Perfluorooctane sulfonate (PFOS) exposure of bovine oocytes affects early embryonic development at human-relevant levels in an in vitro model

Perfluorooctane sulfonate (PFOS) has been added to Stockholm Convention for global phase out, but will continue to contribute to the chemical burden in humans for a long time to come due to extreme persistence in the environment. In the body, PFOS is transferred into to the ovarian follicular fluid that surrounds the maturing oocyte. In the present study, bovine cumulus oocyte complexes were exposed to PFOS during 22 h in vitro maturation. Concentrations of 2 ng g -1 (PFOS-02) representing average human exposure and 53 ng g -1 (PFOS-53) relevant to highly exposed groups were used. After exposure, developmental competence was recorded until day 8 after fertilisation. Blastocysts were fixed and either stained to evaluate blastomere number and lipid distribution using confocal microscopy or frozen and pooled for microarray-based gene expression and DNA methylation analyses. PFOS-53 delayed the first cleavage to two-cell stage and beyond at 44 h after fertilisation ( p < .01). No reduction of proportion blastocysts were seen at day 8 in either of the groups, but PFOS-53 exposure resulted in delayed development into more advanced stages of blastocysts seen as both reduced developmental stage ( p = .001) and reduced number of blastomeres ( p = .04). Blastocysts showed an altered lipid distribution that was more pronounced after exposure to PFOS-53 (increased total lipid volume, p = .0003, lipid volume/cell p < .0001) than PFOS-02, where only decreased average lipid droplet size ( p = .02) was observed. Gene expression analyses revealed pathways differently regulated in the PFOS-treated groups compared to the controls, which were related to cell death and survival through e.g., P38 mitogen-activated protein kinases and signal transducer and activator of transcription 3, which in turn activates tumour protein 53 (TP53) . Transcriptomic changes were also associated with metabolic stress response, differentiation and proliferation, which could help to explain the phenotypic changes seen in the blastocysts. The gene expression changes were more pronounced after exposure to PFOS-53 compared to PFOS-02. DNA-methylation changes were associated with similar biological functions as the transcriptomic data, with the most significantly associated pathway being TP53. Collectively, these results reveal that brief PFOS exposure during oocyte maturation alters the early embryo development at concentrations relevant to humans. This study adds to the evidence that PFOS has the potential to affect female fertility.


Introduction
Per-and polyfluoroalkyl substances (PFASs) are a large group of man-made chemicals used in a range of industrial applications and consumer products (Gluge et al., 2020).Due to their wide use and persistent nature, PFASs are found worldwide, exposing humans through food, drinking water, inhalation of dust and dermal contact (DeLuca et al., 2021;Poothong et al., 2020).During the last 60 years, perfluorooctane sulfonate (PFOS) and perfluorooctanoic acid (PFOA) have been the most widely used PFASs.Since potential health effects have been associated with exposure, phase-out of production and distribution of PFOS was initiated in the beginning of 21 st century and PFOS was included in Annex B of the Stockholm Convention of Persistent Organic Pollutants in 2009(UNEP, 2009).This led to substitution with other PFASs in production.In general, phase-out of PFOS seems to result in decreased human serum levels (Bjerregaard-Olesen et al., 2016;Calafat et al., 2007;Glynn et al., 2012).Yet, due to their long half-life (Li et al., 2018;Zhang et al., 2013), it is inevitable that the exposure will persist for decades to come.PFOS is still the most detected PFAS in environmental and biological samples and present in serum in humans worldwide (Kannan et al., 2004).The average PFOS serum concentration measured in humans ranges from 4− 20 ng mL − 1 (Bjerregaard- Olesen et al., 2016;Calafat et al., 2007;Kim et al., 2020).Occupational exposure, as well as local contamination of drinking water can widely exceed the exposure of the general population (Li et al., 2018), and fire fighters with occupational exposure have significantly higher concentrations compared to controls (Rotander et al., 2015).
PFOS has been associated with reduced birth weight and poorer immunisation in response to vaccines in children, and to higher serum cholesterol in adults (Steenland et al., 2009).Furthermore, PFASs cross the ovarian blood-follicle barrier and have been found in ovarian follicular fluid in similar concentrations as in serum (Heffernan et al., 2018;Kang et al., 2020;McCoy et al., 2017;Petro et al., 2014).It also passes the placenta, exposing the foetus in utero (Mamsen et al., 2019) and is further secreted in milk, exposing the nursing infant (Kärrman et al., 2007).The evidence for adverse health effects related to reproductive outcomes in humans has been found to be insufficient to draw anyconclusions (Schrenk et al., 2020).
The process of forming a healthy offspring starts with the germ cell.In the female, the developmental competence of the oocyte is dependent on the folliculogenesis where the environment surrounding the oocyte consists of, for example, growth factors and sex steroids stimulating oocyte development.The final stage of oocyte maturation (from late antral stage) is an especially sensitive period where many critical events take place.At this point, the oocyte will resume the first meiosis and final nuclear and cytoplasmic maturation will occur in parallel with preparation for the genetic program that will drive early embryo development.Oocyte characteristics, as well as timing of events during early embryo development, are significantly different between species (Ménézo and Hérubel, 2002).The bovine oocyte maturation and early embryo development share many features with human oocyte maturation and early pre-implantation embryo development, making the bovine in vitro model an attractive alternative to rodent studies for toxicological and reproductive studies (Alm et al., 1998;Grossman et al., 2012;Hallberg et al., 2019;Jorssen et al., 2015;Krogenaes et al., 1998;Ménézo and Hérubel, 2002;Santos et al., 2014).The in vitro embryo production (IVP) set-up enables the study of specific windows, including the final 22 h of oocyte maturation (in vitro maturation, IVM) and the consequence for the subsequent early embryo development.IVP allows for phenotypic studies of the pre-implantation embryo to detect alterations during development using morphological assessments (IETS, 2010), including staining for further phenotypic characterisations (Gonzalez and Sjunnesson, 2013;Hallberg et al., 2019;Laskowski et al., 2017).Furthermore, microarray platforms, and more recently RNA sequencing techniques, have been developed to study the early embryo transcriptome (Robert et al., 2011) and methylome profiles (Saadi et al., 2014), which further contributes to the understanding of embryo interaction with their microenvironment, with the benefit of studying pathways rather than individual genes.
Even though PFOS has been manufactured for more than 60 years, information about potential toxicity to the female gamete and the consequences for female fertility is scarce.This is especially true for concentrations relevant to humans.However, PFOS has been associated with outcomes related to ovarian function, as recently reviewed (Ding et al., 2020).Human cohorts have shown associations with PFOS and longer time to pregnancy and effects on the menstrual cycle (Fei et al., 2009;Lopez-Espinosa et al., 2011;Whitworth et al., 2012;Zhou et al., 2017), which is supported by experimental animal studies (Dominguez et al., 2016;Feng et al., 2015).Therefore, the aim of this study was to investigate whether and how PFOS exposure during final oocyte maturation affects the pre-implantation embryo development.Concentrations of PFOS were based on levels previously reported in human ovarian follicular fluid, and the bovine IVP model was used to study its impact on the morphological and phenotypic characteristics of the embryos as well as on gene expression and epigenetic patterns.

