A metabolomic investigation of serum perfluorooctane sulfonate and perfluorooctanoate

Abstract

Background:

Exposures to perfluorooctane sulfonate (PFOS) and perfluorooctanoate (PFOA), environmentally persistent chemicals detectable in the blood of most Americans, have been associated with several health outcomes. To offer insight into their possible biologic effects, we evaluated the metabolomic correlates of circulating PFOS and PFOA among 3,647 participants in eight nested case-control serum metabolomic profiling studies from the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial.

Methods:

Metabolomic profiling was conducted by Metabolon Inc., using ultra high-performance liquid chromatography/tandem accurate mass spectrometry. We conducted study-specific multivariable linear regression analyses estimating the associations of metabolite levels with levels of PFOS or PFOA. For metabolites measured in at least 3 of 8 nested case-control studies, random effects meta-analysis was used to summarize study-specific results (1,038 metabolites in PFOS analyses and 1,100 in PFOA analyses).

Results:

The meta-analysis identified 51 and 38 metabolites associated with PFOS and PFOA, respectively, at a Bonferroni-corrected significance level (4.8x10⁻⁵ and 4.6x10⁻⁵, respectively). For both PFOS and PFOA, the most common types of associated metabolites were lipids (sphingolipids, fatty acid metabolites) and xenobiotics (xanthine metabolites, chemicals). Positive associations were commonly observed with lipid metabolites sphingomyelin (d18:1/18:0) (P=2.0x10⁻¹⁰ and 2.0x10⁻⁸, respectively), 3-carboxy-4-methyl-5-pentyl-2-furanpropionate (P=2.7x10⁻¹⁵, 1.1x10⁻¹⁷), and lignoceroylcarnitine (C24) (P=2.6x10⁻⁸, 6.2x10⁻⁶). The strongest positive associations were observed for chemicals 3,5-dichloro-2,6-dihydroxybenzoic acid (P=3.0x10⁻¹¹² and 6.8x10⁻¹³, respectively) and 3-bromo-5-chloro-2,6-dihydroxybenzoic acid (P=1.6x10⁻¹⁴, 2.3x10⁻⁶). Other metabolites positively associated with PFOS included D-glucose (carbohydrate), carotene diol (vitamin A metabolism), and L-alpha-aminobutyric acid (glutathione metabolism), while uric acid (purine metabolite) was positively associated with PFOA. PFOS associations were consistent even after adjusting for PFOA as a covariate, while PFOA associations were greatly attenuated with PFOS adjustment.

Conclusions:

In this large metabolomic study, we observed robust positive associations with PFOS for several molecules. Further investigation of these metabolites may offer insight into PFOS-related biologic effects.

1. Introduction

Per- and polyfluoroalkyl substances (PFAS) comprise a subclass of halogenated environmentally persistent organic chemicals that have been used commercially since the 1950s.(1) Historically, PFAS were commonly used in producing non-stick cookware, food packaging, textiles, and firefighting foams. Exposure to PFAS is widespread in the U.S., with serum concentrations detected in over 98% of the general population in the early 2000s; major sources of exposure include consumption of contaminated food and drinking water.(1) Of the over 9,000 identified PFAS, evidence regarding exposures, toxicity, and human health effects is available for only a small number of legacy PFAS chemicals, most notably perfluorooctane sulfonate (PFOS) and perfluorooctanoate (PFOA).(2, 3) These chemicals do not undergo metabolism in humans and have half-lives of serum elimination that are estimated to be between 2 and 4 years.(4, 5) Health effects that have been associated with PFOS and/or PFOA exposures in epidemiologic studies include altered immune and thyroid function, liver disease, lipid and insulin dysregulation, kidney disease, adverse reproductive and developmental outcomes, and some cancers (e.g., kidney and testis). (4, 6) However, it is unclear what mechanisms underlie these associations because the biologic effects of these chemicals in humans remain poorly understood.

Epidemiologic studies investigating biomarker associations with PFAS exposures have mainly focused on targeted investigations of specific analytes such as lipids, thyroid hormones, or biomarkers of liver function.(7) Important new insights may come from the use of semi-targeted metabolomic analyses capable of measuring thousands of small low-molecular-weight metabolites, produced by chemical processes and exogenous sources, in biological samples.(8) A recent systemic review including 11 metabolomic investigations of PFOS and/or PFOA identified reported associations with lipids and lipid-like metabolites (e.g., glycerophosphocholines, fatty acids) as well as amino acid-related metabolites (e.g., pyroglutamic acid, glutamate/glutamine).(7, 9-11) However, these investigations had limitations in study power, with sample sizes ranging from 74 to 965 participants.(7) To address this knowledge gap, we conducted a serum metabolomics study of circulating PFOS and PFOA using measurement data from 3,647 participants included in one of eight nested case-control studies conducted within the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO). All samples were analyzed using the same metabolomic platform at the same laboratory.

2. Material and Methods

2.1. Study participants

The PLCO trial enrolled 155,000 men and women randomized to either the intervention or usual-care arm at ten screening centers across the United States from 1993 to 2001.(12) Eligible participants were 55-74 years of age at baseline. The primary exclusion criteria included history of a prostate, lung, colorectal, or ovarian cancer, and current cancer treatment (except basal or squamous cell skin cancer). Non-fasting blood specimens were collected from participants randomized to the PLCO screening arm. Specimens were processed within two hours of collection and stored at −70C. All participants completed a risk factor questionnaire at baseline, and were followed up for incident cancer diagnoses through annual updates, participant reports, medical record abstraction, and death certificates.

Participants in our analysis were included in one of eight nested case-control studies investigating serum metabolomic data in relation to selected cancers (typically diagnosed several years after blood collection) or other endpoints (13-15); the designs of these studies are summarized in Supplementary Table 1. After excluding 35 subjects with missing data on baseline body mass index (BMI) and/or smoking status (n=35), the pooled analysis included 3,647 participants, 1,818 of whom had been originally selected as study cases and 1,829 as controls. Institutional review boards of the National Cancer Institute (NCI) and the 10 study centers approved the trial, and all participants provided written informed consent in accordance with the Declaration of Helsinki.

