Perfluoroalkyl Acids in Follicular Fluid and Embryo Quality during IVF: A Prospective IVF Cohort in China

Background: Perfluoroalkyl acids (PFAA) have been measured in ovarian follicular fluid from women using in vitro fertilization (IVF), although associations between follicular fluid PFAA and IVF outcomes have been inconsistent. Objectives: We investigated the association between follicular fluid PFAA and embryo quality in women undergoing IVF. Methods: We prospectively enrolled 729 women undergoing IVF treatment in Guangxi province, China, from July 2018 to December 2018. We measured 32 PFAA, including branched isomers, in follicular fluid using ultra-performance liquid chromatography coupled to tandem mass spectrometry. We applied restricted cubic splines, linear regression, and log-binominal regression models to investigate associations between follicular fluid PFAA and embryo quality, adjusting for confounding variables and investigated oocyte maturity as an intervening variable using causal mediation analysis. We further estimated the overall effect of the PFAA mixture on outcomes using Bayesian kernel machine regression (BKMR). Results: We detected 8 of 32 measured PFAA in >85% of follicular fluid samples. Higher PFAA concentrations were associated with fewer high-quality embryos from IVF. The high-quality embryo rates at the 50th percentile of linear perfluoro-1-octanesulfonate acid (n-PFOS), all branched PFOS isomers (Br-PFOS) and linear perfluoro-n-octanoic acid (n-PFOA) were −6.34% [95% confidence interval (CI): −9.45, −3.32%], −16.78% (95% CI: −21.98, −11.58%) and −8.66% (95% CI: −11.88, −5.43%) lower, respectively, than the high quality embryo rates at the reference 10th percentile of PFAA. Oocyte maturity mediated 11.76% (95% CI: 3.18, 31.80%) and 14.28% (95% CI: 2.95, 31.27%) of the n-PFOS and n-PFOA associations, respectively. The results of the BKMR models showed a negative association between the PFAA mixture and the probability of high-quality embryos, with branched PFOS isomers having posterior inclusion probabilities of 1 and accounting for the majority of the association. Discussion: Exposure to higher PFAA concentrations in follicular fluid was associated with poorer embryo quality during IVF. Branched PFOS isomers may have a stronger effect than linear PFOS isomers. More studies are needed to confirm these findings and to directly estimate the effects on pregnancy and live-birth outcomes. https://doi.org/10.1289/EHP10857


Introduction
Infertility, defined as 12 months of unprotected heterosexual intercourse without a successful pregnancy according to the International Committee for Monitoring Assisted Reproductive Technology, 1 affects approximately 48:5 million reproductive-age couples worldwide. 2 More advanced reproductive-age, 3 unhealthy lifestyle factors (e.g., irregular sleeping mode, etc.), and psychosocial stress 4 are likely to contribute to lower fecundity. However, growing evidence also suggests that exposure to environmental endocrine-disrupting chemicals (EDCs) may affect human reproductive health by altering the hormonal milieu that plays an important role in the pathogenesis of infertility. 5 Perfluoroalkyl acids (PFAA) are EDCs and are a large family of synthesized chemicals, which have been applied in myriad industries since the 1940s, such as in the manufacture of nonstick pans, textile coatings, aqueous film-forming firefighting foams, and food packaging. 6 Dietary exposure, drinking water contamination, and inhalation are the dominant PFAA exposure routes in the general human population. 7 Perfluorooctane sulfonate (PFOS) and perfluorooctanoic acid (PFOA), recognized as persistent organic pollutants and human health risks by the Stockholm Convention [United Nations Environment Program (UNEP)] (http://chm.pops. int/TheConvention/ThePOPs/TheNewPOPs), are the most widely studied PFAA to date, although PFOS and PFOA production has been restricted or banned in most Western nations, leading to declining levels in human blood. 