Transcriptome signatures of wastewater effluent exposure in larval zebrafish vary with seasonal mixture composition in an effluent-dominated stream

Abstract

Wastewater treatment plant effluent-dominated streams provide critical habitat for aquatic and terrestrial organisms but also continually expose them to complex mixtures of pharmaceuticals that can potentially impair growth, behavior, and reproduction. Currently, few biomarkers are available that relate to pharmaceutical-specific mechanisms of action. In this experiment, zebrafish (Danio rerio) embryos at two developmental stages were exposed to water samples from three sampling sites (0.1 km upstream of the outfall, at the effluent outfall, and 0.1 km below the outfall) during base-flow conditions from two months (January and May) of a temperate-region effluent-dominated stream containing a complex mixture of pharmaceuticals and other contaminants of emerging concern. RNA-sequencing identified potential biological impacts and biomarkers of wastewater effluent exposure that extend past traditional markers of endocrine disruption. Transcriptomics revealed changes to a wide range of biological functions and pathways including cardiac, neurological, visual, metabolic, and signalling pathways. These transcriptomic changes varied by developmental stage and displayed sensitivity to variable chemical composition and concentration of effluent, thus indicating a need for stage-specific biomarkers. Some transcripts are known to be associated with genes related to pharmaceuticals that were present in the collected samples. Although traditional biomarkers of endocrine disruption were not enriched in either month, a high estrogenicity signal detected upstream in May was associated with disruptions of normal development at one developmental stage and implicates the presence of unidentified chemical inputs not captured by the targeted analysis. This work reveals associations between bioeffects of exposure, stage of development, and the composition of chemical mixtures in effluent-dominated surface water. The work underscores the importance of measuring effects beyond the endocrine system when assessing the impact of bioactive chemicals in wastewater effluent and identifies a need for non-targeted chemical analysis when bioeffects are not explained by the targeted analysis.

1. INTRODUCTION

Wastewater effluent containing complex mixtures of pharmaceuticals and other contaminants of emerging concern (CECs) poses a potential threat to environmental health, and small effluent-dominated streams with minimal dilution are particularly at risk.¹ Conventional wastewater treatment plants (WWTPs) were not designed to remove CECs, resulting in their frequent detection and wide distribution in surface water including drinking water sources.^2–4 Although effluent discharged into small streams typically undergoes substantial dilution once released into higher volume waterways,^5,6 small receiving streams provide important habitat in aquatic and riparian ecosystems.^7,8 The continual presence of bioactive chemical inputs can produce exposure conditions in which individual chemical concentrations exceed thresholds documented to harm aquatic organisms,^9,10 particularly during early life stages.^11,12 However, less is known regarding the potential impacts of chemical mixtures known to be prevalent in streams.¹³

Although the chemical contaminants typically found in WWTP effluent represent many use classes, pharmaceuticals warrant heightened scrutiny due to their designed bioactivity and the diversity of their protein targets, most of which are relevant to fish.¹⁴ In lab-based exposures with fish, deleterious impacts of exposure to pharmaceuticals have been documented at sublethal concentrations (ng/L-μg/L)¹⁵ that would otherwise be considered safe by conventional toxicological endpoints.^16–18 Impacts include reproductive impairments, altered stress responses, behavioral changes, and decreased disease resilience.^18–22 For example, the insulin-sensitizing diabetes drug, metformin, impairs growth in fish (1000–3000 ng/L)^23,24 and can induce intersex condition in fathead minnow testes (40000 ng/L).²⁵ Furthermore, in many fish species, environmentally relevant concentrations of antidepressants of less than 1000 ng/L are reported to disrupt stress responses (venlafaxine),²⁶ reproductive and predator avoidance behaviors (fluoxetine),²⁷ brain monoamine levels (venlafaxine),²⁸ and diurnal activity patterns (fluoxetine/sertraline/venlafaxine mixtures).²⁹ Wild fish in effluent dominated streams may experience both chronic and pulse exposures, both of which have been shown in zebrafish embryos to cause physiological and behavioral problems later in development^30–32 and across generations.^33,34

Nevertheless, applying these findings to biomonitoring effluent-dominated surface water in the real world presents significant challenges. Although many sublethal exposure impacts are likely shared, pollution sensitivity can vary substantially by species.³⁵ Furthermore, complex chemical mixtures often induce additive, synergistic, and other kinds of interactive effects that a simple aggregation of effects observed in lab-based exposures to singular chemicals would not capture.^36,37 In effluent-dominated streams, chemical mixture composition evolves over a downstream gradient and varies spatiotemporally as human use patterns change.³⁸ Wild fish are likely subjected to a variety of chronic and pulse exposure conditions related to their travel patterns and may experience greater vulnerability during reproductive periods and early life stages. Identifying relevant sublethal exposure impacts to monitor in wild fish populations is also complicated by the fact that although exposure to effluent may not result in readily observable changes or be associated with conventional biomarkers of fish health, it may nevertheless contribute to the vulnerability of fish populations facing additional stressors.^39–41 Thus, biomarkers that can be used to monitor surface water and wild fish populations for evidence of these impacts are extremely important.