2.2. Metabolite assessment

All metabolomic profiling was conducted by Metabolon Inc. according to methods which have been described previously.(16) Metabolon platforms used in each study are described in Supplementary Table 1. Briefly, a single non-targeted extraction with methanol was used, followed by protein precipitation, to recover a diverse set of metabolites. Serum samples were analyzed using ultra high-performance liquid chromatography/tandem accurate mass spectrometry (UHPLC-MS/MS) in all studies and one study (esophageal cancer study) used GC-MS together with UHPLC-MS/MS. The mass spectra were compared to a chemical reference library generated from >3,300 standards to identify individual metabolites (>2,400 standards for the esophageal cancer study).

In this investigation, we restricted to 1,409 metabolites identified by the chemical reference library, including the linear isomers of PFOS and PFOA (n-PFOs and n-PFOA, respectively). The metabolites were grouped into eight chemical classes by Metabolon Inc. (amino acids, carbohydrates, cofactors and vitamins, energy metabolites, lipids, nucleotide metabolites, peptides, and xenobiotics) based on the classifications of the Kyoto Encyclopedia of Genes and Genomes.(17)

To account for measurement variability for studies involving samples run over multiple days, within each study we rescaled the detectable values to set the median equal to 1, transformed metabolite values to their natural logarithm and rescaled them to a standard normal distribution (mean=0, standard deviation=1). For each metabolite, values below the limit of detection (LOD) were imputed using the lowest detected values.

To establish the validity of semi-quantitative PFOS and PFOA quantitation using the Metabolon platform, we calculated Spearman correlations comparing Metabolon-measured PFOS and PFOA serum levels to both linear and total PFOS and PFOA measured by the Centers for Disease Control and Prevention (CDC, Atlanta, GA) using online solid phase extraction liquid chromatography isotope dilution tandem mass spectrometry among 617 participants (282 prostate cancer cases, 335 controls) with sera analyzed at both laboratories.(18) We found that the Metabolon-measured PFOS and PFOA were strongly correlated with the CDC measurements (n-PFOS, rho=0.74; total PFOS, rho=0.76; n-PFOA, rho=0.77; total PFOA, rho=0.76; Figure 1).

Figure 1.

Scatter plots of serum PFAS measured at Metabolon and CDC. Spearman correlations are 0.77 for n-PFOA and 0.74 for n-PFOS.

PFAS, per- and polyfluoroalkyl substances; PFOA, perfluorooctanoate; PFOS, perfluorooctane sulfonate; CDC, Centers for Disease Control and Prevention Serum PFOA and PFOS are for the linear isomer. All measurements were log-transformed.

2.3. Statistical analysis

As a preliminary analysis, we explored associations between the Metabolon PFAS measurements and selected demographic and lifestyle factors using simple linear regression models. As the main analysis, we first conducted multiple linear regression modeling to estimate study-specific associations between metabolite levels and n-PFOS (or n-PFOA) as a continuous dependent variable, adjusted for age at specimen collection (continuous), race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, Asian, Pacific Islander, American Indian), BMI (continuous), smoking status (never, former, current), case-control status, study center, year of specimen collection (continuous) and sex (for studies that included both men and women). We then combined the study-specific findings by meta-analysis using DerSimonian and Laird random effects models.(19) As each study did not measure the entire list of 1,409 metabolites, meta-analyses were conducted for metabolites assessed in three or more studies (1,038 metabolites for n-PFOS analysis, 1,100 metabolites for n-PFOA. Note: one of the studies did not measure n-PFOS; Supplementary Table 1). We used a Bonferroni-corrected significance threshold (i.e., alpha=0.05/1,038=4.82 x 10⁻⁵ for n-PFOS; alpha=0.05/1,100=4.55 x 10⁻⁵ for n-PFOA) to account for multiple testing. To examine n-PFOS-metabolite associations independent of n-PFOA and vice versa, we ran models additionally adjusting for the other PFAS as a continuous covariate (Pearson correlation coefficient for n-PFOS and n-PFOA=0.32). As a sensitivity analysis, we further adjusted for metabolites associated with lower estimated glomerular filtration rate (eGFR) [creatinine, pseudouridine, N-acetylalanine, erythronate, myo-inositol, and glutarylcarnitine (20, 21)] to attempt to indirectly control for kidney function in the association with PFAS. We additionally adjusted for education attainment (less than high school, high school graduate or some college, college graduate or post-graduate) as one of socioeconomic factors. Given that BMI might play a role as a mediator in the association between PFAS and health outcomes, we also repeated analyses excluding BMI as a covariate. Lastly, we repeated analyses stratified by case status and sex. We evaluated the extent of heterogeneity in PFAS-metabolite associations across studies using the Q statistic.(22). All analyses were performed with SAS software version 9.4 (SAS Institute, Cary, NC).

3. Results

Selected characteristics of the participants are presented in Supplementary Table 2. The median age at specimen collection was 63 years (interquartile range=60, 67 years), half of the participants were women (51%), and the majority were non-Hispanic White (81%), ever smokers (53%), and had BMI ≥25 kg/m² (71%). Among female participants, 99.7% of them were post-menopausal. Serum n-PFOS levels were significantly elevated in relation to male sex, non-Hispanic Black race/ethnicity, higher BMI and smoking, while n-PFOA levels were significantly lower in relation to older age at serum collection and for parous vs non-parous women. Previous investigations within PLCO reported that absolute serum concentrations of PFOS and PFOA are comparable to those observed in the general U.S. population. (23, 24)

The complete meta-analysis results for n-PFOS and n-PFOA are summarized in the Supplementary Data. We identified 51 and 38 metabolites associated with n-PFOS and n-PFOA, respectively, at the Bonferroni-corrected significance threshold (Tables 1 and 2). The Pearson correlations between these metabolites are shown in Supplementary Data.

Table 1.

