Cumulative maternal and neonatal effects of combined exposure to a mixture of perfluorooctanoic acid (PFOA) and perfluorooctane sulfonic acid (PFOS) during pregnancy in the Sprague-Dawley rat

Abstract

Globally, biomonitoring data demonstrate virtually all humans carry residues of multiple per- and polyfluoroalkyl substances (PFAS). Despite pervasive co-exposure, limited mixtures-based in vivo PFAS toxicity research has been conducted. Perfluorooctanoic acid (PFOA) and perfluorooctane sulfonic acid (PFOS) are commonly detected PFAS in human and environmental samples and both produce adverse effects in laboratory animal studies, including maternal and offspring effects when orally administered during pregnancy and lactation. To evaluate the effects of combined exposure to PFOA and PFOS, we orally exposed pregnant Sprague-Dawley rats from gestation day 8 (GD8) to postnatal day 2 (PND2) to PFOA (10–250 mg/kg/d) or PFOS (0.1–5 mg/kg/d) individually to characterize effects and dose response curve parameters, followed by a variable-ratio mixture experiment with a constant dose of PFOS (2 mg/kg/d) mixed with increasing doses of PFOA (3–80 mg/kg/d). The mixture study design was intended to: 1) shift the PFOA dose response curves for endpoints shared with PFOS, 2) allow comparison of dose addition (DA) and response addition (RA) model predictions, 3) conduct relative potency factor (RPF) analysis for multiple endpoints, and 4) avoid overt maternal toxicity. Maternal serum and liver concentrations of PFOA and PFOS were consistent between the individual chemical and mixture experiments. Combined exposure with PFOS significantly shifted the PFOA dose response curves towards effects at lower doses compared to PFOA-only exposure for multiple endpoints and these effects were well predicted by dose addition. For endpoints amenable to mixture model analyses, DA produced equivalent or better estimates of observed data than RA. All endpoints evaluated were accurately predicted by RPF and DA approaches except for maternal gestational weight gain, which produced less-than-additive results in the mixture. Data support the hypothesis of cumulative effects on shared endpoints from PFOA and PFOS co-exposure and dose additive approaches for predictive estimates of mixture effects.

1. Introduction

Monitoring studies across the globe have consistently reported the detection of multiple per- and polyfluoroalkyl substances (PFAS) in human samples and exposure media including drinking water and foods, as well as ecological components including fish, wildlife, surface waters, and soils/sediments (Evich et al. 2022; Sunderland et al. 2019). Exposure to PFAS is widespread and nearly ubiquitously involves multiple compounds including those that are slowly eliminated and bioaccumulative, and those that are more rapidly excreted but may have chronic exposure (Bangma et al. 2022). The assortment of detected PFAS also continues to expand as analytical methods incorporate more target analytes and researchers advance the use of non-targeted screening for organo-fluorine compounds (Koelmel et al. 2022; McDonough et al. 2022). Currently available data on PFAS detections in humans and the environment are dominated by straight-chain carboxylates and sulfonates, but recent studies have reported a range of emerging ether-linked PFAS as well as sulfonamides and chloro perfluoroether compounds, among others (Kotlarz et al. 2020; Washington et al. 2020).

There is a critical need for mixtures-based studies designed to evaluate the joint toxicity and chemical interactions of in vivo exposure to combinations of PFAS to advance PFAS mixtures science and support risk assessment and regulatory actions (EFSA et al. 2020; Ojo et al. 2021). One major factor in study design and health-based agency decision making is whether mixture effects can be most accurately modeled by dose addition (DA) or response addition (RA, also referred to as independent action) (Kortenkamp and Faust 2018; USEPA 2000). Dose addition assumes that components of a mixture are toxicologically similar and contribute to a combined effect based on the ratio of dose to potency for each component chemical. In contrast, RA assumes that components act independently and the mixture effect is estimated from the combined individual component responses based on probability theory. Empirical support for an appropriate model of additivity is necessary for evaluation of mixture interactions including synergy (greater-than-additive effects) and antagonism (less-than-additive effects) (Cedergreen 2014; Martin et al. 2021). Importantly, it is only possible to statistically evaluate model accuracy and interactions if empirical data on individual chemical dose response parameters (i.e., slope and potency) are available (or reasonably estimated) for mixture components to generate mixture model estimates for comparison to empirical mixture data. Broadly, there is considerable support for the use of DA approaches as the default for estimating cumulative effects from combined exposure to multiple similar-acting compounds (ATSDR 2018; EFSA 2019; Kortenkamp 2022; Kortenkamp and Faust 2018; NRC 2008; USEPA 2000). Specifically, the relative potency factor (RPF) approach is a well-established method based on dose addition that has recently been employed in PFAS risk assessment (Bil et al. 2021; Bil et al. 2022; RIVM 2018; USEPA 2021). The RPF approach assumes that each chemical acts as a dilution of every other chemical in the mixture and the mixture response can be estimated by scaling the component chemical doses based on the relative potency to a designated index chemical (USEPA 2000).