In the past two decades, a new paradigm has emerged for investigating the adverse impacts of chemical exposures as the result of progressive changes along biological pathways at the molecular, cellular, tissue, and organ level. The shift to pathway-based systems toxicology was made possible by the development of the adverse outcome pathway (AOP) framework⁴² and the widespread use within ecotoxicology of transcriptomics and other biomolecule “omics” technologies that identify systematic changes across known biological pathways and can be used to uncover mechanistic relationships when integrated with phenotypic and other biomolecular data.^42,43 Sublethal pathway-based impacts of exposures have been particularly well characterized for endocrine active chemicals, most notably for 17α-ethinylestradiol (EE2)^41–43, the synthetic estrogen used as birth control.^44–46 “Omics” tools have proved invaluable to the task of assessing sublethal exposure impacts and provide opportunity to expand knowledge of biomarkers associated with specific chemicals and AOPs.

Although transcriptomics has been successfully used to distinguish “clean” and contaminated field sites,⁴⁷ reliable biomarkers of specific pharmaceuticals and pharmaceutical classes have not been established^48,49 in part because pathway-based impacts of exposure in fish are not well characterized beyond those that involve steroidal reproductive hormones.^18,35,50 Of 973 currently marketed pharmaceuticals that act on small protein targets, 88% lack comprehensive ecotoxicity testing data.¹⁴ Without pathway-based genetic biomarkers, predicting the net impact of exposure to chemical mixtures containing pharmaceuticals becomes problematic.

In this study, we evaluated the response of zebrafish embryos exposed to water samples from strategic locations in an effluent-dominated stream and its evolving chemical composition using RNAseq with the goal to identify biomarkers of effluent exposure beyond traditional stress or estrogenic biomarkers and identify the range of biological impacts that could be monitored in surface water containing wastewater-derived CECs. Zebrafish embryos at 3 and 6 days post-fertilization (dpf) were exposed to water from select points of an effluent-dominated stream reach. These two developmental stages capture distinct vulnerabilities to environmental insult before and after the onset of free-feeding, which is marked by activation of the cytochrome p450 enzymes responsible for metabolizing xenobiotics.⁵¹ Water samples used for exposures in this study were taken in two seasons (winter and spring) and selected from a suite on monthly baseflow samples over a 12-month time period due to their contrasting pharmaceutical profiles. We hypothesized that effluent exposure would reveal biological effects beyond current endocrine system related biomarkers, that gene expression would vary with chemical exposure over two contrasting seasonal conditions, and that disrupted biological pathways would differ by stage of development.

2. MATERIALS AND METHODS

2.1. Study Sites and Water Sampling

Muddy Creek is a small effluent-dominated stream near North Liberty, Iowa (latitude 41°42′00″, longitude 91°33′46″) that receives approximately 5,300 m³ of effluent per day from the North Liberty WWTP⁵² and discharges to the Iowa River. The stream’s 22.5 km² drainage area encompasses a mix of suburban (60%) and agricultural land uses (24.5%).³⁸ Three previously established USGS sampling sites were used: (1) 0.1 km above the WWTP outfall (US1; 0545405), (2) the effluent outfall (effluent; 05454051), and (3) 0.1 km below the outfall (DS1; 05454052) (Figures S.1 and S.2). Monthly grab samples using the single vertical at centroid-of-flow method⁵³ were taken in triplicate from each site over 12 baseflow sampling events between September 2017 and August 2018. Replicate samples were used for measurement of 113 pharmaceuticals using a previously published USGS method,⁵⁴ a bioluminescent yeast estrogenicity assay,⁵⁵ and zebrafish embryo exposures.⁵⁶ Water samples were shipped on ice to the University of Wisconsin-Milwaukee where they were stored at −80 °C prior to use in zebrafish embryo assays and to the USGS Organic Geochemistry Research Lab (Lawrence, KS) for solid phase extraction⁵⁷ prior to estrogenicity analysis at the Eastern Ecological Science Center (Kearneysville, WV). Further characterization of the field sites, detailed methods for water sampling, and analysis of the monthly chemical data are available in Zhi et al. (2020).³⁸

2.2. Animals

Zebrafish embryos were obtained from group spawning events of a wild type 5D zebrafish lab culture (University of Wisconsin-Milwaukee). Adult zebrafish of mixed sexes were housed in a flow-through aquatic system (Aquaneering, San Diego, CA) with recirculating dechlorinated municipal water and fed TetraMin flake twice daily. The system was maintained at 27°C in a 16:8-hr light/dark cycle. All procedures were conducted in accordance with animal use and care protocols approved by the Institutional Animal Care and Use Committee of the University of Wisconsin-Milwaukee.

