Plasma concentrations of per- and polyfluoroalkyl substances are associated with perturbations in lipid and amino acid metabolism

Abstract

Exposure to per- and polyfluoroalkyl substances (PFAS) through the environment can lead to harmful health outcomes and the development of disease. However, little is known about how PFAS impact underlying biology that contributes to these adverse health effects. The metabolome represents the end product of cellular processes and has been used previously to understand physiological changes that lead to disease. In this study, we investigated whether exposure to PFAS was associated with the global, untargeted metabolome. In a cohort of 459 pregnant mothers and 401 children, we quantified plasma concentrations of six individual PFAS- PFOA, PFOS, PFHXS, PFDEA, and PFNA- and performed plasma metabolomic profiling by UPLC-MS. In adjusted linear regression analysis, we found associations between plasma PFAS and perturbations in lipid and amino acid metabolites in both mothers and children. In mothers, metabolites of 19 lipid pathways and 8 amino acid pathways were significantly associated with PFAS exposure at an FDR<0.05 threshold; in children, metabolites of 28 lipid pathways and 10 amino acid pathways exhibited significant associations at FDR<0.05 with PFAS exposure. Our investigation found that metabolites of the Sphingomyelin, Lysophospholipid, Long Chain Polyunsaturated Fatty Acid (n3 and n6), Fatty Acid- Dicarboxylate, and Urea Cycle showed the most significant associations with PFAS, suggesting these may be particular pathways of interest in the physiological response to PFAS. To our knowledge, this is the first study to characterize associations between the global metabolome and PFAS across multiple periods in the life course to understand impacts on underlying biology, and the findings presented here are relevant in understanding how PFAS disrupt normal biological function and may ultimately give rise to harmful health effects.

Graphical Abstract

1. Introduction

Polyfluoro- and perfluoroalkyl substances (PFAS) are organic pollutants that have garnered attention in recent years due to their impact on public health. PFAS have been used extensively for commercial applications, and while many have been banned or phased out of commercial use, exposure to these substances remains widespread.^1–3 PFAS have been previously linked to increased risk of certain cancers,⁴ adiposity and diabetes,⁵ cardiovascular disease,⁶ reduced birth weight,⁷ and impaired postnatal growth.⁸ Consequently, PFAS are generally considered unsafe by regulating agencies such as the U.S. Environmental Protection Agency⁹ and U.S. Department of Health and Human Services.¹⁰ However, the results of many existing studies are complicated by inconsistent or nonconclusive findings,^11–14 that limit interpretation of their direct impact on biology. Exposures to PFAS vary widely based on socioeconomic status, geographic location, and other factors,^15,16 which could contribute to this inconsistency. Currently, there is a gap in the literature regarding the specific biological pathways that are disrupted by PFAS and contribute to adverse health outcomes in large, heterogeneous cohorts, but examining associations between the metabolome and PFAS exposures may elucidate underlying biology.

Recent evidence suggests PFAS exposure can perturb the metabolome, with disruptions to lipid and amino acid metabolism as classes of particular interest. Potential metabolomic consequences of PFAS exposure have been suggested to include increased fatty acid oxidation,¹⁷ hypolipidemia,^18,19 and suppressed amino acid metabolism.^20,21 While the results of these studies have proven invaluable in establishing a role for the metabolome in PFAS environmental exposure response, many previous studies have relied on targeted panels for investigation rather than a global examination through untargeted metabolomic profiles. Global, untargeted metabolomic profiles may allow discovery of novel pathways and biological mechanisms that underlie diseases and provocation of physiological response. To better understand relationships between PFAS exposure and the metabolome, we performed an exploratory analysis of associations between PFAS concentrations and metabolites in mothers and children from the Vitamin D Antenatal Asthma Reduction Trial (VDAART) using a global, untargeted metabolomic approach. The VDAART cohort was heterogeneous with respect to race, socioeconomic status, and geographical location, and the findings presented here provide insight into metabolomic alterations in response to environmental PFAS exposure.

2. Materials and Methods

2.1. Vitamin D Antenatal Asthma Reduction Trial Study Participants

VDAART is a randomized, double-blind, parallel-design study conducted at three study sites across the United States (ClinicalTrials.gov identifier: NCT00920621), originally described by Litonjua et al.²² Informed consent want obtained from all individuals, and the work described here was conducted in accordance with The Code of Ethics of the World Medical Association. VDAART recruited pregnant mothers and randomized them to Vitamin D supplementation at 4000 IU/day or placebo to better understand the relationship between vitamin D levels and childhood outcomes; all women received 400 IU/day vitamin D supplementation as part of usual pregnancy care. Mothers returned for a follow-up study visit, and blood was collected. Follow up of the offspring, including periodic blood collection, is ongoing.