Experimental design
Bovine cumulus oocyte complexes (COCs) were exposed to PFOS for 22 h during the final stage of maturation in vitro using two different concentrations.Bovine COCs (n = 1,945) were collected from abattoirderived ovaries, randomly divided between exposed and control group, and subsequently kept separated during all following in vitro experiments.COCs (average 26, range 19-47) were matured, fertilised and cultured in vitro until day 8 post fertilisation (pf) under serum-free conditions.Developmental parameters were assessed during embryo culture (embryos cleaved and embryos cleaved beyond 2-cell stage 44 h pf, embryo development at day 7 and 8 pf, as well as stage and grade of developed embryos at day 7 and 8 pf).
At day 8 pf, blastocysts were fixed and stained for evaluation of blastomere count and lipid content using a confocal microscope (n = 78) or individually snap-frozen and pooled with 10 embryos in each group for later RNA and DNA extraction (n = 120).Four replicates for each experimental group were used for gene expression and DNA methylation analyses (Fig. 1).

Media and chemicals
Media were produced at the IVF laboratory at the Department of Clinical Sciences (Swedish University of Agricultural Sciences, Uppsala, Sweden) according to standard protocols (Gordon, 1994) as described previously in related projects at the facilities (Abraham et al., 2012;Hallberg et al., 2019;Laskowski et al., 2017).All chemicals and media used were purchased from Sigma Aldrich (Stockholm, Sweden) unless otherwise stated, details can be found in Tables S1 to S6 of the supporting information (SI).
PFOS (CAS no 1763-23-1, 77282, potassium salt ≥98.0 %) dissolved in molecular grade water (due to concentrations not exceeding water solubility of PFOS no vehicle other than water was used) was added to maturation media of the exposed groups to achieve 100 ng mL − 1 and 10 ng mL − 1 respectively.In the control, only vehicle (water) was added.PFOS stock solutions were stored at 4 • C protected from light; we did not expect degradation during the course of the experiments due to persistence of the compound.Due to PFOS loss during filtration of media, the final concentration was 2 ng mL − 1 and 53 ng mL − 1 , which is approximately 2-5 times under the intended dose, for details see Section 2.3.After in vitro maturation, media was frozen in − 20 • C and samples of media stored for later analyses.

Validation of exposure concentration
Samples from the maturation media (one sample from the control, three samples from for PFOS-02 and PFOS-53 respectively) were sent for quantification of PFOS concentration to SGS, Linköping, Sweden.Due to protein content (bovine serum-albumin, BSA), analysis procedure was set up based on in-house methods and as described by Gyllenhammar et al., 2017 for serum samples (Gyllenhammar et al., 2017) with small modifications.Samples (0.5 g) were weighted into 12 mL centrifuge tubes of polypropylene and spiked with isotopically labelled internal standards.A blank and control were prepared in parallel to the media.After addition of acetonitrile for extraction, samples were vortexed-mixed, sonicated and centrifuged for 10 min at 2000×g.The extract was transferred to a new 12 mL centrifuge tube and the extraction procedure was repeated on the remaining media.The extracts were combined and diluted with molecular grade water and loaded onto weak anion exchange (WAX, Oasis® WAX6 mL, 6 mL, 500 mg, 60 μm, Waters) which was preconditioned with 0.1 % NH 4 OH in methanol, methanol and water.The cartridges were rinsed with 1% formic acid in water, water and 0.1 % formic acid in methanol and then dried under vacuum.The samples were eluted with methanol and 0.1 % NH 4 OH and the cartridges were dried under vacuum.Following evaporation of the eluates using N 2 steam, the extracts were dissolved in methanol and further diluted in ammonium acetate before being analysed using a LC-MS/MS system (Agilent 1290 Infinity II/6495A).Results were adjusted to blank signal and internal standard recoveries were used to compensate for potential loss during processing.In the media from the control, the PFOS concentration was below detection limit (<0.1 ng g − 1 ).In PFOS-02 the concentration of PFOS was 2.05 ng g − 1 (range 2.0-2.1 ng g − 1 ) and in PFOS-53 the concentration was 53 ng g − 1 (52− 54 ng g − 1 ) which is the equivalent of 4 nM and 106 nM PFOS respectively (details can be found in supporting information, SI, Table S7).The lower concentration is equal to concentrations measured in human follicular fluid (Kim et al., 2020).The higher concentration is within the range of human relevance (Kang et al., 2020;Petro et al., 2014).

Oocyte recovery, in vitro maturation and exposure to PFOS
Bovine in vitro embryo production (IVP) was performed according to standard protocols (Gordon, 1994), and as previously described in related projects from our facilities (Abraham et al., 2012;Hallberg et al., 2019;Laskowski et al., 2016).Abattoir-derived ovaries from non-stimulated heifers and cows were collected and transported at 29 • C (range 28-30) within 3 h to the IVF laboratory facilities (Department of Clinical Sciences, Swedish University of Agricultural sciences, Uppsala).As the animals were not killed for the sake of this experiment, no ethical permission was required according to Swedish legislation.Acceptable COCs (compact multilayer cumulus investment and homogenous ooplasm, selection criteria according to (Gordon, 2003)) were randomly and equally distributed between the treatment/exposure groups.Incubation for 22 h for in vitro maturation followed, in bicarbonate buffered tissue culture medium 199 (TCM 199) supplemented with 0.68 mM L-Glutamine, 0.4 % w/v bovine serum albumin (BSA), Fraction V, 0.5 μg mL − 1 FSH, 0.1 μg mL -1 LH (Stimufol, PARTNAR Animal Health, Stoumont, Belgium) and 50 μg mL − 1 gentamicin sulfate (Table S2, SI).PFOS was added to the exposed groups and vehicle to the control in a volume of 10 μL mL -1 .