2.3. Embryo exposures

Fertilized embryos were screened for uniformity in developmental stage progression. Simultaneous 3- and 6-day exposures were performed separately using five replicate petri dishes containing groups of 20 embryos (6 hours post-fertilization) immersed in 30 mL of sample water from each site. The pH of samples ranged from 7.3 to 7.6. Parafilm-sealed petri dishes were incubated at 28°C under a 16:8-hour light/dark schedule. Upon completion, surviving larval fish from each replicate dish were pooled into 1.5 mL tubes and snap frozen in liquid nitrogen for RNA extraction. Samples exposed to the January and May 2018 waters from the US1, effluent, and DS1 sites were selected for RNA sequencing in this study. The selection of January and May sampling events was based on prior published work³⁸ at this site in which these months captured seasonal pharmaceutical use patterns with higher concentrations of antibiotics in January and higher antihistamines in May (Figure S.3). The site above the outfall (US1) was used as a point of comparison to assess the relative impact of wastewater effluent exposure at the effluent and DS1 site. Using an upstream field site as a reference rather than a municipal tap water control effectively limits variability in gene expression associated with the ambient conditions common to all sites. Significant changes at the effluent and DS1 sites compared to the US1 are thus more likely to reflect the characteristics of the effluent rather than difference between the basal water chemistry of the stream and a more “pristine” municipal source.

2.4. RNA Sequencing

2.4.1. RNA Extraction, Library Prep, and Sequencing

Total RNA was isolated using standard protocol for Direct-zol RNA MiniPrep (Zymo Research, R2051). Whole embryos were homogenized in TRIzol with a pestle in a microfuge tube, and RNA was purified on Zymo-Spin IIC columns. Sample purity was assessed with a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA) with acceptable wavelength ratios of 1.8–2.0 for 260/280nm and 2.0–2.2 for 260/230 nm. RNA integrity (RIN) was measured on an Agilent Bioanalyzer 2100 (Agilent, Santa Clara, CA), and samples with a RIN>7 were used. RNA was quantified on a Qubit 2.0 fluorometer (Invitrogen, Thermo Fisher Scientific, Waltham, MA). Two samples (January US1 3 dpf and May US1 6 dpf) had RIN scores below 7 and were not used for RNA sequencing. Libraries were prepared using Illumina TruSeq Stranded mRNA sample preparation kit (Illumina, RS-122–2102) and IDT for Illumina – TruSeq RNA UD Indexes (Illumina, 20022371) following standard protocol, using 200ng of total RNA. Libraries were sequenced on an Illumina NovaSeq6000 (paired-end 150 bp reads).

2.4.2. Processing of RNAseq Data

The total genomic yield surpassed 2.104 billion paired-end reads with a median per-sample yield of 51 million fragments and a population standard deviation of 14 million fragments. Sequence data was quality-assessed using FastQC v0.11.5,⁵⁸ and sequencing adapters were clipped using Cutadapt v1.18. The resulting quality-controlled data was pseudoaligned and sample-quantified against the GRCz11 Ensembl release of the zebrafish reference transcriptome using Kallisto v0.45.0.^59,60

DaMiRseq was used to filter and normalize raw count data.⁶¹ Transcripts were removed if they had fewer than 10 counts across 70% of samples or were hypervariant (coefficient of variance threshold of 3). Raw counts were normalized to library size using variance stabilizing transformation (vst), which reduces the dependence of the mean on variance. DESeq2 was used to perform analysis of differential expression between the upstream baseline (US1) and the effluent and DS1 samples.⁶² Resulting tables of differentially expressed transcripts (DETs) were re-annotated with Ensembl reference information and relationally joined with Kallisto sample quantification counts using custom tooling. Transcripts differentially expressed at the effluent and DS1 sites (vs. US1) were considered significant at a Benjamini-Hochberg (B-H) adjusted p-value<0.01 and |log₂ fold change|>1. To characterize developmental differences at the upstream reference site, DESeq2 was also used to perform analysis of differential expression between the 3 and 6 dpf exposures to US1 samples from the same month. Transcripts were considered to have significantly higher expression at 3 dpf (vs. 6 dpf) and 6 dpf (vs. 3 dpf) at a B-H adjusted p-value<0.01 and |log₂ fold change|>1. RNA-seq data are available in the National Center for Biotechnology Information’s Gene Expression Omnibus under accession number GSE179335.

2.5. PLS-DA and Functional Enrichment

MixOmics⁶³ was used for partial least squares-discriminant analysis (PLS-DA) to determine the relative influence of month, exposure length, and site on the overall transcriptome. PLS-DA was performed over all exposures to compare the influence of month and exposure length and performed separately over the 3 and 6 dpf exposures to compare the influence of month and site.

