Article pubs.acs.org/est
Transcriptional and Physiological Responses Induced by Binary Mixtures of Drospirenone and Progesterone in Zebrafish (Danio rerio) Sara Zucchi,†,∥ Leda Mirbahai,†,∥,⊥ Sara Castiglioni,‡ and Karl Fent*,†,§ †
University of Applied Sciences Northwestern Switzerland, School of Life Sciences, Gründenstrasse 40, CH−4132 Muttenz, Switzerland ‡ IRCCS − Istituto di Ricerche Farmacologiche “Mario Negri”, Environmental Biomarkers Unit, Department of Environmental Health Sciences, Via La Masa 19, I-20156, Milan, Italy § Swiss Federal Institute of Technology (ETH Zürich), Institute of Biogeochemistry and Pollution Dynamics, Department of Environmental Systems Science, CH−8092 Zürich, Switzerland S Supporting Information *
ABSTRACT: Drospirenone (DRS) is a synthetic progestin increasingly used in oral contraceptives with similar effects to progesterone (P4). Wild fish are exposed to DRS and P4 through wastewater. However, the effects of DRS on fish, both as an individual compound and in mixtures, have not been extensively studied. Therefore, in this study, global gene expression profiles of ovary and brain of female zebrafish (Danio rerio) were characterized after exposure to 55, 553, and 5442 ng/L DRS for 14 days. The effects were then compared to the observed responses after exposure to mixtures of DRS and P4 (DRS+P4: 27 + 0.8, 277 + 8 and 3118 + 123 ng/L). Transcriptomics findings were related to the changes in vitellogenin protein concentrations in the blood, morphology, and histology of gonads. Multivariate analysis indicated tissue-, dose-, and treatment-dependent expression profiles. Genes involved in steroid hormone receptor activity and circadian rhythm were enriched in DRS and mixture groups, among other pathways. In mixtures, the magnitude of response was dose- and transcript-dependent, both at the molecular and physiological levels. Effects of DRS and P4 were additive for most of the investigated parameters and occurred at environmentally relevant concentrations. They may translate to adverse reproductive effects in fish.
■
respectively (www.imshealth.com).10 Residues of P4 and metabolites as well as synthetic progestins originating from sewage are detected at ng/L in aquatic environments. P4 also originates from livestock farms, where up to 350 ng/L P4 was detected in surface- and ground-waters located downstream.11 Although DRS levels have not been reported in aquatic environments, it is thought that their levels are similar to other steroid hormones in surface water (low ng/L range). An increase in DRS levels is expected due to its increased usage in the new generation of contraceptives (e.g., Yasmin). As a derivative of spironolactone, DRS exhibits antimineralcorticoid (antialdosterone) and slight antiandrogenic activity in mammals12−14 and androgenic activity in recombinant yeast.3 Furthermore, DRS altered fecundity and gonad histology in fathead minnows at 6.5 and 70 μg/L,4 but not at 100 ng/L.3
INTRODUCTION Natural and synthetic steroid hormones are among the most active endocrine disrupters, entering aquatic systems via excretion by humans and livestock. They may adversely affect reproduction of fish at environmental levels.1,2 Less investigated synthetic progestins (gestagens) used in oral contraceptives and hormone replacement therapies equally display strong hormonal activities in fish.3 Adverse outcomes in reproduction have been found in fathead minnows (Pimephales promelas) 4 and sticklebacks 5 exposed to levonorgestrel or norethindrone,6 gestodene,3 and desogestrel.3 Adverse effects were also observed in zebrafish exposed to ng/L concentrations of progesterone (P4)7,8 and the antiprogestin mifepristone.8,9 To date, knowledge on the modes of action (MOA) of progestins is scarce, and the effects of progestin mixtures are unknown. The consumption of P4 in medicine was 500 kg in Switzerland in 2010, and it is higher than the sum of all synthetic progestins (213 kg) mainly used in contraceptives (www.imshealth.com). Of the synthetic progestins, drospirenone (DRS) shows the highest consumption of over 100 and 153 kg in Switzerland in 2009 and in the U.K. in 2006, © 2014 American Chemical Society
Received: Revised: Accepted: Published: 3523
December 17, 2013 February 21, 2014 February 25, 2014 February 25, 2014 dx.doi.org/10.1021/es405584f | Environ. Sci. Technol. 2014, 48, 3523−3531