Association of serum metabolites with PFOS in a cross-sectional analysis of 3,647 participants from eight nested case-control serum metabolomic studies from the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (metabolites ordered by chemical class and pathway)

			All Subjects			Adjustment for PFOAa
Metabolite (HMDB ID)	Chemical class	Pathway	Overall effect estimate (β) (95% CI)	P-valueb	P heterogeneity by study	Overall effect estimate (β) (95% CI)	P-value	P heterogeneity by study
L-Alpha-aminobutyric acid (HMDB00650)	Amino Acid	Glutathione Metabolism	0.16 (0.09,0.23)	2.24E-06	0.51	0.06 (0.01,0.11)	0.02	0.82
2-Ketobutyric acid (HMDB00005)	Amino Acid	Methionine, Cysteine, SAM and Taurine Metabolism	0.06 (0.04,0.08)	1.95E-09	0.74	0.04 (0.02,0.05)	4.92E-06	0.42
Indoxyl sulfate (HMDB00682)	Amino Acid	Tryptophan Metabolism	0.08 (0.04,0.11)	4.00E-05	0.76	0.05 (0.02,0.07)	9.89E-04	0.45
D-Glucose (HMDB00122)	Carbohydrate	Glycolysis, Gluconeogenesis, and Pyruvate Metabolism	0.27 (0.15,0.39)	1.62E-05	0.14	0.17 (0.11,0.24)	3.22E-07	0.51
L-Urobilin (HMDB04159)	Cofactors and Vitamins	Hemoglobin and Porphyrin Metabolism	0.05 (0.03,0.06)	5.87E-11	0.5	0.03 (0.02,0.04)	6.10E-07	0.39
Trigonelline (HMDB00875)	Cofactors and Vitamins	Nicotinate and Nicotinamide Metabolism	0.04 (0.02,0.06)	7.06E-06	0.67	0.01 (0,0.02)	0.21	0.87
Gamma-CEHC (HMDB01931)	Cofactors and Vitamins	Tocopherol Metabolism	0.09 (0.05,0.12)	5.48E-07	0.44	0.05 (0.02,0.08)	0.001	0.25
Gamma-Tocopherol	Cofactors and Vitamins	Tocopherol Metabolism	0.07 (0.04,0.11)	4.53E-05	0.38	0.06 (0.01,0.11)	0.01	0.02
Carotene diol (3)	Cofactors and Vitamins	Vitamin A Metabolism	0.22 (0.12,0.31)	1.60E-05	0.22	0.12 (0.06,0.18)	5.67E-05	0.88
Lignoceroylcarnitine (C24)	Lipid	Fatty Acid Metabolism	0.23 (0.15,0.32)	2.64E-08	0.41	0.11 (0.02,0.2)	0.02	0.09
N-Acetylaminooctanoic acid (HMDB59745)	Lipid	Fatty Acid, Amino	0.08 (0.05,0.12)	5.36E-06	0.85	0.04 (0.02,0.07)	0.002	0.97
Eicosenedioate (C20:1-DC)	Lipid	Fatty Acid, Dicarboxylate	0.16 (0.1,0.22)	8.30E-07	0.03	0.06 (0.03,0.09)	5.24E-05	0.38
3-carboxy-4-methyl-5-pentyl-2-furanpropionate (3-CMPFP)	Lipid	Fatty Acid, Dicarboxylate	0.26 (0.19,0.32)	2.67E-15	0.2	0.12 (0.08,0.16)	1.01E-09	0.41
Eicosanodioate	Lipid	Fatty Acid, Dicarboxylate	0.18 (0.12,0.24)	3.20E-09	0.33	0.09 (0.05,0.13)	2.41E-05	0.76
(R)-2-Hydroxycaprylic acid (HMDB02264)	Lipid	Fatty Acid, Monohydroxy	0.13 (0.09,0.17)	6.64E-10	0.81	0.06 (0.03,0.09)	5.04E-04	0.78
Hydroxypalmitoyl SM (d18:1/16:0(OH))	Lipid	Sphingomyelins	0.44 (0.24,0.64)	1.26E-05	0.15	0.30 (0.19,0.41)	7.83E-08	0.37
SM (d18:1/18:0) (HMDB01348)	Lipid	Sphingomyelins	0.35 (0.24,0.46)	1.96E-10	0.39	0.14 (0.06,0.22)	3.56E-04	0.46
SM (d18:2/24:2)	Lipid	Sphingomyelins	0.32 (0.17,0.47)	1.92E-05	0.7	0.14 (0.02,0.27)	0.03	0.26
Lactosylceramide (d18:1/24:1(15Z))	Lipid	Lactosylceramides	0.41 (0.29,0.54)	1.45E-10	0.74	0.20 (0.05,0.36)	0.009	0.07
1-Oleoylglycerophosphoinositol	Lipid	Lysophospholipid	0.13 (0.07,0.19)	1.09E-05	0.03	0.06 (0.03,0.09)	2.20E-04	0.25
LysoPE(18:2(9Z,12Z)/0:0) (HMDB11507)	Lipid	Lysophospholipid	0.11 (0.06,0.16)	9.98E-06	0.9	0.08 (0.05,0.12)	1.52E-05	0.63
1-(1-enyl-stearoyl)-GPE (P-18:0) or 1-stearoylplasmenylethanolamine	Lipid	Lysoplasmalogen	0.20 (0.12,0.29)	4.06E-06	0.04	0.13 (0.07,0.18)	2.71E-06	0.15
Tauro-b-muricholic acid (HMDB00932)	Lipid	Primary Bile Acid Metabolism	0.03 (0.02,0.05)	3.99E-05	0.96	0.03 (0.02,0.04)	1.15E-07	0.66
Deoxycholic acid (HMDB00626)	Lipid	Secondary Bile Acid Metabolism	0.07 (0.05,0.1)	1.65E-07	0.37	0.04 (0.02,0.06)	3.01E-05	0.51
Deoxycholic acid glycine conjugate (HMDB00631)	Lipid	Secondary Bile Acid Metabolism	0.03 (0.01,0.04)	2.51E-05	0.5	0.02 (0.01,0.03)	3.50E-04	0.48
Hyocholic acid (HMDB00760)	Lipid	Secondary Bile Acid Metabolism	0.08 (0.05,0.11)	2.27E-07	0.23	0.05 (0.03,0.07)	3.86E-05	0.22
Lithocholic acid sulfate (1)	Lipid	Secondary Bile Acid Metabolism	0.06 (0.03,0.08)	2.41E-06	0.14	0.03 (0.02,0.05)	8.19E-05	0.18
Taurodeoxycholic acid (HMDB00896)	Lipid	Secondary Bile Acid Metabolism	0.03 (0.02,0.04)	9.86E-06	0.94	0.02 (0.01,0.03)	3.10E-04	0.61
Taurolithocholic acid 3-sulfate (HMDB02580)	Lipid	Secondary Bile Acid Metabolism	0.07 (0.04,0.09)	2.28E-06	0.12	0.04 (0.03,0.06)	6.91E-09	0.48
Guanosine (HMDB00133)	Nucleotide	Purine Metabolism, Guanine containing	0.04 (0.02,0.05)	1.16E-05	0.87	0.02 (0.01,0.03)	0.002	0.74
3-methylcytidine	Nucleotide	Pyrimidine Metabolism, Cytidine containing	0.16 (0.1,0.22)	2.12E-07	0.34	0.05 (0.01,0.09)	0.02	0.67
3-methyl catechol sulfate (2)	Xenobiotics	Benzoate Metabolism	0.03 (0.02,0.04)	6.49E-06	0.41	0.01 (0,0.02)	0.05	0.91
p-Cresol sulfate	Xenobiotics	Benzoate Metabolism	0.04 (0.02,0.05)	3.02E-05	0.92	0.02 (0,0.03)	0.02	0.96
3,5-dichloro-2,6-dihydroxybenzoic acid	Xenobiotics	Chemical	0.74 (0.67,0.8)	3.01E-112	0.58	0.42 (0.34,0.5)	2.56E-24	0.13
3-bromo-5-chloro-2,6-dihydroxybenzoic acid	Xenobiotics	Chemical	0.57 (0.43,0.72)	1.60E-14	0.01	0.35 (0.23,0.47)	7.35E-09	0.03
4-hydroxychlorothalonil	Xenobiotics	Chemical	0.39 (0.3,0.49)	2.22E-16	0.00002	0.24 (0.2,0.28)	3.39E-27	0.1
Thioproline	Xenobiotics	Chemical	0.19 (0.11,0.28)	5.42E-06	0.33	0.14 (0.06,0.21)	0.0003	0.16
3D,7D,11D-Phytanic acid (HMDB00801)	Xenobiotics	Food Component/Plant	0.14 (0.07,0.2)	2.18E-05	0.66	0.06 (−0.01,0.13)	0.07	0.1
3-indoleglyoxylic acid	Xenobiotics	Food Component/Plant	0.22 (0.13,0.32)	3.73E-06	0.56	0.12 (0.04,0.19)	0.002	0.81
Dihydrocaffeate sulfate (2)	Xenobiotics	Food Component/Plant	0.03 (0.02,0.04)	1.39E-05	0.57	0.01 (0,0.02)	0.05	0.26
Glucuronide of piperine metabolite C17H21NO3 (3)*	Xenobiotics	Food Component/Plant	0.03 (0.02,0.05)	7.69E-06	0.45	0.02 (0.01,0.03)	0.006	0.35
Quinic acid (HMDB03072)	Xenobiotics	Food Component/Plant	0.03 (0.02,0.04)	2.65E-06	0.46	0.01 (0,0.02)	0.09	0.84
Sulfate of piperine metabolite C16H19NO3 (3)*	Xenobiotics	Food Component/Plant	0.06 (0.03,0.08)	3.23E-05	0.09	0.04 (0.02,0.06)	6.46E-05	0.08
Sulfate of piperine metabolite C18H21NO3 (1)*	Xenobiotics	Food Component/Plant	0.05 (0.03,0.06)	1.23E-10	0.9	0.04 (0.02,0.05)	3.26E-06	0.32
1,3,7-Trimethyluric acid (HMDB02123)	Xenobiotics	Xanthine Metabolism	0.03 (0.02,0.04)	1.05E-07	0.74	0.01 (0,0.02)	0.04	0.9
1,3-Dimethyluric acid (HMDB01857)	Xenobiotics	Xanthine Metabolism	0.03 (0.02,0.05)	2.88E-07	0.7	0.01 (0,0.02)	0.006	0.95
1,7-Dimethyluric acid (HMDB11103)	Xenobiotics	Xanthine Metabolism	0.04 (0.02,0.05)	1.87E-07	0.58	0.01 (0,0.02)	0.02	0.56
1-Methylxanthine (HMDB10738)	Xenobiotics	Xanthine Metabolism	0.04 (0.03,0.06)	1.98E-07	0.73	0.01 (0,0.03)	0.03	0.74
Caffeine (HMDB01847)	Xenobiotics	Xanthine Metabolism	0.03 (0.02,0.05)	2.56E-07	0.75	0.01 (0,0.02)	0.05	0.73
Paraxanthine (HMDB01860)	Xenobiotics	Xanthine Metabolism	0.04 (0.02,0.05)	1.84E-07	0.7	0.01 (0,0.02)	0.04	0.56
Theophylline (HMDB01889)	Xenobiotics	Xanthine Metabolism	0.04 (0.02,0.05)	1.61E-07	0.66	0.01 (0,0.02)	0.04	0.61