Overrepresentation analysis was conducted on DETs from each comparison using the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database⁶⁴ and Gene Ontology (GO) annotations from the Gene Ontology Consortium.^65,66 GO terms and KEGG pathways overrepresented among significant DETs (vs. US1) were identified using clusterProfiler,⁶⁷ which employed hypergeometric enrichment tests with a B-H adjusted p-value<0.05 to control for multiple testing and an FDR<0.1. GO terms and KEGG pathways overrepresented among significant US1 3 dpf DETs (vs. 6 dpf) and 6 dpf DETs (vs. 3 dpf) were identified through hypergeometric enrichment tests in g:Profiler using a custom g:SCS significance threshold to control for multiple comparisons and a corrected p-value cutoff of 0.05.⁶⁸ Gene targets of pharmaceuticals detected in Muddy Creek water samples were pulled from the manually curated Comparative Toxicogenomic Database.⁶⁹ Gene targets from all vertebrate organisms were considered. Genes were converted to Danio rerio emsembl transcript IDs using g:Profiler.

2.6. Bioluminescent Yeast Estrogenicity

Total estrogenicity of sample extracts⁵⁷ was determined using the bioluminescent yeast estrogen screen (BLYES) as previously described,^55,70 but with minor modifications detailed in the Supporting Information. The detection limit for this assay was 0.18 ng/L E₂Eq_(BLYES).

3. RESULTS AND DISCUSSION

3.1. Transcriptome profiles differ with shifting effluent chemistry across exposure months and relate to known pharmaceutical targets.

Signatures of WWTP effluent were present in both months at each developmental stage, evident in the full transcriptome and in the proportion of DETs (vs. US1) shared between the effluent and DS1 sites, whose similar chemical profiles are the result of the minimal dilution of the effluent at DS1 (≥80% effluent during both months).³⁸ Developmental stage was a strong determinant of the transcriptome profile across all samples (Figure 1a). Similarity between the effluent and DS1 sites was evident in PLS-DA performed separately on the 3 and 6 dpf exposures where the effluent and DS1 sites of each month were more similar to each other than to the US1 site (Figure 1b and c). The influence of effluent exposure was also evident in the proportion of DETs shared between the effluent and DS1 sites within a month and developmental stage. In the January 3 and 6 dpf exposures, 40–42% of DETs were shared between the effluent and DS1 sites. In the May exposures, 76–82% of DETs were shared between sites (Figure 1e). All DETs shared between sites within the same month and developmental stage had consistent fold change directions (positive or negative). The overlap between the effluent and DS1 transcriptomes likely reflects the high proportion of effluent to DS1 streamflow in both months: 89% in January and 80% in May (Table S.1).⁷¹

Figure 1.

Partial least squares discriminant analysis (PLS-DA) of January and May samples exposed to water from the effluent, DS1, and US1 sites at 3 dpf and 6 dpf. Ovals represent 95% confidence intervals. Sites: + = US1, Δ = effluent, and ◯ = DS1. (A) PLS-DA using samples grouped by month of exposure and developmental stage: Jan 3 dpf (n=14), Jan 6 dpf (n=15), May 3 dpf, (n=15), May 6 dpf (n=14).

(B) PLS-DA of 3 dpf samples classified by month and site: Jan 3 dpf US1 (n=4), Jan 3 dpf effluent (n=5), Jan 3 dpf DS1 (n=5), May 3 dpf US1 (n=5), May 3 dpf effluent (n=5), May 3 dpf DS1 (n=5).

(C) PLS-DA of 6 dpf samples classified by month and site: Jan 6 dpf US1 (n=5), Jan 6 dpf effluent (n=5), Jan 6 dpf DS1 (n=5), May 6 dpf US1 (n=4), May 6 dpf effluent (n=5), May 6 dpf DS1 (n=5).

(D) The total number of upregulated (red) and downregulated (green) transcripts with significant differential expression (DETs) from the US1 baseline are represented for each month, developmental stage, and site. DETs were defined as protein-coding transcripts with |log₂ fold change| >1 and adjusted p-value <0.01.

(E) Number and percent of DETs shared between and unique to the effluent and DS1 sites in each month and developmental stage.

(F) Number and percent of DETs shared between and unique to January and May at both developmental stages and sites.

The comparatively few DETs (vs. US1) shared between months at the same site and developmental stage highlights the likely influence of seasonal differences in stream chemistry. Only 8 to 12% of DETs were shared across month (Figure 1f). Compared to January, total concentrations of H₂ antagonists (917 ng/L) and antivirals (841 ng/L) were three times higher in the May effluent, and antidepressants (4152 ng/L) and antihistamines (3838 ng/L) were 30 and 36% higher, respectively. In January, effluent concentrations of antibiotics (992 ng/L) and methyl-1H-benzotriazole (713 ng/L) were three times higher than in May and the total concentration of beta-blockers was 28% higher (980 ng/L) (Figure S.3, Table S.2).

The frequency of pharmaceutical gene targets from the Comparative Toxicogenomic Database among differentially expressed transcripts from the two months suggests a greater influence in May of nearly every pharmaceutical. Consistent with the seasonal stream chemistry, more CTD gene targets associated with antidepressants, antihistamines, and H₂ antagonists were differentially expressed in May (Table 1). However, other pharmaceuticals with measured concentrations that did not notably differ between months, like carbamazepine and metformin, also appeared to have an outsized impact on the May exposures. CTD targets of beta-blockers and other pharmaceuticals detected at higher concentrations in January also had more differential expression in May, although the difference between months was less dramatic than with other pharmaceuticals.

