Dose additive maternal and offspring effects of oral maternal exposure to a mixture of three PFAS (HFPO-DA, NBP2, PFOS) during pregnancy in the Sprague-Dawley rat

Abstract

Simultaneous exposure to multiple per- and polyfluoroalkyl substances (PFAS) is common in humans across the globe. Individual PFAS are associated with adverse health effects, yet the nature of mixture effects after exposure to two or more PFAS remains unclear. Previously we reported that oral administration of hexafluoropropylene oxide-dimer acid (HFPO-DA, or GenX), Nafion byproduct 2 (NBP2), or perfluorooctane sulfonate (PFOS) individually during pregnancy produced maternal and F1 effects. Here, we hypothesized that responses to the combined exposure to these three PFAS would be dose additive. Pregnant Sprague-Dawley rats were exposed to a fixed-ratio equipotent mixture where the top dose contained each PFAS at their ED50 for neonatal mortality (100% dose=PFOS 3 mg/kg; NBP2 10 mg/kg; HFPO-DA 110 mg/kg), followed by a dilution series (33.3, 10, 3.3, and 1%) and vehicle controls (0% dose). Consistent with the single chemical studies, dams were exposed from gestation day (GD)14-18 or from GD8-postnatal day (PND2). Fetal and maternal livers on GD18 displayed multiple significantly upregulated genes associated with lipid and carbohydrate metabolism at all dose levels, while dams displayed significantly increased liver weight (≥3.3% dose) and reduced serum thyroid hormones (≥33.3% dose). Maternal exposure from GD8-PND2 significantly reduced pup bodyweights at birth (≥33.3% dose) and PND2 (all doses), increased neonatal liver weights (≥3.3% dose), increased pup mortality (≥3.3% dose), and reduced maternal bodyweights and weight gain at the top dose. Echocardiography of adult F1 males and females identified significantly increased left ventricular anterior wall thickness (~10% increase), whereas other cardiac morphological, functional, and transcriptomic measures were unaffected. Mixture effects in maternal and neonatal animals conformed to dose addition using a relative potency factor (RPF) analysis. Results support dose addition-based cumulative assessment approaches for estimating combined effects of PFAS co-exposure.

1. INTRODUCTION

Many state, federal, and international human and environmental health-based agencies have recently moved forward with decisions to assess and regulate per- and polyfluoroalkyl substances (PFAS) using cumulative, mixture-based approaches due to pervasive human and ecological exposure to multiple PFAS (EFSA et al. 2020; RIVM 2018; USEPA 2023c). We recently reported dose additive mammalian maternal and neonatal effects from combined exposure to perfluorooctanoate (PFOA) and perfluorooctanesulfonate (PFOS) during pregnancy in a Sprague-Dawley rat model (Conley et al. 2022a). These effects were more accurately predicted using mixture models based on principles of dose addition (assuming toxicological similarity), including the relative potency factor (RPF) approach, than a model based on response addition (assuming toxicological independence). Additional experimental investigations of cumulative effects from combined exposure to multiple PFAS are needed due to the expansive structural diversity of compounds categorized as PFAS (USEPA 2023a; Wang et al. 2017), myriad adverse effects that have been reported in studies of individual compounds (ATSDR 2021; Fenton et al. 2021; Lau et al. 2007), and widespread co-occurrence (USEPA 2023b).

The universe of PFAS contains thousands of diverse chemical structures (Evich et al. 2022; Wang et al. 2017), however most exposure and toxicity studies have focused on straight-chain and ether-linked compounds containing carboxylate or sulfonate functional groups. Further, many PFAS compounds with alternate functional groups (e.g., fluorotelomers, sulfonamides) can (bio)degrade to form perfluoroalkyl acids (Buck et al. 2011; Suthersan et al. 2016). The mode(s) of action for maternal and developmental effects of PFAS carboxylates and sulfonates are not fully characterized, but data indicate multiple nuclear receptors are involved including peroxisome proliferator-activated receptor alpha (PPARα) and gamma (PPARγ), constitutive androstane receptor (CAR), liver X receptor (LXR), farnesoid X receptor (FXR), and pregnane X receptor (PXR) (Bjork et al. 2011; Evans et al. 2022; Houck et al. 2021; Rosen et al. 2017) and multiple target tissues including the liver and kidneys. Further, data indicate considerable variation in effects across lifestages with neonates more affected than maternal animals for some endpoints. Studies from our lab and others have demonstrated that carboxylate and sulfonate PFAS produce different effects in maternal and F1 animals from exposure during pregnancy, however there are multiple common effects shared across compounds including reduced F1 body weights, increased liver weights, reduced serum thyroid hormones and lipids, and increased liver gene expression (Abbott et al. 2009; Abbott et al. 2007; Butenhoff et al. 2009; Butenhoff et al. 2004; Chang et al. 2018; Conley et al. 2021; Conley et al. 2022a; Conley et al. 2022b; Conley et al. 2019; Gilbert et al. 2021; Lau et al. 2020; Lau et al. 2006; Lau et al. 2003; Luebker et al. 2005a; Luebker et al. 2005b; Rosen et al. 2017; Rosen et al. 2009; Rosen et al. 2007; Thibodeaux et al. 2003). Despite slight differences in the overall spectra of effects, for these shared endpoints the data indicate that effects are dose additive when there is exposure to multiple PFAS (Conley et al. 2022a).

Previously, we published individual chemical data on the developmental toxicity of hexafluoropropylene oxide-dimer acid (HFPO-DA or GenX) (Conley et al. 2021; Conley et al. 2019), Nafion byproduct 2 (NBP2) (Conley et al. 2022b), or PFOS in the Sprague-Dawley rat (Conley et al. 2022a; Conley et al. 2022b). HFPO-DA is a perfluoroalkyl ether carboxylic acid that was developed to replace PFOA in manufacturing processes (DuPont 2010). HFPO-DA has been detected in drinking water at elevated concentrations (Gebbink et al. 2017; Sun et al. 2016) and in human serum (Petriello et al. 2022), and there are recently published human health toxicity assessments by the US EPA (USEPA 2021) and ECHA (ECHA 2019). Nafion byproduct 2 is a polyfluoroalkyl ether sulfonic acid that was recently detected in 99% of serum samples from a human biomonitoring cohort in eastern North Carolina, USA (Kotlarz et al. 2020) and found to have relatively high liver accumulation in rodents (Bangma et al. 2022). PFOS is a legacy perfluoroalkyl sulfonic acid that has been phased out of production in some countries, including the United States, but remains a global contaminant (Cousins et al. 2022;Jian et al. 2017) with high detection frequencies in human serum and drinking water across the globe (Berg et al. 2014; Calafat et al. 2007; Kabore et al. 2018). Developmental toxicity studies of PFOS by our group and others (Lau et al. 2003; Thibodeaux et al. 2003) reported dose-responsive effects in maternal and F1 animals, including reduced maternal and pup bodyweights, increased pup mortality, increased maternal liver weights, reduced maternal and pup thyroid hormones, altered serum lipids, and altered liver expression of genes associated with lipid and carbohydrate metabolism genes, among others. The data reported in the individual chemical studies provided the basis for the present mixture experiments.

Building on prior studies, here we hypothesized that a mixture of HFPO-DA, NBP2, and PFOS would produce cumulative maternal and neonatal effects that were dose additive and could be accurately predicted using the simple RPF approach. The goal was to test additivity of effects from in vivo exposure to a mixture of PFAS that have been reported to co-occur in human serum and/or drinking water (Hopkins et al. 2018; Kotlarz et al. 2020; Sun et al. 2016), but not to specifically mimic the exact exposure levels that have been reported for these three PFAS. Here, F1 animals were aged to adulthood to investigate potential developmental and permanent effects from in utero exposure to all 3 PFAS. Studies have reported effects of exposure to various PFAS on reproductive tissues and steroid hormone signaling pathways (Carlson et al. 2022; Houck et al. 2021; Radke et al. 2022), thus we investigated markers of male and female reproductive development. Further, given the epidemiological evidence of cardiovascular effects associated with PFAS exposure (Ou et al. 2021) and cardiac remodeling from in utero exposure reported in chicken studies of PFOA (Jiang et al. 2012; Ni et al. 2023), we investigated a broad range of cardiovascular functional and morphological endpoints in male and female adult F1 animals in control and 33% dose groups due to complete neonatal litter loss in the top dose level (100%).

2. METHODS

2.1. Dosing solutions

Dosing solutions were prepared using high performance liquid chromatography-grade water purchased from Honeywell Research Chemicals (Muskegon, MI, USA). NBP2 (7H-Perfluoro-4-methyl-3,6-dioxaoctanesulfonic acid; CASRN: 749836-20-2; Product #: 6164-3-3J; Lot: 512400) and HFPO-DA (ammonium perfluoro(2-methyl-3-oxahexanoate); CASRN: 62037-80-3; Product #: 2122-3-09; Lot: 0000887; Purity ≥97%) were purchased from SynQuest Laboratories (Alachua, FL, USA). PFOS (heptadecafluorooctanesulfonic acid potassium salt; CASRN: 2795-39-3, Product #: 77282, Lot: BCBX5798, Purity ≥98%) was purchased from Sigma-Aldrich. Dosing was administered once daily (06:00-08:00 EST) via oral gavage at 2.5 mL/kg-body weight across a range of fixed ratio dilutions of the mixture. Dosing solutions were stored at room temperature and prepared fresh every 5-6 days.