PFOS, perfluorooctane sulfonate; PFOA, perfluorooctanoate; HMDB, Human Metabolome Database; SM, sphingomyelin

Results are based on a population of 3,647 participants [prostate cancer among non-Hispanic Black men (cases=220, controls=219), prostate cancer (cases=356, controls=372), esophageal cancer (cases=131, controls=127), glioma cancer (cases=161, controls=163), breast cancer (cases=604, controls=605), pancreatic cancer (cases=107, controls=103), endometrial cancer (cases=190, controls=191), and multivitamin exposure study (multivitamin users=49, non-users=49)].

Overall effect estimate (β) indicates the association between metabolite levels and PFOS was modeled using linear regression, adjusted for age at specimen collection, race/ethnicity, body mass index, smoking status, case status, study center, year of specimen collection. Sex was additionally adjusted for studies of pancreatic cancer, glioma, and multivitamin study. The combined model was conducted using random effects meta-analysis.

^aPFOA was further adjusted in linear regression models.

^bOnly the metabolites that met the Bonferroni corrected threshold of statistical significance in the combined models (0.05/1,038 = 4.82x10⁻⁵) are shown here.

Table 2.

Association of serum metabolites with PFOA in a cross-sectional analysis of 3,647 participants from eight nested case-control serum metabolomic studies from the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (metabolites ordered by chemical class and pathway)