8 However, China continues to produce PFOS, PFOA, and other PFAA, so exposure to PFAA in Chinese populations is likely to grow. 9 Experimental studies have demonstrated that PFAA have the potential to impair female fertility by altering hypothalamicpituitary-ovarian regulation (as reviewed by Ding et al. 10 ). However, the findings of observational studies of PFAA exposure and female fecundability, the probability of conceiving a pregnancy in any "at risk" menstrual cycle, have been inconsistent (as reviewed by Bach et al. 11 and Negri et al. 12 ). For example, in the Danish National Birth Cohort, Fei et al. found strong associations between greater maternal plasma PFOS and PFOA concentrations and longer waiting time to pregnancy, a measure of fecundability, 13 although concerns were expressed about the impact of parity on the results. 14,15 In contrast, Bach et al. later reported null associations in a smaller subpopulation from the same cohort. 16 EDCs, including PFAA, have been measured in ovarian follicular fluid, which bathes the developing oocyte. 17 Endocrine disruptor-associated changes in follicular hormone activities may lead to diminished oocyte quality, which may further influence embryo development and lead to poorer reproductive outcomes. 18 However, few epidemiological studies have investigated the potential relationship between follicular fluid PFAA and oocyte maturation and embryo quality. 5 Given its highly invasive nature, follicular fluid is generally obtained from women using in vitro fertilization (IVF) treatment for infertility, so previous studies have been limited by small numbers of participants. [19][20][21][22][23][24] To further clarify the potential impact of follicular fluid PFAA on female fertility, we conducted a hypothesis-generating study of associations between follicular fluid PFAA and IVF outcomes in a large IVF cohort from China.

Study Participants and Data Collection
Participants were enrolled into the Guangxi In Vitro Fertilization and the Environment Study (GIVES), a prospective investigation of environmental factors and IVF outcomes among couples from an infertility center in Guangxi province, China. Guangxi province is the centralized region for the Zhuang ethnic minority located in south China. The Zhuang population accounts for approximately 36% of Guangxi province residents, with 6.5% from other ethnic minority groups, such as Yao, Miao, Dong, and other ethnicities, and 57.5% are of the Chinese ethnic Han majority. 25 Heterosexual couples initiating a first IVF cycle at the Reproductive Medicine Center, the People's Hospital of Guangxi Zhuang Autonomous Region, were enrolled into GIVES from July 2018 to December 2018 (n = 736). Of the enrolled subjects, no exclusions were applied. Participants completed a comprehensive self-administered study questionnaire, and women agreed to provide follicular fluid samples for analysis.
All participants provided informed consent prior to enrollment. This study was approved by the Ethics Committee Board of the People's Hospital of Guangxi Zhuang Autonomous Region (No. 2018-42).

Clinical Protocols and Outcomes
Initial infertility diagnoses were evaluated by a physician according to Society for Assisted Reproductive Technology (SART) definitions. 26 Based on the infertility evaluation and other clinical history, including baseline sex-hormone measurements, women underwent one of three ovarian controlled hyperstimulation protocols: a) long gonadotrophin releasing hormone (GnRH) agonist protocol (start at luteal or follicular phase); b) GnRH-antagonist protocol; or c) other protocols (including short GnRH agonist protocol, mild stimulation with addition of clomiphene citrate (CC)/letrozole to gonadotropins, modified natural cycle, and progestin for luteinizing hormone (LH) peak suppression). Serum estradiol (E 2 ) was assayed, and follicle size was monitored throughout ovarian stimulation using transvaginal ultrasound. When the two largest follicles reached 18 mm diameter or ≥3 follicles reached 17 mm, ovulation was triggered by the administration of 0:5 mL 250 lg OVIDREL (Recombinant Human Choriogonadotropin alfa Solution for Injection; Merck Serono). Oocytes were retrieved 36 h later by transvaginal needle aspiration under transvaginal ultrasound guidance.
