antihistamine	4
Propoxyphene	Opioid analgesic	4
Propranolol	Cardiovascular modulating agent	4
Ranitidine	Antacid	4
Sertraline	Antidepressant	4
Sulfamethazine	Antimicrobial	4
Sulfathiazole	Antimicrobial	4
Testosterone	Hormone	4
Triamterene	Cardiovascular modulating agent	4
Triclocarban	Disinfectant	4
Trimethoprim	Antimicrobial	4
Venlafaxine	Antidepressant	4

Table 3.

High priority contaminants detected in sediment samples. Maximum possible priority level = 5.

Chemical	Primary use	Priority level
Clotrimazole	Antimicrobial	5
Fluoxetine	Antidepressant	5
Triclocarban	Disinfectant	5
17 alpha-Ethinyl-Estradiol	Hormone	4
Androsterone	Hormone	4
Atorvastatin	Cardiovascular modulating agent	4
Azithromycin	Antimicrobial	4
Benztropine	Anticholinergic	4
Bisphenol A	Plastic residue	4
DEET	Insect repellant	4
Diphenhydramine	Antihistamine	4
Estrone	Hormone	4
Gemfibrozil	Cardiovascular modulating agent	4
Miconazole	Antimicrobial	4
Paroxetine	Antidepressant	4
Progesterone	Hormone	4
Promethazine	Antihistamine	4
Sertraline	Antidepressant	4
Tamoxifen	Antineoplastic	4
Triclosan	Disinfectant	4
Verapamil	Cardiovascular modulating agent	4

Table 4.

High priority contaminants detected in fish tissue samples. Maximum possible priority level = 6.

Fish	Primary use	Priority level
DEET	Insect repellant	5
Azithromycin	Antimicrobial	4
Diphenhydramine	Antihistamine	4
Fluoxetine	Antidepressant	4
Miconazole	Antimicrobial	4
Roxithromycin	Antimicrobial	4
Verapamil	Cardiovascular modulating agent	4

We grouped the detected CECs into 23 primary use categories to assess the number of high priority contaminants within primary use categories (Table 5). For high priority contaminants, there was more diversity among primary use categories detected in water (n=17) than in sediment (n=10) or fish (n=5).

Table 5.

Number of high priority contaminants (priority level 4, 5, or 6) in each primary use category and media. N = the number of possible compounds detected in each category.

Primary use category	N	Water	Sediment	Fish
Antacid	2	1	0	0
Antianxiety	4	1	0	0
Anticholinergic	1	1	1	0
Anticoagulant	1	0	0	0
Antidepressant	8	6	3	1
Antidiabetic	3	1	0	0
Antiepileptic	1	1	0	0
Antigout	1	0	0	0
Antihistamine	2	2	2	1
Antimicrobial	28	7	3	3
Antineoplastic	5	1	1	0
Bronchodilator	2	1	0	0
Cardiovascular modulating agent	18	8	3	1
Contrast agent	2	0	0	0
Disinfectant	2	2	2	0
Hormone	20	11	4	0
Immunosuppressant	1	0	0	0
Insect repellant	1	1	1	1
Nicotine metabolite	1	0	0	0
Non-opioid analgesic	4	0	0	0
Opioid analgesic	4	3	0	0
Plastic residue	1	1	1	0
Stimulant	5	2	0	0
Total high priority contaminants	-	50	21	7

We detected high priority contaminants in all categories of sites we sampled (i.e., undeveloped, developed, and wastewater effluent-impacted) (Figure 1). Wastewater effluent-impacted sites contained the most contaminants ranked as high priority contaminants, with sites ranging from 11 to 30 high priority contaminants. Developed and undeveloped sites contained a range of 2 to 13 high priority contaminants per site. Within all priority levels, wastewater effluent-impacted sites contained a significantly higher mean number of detections than both developed and undeveloped sites (Table 6). There were no significant differences between developed and undeveloped sites within all priority levels (Kruskal-Wallis P = 0.4208, P = 0.4684, P = 0.3535 for high, intermediate, and low priority levels, respectively). Within developed and undeveloped sites, there was a significantly higher mean number of detections of high priority level contaminants than both intermediate and low priority contaminants (Table 6). There were no significant differences among any priority levels within wastewater effluent-impacted sites (P = 0.4130).

