Association between Perfluorooctanoic Acid-Related Poor Embryo Quality and Metabolite Alterations in Human Follicular Fluid during IVF: A Cohort Study.

A flowchart of the patients included in this study is shown in Figure 1 . The baseline characteristics of the 378 women in the present study undergoing IVF/ICSI are listed in Table 1 . The average age was 31.11 ± 3.33 years and average BMI was 21.85 ± 3.22   kg / m 2 . The patients had been infertile for about 2 years, and 68.8% and 31.2% of them exhibited primary and secondary infertility, respectively. Female factors accounted for 66.9% of all causes of infertility, and the GnRH antagonist ovarian stimulation protocol was administered in 69.8% of all cases. Of all couples, 61.1% used conventional IVF while 34.9% underwent ICSI. The basal follicle-stimulating hormone, luteinizing hormone, and estradiol levels are listed in Table 1 . The flowchart of the participants from a hospital-based study in Shanghai, China (2023). Note: ART, assisted reproductive technology; ICSI, intracytoplasmic sperm injection; IVM, in vitro maturation; PGT, preimplantation genetic testing; RSA, recurrent spontaneous abortion. We identified 29 types of PFAS in the follicular fluid samples. Of these, PFOA, PFDA, PFHpS, n-PFHxS, PFNA, n-PFOS, 6:2 Cl-PFESA, PFPeS, PFUdA, PFTrDA, 8:2 Cl-PFESA, PFBS, Br-PFHxS, FOSA-I, and PFHxA had detection frequencies > 85 % ( Table 2 ). PFOA had the highest median concentration ( 6.54  ng / mL ), followed by n-PFOS ( 3.13  ng / mL ), n-PFHxS ( 1.17  ng / mL ), and 6:2 Cl-PFESA ( 1.08  ng / mL ) ( Table 2 ). In general, the concentrations of the individual PFAS were mutually positively correlated (Figure S2; Excel Table S5). To determine the potential confounders (age, BMI, menstrual cycle, menstrual day, factors of infertility, stimulation protocol, fertilization method, type of infertility, duration of infertility, employment situation, and education level), a directed acyclic graph was made, as shown in Figure S1. These confounders were adjusted for the following analysis. Associations between higher concentrations of PFAS in follicular fluid and lower likelihoods for the high-quality embryo rates were detected using beta regression models adjusted for confounders {adjusted relative risk (aRR) = 0.965 [95% confidence interval (CI): 0.950, 0.981], per log PFOA; aRR = 0.933 (95% CI: 0.874, 0.996), per log PFDA; aRR = 0.885 (95% CI: 0.828, 0.946), per log PFHpS; aRR = 0.892 (95% CI: 0.839, 0.948), per log n-PFHxS; aRR = 0.880 (95% CI: 0.810, 0.955), per log PFNA; aRR = 0.935 (95% CI: 0.875, 0.998), per log n-PFOS; aRR = 0.955 (95% CI: 0.915, 0.996), per log 6:2 Cl-PFESA} ( Table 3 ). Similar to the high-quality embryo rates, we observed a significant negative association between five of these PFAS and clinical pregnancy outcomes after adjustment [adjusted odds ratio (aOR) = 0.908 (95% CI: 0.844, 0.972), per log PFOA; aOR = 0.727 (95% CI: 0.541, 0.968), per log PFDA; aOR = 0.604 (95% CI: 0.420, 0.855), per log PFHpS; aOR = 0.612 (95% CI: 0.418, 0.888), per log PFNA; and aOR = 0.728 (95% CI: 0.541, 0.968), per log n-PFOS] ( Table 4 ). Stratified analysis was then performed on the infertility population caused by female factor or male factor. Associations between PFAS level and high-quality embryo rate or clinical pregnancy outcomes in infertile women caused by female factors remained similar as the total population after adjusting for confounders (Tables S6 and S7), while there were less associations in infertile women caused by male factors, which may be related to the small sample size of the subgroups (Tables S8 and S9). Association between the PFAS level and high-quality embryo rate for participants from a hospital-based study in Shanghai, China, 2023–2024 ( n = 378 ). Note: Values are presented as RR (95% CI) or aRR (95% CI). 6:2 Cl-PFESA, potassium 9-chlorohexadecafluoro-3-oxanonane-1-sulfonate; 8:2 Cl-PFESA, potassium 11-chloroeicosafluoro-3-oxaundecane-1-sulfonate; aRR, adjusted relative risk; BMI, body mass index; Br-PFHxS, sum of all branched isomers PFHxS; CI, confidence interval; FOSA-I, perfluoro-1-octanesulfonamide; n-PFHxS, potassium perfluorohexanesulfonate; PFAS, perfluoroalkyl and polyfluoroalkyl substances; PFBS, potassium perfluoro-1-butanesulfonate; PFDA, perfluoro- n -decanoic acid; PFHpS, sodium perfluoro-1-heptanesulfonate; PFHxA, perfluoro- n -hexanoic acid; n-PFOS, potassium perfluoro-1-octanesulfonate; PFNA, perfluoro- n -nonanoic acid; PFOA, perfluoro- n -octanoic acid; PFPeS, sodium perfluoro-1-pentanesulfonate; PFTrDA, perfluoro- n -tridecanoic acid; PFUdA, perfluoro- n -undecanoic acid; RR, relative risk. Univariate and multivariable beta regression model were used. RR (95% CI) and aRR (95% CI) were estimated to represent the ratio of the altered high-quality embryo rate to the baseline high-quality embryo rate, indicating the risk associated with high-quality embryo rates for each one-unit increase in the log-transformed PFAS levels. Adjustment included female age, BMI, menstrual cycle, menstrual day, factors of infertility, stimulation protocol, fertilization method, type of infertility, duration of infertility, employment, and education level. p -Value for Wald test. Association between the PFAS level and clinical pregnancy outcome for participants from a hospital-based study in Shanghai, China, 2023–2024 ( n = 378 ). Note: Values are presented as OR (95% CI) or aOR (95% CI). 