Evaluation of developmental competence to blastocyst stage
During IVP, developmental parameters were assessed as previously described (Hallberg et al., 2019).At 44 h pf, cleaved zygotes and zygotes cleaved beyond 2-cell stage were recorded.At day 7 and day 8 pf, proportion blastocysts as well as morphology and developmental stage of the embryos were recorded.The stage was assessed according to IETS grading system developed for assessing bovine embryos for trade, transfer and research (IETS, 2010), but modified into 3 stages; early blastocyst or blastocyst stage (stage 1).Expanding or expanded blastocysts where the zona pellucida has thinned due to blastocysts expansion (stage 2) and hatching or hatched blastocysts, where zona pellucida has started to or ruptured (stage 3).The quality of each embryo was graded according to IETS grading scheme (IETS, 2010) from grade 1-4, where grade 1 corresponds to a top quality embryo.An embryologist trained in the grading scored the embryos, and within the same replicate the same person scored all embryos but were not blinded to the treatments, as the practical laboratory work did not allow this.

Fixation and staining of blastocysts
Blastocysts from day 8 pf (n = 78) were fixed in 2 % paraformaldehyde at 4 • C overnight, followed by rinsing in phosphate buffered saline with 0.1 % polyvinyl alcohol (PBS-PVA).DNA was visualized by fluorescent labelling of DNA in 2.5μM Hoechst 33342 (B2261) incubated in RT for 15 min and rinsed with PBS-PVA followed by staining for visualising neutral lipids using LipidTOX™ (HCS Lip-idTOX™ Green Neutral Lipid Stain H34475, Thermo Fisher Scientific, Waltham, USA) for 30 min at RT according to the manufacturer's Fig. 1.Overview of the experimental design.Bovine oocyte cumulus complexes (COCs) were retrieved from the slaughterhouse after the animals have been killed.Exposure of PFOS occurred during in vitro maturation (IVM).Developmental competence of the oocyte to blastocyst stage were recorded during development until day 8 when phenotypic evaluation of the blastocysts were conducted.Stage and grade were evaluated using a light-microscope.A subset of blastocysts were stained for confocal imaging and DNA/RNA extraction for microarray analyses were performed on another subset of blastocysts (matched for stage and grade).instructions.To ensure absence of background fluorescence, a negative control accompanied the blastocyst during the staining procedures but was excluded from fluorescent labelling.Embryos were mounted on a microscope-slide (ER-201B-CE24, Thermo Fisher Scientific) in approximately 2 μL Vectashield (Vector Laboratories Inc., CA, US).

Confocal microscopy and image analysis
Stained embryos (n = 78) and negative control embryos (n = 3) were scanned using standard magnification (20×) in sections at 2 μm intervals (z-stack) using a confocal microscope (Zeiss LSM 800) equipped with He/Ne 543 and Ar 450/530 lasers.All steps of image processes were performed by an operator blinded to the codes and thereby the treatments of the individual blastocysts.Negative controls did not show detectable florescence.Images were analysed using Fiji for ImageJ (https://imagej.net/Fiji/)software.To separately detect nuclei and lipids in their respective colour channels, a local adaptive threshold approach was used that computes the optimal threshold level for each object to optimise an ellipse fit (Bombrun et al., 2017;Ranefall et al., 2016).The channels were aligned in each section as well as between sections.Random background noise (due to intensity threshold used for the signal detection) was reduced by applying a quality check based on the general stain after the normalisation.After computed-assisted image analysis, the results were manually validated by operator blinded to the codes, and images were excluded when extensive deviations were recorded, i.e. when the macro failed to recognise lipids and/or nuclei because of image quality (n = 4).The average volume of all lipid droplets within an embryo (average lipid droplet size, μm 3 ); the sum of all lipid droplet volume in each embryo (total lipid volume, μm 3 ); and total volume of lipids divided by number of cells (lipid volume/cell, μm 3 ) were recorded.

Statistical analysis
RStudio for R (R i386, 4.0.5,) was used for statistical comparisons.The effect of treatment on developmental parameters (proportion of cleaved, cleaved beyond two-cell stage and blastocysts on day 7 and day 8) was calculated using mixed effect logistic regression with binary distribution with replicate as a random factor and weighted for group size to calculate the odds ratio (glmer model of the lme4 package, R i386, 4.0.5).Day 7 and day 8 blastocysts were treated as repeated measurements and therefore results for blastocyst development are presented as one outcome (glmmSQL model of the MASS package, R i386, 4.0.5).Odds ratio (OR) <1 indicate a negative effect of treatment.The effects of treatment on ordinal variables (stage, grade) were analysed using cumulative link mixed-effect models with multinomial distribution (clmm model of the ordinal package, R i386, 4.0.5) with replicate as random factor.This resulted in one model for effect of treatment on stage.Linear mixed effect models (lmer model of the lme4 package, R i386, 4.0.5)were used to calculate the effect of treatment on the continuous variables total lipid volume, average lipid droplet size, and lipid volume/cell.Replicate (random factor) and treatment (fixed factor) were included in the models.Due to the high energy-demand in the transition from morula to blastocyst-stage, the lipid content is decreased in blastocysts (Sudano et al., 2016).Therefore, it is expected that lipid content is different in different blastocyst-stages.To account for this difference, stage (fixed factor) and blastomere count (fixed factor) and their interactions were considered for inclusion in the models for lipid distribution.Stage and blastomere count were highly correlated, and therefore only stage was included in the final models based on best model fit.The interactions were removed based on p-values and model AIC.Blastomere count provided an objective and blinded measurement of blastocyst size (and indirectly stage).The results from this model are therefore also presented in addition to the model of blastocyst stage.The effect of treatment on blastomere count in blastocysts at day 8 blastocysts was calculated using linear mixed effect models with replicate as a random factor.For degraded embryos or embryos not reaching blastocyst stage, square root transformed values were used to obtain normal distribution.Results are presented as mean ± SEM if not stated otherwise and p < .05were considered significant.

