An ‘Omics Approach to Unraveling the Paradoxical Effect of Diet on Perfluorooctanesulfonic Acid (PFOS) and Perfluorononanoic Acid (PFNA)-Induced Hepatic Steatosis

Abstract

Perfluoroalkyl substances (PFAS) are a family of toxicants universally detected in human serum and known to cause dyslipidemia in animals and humans. Hepatic steatosis, which is defined as lipid deposition in the liver, is known to be a consequence of poor diet. Similarly, PFAS are known to induce hepatic steatosis in animals on a low-fat chow. This study explored diet-PFAS interactions in the liver and their potential to modulate hepatic steatosis. Male C57BL/6J mice were fed with either a low-fat diet (10% kcal from fat, LFD) or a moderately high-fat diet (45% kcal from fat, HFD) with or without perfluorooctanesulfonic acid (3 ppm, PFOS) or perfluorononanoic acid (3 ppm, PFNA) in feed for 12 weeks. Livers were excised for histology and quantification of PFAS and lipids. The PFOS and PFNA coadministration with HFD reduced the hepatic accumulation of lipid and PFAS relative to the LFD treatment groups. Furthermore, transcriptomic analysis revealed that PFAS administration in the presence of an HFD significantly reduces expression of known hepatic PFAS uptake transporters, organic anion transporter proteins. Transcriptomics and proteomics further revealed several pathways related to lipid metabolism, synthesis, transport, and storage that were modulated by PFAS exposure and further impacted by the presence of dietary fat. Both dietary fat content and the chemical functional head group exerted significant influence on hepatic PFAS accumulation and the resulting biochemical signature, suggesting that diet and structure should be considered in the design and interpretation of research on PFAS induced hepatic steatosis.

According to the American Liver Foundation, about one-quarter of the American population suffers from diet-induced hepatic steatosis, or fatty liver, representing the most common form of liver disease. Furthermore, fatty liver serves as a risk factor for other adverse health outcomes, such as cardiovascular disease (Ismaiel and Dumitraşcu, 2019). Obesity, metabolic syndrome, and diabetes are all known risk factors for hepatic steatosis, further environmental toxicants have been known to induce fatty liver onset (Joshi-Barve et al., 2015). The lack of serum biomarkers to diagnose hepatic steatosis and the need to acquire thin needle biopsy samples present a challenge in assessing the human exposures that cause hepatic steatosis. Thus, rodent models are often used as surrogates to understand the relationship between environmental exposures and fatty liver endpoints.

Perfluoroalkyl substances are a family of man-made chemicals used in manufacturing of Teflon, aqueous film-forming foams, food packaging, and stain resistant sprays for their unique surfactant and antistick properties (Buck et al., 2011). These long-chain fluorinated structures are extremely resistant to degradation leading to their accumulation in water sources, dust, and even ambient air (Hu et al., 2016; Winkens et al., 2018; Barber et al., 2007). In a study conducted by the Centers for Disease Control and Prevention (CDC), it was revealed that four PFAS members have been nearly ubiquitous in human serum since at least 1999 including, among others, perfluorooctanesulfonic acid (PFOS) and perfluorononanoic acid (PFNA) (National Report on Human Exposure to Environmental Chemicals | CDC, 2020; Olsen et al., 2017). The PFOS and PFNA are slow to excrete from the human body with half-lives spanning several years, 5.4 and 4.3 years, respectively (Olsen et al., 2007; Zhang et al., 2013). Once in the body, PFOS and PFNA distribute mainly to protein-rich body compartments, such as serum and liver (Jian et al., 2018). This distribution pattern leads to relatively high exposure of PFAS to hepatocytes, increasing potential risk for hepatic toxicity.

Perfluoroalkyl substances have been associated with the following adverse human health effects: Impaired immune function (Grandjean et al., 2012) elevated serum cholesterol (Nelson et al., 2010), thyroid disease (Ballesteros et al., 2017), low fetal birth weight (Shoaff et al., 2018), elevated serum markers of liver injury (Salihovic et al., 2018), and even kidney and testicular cancer (Nicole, 2013). The PFAS are significantly associated with elevated serum alanine aminotransferase, however, the mechanism of liver injury remains uncertain. Studies conducted in mice (Das et al., 2017), rats (Curran et al., 2008), and cynomolgus monkeys (Seacat et al., 2002) have demonstrated that some PFAS members augment lipid accumulation and steatosis in the liver at high doses. Likewise, PFAS have been demonstrated to induce lipid accumulation and perturb lipid metabolism in human hepatocytes (Bjork et al., 2011; Rosen et al., 2013). To date there is only one study concerning the long-term effects of relatively low-level exposure to common PFAS, in combination with dietary risk factors, and their potential role in augmenting the onset of hepatic steatosis (Huck et al., 2018). Huck et al. reported that PFOS exhibited a surprising preventative effect against fatty liver in the presence of a high-fat diet (HFD). The present study is the first to confirm this finding, expand it to include PFNA, and delve further into the mechanistic drivers of diet-PFAS interactions.

Understanding the potential health effects of PFAS members plays a vital role in guiding the risk assessments that determine federal health advisory levels in drinking water. Due to emerging toxicity data, the EPA health advisory for PFOS in drinking water was lowered to 70 ppt in 2016 (US EPA, 2016). This health advisory level was derived from a no observable adverse effect level (NOAEL) dose of 0.1 mg/kg/day obtained from rodent toxicity studies evaluating the effects of developmental PFOS exposure on pup weight and mortality (Luebker et al., 2005a,b). The PFAS manufacturing companies, such as 3M, have voluntarily removed PFOS from manufacturing. However, PFNA can still legally be used as a PFAS replacement in manufacturing and consumer products. Despite the high potential for similar toxicity, the EPA has not yet released a health advisory for PFNA. Both PFOS and PFNA continue to be highly prevalent in the tissues of humans, wildlife, and the environment today (Jian et al., 2018). This study investigates the potential consequences of chronic relatively low-level PFOS or PFNA exposure concurrently with diet-induced hepatic steatosis. Our work also characterizes and compares the hepatic signatures of the sulfonic acid, PFOS, and the carboxylic acid, PFNA while exploring the additional impact of diet interactions.

MATERIALS AND METHODS

 

 

Chemicals and reagents

The PFOS (catalog number 33829-100 mg) and PFNA (catalog number 394459-5G) chemical stocks were purchased from Sigma-Aldrich (St Louis, Missouri). The internal standards used for liquid chromatography with tandem mass spectrometry (LC-MS/MS) included C9 labeled PFNA (catalog number CLM-8060-1.2, Cambridge Isotope Laboratories, Andover, Massachusetts) and C4 labeled PFOS (catalog number MPFOS, Wellington Laboratories, Ontario, Canada).

Animals

The study was conducted at the University of Rhode Island under an approved Institutional Animal Care and Use Committee (IACUC) protocol. C57BL/6J mice were acquired from Jackson Labs (Bar Harbor, Maine) at 6-weeks-old and were acclimated for 2 weeks prior to being weight paired and housed four per cage. The mice were housed in a temperature-controlled room and kept on a strict 12-hour dark/light cycle with access to food and water ad libitum. Body weights and food consumption were monitored weekly. Following 12 weeks of diet administration, all mice were anesthetized using isoflurane and euthanized by cervical dislocation. Tissues were immediately harvested and snap frozen in liquid nitrogen. Gross liver weight was recorded prior to sectioning in 10% formalin for histology. The remaining liver was snap frozen for downstream analysis.