			All Subjects			Adjustment with PFOSa
Metabolite (HMDB ID)	Chemical class	Sub-pathway	Overall effect estimate (β) (95% CI)	P-value	P heterogeneity by study	Overall effect estimate (β) (95% CI)	P-value	P heterogeneity by study
Cysteine-S-sulfate (HMDB00731)	Amino Acid	Methionine, Cysteine, SAM and Taurine Metabolism	0.07 (0.04,0.09)	5.18E-07	0.41	0.06 (0.04,0.09)	6.52E-06	0.12
1,5-Anhydrosorbitol (HMDB02712)	Carbohydrate	Glycolysis, Gluconeogenesis, and Pyruvate Metabolism	0.07 (0.05,0.1)	8.00E-08	0.50	0.05 (0.02,0.07)	2.43E-05	0.64
Trigonelline (HMDB00875)	Cofactors and Vitamins	Nicotinate and Nicotinamide Metabolism	0.04 (0.02,0.05)	3.94E-08	0.53	0.02 (0.01,0.03)	4.84E-04	0.76
4-androsten-3alpha,17alpha-diol monosulfate (2) or Androstenediol (3alpha, 17alpha) monsulfate (2)	Lipid	Androgenic Steroids	0.03 (0.02,0.04)	3.13E-06	0.53	0.02 (0,0.03)	0.01	0.41
5alpha-Androstan-3alpha,17beta-diol disulfate	Lipid	Androgenic Steroids	0.05 (0.03,0.07)	2.38E-05	0.02	0.02 (0.01,0.03)	0.003	0.26
SM(d18:0/18:0) (HMDB12087)	Lipid	Dihydrosphingomyelins	0.20 (0.12,0.27)	8.61E-08	0.67	0.08 (0.02,0.13)	0.005	0.89
SM(d18:0/20:0)	Lipid	Dihydrosphingomyelins	0.21 (0.15,0.28)	4.13E-10	0.61	0.08 (0.03,0.13)	0.002	0.72
Lignoceroylcarnitine (C24)	Lipid	Fatty Acid Metabolism	0.16 (0.09,0.23)	6.23E-06	0.58	0.03 (−0.03,0.09)	0.38	0.25
Hexanoylcarnitine (HMDB00705)	Lipid	Fatty Acid Metabolism (Acyl Carnitine, Medium Chain)	0.07 (0.04,0.11)	3.47E-05	0.29	0.04 (0.02,0.07)	2.91E-04	0.46
9-Decenoylcarnitine	Lipid	Fatty Acid Metabolism (Acyl Carnitine, Monounsaturat	0.09 (0.05,0.13)	2.98E-05	0.11	0.04 (0.02,0.06)	1.57E-04	0.44
3-carboxy-4-methyl-5-pentyl-2-furanpropionate (3-CMPFP)	Lipid	Fatty Acid, Dicarboxylate	0.16 (0.12,0.2)	1.11E-17	0.42	0.04 (0.01,0.07)	0.01	0.81
Eicosanodioate	Lipid	Fatty Acid, Dicarboxylate	0.12 (0.07,0.16)	8.66E-07	0.37	0.02 (−0.02,0.05)	0.37	0.33
Maleic acid (HMDB00176)	Lipid	Fatty Acid, Dicarboxylate	0.05 (0.03,0.08)	4.81E-06	0.92	0.03 (0.01,0.04)	0.004	0.89
(R)-2-Hydroxycaprylic acid (HMDB02264)	Lipid	Fatty Acid, Monohydroxy	0.1 (0.06,0.13)	8.36E-07	0.28	0.03 (0,0.06)	0.08	0.20
SM(d18:1/18:0) (HMDB01348)	Lipid	Sphingomyelins	0.25 (0.16,0.34)	2.01E-08	0.49	0.06 (−0.01,0.12)	0.09	0.45
SM(d18:1/18:1(9Z)) (HMDB12101)	Lipid	Sphingomyelins	0.24 (0.16,0.32)	1.10E-09	0.35	0.12 (0.06,0.17)	1.60E-05	0.93
Glycosyl-N-behenoyl-sphingadienine (d18:2/22:0)*	Lipid	Hexosylceramides (HCER)	0.26 (0.14,0.37)	9.62E-06	0.24	0.06 (−0.02,0.13)	0.13	0.88
Lactosylceramide (d18:1/24:1(15Z))	Lipid	Lactosylceramides (LCER)	0.22 (0.12,0.32)	0.00003	0.71	0.01 (−0.06,0.09)	0.72	0.52
1-linoleoylglycerophosphoinositol* or 1-linoleoyl-GPI (18:2)*	Lipid	Lysophospholipid	0.13 (0.07,0.19)	6.48E-06	0.04	0.04 (0,0.08)	0.04	0.15
3 beta-Hydroxy-5-cholestenoate	Lipid	Sterol	0.24 (0.14,0.33)	4.74E-07	0.15	0.08 (0.02,0.13)	0.004	0.38
Uric acid (HMDB00289)	Nucleotide	Purine Metabolism, (Hypo)Xanthine/Inosine containing	0.28 (0.16,0.41)	1.23E-05	0.02	0.12 (0.05,0.19)	0.0009	0.25
3-methylcytidine	Nucleotide	Pyrimidine Metabolism, Cytidine containing	0.15 (0.08,0.22)	3.93E-05	0.09	0.05 (0,0.1)	0.04	0.13
Leucyl-Alanine (HMDB28922)	Peptide	Dipeptide	0.07 (0.04,0.11)	4.31E-05	0.33	0.05 (0.02,0.07)	2.25E-04	0.96
O-methoxycatechol-O-sulphate (HMDB60013)	Xenobiotics	Benzoate Metabolism	0.05 (0.03,0.07)	4.38E-06	0.42	0.02 (0,0.04)	0.02	0.27
3,5-dichloro-2,6-dihydroxybenzoic acid	Xenobiotics	Chemical	0.48 (0.35,0.61)	6.82E-13	0.003	0.06 (−0.16,0.28)	0.6	9.09E-10
3-bromo-5-chloro-2,6-dihydroxybenzoic acid	Xenobiotics	Chemical	0.39 (0.23,0.55)	2.29E-06	0.003	0.08 (−0.21,0.36)	0.6	9.63E-10
4-hydroxychlorothalonil	Xenobiotics	Chemical	0.21 (0.13,0.3)	1.26E-06	1.18E-06	0.00 (−0.06,0.06)	0.96	0.0005
Quinic acid (HMDB03072)	Xenobiotics	Food Component/Plant	0.03 (0.01,0.04)	6.88E-06	0.32	0.01 (0,0.02)	0.002	0.63
1,3,7-Trimethyluric acid (HMDB02123)	Xenobiotics	Xanthine Metabolism	0.03 (0.02,0.04)	7.40E-10	0.70	0.01 (0.01,0.02)	0.001	0.60
1,3-Dimethyluric acid (HMDB01857)	Xenobiotics	Xanthine Metabolism	0.02 (0.01,0.03)	5.23E-07	0.66	0.01 (0,0.02)	0.03	0.76
1,7-Dimethyluric acid (HMDB11103)	Xenobiotics	Xanthine Metabolism	0.03 (0.02,0.04)	1.32E-09	0.86	0.01 (0,0.02)	0.003	0.43
1-Methyluric acid (HMDB03099)	Xenobiotics	Xanthine Metabolism	0.04 (0.02,0.05)	1.37E-06	0.36	0.01 (0,0.02)	0.01	0.71
1-Methylxanthine (HMDB10738)	Xenobiotics	Xanthine Metabolism	0.04 (0.02,0.05)	2.31E-09	0.82	0.01 (0.01,0.02)	0.002	0.49
5-Acetylamino-6-amino-3-methyluracil (HMDB04400)	Xenobiotics	Xanthine Metabolism	0.02 (0.01,0.03)	3.71E-08	0.52	0.01 (0,0.02)	0.01	0.61
Caffeine (HMDB01847)	Xenobiotics	Xanthine Metabolism	0.03 (0.02,0.04)	3.51E-10	0.63	0.01 (0,0.02)	0.02	0.16
Paraxanthine (HMDB01860)	Xenobiotics	Xanthine Metabolism	0.03 (0.02,0.05)	1.52E-10	0.86	0.02 (0,0.03)	0.01	0.29
Theobromine (HMDB02825)	Xenobiotics	Xanthine Metabolism	0.03 (0.02,0.04)	2.71E-05	0.60	0.01 (0,0.03)	0.11	0.15
Theophylline (HMDB01889)	Xenobiotics	Xanthine Metabolism	0.03 (0.02,0.05)	1.96E-10	0.96	0.02 (0.01,0.02)	0.001	0.49