Each of the three PFAS included here have been shown to individually produce neonatal mortality with steep dose response curve slopes. Early pup mortality precludes the measurement of all subsequent endpoints and thus we structured the mixture based on the component chemical potencies for neonatal mortality to evaluate the mixture-based effects on this endpoint and to prevent extensive litter loss across the experiment. The top mixture dose (100% dose) contained each chemical at their respective effective dose 50% (ED50) for neonatal mortality (HFPO-DA=110 mg/kg-d, NBP2=10 mg/kg-d, and PFOS=3 mg/kg-d) based on data we previously published in individual chemical studies with maternal oral exposure from GD8-PND2 (Conley et al. 2021; Conley et al. 2022a; Conley et al. 2022b). The top dose solution was then serially diluted at 33.$\overline{3}$, 10, 3.$\overline{3}$, and 1% dose and a 0% vehicle control (HPLC-grade water) (Table 1).

Table 1.

Mixture dosing structure for 3 PFAS mixture

Mixture dose	100%	33.$\overline{3}$%a	10%	3.$\overline{3}$%a	1%
HFPO-DA (mg/kg/d)	110	36.7	11	3.67	1.1
NBP2 (mg/kg/d)	10	3.3	1	0.33	0.1
PFOS (mg/kg/d)	3	1	0.3	0.1	0.03

^aThe 33.$\overline{3}$% and 3.$\overline{3}$% dose levels were 1/3 serial dilutions of the next higher dose levels

2.2. Animals

Time-mated Sprague-Dawley rats (Crl:CD(SD)), ~11-12 weeks old, were purchased from Charles River Laboratories (Raleigh, NC, USA) and shipped to USEPA (Research Triangle Park, NC, USA) on gestation day 2 (GD2; GD0 = bred date; GD1 = plug positive date). Dams and their offspring were housed individually in clear polycarbonate cages (20 x 25 x 47 cm) with heat-treated, laboratory-grade pine shavings (Northeast Products, Warrensburg, NY) and fed NIH07 Rat Chow and filtered (5 μm) municipal tap water (Durham, NC) ad libitum. Dams were weight-ranked and stratified based on the number of treatment groups then randomly assigned to treatment groups using a random sequence generator (www.random.org/sequences/) to produce similar mean weights and variances given the range of dam body weights (typically ~10% coefficient of variation). This study was conducted in accordance with a protocol approved by the USEPA Center for Public Health and Environmental Assessment Institutional Animal Care and Use Committee. Animals were housed in a facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care and maintained at 20-22°C, 45-55% humidity, and a 12:12 h photoperiod (06:00 – 18:00 EST).

Two experiments were conducted with dosing schedules consistent with our prior published studies on PFAS maternal and developmental effects (Conley et al. 2021; Conley et al. 2022a; Conley et al. 2022b; Conley et al. 2019). The first experiment referred to as the “Postnatal study” (section 2.3) included maternal oral exposure from embryo post-implantation (GD8) through delivery to postnatal day 2 (PND2) and was intended to identify maternal and offspring adverse effects from exposure during pregnancy and early lactation. The second experiment referred to as the “Fetal study” (section 2.4) included maternal oral exposure from GD14-18 and was intended to identify maternal and fetal adverse effects and biomarkers from short term exposure during pregnancy.

2.3. Postnatal study (GD 8 – PND 2 maternal exposure)

The Postnatal study consisted of one block of 30 rat dams to assess the maternal and postnatal effects of maternal exposure to the PFAS mixture during pregnancy and early lactation (Figure S1a). Dams were exposed daily via oral gavage to either 0 (water vehicle), 1, 3, 10, 33, or 100% of the top dose (n=5 dams per dose) from GD8 – postnatal day 2 (PND2; PND0 = day of parturition). Dams gave birth naturally and were checked for parturition hourly beginning at 6AM on the morning of GD22 (i.e., PND0) until all dams had delivered. Hourly checks on maternal and neonatal pup health continued until 5PM on PND1 and dead pups were removed and those that were moribund were removed and euthanized via decapitation. As dams delivered, we recorded the onset of parturition as the time the first pup was observed in a cage and the completion of delivery was determined as the time of maternal behaviors (e.g., retrieving pups to the nest, licking/grooming of pups, hovering over nest and nursing). Upon completion of delivery, all pups were briefly removed, counted and total litter weight was recorded. All pups were returned to their nest except for two pups randomly selected by blind grab from the litter, which were euthanized via decapitation and pooled trunk blood was collected. Of the two randomly selected pups, one whole pup was fixed in 10% formalin after the thoracic and abdominal cavities were exposed. Subsequently, livers from formalin fixed pup carcasses were removed and shipped to Experimental Pathology Laboratories, Inc. (Durham, NC) where they were embedded, sectioned, stained with hematoxylin and eosin (H&E) and periodic acid-Schiff (PAS), and evaluated by a Diplomate of the American College of Veterinary Pathology. For the second euthanized pup we removed liver tissue for RNA extraction and gene expression (described below).

F1 animals were evaluated for reproductive developmental landmarks using previously described methods (Gray et al. 2009; Gray et al. 2022), but with minor variation as described here. On PND2 all pups were sexed, weighed, anogenital distance (AGD) was measured using a stereoscope and ocular micrometer and the technician measuring AGD was blinded to treatment group. Measurement of AGD deviated slightly from prior studies in that the midline of the genital papilla was used as the anterior landmark instead of the posterior base of the genital papilla (Figure S2), which resulted in AGD measurements ~15% greater than those historically reported by our lab. In this instance the midline of the genital papilla was an easier landmark to identify than the posterior base for the technician measuring AGD and measurements within sex had similar variance to prior published AGD measurements. After AGD measurement, litter sizes were reduced to n=4 male and n=4 females where possible with pups randomly selected for euthanasia by blind grab from the litter. The randomly selected pups were euthanized via decapitation and trunk blood was collected for serum isolation and for one pup per litter, where available, liver weight was recorded, and a liver sample was collected for determination of PFAS concentrations. Due to complete litter losses, dams in the 100% dose group were euthanized via decapitation on PND2, trunk blood was collected for serum isolation, and liver weight and uterine implants were recorded.

On PND9 litter size and total litter weight were recorded. On PND13 all pups were weighed and scored for areolae/nipple retention. On PND21 pups were weighed and weaned to two animals per cage by sex and treatment group, and food was changed to NTP2000 rodent diet. Following weaning, dams were euthanized, trunk blood was collected for serum isolation, liver weight was recorded, and uterine implantation sites were scored. Beginning on PND31 for female offspring and PND41 for male offspring, individuals were evaluated at the same time every day for markers of pubertal onset, vaginal opening (VO) for females and preputial separation (PPS) for males.

Beginning PND156 a subset of offspring (n=2 males and n=2 females per litter) from the 0% (vehicle control) and top surviving dose group (33.3%) were evaluated for cardiac endpoints (see below). Also beginning on PND156 female offspring not undergoing cardiac evaluation were weighed, euthanized via decapitation, examined for any gross reproductive tract abnormalities, and organ weights were collected for paired ovaries, liver, paired kidneys, heart, and paired adrenal glands. Similarly, beginning PND176 males were weighed, euthanized, and examined for reproductive tract malformations and organ weights were collected for paired testes, paired epididymides, liver, paired kidneys, heart, and paired adrenal glands.

2.3.1. Cardiac evaluation of adult F1 animals

Echocardiography, a non-invasive ultrasound imaging technique, was used to obtain non-invasive measures of left ventricular systolic and diastolic function and wall dimensions (Ram et al. 2011). Cardiac evaluation was a 2x2 factorial design (sex x dose) with two male and two female littermates evaluated from each of 10 litters (5 litters each from control and 33.3% dose) for a total of 40 rats. Cardiac function was determined using a high-frequency echocardiography ultrasound system (Vevo 2100, Fujifilm Visual Sonics Inc., Toronto, Canada). Rats were anesthetized with isoflurane (1%-3% delivered in 100% O₂ at 0.8-1 L/min) then transferred to a heated ECG-monitoring table in dorsal recumbency where anesthesia levels were maintained with 1%-3% isoflurane via nose cone. Eye lubricant was applied to avoid ocular drying, and each paw was ground to ECG electrodes coated with electrode cream (Cat. No. 600-0001-01-S, Indus Instruments, Webster, TX, USA) for physiological monitoring/recording of ECG, and heart and respiratory rates. The surface of the Vevo Rat Handling Table was set to 38°C to support animal core temperature. There chest and upper abdomen were shaved using an electric razor and Nair gel was used to remove any remaining fur from the imaging location. The application area was wiped with clean gauze to remove any residual Nair from the skin. Prewarmed ultrasound gel was applied to the chest prior to imaging. Electrocardiogram and respiratory signals were recorded via paw electrodes in Lead II configuration. To minimize the effect of heart rate on the values acquired during assessment, heart rates were targeted to 350 beats/min by titrating anesthesia. A MS-201 transducer (FUJIFILM VisualSonics Inc., Toronto, Canada) was used to obtain parasternal long axis views of the left ventricle in M-mode (15 MHz) for functional measurements and pulsed wave Doppler (12.5 MHz) of transmitral and pulmonary artery blood flow. The sonographer was blinded to treatment group identities.