The PFOS and PFNA dosing in feed

Mice were fed either a 10% kcal, low-fat diet (LFD, D12450B Research Diets, New Brunswick), or a matched 45% kcal, moderately HFD (D12451 Research Diets, New Brunswick). To mimic the major route of PFAS exposure, oral consumption, mice were assigned to diet alone or diet containing either 0.0003% PFOS or 0.0003% PFNA and fed ad libitum for 12 weeks. The resulting treatment groups were as follows: LFD, HFD, LFD + PFOS (LPFOS), HFD + PFOS (HPFOS), LFD + PFNA (LPFNA), and HFD + PFNA (HPFNA) with n = 8 per treatment group. The daily exposure to PFAS via diet was roughly approximately 0.24 mg/kg/day based on an overall average daily food consumption of 2.37 g and average body weight of 29.89 g over the course of the study. In the current EPA health advisory document for PFOS, 0.1 mg/kg/day was considered the NOAEL dose for PFOS-induced developmental toxicity (US EPA, 2016; Luebker et al., 2005a).

Hepatic lipid isolation and analysis

Liver lipids were isolated from approximately 50 mg of liver tissue using the Folch chloroform-methanol extraction (Folch et al., 1957). Triacylglyceride (TAG) and total cholesterol were measured using kits from Pointe Scientific (Ann Arbor, Michigan). Total non-esterified free fatty acids (FFAs) were quantified using a kit from Wako Chemicals (Richmond, Virginia). Phospholipids were quantified using the EnzyChrom phospholipid colorimetric assay kit (BioAssay Systems, Hayward, California) according to kit instructions. Liver tissue sections were fixed in 10% buffered formalin prior to paraffin embedding. Paraffin sections (5 μm) were cut and stained with hematoxylin and eosin (H&E). Stains and scoring were conducted by Rutgers University Research Pathology Services (Piscataway, New Jersey). The histopathological classification was made by a board-certified pathologist. Scores ranged from 0 to 5, with 0 being the least and 5 the most severe.

Quantification of PFAS content in liver

The LC-MS/MS was used to quantify hepatic PFAS concentrations and to further explore the effect of diet on internal distribution of PFAS to the liver. The PFAS were isolated from liver using an adapted 3M method published by Chang et al. (2017). Roughly 100 mg of tissue was homogenized in 4× LC-MS grade water spiked with internal standards. Of the homogenate, 250 μl was transferred to a new tube containing 10% 1 N KOH and digested overnight. Following digestion, 100 μl of sample was combined with 100 μl 1 N formic acid, 500 μl 2 N HCL, 500 μl saturated ammonium sulfate, and 5 ml LC-MS grade MTBE. The samples were shaken and then centrifuged for 5 min at 2500 × g. For each sample, 4.5 ml of the organic layer was transferred to a fresh tube and evaporated overnight. The dried samples were resuspended in 10 ml acetonitrile:water and shaken for 30 min at room temperature. The samples were filtered through a 0.2 μm syringe filter prior to injection on the instrument. The LC-MS/MS analysis was run in negative mode on a QTRAP 4500 LC-MS/MS System coupled to an SHIMADZU Prominence UPLC (SCIEX, Framingham, Massachusetts). A Waters XBridge C18 (100 mm × 4.6 mm, 5 μm) column was used. Sample injection volume was 10 μl with a flow rate of 0.6 ml/min.

RNA preparation and transcriptomics

RNA was isolated from roughly 50 mg of hepatic tissue using the Trizol method. The RNA was quantified and checked for purity on a ThermoFisher Nanodrop 1000 and diluted with DEPC water to equal concentrations. RNA integrity was measured on an Agilent Bioanalyzer using an Agilent RNA 6000 Nano kit. Only pure, intact RNA samples were used for downstream analysis, RIN score of ≥8. An Affymetrix mouse ST 2.0 global array (Affymetrix, Waltham, Massachusetts) was conducted according to the manufacturer’s protocols. The array was conducted by the Genomics Core Facility at Brown University (Providence, Rhode Island). Raw .cel files were uploaded into the Transcriptome Analysis Console (TAC) version 4.0.1 (ThermoFisher, Waltham, Massachusetts). The TAC software was used to conduct data normalization, quality control, and differential expression analysis (GEO accession number GSE138602). The data were filtered using the criteria of ≥1.5-fold change and p-value <.05. Pathway analysis was conducted using the ingenuity pathway analysis (IPA) database from Qiagen (Krämer et al., 2014). The upstream analysis feature was used to predict upstream regulators. The IPA’s comparison analysis tool was used to compare predicted activation and inhibition between treatment groups. In addition to the untargeted global array, a targeted assessment of key genes was conducted.

QuantiGene plex targeted gene expression

Targeted gene expression analysis was conducted using a custom QuantiGene Luminex xMAP gene expression panel (ThermoFisher) using 0.5 µg of total RNA input. The multiplex panel was performed according to the manufacturer’s protocols with mean fluorescence intensity quantified using a Bio-plex 200 instrument (Bio-rad, Hercules, California). Intensity values were normalized to housekeeper gene beta-actin (Actb) and converted to fold change relative to the control. Hepatic genes involved in inflammation, lipid uptake, lipid metabolism, and lipid regulation were analyzed. Genes of interest in the liver included Acaca, Acot2, Ccl2, Cd36, Cpt1a, Cpt1b, Csf2ra, Cyp4a14, Ehhdh, Gstm3, Fabp1, Cidea, Fabp4, Fas, Gapdh, Gclc, Gpam, Gusb, Hmgsc1, IL6, Lpl, Nrf2, Nqo1, Pparα, Pparγ, Scd1, Slc27a1, Sod1, Srebf1, Tnfα, Actb, Hprt, Ppia, and Ppib (Supplementary Table 1).

Protein digestions and SWATH-MS proteomics

Protein samples were obtained from 50 mg of liver tissue homogenized in 1 ml of homogenization buffer (0.25 µM sucrose, 10 mM Tris-HCL [pH 7.8], 0.5 mM ethylenediaminetetraacetic acid, 10% vol/vol glycerol, 1 mM dithiothreitol [DTT], 20 µM butylated hydroxytoluene). Protein samples were centrifuged at 10 000 × g for 20 min at 4°C. The supernatant containing the S9 fraction enriched for microsomal and cytosolic proteins was transferred to a fresh tube. The S9 protein was quantified using the Pierce BCA Protein Assay Kit (ThermoFisher Scientific, Rockford, Illinois) according to the manufacturer’s instructions. Stock S9 protein samples (5 mg/ml) were diluted to 2.5 mg/ml. Protein (200 µg) was spiked with 2 µg BSA and denatured with 20 µl DTT (100 mM) at 95°C for 15 min in a shaking water bath (100 rpm). After denaturation, samples were alkylated in the dark with 20 µl indole-3-acetic acid (200 mM) for 30 min at room temperature. Subsequently, 40 µl of 50 mM ammonium bicarbonate was added to each sample was pH was confirmed to be alkaline. Furthermore, N-tosyl-L-phenylalanine chloromethyl ketone (TPCK)-treated trypsin (10 µg) was added to samples at a ratio of 1:20 (trypsin:protein) and 150 µl of the resulting solution was transferred into digestion tube (PCT MicroTubes, Pressure Biosciences Inc, Easton, Massachusetts). The barocycler was run at 50°C, for 90 cycles with 60 s per pressure-cycle (50 s high pressure, 10 s ambient pressure, 25 kpsi). Furthermore, to 145 µl of digested peptide sample, 5 µl of acetonitrile (ACN)(1:1, vol/vol containing 5% formic acid) was added to acidify the samples and quench trypsin reaction. Samples were spun to remove any precipitate and 125 µl supernatant was collected (10 000 rpm for 5 min at 10°C). Subsequently, 20 µl of the resulting peptide solution was injected into the analytical column and samples were analyzed using LC-MS/MS. SWATH-DIA proteomics was conducted as previously published (Jamwal et al., 2017) The proteomics was run on an Acquity UHPLC HClass system (Waters Corp, Milford, Massachusetts) coupled to an SCIEX 5600 TripleTOF mass spectrometer (SCIEX, Concord, Canada). The method used a run time of 60 min at 100 μl/min and a linear gradient. Global protein changes were assessed using MaxQuant (Tyanova et al., 2016a) and Perseus (Tyanova et al., 2016b). Targeted data analysis was conducted on the open-source software Skyline (MacLean et al., 2010) (MacCoss Lab Software) and normalized to peptide concentration and BSA spiked control. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD015977.