Gene-expression and DNA methylation analyses 2.7.1. Parallel gDNA and total RNA extraction
Blastocysts from day 8 pf (n = 120) were individually snap-frozen in liquid nitrogen in a fixed volume of PBS-PVA.The embryos were then stored in − 80 • C until further analysis.Pools of 10 embryos of equivalent stage and grades (in order to avoid confounding effects due to different embryo characteristics) were analysed in four replicates for both controls and treated groups (see SI Table S8 for details on groups).Parallel genomic (g)DNA and total RNA extraction was conducted using the AllPrep DNA/RNA micro kit (Qiagen Cat no 80284) according to the manufacturer's instructions with slight modifications (Page-Lariviere et al., 2017).Washed DNA was eluted in elution buffer, evaluated using NanoDrop (ND-1000, NanoDrop Technologies, Wilmington, DE, USA) and stored at − 80 • C for further use.RNA was eluted in 15 μL RNAse-free water and purity, quality and concentration were evaluated using the Agilent Bioanalyzer 2100 (Agilent technologies Inc., Santa Clara, CA, USA) where all samples had an RNA integrity number ≥8.9 (range 8.9-9.7)which was considered as good quality.RNA was stored at − 80 • C.

RNA amplification and microarray hybridisation
RNA was amplified using the 2-round RiboAmp® HS PLUS RNA Amplification kit (Applied Biosystems, Foster City, California, USA), which provides amplification of small amounts of RNA and yields sufficient material (antisense (a)RNA) for microarray hybridisation using a T7 RNA polymerase amplification approach (T7-IVT).Amplified aRNA was quantified (Nano-Drop ND-1000, NanoDrop Technologies, Wilmington, DE, USA) and labelled with Cy3 and Cy5 using the Universal Linkage System (ULS™) Fluorescent Labelling Kit (Leica, Wetzlar, Germany).Exposed groups (PFOS-53 and PFOS-02) were compared to the control using a four-array dye-swap design.For a duration of 17 h at 65 • C, 825 ng aRNA per replicate was hybridised to an Agilentmanufactured EmbryoGENE slide in 10× blocking agent and 25× fragmentation buffer (Agilent).After hybridisation, the slide was washed (0.005 % Triton X-102 repeated twice followed by acetonitrile and drying and stabilising solution, provided by Agilent) and scanned using a PowerScanner (Tecan Group Ltd., Mannedorf, Switzerland).Feature extraction was performed with Array-pro 6.3 (Media Cybernetics, Rockville, Maryland, USA).

Analysis of gene expression profile
In the microarray, 42,242 probes were covered including 21,139 reference genes, 9,322 probes for novel transcribed regions, 3,677 alternatively spliced exons, 3,353 3 ′ -tiling probes and 3,723 control probes (Robert et al., 2011).Relative transcript abundance was analysed using Flexarray (Blazejczyk et al., 2007).Background was subtracted from the raw fluorescence intensity.Normalisation within the fluorophores (Cy3/Cy5) and between each array was conducted using Loess and quantile respectively.Comparison between treatments (exposed groups and control) was performed with the limma algorithm.Each probe was attributed a probability to fold-change between treatment and control and differentially expressed genes (DEGs) defined as significant (p < .05)probes with fold change > 1.5.Applying fold change or false-discovery rate (FDR) by Benjamini Hochberg correction did not generate enough transcripts for pathway analysis.All significant probes (p < .05)were therefore analysed by Ingenuity Pathway analysis (QIAGEN Ingenuity Pathway Analysis (IPA) software).In order to obtain a general overview of pathways affected on transcriptomic level, all significant (p<0.05)transcripts were included.The IPA analysis consider the amount of transcripts consistent with the effects and adjust its p-value accordingly.In this way, if several transcripts are linked to the same pathway in the same direction, this represent more important information at the cellular level and the analysis gives higher significance/importance to it.

gDNA fragmentation, amplification and hybridisation
gDNA from the same pools used for transcriptome microarray was amplified and enzyme fragmented before hybridisation for Agilent arrays as described previously (Saadi et al., 2014).The gDNA was fragmented using MseI enzyme (NEB-R0525S) to enable standardised fragmentation between experiments regardless of methylation status.gDNA was then treated with methylation sensitive enzymes (AciI (NEB-R0551S), Hinp1I (NEB-R0124S), HpaII (NEB-R0171 L)) to enable fragmentation according to methylation status, whereby methylated regions are protected from cleavage.After enzymatic digest, samples were amplified by ligation-mediated PCR resulting in exponential amplification of the methylated regions.Labelled (Cy3/Cy5) samples of exposed groups (PFOS-02 and PFOS-53) were compared to the control using a four-array dye-swap design.Hybridisation was performed according to the manufacturer of the microarrays instruction (Agilent).Labelled samples were hybridised in 100× Blocking agent and 2× hybridisation buffer (Agilent) and for 40 h at 65 • C. Hybridised samples were then washed according to the manufacturer's instructions and scanned with the PowerScanner (Tecan) and features were analysed using Array-Pro Analyser 6.3 Software (Media Cybernetics).

Quantification of DNA methylation patterns
The bovine EmbryoGENE DNA Methylation Array (EDMA) was used to analyse the DNA methylation patterns of the embryos as previously described (Montera et al., 2013;Saadi et al., 2014).The microarray containing methylation sites within the bovine genome without topological bias contained a total of 414,566 probes, surveying 20,355 genes and 34,379 CpG islands.Controls accounted for 2.5% of the total probe number and represented Agilent proprietary controls, genomic cleavage controls and EDMA spiked-in controls (Saadi et al., 2014).Quantile inter-array scale normalisation followed loess intra-array normalisation within the limma-package of a pipeline described previously to obtain Bayesian statistics of different methylations, including controls and validations (Saadi et al., 2014).For comparative analyses, differentially methylated regions (DMRs) were defined when p < .05 and fold change > 1.5.Functional analysis was performed by IPA on significantly different probes (p < .05)without applying fold change or FDR adjustment to analyse altered pathways after PFOS exposure (see Section 2.7.3).