Environmental Science & Technology
Article
In the natural environment, fish are exposed to mixtures of anthropogenic compounds, including progestins. There is a need for better understanding of the MOA in particular within the context of the combined effects of mixtures. Effects of mixtures of estrogens,15 estrogen mimicking UV-filters,16 and antiandrogenic phthalates have been studied in vitro.17 However, there are only a few studies on mixture effects at the transcription and physiological levels. Mixtures of androgenic trenbolone and antiandrogenic flutamide18 and binary mixtures of estrogen and antiestrogen have been studied in fathead minnows.19 However, studies on effects of mixtures of progestins on transcriptome and physiological endpoints in fish are lacking. Combined effects of compounds that act through similar mechanisms are additive and follow the concentration addition (CA) model.15−17 Here, we hypothesize that DRS and P4 act by a similar MOA resulting in similar transcriptomic and physiological effects. Second, we hypothesize that the combinatory effects of the two progestins follows the CA model reflecting additive interaction.20 To test these hypotheses, we compared effects of DRS to P4 at several endpoints, including transcriptomics, effects on vitellogenin (VTG) protein content, as well as histological alterations and morphological effects. Second, we tested the CA hypothesis by analyzing the magnitude of response to binary mixtures of P4 and DRS in zebrafish. Using data on P48 and the current DRS experiment, we demonstrate a similar MOA for both progestins as well as demonstrating a CA effect for the mixtures. Our data clearly highlight the endocrine disrupting activity of these compounds even at environmental levels. The data illustrate that mixture effects follow the CA model and the effects are both transcript- and dose-dependent.
replicates were dissected immediately. Ovary, brain (including pituitary gland), and liver samples were transferred to RNAlater and stored at −80 °C for subsequent RNA extraction. Ovaries were weighed in order to assess the gonadosomatic index (GSI = gonad weight (g)/body weight (g) ×100). Chemical Analysis. The same analytical methods described in our previous publications8,21 were used to determine the concentrations of DRS and P4 in the exposure waters (SI Table S2A). The recoveries for DRS and P4 were 89.5 ± 4.2% and 85.7 ± 1.5%, respectively. Detailed information about the analytical procedures (solid-phase extraction and liquid chromatographic-tandem mass spectrometry) used for chemical analysis is provided in the SI. RNA Isolation and Microarray Hybridization. Total RNA was extracted from ovaries of individual zebrafish. To obtain sufficient amounts of RNA, brain and liver samples from 6 technical replicates were pooled and RNA was extracted using the RNeasy Mini Kit (Qiagen, Switzerland; four biological replicates per condition), as previously described.8 Brain (including pituitary gland) and ovary samples were used for microarray analysis, while liver samples were only used for RTPCR analysis. The tissue selections were done according to our previous P4 study,8 allowing us to compare responses between our previous female zebrafish studies. Microarray experiments were carried out by the Functional Genomics Centre (ETH and University of Zürich, Switzerland) using the single-color zebrafish gene expression oligonucleotide microarray in a 4 × 44K slide format (Agilent, design ID G2519F-026437) as described previously.8 In total, 16 microarrays were analyzed for ovary (DRS treated; 4 biological replicates per dose group and 4 controls) and 32 for brain (DRS and DRS+P4 treated; 4 biological replicates per dose group and 4 controls). Due to resource limitations, only brain tissue samples were used for transcriptomics analysis in mixtures. The tissue selection was based on our previous work, whereby the most significant changes were observed in the zebrafish brain rather than ovaries after exposure to P4.8 RT-PCR Analysis. RT-PCR analyses were performed according to the method described previously.8 Briefly, RNA samples were reverse-transcribed using the iScript cDNA Synthesis Kit (BIORAD, Switzerland). RT-PCR was conducted on