Statistical analysis

Body weight, tissue weight, and hepatic PFAS concentrations are shown as mean ± standard error (SEM). Unless otherwise indicated, data were analyzed using 1-way analysis of variance (ANOVA) followed by the Bonferroni post hoc test to correct for multiple comparisons, where p < .05 was denoted as statistically significant. Calculations were performed using Graphpad Prism (Graphpad Prism Software for Windows Ver 8.0, La Jolla, California).

RESULTS

 

Perfluorooctanesulfonic Acid and PFNA Exert Differential Effects on White Adipose Tissue, Liver, and Body Weight

Mice were weight matched prior to assignment to treatment group. Body, liver, and white adipose tissue (WAT) weight increased approximately 41%, 36%, and 144%, respectively, within the HFD controls relative to the LFD controls (Table 1). Interestingly, both PFOS and PFNA caused significant modulation of tissue and body weights, but the effects differed based on the compound. For example, the LPFOS group produced a 44% increase in WAT weight whereas LPFNA treatment resulted in a 44% decrease relative to the LFD controls. Furthermore, PFNA produced a dramatic change in liver weight with increases of 155% and 73%, relative to the LFD and HFD control, whereas PFOS only increased liver weight by 18% and 13%, respectively. Lastly, we observed a significant discrepancy in weight gain between the PFOS- and PFNA-treated mice within both diets, with HPFNA-treated mice gaining 5.5 g less and LPFNA gaining 2.9 g less relative to their PFOS-treated counterparts.

Table 1.

Body and Tissue Weights

	Unit	LFD	LPFOS	LPFNA	HFD	HPFOS	HPFNA
Body weight	g	30.2 ± 1.1#	30.6 ± 1.2#,&	27.7 ± 0.7#,&	42.6 ± 1.4*	40.5 ± 0.3*,$,&	35.0 ± 1.0*,$,&
WAT weight	g	0.9 ± 0.2#	1.3 ± 0.1*,#,$,&	0.5 ± 0.1*,#,$,&	2.2 ± 0.1*	2.3 ± 0.1*,$,&	1.6 ± 0.1*,#,$,&
Liver weight	g	1.1 ± 0.1	1.3 ± 0.1$	2.8 ± 0.3*,#,$	1.5 ± 0.1	1.7 ± 0.0$	2.6 ± 0.2*,#,$
Liver: BW	%	4.4 ± 0.7	4.3 ± 0.1$,&	10.2 ± 1.0*#$&	3.4 ± 0.3	4.1 ± 0.1$,&	7.4 ± 0.5*,#,$,&
WAT: BW	%	2.9 ± 0.4#	4.3 ± 0.2*,#,$,&	1.7 ± 0.3*#$&	5.3 ± 0.3*	5.7 ± 0.3*,$,&	4.6 ± 0.2*,$,&
Weight gain	%	25.1 ± 4.5#	26.8 ± 4.3#,$,&	18.7 ± 1.9#$&	79.1 ± 5.2*	66.4 ± 2.6*,$,&	48.2 ± 7.0*,$,&

Male C57BL/6 mice were fed with either a low-fat diet or a moderately high-fat diet with or without PFOS or PFNA (0.0003% wt/wt in feed) for 12 weeks (n = 8). After euthanization, gross body and organ weights were recorded and analyzed.

^* p < .05, significant in comparison to the LFD control.

^# p < .05, significant in comparison to the HFD control.

^$ p < .05, significance between PFOS and PFNA within the same diet (ie, L/HPFOS vs L/HPFNA).

^& p < .05, significance between diet treatment within the same compound (ie, LPFAS vs HPFAS).

Dietary Fat Has a Significant Impact on PFAS Modulation of Hepatic Lipid Content

An excess of dietary fatty acids within the liver can augment formation of hepatic lipid droplets and trigger the onset of hepatic steatosis. To explore the potential interaction of dietary fat with PFOS and PFNA modulation of hepatic lipid content, liver lipids were extracted and TAG, FFA, cholesterol, and phospholipids were quantified (Figure 1). In LFD mice, the FFA levels in liver were similar between all groups. As expected, the HFD increased FFA concentrations in liver by 60.6%. However, this increase in FFA content was not observed in HPFOS or HPFNA livers, which were similar to LFD controls (Figure 1A). Likewise, TAG concentrations were similar between the LFD treatments. The HFD increased TAG content by 54.3% and HPFOS by 92.6% relative to LFD and LPFOS, respectively. An HPFNA decreased triglyceride content by 28.4% relative to the HFD control (Figure 1B). Liver cholesterol was similar between all treatment groups (Figure 1C). However, PFAS treated reduced hepatic phospholipids. Within the LFD, PFOS, and PFNA decreased hepatic phospholipid content, by 34.3% and 15.6%. This effect was conserved within the HFD with reduction by 31.6% and 33.6%, respectively (Figure 1D). To tabulate hepatic steatosis, liver sections were collected at necropsy and scored by a board-certified pathologist for lipid accumulation (Table 2).

Figure 1.

The PFOS and PFNA reduce hepatic phospholipid concentrations. Lipid moieties were quantified from approximately 50 mg hepatic tissue and normalized to liver weight, expressed as milligrams per grams. Lipids measured included: A, Non-esterified fatty acids (NEFA); B, triglycerides; C, cholesterol; and D, phospholipids. Statistical significance was calculated using 1-way ANOVA where *: significant from the LFD control; #: significant from the HFD control; $: significance between PFOS and PFNA within each diet; and &: significance between diet within each compound.

Table 2.

Lipid Accumulation Scores

Treatment	LFD	LPFOS	LPFNA	HFD	HPFOS	HPFNA
0	3/8	2/8	0/8	0/8	0/7	0/8
1	5/8	3/8	4/8	0/8	1/7	2/8
2	0/8	3/8	4/8	2/8	3/7	5/8
3	0/8	0/8	0/8	5/8	3/7	1/8
4	0/8	0/8	0/8	1/8	0/7	0/8
% ≥2	0%#	38%#,&	50%*,#	100%*	86%*,&	75%*,#

The lipid accumulation score assigned out of total mice per group. Assigned scores ranged from 0 to 4, with 0 being the least and 4 the most severe. Statistical analysis was performed using Kruskal–Wallis test to derive ranks followed by Dunn's multiple comparison test for multiple comparisons using GraphPad Prism software v8.2.0 (La Jolla, CA).