Previously published in vivo studies of rodents exposed to perfluorooctanoic acid (PFOA) and perfluorooctane sulfonic acid (PFOS) have reported a range of effects in maternal and F1 animals from maternal oral exposure during pregnancy and/or lactation (Abbott et al. 2009; Abbott et al. 2007; Blake et al. 2020; Butenhoff et al. 2004; Lau et al. 2003; Luebker et al. 2005a; Luebker et al. 2005b; Thibodeaux et al. 2003). These effects occur in the absence of overt maternal toxicity and include reduced maternal gestational weight gain, reduced pup bodyweights, reduced pup survival, increased maternal and pup liver weight, reduced maternal thyroid hormones, and altered serum lipids, among others. The modes of action for PFOA and PFOS maternal and pup effects are not fully characterized but several studies demonstrate that they involve multiple molecular initiating events (Abbott et al. 2009; Abbott et al. 2007; Bjork et al. 2011; Rosen et al. 2017; Rosen et al. 2007) and target multiple tissues with lifestage-specific effects. Alteration of lipid and carbohydrate metabolic homeostasis across the maternal-placental-fetal unit during pregnancy appears to contribute to the toxic effects in maternal and F1 animals for PFOA, PFOS, and other PFAS (Blake et al. 2020; Blake and Fenton 2020; Conley et al. 2021b; Conley et al. 2022; Cope et al. 2021). Animal data for metabolic alterations during pregnancy support similar observations in human epidemiological studies of pregnancy/birth cohorts (Birru et al. 2021; Blomberg et al. 2021; Chang et al. 2022). These data indicate that exposure to PFOA and PFOS individually result in similar effects, however data are lacking whether these effects are additive or independent when exposure occurs as a mixture.

Here we hypothesized that a mixture of PFOA and PFOS would produce cumulative effects of combined exposure and that the combined effects would be more accurately predicted using DA-based models than a RA model. Individual dose response experiments were conducted with PFOA and PFOS and data were analyzed to detect significant effects from exposure to PFOA and/or PFOS, model individual dose response data to generate individual chemical dose response parameters (e.g., slope and potency) necessary for mixture model analyses, and determine if dose response curve slope parameters were congruent between PFOA and PFOS. Then, individual chemical study data were used to design a variable-ratio binary mixture study where PFOS was held constant across increasing doses of PFOA. This study design was intended to 1) shift the PFOA dose–response curves towards greater effects at lower doses due to combined exposure with PFOS (for example of dose response shift see Blystone et al. (2009)), 2) allow for comparison of observed mixture effects with mixture model predictions based on DA and RA (similar to prior mixture studies from our group (Conley et al. 2018; Conley et al. 2021a; Rider et al. 2008; Rider et al. 2010)), 3) conduct RPF analysis for a range of endpoints (similar to Van Der Ven et al. (2022)), and 4) avoid overt maternal toxicity.

2. Methods

2.1. Chemicals and dosing solutions

PFOA (pentadecafluorooctanoic acid ammonium salt; CAS: 3825–26-1; Product #: 77262; Lot: BCBW7054; Purity ≥ 98 %) and PFOS (heptadecafluorooctanesulfonic acid potassium salt; CAS: 2795–39-3, Product #: 77282, Lot: BCBX5798, Purity ≥ 98 %) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Dosing vehicle was 0.5 % (by weight) Tween-80 (polyethylene glycol sorbitan monooleate; CAS: 9005–65-6; Product #: P1754; Lot: BCCB5237; Sigma) in high performance liquid chromatography-grade water purchased from Honeywell Research Chemicals (Muskegon, MI, USA). Oral doses reported in mg/kg/d represent nominal exposures to the reported standards (i.e., PFOA ammonium salt and PFOS potassium salt) and were not corrected for counterion content, which account for 4 % and 7 % of the molecular weights for PFOA ammonium salt and PFOS potassium salt, respectively. Dosing solutions were prepared fresh every 5–6 days and aliquots of each solution were analytically verified for PFOA and PFOS concentration. Average PFOA accuracy ranged from 89 to 104 % of nominal, while PFOS accuracy ranged from 71 to 122 % of nominal (Table S1). All vehicle control solutions had no detections of PFOA or PFOS above analytical detection limits. Dosing was administered once daily via oral gavage at 2.5 mL/kg-body weight between 07:00 – 09:00 EST.