BIORAD CFX96 Real-Time PCR Detection System (BIORAD, Switzerland) using SYBR Green Fluorescence (BIORAD, Switzerland) as recommended by the manufacture’s guidelines (primer details are presented in the SI). For RTPCR analyses, five biological replicates, with two technical replicates, were analyzed. Melting curves were analyzed to ensure only a single product was amplified. Primer efficiencies were calculated to ensure no significant change between the primer efficiencies of the target genes and the reference gene (18S rRNA). Threshold cycle (CT) values were recorded in the linear phase of amplification and the data were analyzed using the delta−delta CT method of relative quantification.22 Vitellogenin Protein Analysis. Blood was collected from individual fish (n = 30 samples per dose group; 6 technical replicates per 5 biological replicates) immediately after anesthesia, as described in the SI and our previous publication.8 Plasma VTG concentrations were measured using a zebrafishspecific kit (Biosense Laboratories AS, Norway) according to manufacture’s guidelines. Histology. After anesthesia, one fish per biological replicate (total of n = 5 per treatment) was dissected at the abdominalsite and fixed in Bouin’s solution for about 24 h. Slides were
■
MATERIALS AND METHODS Chemicals and Maintenance of Zebrafish. This information can be found in the Supporting Information (SI). Experimental Design. The experimental design is explained in detail in the SI. Briefly, adult female zebrafish (10 months old; 5 replicates per dose and 7 fish per replicate) were exposed to either increasing concentrations of DRS (nominal concentration of 50, 500, and 5000 ng/L) or a mixture of DRS and P4 (nominal concentration of 50 + 4, 500 + 40, and 5000 + 400 ng/L) for 14 days with both water (reconstituted water) and solvent controls (0.01% DMSO). Concentrations of P4 were within the range of reported environmental levels. DRS concentrations were chosen based on the assumed environmental levels (low and intermediate dose)11 and the highest dose as effective dose.4 The concentrations in mixtures were chosen to test the CA model on the basis of the single compounds; measured concentrations were about 50% of those of the single progestins. For practical reasons, no full-dose−response curves could be generated and therefore, the fixed equi-effective concentration approach16 could not be used in the mixtures. The flow-through system and fish were monitored daily (details in the SI). The study was conducted according to the OECD guideline 204. At the end of exposures, fish were anesthetized by KoiMed Sleep (1.5 mL/L water; Koi & Bonsai Zimmermann, Switzerland). Weight (g) and length (mm) of each fish were recorded and used to calculate the condition factor. Furthermore, blood samples were collected from individual fish, as described in the SI and stored at −80 °C for VTG protein analysis. Six fish from each of the five biological 3524
dx.doi.org/10.1021/es405584f | Environ. Sci. Technol. 2014, 48, 3523−3531
Environmental Science & Technology
Article
Figure 1. Multivariate analysis of the transcriptomics data. Concentrations in binary mixtures were about 50% as in single exposures of P4 and DRS. (A) Hierarchical clustering analysis of 37 507 entities. Red: over-expressed; Yellow: no change; Blue: under-expressed, all compared to control. (B) Principal component analyses of the 35 452 entities that were used for differentially expressed gene analysis. M: mixture; P4: progesterone; DRS: drospirenone; _L: lowest concentration; _I: intermediate concentration; _H: highest concentration.
marginal and absent entities. Statistically significantly changing genes were identified by one-way ANOVA with a multiple testing correction25 for a false-discovery rate (FDR) of 1.5 was additionally applied to the list of differentially expressed genes (DEG). The Gene Ontology (GO) function in GeneSpring and the “Functional Annotation Clustering” tool in Database for Annotation, Visualization, and Integrated Discovery (DAVID; v6.7)26,27 were used to aid interpretation of the DEGs and to determine significantly enriched biological functions and canonical pathways represented by these gene sets. SPSS v16 software was used for statistical analysis of the nontranscriptomic data, as described in the SI.