2.2. Plasma collection and metabolomic profiling

Blood draws were collected in EDTA tubes during study visits (3^rd trimester between 32–38 gestational weeks for mothers; age 6 years for children); plasma was separated through centrifugation at 2000 RPM at 4°C, and samples were stored at −80°C until metabolomic profiling. Metabolomic profiles of untargeted metabolites were generated by UPLC-MS/MS at both maternal and child time points in VDAART by Metabolon, Inc. (NC, USA). All methods utilized a Waters ACQUITY ultra-performance liquid chromatography (UPLC) and a Thermo Scientific Q-Exactive high resolution/accurate mass spectrometer interfaced with a heated electrospray ionization (HESI-II) source and Orbitrap mass analyzer operated at 35,000 mass resolution. A total of 828 global metabolites in mothers and 804 metabolites in children passed QC and were included in analysis. Metabolites were assigned to sub-pathways by Metabolon based on the existing literature. Of these global metabolites, there were 586 lipid and amino acid metabolites in mothers and 573 lipid and amino acid metabolites in children; 97.1% of metabolites were detected at both time points. All lipid and amino acid pathways evaluated included at least one metabolite at both time points. Biological pathway assignments from Metabolon were used. Details of sample preparation, data collection, and data processing/quality control are provided in the Supplementary Methods and described elsewhere.²³

2.3. Measurement of PFAS in VDAART plasma samples

A subset of VDAART samples were selected for PFAS quantification, based on availability of suitable plasma sample volume and metabolomic profiling; this included 459 VDAART mothers and 401 VDAART children. PFAS were measured in an aliquot of the same sample as metabolomic profiling. PFAS quantification was performed by the Human Health Exposure Analysis Resource (HHEAR), and more information about HHEAR initiatives for PFAS measurement can be found at: https://hhearprogram.org/exposure-and-polyfluoroalkyl-substances-pfas. Quantitative, targeted measurement of six PFAS were included for this study, using methods originally developed by Kato et al.²⁴ In this analysis, the following PFAS were included: Perfluorooctanoic acid (PFOA), Perfluorooctanesulfonic acid (PFOS; linear and branched), Perfluorohexane-1-sulphonic acid (PFHXS), Perfluorononanoic acid (PFNA), Perfluorodecanoic acid (PFDEA), and Perfluoroundecanoic acid (PFUA). Concentrations were log-transformed, and values below LOD were imputed as the limit of detection (LOD) divided by the square root of 2. Due to a high percent of values below LOD for the compound PFUA, exposure was treated as a binary variable in which detected=1 and not detected=0. The LOD for all PFAS species was 0.1 ng/mL.

2.4. Statistical analyses

PFAS concentrations were log-transformed prior to statistical analysis, but pre-transformed concentrations are shown in Table 2. Correlations between individual PFAS were assessed separately in mothers and children using Pearson correlations (Supplementary Figure 1). Maternal and child VDAART samples with both untargeted metabolomic profiling and PFAS measurement were included for analysis, resulting in a total of 459 VDAART mothers and 401 VDAART children.

Table 2.

Descriptive statistics for six individual PFAS species included in analysis. Statistics are shown separately for VDAART mothers and children.

PFAS Analyte	N	Limit of Detection (ng/mL)	Percent detection above LOD (%)	Median (ng/mL)	25th–75th Percentiles
VDAART Mothers at 32-38 gestational weeks
PFOA	459	0.1	99.7	1.3	0.9–1.8
PFOS a	459	0.1	100	3.9	2.7–5.2
PFDEA	459	0.1	90.2	0.2	0.2–0.3
PFHXS	459	0.1	99.6	0.8	0.5–1.4
PFNA	459	0.1	99.3	0.5	0.4–0.7
PFUA b	459	0.1	68.0	0.2	0.1–0.2
VDAART Children at age 6 years
PFOA	401	0.1	99.0	1.4	1.1–2.0
PFOS a	401	0.1	99.8	2.3	1.7–3.4
PFDEA	401	0.1	78.3	0.2	0.1–0.2
PFHXS	401	0.1	99.8	0.7	0.5–1.1
PFNA	401	0.1	96.3	0.4	0.3–0.5
PFUA b	401	0.1	24.2	0.1	0.1–0.2

^aLinear and branched PFOS were combined

^bPFUA was converted to a binary variable for yes/no detected due to low percent of detection

2.5. Linear regression models

Linear regression analyses were performed with log-transformed PFAS concentrations as continuous (PFOA, PFOS, PFDEA, PFHXS, and PFNA) or binary (PFUA) measures based on the distribution of the data. Metabolites were treated as outcomes, and models adjusted for variables relevant for PFAS exposure based on scientific literature^25,26 as well as covariates relevant specifically for the VDAART cohort.²³ Models were adjusted for race, ethnicity, vitamin D level (nanograms per milliliter of blood [ng/ml]), body mass index (BMI), study site, household income, and maternal education. Maternal models included an additional adjustment for age, and child models included an additional adjustment for sex. Household income was categorized based on reported household yearly income in USD: Low (<$50,000/year), Medium ($50,000-$100,000/year), or High (>$100,000/year). Maternal education was categorized based on maximum education level reported: Low (primary school, secondary school, or some college/junior college), Medium (technical/trade school or bachelor’s degree), or High (graduate degree). Linear regression analyses were performed using the glm package in R,²⁷ and the Benjamini-Hochberg Procedure was used to correct for multiple testing. Metabolite-PFAS associations were considered to be statistically significant if they reached a threshold of FDR-corrected P-value<0.05.

2.6. Mixture analysis using quantile g-computation

In order to estimate the joint association of all PFAS as a mixture on the metabolome, we utilized quantile g-computation. Quantile g-computation estimates the effect of simultaneously increasing all exposure variables in a mixture by one quartile. All metabolites that demonstrated FDR-corrected significant associations (FDR<0.05) with an individual PFAS were included the mixture model as outcomes. The same covariates as individual PFAS models were used as adjustments, and PFOA, PFOS, PFDEA, PFHXS, and PFNA were used as exposure variables in the mixture. PFUA could not be included due to its low percent detected that prohibited transformation into a quartile. The qgcomp package version 2.10.1²⁸ was utilized in R, which implements a parametric, generalized linear model of gcomputation to provide an estimate of the association between metabolites and a one quartile increase in the total PFAS mixture.