Table 1.

Number of genes associated with each chemical in the Comparative Toxicogenomic Database (CTD) detected in either the 3 or 6 dpf D. rerio exposures, number of CTD genes with significant differential expression (DE) (log2 fold change >1, adjusted p-value<0.01), and ratio of differentially expressed genes to CTD genes (DE/CTD) as percent. VEN, venlafaxine; FLX, fluoxetine; CTL, citalopram; BUP, bupropion; CBZ, carbamazepine; GAB, gabapentin; MET, metformin; MTP metoprolol; ATE, atenolol; DPH, diphenhydramine; FAM, famotidine; FEX, fexofenadine; RAN, ranitidine; SFX, sulfamethoxazole; ACV, acyclovir; OXY, oxycodone.

			January	3 dpf	May	3 dpf	January	6 dpf	May	6 dpf
	Chm	Total CTD	DE CTD	DE:Total	DE CTD	DE:Total	DE CTD	DE:Total	DE CTD	DE:Total
Antidepressants	VEN	192	26	14%	85	44%	25	13%	81	42%
	FLX	355	36	10%	141	40%	40	11%	133	37%
	CTL	28	3	11%	9	32%	3	11%	8	29%
	BUP	39	1	3%	14	36%	2	5%	13	33%
Other neuro- pharmaceuticals	CBZ	2174	229	11%	766	35%	217	10%	616	28%
	GAB	309	18	6%	121	39%	25	8%	107	35%
Diabetes	MET	835	74	9%	274	33%	80	10%	252	30%
Cardiovascular	MTP	47	7	15%	18	38%	10	21%	15	32%
	ATE	54	5	9%	18	33%	8	15%	18	33%
Histaminergic	DPH	23	2	9%	5	22%	3	13%	7	30%
	FAM	48	3	6%	17	35%	6	13%	16	33%
	FEX	18	4	22%	9	50%	3	17%	7	39%
	RAN	129	8	6%	43	33%	9	7%	40	31%
Antibiotic	SFX	9	0	0%	3	33%	0	0%	1	11%
Antiviral	ACV	10	1	10%	1	10%	2	20%	1	10%
Analgesic	OXY	197	22	11%	72	37%	15	8%	60	30%

3.2. Common transcripts suggest metabolic impacts of sewage effluent.

Biomarkers of cellular energy production were upregulated at all sites downstream from the WWTP in the January and May 3 dpf exposures. Growth inhibition and disruption of metabolic pathways are commonly observed responses to pharmaceutical exposures.⁷² Of the 39 transcripts upregulated at all sites, 9 are involved in metabolic processes that generate cellular energy including suclg2, gnpda2, slc2a2, g6pca.2, tmem86b, aldh1l1, abhd14b, cers3a, and abcb11b (Figure S.4). Succinate-CoA ligase (suclg2) and glucose-6-phosphatase a (g6pca.2) are involved in the citric acid cycle and have been identified in exposures to neuro-pharmaceuticals and antibiotics. Previous research has documented exposure to antidepressants downregulates g6pca.2 and energy metabolism KEGG pathways⁷³ in 3 dpf larval zebrafish. Exposure to 1000 ng/L carbamazepine downregulated s uclg2 in adult Chinese rare minnow (Gobiocypris rarus)²² and in 3 dpf zebrafish larvae (antibiotic mixtures of 100–100000 ng/L), where it was identified as an influential hub gene.⁷⁴

3.3. Significant differences in expression signatures across months reflect differences in chemical mixtures

Impacts related to glycolysis, the musculoskeletal system, and cardiac functioning were indicated in both months. However, many other biological processes and pathways were uniquely or primarily enriched in exposures to May water samples and likely reflect influence of specific chemical mixtures at higher concentrations relative to January. Unique KEGG pathways and biological process Gene Ontology (GO:BP) terms included adrenergic signaling in cardiomyocytes, visual system functions, neurological development, the MAPK, FoxO and PPAR signaling pathways, and histidine metabolism (Figure 2).

Figure 2.

Statistically significant (adjusted p-value <0.05) overrepresentation of biological pathways and processes of differentially expressed transcripts (DETs, |log₂ fold change| >1 and adjusted p-value <0.01) up- and downregulated at effluent and DS1 (vs. US1, the upstream reference) in zebrafish embryo exposures (Jan 3 dpf, Jan 3 dpf, May 6 dpf). Overrepresentation analysis of Kyoto Encyclopedia of Genes and Genomes⁶⁴ (KEGG) pathways and Gene Ontology (GO:BP) was performed in clusterProfiler (R v4.0.1).

(A) Overrepresented KEGG pathways from each set of DETs (vs. US1).