Oocyte developmental competence until blastocyst stage after exposure to PFOS during IVM
Exposure to 53 ng mL -1 PFOS during oocyte maturation in vitro significantly decreased the chance of the fertilised embryo reaching cleaved stage at 44 h pf, which was also observed as a reduced chance of cleaving beyond two-cell stage (Fig. 2, exact proportions, OR and pvalues shown in Table 1).This was not observed in the PFOS-02 group.The initial delay in development in the PFOS-53 group could not be seen in later stages of early embryonic development.The proportion of embryos reaching blastocysts stage at day 7 or 8 did not differ from the control group (Fig. 2, Table 1).Phenotypic variation in blastocysts upon exposure to PFOS during IVM

Blastocyst stage and cell count
PFOS exposure led to impaired blastocyst development, whereby a smaller proportion of embryos reached more advanced stages at day 8 compared to the control (Fig. 3A).The reduction in stage was significant in PFOS-53 (p=.01) but not in PFOS-02 (p=.06).This corresponded to a reduced number of blastomeres in PFOS-53 (mean 70.1 (SEM 4.7) p = .04,Fig. 3B) compared to the control (84.8 (5.0)) in the subset of embryos stained for confocal imaging.There was no difference in blastomere count in the PFOS-02 group (79.5 (5.5) p=.70) compared to the control.The embryos that did not develop into blastocyst stage or were degraded at day 8 pf were also assessed for blastomere count.There was no difference in how far the development proceeded (mean nuclei in control 7.8 (SEM 0.4), PFOS-02 7.9 (0.4), p=.63, PFOS-53 8.0 (0.4), p = .56).We can observe that the oocytes not developing into blastocyst stage halt development at an early stage, before embryo genome activation (8-16 cell stage in bovines, (Barnes and First, 1991)).
The quality grade of the embryos, scored according to the IETS system, was not altered by PFOS treatment (PFOS-53 p = .20,PFOS-02 p = .69).

Lipid distribution in blastocysts
To evaluate lipid storage and metabolic status of the blastocysts, lipid distribution (average lipid droplet size, total lipid volume, lipid volume/cell) was measured in the blastocysts.Blastocyst size were added in the models to account for differences in lipid content depending on developmental stage (see Section 2.6 for detail).Blastocysts that developed from oocytes exposed to PFOS during in vitro maturation showed an increased lipid volume/cell in PFOS-53 (control: mean 607.6 (SEM 44.3) μm 3 , PFOS-53: 853.0 (91.4) μm 3 , p < .0001)as well as an increased total volume of lipids in the embryo (control 53371 (5678) μm 3 , PFOS-53 61543 (9251) μm 3 , p=.0003) (Fig. 4).This was not seen in the PFOS-02 group (lipid volume/ cell 621.1 (57.3) μm 3 , p=.62 and total volume of lipids in the embryo 50175 (6212) μm 3 , p=.65).

Gene expression changes in blastocysts upon oocyte exposure to PFOS during IVM
To determine the changes in gene expression after PFOS exposure during IVM, transcriptomic analyses using microarrays on pools of day eight blastocysts of equivalent stage and grade (to avoid confounding effects from different embryo characteristics at different stages of development) were performed.Following Benjamini-Hochberg adjustment of p-values, for multiple comparisons, no differentially expressed genes (DEGs) were identified.When applying fold change threshold to non-adjusted p-values, 16 DEGs (fold change > 1.5, p < .05)were identified in PFOS-02, of which 12 were downregulated and four upregulated (Fig. 5A).In PFOS-53, 223 DEGs (fold change > 1.5, p < .05)were found, with the majority being downregulated (218) and five genes upregulated (Fig. 5B).Five genes overlapped between the treatments (Fig. 5C).The transcripts differentially expressed in both treatment were ubiquitin D (UBD), which was upregulated and Histone H2B type 1-N (HIST1H2BN), sperm-adhesion molecule-1 (SPAM1), transcription factor 15 (TCF15) and charged multivesicular body protein 4B (CHMP4B), all downregulated (Fig. 5D).

Functional analysis of gene expression changes
In a functional analysis of gene expression changes in PFOS treated groups compared to control, DEGs with > 1.5 fold change in expression and p < .05(16 for PFOS-02 and 223 for PFOS-53) showed no enrichment for specific biological functions using Ingenuity pathway (IPA) analysis.Instead, in order to obtain a general overview of pathways in the blastocysts associated with oocyte exposure, we considered also small changes in expression if transcripts belonged to the same biological pathways.Using this larger set of genes (n=668 in PFOS-53, n=300 in PFOS-02), IPA analysis revealed the top biological functions altered after PFOS treatment at both doses to be related to proliferation, cell death and survival, cell differentiation, export and transport of RNA or proteins, and initial expression or transcription of RNA.There was a higher number of transcripts included in each biological function for PFOS-53 (668) compared to PFOS-02 (300).This resulted in an overall stronger prediction of altered biological functions in the higher concentration (Fig. 6).
Regulatory pathways upstream of the significant transcripts (p < .05,no fold change criteria) were identified by IPA.This analysis revealed pathways differently regulated in PFOS-treated groups, compared to the control group, which were related to apoptosis, cell stress and metabolism.Alterations were more pronounced in the PFOS-53 group compared to the PFOS-02 group (full table of activated and inhibited upstream regulators can be found in SI, Tables S9 and 10).More specifically, both treatments were predicted to alter biological functions related to cell death and survival which occurred most significantly through P38 MAPK (Fig. 6 CTNNB1 and RELA inhibition was as also predicted to be inhibited after exposure to PFOS-53.These pathways, as well as STAT3, are pathways with functions that include embryo development and blastocyst formation.

Assessment of DNA methylation patterns
In order to address if PFOS treatment affected DNA methylation, and whether this could be linked to transcriptional changes, the same groups of pooled day eight blastocyst as used in the transcriptomic microarray were analysed using the Agilent EDMA microarray.When applying a Benjamini Hochberg adjustment of p-values, no DMRs were identified.Using unadjusted p-values, 189 DMRs were identified in the PFOS-53 group (fold change >1.5, p <.05), of which 69 were hypermethylated and 120 hypomethylated.In the PFOS-02 group, 380 DMRs were identified, of which 81 were hypermethylated and 299 hypomethylated (Fig. 7).In PFOS-02, one probe was identified as both DMR and DEG; uromodulin like 1 (UMODL1).For PFOS-53, there were four probes identified as both DEGs and DMRs, namely neurexin 2 (NRXN2), Interleukin-2 receptor alpha (IL2RA), zinc finger protein 532 (ZNF532) and methyl CpG binding protein 2 (MECP2).There were 83 and 193 significant (p < 0.05) probes overlapping between the transcriptomic and DNAmethylation microarray for PFOS-02 and PFOS-53, respectively (Fig. 8).
With the same data set, regulatory pathways upstream of the significant DMRs were identified by IPA.This analysis revealed pathways differently regulated in PFOS-treated groups compared to the control group that were similar between to the ones identified with the transcriptomic data set.Full data on upstream regulators of PFOS-02 and PFOS-53 can be found in Tables S11 and 12 -53, p=1.37e-08) were also persistent following PFOS-exposure and present in both methylation and transcriptomic data sets.For PFOS-53 treatment, 59 genes belonging to these 3 pathways were overlapping between the transcriptomic and DNA methylation data set (Table 2).