Echocardiography data analyses were also performed while blinded to treatment using Vevo LAB software (Fujifilm Visual Sonics Inc.). Two beats between breaths from each of the 3 cine loops were collected. This yielded a total of 6 beats analyzed per animal. M-mode data was analyzed to determine stroke volume (SV), cardiac output (CO), ejection fraction (EF), fractional shortening (FS), end diastolic volume (EDV), end systolic volume (ESV), and left ventricular anterior and posterior wall thickness (LVAW and LVPW). Pulsed wave Doppler of transmitral flow was analyzed to determine isovolumic contraction time (IVCT), aortic ejection time (AET), and isovolumic relaxation time (IVRT). The Tei index of myocardial performance was calculated with the following equation: (IVCT + IVRT)/AET. Pulmonary artery flow was analyzed to determine pulmonary acceleration time (PAT), pulmonary ejection time (PET) and PAT/PET.

At necropsy for the cardiac evaluation cohort, organ weight measurements were recorded similar to the remaining F1 animals and serum was collected in addition to a subsample of cardiac tissue for RNA isolation. Right tibia length was also measured for normalization of heart weight. Serum samples were run on Acute Phase Protein Panel 1 (rat) Kit, Proinflammatory Panel 2 (rat) Kit, and Cardiac Injury Panel 3 Kit from Meso Scale Discovery (Meso Scale Diagnostics, Rockville, MD, USA) according to manufacturer specifications. Heart RNA samples were prepared similar to maternal and fetal livers described below in Section 2.4.1 and evaluated using the RT2 Profiler PCR Array for Rat Cardiovascular Disease by Qiagen (Cat. no. 330231 PARN-174ZA), which contains 84 key genes linked to cardiac disease. PCR reactions were run using RT2 SYBR Green qPCR Master Mix (SABiosciences Corp., Frederick, MD, USA) on a CFX96 Touch Real-Time Detection System (Bio-Rad, Hercules, CA, USA).

2.4. Fetal study (GD 14-18 maternal exposure)

The Fetal study was comprised of a total of 30 rat dams exposed to the PFAS mixture from GD 14-18 across the same dose range as the GD8-PND2 experiment with n=6 for controls and n=5 for treated except n=4 at 1% dose (Figure S1b). We evaluated maternal weight gain from GD14-18, reproductive output (# fetuses, resorptions), maternal liver weight, body weight of 3 randomly selected fetuses, collected maternal and fetal liver mRNA, and collected maternal serum. On GD18 dams and fetuses were euthanized by decapitation 2-4 hours after the final oral dose (~8:30-10:30AM EST). Euthanasia order was stratified such that the timing of necropsy was equally distributed across dose groups.

2.4.1. Liver gene expression in GD18 maternal and fetal rats

Subsamples of fetal liver (~30-50 mg) were collected on GD18 into polypropylene microcentrifuge tubes containing 500 μL TRIzol Reagent (Invitrogen, Carlsbad, CA, USA) on ice, then individually homogenized using a Bullet Blender (Next Advance, Troy, NY, USA) with 1mm zirconium oxide beads and stored at −80°C prior to RNA extraction. RNA extraction was conducted according to TRIzol Reagent manufacturer specifications using chloroform and isopropanol. Following extraction, RNA was purified using GeneJet RNA clean up and concentration kit (Cat no. K08401; Thermo Fisher Scientific, Waltham, MA, USA). RNA concentration and purity (260:280 ratio ≥1.8) were determined with a NanoDrop 2000 spectrophotometer (Thermo Scientific). cDNA was synthesized from purified RNA using the RT2 First Strand kit (Qiagen, Hilden, Germany) and gene expression was assessed using reverse transcriptase real-time polymerase chain reaction (RT-qPCR).

Fetal livers (GD18, one fetus per litter) were evaluated using the RT2 Profiler PCR Array for Rat PPAR Targets by Qiagen (Cat. no. 330231 PARN-149ZA), which contains 84 target genes relevant to PPAR alpha (α), beta/delta (β/δ), and gamma (γ) signaling pathways. Maternal GD18 livers from the Fetal study were evaluated using the Qiagen RT2 Profiler PCR Array for Rat Fatty Acid Metabolism (Cat no. 330231 PARN-007ZA), which contains 84 key genes involved in the regulation and enzymatic pathways of fatty acid metabolism. Similar to heart gene expression, PCR reactions were run using RT2 SYBR Green qPCR Master Mix (Qiagen) on a CFX96 Touch Real-Time Detection System (Bio-Rad).

2.4.2. Determination of thyroid hormone concentrations and clinical chemistry parameters in GD18 dams

Maternal serum was analyzed for thyroid hormone concentrations (total triiodothyronine (T₃) and thyroxine (T₄)) by radioimmunoassay according to manufacturer specifications (IVD Technologies) and as previously described (Conley et al. 2019). Clinical chemistry was analyzed in maternal serum using a Rx daytona+ (Randox Laboratories, Kearneysville, WV) according to manufacturer specifications. Clinical chemistry parameters included: ALT, AST, triglycerides, cholesterol, albumin, and glucose (non-fasted).

2.5. Analytical chemical determination of PFAS concentrations

Maternal (at GD18 in Fetal study and at PND21 weaning in Postnatal study) and offspring (PND0 in Postnatal study) serum and liver samples were analyzed for HFPO-DA, NBP2, and PFOS concentrations using similar methodology to that reported by (Conley et al. 2019; McCord et al. 2018). Serum was isolated from trunk blood via centrifugation (10,000 x g for 15 min at 4°C) in vacutainer tubes, transferred to 1.5 mL microcentrifuge tubes and stored at ~80°C. Liver tissue subsamples were individually weighed then transferred to clean polypropylene microcentrifuge tubes and HPLC-grade water was added at a ratio of 3:1 (water:liver, weight basis). Tissues were then homogenized and stored at 4°C prior to chemical analysis. Briefly, samples were denatured using 100 μL 0.1 M formic acid followed by a 500 μL cold (−20°C) acetonitrile protein crash. Denatured samples were centrifuged at 10,000 x g for 5 min and 200 μL supernatant was transferred to screw-cap chromatography vials with 200 μL ammonium acetate buffer. Sample extracts were separated using a Waters ACQUITY ultra performance liquid chromatograph (UPLC) (Waters Corporation, Milford, MA, USA) fitted with a Waters ACQUITY UPLC BEH C18 column (2.1 mm x 50 mm; 1.7 μm; 130 Å). Detection was performed using a Waters Quattro Premier XE tandem quadrupole mass spectrometer in negative ionization mode. Stable isotopes of HFPO-DA (¹³C₃, Cat. no. M3HFPO-DA) and PFOS (¹³C₄, Cat. no. MPFOS) were purchased from Wellington Laboratories (Guelph, Ontario, Canada) and used as an internal standard for quantitation. MPFOS was used as the internal standard for NBP2 quantitation as no isotopically labelled standard was available for NBP2. Calibration curves were prepared to encompass the range of concentrations in serum and liver homogenates for each chemical. Limits of quantitation (LOQ) were 5 ng/mL (serum) and 0.1 μg/g (liver). Samples below LOQ were estimated at LOQ/2 for the purpose of calculating means and standard errors (only vehicle controls had samples <LOQ).

2.6. Data analyses

All data analyses were conducted using GraphPad Prism (version 9; GraphPad, Inc.; La Jolla, CA, USA) and SAS (version 9.4; SAS Institute; Cary, NC, USA). All values are reported as mean ± standard error of the mean (SEM). All mixture endpoints except echocardiography were analyzed by analysis of variance (ANOVA) using PROC GLM in SAS, followed by pairwise comparison of individual dose levels to vehicle controls using LSMEANS (α=0.05). For all pup data, litter means were used as the statistical unit to account for the nested effects of individuals within litters. For F1 adult echocardiography and cardiac mRNA expression, PROC MIXED was used to evaluate the main effects of dose and sex and the dose*sex interaction with litter included as a random variable. Histopathological changes were analyzed with the Cochran-Armitage Trend Test using PROC FREQ in SAS.

Maternal and fetal/neonatal liver RT-qPCR gene expression data were analyzed using the comparative cycle threshold (C_T) method. Melt curve analyses were conducted for all genes at all doses, and any well not displaying a clear peak was assigned a C_T value of 35. This is critical for the evaluation of genes that are typically unexpressed in control liver, but are upregulated as a result of exposure. Delta C_T values were calculated using the equation 2^−ΔΔC_T and normalized to the mean C_T value of at least 2 housekeeping genes for each tissue and gene array. We selected housekeeping genes that did not display significant treatment effect of exposure (fetal liver: Actb, B2m; maternal liver: Actb, Hprt1, Rplp1). Delta C_T values were then converted to fold-induction by dividing the treated replicate delta C_T by the mean delta C_T of the control replicates for each gene. Fold induction values were log₁₀ transformed prior to ANOVA to correct for heterogeneity of variance. ANOVA p-values were then corrected using False Discovery Rate adjustment in Prism (“Two-stage step-up method of Benjamini, Krieger, and Yekutieli”; Desired FDR = 5%).