Retrieved oocytes in metaphase-II (MII) arrest (i.e., "mature" oocytes) were fertilized by incubation with fresh sperm using conventional IVF or by intracytoplasmic sperm injection (ICSI) for couples with severe male factor infertility. Rescue ICSI was performed if fertilization failure occurred at early cumulus cell removal after insemination of 4-6 h. Fertilization was confirmed 16-18 h after insemination by the appearance of an oocyte with two pronuclei (i.e., "2PN"). Embryo quality was evaluated by an embryologist according to morphology and the number of blastomeres on the second and third days post fertilization based on the SART guidelines. 26 A high-quality embryo, suitable for transfer, was defined as an embryo with: a) no multinucleated blastomeres, b) 7-9 blastomeres on day 3, and c) <20% anucleated fragments (Alpha Scientists in Reproductive Medicine ESHRE Special Interest Group of Embryology, 2011). 27 All other embryos were classified as low quality. Oocyte maturity (MII rate) was calculated as the number of MII oocytes divided by the total number of retrieved oocytes. High-quality embryo rate was calculated as the number of high-quality embryos divided by the number of 2PN zygotes per woman. The presence of high-quality embryos was defined as >1 high-quality embryo available for transfer per oocyte retrieval cycle. Each woman contributed one IVF cycle (all women initiating their first IVF cycle) to the study.

Follicular Fluid Sample Collection and PFAA Measurement
Pooled follicular fluid (approximately 5 mL in total) was aspirated from two to four follicles from an individual woman during oocyte retrieval and collected in a 15 mL Falcon tube. Samples were examined for visual evidence of blood contamination, and only clear fluid was analyzed. 28 The sample was centrifuged at 1,500 × g for 20 min and the supernatant was stored in polypropylene tubes at −80 C until analysis. Follicular fluid specimens were collected from 729 women (follicular fluid was not collected from 7 women).
We analyzed 32 PFAA concentrations in follicular fluid using ultra-performance liquid chromatography coupled with triple quadrupole tandem mass spectrometry (UPLC-MS/MS) with electrospray ionization in a negative mode (Agilent 6495B, Agilent Technologies). All PFAA standards were purchased from Wellington Laboratories (Guelph, Ontario, Canada). The abbreviations for the measured PFAA are listed in Table S1. The method was based on minor modifications to our previously described method. 29 In brief, 0:2 mL follicular fluid was mixed with 2 mL of 0:1 M formic acid, followed by spiking with 0:5 ng of mass labeled PFAA internal standards and extracted using a solid phase extraction cartridge (200 mg=6 cc; Oasis-HLB). The extracts were centrifuged at 12,000 × g for 10 min at 4°C and analyzed using UPLC-MS/MS. Procedural blanks (saline) were prepared and included in each interval of 20 samples to monitor for method contamination. Solvent blanks containing methanol and Milli-Q water (3:7 v/v) were prepared and included after every 12 samples to monitor for background contamination. Duplicate injections and calibration check standards were run after every 20 samples to ensure the precision and accuracy of each run. We also checked for PFAA contamination of the sterile needles used for oocyte retrieval by measuring the concentration of PFAA in saline run through the needles. The method detection limit (MDL) was defined as the mean concentrations of PFAA plus three times the standard deviation (SD) of procedural blanks. 30 Values below the MDL were imputed as MDL=2 prior to analysis. 31 The MDL of each PFAA is listed in Table S2.

Potential Confounding Variables
We identified potential confounding variables in this study a priori based on a direct acyclic graph (DAG; Figure S1), with reference to the previous literature. 11,19 The minimally sufficient adjustment set of variables for estimating unbiased associations of follicular fluid PFAA with embryo quality included: women's ages, socioeconomic status, prepregnancy body mass index (BMI), ovarian stimulation protocol, parity, infertility diagnosis, and seafood consumption. Information on age (years), family income in Chinese Yuan (CNY) (<5,000 CNY/month; 5,000-10,000 CNY/month; >10,000 CNY/month), prepregnancy BMI (kilograms per square meter), ovarian stimulation protocol (GnRH agonist or GnRH antagonist), and female infertility factors (tubal, ovulation disorder, poor ovarian response, advanced reproductive age, or "other") were collected from the medical record and the self-administered study questionnaire. Regular seafood consumption was defined as a positive answer to the question "Eat seafood at least one time per week." Approximately 14.4% participants did not respond to the seafood consumption question (n = 105) and were assigned as a negative answer (no regular seafood consumption) for this question. There were no other missing data. In this study, we categorized the ethnicity into Han majority group and ethnic minority groups; the latter included Zhuang, Yao, Miao, Dong, and other minority ethnicities. This information was collected from participant's medical record.