prepared and histological analyses were conducted as described.8,23 Ovaries from five females were selected per treatment, and two sections (out of three) from different regions (depth) of the ovary were examined in each individual. Ovary staging and histological alterations were evaluated according to guidelines (OECD, 2010) and our previous studies.8,23 Investigating Additive Mixture Effects of Progestins. Additive effects in the mixtures were investigated for selected transcripts (nr1d2b, per1b, and vtg1), VTG protein levels and histological endpoints. The measured response was plotted against concentration for individual compounds. A logarithmic curve was fitted to the data, and the response for the concentrations used in individual exposures was predicted based on the equation. The difference between the observed and predicted responses for individual compounds was quantified and used as margin of error (MOE, the difference between observed and predicted response) for the equations. The expected response was predicted for the concentrations used in the mixture studies ± MOE. The expected additive effect (A+B) was calculated based on the predicted response and compared to the observed response (described in detail in SI Tables S10, S11, S13). Statistical Analysis. MIAME-compliant raw microarray data were submitted to ArrayExpress at EMBL-EBI and can be found under accession E-MTAB-2077. GeneSpring vGX11.5 (Agilent) and MultiExperimental Viewer v4.724 were used for analyzing the data. Data were quantile normalized and analyzed at the gene level following initial quality checks and removal of
■
RESULTS AND DISCUSSION Chemical Analysis. Measured DRS concentrations were close to nominal (SI Table S2B) with mean exposure concentrations of 55, 553, and 5442 ng/L. Measured mixture (DRS+P4) concentrations differed from nominal (SI Table S2C). DRS concentrations in the mixture were 40% lower than nominal with mean concentrations of 27, 227, and 3118 ng/L, while mean P4 concentrations of 0.8, 8, and 123 ng/L in the mixture were 70−80% lower than nominal. Detailed results and discussion of the lower than nominal concentrations (due to apparatus, sorption to tubing, fish debris etc.) and compound stability are provided in the SI. Morphometric Response to DRS and Mixture. At the end of the exposures, somatic endpoints were assessed (SI Figure S1A). No significant change was detected in the 3525
dx.doi.org/10.1021/es405584f | Environ. Sci. Technol. 2014, 48, 3523−3531
Environmental Science & Technology
Article
concentrations led to lower gonado-somatic index (GSI) (SI Figure S1B), being significant at 5442 ng/L of DRS. The reduction may be related to the effects of DRS and the mixture on gonad morphology and histology, as discussed below. Multivariate Analysis of Transcriptomics Data Reveals Tissue-, Treatment- and Dose-Dependent Transcription Profiles. Transcription profiles of ovary and brain of DRS exposed zebrafish were determined by microarrays. To test the combined effects of DRS and P4, we conducted a transcriptomics study in the brain of adult zebrafish exposed to three mixture (DRS+P4) concentrations and compared the data with those of the single DRS and P4 8 exposures. Hierarchical clustering (HAC) of the genes (total number of 35 452) used for differentially expressed gene (DEG) analysis, demonstrated a clear separation and clustering of the data set based on tissue type and treatment (Figure 1A). The MOA of DRS is demonstrated to be similar to P4. 8 Principal Component Analysis (PCA) showed a clear separation of the DRS and mixture groups from the P4 groups along the PC1 axis with the mixture separating from the DRS group along the PC2 axis (Figure 1B). Both PCA and HAC demonstrated that the transcriptional response of the mixture resembles more to that of DRS than P4 (Figure 1). This was expected due to the higher comparative dose for DRS than P4 in the mixture. Gene Expression Profiling and Functional Genomics Analysis: Dose−Response and Cues to Mechanisms of Mixture Activity. The results of the microarray analysis revealed that exposure to 55, 553, and 5442 ng/L of DRS resulted in statistically significant differential expression of 38, 22, and 19 genes (FC ≥1.5, FDR < 0.05; SI Table S3), respectively, in the ovary. In the brain, transcriptomics analysis