Several endpoints were also analyzed by analysis of covariance (ANCOVA). Dam bodyweight on PND2 and PND21 were analyzed using initial (GD8) bodyweight as a covariate with dose to account for maternal baseline bodyweights. Cumulative maternal gestational weight gain was analyzed with litter size as a covariate with dose. Pup birthweight (i.e., adjusted birthweight) was analyzed with total litter size and gestational age (birthtime) as covariates with dose. AGD on PND2 was analyzed with body weight as a covariate with dose to account for the influence of larger body size on larger AGD. Statistical evaluation and interpretation of organ weight effects can be difficult for organs that show a strong association with body weight, particularly when body weight is also affected by dose (Haseman et al. 2001; Lazic et al. 2020). Here, organ weights (maternal and offspring) were analyzed on an absolute weight basis, with terminal body weight as a covariate with dose in ANCOVA (i.e., adjusted liver weight), and as relative organ weights (mg organ per g bodyweight).

Mixture analyses were conducted using the RPF approach as reported in our prior PFAS mixture study (Conley et al. 2022a) and similar to the method described by Van Der Ven et al. (2022). Mixture evaluation was limited to endpoints that were similarly recorded in the present mixture experiment and from our previously published individual chemical studies for HFPO-DA, NBP2, and PFOS (Conley et al. 2021; Conley et al. 2022a; Conley et al. 2022b; Conley et al. 2019) (Summary data for individual chemical endpoints provided in Tables S1, S2). Data were normalized to vehicle controls and fit with four parameter logistic regressions (“Sigmoidal, 4PL, X is log(concentration)” in Prism) with top constrained to 100% (for increasing effects, as % increase from control) or bottom constrained to 0% (for decreasing effects, as % of control) and the remaining three parameters unconstrained (i.e., slope, ED₅₀, and top or bottom). Endpoints where the maximum (100%) or minimum (0%) responses were not defined were fit with four parameter logistic regressions with the top constrained to a high value for increasing effects (e.g., 1000%, because the top cannot be estimated and so that Prism would not attempt to “assign” an arbitrary top value) or the bottom unconstrained for decreasing effects. We then statistically compared slope parameters between HFPO-DA, NBP2, and PFOS using extra sums-of-squares F test and estimated equivalent effective dose levels (ED) based on the extent of the dose responses (e.g., ED₂₀). For endpoints with statistically similar slopes (p>0.05), the ED values were used to calculate RPFs (HFPO-DA = index chemical). The NBP2 and PFOS RPFs were then used to adjust the NBP2-only, PFOS-only and HFPO-DA+NBP2+PFOS mixture doses into HFPO-DA equivalents. For example, if NBP2 RPF = 10 and PFOS RPF = 20, then given a 1 mg/kg PFOS dose (20 mg/kg HFPO-DA equivalent) and 3 mg/kg NBP2 dose (30 mg/kg HFPO-DA equivalent) in the mixture, a total of 50 mg/kg HFPO-DA equivalent was added to the HFPO-DA mixture dose level. The RPF adjusted data for HFPO-DA-only, NBP2-only, PFOS-only, and the HFPO-DA+NBP2+PFOS mixture dose response data were all graphed together, and Prism was used to determine if one curve adequately fit all of the data sets using the Extra sum-of-squares F test. Assuming dose additivity, the RPF-adjusted dose response curves for the 3 individual PFAS and the PFAS mixture could be fit by a single nonlinear regression curve.

Maternal liver weight from GD8-PND2 exposure was not evaluated for NBP2 or the present mixture and thus could not be analyzed for mixture effects. Further, for maternal liver weight from GD14-18 exposure and pup liver weight from GD8-PND2 exposure there were effects from the mixture and from HFPO-DA alone, but no significant effects of PFOS or NBP2 alone and thus RPFs could not be calculated. However, to assess if there was any effect of mixture exposure on liver weights (absolute and relative), we statistically compared the dose response curves for HFPO-DA alone and HFPO-DA in the present mixture. A significant shift in the HFPO-DA dose response curve between the individual study and the mixture indicates a cumulative effect of co-exposure to NBP2 and PFOS. Pup absolute liver weight displayed a high dose non-monotonic response and was fit with a 5 parameter regression (“Bell shaped, x is log(concentration)”), while maternal absolute liver weight and maternal and pup relative liver weights were fit with 4 parameter logistic regressions as described above.

3. RESULTS

3.1. Postnatal study (GD8 – PND2 maternal exposure)

3.1.1. F1 endpoints from birth through puberty

Gestation length (measured as pup delivery time after 12:00AM on GD22/PND0) was shorter than control in all dose groups but not statistically significant (p=0.065 in the 100% dose group, ~0.5-day earlier delivery time than controls, Tables 2, S3). Pup birthweight was lower than control in all dose groups (ANOVA p<0.0001) and statistically significant at ≥33.3% dose (p=0.022, Table 2). Many pups in the 100% dose group were either stillborn or died shortly after birth, prior to the completion of litter delivery. Mean total litter size of delivered pups was similar across dose groups (range 10.6±1.1 – 13±0.3). Newborn pup liver histopathological evaluation identified a significant reduction in hepatic glycogen accumulation score based on H&E staining (p<0.0001; Table S4) but not PAS staining. Pup serum on PND0 displayed significantly reduced total protein (24% reduction) and ALT (37% reduction), and significantly elevated cholesterol (57% increase) and triglycerides (54% increase) in the 100% dose group (Table S3).

Table 2.

Maternal and neonatal endpoints (mean ± standard error) from GD8-PND2 maternal oral exposure to HFPO-DA+NBP2+PFOS mixture

	Control	1%	3.3%	10%	33.3%	100%
Maternal endpoints	n=5	n=5	n=5	n=5	n=5	n=5
GD8 bodyweight (g)	270.0 ± 5.3	274.4 ± 4.3	272.1 ± 4.6	278.3 ± 4.4	278.6 ± 5.3	281.8 ± 8.6
GD22 bodyweight (g)	406.2 ± 15.0	437.4 ± 8.9	431.1 ± 11.0	439.8 ± 7.2	418.0 ± 14.9	357.0 ± 14.3 **
GD8-GD22 weight gain (g)	136.1 ± 11.4	163.0 ± 6.7	159.0 ± 7.8	161.5 ± 4.4	139.4 ± 12.0	75.2 ± 16.4 **
PND2 bodyweight (g)	322.3 ± 10.4	332.0 ± 8.5	339.5 ± 10.0	340.2 ± 8.9	318.7 ± 8.9	249.6 ± 15.0 **
GD8-PND2 weight gain (g)	52.3 ± 5.9	57.6 ± 5.7	67.5 ± 6.8	61.9 ± 5.6	40.1 ± 7.1	−32.1 ± 18.3 **
Uterine implants	12.2 ± 0.9	13.4 ± 0.5	11.6 ± 0.5	11.8 ± 0.7	14.2 ± 0.4 *	12.4 ± 0.9
PND21 bodyweight (g)	317.6 ± 13.4	313.0 ± 5.2	326.6 ± 15.4	330.1 ± 9.9	325.3 ± 10.4	. ± .
GD8-PND23 weight gain (g)	47.5 ± 9.5	38.6 ± 1.2	54.5 ± 12.0	51.8 ± 7.7	46.7 ± 11.7	. ± .
Liver weight PND23 (g)	14.1 ± 0.8	14.2 ± 0.4	14.8 ± 1.1	15.1 ± 0.8	16.1 ± 0.9	. ± .
Adjusted Liver weight PND23 (g) a	14.4 ± 0.4	14.8 ± 0.4	14.6 ± 0.4	14.6 ± 0.4	16.0 ± 0.4 **	. ± .
Neonatal endpoints
Delivery time b	29.7 ± 5.1	22.8 ± 4.0	23.9 ± 4.7	24.3 ± 4.9	20.7 ± 3.9	18.0 ± 2.2
Litter size (# pups)	10.6 ± 1.1	13.0 ± 0.3	11.0 ± 1.0	11.8 ± 0.7	11.6 ± 1.0	12.0 ± 0.6
Birthweight (g)	7.0 ± 0.2	6.6 ± 0.3	6.8 ± 0.2	6.9 ± 0.3	6.2 ± 0.1 *	5.0 ± 0.2 **
Adjusted birthweight (g) c	6.8 ± 0.2	6.7 ± 0.2	6.7 ± 0.2	6.9 ± 0.2	6.3 ± 0.2	5.2 ± 0.2 **
PND2 bodyweight (g)	8.90 ± 0.20	8.21 ± 0.25 *	8.23 ± 0.32 *	8.20 ± 0.20 *	6.85 ± 0.14 **	. ± .
Survival (% implants)	81.5 ± 5.1	96.9 ± 3.1	80.7 ± 12.4	100.0 ± 0.0 *	67.5 ± 5.5	0.0 ± 0.0 **
Survival (% live born)	98.2 ± 1.8	100.0 ± 0.0	83.8 ± 6.7 *	100.0 ± 0.0	87.5 ± 5.9 *	0.0 ± 0.0 **
Female AGD	2.94 ± 0.09	2.87 ± 0.06	2.82 ± 0.07	2.76 ± 0.06	2.86 ± 0.14	. ± .
Male AGD	5.25 ± 0.14	5.23 ± 0.12	5.22 ± 0.08	5.23 ± 0.12	4.94 ± 0.05	. ± .
Liver weight PND2 (g)	0.35 ± 0.01	0.35 ± 0.02	0.40 ± 0.03	0.49 ± 0.02 **	0.44 ± 0.03 *	. ± .
Adjusted liver weight PND2 (g) a	0.32 ± 0.02	0.35 ± 0.01	0.39 ± 0.01 **	0.47 ± 0.01 **	0.49 ± 0.02 **	. ± .