6:2 Cl-PFESA, potassium 9-chlorohexadecafluoro-3-oxanonane-1-sulfonate; 8:2 Cl-PFESA, potassium 11-chloroeicosafluoro-3-oxaundecane-1-sulfonate; aOR, adjusted odds ratio; BMI, body mass index; Br-PFHxS, sum of all branched isomers PFHxS; CI, confidence interval; FOSA-I, perfluoro-1-octanesulfonamide; n-PFHxS, potassium perfluorohexanesulfonate; n-PFOS, potassium perfluoro-1-octanesulfonate; OR, odds ratio; PFBS, potassium perfluoro-1-butanesulfonate; PFDA, perfluoro- n -decanoic acid; PFHpS, sodium perfluoro-1-heptanesulfonate; PFHxA, Perfluoro- n -hexanoic acid; PFNA, perfluoro- n -nonanoic acid; PFOA, perfluoro- n -octanoic acid; PFPeS, sodium perfluoro-1-pentanesulfonate; PFTrDA, perfluoro- n -tridecanoic acid; PFUdA, perfluoro- n -undecanoic acid. Univariate and multivariate logistic regression models were used. OR (95% CI) and aOR (95% CI) were estimated by one standard deviation higher difference in log-transformed PFAS as continuous variables. Adjustment included female age, BMI, menstrual cycle, menstrual day, factors of infertility, stimulation protocol, fertilization method, type of infertility, duration of infertility, employment, and education level. p -Value for Wald test. A causal mediation analysis revealed that embryo quality potentially mediated the relationship between clinical pregnancy outcome and PFOA (proportion mediated: 0.181; 95% CI: 0.024, 0.755), PFNA (proportion mediated: 0.148; 95% CI: 0.022, 0.656), or PFTrDA (proportion mediated: 0.130; 95% CI: 0.005, 0.693) ( Table 5 ). Causal mediation analysis of high-quality embryo rates in associations between PFAS and clinical pregnancy outcomes for participants from a hospital-based study in Shanghai, China, 2023–2024 ( n = 378 ). Note: Values are presented as 95% CI. 6:2 Cl-PFESA, potassium 9-chlorohexadecafluoro-3-oxanonane-1-sulfonate; 8:2 Cl-PFESA, potassium 11-chloroeicosafluoro-3-oxaundecane-1-sulfonate; BMI, body mass index; Br-PFHxS, sum of all branched isomers PFHxS; FOSA-I, perfluoro-1-octanesulfonamide; n-PFHxS, potassium perfluorohexanesulfonate; n-PFOS, potassium perfluoro-1-octanesulfonate; PFBS, potassium perfluoro-1-butanesulfonate; PFDA, perfluoro- n -decanoic acid; PFHpS, sodium perfluoro-1-heptanesulfonate; PFHxA, perfluoro- n -hexanoic acid; PFNA, perfluoro- n -nonanoic acid; PFOA, perfluoro- n -octanoic acid; PFPeS, sodium perfluoro-1-pentanesulfonate; PFTrDA, perfluoro- n -tridecanoic acid; PFUdA, perfluoro- n -undecanoic acid. Causal mediation effect model was performed and adjusted for female age, BMI, menstrual cycle, menstrual day, factors of infertility, stimulation protocol, fertilization method, type of infertility, duration of infertility, employment, and education level. p -Value for quasi-Bayesian approximation. We next used a restricted cubic spline with five nodes (5th, 27.5th, 50th, 72.5th, and 95th percentiles) to explore the potential nonmonotonic response between PFAS exposure and high-quality embryo rates. The results demonstrated that some of the PFAS had a nonlinear relationship with high-quality embryo rates and were, therefore, applicable for the BKMR model (Figure S3; Excel Table S6). The BKMR model was then used to estimate the overall effect of the PFAS mixture, and this model demonstrated that the concentration of PFAS in follicular fluid was negatively associated with the high-quality embryo rate ( Figure 2A ; Excel Table S1). We then estimated the contributions of individual PFAS to the overall effect of the PFAS mixtures and discovered that PFOA was significantly associated with a lower high-quality embryo rate ( Figure 2B ; Excel Table S2). We then calculated the PIPs and measured variable importance in the range of 0–1 (least to most important) to identify individual PFAS that contributed the most to the main effect of the mixture. PFOA was the dominant contributor (conditional PIP = 0.97295 ; Table S10). A quantile-based g-computation model was then constructed to validate the foregoing findings, and this model also showed that PFOA was the main contributor ( weight = 0.278 , β : − 0.052 , 95% CI: − 0.096 , − 0.009 ) (Table S11). Impact of perfluoroalkyl and polyfluoroalkyl substances (PFAS) mixture and individual PFAS on high-quality embryo rate for participants from a hospital-based study in Shanghai, China (2023–2024). Corresponding numerical data are shown in Excel Tables S1 and S2. (A) Overall impact of PFAS mixture on estimated differences (95% confidence interval) in the high-quality embryo rate ( n = 378 ). (B) Impact of individual PFAS on estimated differences (95% confidence interval) in high-quality embryo rate. Bayesian kernel machine regression (BKMR) model was used, and all PFAS in the 0.25–0.75 percentile were compared against those in the 50th percentile. Note: 6:2 Cl-PFESA, potassium 