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 GD2 (GD0 = bred date; GD1 = plug positive date). Dams and their offspring were housed individually in clear polycarbonate cages (20 × 25 × 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. 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). Dams were weight-ranked, stratified into weight blocks based on the number of dose groups, and dams from each weight block were randomly assigned to treatment groups using a random sequence generator (https://www.random.org) to produce similar mean weights and variances (coefficient of variation range 2.5–10.3 %). 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.

2.3. Study design

For the PFOA study, the dose range tested was 0 (vehicle), 10, 30, 62.5, 125, and 250 mg/kg/d, while PFOS was tested at 0 (vehicle), 0.1, 0.3, 1, 2, and 5 mg/kg/d. The dose ranges for PFOA and PFOS were selected based on evaluation of limited data from previously published developmental toxicity studies using Sprague-Dawley rats (Butenhoff et al. 2004; Lau et al. 2003; Thibodeaux et al. 2003). These doses were intended to avoid maternal toxicity and allow for dose response modeling of a broad range of endpoints. Given the lack of rat developmental dose response data for PFOA, and prior preliminary experiments by our lab, we estimated the oral potency for PFOA would be roughly similar to our previously published studies on HFPO-DA (Conley et al. 2021b; Conley et al. 2019); however, as reported here, PFOA was more acutely toxic to the pregnant SD rat than HFPO-DA. Based on the effects from the individual studies, a variable-ratio mixture experiment was designed with a vehicle control and PFOA at 3, 10, 30, 40, 62.5, and 80 mg/kg/d with each dose of PFOA also containing 2 mg/kg/d of PFOS (Table 1). One goal of this design was to shift the PFOA dose response curves for multiple endpoints towards greater effects at lower oral doses in the mixture exposure compared to the PFOA-only exposure. A second goal of this design was to allow for mixture modeling and comparison of DA and RA model predictions to the observed mixture data. DA and RA modeling was conducted using the model equations by Rider and LeBlanc (2005), as well as the RPF approach reported by Van Der Ven et al. (2022).

Table 1.

Experimental design of individual chemical exposures and mixture.

PFOA exposure
PFOA (mg/kg/d)	0	10	30	62.5	125	250a	-
Dams/litters (#)	5	5	5	5	5b	5	–
PFOS exposure
PFOS (mg/kg/d)	0	0.1	0.3	1	2	5	–
Dams/litters (#)	5	5	5	5	5	5	–
PFOA + PFOS exposure
PFOA (mg/kg/d)	0	3	10	30	40	62.5	80
PFOS (mg/kg/d)	0	2	2	2	2	2	2
Dams/litters (#)	5	4	4	5	4	4	4

^aExposure level was overtly toxic and dosing was terminated.

^b2/5 dams lost weight at beginning of exposure and dosing was terminated, resulting in n = 3.

Thirty pregnant rats each were used to assess the maternal and neonatal effects of maternal exposure to PFOA-only, PFOS-only, and the PFOA + PFOS mixture during pregnancy (N = 90 across 3 experiments). For the PFOA-only and PFOS-only studies, sample size was n = 5 dams per dose across vehicle controls and 5 dose levels. In the mixture study with vehicle controls and 6 PFOA dosing levels sample sizes were n = 4 dams per dose except vehicle control and 30 mg/kg PFOA where n = 5. Dams were weighed daily and dosed via oral gavage from GD8 – PND2 (PND0 = day of parturition), which included fetal development from post-implantation through early lactation because prior studies reported the majority of neonatal mortality occurred by PND2 (Lau et al. 2003) (Fig. 1). This dosing interval was consistent with our prior PFAS developmental toxicity studies and intended to cover the period of rat fetal development without potentially affecting uterine implantation of embryos (Conley et al. 2021b; Conley et al. 2022; Conley et al. 2019). 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 on GD23 (PND1). Dead pups were removed, and moribund pups were removed and euthanized via decapitation. As dams delivered, the time the first pup was observed in a cage was recorded and then the time of completion of delivery was recorded as determined by the onset 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 randomly selected pups, which were euthanized via decapitation and pooled trunk blood was collected. One pup had the thoracic and abdominal cavities exposed and the whole pup was fixed in 10 % formalin. Subsequently, livers from formalin fixed pup carcasses were removed and shipped to Experimental Pathology Laboratories, Inc. (Durham, NC) where they were embedded and sectioned. Sections from each pup were stained with hematoxylin and eosin (H&E), periodic acid-Schiff (PAS), and Oil Red O and evaluated for histopathological lesions by a Diplomate of the American College of Veterinary Pathology. Liver tissue was removed from the second pup for RNA extraction and gene expression (data not presented here) and a second piece of liver was weighed and flash frozen in liquid nitrogen for glycogen assay (described below).