Figure 2. Venn diagram of the differentially expressed genes (DEG; FDR
Transcriptional and Physiological Responses Induced by Binary Mixtures of Drospirenone and Progesterone in Zebrafish (Danio rerio) Sara Zucchi,†,∥ Leda Mirbahai,†,∥,⊥ Sara Castiglioni,‡ and Karl Fent*,†,§ †
University of Applied Sciences Northwestern Switzerland, School of Life Sciences, Gründenstrasse 40, CH−4132 Muttenz, Switzerland ‡ IRCCS − Istituto di Ricerche Farmacologiche “Mario Negri”, Environmental Biomarkers Unit, Department of Environmental Health Sciences, Via La Masa 19, I-20156, Milan, Italy § Swiss Federal Institute of Technology (ETH Zürich), Institute of Biogeochemistry and Pollution Dynamics, Department of Environmental Systems Science, CH−8092 Zürich, Switzerland S Supporting Information *
ABSTRACT: Drospirenone (DRS) is a synthetic progestin increasingly used in oral contraceptives with similar effects to progesterone (P4). Wild fish are exposed to DRS and P4 through wastewater. However, the effects of DRS on fish, both as an individual compound and in mixtures, have not been extensively studied. Therefore, in this study, global gene expression profiles of ovary and brain of female zebrafish (Danio rerio) were characterized after exposure to 55, 553, and 5442 ng/L DRS for 14 days. The effects were then compared to the observed responses after exposure to mixtures of DRS and P4 (DRS+P4: 27 + 0.8, 277 + 8 and 3118 + 123 ng/L). Transcriptomics findings were related to the changes in vitellogenin protein concentrations in the blood, morphology, and histology of gonads. Multivariate analysis indicated tissue-, dose-, and treatment-dependent expression profiles. Genes involved in steroid hormone receptor activity and circadian rhythm were enriched in DRS and mixture groups, among other pathways. In mixtures, the magnitude of response was dose- and transcript-dependent, both at the molecular and physiological levels. Effects of DRS and P4 were additive for most of the investigated parameters and occurred at environmentally relevant concentrations. They may translate to adverse reproductive effects in fish.
■
respectively (www.imshealth.com).10 Residues of P4 and metabolites as well as synthetic progestins originating from sewage are detected at ng/L in aquatic environments. P4 also originates from livestock farms, where up to 350 ng/L P4 was detected in surface- and ground-waters located downstream.11 Although DRS levels have not been reported in aquatic environments, it is thought that their levels are similar to other steroid hormones in surface water (low ng/L range). An increase in DRS levels is expected due to its increased usage in the new generation of contraceptives (e.g., Yasmin). As a derivative of spironolactone, DRS exhibits antimineralcorticoid (antialdosterone) and slight antiandrogenic activity in mammals12−14 and androgenic activity in recombinant yeast.3 Furthermore, DRS altered fecundity and gonad histology in fathead minnows at 6.5 and 70 μg/L,4 but not at 100 ng/L.3
INTRODUCTION Natural and synthetic steroid hormones are among the most active endocrine disrupters, entering aquatic systems via excretion by humans and livestock. They may adversely affect reproduction of fish at environmental levels.1,2 Less investigated synthetic progestins (gestagens) used in oral contraceptives and hormone replacement therapies equally display strong hormonal activities in fish.3 Adverse outcomes in reproduction have been found in fathead minnows (Pimephales promelas) 4 and sticklebacks 5 exposed to levonorgestrel or norethindrone,6 gestodene,3 and desogestrel.3 Adverse effects were also observed in zebrafish exposed to ng/L concentrations of progesterone (P4)7,8 and the antiprogestin mifepristone.8,9 To date, knowledge on the modes of action (MOA) of progestins is scarce, and the effects of progestin mixtures are unknown. The consumption of P4 in medicine was 500 kg in Switzerland in 2010, and it is higher than the sum of all synthetic progestins (213 kg) mainly used in contraceptives (www.imshealth.com). Of the synthetic progestins, drospirenone (DRS) shows the highest consumption of over 100 and 153 kg in Switzerland in 2009 and in the U.K. in 2006, © 2014 American Chemical Society
Received: Revised: Accepted: Published: 3523