PFOA, perfluorooctanoate; PFOS, perfluorooctane sulfonate; HMDB, Human Metabolome Database

Results are based on a population of 3,647 participants [prostate cancer among non-Hispanic Black men (cases=220, controls=219), prostate cancer (cases=356, controls=372), esophageal cancer (cases=131, controls=127), glioma cancer (cases=161, controls=163), breast cancer (cases=604, controls=605), pancreatic cancer (cases=107, controls=103), endometrial cancer (cases=190, controls=191), and multivitamin exposure study (multivitamin users=49, non-users=49)]. Within each specific study, the association between metabolite levels and PFOA was modeled using linear regression, adjusted for age at specimen collection, race/ethnicity, body mass index, smoking status, case status, study center, year of specimen collection. Sex was additionally adjusted for studies of pancreatic cancer, glioma, and multivitamin study. The combined model was conducted using random effects meta-analysis.

^aPFOS was further adjusted in linear regression models.

^bOnly the metabolites that met the Bonferroni corrected threshold of statistical significance in the combined models (0.05/1,100 = 4.55x10⁻⁵) are shown here.

Metabolites with a single asterisk (*) have identification level 2 no standards or matching spectra were available for them.

The 51 n-PFOS-associated metabolites, listed in Table 1, include 20 lipids, 20 xenobiotics, five cofactors/vitamins, three amino acids, two nucleotides, and one carbohydrate. The majority of n-PFOS-associated lipids were sphingolipids, fatty acid metabolites, and bile acid metabolites. In particular, we observed strong associations with sphingolipids, including hydroxypalmitoyl sphingomyelin (d18:1/16:0(OH)) [overall effect estimate=0.44 for a 1-standard deviation (SD) increase of the metabolite on the log scale, 95% confidence interval (CI)=0.24,0.64], lactosylceramide (d18:1/24:1(15Z)) [0.41 (0.29,0.54)], sphingomyelin (d18:1/18:0) [0.35 (0.24,0.46)], and sphingomyelin (d18:2/24:2) [0.32 (0.17,0.47)]. Among fatty acid metabolites, we also observed strong associations with 3-carboxy-4-methyl-5-pentyl-2-furanpropionate [3-CMPFP; 0.26 (0.19, 0.32)] and lignoceroylcarnitine (C24) [0.23 (0.15, 0.32)]. Bile acid metabolites generally had smaller effect estimates (<0.10). Most associated xenobiotics comprised xanthine metabolites (n=7 out of 20 xenobiotics) and food components/plants-related metabolites (n=7), although they had weaker associations. The strongest associations among xenobiotics were chemicals; 3,5-dichloro-2,6-dihydroxybenzoic acid [0.74 (0.67,0.80)], 3-bromo-5-chloro-2,6-dihydroxybenzoic acid [0.57 (0.43,0.72)], and 4-hydroxychlorothalonil [0.39 (0.3,0.49)]. Besides lipids and xenobiotics, we found associations in other chemical classes, including D-glucose [carbohydrate metabolite, 0.27 (0.15,0.39)], carotene diol [vitamin A metabolite, 0.22 (0.12,0.31)], 3-methylcytidine [nucleotide metabolite, 0.16 (0.1,0.22)], and L-alpha-aminobutyric acid [amino acid metabolite, 0.16 (0.09,0.23)]. When we adjusted for n-PFOA as a covariate, the associations with n-PFOS generally remained but became weaker (Table 1).

The 38 n-PFOA-associated metabolites, listed in Table 2, were composed of lipids (n=17), xenobiotics (n=15), nucleotides (n=2), an amino acid, a carbohydrate, a cofactor/vitamin, and a peptide. We observed 18 metabolites significantly associated with both n-PFOS and n-PFOA (10 xenobiotics, six lipids, one cofactor/vitamin, and one nucleotide). Similar to the n-PFOS results, lipids made up the majority of associations, including sphingolipids, fatty acid metabolites, and androgenic steroids. The strongest associations were observed for sphingolipids glycosyl-N-behenoyl-sphingadienine (d18:2/22:0) [0.26 (0.14,0.37)], sphingomyelin (d18:1/18:0) [0.25 (0.16,0.34)], sphingomyelin (d18:1/18:1(9Z)) [0.24 (0.16,0.32)], and lactosylceramide (d18:1/24:1(15Z)) [0.22 (0.12,0.32)]. Strong associations were also observed for sterol metabolite [3 beta-Hydroxy-5-cholestenoate, 0.24 (0.14,0.33)] followed by several fatty acid metabolites, including 3-CMPFP and lignoceroylcarnitine (C24), which were also significantly associated with n-PFOS. Among xenobiotics, the strongest associations were observed for the chemicals 3,5-dichloro-2,6-dihydroxybenzoic acid [0.48 (0.35,0.61)] and 3-bromo-5-chloro-2,6-dihydroxybenzoic acid [0.39 (0.23,0.55)], which are strongly correlated with one another (Pearson correlation=0.69; Supplementary Data). Most n-PFOA-associated xenobiotics (10 out of 15) were xanthine metabolites, although their associations were weak. Among metabolites from other chemical classes, associations with n-PFOA were observed for two nucleotide metabolites: uric acid [0.28 (0.16,0.41)] and 3-methylcytidine [0.15 (0.08,0.22)]. However, after adjustment with n-PFOS as a covariate, most of these associations were substantially attenuated and no longer statistically significant (Table 2); there were only three metabolites remained significant after n-PFOS adjustment [sphingomyelin (d18:1/18:1(9Z)), cysteine-S-sulfate, and 1,5-Anhydrosorbitol].