Fig. 1.

Schematic diagram of study design for the individual PFOA and PFOS experiments and the PFOA + PFOS mixture experiment (see dose levels and samples sizes in Table 1). Time-mated Sprague-Dawley rats were dosed via oral gavage from gestation day 8 (GD8) to postnatal day 2 (PND2) with endpoints collected at birth (PND0) and at necropsy on PND2.

On the afternoon (15:00–16:00 EST) of PND1 all litters were counted and weighed. On PND2 all pups were sexed by visual examination of anogenital distance, weighed individually, and euthanized via decapitation. Pooled trunk blood was collected from all pups in a litter for total serum T3, T4, and clinical chemistry analyses. Liver weight was recorded for one male and one female pup per litter, where available. For the mixture experiment, all PND2 pup livers were formalin fixed and processed and evaluated for histopathology as described above.

Dams were euthanized via decapitation on PND2, and trunk blood was collected for serum thyroid hormone analyses, clinical chemistry, and on-going metabolomic analyses. Maternal liver and kidney weights were recorded. A liver sample was excised and formalin fixed and both kidneys were sectioned longitudinally, and formalin fixed for histopathological evaluation. A sample of maternal liver was collected for RNA extraction and gene expression (data not presented here). A second piece of maternal liver was collected and weighed for analytical chemical determination of test chemical. The uterus was removed, and the number of uterine implants and resorptions were counted.

2.4. Clinical chemistry

Clinical chemistry parameters were analyzed in maternal (PND2) and pup (PND0 and PND2) sera using a Rx daytona+ (Randox Laboratories, Kearneysville, WV) according to manufacturer specifications. Clinical chemistry parameters included: alanine aminotransferase (ALT), aspartate aminotransferase (AST), triglycerides, cholesterol, albumin, glucose (non-fasted), total protein, blood urea nitrogen (BUN), creatinine, total bilirubin, urea, globulin, total bile acids, and glutamate dehydrogenase (GLDH). Common parameter ratios were calculated including BUN:Creatinine, albumin:globulin, and AST:ALT.

2.5. Free and total thyroid hormones

PND2 maternal and pup sera were analyzed for total thyroid hormone concentrations (triiodothyronine (T3), reverse triiodothyronine (rT3), and thyroxine (T4)) and PND2 maternal sera were separately analyzed for free T3 and free T4. Total thyroid hormone analyses were conducted as previously described (Hornung et al. 2015; O’Shaughnessy et al. 2018) with minor modifications. Briefly, aliquots of maternal sera (20 μL) and pup sera (40 μL) received 5 μL of a 100 ng/mL solution of isotopically labelled internal standards (¹³C₆-T3, ¹³C₆-rT3, and ¹³C₆-T4; Cerilliant, Round Rock, TX, USA). Samples were extracted using Evolute CX well plates (96-well, 10 mg, 1 mL; Biotage, NC, USA) and analyzed using an AB Sciex 6500 + QTRAP Linear Ion Trap mass spectrometer with a Restek Raptor Biphenyl column (2.1 mm × 100 mm; 2.6 μm). Mobile phase components were 0.1 % acetic acid in water and methanol. The LOQ for total T3, rT3, and T4 was 5 pg/mL each.

For PND2 maternal sera, ultrafiltration was used to separate the bound thyroid hormones from the free fraction in serum. Free thyroid hormone analyses were conducted as previously described (Tanoue et al. 2018) with minor modification. Specifically, 500 μL of serum was added to an Amicon Ultra-0.5 mL centrifugal filter unit (Ultracel-10; Millapore Sigma, Burlington, MA, USA), having a molecular weight cut-off (MWCO) of 10 kDa, and the sample was processed at 7500 g for 60 min at 30 °C. A 100 μL aliquot of ultrafiltrate received 25 μL of a 40 pg/mL internal standard mix (T3-¹³C₆, rT3-¹³C₆, and T4-¹³C₆) and the 125 μL sample was then directly injected onto the LC-MS/MS system. Ultrafiltrates were analyzed on an AB Sciex 6500 + QTRAP Linear Ion Trap mass spectrometer using electrospray ionization in positive ion, multiple reaction monitoring mode. Chromatographic separation was performed on a Restek Raptor Biphenyl column (4.6 mm × 50 mm, 2.6 μm). Mobile phase components were 0.1 % acetic acid in water and methanol. The LOQ for free T3, rT3, and T4 was 0.5 pg/mL each.