Figure 1.

Total number of contaminants, and their priority level across any media, detected among undeveloped, developed, and wastewater effluent-impacted sites sampled from 2016 to 2018 in northeastern Minnesota on the Grand Portage Reservation and 1854 Ceded Territory.

Table 6.

Mean number of contaminants detected across sites by anthropogenic pressure and priority level. Superscripts represent significant differences.

	Mean detections
	Undeveloped	Developed	Wastewater effluentimpacted
Low	1.9	2.1	19.2a
Intermediate	1.6	1.5	17.2b
High	5.7d	5.3e	22.7c

^aWastewater effluent-impacted significantly higher than developed (P = .0011) and undeveloped (P = .0011) within low priority level.

^bWastewater effluent-impacted significantly higher than developed (P = .0007) and undeveloped (P = .0007) within intermediate priority level.

^cWastewater effluent-impacted significantly higher than developed (P = .0009) and undeveloped (P = .0012) within high priority level.

^dHigh priority level significantly higher than intermediate (P = .0026) and low (P = .0040) priority levels within undeveloped sites.

^eHigh priority level significantly higher than intermediate (P = .0003) and low (P = .0038) priority levels within developed sites.

The proportion of high priority contaminants out of the total number of detections by site exemplifies the magnitude of high priority CECs across all anthropogenic pressure categories. This relationship can be seen spatially, with many undeveloped locations containing more than 50% high priority contaminants out of the total CECs detected (Figure 2), particularly on and near the GPIR. Note that 108 of the 117 detected CECs in this study are predicted to readily biodegrade, so the presence of high priority contaminants in remote areas is not simply due to persistence.

Figure 2.

Number of detections relative to percent of high priority contaminants, across any media, per sites sampled from 2016 to 2018 in northeastern Minnesota on the Grand Portage Reservation and 1854 Ceded Territory. There were 38 high priority contaminants detected across any media. Sites are offset for visual representation. The size of the circle indicates the number of detections. The color of the circle represents the percent of high priority contaminants (priority level 4, 5, or 6) in any media. The symbol next to each site symbolizes the respective anthropogenic pressure category: developed, undeveloped, or wastewater effluent-impacted.

We explored how choices in the methodology of the hazard prioritization method used here might impact priority levels and provided results of these assessments in the Supplementary Information. Based on data availability, only one to four criteria could be evaluated for 12 contaminants; thus, these chemicals were ranked as low or intermediate priority and should be interpreted with caution: 10-hydroxy-amitriptyline, 2-hydroxy-ibuprofen, clinafloxacin, desmethyldiltiazem, drospirenone, equilin, fluticasone propionate, lomefloxacin, moxifloxacin, norfluoxetine, norverapamil, virginiamycin. To highlight which chemicals might be ranked differently based on the availability of data, we normalized the scores according to the number of questions that could be answered. For water, four chemicals moved to a high priority level when scores were normalized: 10-hydroxy-amitriptyline, drospirenone, norfluoxetine, and norverapamil (Table S6). Equilin was the only chemical that changed priority level, from intermediate to high, when sediment contaminant scores were normalized (Table S7). For fish, 10-hydroxy-amitriptyline and fluticasone propionate changed from low to intermediate priority when scores were normalized (Table S8). No chemicals changed in priority level from a higher to a lower level after normalization.

We performed a sensitivity analysis on the question “Is the contaminant detected in more than 20% of samples?” by changing the detection frequency threshold to 15% and 25% and determined how this affects priority levels. Thirty-four chemicals changed priority levels (either up or down) when the detection frequency in water samples was adjusted (Table S9). When detection frequency in water samples was 15%, 23 chemicals move up a priority level and when detection frequency was 25%, 11 chemicals move down a priority level. For sediment, 12 chemicals changed priority levels: eight chemicals moved up a priority level when the detection frequency was 15% and four chemicals moved down a priority level when detection frequency was 25% (Table S10). Three chemicals changed priority levels when the detection frequency in fish samples was adjusted; all moving up a priority level when the detection frequency was 15% (Table S11). Changing the detection frequency to 25% did not affect the priority levels of and chemicals detected in fish.