^aOrgan weight adjusted for body weight using ANCOVA;

^bDelivery time = hrs after 12:00AM PND0;

^cBirthweight adjusted for delivery time and littersize using ANCOVA; AGD=anogenital distance; GD=gestation day; PND=postnatal day;

^*p<0.05;

^**p<0.01

By PND2 there were no surviving pups in the 100% dose group. Pup survival based on the number of live born pups was also significantly reduced in the 3.3% (p=0.011, 15% reduction) and 33.3% (p=0.05, 11% reduction) dose groups compared to control (Table 2). Pup mortality relative to the number of uterine implants was approximately twice as high in the 33.3% dose group (32.5±5.5%) as the controls (18.5±5.1%) but was not statistically significant (p=0.12), while the 10% dose had significantly lower mortality than controls (100% survival, p=0.041) (Table 2). Total mean pup bodyweight on PND2 was significantly lower than control in all dose groups (Table 2). There was a significant sex effect (p<0.0001) on pup bodyweight with females smaller than males from the same litter in controls and treated groups (Table S3). Female pup bodyweight on PND2 was significantly lower than control at ≥3.3% dose, while male pup bodyweight was significantly lower at 33.3% dose. Pup absolute liver weight was significantly greater than control at ≥10% dose, while ANCOVA bodyweight-adjusted and relative liver weight were significant at ≥3.3% dose (Tables 2, S3). There was no effect of treatment on male or female AGD with or without including bodyweight as a covariate (Tables 2, S3).

Mean pup body weight on PND9 remained lower than control in all dose groups and was significant at 33.3% dose (18% weight reduction, p=0.0015), despite maternal dosing ending on PND2 (Table S3). By PND13 there was a highly significant sex effect (p<0.0001) with females smaller than males in controls and treated groups, but no significant effect of dose on pup bodyweight compared to control (Table S3). Further, on PND13 there was no effect of treatment on male or female nipple retention (Table S3). In contrast, at weaning on PND23 there was a significant reduction in female pup body weight at 33.3% dose (11% reduction, p=0.019), while male pup bodyweight was lower than control but not significant (7.8% reduction, p=0.11) (Table S3).

There were no significant effects of treatment in F1 females on age at vaginal opening or bodyweight on PND38 at the tested mixture doses. Bodyweight at time of vaginal opening was significantly reduced in the 33.3% dose group (13.6% reduction, p=0.028); however, there was no significant effect after correcting for bodyweight at weaning on PND23 by ANCOVA. In F1 males there were no statistically significant effects of treatment on age at PPS, bodyweight at PPS, or bodyweight on PND43 (Table S3).

3.1.2. Adult F1 echocardiography and cardiac biomarker assessment

Ultrasound echocardiography endpoints were evaluated using a mixed model ANOVA that included main effects of dose and sex, interaction of dose*sex, and random effect of litter. Treated F1 animals (33% dose) displayed a significant main effect of dose compared to control with increased left ventricular anterior wall thickness during systole (8.7% increase, p=0.018) and diastole (10.4% increase, p=0.042), but with no significant main effect of sex or sex*dose interaction (Figure 1, Table S5). The remaining echocardiography endpoints had no significant main effects of dose or dose*sex interactions; however, many had a strong main effect of sex with large differences in male versus female measurements (Table S5). One of 10 control males displayed cardiomyopathy and several of the cardiac measures were obviously different than any other control animals. Statistical significance of mixed model analyses was not affected with the outlier control included or removed.

Figure 1.

Left ventricular anterior wall (LVAW) thickness in adult F1 animals (~6 months old) from the control (0%) and top surviving mixture dose group (33.3%) measured in vivo by echocardiography. There was no effect of sex using a mixed model ANOVA, thus analyses were conducted on litter means irrespective of sex, while graph displays all individual replicate measurements with female measurements in open circles and male measurements in black semicircles with line at treatment mean.

Similarly for serum cardiac biomarkers, there were no significant main effects of dose or the interaction of dose*sex, but many endpoints had a significant main effect of sex (Table S6). Heart gene expression analyses identified 1 of 84 genes exhibited increased expression of natriuretic peptide precursor B (Nppb) greater than control in exposed males (3.2-fold increase relative to controls; p< 0.08), but not in females (p > 0.3) (Table S7).

3.1.3. Adult F1 necropsy

There were no treatment effects on adult F1 female body weight or organ weights including paired ovaries, paired kidneys, liver, heart, and paired adrenal glands (Table S8). There were no treatment effects on adult F1 male body weight or organ weights including paired testes, paired epididymides, liver, paired kidneys, heart, and paired adrenal glands and neither male nor female offspring displayed any reproductive tract or other gross abnormalities (Table S8).

3.1.4. Maternal endpoints

Maternal gravid body weight on GD22 (12% reduction, p=0.0092), body weight on PND2 (22.5% reduction, p<0.0001), and gestational weight gain from GD8-GD22 (45% reduction, p=0.0004) were significantly reduced in the 100% dose group compared to control (Tables 2, S9). In the 100% dose group maternal rats lost 32.1g body weight from GD8-PND2, while controls gained 52.3g (p<0.0001) (Tables 2, S9). In the lower dose groups (1, 3.3, and 10%) body weights on GD22 and PND2, and weight gains from GD8-22 and GD8-PND2 were greater than control, but these differences were not statistically significant (Tables 2, S9). Maternal absolute liver weight was higher than control in all dose groups at necropsy on PND23 but not significant, while ANCOVA body weight-adjusted and relative liver weights were significantly increased in the 33.3% dose group (~11% increase, p=0.0074) despite dosing ending on PND2 (Tables 2, S9). Due to complete litter loss maternal rats in the 100% dose group were euthanized on PND2 and absolute and relative liver weights were remarkably greater (~50% larger) than similar measures from maternal rats on PND23 (Table S9).

3.2. Fetal study (GD14-18 maternal exposure)

3.2.1. Maternal and fetal endpoints on GD18

Dam mean body weight on GD18 was unaffected by treatment, but GD14-18 body weight gain was significantly elevated at 33.3% dose compared to controls (54% increase; Table S10). All dam liver measurements (absolute, ANCOVA body weight-adjusted, and relative) were significantly increased at ≥3.3% dose compared to controls (Table S10). There were no treatment effects on fetal body weights or fetal liver weights (Table S10).

3.2.2. Maternal thyroid hormones and clinical chemistry on GD18

Maternal serum thyroid hormone concentrations (total T3 and T4) on GD18 were significantly reduced at ≥33.3% dose compared to controls (Table S10). Maternal serum total cholesterol (35% reduction), triglycerides (50% reduction), total protein (9% reduction) and globulin (15% reduction) were significantly reduced, while albumin:globulin (13% increase) was significantly elevated at 100% dose (Table S10).

3.2.3. Maternal and fetal liver gene expression – RT-qPCR

GD18 fetal livers were evaluated using a PPAR Targets RT-qPCR array and a total of 17 genes were significantly altered (Figure 2, Table S11). Fabp1 and Hmgcs2 were significantly upregulated at all dose levels, while Acaa2, Acox1, Angptl4, Cpt1a, Cpt1b, Cpt2, Ehhadh, and Pck1 were significantly upregulated at ≥3.3% dose compared to controls. Similarly, maternal livers were evaluated using a Fatty Acid Metabolism RT-qPCR array and 39 genes were significantly altered. Two genes associated with peroxisomal (Acot3, Ehhadh) and one associated with mitochondrial fatty acid metabolism (Cpt2) were significantly upregulated at the lowest dose (Figures S3, S4; Table S12). This array contains genes in common with the PPAR Targets array and many of those that were affected in our HFPO-DA study were significantly upregulated by the present mixture (Acaa2, Acadl, Acadm, Acox1, Acsl1, Acsl3, Cpt1a, Cpt1b, Cpt2, Ehhadh, Fabp1, Hmgcs2, Ldha, and Slc27a2 (Conley et al. 2021; Conley et al. 2019)). Further, we previously identified upregulation of acyl-CoA thioesterases from NBP2 exposure (Conley et al. 2022b) and, as anticipated, 4 of the most highly upregulated maternal liver genes here were Acot2, Acot3, Acot7, and Acot8. Additional highly upregulated genes included Gpd2 (important in glycolysis and glucose sensing), Crat (fatty acid and amino acid oxidation), Crot (peroxisomal fatty acid oxidation), Decr1 (mitochondrial fatty acid oxidation), and Lipe (hydrolysis of diacylglycerols) among others. A few genes were downregulated but only significant in the 100% dose group including Prkaca, Cpt1c, Acadsb, and Prkab1 (Figure S4, Table S12). Overall, approximately half of the genes included on the fatty acid metabolism array were significantly altered by the PFAS mixture after correcting for false discovery and demonstrate extensive maternal liver effects on fatty acid metabolism from exposure to the 3 PFAS mixture (Figure S4, Table S12). Further, comparison of the fetal and maternal liver gene expression profiles with those from individual studies of HFPO-DA, NBP2, and PFOS across similar oral doses indicates that the mixture liver gene profiles were predominantly affected by HFPO-DA (Figures 2, S3).