^* p < .05, significant in comparison to the LFD control.

^# p < .05, significant in comparison to the HFD control.

^$ p < .05, significance between PFOS and PFNA within the same diet (ie, L/HPFOS vs L/HPFNA).

^& p < .05, significance between diet treatment within the same compound (ie, LPFAS vs HPFAS).

Representative images using hematoxylin and eosin (H&E) staining for each treatment group are shown in Figure 2A. Arrows highlight hepatic lipid droplets indicative of steatosis. The HFD controls exhibited clear macrovesicular steatosis and overall significantly higher scoring for lipid accumulation when compared with the LFD control. The LFD controls exhibited healthy liver pathology with 0% of the mice achieving a lipid accumulation score greater than or equal to 2. Within the LFD fed groups, PFOS and PFNA caused an expected increase in liver lipid accumulation, with 38% and 50% scoring ≥2. The observed increase in microvesicular steatosis was significant within the LPFNA group. However, within the HFD, PFNA exposure resulted in a slight decrease in scoring for liver lipid accumulation when compared with HFD controls. Out of 8 mice, 100% of the HFD controls scored ≥2, whereas only 86% for PFOS and 75% for PFNA. Visually, there was a reduction in macrovesicular steatosis within the HPFOS- and HPFNA-treated mice when compared with the HFD controls. Figure 2B depicts the lipid accumulation scores as ranks. All HFD groups, HFD, HPFNA, and HPFOS, exhibited significantly higher lipid accumulation relative to the LFD control. An HPFOS exhibited significantly more lipid accumulation that LPFOS, however, there was no significant difference between LPFNA and HPFNA. Within the HFD, PFNA significantly reduced hepatic lipid accumulation relative to HFD controls.

Figure 2.

Dietary fat determines PFOS and PFNA modulation of hepatic steatosis. A, Representative H&E stained liver sections. B, Scores were converted to ranks and graphed as scatter plots. The horizontal line denotes the mean and the vertical lines represent the standard error from the mean. Statistical significance was calculated using 1-way ANOVA where *: significant from the LFD control, #: significant from the HFD control, $: significance between PFOS and PFNA within each diet and &: significance between diet within each compound.

Perfluorooctanesulfonic Acid and PFNA Modulation of the Hepatic Transcriptome is Significantly Affected by the PFAS Chemical Functional Head Group

Hepatic gene expression was explored to determine the consequence of PFOS and PFNA structure, differing by functional head group, in the liver (Figure 3). Hepatic RNA was prepared for an untargeted transcriptomic array, detecting 34 472 total genes. Between 1303 and 1424 hepatic genes were significantly modulated (p < .05) by PFOS or PFNA by at least 1.5-fold. To explore the top genes impacted by PFAS exposure, the top 5 global genes modulated by diet-PFAS treatment were identified (Table 3). Cyp2b10, a gene involved in the oxidation of fatty acids, steroids, and xenobiotics, was the top impacted gene in 3 out of the 4 PFAS groups: LFD versus LPFOS (59.1-fold), LFD versus LPFNA (27.0-fold), and HFD versus HPFOS (33.5-fold). Cidea, a gene involved in lipid droplet enlargement, was heavily modulated by the HPFNA group with a measured global fold change of 88.1 relative to the HFD control. Although the total number of genes modulated by PFNA or PFOS are similar, the transcriptomic signatures, as measured by overlap between genes impacted, were remarkably unique. Moreover, scatter plots directly comparing the log2 of the average signal intensity for PFOS versus PFNA highlight the structure-specific modulation in red and green.

Figure 3.

The PFOS and PFNA exhibit diet- and structure-dependent effects on the hepatic transcriptome. Global transcriptomic analysis was conducted on n = 3 samples from each group using a Mouse ST 2.0 transcriptomic array from Affymetrix detecting 34 472 total genes. The data were analyzed using Transcriptome Analysis Console (TAC) software by Affymetrix. The cutoff values were set at 1.5-fold change or greater and a p-value <.05. The raw data are available on the GEO database accession # GSE138602. A, Describes the overlap between PFOS and PFNA modulated genes relative to LFD dietary controls; B, Scatter plots were generated using TAC software and show direct differential expression based on the log2 of the average signal for each LPFOS and LPFNA treatment group. C, Each Ven Diagram shows the overlap between PFOS and PFNA modulated genes relative to HFD dietary controls. The genes on the left side are the top 10 induced genes, whereas the genes on the right side are the top 10 suppressed genes. D, Scatter plots were generated using TAC software and show direct differential expression based on the log2 of the average signal for each HPFOS and HPFNA treatment group.

Table 3.

Top Global Hepatic Genes

Gene	Fold Change	ANOVA P-Value	Gene	Fold Change	ANOVA P-Value
L Versus LPFOS			L Versus LPFNA
Cyp2b10	59.14	0.000021	Cyp2b10	26.99	0.005249
Cyp2c55	15.86	0.000059	Mogat1	7.06	0.037019
Cyp2c53	9.83	0.003011	Serpinb1a	7.02	0.001564
Gstm3	8.79	0.00236	Gstm3	6.95	0.003246
Cyp2b13	6.94	0.039922	Slc271	6.45	0.015992
H Versus HPFOS			H Versus HPFNA
Cyp2b10	33.48	0.000018	Cidea	88.09	0.009264
Gstm3	6.76	0.009034	Sptlc3	42.16	0.000055
Cyp2c55	6.23	0.000548	Fabp3-ps3	38.98	0.000156
Cyp4a12b	4.59	0.000737	Gm15441	22.9	0.000037
Gstt3	4.33	0.000259	Otop1	11.12	0.00137

Transcriptomic array fold change and p-values were derived from the Transcriptome Analysis Console (TAC). The top genes were identified through the use of IPA (QIAGEN Inc). The top 5 differentially expressed genes are summarized for each PFAS-diet comparison. Cyp2b10 was the top modulated gene in 3 out of 4 comparisons.

Within the LFD, structure-specific effects were observed. Compared with the LFD, LPFNA, and LPFOS modulated a total of 1405 and 1424 hepatic genes, respectively. Out of the total number of differentially expressed genes, only 799 genes, were shared between PFOS and PFNA. This suggests that there are both shared and unique genes modulated by PFOS and PFNA within in the liver. The top genes induced or repressed within each comparison are summarized within each box. Figure 3B shows a direct comparison of the average signal (log2) of LPFNA to LPFOS revealing 260 differentially expressed genes. Compared with LPFOS, LPFNA upregulated RNA associated genes Ddx3y, Mt-tq, and Supt4b by 4.7-, 3.4-, and 2.5-fold, respectively. Overall, LPFNA suppressed keratinization and immune response-related genes, whereas upregulating genes involved in RNA processing and transcription when compared with LPFOS.

Structure-dependent effects on hepatic gene expression were more pronounced within the HFD treatment groups. Out of 1368 and 1303 genes modulated by HFD versus HPFNA and HFD versus HPFOS, only 647 of those genes were conserved between the treatment groups (Figure 3C). Roughly half of the genes modulated by PFOS or PFNA were unique to the compound’s functional head group, 656 and 721, respectively. Figure 3D depicts 764 differentially expressed genes when comparing HPFNA to HPFOS directly. Cidea was the top differentially expressed gene, induced 73.7-fold times higher within the HPFNA treatment group compared with HPFOS. Furthermore, HPFNA caused relative induction of Fabp3, a known fatty acid and potential PFAS uptake transporter, by 52.3-fold and Spltc3, involved in the rate limiting step of sphingolipid biosynthesis, by 34.5-fold. In contrast, HPFNA reduced expression of Hsd3b5, involved in the biosynthesis of active steroids, by 21.4-fold in comparison to HPFOS. In addition, HPFNA reduced the expression of Oatp1a1 and Oatp1a4, known PFOS uptake transporter genes, by 16.6- and 11.3-fold in comparison to HPFOS. These robust changes in expression within the same base diets confirms a significant and PFAS-specific alteration to the hepatic transcriptome, attributable to the functional head group.