In order to assess potential confounding of vitamin D supplementation during pregnancy on the associations between metabolites and PFAS, we performed a stratified analysis within mothers (and children of mothers) that received vitamin D supplementation and those that received placebo for metabolites included in the PFAS mixture models. Directions of effect were compared between the overall cohort, placebo-only group, and treatment-only group.

3. Results

3.1. Description of cohort characteristics for VDAART mothers and children

Cohort characteristics for mothers and children in VDAART with both metabolomic profiling data and PFAS exposure data are shown in Table 1. A total of 459 VDAART mothers and 401 VDAART children were included. More than half of the parents reported low household income (52%) and education (63%). Concentrations of PFOA, PFDEA, and PFHXS did not significantly differ between VDAART mothers and children at P<0.05. PFOS and PFNA concentrations were significantly higher in VDAART mothers than VDAART children (P<0.0001 and P=0.0055, respectively). Comparison of the binary PFUA exposure variable showed that mothers were more likely to be exposed to PFUA than children (P<0.0001).

Table 1.

Cohort characteristics for VDAART pregnant mothers and children at age 6 years with PFAS and metabolomic profiling.

VDAART Cohort Characteristics and PFAS Exposures
Cohort Characteristics
	Mothers	Children
Number of subjects	459	401
BMI kg/m2, Mean (SD)1	28.8 (8.0)	16.9 (2.9)
Age at collection, Mean (SD)	27.4 (5.5)	6.0 (0.1)
Gender, n (%)
Female	459 (100.0)	185 (46.1)
Male	-	216 (53.9)
Race, n (%)
Black	204 (44.4)	197 (49.1)
White	176 (38.3)	129 (32.2)
Other	79 (17.2)	75 (18.7)
Ethnicity, n (%)
Hispanic or Latino	128 (27.9)	136 (33.9)
Not Hispanic or Latino	331 (72.1)	265 (66.1)
Study site, n (%)
Boston	128 (27.9)	114 (28.4)
San Diego	157 (34.2)	138 (34.4)
St. Louis	174 (37.9)	149 (37.2)
Household Income Category, n (%)2
Low	194 (56.1)	172 (57.0)
Medium	106 (30.6)	87 (28.8)
High	46 (13.3)	43 (14.2)
Maternal Education Category, n (%)3
Low	287 (62.5)	248 (61.8)
Medium	110 (24.0)	102 (25.4)
High	62 (13.5)	51 (12.7)
Vitamin D supplementation in pregnancy, n (%)4	227 (49.5)	198 (49.4)
Vitamin D Level at Time of Collection (ng/mL), Mean (SD)	32.4 (13.9)	28.3 (9.7)

¹BMI in mothers was reported as pre-pregnancy BMI

²Income was categorized based on reported household yearly income in USD: Low (<$50,000/year), Medium ($50,000–$100,000/year), or High (>$100,000/year)

³Maternal educational status was categorized based on maximum education level reported: Low (primary school, secondary school, or some college/junior college), Medium (technical/trade school or bachelor’s degree), or High (graduate degree)

⁴Percent of mothers in the vitamin D supplementation group is shown; for children, the percent of children with mothers that received vitamin D supplementation in pregnancy is shown

3.2. PFAS profiling in mothers and children

All PFAS measured by the method developed by Kato et al.24 had a limit of detection of 0.1 ng/mL. Across all individuals (mothers and children combined), percent of values below LOD were as follows: PFOA (1%), PFOS (0%), PFDEA (22%), PFHXS (1%), PFNA (3%), and PFUA (66%). Further description is provided in Table 2, which shows percent detected, median values, and IQR for each PFAS among mothers only and children only, respectively. All individual PFAS were detected in >90% of VDAART mothers except PFUA, which was only detected in 68% of mothers. PFOA, PFOS, PFHXS, and PFNA were detected in >95% of VDAART children, but PFDEA was only detected in 78.3% of children, and PFUA was detected only in 24.2% of children.

3.3. Correlations between plasma PFAS levels in VDAART

Correlations between the 6 individual log-transformed PFAS species (PFOA, PFOS, PFNA, PFDEA, PFHXS, and PFUA) were assessed in both mothers and children. Pearson correlation coefficients are shown for all nominally significant (P<0.05) correlations (Figure S1). These results revealed PFAS concentrations were significantly positively correlated for most PFAS species, except for PFHXS and PFUA in both mothers and children. Correlation coefficients were similar between mothers and children; coefficients ranged from 0.14–0.76 in mothers and 0.14–0.73 in children. No PFAS species were negatively correlated. Additionally, individual PFAS levels were evaluated between mother-child pairs; coefficients for PFAS correlations between mothers and children ranged from 0.10–0.29.