Figure 2.

Comparison of fetal liver gene expression from maternal oral exposure from gestation day (GD) 14-18 to each PFAS individually and the mixture of all three. The maternal oral doses (mg/kg/d) for the individual PFAS approximately correspond to the doses of each across the fixed-ratio mixture doses. For example, the 100% mixture dose contained 110 mg/kg HFPO-DA, 10 mg/kg NBP2, and 3 mg/kg PFOS (see Table 1). Genes displayed are those that had ANOVA false discovery rate p<0.05 for the PFAS mixture. Data for single PFAS adapted from previous studies (Conley et al. 2019, Conley et al. 2022b).

3.3. Maternal and offspring serum and liver PFAS concentrations

Maternal serum was collected following GD14-18 maternal oral exposure to the PFAS mixture and the concentration of all three PFAS increased linearly by ~100-fold from low (1%; mean ∑FAS = 1.3 μg/mL) to high (100%; mean ∑FAS = 115.1 μg/mL) dose groups (Table S13). Within a given dose level the serum HFPO-DA and NBP2 concentrations were highly similar (range ~0.5 – 40 μg/mL), while the PFOS concentrations were approximately half of the HFPO-DA and NBP2 concentrations (range ~0.2 – 20 μg/mL).

Pup samples were collected on PND0 (liver) and PND2 (serum and liver, no samples from 100% dose due to pup mortality) for determination of PFAS concentrations following maternal GD8-PND2 exposure. Across all pup samples NBP2 concentrations were highest (serum range ~1-20 μg/mL, liver range ~2-80 μg/g) of the three PFAS while PFOS concentrations were lower but within a factor of ~2-fold (serum range ~0.5-14 μg/mL, liver range ~2-30 μg/g) (Figure 3, Table S13). In contrast, the HFPO-DA concentrations in pup serum and liver were 10- to 100-fold lower than NBP2 and PFOS, particularly on PND2 (serum range 0.05-0.5 μg/mL, liver range ~0.04-0.6 μg/g) (Figure 3, Table S13). The pup liver concentrations of HFPO-DA decreased ~3-10-fold from PND0 to PND2, while NBP2 and PFOS pup liver concentrations remained relatively stable. These differences in internal dose may be due to rapid elimination of HFPO-DA and/or lower maternal lactational transfer compared to NBP2 and PFOS.

Figure 3.

Internal dosimetry in newborn (PND0) and neonatal (PND2) pups exposed in utero to the PFAS mixture from the Postnatal study (GD8-PND2 maternal exposure). Serum (open bars) and liver (hatched bars) concentrations are reported as parts per million (ppm) for each of the three compounds. No pups survived to PND2 in 100% dose group so no concentrations reported.

Maternal serum and liver were collected from dams in the 100% dose group on PND2 because all pups had died at that dose (Figure S5, Table S13). Maternal PND2 serum concentrations were relatively similar across all three PFAS with NBP2 and HFPO-DA (~100 μg/mL) slightly greater than PFOS (~80 μg/mL). In comparison to maternal serum levels from the GD14-18 fetal study, concentrations were ~2-3-fold greater in maternal animals exposed GD8-PND2 (Figure S5A). In contrast to serum, maternal liver concentrations of NBP2 (~170 μg/g) and PFOS (~140 μg/g) were approximately 3-fold greater than HFPO-DA liver concentrations (~50 μg/g) on PND2 (Figure S5B). Pup liver concentrations from PND0 mirrored PND2 maternal livers from the 100% dose level (Figure S5B). Comparison of the serum and liver concentrations measured in the present 3 PFAS mixture with previously published maternal levels from similar oral doses and intervals in individual PFAS studies indicates that internal doses were similar and thus co-exposure did not appear to substantial alter internal dosimetry (Figure S6). In contrast, the PND0 pup liver levels for NBP2 were ~10-fold greater and the PND2 levels were ~5-fold greater in the present study with the PFAS mixture than those reported from our prior study with NBP2 alone (Figure S6). We did not measure pup liver PFOS concentrations from the individual chemical study (Conley et al. 2022a)and therefore were unable to compare the results of individual chemical exposure to the present mixture study.

3.4. Mixture evaluation

Similar to our prior PFAS mixture study (Conley et al. 2022a) and the method described by Van Der Ven et al. (2022), our goal was to calculate experimental RPFs across a range of effects that were common to all three PFAS in the mixture in order to evaluate dose additivity.

We plotted individual chemical dose response data for HFPO-DA, NBP2, and PFOS from our prior publications (Conley et al. 2021; Conley et al. 2022a; Conley et al. 2022b; Conley et al. 2019) and statistically compared dose response curve slopes and determined effective dose (ED) levels within the range of available data. Then using HFPO-DA as the index chemical (i.e., RPF=1) we calculated NBP2 and PFOS RPFs for 4 maternal endpoints from GD14-18 exposure (GD18 serum total T3, GD18 serum total T4, GD18 serum total cholesterol, and GD18 serum triglycerides), 3 maternal endpoints from GD8-PND2 exposure (GD22 body weight, GD8-22 body weight gain, and PND2 body weight) and 4 offspring endpoints from GD8-PND2 exposure (survival - % live born, survival - % uterine implants, adjusted birthweight, PND2 bodyweight). Across all 11 endpoints, regardless of dosing interval or life stage, the RPFs for PFOS varied by 13.5-fold (RPF range, 2.9 – 38.6) and NBP2 varied by 16.1-fold (RPF range, 0.9 – 13.7) (Table 3). Specifically, for pup-only endpoints from GD8-PND2 exposure the RPFs ranged 4.9-fold for PFOS (range, 7.8 – 38.6) and 16.1-fold for NBP2 (range, 0.9 – 13.7). Maternal RPFs from the Postnatal study were highly consistent across the range of body weight associated endpoints (PFOS 1.3-fold, NBP2 1.6-fold), while short term GD14-18 endpoints ranged 5.6-fold for PFOS and 3.8-fold for NBP2.

Table 3.

HFPO-DA, NBP2, and PFOS dose response slope, effect dose (ED) levels (in mg/kg/d), and relative potency factors (RPF; HFPO-DA=index chemical)

	Slope	ED level	HFPO-DA ED	PFOS ED	NBP2 ED	HFPO-DA RPF	PFOS RPF	NBP2 RPF
Maternal endpoints (GD14-18 exposure)
Serum triglycerides (GD18)	−0.7	20	51.6(12-222)	18.1(3.1-107)	17.9(3.1-104)	1(0.2-4.3)	2.9(0.5-16.8)	2.9(0.5-16.9)
Serum total T3 (GD18)	−0.93	50	86.5(38.4-195)	12.6(5.3-29.9)	50.7(14.6-177)	1(0.4-2.3)	6.9(2.9-16.3)	1.7(0.5-5.9)
Serum cholesterol (GD18)	−0.7	20	111(54.6-224)	10(4.7-21.4)	33.3(10.1-111)	1(0.5-2)	11(5.2-23.5)	3.3(1-11)
Serum total T4 (GD18)	−1.0	50	200(79-507)	12.6(5.4-29.5)	31(10.2-94.8)	1(0.4-2.5)	15.9(6.8-37.1)	6.5(2.1-19.7)
Maternal endpoints (GD8-PND2 exposure)
Bodyweight (PND2)	−0.72	20	264(157-444)	9.6(4.1-22.2)	50.9(23.2-112)	1(0.6-1.7)	27.5(11.9-63.9)	5.2(2.4-11.4)
Weight gain (GD8-22)	−1.1	20	87.3(47.4-161)	2.8(1.6-4.9)	25.9(13-51.7)	1(0.5-1.8)	31.6(17.7-56.2)	3.4(1.7-6.7)
Bodyweight (GD22)	−0.81	20	381(163-892)	10.3(3.6-30)	119(19.9-708)	1(0.4-2.3)	36.9(12.7-107)	3.2(0.5-19.2)
Neonatal endpoints (GD8-PND2 exposure)
Adjusted birthweight	−0.76	10	57(27.1-119.9)	7.3(2.7-20)	66.7(17.3-258)	1(0.5-2.1)	7.8(2.9-21.5)	0.9(0.2-3.3)
Bodyweight (PND2)	−0.71	20	35.5(21.2-59.5)	3.7(1.6-8.3)	26.3(6.2-112)	1(0.6-1.7)	9.6(4.3-21.6)	1.3(0.3-5.7)
Pup survival (implants)	−18.8	50	121(78.8-187)	3.4(2.1-4.7)	9.4(6.1-14.4)	1(0.6-1.5)	35.9(25.6-56.7)	12.9(8.4-20)
Pup survival (live born)	−7.1	50	117(106-127)	3(2.9-3.2)	8.6(7.4-9.5)	1(0.9-1.1)	38.6(36.7-40.2)	13.7(12.4-15.9)