We conducted additional analyses to assess the sensitivity of our findings to different factors. Our findings for the top identified metabolites were nearly identical when we conducted analysis stratified by case status (Supplementary Table 3-4 and Supplementary Data) and sex (Supplementary Table 5 and 6). Additionally, adjusting for six metabolites previously reported to be associated with eGFR [creatinine, pseudouridine, N-acetylalanine, erythronate, myo-inositol, and glutarylcarnitine (20, 21)] also did not affect our findings (Supplementary Table 7 and 8). When we additionally examined associations of n-PFOS and n-PFOA with these six eGFR associated markers, we observed only weak to null associations overall, and no associations in analyses restricted to controls (Supplementary Table 9). Lastly, additional adjustment with education attainment or excluding BMI as a covariate did not change our findings (Supplementary Table 10 and 11).

4. Discussion

In this serum metabolomics study of PFOS and PFOA conducted among 3,647 adults aged 55-74 years enrolled in the PLCO Trial, we found 51 metabolites to be associated with serum n-PFOS at a Bonferroni-corrected significance threshold, predominantly involving lipid molecules and xenobiotics. We also observed 38 Bonferroni-significant metabolites associated with n-PFOA, although these associations were greatly attenuated after adjustment for n-PFOS.

We observed strong correlations between the Metabolon-based PFOS and PFOA data and CDC measurements of absolute serum PFOS and PFOA concentrations using a targeted analysis method among a subset of participants with both types of data. These correlations, along with our replication of previously-established PFOS/PFOA associations with sex and parity, support the validity of using the Metabolon-based PFAS measurements for this analysis (25-29). Our findings of differences in n-PFOS levels by race/ethnicity, BMI and smoking have not been consistently reported in other study populations, although similar findings were observed in a previous PLCO study involving measured serum PFAS concentrations (24).

Most metabolites associated with n-PFOS and n-PFOA were lipids. Of these, many were sphingolipid pathway metabolites. Sphingolipids, derivatives of amino alcohol sphingosine, are biologically active components of cell membranes (30). Sphingolipids play a significant role in intracellular signal transduction, regulate cellular processes (e.g., proliferation, maturation and apoptosis) and are involved in cellular stress responses. Ceramide is one of the most important sphingolipids and serves as a precursor for sphingosine and sphingosine-1-phosphate (S1P). Elevated serum levels of ceramide and S1P are associated with cardiovascular disease (e.g., obstructive coronary artery diseases, myocardial ischemia) and metabolic syndrome.(31-33) Fatty acid metabolites are also frequently represented among top hits. 3-carboxy-4-methyl-5-pentyl-2-furanpropionate (3-CMPF), associated with n-PFOS but not n-PFOA, is a minor derivative of CMPF [Human Metabolome Database (HMDB) 0061112], which is an endogenous metabolite of furan fatty acids and a well-established marker of fish intake.(34-36) Furan fatty acids are catabolized into dibasic urofuran acids and then excreted in the urine.(37) Elevated levels of blood CMPF have been shown in chronic kidney disease patients.(38) Lignoceroylcarnitine (C24) is a member of the class of acylcarnitines (HMDB0240665), major molecules involved in the energy supply pathway of long chain fatty acid β-oxidation.(7) Impaired fatty acid oxidation, along with tissue lipid accumulation, is critical in the pathophysiology of obesity/insulin resistance and cardiomyopathy.(7) Acylcarnitine is a marker for diagnosing fatty acid oxidation disorders and differentiation between the biochemical phenotypes of medium-chain acyl-CoA dehydrogenase deficiency disorders.(39) Lastly, bile acid metabolites were associated with n-PFOS, although their effect sizes were smaller. A previous in vitro study reported that PFOA and PFOS affect bile acid synthesis and bile canalicular morphology,(40) and another study found a PFAS mixture including PFOA and PFOS to increase bile acids in mice.(41) In a study of 20 healthy individuals, serum concentrations of bile acids, e.g., lithocholic acid, glycolithocholic acid, and taurolithocholic acid were positively correlated with several PFAS, including PFOA and PFOS.(42) Bile acid metabolism is known to play a role in the pathogenesis of type 2 diabetes, atherosclerosis, and non-alcoholic fatty liver disease.(43) Our findings are similar to previous metabolomic analyses of PFAS which noted associations with PFOS and PFOA for several classes of lipid metabolites, including glycerophospholipids, glycosphingolipids, fatty acid activation metabolites, and carnitines/acylcarnitines.(7, 9-11)

Many xenobiotics were associated with n-PFOS and n-PFOA. Metabolites of xanthine, both a xenobiotic and endogenous compound, are commonly associated with both PFOS and PFOA, although their effect sizes were small. Two studies conducted in China also reported significant associations of PFAS with hypoxanthine or xanthine.(10, 44) Most observed xanthine metabolites in our study are predictors of caffeinated coffee intake (e.g., 1-methylxanthine, paraxanthine, theophylline, caffeine, 1,3-dimethyluric acid, 1,3,7-trimethyluric acid).(35) Xanthine is a precursor of uric acid, which is the final oxidation product of purines.(45) Xanthine oxidoreductase catalyzes the oxidative hydroxylation of hypoxanthine to xanthine to uric acid. It has been reported that allopurinol, a xanthine oxidoreductase inhibitor, lowers serum levels of uric acid and exerts protective effects in situations associated with oxidative stress (46); however, other studies observed that uric acid can also act as an antioxidant in vivo. (47) Our observed association between n-PFOA and uric acid is consistent with previous studies that reported a positive PFOA association with risk of hyperuricemia and self-reported gout.(48-53) The PFOA association with uric acid remained in analyses adjusting for BMI, which lowers a possibility of increased uric acid due to obesity (54).