Statistical Analyses
Descriptive analyses. Continuous data were presented as the mean ± SD for normally distributed data or the median [quartile 1 (Q1), quartile 3, (Q3)]. Continuous PFAA concentrations were log 10 transformed to reduce the influence of outliers and to normalize the skewed distributions. We also estimated pairwise Spearman correlations between PFAA concentrations in follicular fluid (Table S3).
Single PFAA predictor models. We first estimated associations between individual follicular fluid PFAA as predictors of embryo quality with and without adjustment for confounders. Restricted cubic splines (RCS) were applied to check the linearity of associations. We chose models with a 3-knot RCS function, because the Akaike Information Criterion (AIC) value was lowest in comparison with models with 4-knot or 5-knot RCS functions for most PFAA (Table S4). 33 We constructed multivariable linear regression models to estimate the difference (b) and 95% confidence interval (95% CI) for the difference in the high-quality embryo rate associated with a one log-unit increase in follicular fluid PFAA concentrations in crude models without confounder adjustment and in main models adjusted for the aforementioned confounding variables. We also calculated the estimated differences (95% CI) in high-quality embryo rate at the 25th, 50th, and 75th percentile of PFAA concentration in follicular fluid against the reference values (the 10th percentile of PFAA concentration) using RCS model. We next used log-binominal regression models to estimate associations between the presence of ≥1 high-quality embryo (yes/no) and the concentrations of PFAA in follicular fluid by the relative risk (RR) and 95% CI. We applied the COPY method to avoid convergence problems in the log-binomial regression models. 34 We additionally stratified the main analyses by parity (nulliparous women vs. parous women), age (<35 y old vs. >35 y old), and infertility diagnosis (only male factor, only female factor, and both female and male factor) because previous studies suggested that these factors may modify the associations. 5, 35 We further stratified the association by ethnicity (Han population vs. minority population) to explore potential disparities for the large Zhuang minority population residing in Guangxi province.
We identified oocyte maturity (MII rate) as a potential mediator using the DAG ( Figure S1) and subsequently found an association between follicular fluid PFAA and MII rate. Thus, we used a counterfactual-based causal mediation analysis, 36 employing the CAUSALMED procedure in SAS (version 9.4; SAS Institute Inc.), to investigate oocyte maturity (MII rate) as an intervening variable to explain the association between PFAA exposure and embryo quality. We employed 1,000 bootstrap samples to calculate the 95% CIs for the estimates.
Multiple PFAA predictors models. Considering the interrelationships among follicular fluid PFAA concentrations, we next estimated the potential joint effect estimates of a PFAA mixture on embryo quality using Bayesian kernel machine regression (BKMR). 37 The BKMR model allows for assessment of the independent associations of individual PFAA mixture components with an outcome, in addition to the overall combined mixed PFAA exposure. 38,39 In the current study, we entered only follicular fluid PFAA with detection frequencies >85%, as well as branched PFOS isomers, into BKMR models. Briefly, we applied two BKMR models to investigate the exposure-response relationship between follicular PFAA exposure (log transformed) and the high-quality embryo rate: a) estimating the overall effect estimates of an increase in a PFAA mixture on the high-quality embryo rate when fixing individual PFAA at their median levels and b) estimating the dose-response relationship between each individual PFAA and the high-quality embryo rate when fixing the other PFAA at their 25th, 50th, and 75th percentiles. We examined the overall effect of PFAA mixtures on high-quality embryo rate via fitting the BKMR models, which fixed all PFAA mixtures at the same percentile concentration (in the range of 25th to 75th percentile, with fifth percentile increments) and compared with all PFAA when they were fixed at their 50th percentile. To implement a hierarchical variable selection method by Markov chain Monte Carlo algorithm, we categorized PFAA groups by perfluorocarboxylic acid and perfluoroalkyl sulfonates, respectively. We then calculated the group-posterior inclusion probabilities (PIPs) and conditional-PIPs, a measure of variable importance ranging from 0 to 1, to determine the major PFAA contributing to the main effect using 0.5 as a threshold. All models were run at 10,000 iterations using the Markov chain Monte Carlo algorithm and adjusted for the aforementioned confounding variables. 