2.6. Pup liver glycogen

Newborn pup liver subsamples were flash frozen in liquid nitrogen and stored at −80 °C until assay. Glycogen was quantified using Glycogen Assay Kit II (Colorimetric; Prod no. LS-K151-100) by LifeSpan BioSciences, Inc. (Seattle, WA, USA) as per manufacturer protocol with additional 1:10 or 1:15 dilutions for sample values to be within calibration curve limits.

2.7. Analytical chemical determination of PFAS concentrations

Maternal serum and liver samples and dosing solutions were analyzed for PFOA and PFOS concentrations using similar methodology as previously reported (Conley et al. 2019; McCord et al. 2018). Serum was isolated from trunk blood via centrifugation (10,000 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, wet weight basis). Tissues were then homogenized using a Kontes pestle homogenizer 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 × 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 × 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 PFOS (¹³C₄) and PFOS (¹³C₈) were purchased from Wellington Laboratories (Guelph, Ontario, Canada) and used as internal standards. Separate calibration curves were prepared for the ranges 10–10,000 ng/mL, and 5,000–500,000 ng/mL. Limits of quantitation (LOQ) for both PFOA and PFOS were 10 ng/mL (serum) and 40 ng/g (liver).

2.8. Data analyses

2.8.1. Individual study 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 PFOA, PFOS, and PFOA + PFOS mixture endpoints 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. Histopathological changes were analyzed with the Cochran-Armitage Trend Test using PROC FREQ in SAS.

Several endpoints were also analyzed by analysis of covariance (ANCOVA) in addition to analysis on absolute measurements. Dam bodyweight on PND2 was analyzed using initial (GD8) bodyweight as a covariate with dose and cumulative maternal gestational weight gain was analyzed with litter size as a covariate with dose. Pup birthweight was analyzed with total litter size and gestational age (birthtime) as covariates with dose. Liver weight (pup and maternal) and paired kidney weight (maternal) were analyzed with terminal body weight as a covariate with dose. As it is commonly reported, we also calculated relative liver and kidney weights (mg organ per g bodyweight).

A multivariate data analysis was used to provide parsimonious description of the effects on serum clinical chemistry variables. Maternal and pup clinical chemistry data from PND2 necropsy were analyzed using PROC Factor in SAS for PFOA and the PFOA + PFOS mixture. PROC Factor combines clusters of correlated data together into “latent” variables in the data set. The output from PROC Factor includes factor loadings for each clinical chemistry variable on each factor (with an eigen value > 1) and factor scores for each dam and PND2 pup. Three factor analyses were conducted including: 1) maternal PFOA and PFOA + PFOS, 2) pup PFOA and PFOA + PFOS and 3) maternal and pup PFOA + PFOS data. The factor scores were then analyzed using PROC GLM in SAS 9.4 for dose related effects and LSMEANS was used to generate dose group means and standard errors (SEs). Since each analysis indicated that a single factor displayed treatment related effects, the means and SEs of the factor scores were entered into GraphPad Prism to compare the dose response curves.

2.8.2. Evaluation of mixture effects

Evaluation of mixture effects was conducted in a series of data-driven step-wise decisions (Fig. 2). Endpoints that had no significant ANOVA effects for both PFOA and PFOS were not evaluated for mixture effects. Pup clinical chemistry parameters were similar but with greater effects on PND2 than PND0, therefore mixture evaluation of pup clinical chemistry focused on PND2 data. First, all endpoints with significant PFOA-only and/or PFOA + PFOS mixture effects were analyzed for dose response shift between the PFOA-only and PFOA + PFOS mixture exposure. 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, calculated as % increase from control) or bottom constrained to 0 (for decreasing effects, calculated as % of control) and the remaining three parameters unconstrained (i.e., slope, ED₅₀, and top or bottom). The vehicle (0 dose) control was not graphed with the mixture data because it does not have a linear relationship with the mixture (i.e., all mixture doses contained 2 mg/kg PFOS to shift the curve). We then determined if there was a significant difference between the slopes of the best fit regressions for the PFOA-only and PFOA + PFOS mixture dose response curves (“Extra sums-of-squares F test” in Prism), or if a single slope could statistically fit both dose responses. For all endpoints with statistically similar slopes (p > 0.05) the regression models were then constrained to the common slope and analyzed for a significant difference between the two remaining parameters (i.e., ED₅₀ and top or bottom) to detect a dose–response shift. For endpoints with dissimilar slopes (p < 0.05), models were analyzed for significant difference between the three unshared parameters.