Perfluorooctanesulfonic Acid and PFNA Modulation of the Hepatic Transcriptome is Significantly Impacted by Dietary Fat

A notable diet effect was observed on hepatic transcriptomic expression. When comparing the PFOS impacted genes within each diet only 617 were shared between LPFOS and HPFOS compared with control diet (Figure 4A). When directly comparing HPFOS to LPFOS gene expression, 824 differentially expressed genes were identified (Figure 4B). Cidea, a gene involved in lipid droplet enlargement, was the top differentially expressed gene, induced 73.7-fold times higher within the HPFOS treatment group compared with LPFOS. Furthermore, HPFOS caused relative induction of Fabp3, a known fatty acid and potential PFAS uptake transporter, by 52.3-fold and Spltc3, involved in the rate limiting step of sphingolipid biosynthesis, by 34.5-fold. In contrast, HPFOS reduced expression of Hsd3b5, involved in the biosynthesis of active steroids, by 21.4-fold in comparison to LPFOS. In addition, HPFOS reduced the expression of Oatp1a1 and Oatp1a4, known PFOS uptake transporter genes, by 16.6- and 11.3-fold relative to LPFOS. Overall, diet modulated PFOS induced expression of genes involved in lipid and PFOS uptake, lipid and steroid synthesis, and lipid droplet enlargement.

Figure 4.

The PFOS and PFNA exhibit diet- and structure-dependent effects on the hepatic transcriptome. The raw data are available on the GEO database accession number GSE138602. A, The Venn diagram summarizes the overlap between LPFOS and HPFOS modulated genes relative to dietary controls. The genes on the left side are the top 10 induced genes, whereas the genes on the right side are the top 10 suppressed genes. B, Scatter plots were generated using TAC software and show direct differential expression based on the log2 of the average signal for LPFOS versus the HPFOS treatment group. C, The Venn diagram summarizes the overlap between LPFNA and HPFNA modulated genes relative to dietary controls. The genes on the left side are the top 10 induced genes, whereas the genes on the right side are the top 10 suppressed genes. D, Scatter plots were generated using TAC software and show direct differential expression based on the log2 of the average signal for LPFNA versus the HPFNA treatment group.

The pronounced diet effect observed with PFOS exposure was conserved in the PFNA-treated groups. LFD versus LPFNA modulated 824, whereas HFD versus HPFNA impacted 787 genes unique to the diet. As measured by scatter plot, the direct comparison of LPFNA to HPFNA based on the log2 of the average signal intensity yielded 279 differentially expressed genes. Compared with LPFNA, HPFNA downregulated RNA genes, Snord92 and Ddx37, by 4.4-fold and metalloprotease, Adam11 by 2.2-fold. In contrast, HPFNA caused relative induction of olfactory-related genes, Obp2a and Vmn2r109, and the RNA gene, Traj58, by 4.5-, 3.0-, and 2.4-fold. Based on the expression data, it is clear that diet exerted a significant influence on PFAS modulation of hepatic transcriptome.

Perfluorononanoic Acid is a Potent Inducer of Lipid Metabolism, Transport, and Accumulation Genes

Targeted gene expression of key genes involved in lipid metabolism, synthesis, transport, and oxidative stress was measured (Figure 5). The targeted panel of hepatic genes, names, and functions is further described in Supplementary Table 1. Within the targeted data, compound and diet-specific effects between PFOS and PFNA were observed. In the case of Acaca, PFOS caused a 0.8-fold reduction in expression yet PFNA caused 1.7-fold induction of the same gene. Furthermore, we observed a clear diet effect for genes such as Lpl and Cidea. Within the LPFOS treatment group a 3.3-fold induction of Lpl was observed, however, the HPFOS mice exhibited a 7.6-fold induction, over twice that of the LPFOS group, despite no observed diet induction within the HFD controls. Likewise, Cidea showed diet-specific effects in both PFOS- and PFNA-treated mice. Within the LPFOS group, there was no significant induction of Cidea, however, the HPFOS group experienced an 8.5-fold induction in Cidea. Both PFOS and PFNA affected lipid homeostasis and oxidative stress related genes and pathways related to perturbed hepatic lipid content on both a targeted and global scale. Compared with PFOS, PFNA exposure produced 10- to 100-fold stronger induction of several genes including Acot2, Cpt1b, and Cidea when administered at the same dose. Overall, PFNA caused stronger induction of hepatic genes relative to PFOS.

Figure 5.

The PFOS and PFNA significantly modulate many gene expression related to lipid accumulation, metabolism, and oxidative stress on the targeted level. A targeted Quantigene Plex panel measuring 36 genes was measured on a Bioplex 2.0 system. Fluorescence intensity values were normalized to beta actin as a housekeeper. The value in each cell represents the average fold change relative to the LFD control for n = 5 samples.

Diet and PFAS interactions cause differential activation of upstream transcriptional drivers and downstream pathways that can modulate the onset and progression of steatosis. Looking further into the mechanistic function of the untargeted hepatic genes perturbed by PFAS exposure, upstream pathway regulators were assessed for potential activation or inhibition (Figure 6). Both PFOS and PFNA significantly modulated pathways regulated by Ppar, Pxr, Car, Zbtb, and Ahr. Perfluorononanoic acid selectively modulated Pml whereas PFOS exerted no effect on this pathway. Interestingly, HPFOS and HPFNA mediated stronger induction of Ppar$\alpha$ compared with the HFD controls as well as LPFOS and LPFNA, respectively. Ppar$\alpha$ is a key upstream regulator for lipid homeostasis in the liver- and diet-specific modulation of this activator could be a driver for differential hepatic lipid outcomes. A diet-specific response was also observed in the coactivator, Ncoa2. Within the HFD, PFOS, and PFNA exposure suppressed signaling of Ncoa2, whereas in the HFD control and LPFOS, an increased z-score was observed. The data suggest that PFNA and PFOS preferentially activate or enhance the activity of key upstream pathway drivers, such as Ppar$\alpha ,$ when combined with the HFD. In addition, activation of Nrf2 was observed only within the PFNA groups, suggesting structure-specific activation. The summary of key transcriptional drivers in response to PFAS exposure and their impact on steatosis progression is summarized in Table 4.

Figure 6.

The PFOS and PFNA significantly modulate many transcriptional drivers related to lipid homeostasis within the liver. Upstream analysis was used to predict transcriptional drivers. Orange signifies activation and blue signifies predicted upstream regulator inhibition. Activation or inhibited or a transcriptional regulator was predicted based on the expression state of the downstream targets and a z-score for activation or inhibition was assigned using Ingenuity Pathway Analysis software by Qiagen.

Table 4.