38 All data analyses were conducted using SAS (version 9.4), with the exception of the BKMR model, which was conducted using R software (version 4.0.2; R Development Core Team) with the "bkmr" package. All hypothesis testing was 2-sided. Statistical significance was set at p < 0:05 following correction for multiple hypothesis testing using the false discovery rate (FDR). 40  There were eight PFAA (n-PFOS, n-PFOA, PFDA, PFHxS, PFHpS, PFNA, PFUnDa, and PFTrDA) with detection frequencies >85% in follicular fluid samples ( Table 2; Table S2). Table 2 presents the concentration of PFAA in follicular fluid samples, with the highest median concentration measured for n-PFOS (1:70 ng=mL), followed by n-PFOA (1:09 ng=mL), total branched PFOS (Br-PFOS, 0:47 ng=mL), PFNA (0:39 ng=mL), and PFUnDa (0:30 ng=mL). In general, each one of PFAA was highly and positively correlated with the others (r = 0:141-0:920, p < 0:001; Table S3).

Results
We found a lower high-quality embryo rate in association with greater follicular fluid PFAA exposure. For example, the highquality embryo rates at the 50th percentile of n-PFOS, Br-PFOS, n-PFOA, and PFHxS were −6:34% (95% CI: −9:45, −3:32%), −16:78% (95% CI: −21:98, −11:58%), −8:66% (95% CI: −11:88, −5:43%), and −10:12% (95% CI: −14:52, −5:72%) lower, respectively, than the high-quality embryo rates at the reference 10th percentile of PFAA (Table 3; Figure S2). The negative associations were consistent in both crude and confounderadjusted regression models (Table S5). When we further dichotomized the high embryo quality rate by the presence of ≥1 high-quality embryo as the outcome, we detected associations between higher concentrations of PFAA in follicular fluid and lower likelihoods for the presence of a high-quality embryo using log-binominal regression models adjusted for confounders, particularly     Table 4). We found associations for n-PFOS, PFDA, PFHpS, PFNA, and PFUnDA among nulliparous (Table S6). However, we found no evidence of modification when we stratified the associations by maternal age (Table S7) and ethnicity (Table S8). When we restricted our analyses in participants with only female factor of infertility diagnosis (n = 569), we found results consistent with our main findings of associations between higher follicular fluid PFAA and lower odds of a high-quality embryo (Table S9). We also found that higher concentrations of PFAA, particularly n-PFOS, Br-PFOS, n-PFOA, and PFHxS, were associated with lower MII rate, adjusted for confounding variables (Table S10). The causal mediation analysis model suggested that the MII rate may partially mediate the relationship between the presence of a high-quality embryo and follicular fluid n-PFOS (11.76%; 95% CI: 3.18, 31.80%), n-PFOA (14.28%; 95% CI: 2.95, 31.27%), and PFHxS (8.13%; 95% CI: 1.53, 20.24%) ( Table 5). Figure 1 shows the overall association of the PFAA mixture with the difference in the high-quality embryo rate using confounder-adjusted BKMR models. Consistent with our findings from the single-PFAA predictor models, the overall effect estimates of the PFAA mixture suggested that greater concentrations of follicular fluid PFAA were monotonically associated with a lower high-quality embryo rate ( Figure 1A; Table S11). For example, we found a lower posterior mean estimate of a high-quality embryo rate (−7:25%; 95% CI: −9:56, −4:94%) when the PFAA mixture concentration was fixed at the 75th percentile of the exposure distribution in comparison with being fixed at the 50th percentile (Table S11). We further characterized the contribution of individual follicular fluid PFAA exposure components to the overall PFAA mixture association and found that branched PFOS isomers were significantly associated with a lower high-quality embryo rate while holding all other PFAA at the 25th, 50th, and 75th percentiles ( Figure 1B and Table S12). Specifically, the PIPs ranging from 0 to 1, indicating the importance of each individual PFAA (from unimportant to most important), showed that Br-PFOS was the dominant contributor (cond-PIP = 1) (Table S13).