December 17, 2013 February 21, 2014 February 25, 2014 February 25, 2014 dx.doi.org/10.1021/es405584f | Environ. Sci. Technol. 2014, 48, 3523−3531
Environmental Science & Technology
Article
In the natural environment, fish are exposed to mixtures of anthropogenic compounds, including progestins. There is a need for better understanding of the MOA in particular within the context of the combined effects of mixtures. Effects of mixtures of estrogens,15 estrogen mimicking UV-filters,16 and antiandrogenic phthalates have been studied in vitro.17 However, there are only a few studies on mixture effects at the transcription and physiological levels. Mixtures of androgenic trenbolone and antiandrogenic flutamide18 and binary mixtures of estrogen and antiestrogen have been studied in fathead minnows.19 However, studies on effects of mixtures of progestins on transcriptome and physiological endpoints in fish are lacking. Combined effects of compounds that act through similar mechanisms are additive and follow the concentration addition (CA) model.15−17 Here, we hypothesize that DRS and P4 act by a similar MOA resulting in similar transcriptomic and physiological effects. Second, we hypothesize that the combinatory effects of the two progestins follows the CA model reflecting additive interaction.20 To test these hypotheses, we compared effects of DRS to P4 at several endpoints, including transcriptomics, effects on vitellogenin (VTG) protein content, as well as histological alterations and morphological effects. Second, we tested the CA hypothesis by analyzing the magnitude of response to binary mixtures of P4 and DRS in zebrafish. Using data on P48 and the current DRS experiment, we demonstrate a similar MOA for both progestins as well as demonstrating a CA effect for the mixtures. Our data clearly highlight the endocrine disrupting activity of these compounds even at environmental levels. The data illustrate that mixture effects follow the CA model and the effects are both transcript- and dose-dependent.
replicates were dissected immediately. Ovary, brain (including pituitary gland), and liver samples were transferred to RNAlater and stored at −80 °C for subsequent RNA extraction. Ovaries were weighed in order to assess the gonadosomatic index (GSI = gonad weight (g)/body weight (g) ×100). Chemical Analysis. The same analytical methods described in our previous publications8,21 were used to determine the concentrations of DRS and P4 in the exposure waters (SI Table S2A). The recoveries for DRS and P4 were 89.5 ± 4.2% and 85.7 ± 1.5%, respectively. Detailed information about the analytical procedures (solid-phase extraction and liquid chromatographic-tandem mass spectrometry) used for chemical analysis is provided in the SI. RNA Isolation and Microarray Hybridization. Total RNA was extracted from ovaries of individual zebrafish. To obtain sufficient amounts of RNA, brain and liver samples from 6 technical replicates were pooled and RNA was extracted using the RNeasy Mini Kit (Qiagen, Switzerland; four biological replicates per condition), as previously described.8 Brain (including pituitary gland) and ovary samples were used for microarray analysis, while liver samples were only used for RTPCR analysis. The tissue selections were done according to our previous P4 study,8 allowing us to compare responses between our previous female zebrafish studies. Microarray experiments were carried out by the Functional Genomics Centre (ETH and University of Zürich, Switzerland) using the single-color zebrafish gene expression oligonucleotide microarray in a 4 × 44K slide format (Agilent, design ID G2519F-026437) as described previously.8 In total, 16 microarrays were analyzed for ovary (DRS treated; 4 biological replicates per dose group and 4 controls) and 32 for brain (DRS and DRS+P4 treated; 4 biological replicates per dose group and 4 controls). Due to resource limitations, only brain tissue samples were used for transcriptomics analysis in mixtures. The tissue selection was based on our previous work, whereby the most significant changes were observed in the zebrafish brain rather than ovaries after exposure to P4.8 RT-PCR Analysis. RT-PCR analyses were performed according to the method described previously.8 Briefly, RNA samples were reverse-transcribed using the iScript cDNA Synthesis Kit (BIORAD, Switzerland). RT-PCR was