Effects of Upstream Transcriptional Regulators on NAFLD Onset

	Ppara	Zbtb20	Pxr	Car	Ahr	Lxr
HFD/LFD	−1.44	1.63	1.98*	2.62**	1.66
LPFOS/LFD			2.55*	1.29	0.27	1.93
LPFNA/LFD	3.17*	1.89	1.41	0.77	−0.73	1.93
HPFOS/LFD	0.17	2.89**	2.95**	2.79**	0.27	2.16*
HPFNA/LFD	4.82**	−0.50	2.34*	−0.03	2.02*	2.17*
HPFOS/HFD	1.49	1.51	1.72	1.32	−1.28	2.00*
HPFNA/HFD	5.00**	−0.91	1.81	−0.43	−2.23*	1.93
Steatosis	(−)	(+)	(+)	(−)	(−)	(+)
	Kim et al. (2017)	Liu et al. (2017)	Zhou et al. (2006)	Yamazaki et al. (2007)	Wada et al. (2016)	Jung et al. (2011)

Treatment activation or inhibition of upstream regulators based on IPA derived z-scores using the upstream analysis feature. Each z-score represents the number of deviations from the mean, whereas a blank well represents no prediction.

^*Significance at ≥ the 95% confidence level.

^**Significance at ≥ the 99% confidence level. The impact of pathway activation on hepatic steatosis is summarized in the steatosis row, (−) denotes attenuation and (+) augmentation of this endpoint. Literature sources are included below.

Downstream pathway outcomes were predicted using the core analysis feature and outcomes were compared using comparison analysis within IPA (Figure 7). Pathways were assigned a z-score indicating predicated activation or inhibition based on the number of overlapping differentially expressed genes and the direction of expression for each pathway. An HPFNA exhibited the highest degree of downstream activity activating competing pathways related to the onset of fatty liver. HPFNA activated genes involved in lipid accumulation, such as fatty acid and lipid synthesis, as well as lipid metabolism pathways, including beta oxidation of fatty acids. An HPFOS increased glucose and carbohydrate synthesis pathways and modulated the activation and quantity of immune cells. LPFOS and LPFNA led to increased synthesis of fatty acids. In addition, LPFNA activated both fatty acid uptake and fatty acid metabolism. The upstream drivers and downstream pathways modulated by either PFOS or PFNA exert competitive influences on the outcome of hepatic lipid accumulation, making the key drivers difficult to interpret. Overall, diet and the PFAS functional head group exerted influence on both upstream transcriptional drivers and downstream pathway outcomes.

Figure 7.

The PFOS and PFNA significantly modulate many pathways related to lipid homeostasis within the liver. The color gradient represents IPA generated pathway inhibition or activation scores, called z-scores. Orange signifies activation and blue signifies predicted pathway inhibition. Activation or inhibition of functional pathways was predicted based on the direction of modulation and the number of genes involved in the pathway.

Diet- and Structure-Induced Alterations in the Transcriptomic Signatures of PFOS and PFNA are Conserved at the Protein Level

Proteomics was performed on mouse liver tissue to derive the expression of >300 hepatic proteins. The global analysis of the liver proteome supports the diet and compound effects observed in the transcriptomic data. A PCA plot was generated to visualize changes in the global liver proteome (Figure 8). Distinct groupings can be observed between treatment groups. Both diet and the PFAS compound caused distinct shifts in the proteomic signature of the liver. Interestingly, PFNA showed a drastic shift in grouping compared with both PFOS and dietary controls. In Figure 9, the total number of differentially expressed proteins out of total proteins detected can be found within the box next to each comparison. Again, both diet- and structure-dependent effects on expression were observed. For example, LPFNA significantly impacted the expression of 197 out of approximately 300 proteins detected whereas LPFOS modulated 65. LPFNA and LPFOS shared 46 proteins, whereas LPFOS retained 19 and LPFNA 151 unique changes.

Figure 8.

Diet interactions shift the hepatic proteome and PFNA is a potent modulator. Global protein analysis was conducted using SWATH-MS DIA proteomics. The raw and processed data are accessible via ProteomeXchange with identifier PXD015977. The data were analyzed using MaxQuant and Persues software. The data were filtered for contaminants and the cutoff for significance was set as p < .05. The principal component analysis was created using Perseus software and represents the treatment effect on global hepatic protein expression.

Figure 9.

Structure-specific interactions shift the hepatic proteomic signature of PFNA and PFOS. Each Venn diagram depicts the overlap between the significantly modulated proteins from each treatment group. Significance was calculated on the SWATH-MS acquired data using MaxQuant software.

To confirm the global changes observed, targeted quantification of relevant lipid and oxidative stress related molecules was performed using open source targeted proteomic analysis software, Skyline (Figure 10). The elevated potency of PFNA can be clearly observed at the protein level in the expression of fatty acid metabolizing, cytochrome P450s, and antioxidant proteins. For the cytochrome P450s, PFOS exhibited stronger induction for CYP2C enzymes including, CYP2C50 and CYP2C37. However, PFNA showed more potent induction for CYP4A enzymes, CYP4A12, and CYP4A14. Lastly, we observed that the relative induction of gene and protein expression was conserved for many of the overlapping gene and protein targets such as EHHDAH and CYP4A14. In conclusion, both PFOS and PFNA significantly modulated the liver proteome and potential metabolic function at a relatively low dose.

Figure 10.

The PFOS and PFNA significantly modulate protein expression related to lipid accumulation, metabolism, and oxidative stress on the targeted level. An SWATH-DIA-acquisition files were imported into Skyline (MacCoss Lab Software) for targeted protein expression analysis. Each protein expression values represents the average expression of 2 library matched peptides normalized to nanodrop concentration and beta-actin digestion controls. Average expression intensity (n = 5) was converted to fold change and presented in each cell. Orange indicates activation, whereas blue indicates inhibition relative to control.

Diet-Specific Modulation of Liver Pathology in the Presence of PFAS Could Be Attributed to Altered Expression of PFAS Uptake Transporters

Diet exerted a dramatic influence on the resulting liver pathology. Perfluorooctanesulfonic acid and PFNA have low passive permeability and a high ($\ge$400) molecular weight (Cheng and Ng, 2017). Perfluoroalkyl substances uptake into the liver and reabsorption from the kidney is known to be driven by a class of transporters called organic anion transporter proteins (OATPs). In the liver, there are 4 major Oatps that mediate hepatic uptake from the bloodstream: OATP1A1, OATP1A4, OATP1B2, and OATP2B1 (Zhao et al., 2017). In the HFD-treated groups, Oatp transporter expression was significantly reduced by PFAS exposure (Figure 11). To examine if diet influences the uptake of PFOS and PFNA into the liver, internal hepatic exposure to PFOS and PFNA was quantified by LC-MS/MS (Table 5). An HFD feeding caused a significant decrease in hepatic concentration of PFOS and PFNA when compared with LFD feeding. In addition, PFNA achieved significantly higher hepatic concentrations than PFOS within both the LFD- and HFD-treated groups when administered at the same dose. When taken together, these findings suggest that the HFD reduced PFAS accumulation within the liver via decreased uptake transporter expression.

Figure 11.

Mechanistic drivers of PFAS-diet interactions. Significance was calculated using 1-way ANOVA followed by Fisher’s LSD test. *p < .05, significance from the LFD controls. #p < .05, significance from the HFD controls. $p < .05, significance between PFOS and PFNA within the same diet (ie, L/HPFOS vs L/HPFNA). &p < .05, significance between diet treatment within the same compound (ie, LPFAS vs HPFAS). A, Expression is presented as the log2 of the average signal ± SEM. The values were generated via transcriptomic array on n = 3 individual samples per treatment group. B, Male C57BL/6 mice were fed with either a low-fat diet or a moderately high-fat diet with or without PFOS or PFNA (0.0003% wt/wt in feed) for 12 weeks. After necropsy, PFOS and PFNA were extracted from liver and quantified using LC-MS/MS. All control groups exhibited PFOS and PFNA concentrations below the lower limit of quantification (LLOQ). Likewise, no quantifiable cross-contamination between PFOS- and PFNA-treated groups was found. All values are means ± SEM; n = 7–8.