3.4. Impact of PFAS on Lipid Metabolites Differed Between Mothers and Children

Metabolites from a total of 19 of 57 lipid pathways were significantly associated with individual PFAS in VDAART mothers (Figure 2a, Table 3), and direction of association differed across lipid sub-pathways. Metabolites from 10 lipid sub-pathways were positively associated with PFAS exposure, while metabolites from 9 sub-pathways were negatively associated with PFAS exposure (Supplementary files S1 and S2). The metabolite that showed the lowest P-value with any individual PFAS was hydroxy-CMPF with PFUA (P=5.03×10⁻¹¹) of the Fatty Acid- Dicarboxylate pathway; another metabolite in this pathway showed the second lowest p-value for association with the PFAS mixture, 3-carboxy-4-methyl-5-propyl-2-furanpropanoate (CMPF; P=1.47×10⁻³). 3-CMPFP was the most or second most strongly significantly associated lipid metabolite with PFUA, PFOS, and PFDEA, with P-values of 6.09×10⁻¹¹, 1.46×10⁻⁵, and 9.99×10⁻⁶, respectively. It was in the top 10 most significant lipid metabolite associations for PFHXS and PFNA, with P-values of 2.21×10⁻³ and 2.82×10⁻³, respectively. 3-CMPFP was significantly associated with PFOA at FDR<0.05 but was not one of the top associations. In VDAART mothers, the Sphingomyelin sub-pathway showed the highest number of individual metabolites associated with PFAS, with 11 sphingomyelin metabolites showing 35 negative associations with individual PFAS (P-value range 3.20 ×10⁻⁴ to 0.030), but no sphingomyelin metabolites were associated with overall PFAS mixture at a nominal level of P<0.05. Lipid metabolites belonging to Long Chain Polyunsaturated Fatty Acid (n3 and n6) showed the top significant association with PFAS mixture, as well as two additional metabolites in the top 10 (P-value range 1.13×10⁻³ to 6.68×10⁻³). We assessed these relationships overall compared to strata of vitamin D treatment and found that 40 of 43 lipid metabolites (93.0%) retained their direction of effect in the treatment-only mothers and 32 of 43 lipid metabolites (74.4%) retained their direction of effect in the placebo-only mothers (Supplementary File S5).

Figure 2.

Lipid metabolite sub-pathways are associated with PFAS exposure in VDAART mothers and children. Lipid sub-pathways containing metabolites significantly associated with PFAS at FDR<0.05 are shown for total PFAS (grey) and 6 individual PFAS species (multiple colors, according to legend), and the count of significant metabolite associations are shown for all plots. In VDAART mothers (A), 19 lipid sub-pathways were significantly associated with either total PFAS, an individual PFAS, or both; in VDAART children (B), 28 lipid sub-pathways were significantly associated with either total PFAS, an individual PFAS, or both.

Table 3.

Summary of lipid metabolites associated with total PFAS and individual PFAS in VDAART mothers.

VDAART Mothers
	PFAS mixture			Individual PFAS
Lipid Sub-Pathway	Number of Metabolites	Number in Pathway	% Significant in Pathway	Number of Total Associations	Number of Unique Metabolites	Number in Pathway	% Significant in Pathway
Sphingomyelins		29		35	11	29	37.93
Fatty Acid, Dicarboxylate	3	27	11.11	12	5	27	18.52
Fatty Acid, Monohydroxy		16		8	3	16	18.75
Long Chain Polyunsaturated Fatty Acid (n3 and n6)	4	16	25	6	4	16	25
Progestin Steroids		7		5	2	7	28.57
Fatty Acid Metabolism (also BCAA Metabolism)	1	4	25	4	1	4	25
Fatty Acid Metabolism (Acyl Carnitine, Monounsaturated)		10		3	2	10	20
Diacylglycerol	1	3	33.33	3	2	3	66.67
Hexosylceramides (HCER)		4		2	2	4	50
Fatty Acid Metabolism (Acyl Choline)		9		2	1	9	11.11
Ceramides		4		2	2	4	50
Sterol		5		1	1	5	20
Phospholipid Metabolism		7		1	1	7	14.29
Phosphatidylethanolamine (PE)	1	12	8.33	1	1	12	8.33
Medium Chain Fatty Acid		9		1	1	9	11.11
Long Chain Saturated Fatty Acid		7		1	1	7	14.29
Fatty Acid Synthesis		2		1	1	2	50
Fatty Acid Metabolism (Acyl Carnitine, Polyunsaturated)	1	6	16.67	1	1	6	16.67
Fatty Acid Metabolism (Acyl Carnitine, Long Chain Saturated)		8		1	1	8	12.5

The number of lipid metabolites significantly associated with PFAS mixture or one of the six individual PFAS species are summarized by lipid sub-pathway. For associations with PFAS mixture, the number of metabolites with associations that met an FDR<0.05 threshold, total number of metabolites tested in the sub-pathway (significant and non-significant), and the percent of significant metabolites within a pathway are shown for each lipid sub-pathway. For associations with individual PFAS, the number of metabolite associations that met an FDR<0.05 threshold, number of unique metabolites that produced associations, total number of metabolites tested in the sub-pathway (significant and non-significant), and the percent of significant metabolites within a pathway are shown for each lipid sub-pathway. Sub-pathways with no significant associations are colored in grey.