Discussion
In this large prospective cohort study of infertile couples undergoing IVF treatment, exposure to individual follicular fluid PFAA (n-PFOS, Br-PFOS, n-PFOA, PFHxS, PFDA and PFUnDA) was inversely associated with embryo quality during IVF. Furthermore, associations of n-PFOS, n-PFOA, and PFHxS were partially mediated by the MII-oocyte rate. A mixture of nine follicular fluid PFAA was associated with a lower highquality embryo rate, and branched PFOS isomers appeared to dominate the association. Thus, exposure to PFAA may further lower IVF success rates.

PFAA Levels in Follicular Fluid
A growing body of literature describes the worldwide distribution of PFAA in human blood and other biospecimens, including blood, urine, hair, nails, and semen. 10,11 The few studies that have measured ovarian follicular fluid showed that PFOS and PFOA were the most prevalent PFAA (Table S14). The range of follicular fluid PFAA concentrations in our study was comparable to other studies. For example, the median follicular fluid linear PFOS (1:70 ng=mL) and PFOA (1:09 ng=mL) concentrations in this study were similar to the geometric mean follicular fluid PFOS (1:8 ng=mL) and PFOA (1:9 ng=mL) concentrations in women without polycystic ovary syndrome (PCOS) (n = 59) from a United Kingdom fertility clinic, 21 and the mean PFOS (4:80 ng=mL) and PFOA (2:40 ng=mL) concentrations in women (n = 97) from an Australian IVF clinic. 20 However, the median follicular fluid PFAA concentrations in our study were lower than those reported in women seeking IVF treatment (4:54 ng=mL for PFOS and 3:38 ng=mL for PFOA, n = 28) at the Peking University People's Hospital in China 41 and were also lower than those in women who underwent the IVF treatment in Yantai, China (4:77 ng=mL for n-PFOS and 6:37 ng=mL for n-PFOA, n = 124). 24 The discrepancies in follicular fluid PFAA concentrations may be related to different sources of environmental PFAA exposure (e.g., PFAA-contaminated drinking water exposure vs. exposure via Table 3. Estimated differences in high-quality embryo rate (percentage) at the 25th, 50th, and 75th percentiles of log-PFAA concentrations (nanograms per milliliter) measured in follicular fluid against the 10th percentile (reference concentration). a

PFAAs and Assisted Reproduction
Among studies evaluating the association between PFAAs and female infertility, only five have examined PFAA concentrations in follicular fluid, a direct indicator of exposure in the oocyte's microenvironment. [20][21][22][23][24] Other studies used serum or plasma PFAA concentrations as an exposure indicator. Governini et al. found a correlation between higher follicular fluid PFAAs with lower fertilization rate in a pilot study (n = 16). 23 However, the authors did not provide detailed information of PFAA levels in the samples. In a U.S.-based study (n = 36), McCoy et al. detected negative relationships of PFDA and PFUnDA concentrations with blastocyst conversion rate (the number of day-6 blastocysts per cultured day-3 embryo) although no statistically significant associations with measures of ovarian response to gonadotropin simulation during IVF were found. 19 In our study, we also detected statistically significant associations between higher follicular fluid PFDA and PFUnDA concentrations and poorer fertilization outcomes, measured as a lower high-quality embryo rate. A more recent study showed that both maternal and paternal plasma PFOA concentrations were negatively associated with IVF outcomes. 43 In a Belgian study (n = 38), Petro et al. observed contradictory findings using a principal component analysis in which overall follicular fluid PFAAs were associated with an unexpected higher fertilization rate after adjusting for covariates and other EDCs. 22 No relationship between eight different follicular fluid PFAAs and fertilization rate was observed among Australian women. 20 Hong et al. also reported no association between PFAA concentrations in follicular fluid and IVF clinical outcomes in China (n = 124). 24 All of the above published studies were based on small sample sizes. Ours is the largest study to date to assess associations of follicular fluid PFAAs with fertility outcomes. In addition, our study comprehensively profiled follicular fluid PFAA concentrations including linear and branched PFOS in follicular fluid, which provided new evidence on the potential health impacts of exposure to PFAA isomers. We found that increased follicular fluid PFAAs were associated with a lower oocyte maturity rate, which is a marker of the efficiency of ovarian stimulation and triggering in IVF treatment. 44 Our findings further suggested that oocyte maturity rate may partially mediate the adverse associations between greater PFAAs and poorer embryo quality. Recent in vitro and in vivo studies have shown that PFAAs can impair oocyte maturation, but the epidemiological evidence to date is limited. 10 Similar to our results, Ma et al. reported that higher maternal serum PFOA concentrations were significantly associated with fewer mature oocytes and fewer good-quality embryos among 97 women using IVF. 43 Still, additional epidemiological studies are needed to more definitively characterize the potential adverse association between PFAA exposure and fertility and the potential mediating role of oocyte quality.