9-chlorohexadecafluoro-3-oxanonane-1-sulfonate; 8:2 Cl-PFESA, potassium 11-chloroeicosafluoro-3-oxaundecane-1-sulfonate; Br-PFHxS, sum of all branched isomers PFHxS; FOSA-I, perfluoro-1-octanesulfonamide; n-PFHxS, potassium perfluorohexanesulfonate; n-PFOS, potassium perfluoro-1-octanesulfonate; PFBS, potassium perfluoro-1-butanesulfonate; PFOA, perfluoro- n -octanoic acid; PFDA, perfluoro- n -decanoic acid; PFHpS, sodium perfluoro-1-heptanesulfonate; PFHxA, perfluoro- n -hexanoic acid; PFNA, perfluoro- n -nonanoic acid; PFPeS, sodium perfluoro-1-pentanesulfonate; PFTrDA, perfluoro- n -tridecanoic acid; PFUdA, perfluoro- n -undecanoic acid. To explore the underlying factors for the correlation between PFOA exposure and embryo quality, the participants were divided into the good embryo group (GE, high embryo rate > 60 % ) and poor embryo group (PE, high embryo rate < 30 % ). Twenty-five samples were enrolled in each group by computerized randomization and used for untargeted metabolomics analyses. We identified 692 and 504 annotated metabolites in positive-ion and negative-ion modes, respectively. OPLS-DA analysis showed a clear distinction between the follicular fluid metabolites in the GE group vs. the PE group ( Figure 3A,B ; Excel Table S3). A permutation test was used to prevent overfitting of the OPLS-DA model ( Figure 3C,D ; Excel Table S3). In positive-ion mode, 41 differentially expressed metabolites were detected between the GE and PE groups based on variable importance for the projection > 1 and p < 0.05 . The p -values were adjusted for multiple comparisons using the Benjamini–Hochberg method. The adjusted and unadjusted p -values of differential metabolites are shown in Table S12. Thirteen and 28 metabolites were significantly decreased and increased, respectively, in the PE group relative to the GE group ( Figure 3E ; Excel Table S3). In the negative-ion mode, 44 differentially expressed metabolites were detected. Nine and 35 metabolites were significantly decreased and increased, respectively, in the PE group relative to the GE group ( Figure 3F ; Excel Table S3). Thirty-nine of the differentially expressed metabolites were lipids and lipid-like molecules. Twelve metabolites were organic acids and their derivatives, nine metabolites were benzenoids, eight metabolites were organoheterocyclic compounds, and six metabolites were organonitrogens ( Figure 3G ; Excel Table S3). Ovarian follicular fluid metabolomics for participants from a hospital-based study in Shanghai, China (2023–2024). Corresponding numerical data are shown in Excel Table S3. (A,B) Orthogonal partial least-squares discriminant analysis (OPLS-DA) model showing metabolites in good embryo (GE) ( n = 25 ) and poor embryo (PE) ( n = 25 ) groups in positive-ion mode (A) and in negative-ion mode (B). (C,D) Permutation test of OPLS-DA model verification in positive-ion mode (C) and in negative-ion mode (D). (E,F) Bar plot of differentially expressed metabolites between GE and PE groups in positive-ion mode (E) and in negative-ion mode (F). (G) Classification of differentially expressed metabolites. Note: FC, fold change; PC, principal component. Spearman’s rank correlation coefficient analyses were performed on the differentially expressed metabolites and high-quality embryo rates between the GE and PE groups. The absolute value of the correlation coefficient ( ρ ) > | 0.5 | and p < 0.05 indicated a significant correlation. The high-quality embryo rate was positively correlated with proline-tryptophan (Pro-Trp), phytosphingosine, N -myristoylsphinganine, N -lauroyl-d-erythro-sphinganine, and lauryldimethylamine oxide, but negatively correlated with benzyl alcohol, p-fluoro-.alpha.-[1-(methylamino)ethyl]-, erythro-(.+/−.)-(benzyl alcohol), and 9-chlorohexadecafluoro-3-oxanonane-1-sulfonate (9cl-pf3ons), as depicted in the linear regression charts ( Figure 4A–G ; Excel Table S4). Pro-Trp, phytosphingosine, N -myristoylsphinganine, N -lauroyl-d-erythro-sphinganine, and lauryldimethylamine oxide levels were significantly lower, while the benzyl alcohol and 9cl-pf3ons levels were significantly higher in the PE group vs. the GE group ( Figure 4H–N ; Excel Table S4). Association between differentially expressed metabolites and high-quality embryo rates in participants from the good embryo and the poor embryo groups for participants in a hospital-based study in Shanghai, China (2023–2024). Corresponding numerical data are shown in Excel Table S4. (A–G) Linear regression of high-quality embryo rates with proline-tryptophan (Pro-Trp) (A), phytosphingosine (B), N -myristoylsphinganine (C), N -lauroyl-d-erythro-sphinganine (D), lauryldimethylamine oxide (E), benzyl alcohol (F), and 9-chlorohexadecafluoro-3-oxanonane-1-sulfonate (9cl-pf3ons) (G). (H–N) Comparison of preceding metabolite levels between good embryo group (GE) ( n = 25 ) and poor embryo group (PE) ( n = 25 ). Student’s t -test was applied to determine the significance of differences between two groups of independent samples. The p -values were adjusted for multiple comparisons using the Benjamini–Hochberg method. Note: AUC, area under the curve.