There were 28 of 56 lipid pathways significantly associated with individual PFAS or PFAS mixture in VDAART children (Figure 2b, Table 3); 19 were positively associated with PFAS, 8 were negatively associated, and one (Fatty Acid- Dicarboxylate) showed mixed positive and negative assocations with PFAS (Supplementary files S3 and S4). Metabolites hydroxy-CMPF and 3-carboxy-4-methyl-5-propyl-2-furanpropanoate of the Fatty Acid- Dicarboxylate pathway showed the smallest P-values of individual associations, both with PFUA (P=8.01×10⁻¹⁰ and 1.06×10⁻⁷, respectively). However, Lysophospholipid and Phosphatidylcholine (PC) pathways showed the largest number of associations with individual PFAS and PFAS mixture. The metabolite with the smallest P-value for the association in VDAART children in the Lysophospholipid pathway was the metabolite, 1-palmitoyl-GPC (16:0) with a P-value of 3.51×10⁻⁵. 1-palmitoyl-GPC (16:0) was not the most significant association with any individual PFAS, but other Lysophospholipid sub-pathway metabolites showed the most or second most significant lipid associations for PFHXS (Metabolite: 2-palmitoyl-GPC* (16:0)*; P-value0.018), PFNA (Metabolite: 2-stearoyl-GPE (18:0)*; P-value=1.65×10⁻³), and PFDEA (Metabolite: 2-palmitoyl-GPC* (16:0)*; P-value=4.45×10⁻⁵). In VDAART children, the Lysophospholipid sub-pathway showed the highest number of individual metabolites associated with PFAS, with 10 unique metabolites (P-value range 3.51×10⁻⁵ to 0.046). The Phosphatidylcholine (PC) pathway also demonstrated 10 unique metabolite associations, for a total of 32 total significant associations at FDR<0.05 (P-value range 5.02×10⁻⁵ to 0.034). We assessed these relationships overall compared to strata of vitamin D treatment and found that 78 of 80 lipid metabolites (97.5%) retained their direction of effect in the children of treatment-only mothers and 75 of 80 lipid metabolites (93.7%) retained their direction of effect in the children of placebo-only mothers (Supplementary File S6).

Of the lipid pathways significantly associated with PFAS in VDAART mothers and children, there were 12 lipid pathways that overlapped between the two groups (Figure 4). Five of these pathways were consistent in direction of association between mothers and children: Phospholipid Metabolism, Long Chain Saturated Fatty Acid, Fatty Acid Metabolism (Acyl Carnitine, Monounsaturated), Medium Chain Fatty Acid, and Long Chain Polyunsaturated Fatty Acid (n3 and n6); Fatty Acid- Dicarboxylate showed a mix of direction in VDAART children but did have some overlap with VDAART mothers. The remaining 7 pathways differed in direction of association; most were negatively associated in mothers but positively associated in children, with the exception of Progestin Steroids, which were positively associated in mothers but negatively associated in children.

Figure 4.

Overlap in significantly associated pathways (FDR<0.05) in VDAART mothers and children. Lipid and amino acid sub-pathways that were associated with individual PFAS or overall PFAS mixture (FDR<0.05) are shown. A total of 47 lipid or amino acid sub-pathways were significantly associated with PFAS in either mothers or children, and 18 (38.3%) of these pathways overlapped at the two time points; 12 belonged to the lipid super-pathway and 6 belonged to the amino acid super-pathway. Nine (19.1%) pathways were only significantly associated with PFAS in VDAART mothers; 8 of these were lipid, and 1 was amino acid. A total of 20 (42.6%) pathways were only significantly associated with PFAS in VDAART children; 16 were lipid and 4 were amino acid.

3.5. Amino Acid Metabolism was Increased in Response to PFAS

A total of 8 of 15 amino acid sub-pathways and 17 metabolites were associated with total PFAS or individual PFAS in VDAART mothers at FDR<0.05, and these associations were predominantly in the positive direction (Supplementary files S1 and S2). Only 2 metabolites produced negative associations with PFAS, including one Glutamate Metabolism metabolite, beta-citrylglutamate, with PFNA (P=7.55×10⁻³), one Glycine, Serine, and Threonine metabolite, dimethylglycine, with PFNA (5.57×10⁻³). The associations for the remaining 15 metabolites were in the positive direction (P-value range=5.44×10⁻⁴ to 0.048). The amino acid sub-pathway with the highest number of associations with individual PFAS was Leucine/Isoleucine/Valine Metabolism (Figure 3a, Table 5). Only two amino acid pathways showed significant associations with PFAS mixture: Methionine, Cysteine, SAM, and Taurine Metabolism (O-acetylhomoserine, P=0.035) and Glycine, Serine, and Threonine metabolism (alphaketobutyrate, P=0.044). O-acetylhomoserine demonstrated FDR-corrected significant associations with PFNA (P=0.020), PFOS (P=0.019), PFOA (P=0.038), and PFUA (0.047). Alphaketobutyrate demonstrated FDR-corrected significant associations with PFHXS (P=0.029) and PFDEA (P=0.028). When we assessed whether these associations were robust across strata, we found that 14 of 17 (82.4%) amino acid metabolites maintained their directions of effect in treatment-only mothers, and 16 of 17 (94.1%) amino acid metabolites maintained their directions of effect in placebo-only mothers.

Figure 3.

Amino acid metabolite sub-pathways are associated with PFAS exposure in VDAART mothers and children. Amino acid sub-pathways containing metabolites significantly associated with PFAS at FDR<0.05 are shown for total PFAS (grey) and 6 individual PFAS species (multiple colors, according to legend), and the count of significant metabolite associations are shown for all plots. In VDAART mothers (A), 8 amino acid sub-pathways were significantly associated with either total PFAS, an individual PFAS, or both; in VDAART children (B), 10 amino acid sub-pathways were significantly associated with either total PFAS, an individual PFAS, or both.

Table 5.