Table 5.

Hepatic Concentration of PFAS

	PFOS (µg/g)	PFNA (µg/g)
LFD	115.7 ± 6.8$,&	169.7 ± 3.8$,&
HFD	87.3 ± 7.4$,&	110.9 ± 11.1$,&

PFOS and PFNA were measured in hepatic tissue following 12 weeks of exposure to chemical in diet. All control groups exhibited PFOS and PFNA concentrations below the lower limit of quantification (LLOQ) of 15 ng/ml. Likewise, no quantifiable cross-contamination between PFOS- and PFNA-treated groups was found. Calculations were performed using a 1-way ANOVA followed by post hoc Bonferroni test. All values represent the average ± SEM; n = 7–8.

^$ p < .05, significance between PFOS and PFNA within the same diet (ie, LPFOS vs LPFNA).

^& p < .05, significance between diet treatment within the same compound (ie, LPFAS vs HPFAS).

Perfluoroalkyl substances are known to interact with not only OATPs, but also with fatty acid binding proteins (FABPs). The FABPs are postulated to serve as reservoir depots in the cell that facilitate PFAS accumulation (Cheng and Ng, 2017). In Figure 5, we observed that the expression of FABPs and transporters, such as Fabp1, Fabp4, Cd36, and Slc27a1, were induced by PFAS within both diets. In Figure 6, a notable increase in Ppar$\alpha$ activity within the HFD + PFAS groups can be observed. Ppar$\alpha$ activation leads to the downregulation of Oatps (Cheng and Klaassen, 2008). This diet-specific downregulation coupled with an elevated influx of dietary fatty acids from the HFD into the serum may lead to increased competition between PFAS and FFAs for hepatic uptake. This competitive uptake could lead to a decrease in both lipid and PFAS accumulation in the liver. A potential mechanism to explain the paradoxical effect of PFAS in combination with HFD on the onset of hepatic steatosis is proposed in Figure 12.

Figure 12.

Hypothesized mechanism of the PFAS-diet effect on hepatic steatosis. An HPFNA and HPFOS enhanced activity of Ppar$\alpha$ relative to dietary controls and LPFAS groups. Ppar$\alpha$ activation causes a reduction in Oatp expression, a known uptake transporter for PFAS. Decreased Oatp expression coupled with increased serum FFAs from diet creates increased competition for fatty acid uptake transporters. This competition may reduce hepatic FFA and PFAS accumulation.

DISCUSSION

The majority of studies examining PFAS induced hepatic steatosis utilize relatively high doses of PFAS, acute exposure windows, and a standard chow (Wan et al., 2012; Kudo et al., 2006; Bagley et al., 2017; Curran et al., 2008). It has been well-established that a high dose of PFAS can incite the onset of fatty liver when administered with a lean standard mouse chow (Armstrong and Guo, 2019). However, the mechanism of this toxicant-associated fatty liver disease is still poorly understood. The current theories include PFAS induced choline deficiency (Zhang et al., 2016), impaired mitochondrial function (Quist et al., 2015; Yao et al., 2016), impaired lipid export via VLDL (Bijland et al., 2011; Wang et al., 2015), and an imbalance between FA synthesis and beta oxidation (Das et al., 2017). Perfluoroalkyl substances exposure is lifelong and there is a need to better understand the potential hepatotoxicity of PFAS under chronic exposure. Moreover, the largest risk factor for hepatic steatosis is metabolic syndrome, often initiated by poor diet. In order to better explore the ability of PFAS to augment fatty liver risk, the experimental design must take into account the most common risk factor for fatty liver, diet induced metabolic syndrome.

Few papers have examined the exposure to PFAS in combination with dietary fat on the development of hepatic steatosis. According to Huck et al., HFD feeding in combination with 0.0001% PFOS (approximately 0.1 mg/kg/day). Using a dose of 0.0003% PFAS in diet (approximately 0.24 mg/kg/day), this study confirms that coadministration with an HFD can attenuate PFOS induced hepatic lipid accumulation. Furthermore, this effect was demonstrated with PFNA for the first time. Other studies have demonstrated that perfluorooctanoic acid (PFOA), an 8-carbon PFAS with a carboxylic acid head group, can have attenuate fatty liver in murine models when combined with dietary sources of lipids. One such study demonstrated that PFOA prevented hepatic accumulation of triglycerides with a fish oil supplemented diet. Furthermore, a study by Li et al. revealed that PFOA (1 mg/kg/day) lessened the severity of hepatic steatosis, triglyceride content, and fibrosis in mice with pre-existing fatty liver induced by HFD feeding (Li et al., 2019). This work goes further to characterize diet effects on hepatic PFAS accumulation and utilized transcriptomic and proteomic techniques to uncover the mechanistic drivers of the diet-specific response to PFAS exposure.

In the present study, we observed significant PFAS effects on mouse body, WAT, and liver weight. Perfluoroalkyl substances are well known to reduce body and WAT weight, whereas inducing liver weight at higher doses in rodents. However, the dose administered in this study was too low to observe significant PFOS effects on body and liver weight gain. In a study of PFOS exposure to dams at 0.3 mg/kg/day there was similarly no effect on these gross weight outcomes, whereas significant changes were observed at the high dose of 3 mg/kg/day (Wan et al., 2014). It has been previously demonstrated that PFOS can induce adipogenesis in WAT (Xu et al., 2016). Likewise, we observed increases in WAT weight and WAT:BW with relatively low-dose PFOS administration. Despite administration at the same low dose, PFNA exposure resulted in significant modulation to body, WAT, and hepatic weights. Unlike PFOS, PFNA significantly reduced WAT weight. Perfluorononanoic acid has been previously reported to cause marked hepatomegaly at higher doses, more so than structurally similar perfluorinated carboxylic acids (PFCAs) (Kudo et al., 2006). Perfluorononanoic acid exerted significant effects on these endpoints at the same dose due to greater hepatic accumulation and its potent induction of PPARα. This induction leads to the increased expression of proteins involved in fatty acid oxidation which can metabolize lipids in WAT. In the current study, PFNA induced several lipid metabolizing proteins at much higher levels than PFOS. The discrepancy in weight adipose tissue and liver weights between PFNA and PFOS may be attributable to the PFOS specific induction of Pparγ observed with HPFOS feeding and increased PFNA potency in inducing lipid metabolizing enzymes. In the case of PFNA, the rate of Pparα driven white adipose lipid metabolism outweighed the rate of adipogenesis. In the case of PFOS, the rate of Pparγ driven adipogenesis outweighed the rate of lipid metabolism within the WAT.