A single-center cross-sectional and prospective cohort study was conducted that included 378 women with infertility who had undergone their first round of IVF/ICSI at the Reproductive Center of Shanghai First Maternity and Infant Hospital (Shanghai, China) between May 2023 and November 2023. Infertility is defined as a disease characterized by the failure to establish a clinical pregnancy after 12 months of regular and unprotected sexual intercourse. The study exclusion criteria were a ) age ≥ 38 years; b ) endocrine system abnormalities including hyperprolactinemia, diabetes mellitus, polycystic ovary syndrome, and thyroid or adrenal dysfunction; c ) couples with abnormal karyotypes; d ) women with recurrent spontaneous abortion or uterine abnormalities, such as Müllerian duct anomaly, adenomyosis, submucous myoma, intrauterine adhesion, or scarred uterus; and e ) preimplantation genetic testing, in vitro maturation, or rescue ICSI. Definitions and concepts of relevant terms are shown in Table S1. All participants provided written informed consent and agreed to provide analytical samples. The protocol of the present study was approved by the Research Ethics Committee of Shanghai First Maternity and Infant Hospital (number KS23197). All participants were subjected to controlled ovarian stimulation according to the standard procedure of the Reproductive Center. A long gonadotrophin-releasing hormone (GnRH) agonist, short GnRH agonist, or GnRH antagonist protocol was used, all of which were performed as previously described in detail. 21 , 22 When at least two follicles measured ≥ 18 mm , human chorionic gonadotropin (5,000 IU or 250 μ g ) was administered to induce oocyte maturation. Oocyte retrieval was performed after 34–36 h. At that time, about 5 mL of follicular fluid was collected from two or three mature follicles with mean diameters ≥ 18 mm . Follicular fluid in the first tube was not collected to avoid contamination caused by flushing the media in the tubing system. The samples were centrifuged at 3,000  rpm for 30 min to separate cells from the supernatants, which were stored at − 80 ° C until use for subsequent detection of PFAS. The oocytes were fertilized 4–6 h after follicular aspiration by IVF or ICSI. After 17–19 h, normal fertilization was confirmed by the presence of two pronuclei. Embryo quality was evaluated based on blastomere number and regularity and the percentages and patterns of anucleate fragments, with reference to our previously published paper. 23 In brief, embryos were graded on day 2 or 3 after insemination as grade 1 to grade 6 according to the evenness of each blastomere and the percentage of fragmentation. Grade 1 embryos had equal-sized blastomeres and < 5 % fragments; grade 2 embryos had equal-sized blastomeres and < 25 % fragmentation; grade 3 embryos had blastomeres of distinctly unequal size and < 5 % fragments; grade 4 embryos had blastomeres of distinctly unequal size and < 25 % fragmentation; Grade 5–6 embryos had more than 25% and 50% fragmentation, respectively. 22 Embryos with two pronuclei on the first day, 4–5 blastomeres on the second day, and 7–9 blastomeres on the third day, and of grades 1 or 2 were regarded as high-quality embryos. All embryo assessments at our center are conducted by embryologists possessing a minimum of 5 years of professional experience, who were subject to stringent monthly quality control to ensure data accuracy. For embryos with indeterminate scores, a minimum of three embryologists verified the final assessment. In addition, embryo evaluation in our embryology laboratory primarily adheres to the Istanbul consensus, 24 established by embryologists around the world and widely adopted by reproductive centers globally to formulate their scoring criteria. The high-quality embryo rate was calculated by dividing the number of high-quality embryos (numerator) by the number of normal fertilizations (denominator) per patient. Transvaginal ultrasonography was performed 4–6 wk after embryo transfer to confirm clinical pregnancy, which was defined as the detection of an intrauterine gestational sac. Twenty-nine PFAS (PFOA, perfluoro- n -octanoic acid; PFDA, perfluoro- n -decanoic acid; PFHpS, sodium perfluoro-1-heptanesulfonate; n-PFHxS, potassium perfluorohexanesulfonate; PFNA, perfluoro- n -nonanoic acid; n-PFOS, potassium perfluoro-1-octanesulfonate; 6:2 Cl-PFESA, potassium 9-chlorohexadecafluoro-3-oxanonane-1-sulfonate; PFPeS, sodium perfluoro-1-pentanesulfonate; PFUdA, perfluoro- n -undecanoic acid; PFTrDA, perfluoro- n -tridecanoic acid; 8:2 Cl-PFESA, potassium 11-chloroeicosafluoro-3-oxaundecane-1-sulfonate; PFBS, potassium perfluoro-1-butanesulfonate; Br-PFHxS, sum of all branched isomers PFHxS; FOSA-I, perfluoro-1-octanesulfonamide; PFHxA, perfluoro- n -hexanoic acid; PFHpA, perfluoro- n -heptanoic acid; PFDoA, perfluoro- n -dodecanoic acid; PFBA, perfluoro- n -butanoic acid; HFPO-DA, 2,3,3,3-tetrafluoro-2-(1,1,2,2,3,3,3-heptafluoropropoxy) propanoic acid; 6:2FTS, sodium 1H, 1H, 2H, 2H-perfluoro-1-octanesulfonate; PFPeA, perfluoro- n -pentanoic acid; PFTeDA, perfluoro- n -tetradecanoic acid; NaDONA, sodium dodecafluoro-3H-4,8-dioxanonanoate; N-MeFOSAA, N -methylperfluoro-1-octanesulfonamidoacetic acid; 4:2FTS, sodium 1H, 1H, 2H, 2H-perfluoro-1-hexanesulfonate; PFNS, sodium perfluoro-1-nonanesulfonate; 8:2FTS, sodium 1H, 1H, 2H, 2H-perfluoro-1-decanesulfonate; PFDS, sodium perfluoro-1-decanesulfonate; and N-EtFOSAA, N -ethylperfluoro-1-octanesulfonamidoacetic acid) were detected quantitatively by ultrahigh-performance liquid chromatography (UHPLC) coupled with triple quadruple tandem mass spectrometry (UPLC-MS/MS) and electrospray ionization in negative mode (AB 5500 + ; AB Sciex LLC). All PFAS standards were purchased from Wellington Laboratories. Follicular fluid ( 0.2 mL ) was mixed with 10  ng / mL PFAS internal standards and 1 mL of 0.1 M formic acid and extracted with a hydrophilic–lipophilic balanced solid phase extraction cartridge ( 60 mg / 3  cc ; Waters Corp.). The extracts were centrifuged at 15,000  rpm and 4°C for 10 min. Then 20 μ L of each supernatant was transferred to an autosampler vial and subjected to UPLC-MS/MS analysis. To monitor contamination, a sheep plasma sample was used as a blank control and handled together with follicular fluid samples. If PFAS were detected in the blank control, it was considered that there was contamination in the operation process. Sheep plasma samples supplemented with PFAS mixed standards were used as a quality control to determine the reliability of the PFAS testing. The actual PFAS concentration was the sample PFAS concentration minus the blank plasma PFAS concentration interpolated from the standard curve. The PFAS mass spec was conducted in Shanghai Key Laboratory of Children’s Environmental Health, Xinhua Hospital. The full chemical names of all PFAS are listed in Table S2. The limits of detection (LOD) and the limits of quantification of the PFAS are listed in Table S3. PFAS concentrations below the LOD value were estimated as the limits of detection divided by the square root of 2. 