Exposure relevant to humans
In the present study, we show how exposure to human relevant concentrations of PFOS during a brief window of exposure during oocyte final maturation in vitro (IVM) alters early embryo development in a bovine model.Concentrations in maturation media were validated by mass-spectrometry (53 ng g -1 and 2 ng g -1 ) and were lower compared to what has commonly been used in experimental in vitro settings.The concentrations reached were comparable to levels measured in follicular fluid in IVF cohorts (average range 2-7.5 ng mL -1 , range 0.6-30.4,(Heffernan et al., 2018;McCoy et al., 2017;Petro et al., 2014)), and the exposure in the in vitro setting thus simulated the exposure in vivo where the follicular fluid surrounded the oocyte during maturation in the mammalian follicle.Even though concentrations of PFOS in follicular fluid are lower compared to serum, the transfer efficiency to follicular fluid is close to one (0.64,) suggesting that serum concentrations in the general public can also be used as an estimate to oocyte exposure.PFOS-02 exposure was generally lower than the average exposure measured in human serum and PFOS-53 approximately ×1-10 higher than levels measured in the general public (Calafat et al., 2007;Kärrman et al., 2007;Mamsen et al., 2017;Velez et al., 2015;Wikström et al., 2020).In addition, local contamination of ground water depots used as drinking water sources may have led to exposures that clearly exceeded the levels measured in the general public.For example, the average serum level in the heavily exposed Ronneby community in Sweden is 245 ng mL -1 (ranging up to 1870 ng mL -1 , (Li et al., 2018)).The range of PFOS to which humans are exposed differs between populations and both PFOS-02 and PFOS-53 can be considered human relevant concentrations.Furthermore, it is important to realize that the window of exposure in our study (22 h) was short compared to the in vivo situation where follicles grow for months.In addition, general populations are continuously exposed to complex mixture of compounds instead of single defined chemicals.
The most important route of exposure to humans is dietary intake primarily through drinking water and seafood (Poothong et al., 2020;Sunderland et al., 2019).Even though contamination of vegetables and plants does occur, it is not considered the main exposure (Ghisi et al., 2019).Hence, the background exposure of cattle (and the oocytes used in this study) is likely to be low under Swedish conditions.This is supported by a study in which PFOS concentrations in blood from cattle were considerably lower (on average 0.110 ng g -1 ) compared to humans and the concentrations used in this experiment (Vestergren et al., 2013).
In addition, the design of the experiments in this study counteracts any differences in exposure or physiology between individual cattle or herds of cattle by random distribution of oocytes between treated groups and control.