Summary of amino acid metabolites associated with total PFAS or individual PFAS in VDAART mothers.

VDAART Mothers
	PFAS mixture			Individual PFAS
Amino Acid Sub-Pathway	Number of Metabolites	Number in Pathway	% Significant in Pathway	Number of Total Associations	Number of Unique Metabolites	Number in Pathway	% Significant in Pathway
Leucine, Isoleucine and Valine Metabolism		31		8	4	31	12.9
Tryptophan Metabolism		18		5	3	18	16.67
Methionine, Cysteine, SAM and Taurine Metabolism	1	20	5	5	3	20	15
Glycine, Serine and Threonine Metabolism	1	10	10	5	2	10	20
Tyrosine Metabolism		14		2	1	14	7.14
Histidine Metabolism		14		2	2	14	14.29
Glutamate Metabolism		12		1	1	12	8.33
Alanine and Aspartate Metabolism		9		1	1	9	11.11

The number of amino acid metabolites significantly associated (FDR<0.05) with PFAS mixture or one of the six individual PFAS species are summarized by amino acid sub-pathway. Categorization is consistent with Table 3.

In VDAART children, 10 of 15 sub-pathways demonstrated significant associations with PFAS at FDR<0.05. Urea cycle; Arginine and Proline Metabolism sub-pathway showed the largest amount of significant metabolites with PFAS mixture, with 3 metabolites all with the top 10 lowest P-values (P-value range 0.013 to 0.029). This pathway also showed the highest number of associations with individual PFAS, representing 5 unique metabolites and 12 total associations (P-value range=4.61×10⁻³ to 6.46×10⁻³). Homocitrulline of this pathway also demonstrated significant positive associations with 4 of the 6 individual PFAS: PFNA (P-value=1.84×10⁻³), PFOS (P-value=7.00×10⁻³), PFUA (P-value=0.015), and PFDEA (P-value=0.033). When we assessed whether these associations were robust across strata, we found that 19 of 20 (95.0%) amino acid metabolites maintained their directions of effect in children of treatment-only mothers, and 17 of 20 (85.0%) amino acid metabolites maintained their directions of effect in children of placebo-only mothers.

A total of 6 amino acid sub-pathways overlapped between VDAART mothers and children, and broadly agreed in direction of association (Figure 4): Alanine and Aspartate Metabolism, Glutamate Metabolism, Histidine Metabolism, Glycine, Serine, and Threonine Metabolism, Tryptophan Metabolism, and Leucine, Isoleucine, and Valine Metabolism. Tables 5 and 6 summarize the number of unique metabolites, total metabolite associations, and total number of metabolites in each amino acid sub-pathway for VDAART mothers and VDAART children.

Table 6.

Summary of amino acid metabolites associated with total PFAS or individual PFAS in VDAART children.

VDAART Children
	PFAS mixture			Individual PFAS
Amino Acid Sub-Pathway	Number of Metabolites	Number in Pathway	% Significant in Pathway	Number of Total Associations	Number of Unique Metabolites	Number in Pathway	% Significant in Pathway
Urea cycle; Arginine and Proline Metabolism	3	21	14.29	12	5	21	23.81
Leucine, Isoleucine and Valine Metabolism		30		5	3	30	10
Glutamate Metabolism	1	12	8.33	4	2	12	16.67
Glycine, Serine and Threonine Metabolism	1	10	10	3	2	10	20
Alanine and Aspartate Metabolism	2	9	22.22	3	2	9	22.22
Tryptophan Metabolism	1	17	5.88	2	2	17	11.76
Lysine Metabolism		17		1	1	17	5.88
Histidine Metabolism		15		1	1	15	6.67
Glutathione Metabolism	1	7	14.29	1	1	7	14.29
Creatine Metabolism		3		1	1	3	33.33

The number of amino acid metabolites significantly associated (FDR<0.05) with PFAS mixture or one of the six individual PFAS species are summarized by amino acid sub-pathway. Categorization is consistent with Table 3.

3.6. Other pathways perturbed by PFAS exposure in the global metabolome

Several non-lipid and non-amino acid metabolites were associated with total PFAS exposure or individual PFAS. In VDAART mothers, significant associations were observed between total PFAS and 1 Cofactor/Vitamin, 1 Nucleotide, and 3 Xenobiotics. In VDAART children, 1 Carbohydrate, 3 Cofactors/Vitamins, 2 Nucleotide, 1 Peptide, and 5 Xenobiotic super-pathway metabolites were associated with PFAS sum. In most cases, there were no more than 2 metabolites belonging to the same sub-pathway (Supplementary Files S1–S4).

4. Discussion

In this study, we utilized global, untargeted metabolomic profiling to investigate novel pathways associated with PFAS exposure in pregnant mothers and their offspring to represent multiple points in the life course. Our results demonstrated disruptions to lipid and amino acid metabolism in both mothers and children in the VDAART cohort, and indicated the Sphingomyelin, Lysophospholipid, Long Chain Polyunsaturated Fatty Acid (n3 and n6), Fatty Acid- Dicarboxylate, and Urea cycle: Arginine and Proline metabolism pathways as particularly important to understanding physiological response to PFAS exposure. Lipid and amino acid metabolism pathways have been previously implicated in PFAS exposures, and previous studies have demonstrated disruptions to multiple lipid- and amino acid-related pathways.²⁹ Recent literature has emphasized the particular importance of the metabolome in response to a range of PFAS exposure concentrations,^29,30 and metabolomic profiling has facilitated improved understanding of the biological mechanisms associated with PFAS exposure. Metabolomic studies have provided insights about the biological changes that underlie disease etiology,^31–33 so further investigation into metabolomic perturbations may improve understanding of the biological response to PFAS.