Fig. 2.

Decision tree of data analyses to evaluate individual and mixture-based effects from combined exposure to PFOA + PFOS.

Next, all endpoints with significant effects for both PFOA and PFOS were analyzed using DA and RA model predictions, and/or RPF evaluation. Mixture model equations for DA and RA were those published by Rider and LeBlanc (2005) and Olmstead and Leblanc (2005) and were performed in Microsoft Excel similar to previous mixture studies we have conducted (Conley et al. 2018; Conley et al. 2021a; Howdeshell et al. 2007; Howdeshell et al. 2008; Rider et al. 2008; Rider et al. 2010). The DA model used here assumes similar slope factors between component chemicals in the mixture equation. Endpoints modeled here were determined to have congruent slope parameters (extra sums-of-squares F test, p > 0.05) for PFOA and PFOS and the global best fit slope from Prism was used in the model calculations. Similar to above, PFOA and PFOS data were fit with four parameter logistic regressions with top constrained to 100 and bottom constrained to 0. For some effects the 100 % or 0 % response could not clearly be defined and thus model constraints and estimation of ED₅₀ were not possible. Further, some endpoints were significantly affected based on ANOVA, but had low level of effect for both PFOA and PFOS and thus estimation of ED₅₀ was ambiguous (i.e., values with infinitely large confidence intervals). Only endpoints with unambiguous slope and ED₅₀ estimates, and congruent slope factors for PFOA and PFOS were used for DA and RA mixture modeling. Where available, we also incorporated data from the PFOS studies by Lau et al. (2003), which used the same species/strain, same route of administration, and a similar dosing interval, to provide more accurate estimates of slope and ED₅₀ parameters for use in mixture model calculations. The following endpoints were determined to have statistically equivalent slope and ED₅₀ parameters (p > 0.05, extra sums-of-squares F test in Prism) and the combined data sets were fit with a single logistic regression: PND2 pup survival (% live born), PND1 and PND2 pup body weight.

To determine whether DA or RA provided a better prediction of the observed mixture data, Prism was used to fit the observed data to a four parameter logistic regression constrained to the parameters for each mixture model and corrected Akaike’s Information Criterion (AICc) scores were calculated. The mixture model with the lowest AICc was determined to be the best fit of the observed data; however, ΔAICc values < 6 indicate that the models are not significantly different from one another in fit. AICc weighted means (Wm) were calculated as a probabilistic approach to determine the mixture model with the best fit compared to the observed best fit regression model (Burnham et al. 2010). Comparing models of additivity to observed mixture data also allowed for investigation of mixture interactions (i.e., greater-than-additive or less-than-additive, also referred to as synergy or antagonism). We calculated effective dose levels (ED_x) based on DA and RA models and compared those estimates to the 95 % confidence intervals for the observed data. Mixture model estimates outside the observed 95 % CI is one indication of greater-than or less-than additive mixture interactions. We also calculated index of prediction quality (IPQ), reported by Martin et al. (2021)) to quantify the percent deviation of model predictions from observed values. Further, we graphed the observed mixture dose response data with 95 % confidence intervals and 95 % prediction intervals. Prediction intervals, as described by Gennings (1995) more accurately reflect between-study variance in modeled data, as compared to 95 % confidence intervals. DA and RA regression models that depart from the 95 % confidence or prediction intervals could be interpreted as departures from additivity and potentially represent mixture interactions.

Finally, we evaluated endpoints using an RPF approach as reported by Van Der Ven et al. (2022). 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 PFOA and PFOS and estimated equivalent effective dose levels (e.g., ED₁₀, ED₂₀) based on the extent of the dose responses for PFOA and PFOS. For endpoints with statistically similar slopes (extra sums-of-squares F test, p > 0.05), the EDx values were used to calculate RPFs (PFOA = index chemical). The PFOS RPFs were then used to adjust the PFOS-only and PFOA + PFOS mixture doses into PFOA equivalents (i.e., if PFOS RPF = 20, then given a 2 mg/kg PFOS dose in the mixture, 40 mg/kg PFOA equivalent was added to the PFOA mixture dose level). The RPF adjusted PFOS-only and PFOA + PFOS mixture dose response data were then graphed with the PFOA-only data and Prism was used to determine if there was a significant difference between the RPF-adjusted PFOA equivalent dose responses across all three experiments. Assuming dose additivity of PFOS and PFOA effects, we hypothesized that the RPF-adjusted PFOA-equivalents could be fit by a single curve.