25 To explore the underlying factors for the correlation between PFOA exposure and embryo quality, the participants were divided into the good embryo (GE) group (high embryo rate > 60 % ) and poor embryo (PE) group (high embryo rate < 30 % ). Twenty-five samples were enrolled in each group by computerized randomization and then subjected to untargeted metabolomics analyses. Untargeted metabolomics was performed by Applied Protein Technology Co. One milliliter of precooled buffer consisting of 2:2:1 (vol/vol/vol) methanol:acetonitrile: H 2 O was added to each follicular fluid sample. The mixture was subjected to low-temperature ultrasound for 30 min, maintained at − 20 ° C for 10 min, and centrifuged at 14,000 × g and 4°C for 15 min. The supernatant was collected, dried, redissolved in 1:1 (vol/vol) acetonitrile:water, and subjected to UHPLC (1290 Infinity LC; Agilent Technologies Inc.). An ACQUITY UPLC BEH amide column (Waters Corp.) was used for separation. The mobile phase consisted of 25  mM ammonium acetate and 25  mM ammonium hydroxide in water (A) and acetonitrile (B). Random injection sequences were used to avoid signal fluctuations during sample processing. Quality control samples were also injected and analyzed to ensure data credibility. The samples were further detected by quadrupole time-of-flight mass spectrometry (QToFMS) (Triple TOF 6600; AB Sciex Inc.) in both the positive and negative ion modes of electrospray ionization. Secondary mass spectra were obtained using the information-dependent acquisition and peak intensity-screening modes. Metabolites were analyzed by comparing their m / z accuracy ( < 25  ppm ) and MS/MS spectra against an internal database of recognized standards. Simple randomization of samples was performed by the sample () function in R (version 4.3.1; R Development Core Team). A dedicated “ropls” package in R was used in the orthogonal partial least-squares discriminant analysis (OPLS-DA) ( https://bioconductor.org/packages/release/bioc/html/ropls.html ). Differentially expressed metabolites between groups were identified based on variable importance for the projection threshold > 1 and validated by adjusted p -value < 0.05 . The results were corrected for multiple comparisons by Benjamini–Hochberg method, a widely used approach for controlling the false discovery rate (FDR). According to previous literature, 26 – 28 the potential confounding variables [including age (linear), body mass index (BMI) (linear), duration of infertility (linear), menstrual cycle (linear), menstrual day (linear), factors of infertility (female factor, male factor, multiple factors, and unexplained infertility), stimulation protocol (long GnRH agonist, short GnRH agonist, and GnRH antagonist), fertilization method (IVF, ICSI, and IVF+ICSI), type of infertility (primary infertility was defined as women who have never had a history of pregnancy or abortion and secondary infertility was defined as women who have had a history of pregnancy), employment situation (employed and unemployed), and education levels (junior college degree or below, college degree, and postgraduate degree)] were determined and plotted into a directed acyclic graph (DAG) (Figure S1) using DAGitty ( http://dagitty.net/ ). The above information was collected from participants’ medical records. Definitions and concepts of relevant terms are shown in Table S1. Continuous data included means ± standard deviation  ( SD ) or medians (quartile 1 or 3) depending on whether the data were normally distributed. Categorical data were expressed as frequencies or percentages. There were no missing variable data in characteristics of participants ( Table 1 ). PFAS concentrations were   log   e (ln)-transformed to improve normality and mitigate the influences of outliers before the association analyses. Spearman’s correlation coefficient analyses were used to detect bivariate correlations among the PFAS congeners in the follicular fluid. Sociodemographic and clinical factors of women undergoing ART for participants from a hospital-based study in Shanghai, China (2023). Note: Values are shown as mean ± SD or number (%). ART, assisted reproductive technology; BMI, body mass index; E 2 , estradiol; FSH, follicle-stimulating hormone; Gn, gonadotrophin; GnRH, gonadotrophin releasing hormone; ICSI, intracytoplasmic sperm injection; IVF, in vitro fertilization; LH, luteinizing hormone. There were no missing variable data in characteristics of participants. Associations between the individual PFAS levels and the high-quality embryo rates were analyzed by beta regression without confounding factor adjustment in the crude model and adjusted for confounders including female age (year), BMI ( kg / m 2 ), menstrual cycle (day), menstrual day (day), infertility factors (female factor, male factor, multiple factor, and unexplained infertility), stimulation protocol (long GnRH agonist, short GnRH agonist, and GnRH antagonist), fertilization method (IVF, ICSI, and IVF+ICSI), infertility type (primary or secondary infertility) and duration of infertility (year), employment situation (employed and unemployed), and education level (junior college degree or below, college degree, and postgraduate degree) in the main model. Tolerance 10 generally indicated multicollinearity 29 – 31 (Table S4). Restricted cubic spline (RCS) models with knots at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles of each PFAS (Table S5) were fitted to further identify potential nonlinearities between PFAS and high-quality embryo rates. In the beta regression model, the relative risk (RR) was estimated to represent the ratio of the altered high-quality embryo rate to the baseline high-quality embryo rate, indicating the risk associated with high-quality embryo rates for each one-unit increase in the log-transformed PFAS levels. The associations between the individual PFAS levels and clinical pregnancy outcomes were analyzed by logistic regression without confounding factor adjustment in the crude model and adjusted for above confounders in the main model. Stratified analysis was then performed on the infertility population caused by female factor or male factor. Considering the interrelationships among PFAS concentrations in follicular fluid, we analyzed the potential joint effect estimates of a PFAS mixture on the high-quality embryo rate using the Bayesian kernel machine regression (BKMR) model. The BKMR model allows for the assessment of independent associations of individual PFAS mixture components with an outcome, in addition to the overall combined mixed PFAS exposure. 32 Simulation studies have demonstrated that BKMR can fully fit the potential complex nonlinear relationship and estimate the difference of study outcomes under different exposure levels and possible interactions between pollutants. Here, only the PFAS in the follicular fluid with detection frequencies > 85 % 26 (PFOA, PFDA, PFHpS, n-PFHxS, PFNA, n-PFOS, 6:2 Cl-PFESA, PFPeS, PFUdA, PFTrDA, 8:2 Cl-PFESA, PFBS, Br-PFHxS, FOSA-I, and PFHxA) ( Table 2 ) were entered into the BKMR models. For all BKMR analyses, the PFAS were subjected to ln-transformations, centered on x ¯ = 0 , and scaled to SD = 1 to account for skewness and mitigate the influences of extreme values and different value scales on the variables. All continuous variables were also standardized using the z -score method prior to fitting the BKMR models. Each BKMR model was adjusted for the corresponding covariates as previously described. The overall PFAS mixture effect was estimated from the average outcome at the 75th percentile of the PFAS concentrations minus the average outcome at the 25th percentile of the PFAS concentrations. The individual effects of each PFAS component were estimated by subtracting the average result of the specified PFAS at the 25th percentile from the average result of the specified PFAS at the 75th percentile. Other PFAS congeners were fixed at their medians, and all covariates remained unchanged. The conditional posterior inclusion property (PIP) was used to indicate the relative importance of each PFAS within the mixture. The variable importance was in the range of 0–1. Individual PFAS with conditional PIP > 0.5 were considered major contributors. 33 – 35 All BKMR models were fitted to 50,000 iterations using the Markov chain Monte Carlo (MCMC) sampler to improve convergence. Chains thinned to every 10th iteration after running the MCMC to reduce autocorrelation. PFAS concentrations in follicular fluid for participants from a hospital-based study in Shanghai, China, 2023–2024 ( n = 378 ). Note: 4:2FTS, sodium 1H, 1H, 2H, 2H-perfluoro-1-hexanesulfonate; 6:2 Cl-PFESA, potassium 9-chlorohexadecafluoro-3-oxanonane-1-sulfonate; 6:2FTS, sodium 1H, 1H, 2H, 2H-perfluoro-1-octanesulfonate; 8:2 Cl-PFESA, potassium 11-chloroeicosafluoro-3-oxaundecane-1-sulfonate; 8:2FTS, sodium 1H, 1H, 2H, 2H-perfluoro-1-decanesulfonate; Br-PFHxS, sum of all branched isomers PFHxS; FOSA-I, perfluoro-1-octanesulfonamide; HFPO-DA, 2,3,3,3-tetrafluoro-2-(1,1,2,2,3,3,3-heptafluoropropoxy) propanoic acid; LOD, limits of detection; NaDONA, sodium dodecafluoro-3H-4,8-dioxanonanoate; N-EtFOSAA, N -ethylperfluoro-1-octanesulfonamidoacetic acid; N-MeFOSAA, N-methylperfluoro-1-octanesulfonamidoacetic acid; n-PFHxS, potassium perfluorohexanesulfonate; n-PFOS, potassium perfluoro-1-octanesulfonate; PFAS, perfluoroalkyl and polyfluoroalkyl substances; PFBA, perfluoro- n -butanoic acid; PFBS, potassium perfluoro-1-butanesulfonate; PFDA, perfluoro- n -decanoic acid; PFDoA, perfluoro- n -dodecanoic acid; PFDS, sodium perfluoro-1-decanesulfonate; PFHpA, perfluoro- n -heptanoic acid; PFHpS, sodium perfluoro-1-heptanesulfonate; PFHxA, perfluoro- n -hexanoic acid; PFNA, perfluoro- n -nonanoic acid; PFNS, sodium perfluoro-1-nonanesulfonate; PFOA, perfluoro- n -octanoic acid; PFPeA, perfluoro- n -pentanoic acid; PFPeS, sodium perfluoro-1-pentanesulfonate; PFTeDA, perfluoro- n -tetradecanoic acid; PFTrDA, perfluoro- n -tridecanoic acid; PFUdA, perfluoro- n -undecanoic acid. The individual and overall effects of the PFAS on the high-quality embryo rates were verified by the quantile-based g-computation method. It combines the simplicity of inference of weighted quantile regression with the flexibility of g-computation and accounts for the nonlinearity and nonadditivity of BKMR. In this way, the biases are decreased, and the robustness of the results are increased. 36 Binomial and Gaussian distributions were specified as link functions. The parameter q represented the specified quantile and was set to 4. We estimated the average indirect effect of PFAS on clinical pregnancy outcomes, mediated by high-quality embryo rates, along with the direct effect, which captures all other causal pathways, using the mediation R package. 37 The mediation effects were assessed using parametric bootstrapping, with p -values obtained through 1,000 Monte Carlo simulations, employing the package’s default settings. This approach allows for robust estimation of both indirect and direct effects while accounting for uncertainty in the causal pathways. Spearman’s rank correlation coefficient was used to identify the associations between the differentially expressed metabolites and the high-quality embryo rates. The key target metabolites were screened using Spearman’s rank correlation coefficient ( ρ ) > | 0.5 | and p < 0.05 . All statistical analyses were conducted in R version 4.3.1 (R Development Core Team). The BKMR, qgcomp, and mediation effect models were constructed with the “bkmr,” “qgcomp,” and “mediation” packages in R, respectively. The statistical significance level was set to p < 0.05 (two-tailed).