VDAART included mothers and children which permitted study of the contrast in metabolomic response across different periods in life. A total of 18 pathways overlapped in associations with either individual PFAS or total PFAS burden, including 12 lipid-associated pathways and 6 amino acid-associated pathways. Sphingomyelins and Fatty Acid- Dicarboxylate lipid metabolites demonstrated a high number of significant associations at both time points, but, notably, Fatty Acid- Dicarboxylate and Long Chain Polyunsaturated Fatty Acid (n3 and n6) metabolites were consistent in direction of effect between mothers and children. Previous in vivo studies of metabolomics response to PFAS have observed a shift from carbohydrate metabolism towards fatty acid oxidation,¹⁷ and our results indicated this shift was present across different stages in the life course, with the most significant results in the Fatty Acid- Dicarboxylate metabolites. Other fatty acid pathways were consistent in direction between both time points to a lesser degree of significance, including Long Chain Saturated Fatty Acids, Medium Chain Fatty Acids, Fatty Acid Metabolism- Acyl Chain Monounsaturated, and Fatty Acid- Monohydroxy. Dysregulated fatty acid oxidation can produce excess reactive oxygen species,³⁴ provoking disruptions to normal physiological function which can lead to cell damage and death.³⁵ Phosphatidylcholine (PC) and Lysophospholipid metabolites showed a high number of significant associations in children only, and the impact of PFAS exposure on this pathway is notable due to the high number of associated metabolites. Lysophospholipids are known to be activated in response to oxidative stress and is likely related to increased fatty acid oxidation.³⁶ The disparate trends in mothers and children may reflect subtle differences in inflammation, since lysophospholipids can exhibit both pro-inflammatory or anti-inflammatory effects, depending on biological context.³⁷ Elevation of Phosphatidylcholine (PC) metabolites may further suggest increased lipid oxidation in response to PFAS exposure, which has been observed in previous studies.³⁸ Sphingolipids metabolites were negatively associated with PFAS in mothers but positively associated with PFAS in children. Increased levels of sphingomyelins, as observed in VDAART children, are associated with increased risk of neurological disorders³⁹ and metabolic disruptions,¹⁹ so exposure in early life may contribute to these adverse health outcomes. Sphingomyelins serve important roles as immune modulators that regulate immune cell trafficking,⁴⁰ and the difference in direction of effect between mothers may also indicate immune suppression or activation in mothers and children, respectively. Overall, our analysis across strata indicated that the observed associations were consistent in the overall cohorts compared to treatment-only and placebo-only groups. However, it is important to consider in VDAART mothers that these associations may be modulated in response to vitamin D level, and vitamin D is an important consideration when evaluating these relationships as it have been previously shown to impact lipid metabolism.⁴¹

Amino acid metabolites were predominantly positively associated with PFAS exposure, and 6 amino acid pathways overlapped between VDAART mothers and children. The directions of effect were consistent between the two time points. Amino acid metabolism is broadly associated with energy generation,⁴² so, taken together, increased metabolism of these pathways in VDAART mothers may reflect increased bioenergetic demands and have important implications for overall health. Elevated levels of catabolic amino acid metabolites observed in VDAART mothers may be related to increase inflammatory processes, which have previously been linked to maturation and plasticity of neurons.⁴³ Methionine, Cysteine, SAM, and Taurine Metabolism has been previously identified for individual PFAS species, including PFOA, PFOS, PFNA, and PFHXS,³⁰ which is consistent with our findings. This pathway is involved in the regulation of a variety of biological functions,^44,45 but is also linked to lipid metabolism.⁴⁶ Increased methionine metabolism may indicate an effort to counteract the increased oxidative products that result from the shift towards oxidative lipid species. Urea cycle metabolism showed a high number of significant associations specifically in VDAART children, and alterations to the urea cycle are some of the most commonly reported metabolomic perturbations in response to PFAS.²⁹ Disruptions to the urea cycle may reflect inability to properly convert toxic ammonia to urea for excretion;⁴⁷ exposure to PFAS may cause a bioenergetic imbalance that negatively impacts normal urea cycle activity. However, Urea cycle metabolites were not significantly associated with PFAS mixture, suggesting that only specific PFAS species may be impact this pathway. Overall, disrupted amino acid metabolism suggested disrupted energy homeostasis, which could have negative impacts for overall health. While vitamin D levels may impact absorption of amino acids and amino acid metabolism,⁴⁸ we observed consistency across strata of vitamin D treatment during pregnancy. However, it is an important variable for consideration in follow-up studies.