We also observed strong associations with the xenobiotics 3,5-dichloro-2,6-dihydroxybenzoic acid and 3-bromo-5-chloro-2,6-dihydroxybenzoic acid. There is limited literature on 3,5-dichloro-2,6-dihydroxybenzoic acid and 3-bromo-5-chloro-2,6-dihydroxybenzoic acid, which belong to the class of benzoic acids, although a structurally related xenobiotic -- 3,5-dichlorobenzoic acid and 3-amino-2,5-dichlorobenzoic acid (3-ADBA) -- is a known component of some herbicides.(55-57) Another similar xenobiotic 2,4-dichlorobenzoic acid is one of the degradation products of Chlorfenvinphos, an organophosphorus insecticide released to the environment from runoff after rainfall and leaching from hazardous waste sites.(58) A plasma metabolomic biomarker study of food intake found that both 3,5-dichloro-2,6-dihydroxybenzoic acid and 3-bromo-5-chloro-2,6-dihydroxybenzoic acid were positively correlated with red meat and milk intake, and negatively correlated with vegetable intake, particularly garlic consumption.(59) Additionally, 4-hydroxychlorothalonil is a xenobiotic metabolite of the widely used fungicide chlorothalonil, which is mainly applied to manage fungal foliar diseases of vegetable, field, and ornamental crops.(60) 4-hydroxychlorothalonil, which has a similar molecular structure to pentachlorophenol, has shown potential endocrine-disrupting activity in an experimental study of zebrafish and has been detected in human milk.(61, 62)

The basis for the associations with PFAS and these other xenobiotic chemicals is unclear; possible explanations include covariation among these chemicals in ingestion/exposure, with potential similar exposure routes from drinking water or diet, and/or excretion. The weak evidence of associations with PFAS within our study for known eGFR-related metabolites, coupled with the similar results we obtained after adjusting for these eGFR-related metabolites, argue against individual differences in renal clearance as an explanation for our findings. A previous PLCO study that measured serum PFAS concentrations and calculated eGFR from clinical measurements similarly found no association between eGFR and PFOS or PFOA.(24) However, it is not known how well these eGFR-related metabolites are predictive of eGFR; without clinically-based eGFR data, we cannot conclusively rule out kidney function as an explanation for findings.

PFAS associations with metabolites of other chemical classes are compatible with some suspected disease effects. Our observed positive association between n-PFOS and D-glucose levels suggests a role of PFOS in insulin resistance. A recent study also reported a positive association between PFOS and D-glucose (63). Some experimental studies have shown that PFAS activates nuclear receptors, including peroxisome proliferator-activated receptor α (PPAR-α), PPAR-γ, and estrogen receptor-α (ERα), which are involved in regulating glucose homeostasis.(64) A review of epidemiologic evidence relating PFAS and insulin resistance reported positive associations with PFOS in most studies.(65-69) An association between PFOA with vitamin A (retinol) metabolism has also been reported in a previous study of pregnant women.(70) Lastly, levels of urinary 3-methylcytidine have been reported to be reduced with increasing whole grain intake and to be higher among breast and gastric cancer patients than cancer-free individuals. (71, 72)

Notably, while the metabolite associations with n-PFOS remained after model adjustment for serum n-PFOA, the n-PFOA-metabolite associations became substantially attenuated after n-PFOS adjustment. It is unclear why we observed the attenuation for PFOA associations with PFOS adjustment; PFOS-related confounding is one possibility. Serum metabolomic measurements of n-PFOA and n-PFOS levels were moderately correlated with one another in our PLCO study population, with a Pearson correlation of 0.32. However, their absolute concentrations differ; in a previous PLCO nested case-control study investigating serum PFAS concentrations and kidney cancer, the median concentrations of PFOS and PFOA among controls were 38.4 μg/L and 5.5 μg/L, respectively (24).

Our study, to our knowledge the largest metabolomic study of PFAS conducted to date, has identified numerous metabolite associations that surpass stringent Bonferroni-corrected significance thresholds. Other strengths of this study include our extensive adjustment for subject and sample characteristics potentially confounding the metabolomic associations. Although sensitivity analyses involving adjustment for eGFR-related metabolites did not affect our key findings, we cannot rule out confounding from kidney function as a source of bias. Another limitation is the cross-sectional nature of our study; with PFOA and PFOS measured along with the other metabolites from the same sample, we are unable to disentangle the temporality of the observed associations, limiting causal inferences. In addition, the duration of storage (~30 years) can affect metabolite levels; however, any such effects are unlikely to differ in relation to PFAS levels, and thus likely to have introduced bias in PFAS-metabolite associations towards the null. As non-fasting serum samples were used in these PLCO studies, it is possible that short-term post-prandial changes in metabolite levels could have also introduced bias, also likely towards the null. The generalizability of our findings to pre-menopausal women is uncertain as the majority were post-menopausal women. As semi-targeted metabolomic analyses measure relative metabolite levels, follow-up studies using targeted analyses are warranted for more direct comparison across populations. Lastly, although we observed higher levels of n-PFOS among non-Hispanic Black persons, we were unable to provide racial/ethnic-specific findings as most participants were non-Hispanic White persons (81%).

5. Conclusions

This large metabolomic investigation has identified numerous metabolites associated with n-PFOS and n-PFOA. Our findings offer further support for some previously reported disease associations (e.g., D-glucose, supporting a PFOS association with insulin sensitivity, and uric acid, supporting a PFOA association with gout) as well as provide new leads into biologic pathways potentially affected by these chemicals (sphingolipid-related pathways in particular) that warrant further investigation.

Abbreviations:

BMI

body mass index

CDC

Centers for Disease Control and Prevention

eGFR

estimated glomerular filtration rate

ERα

estrogen receptor-α

HMDB

Human Metabolome Database

LOD

limit of detection

n-PFOS

linear isomers of perfluorooctane sulfonate

n-PFOA

linear isomers of perfluorooctanoate

NCI

National Cancer Institute

PFAS

Per- and polyfluoroalkyl substances

PFOS

perfluorooctane sulfonate

PFOA

perfluorooctanoate

PPAR-α

peroxisome proliferator-activated receptor α

PLCO

Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial

SM

sphingomyelin

S1P

sphingosine-1-phosphate

SD

standard deviation

UHPLC-MS/MS

ultra high-performance liquid chromatography/tandem accurate mass spectrometry

3-CMPF

3-carboxy-4-methyl-5-pentyl-2-furanpropionate