Values represent mean (95% confidence interval); GD=gestation day; PND=postnatal day

To assess the hypothesis of dose additivity, we plotted data from the individual chemical studies referenced above with the present mixture study in index chemical equivalent concentrations (i.e., HFPO-DA mg/kg/d equivalents) and compared dose response curves (Figure 4). The RPF-adjusted index chemical equivalent doses across the three individual studies and the mixture study displayed statistically equivalent logistic regression parameters for all endpoints except for maternal GD8-22 body weight gain (Figure 4A), GD22 body weight (Figure S7A), and PND2 bodyweight (Figure S7B). Similar to our prior mixture study with PFOA and PFOS, these endpoints displayed less than additive effects in the mixture, which were primarily driven by the lower dose groups (i.e., ≤10%) displaying slightly greater body weights and weight gain than control, while the reduction in weights at the top dose (100%) aligned well with the individual PFAS studies. The statistical alignment of RPF-adjusted dose response curves for the remaining 8 endpoints supports the hypothesis of dose additivity for the majority of maternal and offspring effects evaluated here.

Figure 4.

Relative potency factor (RPF) analyses of maternal and neonatal endpoints shared across the three individual PFAS (HFPO-DA, NBP2, PFOS) and the mixture. HFPO-DA was used as the index chemical and RPFs reported in Table 3 were used to adjust all doses to HFPO-DA equivalents. The index chemical equivalent data sets were then statistically compared for similarity. Endpoints conforming to dose addition could be fit with a single dose response curve (dashed line with 95% prediction intervals in shading). For maternal weight gain the mixture was statistically different from the individual chemical data sets. Individual PFAS data were adapted from previous publications (Conley et al. 2019, 2022a, 2022b). Data points represent mean +/− standard error.

Additionally, we evaluated the dose response curves for increased pup (from GD8-PND2 exposure) and maternal (from GD14-18 exposure) absolute and relative liver weights as a function of maternal HFPO-DA dose. In the NBP2 or PFOS alone studies these PFAS did not increase liver weights at the doses tested and thus RPFs could not be calculated. Maternal liver weight from GD8-PND2 exposure was significantly increased by HFPO-DA alone (Conley et al. 2021) and PFOS alone (Conley et al. 2022a); however maternal liver weight at this dosing interval was not measured here or in the NBP2 alone study (Conley et al. 2022b) and could not be analyzed using the present mixture-based approaches. Instead, to evaluate if the mixture had an impact on GD18 maternal and PND2 pup liver weights, we plotted the HFPO-DA alone and mixture dose response curves on a HFPO-DA oral dose basis. For these liver endpoints exposure to the mixture resulted in significantly different dose response functions that were left shifted (towards effects at lower exposure levels) compared to the HFPO-DA alone data (Figure 5). Pup absolute liver weight decreased at higher dose levels and was fit with a bell curve (similar to our prior PFOA+PFOS mixture study, see Conley et al. (2022a) Figure S1) and the mixture shifted the dose response toward lower doses and increased the upper plateau of the bell-shaped curve. Despite a lack of dose response data for NBP2 or PFOS for these endpoints there are likely molecular alterations that resulted in HFPO-DA dose response curves being shifted towards effects at lower doses than HFPO-DA alone. Further, the stark contrast between the absolute and relative pup liver weight dose response curves demonstrates the difficulty in analyzing and interpreting organ weight data when there is a relationship between body weight and organ weight and when the treatment also affects body weight (see (Haseman et al. 2001; Lazic et al. 2020)).

Figure 5.

HFPO-DA dose response curve shift between HFPO-DA exposure alone (data from Conley et al. 2019, 2021) and in the present 3 PFAS mixture for pup (GD8-PND2 maternal exposure) and maternal (GD14-18 exposure) absolute and relative liver weights. Change in liver weight calculated as percent increase relative to concurrent vehicle controls and dose represents maternal oral exposure either alone or the concentration of HFPO-DA in the present mixture. For these endpoints the individual studies for NBP2 and PFOS indicated no significant effect at the doses tested (Conley et al. 2022a, 2022b). Despite a lack of effect for NBP2 and PFOS, both pup and maternal liver weights were significantly elevated at lower maternal oral doses in the 3 PFAS mixture than for HFPO-DA alone. Pup absolute liver weight decreased at higher doses and data were fit with a 5-parameter bell curve, all other data fit with 4-parameter logistic regressions. Data points represent mean +/− standard error.

DISCUSSION

Occurrence of multiple PFAS in human and environmental matrices is widespread and several international efforts are underway to address potential human and ecological health risks from exposure to mixtures of PFAS (EFSA et al. 2020; RIVM 2018; USEPA 2023c). Limited toxicology research has been conducted across the broad spectrum of structures that comprise the thousands of compounds categorized as PFAS and even less work has been conducted to directly address hypotheses of additivity from mixture exposures in mammalian systems or other taxonomic groups. Recently, we demonstrated that exposure to a combination of PFOA+PFOS produced mixture effects that conformed to dose additivity, but often exceeded estimates modeled by response additivity (Conley et al. 2022a). Here we extended this line of inquiry to a more complex mixture containing 3 PFAS, including two non-traditional, ether-linked compounds (HFPO-DA and NBP2). Similar to our prior mixture experiment with PFOA and PFOS, the present mixture contained both sulfonate (PFOA, NBP2) and carboxylate (HFPO-DA) PFAS. Further, instead of terminating the study at PND2 like our previous experiment (although maternal dosing ended at PND2), we aged F1 animals to adulthood to investigate a range of endpoints with a focus on reproductive development and cardiovascular effects. Exposure to the fixed-ratio mixture of HFPO-DA, NBP2, and PFOS exhibited numerous effects consistent with the effects of individual PFAS alone including reduced F1 body weights, increased liver weights, reduced serum thyroid hormones, altered serum lipids, and changes in liver gene expression. Shared endpoints across individual PFAS again conformed to dose addition using an RPF approach similar to the method described by Van Der Ven et al. (2022); however, RPFs spanned a wide range across endpoints for the compounds. Further, in the absence of dose response data for estimating RPFs for all three PFAS, we demonstrate that for pup (GD8-PND2 exposure) and maternal (GD14-18 exposure) liver weights, co-exposure to NBP2 and PFOS significantly shifted the HFPO-DA dose response curves towards effects at lower maternal oral doses. As F1 animals aged, there were no statistically significant effects on reproductive developmental endpoints (i.e., AGD, nipple retention, onset of puberty, reproductive tissue weights at maturity or gross malformations). Cardiac morphology in the form of increased left ventricular anterior wall thickness appears to be an effect from fetal exposure to the PFAS mixture although no other significant cardiac morphological, functional or transcriptomic changes were noted.

The present mixture was designed to test a hypothesis of additivity across a range of in vivo key events and adverse outcomes in maternal and F1 animals based on the individual chemical dose responses. The study was not designed to represent or mimic a specific human exposure scenario; however, it is highly plausible that human exposure to this assortment of 3 PFAS has occurred in some geographic locations. PFOS remains nearly ubiquitously detected in human and environmental samples from various monitoring studies across the globe. Further, Kotlarz et al. (2020) reported that NBP2 (along with PFOS and other PFAS) was detected in 99% of human serum samples (n=341) from a Wilmington, NC cohort. The 95^th percentile adult serum levels of PFOS (28.2 ng/mL) and NBP2 (8.5 ng/mL) from Kotlarz et al. (2020) were 8-and 63-fold lower than the maternal rat GD18 serum levels from the lowest dose (1%) used in the Fetal study detected here, respectively. Reported detections of HFPO-DA in human serum are rare, but in impacted communities detection of HFPO-DA in drinking water has been common, for example in the same eastern NC study area as the NBP2 serum detections (Sun et al. 2016). The Anniston Alabama Community Health Surveys (ACHS) reported maximum HFPO-DA serum levels of 1.0 ng/mL, which was 600-fold lower than the maternal rat GD18 serum levels from the lowest dose (1%) used in the Fetal study detected here. It is important to consider the short exposure intervals used in the present studies (5 days and 17 days dosing) when putting these margins of exposure into context from a human exposure perspective. More broadly, across the PFAS class, targeted studies with limited analyte sets specifically focused on pregnant women and/or infants have reported nearly ubiquitous detection of multiple PFAS, including PFOA, PFOS, PFNA, and PFHxS, with additional detections depending on specific regional exposure factors (Bao et al. 2022; Berg et al. 2014; DeLuca et al. 2023; Dereumeaux et al. 2016; Kashino et al. 2020; Woodruff et al. 2011).