To ensure individual questions that made up the profiles were not highly correlated, we assessed correlation coefficients among profile questions and found little correlation (Figures S1 – S3). Among significant relationships, correlation coefficients between water profile questions ranged from 0.24 – 0.30; for sediment profiles, correlation coefficients ranged from 0.28 – 0.39; for fish, correlation coefficients ranged from 0.39 – 0.47.

To assess the conservative nature of our screening process, we compared TEST toxicity value predictions to the ECOSAR values used in the final profiles (see Table S1 for TEST toxicity values in comparison to ECOSAR values). There were six chemicals that would have been high priority if we would have used TEST values instead, so we have flagged these chemicals for more follow-up actions (marked with an asterisk in Tables S2 – S4).

4. Discussion

We report a case study applying the Minnesota Pollution Control Agency’s ATP framework to a research study with the goal of a rapid-screening assessment for the prioritization of chemical hazards detected in freshwater ecosystems relied on for subsistence. We identified high priority chemicals across all media, sites, and varying primary use categories, ranging from pharmaceuticals to insect repellent. Remote, undeveloped lakes often contained a larger proportion of high priority contaminants than developed and wastewater effluent-impacted sites. Due to the presence of high priority contaminants throughout the GPIR and 1854 Ceded Territory, we recommend future monitoring, rigorous evaluation of biological effects, and if warranted, the development of a risk assessment to better understand the risk posed by the high priority compounds we have identified.

The hazard profiles presented here address the potential for exposure and biological effects through the incorporation of available data including detection frequency, persistence, endocrine disruption, toxicity, and bioaccumulation of detected chemicals. While some of the questions that make up the chemical profiles may be correlated (e.g., detection frequency and persistence), they encompass persistence, bioaccumulation, and toxicity, which are common factors in many hazard assessments (Arnot & Mackay, 2008). Given their distance from known CEC point sources, we would predict that remote regions would contain a larger percentage of high priority CECs than low priority CECs as a result of the former chemicals generally being more persistent; however, all CECs detected across all sites (except nine unknowns because of data limitations) were persistent. The identification of mostly persistent chemicals is important, as a greater emphasis might be given to highly persistent CECs in chemical assessments and decision making (Cousins et al., 2019). Since most of our high priority CECs were persistent and often detected in more than 20% of the study sites, bioaccumulation, potential endocrine disruption, and potential toxicity in water were the deciding factors in whether a chemical would be classified as high priority or not.

All of the high priority contaminants identified in fish have been shown to affect aquatic biota at the genetic, physiological, or behavioral level: fluoxetine, diphenhydramine, azithromycin, roxithromycin, miconazole, verapamil, and N,N-Diethyl-m-toluamide (DEET). For example, several studies have demonstrated that the selective serotonin reuptake inhibitor and antidepressant fluoxetine can alter reproductive and antipredator behaviors in freshwater fish (Dzieweczynski et al., 2016; Fursdon et al., 2019; Martin et al., 2017, 2019; Pelli & Connaughton, 2015) at environmentally relevant concentrations (De Abreu et al., 2014). Additionally, diphenhydramine, an antihistamine, can be toxic to aquatic organisms (Berninger et al., 2011), is often detected in the environment (Burket et al., 2020; Du et al., 2016; Ramirez et al., 2009; Scott et al., 2019; Wang & Gardinali, 2012), and has been detected in marketed fish filets (Foltz et al., 2014). Antimicrobials, including antibiotics and an antifungal medication, were also among those chemicals found to be high priority in fish in the current study. Antibiotics can be toxic to aquatic biota (Liu et al., 2014; Yan et al., 2019) and have the potential for transfer and biomagnification within aquatic food chains (Ding et al., 2015). Importantly, antibiotics in the environment may also lead to the increased abundance and diversity of antibiotic resistance genes or antibiotic resistant bacteria in the environment, affecting aquatic ecosystem (Bueno et al., 2019; Pazda et al., 2019; Reichert et al., 2019; Szekeres et al., 2018). Similarly, antifungal drugs, such as miconazole, which was found to be toxic to common water fleas (Minguez et al., 2016) and inhibits fungal cytochrome P450 enzymes (Beijer et al., 2018), can lead to the development of antifungal drug resistance (Assress et al., 2020).