Past research showed higher PFAS concentrations are closely correlated with female infertility 27 ; however, it remained unclear which of the many different types of PFAS exert reproductive toxicity. The current large cohort study comprised 29 different PFAS in ovarian follicular fluid samples from 378 women with infertility undergoing IVF/ICSI. This study demonstrated that the follicular fluid concentrations of PFOA, PFDA, PFHpS, n-PFHxS, PFNA, n-PFOS, 6:2 Cl-PFESA, PFTrDA, 8:2 Cl-PFESA, and PFUdA were inversely correlated with embryo quality. PFOA was the major contributor to the overall adverse effects of mixtures of PFAS. To the best of our knowledge, the present study is one of the first to show that PFOA-related lower embryo quality was associated with the reduction of the organonitrogens (Pro-Trp and lauryldimethylamine oxide) and sphingolipids metabolites (phytosphingosine, N -myristoylsphinganine, and N -lauroyl-d-erythro-sphinganine) in follicular fluid. PFAS comprise a large class of synthetic fluorinated compounds. They are commonly occurring environmental toxins, and humans routinely make contact with them. 38 PFOA and PFOS are the most extensively manufactured of all PFAS compounds. 20 According to the National Biomonitoring Program of the United States, at least one PFAS chemical was detected in the blood of nearly every sample examined. 20 PFAS are often found in human biological samples such as blood, urine, semen, nails, and hair. 27 The present work identified 29 PFAS in ovarian follicular fluid samples from women with infertility residing in Shanghai, China, and its environs. PFOA, PFDA, PFHpS, n-PFHxS, PFNA, n-PFOS, 6:2 Cl-PFESA, PFPeS, PFUdA, PFTrDA, 8:2 Cl-PFESA, PFBS, Br-PFHxS, FOSA-I, and PFHxA had detection frequencies exceeding 85%. Of these, the PFOA and then n-PFOS had the highest concentrations. Zeng et al. 26 measured 23 different types of perfluoroalkyl acids (PFAA) in the follicular fluid of women with infertility living in Guangxi Province, China. Detection frequencies surpassed 85% for n-PFOS, n-PFOA, PFDA, PFHxS, PFHpS, PFNA, PFUnDa, and PFTrDA, and n-PFOS and n-PFOA had the highest concentrations. The levels of all other PFAA with high detection frequencies were consistent with those measured in the present study. Kang et al. 39 reported similar findings for the follicular fluid of women with infertility from Beijing, China. The PFOA and PFOS concentrations were highest in the follicular fluid samples from the Shanghai (Yangtze River Delta, n-PFOS: 3.13  ng / mL ; PFOA: 6.54  ng / mL ) region, moderate in those from the Beijing region (Northern China, PFOA: 3.38  ng / mL ; PFOS: 4.54  ng / mL ), and lowest in the samples from the Guangxi region (Southern China, n-PFOS: 1.70  ng / mL ; n-PFOA: 1.09  ng / mL ). Hence, environmental differences among the various regions of China may explain the observed variability in PFAS concentration of the follicular fluid samples collected from these areas. PFAS are immunotoxic, neurotoxic, and carcinogenic and induce developmental, reproductive, and genetic toxicity in animals and humans. 40 – 42 Thus, PFAS exposure could be a contributing factor in the steady decline in human fertility. Exposure to PFOA may reduce circulating sex hormone levels, 43 dysregulate the menstrual cycle, 12 cause premature ovarian insufficiency, 13 and induce premature menopause. 14 Recently, Cohen et al. 44 showed that higher PFAS exposures were associated with decreased fertility in women, and PFDA followed by PFOS, PFOA, and PFHpA were the biggest contributors in a Singaporean population-based preconception cohort. The ovaries produce sex hormones and are the sites of folliculogenesis and oocyte maturation. Follicular fluid reflects the oocyte microenvironment, and the proportions of biomolecules in follicular fluid may directly influence oocyte development. Governini et al. 45 measured the perfluorinated compounds in the follicular fluid of 18 patients undergoing IVF and found that those with higher levels of these compounds had relatively fewer transferred embryos and lower fertilization rates. However, the authors did not specify the types of perfluorinated compounds or their levels in the follicular fluid samples. In other studies, the PFDA and PFUnDA concentrations in follicular fluid were negatively correlated with the rates of blastocyst formation and high-quality embryos, as well as fertilization outcome. 26 , 28 The ovarian and plasma PFAA concentrations were strongly correlated with each other. However, there was no significant relationship between the PFAA concentration and the ovarian response. 28 Here, we discovered that mixtures of PFAS were negatively correlated with embryo quality and clinical pregnancy outcomes. The associations between the foregoing parameters with PFOA, PFNA, PFTrDA, and 8:2 Cl-PFESA were all statistically significant, but the correlations between PFOA and these factors were the strongest. Ma et al. 46 demonstrated that the PFOA concentration in maternal plasma was inversely correlated with the numbers of retrieved oocytes, mature oocytes, and two-pronucleus zygotes, as well as embryo quality. In vitro and in vivo mouse studies showed that PFOA exposure inhibited follicle growth and disrupted ovarian function. 47 Several studies have demonstrated a correlation between the presence of PFAS in follicular fluid and fertility outcomes. 20 , 26 Given that follicular fluid constitutes the immediate microenvironment of the oocyte, we utilized untargeted metabolomics to systematically investigate whether PFOA-associated poor embryo quality can be explained, at least partially, by metabolite alterations. Participants with GE had low PFOA levels and high embryo quality, while participants with PE had high PFOA levels and low embryo quality. High-quality embryo rates were positively correlated with Pro-Trp, phytosphingosine, N -myristoylsphinganine, N -lauroyl-d-erythro-sphinganine, and lauryldimethylamine oxide, but negatively correlated with benzyl alcohol and 9cl-pf3ons. Phytosphingosine, N -myristoylsphinganine, and N -lauroyl-d-erythro-sphinganine are sphingolipid metabolites, while Pro-Trp and lauryldimethylamine oxide are organonitrogens. India-Aldana et al. 48 showed that the most frequently reported associations across studies were observed between PFAS and amino acids, fatty acids, glycerophospholipids, glycerolipids, phosphosphingolipids, bile acids, ceramides, purines, and acylcarnitines. Sphingolipid metabolism is vital for the maintenance of normal follicle and oocyte maturation, as it regulates steroid hormone biosynthesis, cell proliferation, energy metabolism, and apoptosis. 