Recent literature reviews suggest that health risks may differ for exposure to linear and branched PFAA isomers. 45,46 Although a large number of studies have reported adverse effects of exposure    Table S11 and Table S12 for numeric data. to total PFAA in early life, including differences in hormones, very few have investigated associations with PFAA isomers to date. 47 For example, recent studies have shown that exposure to branched PFAA were associated with subclinical maternal hypothyroidism 47 and with lower estradiol levels in young men. 48 We found branched PFOS isomers to be the major contributor to the association between a mixture of follicular fluid PFAA and a lower highquality embryo rate in the current study. On the contrary, another recent study from China observed no relationship of exposure to branched PFOS isomers in follicular fluid with embryo quality and implantation in women who underwent IVF treatment. 24 However, the number of participants in the aforementioned study was smaller (n = 124 vs. n = 729), and the concentration of branched PFOS was also lower than that in our study (< limit of quantitation vs. 0:47 ng=mL). PFAA isomer-specific epidemiological and toxicological studies will be needed to more definitively characterize potential differences in reproductive health risks given broad exposure to PFAA isomers.

Biologic Mechanisms
The results of experimental studies suggest that PFAA have adverse effects on ovarian folliculogenesis by altering oocyte development. 10,49,50 Exposure to PFAA induced apoptosis in human ovarian granulosa cells 51 and disrupted steroidogenic secretion in both granulosa and porcine theca cells with or without gonadotropic stimulation. 52 Using an ex vivo and transcript sequencing approach, Khan et al. found that PFOS was the active component in a mixture of PFOS, PFOA, and PFNA that altered normal ovarian function in Atlantic cod fish. 53 Neonatal exposure to high-dose PFOS or PFOA (0.1, 1, and 10 mg=kg=d) reduced follicle numbers in female rats, although no altered ovarian morphology was observed. 54 Recent studies showed that PFOS 55 and PFNA 56 could impede oocyte maturation by damaging mitochondrial functions in vitro, inducing oxidative stress and causing apoptosis of oocytes. In an in vitro bovine oocyte model, Hallberg et al. also found that exposure to human-relevant PFOS concentrations altered early embryo development, which may have negative consequences for further development. 57 A plausible mechanism for the relationship between PFAA exposure and lower high-quality embryo rate might be attributed to activation of peroxisome proliferator-activated receptors (PPARs) 58 that play a crucial role in gamete function and oocyte development. 59 PFAA binding to PPARs could interact with steroidogenesis response elements and have deleterious impacts on follicular and subsequently embryo development. In ovo exposure to environmentally relevant levels of PFOS suppressed transcription of genes involved in PPAR-mediated transcription. 60 Exposure of bovine oocytes to 0:01 lg=mL-100 lg=mL PFHxS during in vitro maturation affected pathways downstream of PPARc and estrogen signaling and decreased oocyte developmental competence, leading to compromised embryo quality and development. 49 Sant et al. reported that PFOS induced apoptosis and oxidative stress in zebrafish embryos, probably by activating potential nuclear factorerythroid 2-related factor 2 (Nrf2)-PPAR crosstalk. 61 Another study showed that a mixture of linear and branched PFOS elicited a greater transcriptional response than linear PFOS alone elicited, including PPARc signaling, in chicken embryo hepatocytes, suggesting that the isomer-specific toxicological properties of PFOS should be considered. 62 However, the PFAA doses used in animal studies were much higher than the PFAA concentrations encountered by most human populations, which may activate different toxicity pathways and toxic effects. Experimental studies using PFAA doses equivalent to human background exposures and using PFAA mixtures are required in future studies to elucidate the potential effects of PFAA on human fertility more clearly. 10

Strengths and Limitations