Altered embryo development after PFOS exposure
Following PFOS-02 exposure, there were no statistically significant negative effects on the proportion of embryos developing or on the quality or stage of the embryo using morphological assessments.However, after exposure to PFOS-53, we observed a decreased likelihood of cleaving to and beyond the two-cell stage 44 h after fertilisation.This suggests a direct effect of PFOS on the oocyte/cumulus cells that, in turn, affects the timing of the first cleavage divisions.Similarly, PFOS-53 embryos were delayed in the development into more advanced stages of blastocysts by day 8 (seen in PFOS-53 as both reduced stage and blastomere count).However, there was no reduction in the proportion of embryos developed compared to control, which indicate that the observed changes are more likely to be attributed to decreased developmental rate rather than early embryo lethality in this system.
Changed, and especially delayed development could have negative consequences for further development.Timing of events during early embryo development is important in bovines as well as in humans, and follows a strict and predictable timeline where deviations or delays are negative for further blastocyst development (Wong et al., 2010), and the timing of the first cleavage is predictive of blastocysts quality (Wang et al., 2014).Furthermore, the implantation window in humans is only about 48 h (Aghajanova et al., 2008).For successful implantation, the embryo and the endometrium must partake in a delicate communication where embryo as well as endometrial factors are of high importance (Diedrich et al., 2007) and delayed development might hamper this communication.Exposure to PFOS-53 was associated with inhibited CTNNB1 signalling.This pathways seem to be important for blastocyst formation as Ctnnb1 knockout mice embryos have abnormalities in shape and size (Messerschmidt et al., 2016).The exact role of CTNNB1 in early embryo development is not fully understood but is linked to its involvement in WNT-signalling, which is crucial to a number of developmental processes (Tribulo et al., 2017).Furthermore, alteration of the RELA pathway after PFOS-53 exposure observed in this study might also play a role in the mechanisms of delayed/inhibited blastocyst formation.RELA is a subunit of the nuclear factor kappa B (NF-κB).NF-κB activation is required for the development of mouse embryos beyond the two-cell stage (Nishikimi et al., 1999).NFKB1A (inhibitor of NF-κB) transcript has been shown to decrease during oocyte IVM, but later increase as a consequence of embryo genome activation in the bovine model (Paciolla et al., 2011).There were further signs of impaired blastocyst formation and possibly also implantation by changed transcripts associated with PTEN signalling in blastocysts, which seem to be one of the important players for blastocyst formation and trophoblast invasion (Xue et al., 2020;Zhang et al., 2015).However, the process of implantation is complex and in many ways species-specific so to fully understand whether and how PFOS exposure during oocyte maturation could possibly affect the ability of embryos to implant, further experiments with an experimental set-up to investigate the effects on implantation are necessary.
During early development, mammalian embryos rely on pyruvate, fatty acids and amino acids as sources of energy.In bovines, the storage of lipids in droplets increases at morula-stage and decreases at blastocyst stage (Sudano et al., 2016) when the demand for lipids as an energy supply is high.Embryos respond to stress from environmental insults in different ways (see review by (Cagnone and Sirard, 2016)), and metabolic effects of stress include lipid accumulation (Abe et al., 2002;Romek et al., 2009).The increased lipid accumulation upon PFOS treatment observed in our study is supported by previous experiments in zebrafish embryos and larvae (Sant et al., 2021;Yi et al., 2019) and also in the amphibian Daphnia magna (Seyoum et al., 2020).The lipid accumulation after PFOS exposure in the livers of zebrafish fry has also been seen to be persistent after a period of recovery (Du et al., 2009) which stresses the fact that altered lipid metabolism in embryos might persist during further development.The altered lipid distribution during blastocyst stage is also consistent with findings from a previous study from our laboratory showing larger lipid droplets in bovine embryos  after exposure during IVM to sub-lethal concentrations of perfluorononanoic acid (PFNA) (Hallberg et al., 2019).Mechanisms underlying this altered lipid metabolism based on our molecular studies could involve e.g., altered NONO, APOE and JUN pathways.Caveolin 1 (CAV1), which is a key player in the JUN pathway, plays an important role in lipid droplet maturation and biogenesis through cholesterol balance and lipid transport (Pol et al., 2001), and in lipid droplet size regulation in adipocytes (Blouin et al., 2010).It was recently shown that CAV1 expression in bovine blastocysts is lower compared to porcine, and the difference was attributed to different lipid metabolisms between these species (Kajdasz et al., 2020).Furthermore, Acetyl-CoA carboxylase (ACC) coded by ACACA and was associated with JUN signalling upon PFOS exposure.This is the rate-limiting step in fatty acid synthesis and is increased in bovine embryos with higher oxidative stress and altered metabolism (Dos Santos et al., 2019).An increase in the other isoform of ACC, acacab, which mainly regulates fatty acid oxidation, and acox1 gene expression has previously been demonstrated after developmental exposure of PFOS in zebrafish (Yi et al., 2019).The altered pathways in lipid metabolism imply that the surviving, morphologically normal embryos suppress lipid synthesis, possibly as an adaptive mechanism counteracting the PFOS induced lipid accumulation.The preimplantation embryo, although having limited capacity to respond to environmental insults, may alter transcriptome in order to survive (Puscheck et al., 2015).PFOS has a known potential to activate the peroxisome-proliferator activated receptors (PPARs), (Behr et al., 2020;Khazaee et al., 2021;Shipley et al., 2004;Takacs and Abbott, 2007), which affect the lipid/glucose-metabolism, amongst other functions, leading to increased liver weight as a hallmark of PFOS toxicity in animal models.The mechanism of PFOS toxicity has been attributed partly, but not fully, to the activation of PPARα, and other nuclear receptors have also been shown to participate in PFOS response in both rat and human liver cells (Bjork et al., 2011).Apart from PPARα, PPARɤ and estrogen receptor-α (ERα) activation has been shown in mice (Rosen et al., 2017).Our results of delayed or possibly failed embryo development, as well as the lipid accumulation are similar to results in a study on mouse oocytes after exposure during IVM to a PPARɤ agonist (rosiglitazone, 20 μM).In that study, fatty acid oxidation was altered and, despite increasing the mitochondrial membrane potential, treatment was associated with impaired developmental competence and lower proportion of hatched blastocyst at day 5 (but no reduction in the proportion of blastocysts) (Dunning et al., 2014).
In both PFOS concentrations, pathways associated with apoptosis, cell proliferation and differentiation were altered showing clear signs of embryos coping with the previous insult of PFOS during IVM.This suggest that the decreased blastomere count and delayed developmental rate observed in PFOS-53 could be explained with increased apoptosis and/or decreased/delayed proliferation.This corresponds to previous studies of PFAS toxicity, where increased apoptosis during development has been demonstrated after PFOS exposure in i.e. zebrafish larvae (Shi et al., 2008) and xenopus leavis embryos (San-Segundo et al., 2016).p53 signalling seems to be the key-player in apoptosis after both PFOA (Chen et al., 2017) and PFOS (Shi et al., 2008) exposure.In zebrafish, this PFOS induced apoptosis and p53 activation has been attributed to ROS generation and MAPK signalling (Shi and Zhou, 2010).P38 MAPK is crucial for pre-implantation development, and inhibition has been associated with decreased expansion and hatching in the mouse (Bell and Watson, 2013), as well as increased apoptosis at blastocyst stage together with altered glucose metabolism (Sozen et al., 2015).
To interpret whether changes in gene expressions are more likely to be a transient adaptation to the surrounding environment or have more long-term effect on development, studies on the epigenome are necessary (Messerlian et al., 2017).The epigenome coordinates the responses that are adaptive in nature (Peters et al., 2021) and include histone tail modifications, chromatin remodelling and DNA methylation.DNA methylation is believed to have a major impact on gene silencing.Using a microarray platform developed for the bovine species (Montera et al., 2013), alterations in methylation patterns have been seen after exposure to altered metabolic situations (Laskowski et al., 2018) and environmental insults (Page-Lariviere et al., 2016).Following high DNA methylation during oocyte development (and during PFOS exposure), the epigenome, including the DNA methylome, is reprogrammed during the early stages of development in bovine and humans (Dean et al., 2003).During the period when the DNA methylation was studied (day eight of culture) the bovine genome starts to become remethylated after the active demethylation, and is at this point mostly related to the conservation of specific methylation sites from the parental chromosomes (Wang et al., 2014).During the following week of development, the embryo will continue the remethylation leading to a significantly higher methylation level on day twelve (Montera et al., 2013).The changes in methylation that we observe after exposure during the last stage of oocyte maturation are hence not only present a week later during development, but also conserved through DNA methylation erasure and reprogramming.The higher number of DMRs compared to DEGs can be explained by the fact that 1) the methylation microarray interrogates 10 times more probes compared to transcriptomic array (400,000 vs 40,000) and 2) methylation changes do not necessarily translate into changes in gene expression at this stage.We could observe similar biological functions altered as in the transcriptome profile related to cell death, apoptosis proliferation and differentiation.In fact, 59 of the genes in the in these pathways showed both transcriptional and DNA methylation changes upon PFOS-53 treatment.About half of these showed an inverse relationship between DNA methylation and gene expression change (18 of 33 TP53, 15 of 32 TNF and 8 of 18 CTNNB1) including promoter regions in CpG islands.This is line with the general suppressive function of DNA methylation.However, DNA methylation in gene bodies (Teissandier and Bourc'his, 2017) has been linked to increased gene transcription and even in promoter regions direct correlation between methylation and transcriptional activity can occur (Smith et al., 2020).Likewise, DNA methylation preventing transcriptional repressors to bind can lead to increased gene transcription (Ehrlich and Lacey, 2013).Such functions of DNA methylation might explain the cases where gene transcription and methylation was changed in the same direction by PFOS-53 exposure.Thus, this suggests that changes seen in the gene expression profile might be due to epigenetic alterations, and thus potentially persist during development.However, even though there is convincing evidence of developmental origin of later disease (Sinclair et al., 2007), the epigenetic mechanisms of gene programming during early embryo development are far from fully understood.
The phenotypic changes seen in the blastocysts upon PFOS exposure were modest and there were no evident embryo lethality at these exposure concentrations.This is likely reflected by the transcriptomic and the DNA methylation microarray results.Higher doses inducing larger morphological effects in the blastocyst would most likely have more impact with increased numbers of DEGs and DMRs.However, this would not reflect real-life exposures and thus not be within the scope of this study.