3. Results

3.1. PFOA – Maternal and F1 endpoints

Complete study data for all effect endpoints and dose levels are depicted in heatmaps (Figs. 3 and 4) and reported in Supplemental Tables S2–S9. Dams in the 250 mg/kg dose group lost significant weight (~40–50 g; Table S2) within the first 3–4 days of dosing and exposure at this level was terminated and animals were euthanized. In the 125 mg/kg dose group 2 of 5 dams lost significant weight within the first 4–5 days of dosing and exposure was terminated for these animals but continued in the remaining 3 of 5 dams (Table S2). At term (GD22) dam mean bodyweight and cumulative gestational weight gain (from GD8-GD22) were significantly reduced at ≥ 62.5 mg/kg (Table 2). Maternal bodyweight at necropsy (PND2) and bodyweight gain (GD8-PND2) were significantly reduced at 125 mg/kg (Table 2). The 62.5 mg/kg dose group had significantly smaller PND0 litter sizes, but there was no effect on litter size at birth in the other dose groups including 125 mg/kg (Table 2). Mean uterine implants were lower than control in all treatment groups and reached significance at 30 and 62.5 mg/kg (Table 2); however, dosing was initiated after implantation.

Fig. 3.

Heatmaps of maternal endpoints in individual PFOS (left) and PFOA (middle) experiments, and the PFOA + 2 mg/kg PFOS (right) mixture experiment. Endpoints were normalized to control and heatmap cells report mean percent of control values per dose level (measured values reported in Tables 2, 3, 4 and S2, S6, S8).

Fig. 4.

Heatmaps of neonatal pup endpoints in individual PFOS (left) and PFOA (middle) experiments, and the PFOA + 2 mg/kg PFOS (right) mixture experiment. Endpoints were normalized to control and heatmap cells report mean percent of control values per dose level (measured values reported in Tables 2, 3, 4 and S4, S7, S9). Clinical chemistry parameters represent PND2 measurements at necropsy (PND0 clinical chemistry reported in Tables S4, S7, S9).

Table 2.

Maternal and neonatal endpoints (mean ± standard error) from GD8-PND2 maternal oral exposure to PFOA.

	Control	10 mg/kg	30 mg/kg	62.5 mg/kg	125 mg/kg
Maternal endpoints	n = 4	n = 5	n = 5	n = 4	n = 3
GD8 bodyweight (g)	281.2 ± 10.0	264.0 ± 11.0	265.0 ± 7.0	270.8 ± 8.9	265.5 ± 10.5
GD22 bodyweight (g)	427.9 ± 11.3	407.1 ± 18.4	396.9 ± 12.0	360.9 ± 15.2 **	359.3 ± 28.7 **
GD8-GD22 bodyweight gain (g)	146.7 ± 3.2	143.1 ± 10.1	131.9 ± 5.8	96.5 ± 9.8 **	85.3 ± 14.3 **
PND2 bodyweight (g)	318.4 ± 12.4	309.7 ± 13.5	300.3 ± 10.6	287.6 ± 13.8	270.6 ± 21.8 **
GD8-PND2 bodyweight gain (g)	37.2 ± 6.1	45.7 ± 4.6	35.3 ± 5.0	23.2 ± 6.7	−3.4 ± 9.3 **
Uterine implants (#)	14.8 ± 0.9	13.2 ± 0.5	11.8 ± 0.7 *	10.0 ± 0.9 **	12.0 ± 1.5
Liver weight (g)	13.4 ± 0.4	12.5 ± 0.8	13.0 ± 0.7	13.6 ± 0.8	15.2 ± 1.3
Adjusted liver weight (g) a	12.5 ± 0.4	12.0 ± 0.4	13.0 ± 0.4	14.2 ± 0.4 *	16.6 ± 0.5 **
Paired kidney weight (g)	1.92 ± 0.06	1.89 ± 0.07	1.84 ± 0.10	1.86 ± 0.11	2.08 ± 0.09
Adjusted kidney weight (g) a	1.85 ± 0.08	1.85 ± 0.07	1.84 ± 0.06	1.90 ± 0.07	2.19 ± 0.09 *
Free T3 (pg/mL)	0.82 ± 0.11	0.47 ± 0.14 *	0.70 ± 0.14	0.25 ± 0.00 **	0.34 ± 0.09 *
Free T4 (pg/mL)	42.2 ± 7.7	25.9 ± 2.0 *	31.1 ± 6.7	14.9 ± 1.0 **	13.4 ± 1.4 **
Total T3 (ng/mL)	0.56 ± 0.04	0.44 ± 0.02 *	0.34 ± 0.02 **	0.37 ± 0.04 **	0.29 ± 0.05 **
Total T4 (ng/mL)	29.1 ± 1.0	16.2 ± 0.9 **	10.8 ± 0.7 **	9.6 ± 1.5 **	6.3 ± 0.8 **
Neonatal endpoints
Delivery time b	31.3 ± 4.5	16.8 ± 0.6	25.6 ± 4.3	23.1 ± 3.7	25.6 ± 6.5
Litter size (# pups)	13.8 ± 0.5	12.8 ± 0.4	11.8 ± 0.7	9.3 ± 1.0**	11.7 ± 1.9
Birthweight (g)	7.05 ± 0.32	6.20 ± 0.15 *	6.50 ± 0.26	5.82 ± 0.04 **	5.21 ± 0.23 **
Adjusted birthweight (g) c	6.86 ± 0.19	6.54 ± 0.15	6.43 ± 0.14	5.72 ± 0.19 **	5.14 ± 0.18 **
PND2 bodyweight (g)	8.22 ± 0.15	7.81 ± 0.25	7.79 ± 0.23	6.98 ± 0.20 *	5.02 ± 0.51 **
Survival (% live born)	100.0 ± 0.0	100.0 ± 0.0	95.3 ± 2.9	97.2 ± 2.8	55.1 ± 19.7 **
Survival (% implants)	92.7 ± 2.8	96.6 ± 2.1	95.3 ± 2.9	87.9 ± 7.2	51.9 ± 19.0 **
Liver weight (mg)	350 ± 20	350 ± 10	400 ± 9 *	410 ± 20 *	300 ± 10 *
Adjusted liver weight (mg) a	330 ± 20	340 ± 10	390 ± 10 **	420 ± 10 **	350 ± 30
Liver glycogen (ug/mg)	38.5 ± 3.3	23.5 ± 2.8 **	16.1 ± 2.3 **	15.1 ± 1.8 **	9.0 ± 1.5 **
Total T3 (pg/mL)	120 ± 10	90 ± 7 *	50 ± 4 **	60 ± 10 ** d	40 ± 7 ** e
Total reverseT3 (pg/mL)	70 ± 10	40 ± 4 **	40 ± 3 *	50 ± 20 d	120 ± 4 ** e
Total T4 (ng/mL)	5.87 ± 0.49	3.69 ± 0.35 **	1.79 ± 0.20 **	2.29 ± 0.61 ** d	2.51 ± 1.16 ** e