49 – 51 Sphingolipid metabolites, such as sphinganine, sphingosine, and ceramide, may act as second messengers or regulatory factors in signal transduction, thereby modulating apoptosis and cell cycle regulation. 52 Biogenesis of the abundant microstructural domains in sphingolipids enables cells to localize the molecular machinery involved in membrane-initiated cellular functions. 52 The dipeptide Pro-Trp is composed of proline and tryptophan. Proline-rich tyrosine kinase 2 is upregulated in oocytes and plays essential roles in mouse oocyte fertilization and early embryo development. 53 Bahrami et al. 54 reported that culture media containing only glutamine, proline, and isoleucine were sufficient for bovine oocyte maturation and fertilization. Lauryldimethylamine oxide significantly increases the ATPase activity of ATP synthase, catalyzes ATP biosynthesis from ADP, and may, therefore, help maintain normal cellular metabolism and energy status. 55 The present study had certain strengths. First, this study examined a large cohort consisting of 378 women with infertility undergoing IVF/ICSI treatment, and a broad PFAS panel was detected in the follicle microenvironment. This allowed detailed investigation of the effects of PFAS exposure on embryo quality and clinical pregnancy outcomes. Second, both the BKMR model and quantile-based g-computation model were used to evaluate the impact of PFAS mixtures on embryo quality and to identify the main PFAS contributing to the exposure–response relationship. Third, to the best of our knowledge, the present study is one of the first to conduct untargeted metabolomics on follicle fluid and to identify specific metabolites that could associate the toxic effects of PFOA on embryo quality. The current study also had certain limitations. This research collected samples from humans; thus, it was not possible to measure PFAS and metabolites at different time points to investigate the causal mediation assumptions. Therefore, further animal research is required to validate this association and elucidate the mechanisms involved. Additionally, the untargeted metabolomics profiling in the association between PFOA and high embryo quality were performed in a small sample ( n = 50 ). This study investigated the relationship between PFAS exposure in follicular fluid and embryo quality in women with infertility, which may introduce a population selection bias. Consequently, further research is necessary to ascertain whether this association is applicable to the general population. Lastly, since PFAS in follicular fluid after ovarian stimulation primarily originates from blood, the metabolic profile and PFAS concentrations in follicular fluid in this study may not be specific to the ovary alone. However, compared to previous studies that analyzed PFAS levels exclusively in blood, our study offers a more organ-targeted approach. This allows for a more accurate reflection of local microenvironmental changes within the ovary associated with fertility decline, providing stronger and more direct evidence for etiologic exploration. In conclusion, this study revealed that PFAS in follicular fluid exhibited negative correlations with embryo quality. Among all of the PFAS compounds detected and analyzed, PFOA showed the most pronounced association with the embryo quality. The organonitrogens (Pro-Trp and lauryldimethylamine oxide) and the sphingolipid metabolites (Phytosphingosine, N -myristoyl-D-sphinganine, and N -lauroyl-d-erythro-sphinganine) in follicular fluid maintain normal cellular growth and metabolism, and their reduction may contribute to PFOA-related poor embryo quality. The findings made herein will help clarify certain environmental risk factors related to unfavorable outcomes of assisted reproductive technology (ART) and lay theoretical and practical foundations upon which regulatory bodies and policymakers can develop and implement measures restricting the industrial application of PFAS.

Infertility is a common reproductive condition currently defined as unsuccessful pregnancy after over 1 year of normal sexual activity in the absence of any contraceptive measures. 1 About 15% of all couples of childbearing age worldwide exhibit infertility. 2 Women are relatively more prone to infertility than men. 3 An unhealthy lifestyle, psychosocial stress, and environmental contaminants can increase the risk of infertility. 4 Exposure to ubiquitous environmental contaminants, such as classes of endocrine disruptors, harms animal and human reproductive health and has been associated with infertility. 5 Perfluoroalkyl and polyfluoroalkyl substances (PFAS) currently include over 10,000 different synthetic fluorinated chemicals. 6 Most PFAS resist biodegradation, and their average estimated half-lives have been observed to last from a few years to decades. 7 The half-lives of perfluorooctanoic acid (PFOA) and perfluorooctane sulfonate (PFOS) were estimated to be 6.5 and 5.0 years, respectively. 8 Most PFAS can be regionalized and transported over long distances. 9 Hence, they are globally distributed pollutants to which many humans may be exposed. 10 The main PFAS exposure pathways are inhalation or ingestion of contaminated drinking water and foods (such as food packaging material, nonstick cookware, and seafood). 11 Exposure to high PFAS levels is closely associated with premature ovarian failure, menstrual cycle disorders, and early menopause. 12 – 14 Animal experiments showed that PFAS impair follicular development and ovulation in pigs, mice, cattle, and other mammals. 15 – 17 Follicular fluid originates mainly from plasma in the vascular compartment of the follicle wall. This complex extracellular fluid accumulates in the antrum of ovarian follicles during follicular growth. 18 It furnishes a microenvironment for the maturation of the cumulus-oocyte complex and the differentiation of granulosa cells, plays vital roles in oocyte development, and directly influences oocyte quality. 18 A variety of metabolites in follicle fluid are essential for oocyte growth. 19 PFAS may accumulate in follicular fluid and can cross the blood–follicle barrier. 20 However, it is unknown which specific PFAS in follicular fluid negatively impact oocyte development and embryo quality or whether any of the metabolites present in the follicular fluid contribute to these adverse effects. The present study collected follicular fluid from 378 women with infertility undergoing in vitro fertilization (IVF) or intracytoplasmic sperm injection (ICSI), subjected the samples to PFAS measurement and untargeted metabolomics, and conducted correlation analyses on PFAS, embryo quality, and clinical pregnancy outcomes. We hypothesized that PFAS may negatively affect oocyte development and embryo quality, and specific metabolites in the follicular fluid may be involved in this effect.