Although we made some modifications to the ATP framework to accommodate our dataset, there are notable limitations. While toxicity reference values based on water column concentration data are somewhat available, apical effect data are more limited regarding sediment and fish tissue values, restricting our evaluation to a subset of detected CECs. Having only POCIS detections, and not water concentrations, also limits the extent of the current prioritization. Further, due to the weight-of-evidence approach used to create chemical profiles, a low priority level might reflect a lack of information rather than truly indicating low impact. For example, the chemical 10-hydroxy-amitriptyline, a metabolite of the antidepressant amitriptyline, has the potential to bioaccumulate, was neutral at water pH at least 10% of the time, and was present in more than 20% of samples. Therefore, it received a priority level of 3 for its water chemical profile. However, data were not available regarding its toxicity, biodegradation, or endocrine disruption potential, so this chemical was flagged as a contaminant warranting further evaluation. The question “Is the chemical neutral at the pH of the water > 10% of the time?” was added to the chemical profiles developed in this study because water chemistry data, which plays an important role in the toxicity of chemicals (Alsop & Wilson, 2019; Escher et al., 2020), was available for most of the study locations. While this evaluation employed a cutoff of 10% to conservatively prioritize chemicals and sites, this approach could mask effects at sites that are uniquely threatened. Therefore, if ranking individual sites instead of chemicals is the primary objective, the approach may need some modification. Lastly, the growing abundance of high-throughput bioactivity data (e.g., EPA’s Toxicity Forecaster (ToxCast)) may be used to further sort the currently detected chemicals by their potential to exert specific biological activity as in Corsi et al. (2019); although, such approaches also contain uncertainties, such as a somewhat unknown relevance to apical outcomes in the variety of species present in freshwater ecosystems.

This study utilized a screening-level tool that is conservative and likely to avoid type II error (i.e., false negatives); however, it is possible that some low or intermediate priority chemicals were misidentified and are of potential concern. We chose to use ECOSAR toxicity values but acknowledge that other available in silico toxicity models (Melnikov et al., 2016) or measured data could have led to different conclusions. For example, we explored Toxicity Estimation Software Tool (TEST) (Martin, 2016) values post hoc and noted some differences in toxicity values that would impact the answer to the toxicity question. Based on this post-hoc assessment, six chemicals would move to high priority using TEST toxicity values so they were flagged for further follow-up, such as a literature search or ToxCast evaluations. We used this method to increase the range of applicable domains and decrease the chance for misclassification of a chemical into a lower priority bin, with the ultimate goal of increasing the odds that our method is conservative while maintaining its rapid pace. While the purpose of this paper was not to compare methods, it is important to note that priority levels are dependent, in part, on the adopted prioritization method.

This study resulted in a prioritized list of chemicals that guided Phase II of this project, an investigation of anthropogenic factors associated with the detection of high priority CECs in northeastern Minnesota (Servadio et al., under review, this issue). Determining key sources of spillover into and transport through the aquatic environment is critical to the mitigation of these high priority chemical contaminants. Further, the identification of high priority chemicals on tribal lands provides information to natural resource managers and stakeholders developing best management practices for water pollution and wastewater treatment processes. However, we note that an important next step should include a risk-based assessment of the prioritized contaminants.

5. Conclusions

This study adds to the understanding of the potential hazards of 117 CECs detected in northeastern Minnesota and prioritizes chemicals for further study or mitigation, particularly in a region that is important for sustaining indigenous culture through subsistence fishing. We performed a rapid assessment of the detected chemicals in order to prioritize further research and management efforts in the region. Where universal standards, benchmarks, or individual toxicity assessments for CECs are lacking, chemical profiles provide a broad understanding of the potential hazards these chemicals pose to aquatic ecosystems and highlight the need for more research in these areas.

Highlights.

• Contaminants of emerging concern persist in northeastern Minnesota

• 117 contaminants were prioritized based on their relative hazard

• 50 contaminants in water, 21 in sediment, and 7 in fish were high priority

• We recommend future monitoring of high priority contaminants