There were several strengths and limitations to this study. Assignments of lipid and amino acid sub-pathways were performed solely by Metabolon and do not consider metabolite redundancy (i.e., the same metabolite participating in multiple biological pathways), a well-described problem in the field of metabolomics,⁴⁹ which may lead to over-emphasis of certain pathways. Additionally, while an untargeted analytical approach provides breadth, it does not ensure complete coverage of all metabolites within a pathway and cannot compare to the sensitivity provided by targeted assays. Our goal was to summarize global metabolic alterations rather than focusing on a set number of pathways in order to identify new pathways for consideration in PFAS exposures. Further, the focus of this manuscript was to highlight the global metabolomic response, which was dominated by lipid- and amino acid-related metabolites, but metabolites belonging to Carbohydrate, Cofactors/Vitamins, Energy, Nucleotide, Peptide, and Xenobiotics super-pathways generated some significant associations with PFAS exposures. In most cases, only a single metabolite was represented, so we chose to focus our analysis on lipid and amino acid metabolites. These pathways may warrant future consideration of the implications of PFAS exposure in targeted panels. Finally, we did not incorporate effects of diet into this study and did not have a sufficient sample size of mothers with complications such as gestational diabetes and preeclampsia that may affect the relationship between PFAS and metabolites, which may limit our results. Despite these limitations, these results successfully identified several lipid and amino acid biological pathways for consideration. Strengths of the study were concurrent collection of metabolomics and PFAS in the same plasma sample and the ability to adjust for important confounders related to PFAS exposure and metabolomics.

5. Conclusions

This study represents an investigation into untargeted, global metabolomic profiles of a pre-birth cohort of mothers and children in response to individual PFAS and total PFAS exposure burden. Our results indicated disrupted metabolism of both lipid and amino acid classes, and specifically implicated Sphingomyelins, Lysophospholipid, Long Chain Polyunsaturated Fatty Acid (n3 and n6), Fatty Acid- Dicarboxylate, and Urea cycle metabolism as particular pathways of interest that may play a large role in adverse health effects related to PFAS exposures. Further, as Sphingomyelins showed opposite trends in VDAART mothers and children, our results support age- and exposure-dependent trends in metabolic perturbations. This systematic investigation into mother-child pairs could have important implications to understand biological pathways and metabolomic drivers that are disrupted in response to PFAS.

Table 4.

Summary of lipid metabolites associated with total PFAS or individual PFAS in VDAART children.

VDAART Children
	PFAS mixture			Individual PFAS
Lipid Sub-Pathway	Number of Metabolites	Number in Pathway	% Significant in Pathway	Number of Total Associatio ns	Number of Unique Metabolite s	Number in Pathway	% Significant in Pathway
Phosphatidylcholine (PC)	9	14	64.29	32	10	19	52.63
Lysophospholipid	9	19	47.37	28	10	26	38.46
Sphingomyelins	8	29	27.59	20	8	29	27.59
Phosphatidylinositol (PI)	3	6	50	9	3	6	50
Fatty Acid, Dicarboxylate	2	27	7.41	9	6	27	22.22
Long Chain Polyunsaturated Fatty Acid (n3 and n6)	6	15	40	7	6	15	40
Fatty Acid Metabolism (Acyl Carnitine, Long Chain Saturated)	3	8	37.5	7	3	8	37.5
Plasmalogen	2	11	18.18	4	2	11	18.18
Lysoplasmalogen	4	4	100	4	4	4	100
Fatty Acid, Monohydroxy	1	16	6.25	4	3	16	18.75
Dihydrosphingomyelins	1	5	20	4	1	5	20
Phosphatidylethanolamine (PE)	1	12	8.33	3	1	12	8.33
Long Chain Saturated Fatty Acid	2	7	28.57	3	3	7	42.86
Carnitine Metabolism	1	2	50	3	1	2	50
Androgenic Steroids		14		3	3	14	21.43
Progestin Steroids	1	5	20	2	1	5	20
Primary Bile Acid Metabolism		10		2	2	10	20
Pregnenolone Steroids	2	6	33.33	2	2	6	33.33
Fatty Acid, Dihydroxy	1	5	20	2	1	5	20
Fatty Acid, Branched	1	4	25	2	1	4	25
Fatty Acid, Amino	1	3	33.33	2	1	3	33.33
Fatty Acid Metabolism (Acyl Carnitine, Monounsaturated)		4		2	1	10	10
Ceramides		2		2	2	4	50
Sphingosines	1	2	50	1	1	2	50
Phospholipid Metabolism	1	7	14.29	1	1	7	14.29
Mevalonate Metabolism	1	1	100	1	1	1	100
Medium Chain Fatty Acid	1	9	11.11	1	1	9	11.11
Eicosanoid	1	1	100	1	1	1	100

The number of lipid metabolites significantly associated (FDR<0.05) with PFAS mixture or one of the six individual PFAS species are summarized by lipid sub-pathway. Categorization is consistent with Table 3.

Highlights.

• Age- and exposure-dependent trends were observed in metabolic perturbations to PFAS

• Seven lipid pathways were consistently altered across mothers and offspring

• Eight lipid pathways differed in direction of effect between mothers and offspring

• Amino acid metabolism pathways were consistently increased in mothers and children

Abbreviations:

EDTA

Ethylenediaminetetraacetic acid

FDR

False discovery rate

HHEAR

Human Health Exposure Analysis Resource

LOD

Limit of detection

PFAS

Per- and polyfluoroalkyl substances

PFOA

Perfluorooctanoic acid

PFOS

Perfluorooctanesulfonic acid

PFHXS

Perfluorohexane-1-sulphonic acid

PFNA

Perfluorononanoic acid

PFDEA

Perfluorodecanoic acid

PFUA

Perfluoroundecanoic acid

RPM

Revolutions per minute

SD

Standard deviation

UPLC-MS

Ultrahigh performance liquid chromatography-mass spectrometry

VDAART

Vitamin D Antenatal Asthma Reduction Trial