One of the novel findings from the present study was the statistically significant increase in left ventricular anterior wall thickness in both male and female adult F1 animals following in utero exposure to the PFAS mixture. To our knowledge there are no other mammalian studies reporting PFAS-induced effects on cardiac development, but right ventricular hypertrophy (along with elevated heart rate, cardiomyocyte hypertrophy, and increased myocardial fibrosis) has been described in aged chickens (at 1 month and 3 months old) exposed in ovo to PFOA during fetal development (Jiang et al. 2012; Ni et al. 2023). Here there was a ~10% and ~12% increase in left ventricular anterior wall thickness during systole and diastole, respectively. PFAS exposure has been linked to increased risk of hypertensive disorders of pregnancy in humans (Preston et al. 2022) and animal models have reported similar findings of increased gestational blood pressure (Crute et al. 2022; Crute et al. 2023) and placental insufficiency (Blake et al. 2020; Blake and Fenton 2020) from PFAS exposure, which can impact fetal cardiovascular development and offspring cardiovascular health (Burton and Jauniaux 2018; Thornburg et al. 2010). These findings are also aligned with a recent report demonstrating that PFAS exposure in women is linked with altered DNA methylation of genes associated with cardiac hypertrophy (Xu et al. 2020).

While the mechanisms driving these observed cardiac responses are unclear, left ventricular hypertrophy, especially concentric hypertrophy characterized by wall thickening, is frequently caused by pressure overload due to hypertension and aortic stenosis and increases risk for cardiovascular morbidity and mortality (Lorell and Carabello 2000). Left ventricular hypertrophy has been linked with exposure to a broad array of drug and chemical toxicants (Chen et al. 2001) as well as PPAR agonists. For example, Engle et al. (2010) demonstrated that prolonged exposure to suprapharmacological doses of a PPAR α/γ dual agonist in adult rats caused left ventricular hypertrophy, characterized by increased heart weight (~30% increase), left ventricular anterior and posterior wall thickness and inner diameter, increased cardiac output, decreased arterial and left ventricle blood pressure, and increased heart rate. Here, cardiac output and heart rate were not affected by treatment and blood pressure was not measured. The fetal heart accumulates glycogen during mid-gestation, which is critical for cardiogenesis and glycogen synthase knockout mice have high rates of neonatal mortality and abnormal cardiac development (Pederson et al. 2004). We did not measure fetal or neonatal heart glycogen here, however we have observe significantly reduced neonatal liver glycogen from in utero exposure to HFPO-DA (Conley et al. 2021), NBP2(Conley et al. 2022b), PFOA, PFOS, a mixture of PFOA+PFOS (Conley et al. 2022a) and the present mixture. Fetal lung, skeletal muscle and the placenta also store glycogen during gestation and are critical for fetal health and development (Tye and Burton 1980).

Exposure to the PFAS mixture in the present study also appeared to mildly increase cardiac expression of Nppb, a peptide that promotes salt and water excretion and whose abnormal levels may indicate ventricular dysfunction resulting from volume and pressure overload (Yoo 2014). Although the Nppb responses were evident only in males and contrast with the findings by Engle et al. (2010), who found PPAR agonist-induced increases in Nppa and Nppb only in female rats (transcripts increased 130-fold and 8.4-folds, respectively). The data on PFAS and other PPAR agonists, along with the present findings of increased left ventricular anterior wall thickness, suggest that fetal cardiac development may be a target of PFAS gestational exposure; however, given that few statistically significant treatment-related changes were seen here it would be important to confirm and expand examination of cardiac development in future studies.

In addition to glycogen disruption discussed above, data from the present study and our prior PFAS developmental experiments indicate dysregulation of multiple maternal and/or fetal/neonatal metabolic parameters. Here, newborn pups exposed to the mixture displayed elevated serum cholesterol and triglycerides, similar to PFOA (more pronounced on PND2 (Conley et al. 2022a)) and HFPO-DA (Conley et al. 2021), while NBP2 reduced newborn serum triglycerides but not cholesterol (Conley et al. 2022b) and PFOS had no effect on either endpoint (Conley et al. 2022a). Here, the three PFAS in the present mixture reduced maternal serum cholesterol and triglycerides from GD14-18 exposure; however, GD8-PND2 exposure increased maternal triglycerides from HFPO-DA (Conley et al. 2021), PFOA and PFOS exposure (Conley et al. 2022a), while cholesterol was either unaffected (HFPO-DA and PFOA) or reduced (PFOS) (not measured here or for NBP2 alone). PFAS lipid dysregulation appears to have life stage-specific effects that may also be impacted by dosing interval and some differences between PFAS functional groups (carboxylate versus sulfonate). Additional data indicate dysregulation of amino acid metabolism (significantly reduced maternal tryptophan and the tryptophan metabolite indole-3-propionic acid (Conley et al. 2022b)) and highly elevated maternal and neonatal bile acids (Conley et al. 2022a; Conley et al. 2022b). Recent studies have also identified dysregulation of bile acid homeostasis from PFAS exposure (Guo et al. 2021; Wang et al. 2023), as well as markers of altered tryptophan metabolism in young adults (Guo et al. 2022). Future work will continue to investigate the dose responsive changes to these endpoints across PFAS and seek to elucidate mechanistic and key event drivers of these effects.

The carboxylate PFAS (HFPO-DA) appeared to contribute to significant effects in a greater number of endpoints than the sulfonate PFAS (NBP2 and PFOS). For example, at the doses utilized, NBP2 and PFOS contributed very little to the maternal and fetal liver gene expression changes in the PPAR targets array from GD14-18 exposure (Figures 2 and S3). Similarly, NBP2 or PFOS exposure increased maternal liver weight (at PND2), but had no effect on neonatal liver weight (Conley et al. 2022a; Conley et al. 2022b), while HFPO-DA strongly increased liver weight in both maternal and neonatal animals (Conley et al. 2021; Conley et al. 2019). In contrast, all compounds produced neonatal mortality, reduced thyroid hormone concentrations, and reduced pup body weight. A common limitation of evaluating mixture studies is the need for individual chemical dose response data across endpoints. We previously reported that NBP2 increased maternal liver Acot genes, but we do not currently have data for HFPO-DA and PFOS to evaluate mixture effects. Further, because the F1 animals were aged to adulthood, we were not able to measure maternal liver weight at the end of dosing on PND2 to evaluate mixture effects on this endpoint similar to our prior work.

Due to the study design used here, we were not able to collect samples across lifestages at the same timepoints to our prior studies for comparison of internal dosimetry. We did collect maternal serum and liver from the top dose group at PND2 due to complete litter loss, and the concentrations of the three mixture components were similar to those we previously measured at comparable oral doses in single chemical studies. Mixture modeling and analysis based on oral administered dose is dependent on the mixture exposure not impacting toxicokinetics of the mixture components. For example, if the mixture components competed for elimination and thus the same oral dose resulted in greater serum or tissue levels of a given component than what is observed in an isolated exposure, then the mixture predictions based on oral administered dose would underestimate the mixture effects because the internal doses are actually greater than those from the individual chemical study. We did observe an apparent mixture effect on the neonatal liver concentration of NBP2, where the concentration in the mixture study on PND0 was ~4 fold greater (5.6 vs. 24.5 μg/g at maternal dose of 1 mg/kg) and on PND2 ~3-fold greater (10.9 vs. 29.3 μg/g at maternal dose of 1 mg/kg) than the individual chemical study. It is possible that this impact on NBP2 toxicokinetics contributed to the significant shift in the HFPO-DA dose response curve for PND2 pup relative liver weight (Figure 5), but it does not appear that this toxicokinetic effect impacted the accuracy of the RPF approach to predict the mixture-based effects of the other endpoints evaluated here.

In summary, this is the second in vivo mammalian developmental study to report evaluation of dose additive effects of exposure to a PFAS mixture. The results largely conformed to dose addition using an RPF-based approach, which supports both federal and international efforts to conduct cumulative assessment of exposure to multiple PFAS (EFSA et al. 2020; RIVM 2018; USEPA 2023c). Recent work on PFAS RPFs for liver and immune effects (Bil et al. 2023; Bil et al. 2022) have indicated that consensus estimates of PFAS toxic equivalence factor (TEFs) may be warranted, analogous to the approach used for dioxins and dioxin-like compounds (Van den Berg et al. 2006). However, comparable to our prior study, the RPFs estimated from experimental data here varied ~5-fold for PFOS and ~15-fold for NBP2 for the neonatal endpoints from GD8-PND2 exposure. It is clear that PFAS, particularly those with sulfonate and carboxylate functional groups, activate overlapping signaling pathways and are able to act in combination toxicologically to produce additive effects on critical key events and adverse outcomes and assessment and regulatory efforts to protect human and environmental health must account for these combined effects.