(B) Select overrepresented GO:BP terms from each set of DETs (vs. US1).

Although KEGG pathways and GO:BP terms involving the musculoskeletal system, heart, cell adhesion, metabolism, and embryo and larval development were overrepresented in both months and developmental stages, the breadth and quantity of those functional enrichments was greater in May (Figure 2). Contrast between the US1 reference and the effluent and DS1 sites was most pronounced in May with at least three times as many DETs (vs. US1) at each site and developmental stage compared to January (Figure 1d). It is noteworthy that more than 95% of enriched GO:BP terms in 3 and 6 dpf May exposures resulted from upregulated transcripts and 100% of GO:BP enrichments resulted from downregulated transcripts in January (Table S.3).

The dysregulation of cardiac processes across month and developmental stage is consistent with the presence beta-blockers. In the May 3 dpf exposure, 54 transcripts were impacted in the adrenergic signaling in cardiomyocytes pathway, including ion membrane transporters, the actin-myosin crossbridge, and a β2-adrenergic receptor (adrb2b) involved in regulation of heart rate (Figure S.5b). Adrenergic signaling receptors are involved in a wide variety of biological processes including heart contractions, lipolysis (breakdown of triglycerides), blood flow, and regulation of metabolism,⁷⁵ which involve interaction with the FoxO and MAPK pathways, both disrupted in the May 6 dpf exposures. Although few studies have examined sublethal impacts of exposure to beta-blockers on fish, propranolol (80 ng/L) is reported to reduce larval zebrafish heart rate⁷⁶ and atenolol (2 ng/L) can block epinephrine-stimulated glucose production in trout hepatocytes.⁷⁷ Total concentrations of the three beta blockers measured at the Muddy Creek effluent site (metoprolol, atenolol, and propanolol) exceed these levels by orders of magnitude (i.e., 980 ng/L in January and 701 ng/L detected in May) (Figure S.3). Although enrichment of the adrenergic signaling pathway was unique to the May exposures, the enrichment of the cardiac muscle contraction pathway in the January exposures involved many of the same transcripts (Figure S.6).

The antidepressant venlafaxine has been shown to impact many interconnected pathways and biological processes and was one of the top pharmaceuticals detected in Muddy Creek with effluent concentrations at 1240 ng/L (January) and 1550 ng/L (May) (Figure S.3). Venlafaxine is known to disrupt neurological development,⁷⁸ MAPK signaling,⁷⁹ metabolism,²⁶ stress response, locomotor activity,⁸⁰ and adrenergic signaling.^77,78 Although the exact mechanism of action on the MAPK pathway is unclear, the direct impact of venlafaxine may result from upregulation of brain derived neuroeffector (BDNF), which initiates a phosphorylation cascade that reaches the MAPK pathway.⁷⁹ Impacts to larval fish have been documented at low concentrations; an 80-hour exposure of larval zebrafish to just 80 ng/L of venlafaxine was sufficient to increase embryonic malformations, including loss of pigmentation.⁸¹ Pigmentation abnormalities in larval zebrafish have been associated with delayed development and reduced fitness in other studies, which has been implicated as a possible link to beta-adrenergic signaling given the presence of adrenergic receptors on pigment-producing melanophores.⁸² The lack of similar biological pathway and process enrichments in January, however, suggests that venlafaxine is not driving enrichment patterns of DETs between the months examined during this study.

Interestingly, the visual system was also the target of transcriptomic changes. Impacts occurred primarily in the May 3 dpf exposures through enrichment of the phototransduction KEGG pathway and several other visual GO:BP terms among downregulated DETs (Figure 2). Impaired eye development in fish necessarily impacts neurological functioning, behavior and metabolic processes, and thus may later impair growth, reproduction, and survival.¹⁶ Many chemicals found in wastewater effluent, including progestins,^83,84 antidepressants,^85,86 flame-retardants,⁸⁷ and pesticides,^88,89 have been associated with visual system disruptions including altered gene expression in the phototransduction cascade. In the 3 dpf May exposure, transcripts were downregulated across the phototransduction cascade and included light-sensitive pigments, transducins involved in G-protein signaling, Ca²⁺ and Na⁺ voltage-gated channel proteins, and rhodopsin kinases that deactivate phototransduction in dark conditions (Figure S.5a). In total, 71 unique genes were enriched among vision-related biological processes at the effluent site and 61 at DS1.