conducted on BIORAD CFX96 Real-Time PCR Detection System (BIORAD, Switzerland) using SYBR Green Fluorescence (BIORAD, Switzerland) as recommended by the manufacture’s guidelines (primer details are presented in the SI). For RTPCR analyses, five biological replicates, with two technical replicates, were analyzed. Melting curves were analyzed to ensure only a single product was amplified. Primer efficiencies were calculated to ensure no significant change between the primer efficiencies of the target genes and the reference gene (18S rRNA). Threshold cycle (CT) values were recorded in the linear phase of amplification and the data were analyzed using the delta−delta CT method of relative quantification.22 Vitellogenin Protein Analysis. Blood was collected from individual fish (n = 30 samples per dose group; 6 technical replicates per 5 biological replicates) immediately after anesthesia, as described in the SI and our previous publication.8 Plasma VTG concentrations were measured using a zebrafishspecific kit (Biosense Laboratories AS, Norway) according to manufacture’s guidelines. Histology. After anesthesia, one fish per biological replicate (total of n = 5 per treatment) was dissected at the abdominalsite and fixed in Bouin’s solution for about 24 h. Slides were
■
MATERIALS AND METHODS Chemicals and Maintenance of Zebrafish. This information can be found in the Supporting Information (SI). Experimental Design. The experimental design is explained in detail in the SI. Briefly, adult female zebrafish (10 months old; 5 replicates per dose and 7 fish per replicate) were exposed to either increasing concentrations of DRS (nominal concentration of 50, 500, and 5000 ng/L) or a mixture of DRS and P4 (nominal concentration of 50 + 4, 500 + 40, and 5000 + 400 ng/L) for 14 days with both water (reconstituted water) and solvent controls (0.01% DMSO). Concentrations of P4 were within the range of reported environmental levels. DRS concentrations were chosen based on the assumed environmental levels (low and intermediate dose)11 and the highest dose as effective dose.4 The concentrations in mixtures were chosen to test the CA model on the basis of the single compounds; measured concentrations were about 50% of those of the single progestins. For practical reasons, no full-dose−response curves could be generated and therefore, the fixed equi-effective concentration approach16 could not be used in the mixtures. The flow-through system and fish were monitored daily (details in the SI). The study was conducted according to the OECD guideline 204. At the end of exposures, fish were anesthetized by KoiMed Sleep (1.5 mL/L water; Koi & Bonsai Zimmermann, Switzerland). Weight (g) and length (mm) of each fish were recorded and used to calculate the condition factor. Furthermore, blood samples were collected from individual fish, as described in the SI and stored at −80 °C for VTG protein analysis. Six fish from each of the five biological 3524
dx.doi.org/10.1021/es405584f | Environ. Sci. Technol. 2014, 48, 3523−3531
Environmental Science & Technology
Article
Figure 1. Multivariate analysis of the transcriptomics data. Concentrations in binary mixtures were about 50% as in single exposures of P4 and DRS. (A) Hierarchical clustering analysis of 37 507 entities. Red: over-expressed; Yellow: no change; Blue: under-expressed, all compared to control. (B) Principal component analyses of the 35 452 entities that were used for differentially expressed gene analysis. M: mixture; P4: progesterone; DRS: drospirenone; _L: lowest concentration; _I: intermediate concentration; _H: highest concentration.
marginal and absent entities. Statistically significantly changing genes were identified by one-way ANOVA with a multiple testing correction25 for a false-discovery rate (FDR) of 1.5 was additionally applied to the list of differentially expressed genes (DEG). The Gene Ontology (GO) function in GeneSpring and the “Functional Annotation Clustering” tool in Database for Annotation, Visualization, and Integrated Discovery (DAVID; v6.7)26,27 were used to aid interpretation of the DEGs and to determine significantly enriched biological functions and canonical pathways represented by these gene sets. SPSS v16 software was used for statistical analysis of the nontranscriptomic data, as described in the SI.