Implications for human health
PFOS exposure has been associated with decreased birth weight in humans, as well as an increased time to pregnancy (Fei et al., 2009;Whitworth et al., 2012).Decreased birth weight has also been demonstrated in rats (Luebker et al., 2005;Xia et al., 2011), although in combination with decreased weight gain in pregnant dams.Even though the processes are complex, delayed growth as early as the time for the first divisions of the zygote is predictive of later blastocyst quality in humans (Wang et al., 2014).Hence, the reduced growth during early embryo development resulting in inhibited or delayed blastocyst formation in day eight blastocyst might provide an insight to the mechanisms of the reduction in intrauterine growth and/or failed or impaired implantation, resulting in reduced fecundity in humans.
There is evidence linking PFOS exposure to increased serum lipids and cholesterol in epidemiological studies, which is a risk factor for cardiovascular disease (Nelson et al., 2010;Steenland et al., 2009).Altered glucose and lipid metabolism has been seen in the adult offspring after exposure to PFOS in utero in experimental studies in mice (Lee et al., 2015;Wan et al., 2014), and PFOA has been suggested to have an obesogenic effect on offspring in human cohorts after exposure in utero (Halldorsson et al., 2012).The evidence for a causal association of in utero exposure to PFOS with metabolic effects such as diabetes and obesity is insufficient.However, our results of altered lipid metabolism during pre-implantation embryo development, together with previous in vivo studies showing disruption in fatty acid homeostasis in livers after oral exposure in rats (Curran et al., 2008) and altered glucose metabolism after in utero exposure in mice (Wan et al., 2014), provide further insight into PFOS toxicity influencing lipid metabolism of human relevance.

Conclusion
To conclude, we show that human relevant concentrations of PFOS alter early pre-implantation embryo development after short exposure during final oocyte maturation in a bovine model.Exposure to PFOS led to delayed development to more advanced blastocyst stages at concentrations of 53 ng mL − 1 and transcriptomic alterations related to apoptosis, differentiation and proliferation, which was corroborated by epigenomic data one week after exposure.Similar effects were observed in the transcriptome and DNA methylation in blastocysts using 2 ng mL − 1 exposure, suggesting that adverse effects might occur even at general population exposure levels.Effects on the epigenome suggest that if the embryo survives the PFOS insult, the effects may persist during further embryo and foetal development.This study emphasises the urgent need for remediation measures, adds to the evidence that PFOS exposure affects fertility and reproduction and may be useful for human health risk assessment.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 2 .
Fig. 2. Developmental competence of bovine oocytes after PFOS exposure during in vitro maturation.Boxplots showing the proportion of developed embryos in control and after treatment of oocytes with 2 ng mL -1 or 53 ng mL -1 PFOS (PFOS-02 and PFOS-53, respectively).Proportion of cleaved embryos and cleaved embryos beyond two-cell stage 44 h post fertilisation (pf), and proportion of blastocysts at day 7 and day 8 pf calculated from cultured oocytes are shown.The data are presented as boxplots where line represent median, the box is the interquartile range (IQR), and whiskers are 1.5 × IQR, outliers depicted with dots and represent values > 1.5 × IQR.Non-significant p-values annotated with ns.

Fig. 4 .
Fig. 4. Variation in lipid distribution after PFOS exposure.Boxplots showing lipid distribution in the embryos after PFOS exposure during oocyte maturation (average lipid droplet volume, lipid volume/cell and total volume lipids, um 3 ), p-values < .05are shown and values not significant are indicated by n.s.The data are presented as boxplots where the line represents the median and bullet the mean, the box is the interquartile range (IQR) and the whiskers are 1.5 × IQR.Outliers are depicted with dots and represent values > 1.5 × IQR.Confocal imaging of day eight blastocysts stained for neutral lipids (LipidTOX, green) and nuclei (Hoechst, blue) showing representative embryos from the control and the PFOS treated groups.Note larger lipid droplets in embryos from control (arrow) compared to PFOS-02 group (arrowhead).Note also the increased amount of total lipids in PFOS-53 compared to the control.

Fig. 6 .
Fig. 6.Enrichment of top five molecular biofunctions altered after both PFOS exposures (PFOS-53 and PFOS-02) compared to control, derived from IPA analysis of all significant differentially expressed transcripts regardless off fold change and ranked by p-values (lowest to highest) in PFOS-53.-log(p-values) of the top functions are presented.The number to the right of each bar represents the amount of transcripts predicting the function.

Fig. 7 .
Fig. 7. Downstream signalling after upstream inhibition of P38 MAPK.Predicted inhibition of P38 MAPK signalling and transcripts associated with the prediction.Upregulated transcripts are in red and downregulated transcripts are in green.Arrows showing relationships where blue lead to inhibition, red to activation, yellow indicates findings inconsistent with state of downstream molecule and grey indicates when effect is not predicted.

Fig. 8 .
Fig. 8. Volcano plot of differentially methylated regions compared to the control in PFOS-02 and PFOS-53 respectively in the Agilent EDMA microarray.The y-axis represents -log10(expr p-value) and the x-axis log2(fold change) calculated based on different methylation assessed from methylation-sensitive enzymes.

Table 1
Developmental competence of oocytes after exposure to PFOS during maturation in vitro.
a Two ng mL -1 PFOS during in vitro maturation.b 53 ng mL -1 PFOS during in vitro maturation.c Calculated from the number of cultured oocytes at the start of the experiment.d Odds ratio and confidence interval (CI) calculated by logistic regression.e Cleaved and cleaved beyond two-cell stage assessed 44 h after fertilisation.f Day 7 and 8 blastocyst rates were considered repeated measurements resulting in one outcome from the model.

Table 2
Genes showing significant (p < 0.05) changes in both expression and DNA methylation in common pathways altered by PFOS-53.