Perfluoroalkyl substances are well known to cause dyslipidemia in rodents as well as humans. In rodents PFAS are known to cause accumulation of hepatic TAGs, increased hepatic lipid deposition, and hypocholesterolemia. Interestingly, we observed significant attenuation of hepatic phospholipid content by PFOS and PFNA within the LFD and HFD. Previous work has shown that PFAS cause increased concentrations of oxidized phospholipids in the blood (Pfohl et al., 2020). Phospholipids are known to have a hypolipidemic effect within the liver and have even been used to attenuate fatty liver disease (Gundermann et al., 2016). Phospholipids are also known to predict PFAS partitioning and the ability of phospholipids to associate with PFAS has been confirmed (Dassuncao et al., 2019; Sanchez Garcia et al., 2018). Furthermore, PFAS are able to intercalate into cellular membranes and increase membrane permeability (Nouhi et al., 2018; Fitzgerald et al., 2018). Phospholipids act as a surfactant to stabilize lipid droplets, are the major constituent of cell membranes, and aid in vesicle formation for lipid export. The implication of PFAS reduction in hepatic phospholipids and its implications in hepatic steatosis has not been adequately explored.

A reduction in TAG content was produced by PFNA and PFOS supplementation within the HFD. Similarly, previous reports demonstrated that PFOA reduced hepatic triglycerides in the presence of dietary lipids (Kudo and Kawashima, 1997; Li et al., 2019). This result supports the histopathology showing significant reduction in hepatic lipid accumulation with coadministration of PFAS and HFD. The present work is the second study to report that a subchronic exposure to PFOS attenuates the onset of HFD induced hepatic steatosis. Furthermore, this is the first paper to demonstrate this preventative effect using PFNA. Our group went further to quantify the hepatic concentration of PFAS within each treatment group.

Diet and structure related effects on hepatic accumulation of PFAS were observed. Interestingly, we observed significant reduction in hepatic PFAS concentration with administration of an HFD. A developmental study reported that an HFD significantly increased hepatic PFAS concentrations in pups with perinatal exposure to PFAS via dams (Wan et al., 2014). However, a study by Pfohl et al. (2020) demonstrated that direct coadministration with an HFD results in a significant reduction of PFOS and PFHxS hepatic accumulation. The present work confirms this finding with PFOS and reports the conservation of this effect with PFNA. Lastly, we confirmed previous reports that PFNA has more bioaccumulative potential within the murine liver when compared with the sulfonic acid, PFOS. It is also known that PFNA accumulates in the liver at higher levels than similar PFCAs of varying chain lengths (Kudo and Kawashima, 2003). Furthermore, it has been demonstrated PFCAs alter hepatic fatty acid profiles in a hepatic concentration-dependent manner (Kudo et al., 2011). The increased hepatic uptake relative to PFOS within the same diet could be in part due to the PFNA-specific induction of Fabp4, a potential PFAS uptake transporter (Ng and Hungerbühler, 2014).

The mechanism of diet-PFAS interactions has not been adequately explored. The previously published work by Huck et al. suggests that the driver of the diet-specific response is differential regulation of Ppar$\gamma$ and its down-stream target, Cd36. However, in the current study no diet-specific response in Ppar$\gamma$ nor the fatty acid uptake transporter, Cd36, was observed. This finding was confirmed by Western blot (Supplementary Figure 3). Li et al. suggested that PFAS lessened the severity of pre-existing fatty liver because PFOA potentiated genes related to lipid metabolism within the HFD. In contrast, the present study revealed that gene and protein expression of lipid metabolizing enzymes was potentiated more within the low fat + PFAS diets, especially the LPFNA group. The comparison of the PFOS and PFNA outcomes and hepatic signatures relative to diet suggests that there may be an alternative mechanism for the protective effect of PFAS combined with an HFD.

Ppar $\alpha$ induces the metabolic breakdown of fatty acids and is known to have a protective effect against the onset of hepatic steatosis. Perfluoroalkyl substances are generally accepted to be robust Ppar$\alpha$ inducers and the findings presented here confirm this. Overall, PFNA was a more potent inducer of Ppar$\alpha$ and the overall transcriptome when compared with PFOS. Given that PFNA achieves greater hepatic concentrations relative to PFOS it is not surprising that it would result in a greater level of induction. However, in the present study HPFNA and LPFOS concentrations were very similar in the liver suggesting that the increased level of induction observed with PFNA is not solely dependent on increased hepatic deposition and is likely attributable to the difference in functional head group. This potent effect on both gene and protein level expression suggests that PFNA alters the hepatic signature at lower doses than PFOS and therefore may require stricter regulation.

Diet imposed significant influence on the overall hepatic signatures for both PFOS and PFNA. OATPs are known transporters of PFAS and have been demonstrated to mediate PFAS deposition and reabsorption from the liver. The HFD + PFAS groups had much greater predicted activation of PPARα relative to their matched LFD + PFAS groups. Furthermore, PFAS have been shown to reduce the expression of Oatps via Ppar$\alpha$ mediated suppression (Cheng and Klaassen, 2008). High-fat diet feeding is known to initially induce PPARα, given that its endogenous ligands include fatty acids and triglycerides (Patsouris et al., 2006). This activation can serve a protective role by inducing fatty acid oxidation and improving insulin sensitivity. With chronic HFD feeding there is often a shift to downregulation of PPARα due to the resulting development of diet-related health outcomes such as dyslipidemia or insulin resistance (Domínguez-Avila et al., 2016). Similarly, the HFD controls in the present study did exhibit downregulation of PPARα and increased hepatic lipid accumulation, therefore, PPARα mediated suppression of OATPs would not be expected within the HFD controls. Given the attenuating effect of PFAS on the onset of chronic HFD-induced outcomes, such dyslipidemia and fatty liver, this may result in a synergistic increase in both diet and PFAS mediated upregulation of PPARα. This could explain the diet-specific activation of PPARα and resulting suppression of OATPs. Both PFAS and fatty acids are known to interact with FABPs for hepatic uptake/accumulation. The decreased expression of a primary PFAS transporter coupled with increasing competition between FFAs and PFAS for uptake could present a novel mechanism for the underlying cause of the diminished hepatic deposition of both lipid and PFAS within the HFD treatment groups.

CONCLUSIONS

Using a diet-induced obesity rodent model, we sought to explore the potential role of diet in PFOS and PFNA induced fatty liver. Despite the lack of federal regulation, PFNA was significantly more potent than PFOS in altering hepatic molecular pathways and exhibited increased hepatic deposition. These findings suggest that PFNA may require stricter regulation than PFOS despite its structural similarity. Importantly, dietary fat exhibits a prominent influence on PFAS hepatic accumulation, pathology, and biochemical pathways. This effect may be caused by Ppar$\alpha$ mediated suppression of PFAS uptake transporters, Oatps. Further studies are warranted to further confirm and explore PFAS-diet interactions and their potential implications in assessing internal exposure and risk.

SUPPLEMENTARY DATA

Supplementary data are available at Toxicological Sciences online.

DECLARATION OF CONFLICTING INTERESTS

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

FUNDING

This work was supported by National Institute of Health (NIH) grants 1R15ES025404-01 and P42ES027706. This material is based upon work conducted at University of Rhode Island at a Rhode Island National Science Foundation's Established Program to Stimulate Competitive Research (NSF EPSCoR) research facility, the Genomics and Sequencing Center, and the Molecular Characterization Facility, supported in part by the NSF EPSCoR Coopoerative Agreement Number OIA-1655221. The transcriptomic array was conducted at the Brown University Genomics Facility with partial support from the National Institutes of Health (NIGMS Grant No. P30GM103410, NCRR Grant No. P30RR031153, P20RR018728, and S10RR02763), National Science Foundation (EPSCoR Grant No. 0554548), Lifespan Rhode Island Hospital, and the Division of Biology and Medicine, Brown University.