In the 6 dpf May exposure, the overrepresentation of histidine metabolism may relate to the influence of antihistamines, which are inverse agonists of the H₁ histamine receptor, and H₂ antagonists, both of which were detected at higher concentrations in May compared to January. In fish, the synthesis of histamine through histidine metabolism primarily occurs within histaminergic neurons in the hypothalamus. The histaminergic system in zebrafish has been shown to play a role in modulating swimming behavior,⁹⁰ arousal⁹¹ and behaviors important in establishing social hierarchies (e.g., aggression and boldness).^92,93 Both H₁ and H₂ histamine receptors (HRH1 and HRH2) are expressed early in zebrafish development, with HRH1 peaking at 5–7 dpf. Most antihistamines do not directly prevent the production of histamine; rather, as inverse agonists of the H₁ histamine receptor,⁹⁴ they inactivate HRH1 in smooth muscle cells preventing the vascular dilation that causes swelling and redness associated with allergic reactions.⁹⁵ Recently, however, diphenhydramine has been shown to reduce allergic response not just by inactivating HRH1 but also by inhibiting production of histamine in the first place by downregulating the rate limiting enzyme that synthesizes histamine, histidine decarboxylase (HDC).⁹⁶ Notably, there was a −4 log₂ fold change (p<0.01) in hdc expression at both the effluent and DS1 sites in May (at both 3 and 6 dpf) but no significant differential expression of the gene in January (Figure 3). Diphenhydramine was the only antihistamine detected with the ability to downregulate hdc expression.⁹⁶ Diphenhydramine concentrations at the effluent and DS1 sites were 113 and 81 ng/L in January and 150 and 109 ng/L in May (Figure S.3).

Figure 3.

Gene expression from RNA-sequencing of 6 dpf larval zebrafish exposed to water from the May effluent site. Log2 fold changes relative to the US1 site are shown on the KEGG pathway for histidine metabolism (dre00340).⁶⁴ Ovals indicate metabolites and rectangles indicate genes. Pathway rendered by Pathview.⁹⁷

The LC50 of diphenhydramine is high in fish (692 mg/L for 3 dpf zebrafish and 262 for 6 dpf),⁹⁵ and as a result antihistamines are often considered to pose low risk to aquatic vertebrates. However, sublethal impacts on behavior have been documented at much lower concentrations of diphenhydramine, possibly owing to its inhibition of serotonin reuptake.⁹⁸ For example, an LOEC of 5.6 μg/L was established for reduced feeding rate in fathead minnows (Promelas pimephales).⁹⁹ More recently, the histaminergic system has drawn attention for the role it plays in modulating aggression and other behaviors important in establishing social hierarchies in zebrafish.^92,93 Histidine decarboxylase (hdc) was shown to be upregulated in dominant zebrafish, along with histamine receptors hrh1 and hrh2.¹⁰⁰ Elevated hdc expression was maintained in adult offspring of dominant (male and female) zebrafish pairings in a transgenerational study that identified inherited dominant and subordinate behaviors.⁹³ In addition to the production of histamine for allergic responses, hdc also plays a role during embryonic brain development where histamine regulates the number of hypocretin/orexin neurons that are hypothesized to eventually regulate the number of mast cells producing histamine in adults.¹⁰¹ The expression of hdc in larval zebrafish thus may contribute to the plasticity of the histaminergic system in the brain later in adulthood. Finally, the histamine/H₁ receptor axis is now also recently known to play an essential role during cardiac development in larval zebrafish in promoting cardiomyocyte differentiation through activation of the ERK 1/2-STAT3 pathway.¹⁰²

3.3. Estrogenicity still an important signal identified by transcripts but with unknown origins

Notable estrogenicity was observed at all three sites during monthly collections. This included an estrogenic signal with unknown origins detected by the BLYES assay in May at US1 (Figure 4a). The greatest estrogenic measurement of 3.7 ng/L E₂Eq_(BLYES) occurred at the US1 site in May, the highest at any of the Muddy Creek field sites over the 12-month study period. This measurement was more than double the estrogenicity of US1 in January at 1.52 E₂Eq_(BLYES) ng/L, which had the lowest measured estrogenicity among stream sample exposures used for RNA sequencing. Estrogenicity was most variable at US1, which had the highest and lowest E₂Eq values of all sites over the 12-month sampling period (1.03 E₂Eq ng/L in November 2017 and 3.7 ng/L in May 2018) (Figure 4a). In contrast, estrogenicity measurements below US1 ranged from 1.54 to 1.88 E₂Eq_(BLYES) ng/L at the effluent and from 1.55 to 2.37 E₂Eq_(BLYES) ng/L at DS1 (Figure 4a). The January and May US1 samples thus represent contrasting estrogenic stream gradients with US1 estrogenicity lowest of the three sites in January and highest in May.

Figure 4.

(A) Calculated estrogen equivalents (E₂Eq in nanograms per liter) relative to 17 β-estradiol of extracts from monthly 1-liter grab samples collected at baseflow, September 2017-August 2018.

(B) Percent of total differentially expressed transcripts (DETs) with higher upstream reference (US1) expression at 3 dpf (vs. 6 dpf) and 6 dpf (vs. 3 dpf) unique to and shared between January and May (log₂ fold change >1 between developmental stages and B-H adjusted p-value <0.01). Overrepresentation of biological process (GO:BP), molecular function (GO:MF), and cellular component (GO:CC) Gene Ontology terms and KEGG terms (adjusted p-value <0.05) among DETs with significantly higher expression at 3 or 6 dpf at US1 in January and May. Colors represent the proportion of DETs and enriched terms or unique to or shared between months.