prepared and histological analyses were conducted as described.8,23 Ovaries from five females were selected per treatment, and two sections (out of three) from different regions (depth) of the ovary were examined in each individual. Ovary staging and histological alterations were evaluated according to guidelines (OECD, 2010) and our previous studies.8,23 Investigating Additive Mixture Effects of Progestins. Additive effects in the mixtures were investigated for selected transcripts (nr1d2b, per1b, and vtg1), VTG protein levels and histological endpoints. The measured response was plotted against concentration for individual compounds. A logarithmic curve was fitted to the data, and the response for the concentrations used in individual exposures was predicted based on the equation. The difference between the observed and predicted responses for individual compounds was quantified and used as margin of error (MOE, the difference between observed and predicted response) for the equations. The expected response was predicted for the concentrations used in the mixture studies ± MOE. The expected additive effect (A+B) was calculated based on the predicted response and compared to the observed response (described in detail in SI Tables S10, S11, S13). Statistical Analysis. MIAME-compliant raw microarray data were submitted to ArrayExpress at EMBL-EBI and can be found under accession E-MTAB-2077. GeneSpring vGX11.5 (Agilent) and MultiExperimental Viewer v4.724 were used for analyzing the data. Data were quantile normalized and analyzed at the gene level following initial quality checks and removal of
■
RESULTS AND DISCUSSION Chemical Analysis. Measured DRS concentrations were close to nominal (SI Table S2B) with mean exposure concentrations of 55, 553, and 5442 ng/L. Measured mixture (DRS+P4) concentrations differed from nominal (SI Table S2C). DRS concentrations in the mixture were 40% lower than nominal with mean concentrations of 27, 227, and 3118 ng/L, while mean P4 concentrations of 0.8, 8, and 123 ng/L in the mixture were 70−80% lower than nominal. Detailed results and discussion of the lower than nominal concentrations (due to apparatus, sorption to tubing, fish debris etc.) and compound stability are provided in the SI. Morphometric Response to DRS and Mixture. At the end of the exposures, somatic endpoints were assessed (SI Figure S1A). No significant change was detected in the 3525
dx.doi.org/10.1021/es405584f | Environ. Sci. Technol. 2014, 48, 3523−3531
Environmental Science & Technology
Article
concentrations led to lower gonado-somatic index (GSI) (SI Figure S1B), being significant at 5442 ng/L of DRS. The reduction may be related to the effects of DRS and the mixture on gonad morphology and histology, as discussed below. Multivariate Analysis of Transcriptomics Data Reveals Tissue-, Treatment- and Dose-Dependent Transcription Profiles. Transcription profiles of ovary and brain of DRS exposed zebrafish were determined by microarrays. To test the combined effects of DRS and P4, we conducted a transcriptomics study in the brain of adult zebrafish exposed to three mixture (DRS+P4) concentrations and compared the data with those of the single DRS and P4 8 exposures. Hierarchical clustering (HAC) of the genes (total number of 35 452) used for differentially expressed gene (DEG) analysis, demonstrated a clear separation and clustering of the data set based on tissue type and treatment (Figure 1A). The MOA of DRS is demonstrated to be similar to P4. 8 Principal Component Analysis (PCA) showed a clear separation of the DRS and mixture groups from the P4 groups along the PC1 axis with the mixture separating from the DRS group along the PC2 axis (Figure 1B). Both PCA and HAC demonstrated that the transcriptional response of the mixture resembles more to that of DRS than P4 (Figure 1). This was expected due to the higher comparative dose for DRS than P4 in the mixture. Gene Expression Profiling and Functional Genomics Analysis: Dose−Response and Cues to Mechanisms of Mixture Activity. The results of the microarray analysis revealed that exposure to 55, 553, and 5442 ng/L of DRS resulted in statistically significant differential expression of 38, 22, and 19 genes (FC ≥1.5, FDR < 0.05; SI Table S3), respectively, in the ovary. In the brain, transcriptomics analysis
Figure 2. Venn diagram of the differentially expressed genes (DEG; FDR