Our study has some major strengths and novel features. First, we measured PFAA concentrations in the follicle microenvironment, which may more closely approximate the biologically effective dose to a developing oocyte than blood PFAA. In addition, our sample size of 729 follicular fluids is the largest reported to date, far eclipsing the sample sizes of previous studies, which measured no more than 100 follicular fluids. Second, we applied PFOS isomer-specific analysis to capture isomer-specific health associations, because crude measures of "total-PFOS" may have masked associations with branched isomers in previous work. 47 Third, we collected a large number of covariates, allowing us to adjust for confounding variables and to assess potential biases using sensitivity analyses, including the possibility for confounding and reverse causation by parity. 11 Fourth, we used a mediation analysis to investigate successful completion of meiosis 1 (i.e., MII-oocyte rate) as a potential biological mechanism driving the associations between PFAAs and embryo quality. Finally, we used BKMR models, which do not constrain the form or nature of associations a priori, to assess the overall effect of the PFAA mixture on embryo quality and to identify the major contributor of PFAA effects in the exposure-response relationship.
However, our findings should be interpreted as hypothesisgenerating, given some limitations. First, we collected and pooled two to four follicular fluid specimens from each participant, which may have misclassified exposure for some women if there is significant follicle-to-follicle variability in PFAA concentrations. However, plasma/serum and follicular fluid PFAA were strongly correlated in previous studies 19,21,23 with no difference 22 or only modestly greater plasma relative to follicular fluid PFAA concentrations, 19 suggesting that follicular fluid PFAA reflect the blood compartment. Still, follicle-to-follicle variability may have misclassified exposure in some follicles, and the pooled exposure estimate may have underestimated the associations with embryo quality. A future investigation using a "one-follicle, oneoocyte" design will be necessary for a more definitive result. We also did not incorporate seminal PFAA concentrations from the male partner into the current analysis, which may have further misclassified exposure or introduced unmeasured copollutant confounding. A recent study showed that the seminal PFOS and PFOA concentrations were associated with a lower percentage of progressive sperm and higher percentage of DNA fragmentation in men (n = 664), suggesting deleterious effects of PFAA exposure on sperm quality. 63 This finding may indicate a potential contribution of PFAA exposure from seminal PFAA concentrations on embryo fertilization. We plan to validate our findings using male partners' semen PFAA data in the next step. Second, we cannot rule out the possible contamination of blood PFAA in follicular fluid, which may introduce exposure measurement error. However, any occult blood contamination was unlikely to have been differential by embryo quality outcome and will most likely bias the results toward the null hypothesis. Third, IVF patients tend to be highly selected, often comprising older couples actively trying to conceive and with higher socioeconomic status than the general population, 64 which may restrict generalizability of the study results. However, IVF patients may also be a vulnerable population, with heightened sensitivity to environmental chemical exposures providing sentinel indicators of reproductive toxicity. 65

Conclusions
This study found significant adverse associations between follicular fluid PFAA and embryo quality. Oocyte maturity may partially mediate the associations. Important PFAA isomeric associations with poorer embryo quality were also suggested, which may subsequently impact pregnancy and live birth from IVF. However, a future follow-up study of the associations between follicular fluid PFAA with pregnancy and live-birth outcomes is needed to estimate the impact directly. These findings may have important public health implications that help contribute to understanding potential environmental risk factors for unfavorable IVF outcomes and offer regulators and policy makers additional evidence to manage use of PFAA.