Current predicted no effect concentrations (PNOECs) range from 0.1 to 0.73 ng/L E₂Eq_(BLYES)^103,104 thus identifying the baseline estrogenicity of Muddy Creek as high and of biological significance. PLS-DA and differential gene expression suggest that exposure to the estrogenicity signal at US1 affected 3 and 6 dpf larvae differently with observable differences between the months at 6 dpf. The transcriptomes of January and May US1 samples overlapped in PLS-DA of the 3 dpf samples but were completely distinct in PLS-DA of the 6 dpf samples (Figure 1b and c).

The impact of estrogenicity on the US1 exposures was also important when comparing January and May by functional enrichments from US1 transcripts with significantly higher expression at 3 dpf (vs 6 dpf) and 6 dpf (vs 3 dpf) (|fold change|>1, B-H p-value<0.01). The January and May US1 samples exhibited a high degree of functional similarity with respect to transcripts that had highest expression at 3 dpf (Figure 4b). These shared functional enrichments reflected biological processes associated with the 3 dpf stage, including rapid cell proliferation and differentiation (Table S. Nevertheless, enrichments of high expression 6 dpf transcripts differed markedly between months and only reflected biological processes specific to the 6 dpf stage in January. January enrichments encompassed a wide variety of processes known to be more relevant to the 6 dpf life stage (vs. 3 dpf), including functions related to digestion and metabolism, the immune system, response to the environment, circadian rhythm, and the visual system, all of which are required for hunting and free-feeding. Although January and May had a comparable number of transcripts with significantly higher expression at 6 dpf (3234 and 2663), May enrichments were narrow in scope and fewer in number. Pathways related to the visual system and muscle development were absent and there were notably few related to digestion and metabolism, which represented fewer than 50 genes total. Although synchronicity across individuals during development lessens as development progresses into the larval stage,¹⁰⁵ this analysis opens the possibility that the source of the estrogenic signal in May could have a wide range of impacts on biological processes at 6 dpf larvae.

There were several pharmaceutical detections at US1 in both months;³⁸ however, not at levels that could explain a strong estrogenicity signal or 6 dpf developmental impacts. In May, more chemicals were detected at US1 and concentrations tended to increase relative to January, but none of the detections in either month were at levels known to be bioactive (Figure S.3). An alternative explanation of the high estrogenicity signal at US1 could be the presence of chemicals not captured through the targeted analysis. Strategic non-targeted chemical analysis could help resolve cases in which distinct bioeffects occur but are not explained by the chemicals detected in a targeted analysis, which would be particularly useful for field reference sites.¹⁰⁶

4. CONCLUSIONS

Variation in chemical signatures across months is recapitulated in gene expression, and even with estrogenic input observed in the US1 reference site, the transcriptome still reveals key relationships between pathways and processes relevant to understanding environmental effects of chemical exposures. While shared transcripts related to cellular energy metabolism likely represent the off-target impacts that many pharmaceuticals have on metabolic functioning, other transcriptome signatures uniquely associated with the January and May effluent are reflective of chemicals that were in higher concentrations in each of those time periods. The disruption of normal developmental pathways in the 6 versus 3 dpf larval fish underscores the need for biomarkers specific to developmental stage and, along with supporting estrogenicity data, suggests developmental pathways may serve as biomarkers of endocrine disruption in larval zebrafish assays. However, the multiple other biological processes and pathways associated with exposure to effluent chemical mixtures (e.g., histidine metabolism, eye development, neurological function, and MAPK signaling) indicate that related biomarkers could provide more specific information about exposures to emerging contaminants in real world scenarios.

ABBREVIATIONS:

ACV

acyclovir

ATE

atenolol

BUP

bupropion

CBZ

carbamazepine

CTL

citalopram

DPH

diphenhydramine

FAM

famotidine

FEX

fexofenadine

FLX

fluoxetine

GAB

gabapentin

MET

metformin

MTP

metoprolol

OXY

oxycodone

RAN

ranitidine

SFX

sulfamethoxazole

AOP

Adverse Outcome Pathway

B-H

Benjamini-Hochberg

BLYES

bioluminescent yeast estrogen assay

BP

Biological Process

CC

Cellular Component

CEC

contaminant of emerging concern

CTD

Comparative Toxicogenomic Database

DET

differentially expressed transcript

dpf

days post fertilization

ECM

extracellular matrix

FDR

false discovery rate

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

MF

Molecular Function

PLS-DA

partial least squares discriminant analysis

PNOEC

predicted no effect concentration

RIN

RNA Integrity Number

SSNRI

Selective Serotonin and Norepinephrine Reuptake Inhibitor

SSRI

Selective Serotonin Reuptake Inhibitor

USGS

United States Geological Survey

VEN

venlafaxine

vst

variance stabilizing transformation

WWTP

wastewater treatment plant