^aOrgan weight adjusted for body weight using ANCOVA

^bDelivery time = hrs after 12:00AM PND0

^cBirthweight adjusted for delivery time and littersize using ANCOVA

^dn = 3 and

^en = 2 litters due to limited serum volume; GD = gestation day; PND = postnatal day; T3 = triiodothyronine; T4 = thyroxine

^*p < 0.05

^**p < 0.01.

Absolute maternal liver and kidney weights were elevated but these effects were not significant, whereas ANCOVA bodyweight-adjusted liver weight and relative liver weight were significantly greater at ≥ 62.5 mg/kg and ANCOVA bodyweight-adjusted kidney weight and relative kidney weight were significantly elevated at 125 mg/kg (Table 2, S2). Dam livers at 125 mg/kg had hepatocyte hypertrophy, while 1/5 controls and 1/5 30 mg/kg had hepatocyte necrosis (Table S3). Dam liver Oil Red O staining indicated a slight trend in increasing lipid accumulation with dose. Dam kidneys in the 125 mg/kg dose group displayed several lesions not identified in controls (i.e., retrograde nephropathy, tubule mineralization, tubule dilation; severity grade = 1 and 2) (Table S3).

Maternal serum total T3 and T4 were significantly reduced in all dose levels, while free T3 and T4 were reduced in all dose levels compared to control and significant at 10, 62.5, and 125 mg/kg (Table 2, S2). Maternal serum glutamate dehydrogenase (GLDH) was elevated in all dose levels and significant at 10–62.5 mg/kg (Table S2). Maternal serum ALT was significantly elevated at 62.5 mg/kg and blood urea nitrogen (BUN), urea, and BUN:creatinine were significantly elevated at ≥ 62.5 mg/kg (Table S2). Maternal serum triglycerides, albumin, and total protein were significantly elevated and serum globulin was reduced at 125 mg/kg (Table S2). Maternal cholesterol was unaffected by PFOA-only exposure.

Pup birthweight was reduced in all dose groups and significant at 10, 62.5, and 125 mg/kg, however after adjusting with ANCOVA for birthtime and litter size, pup birthweight was only significantly reduced at ≥ 62.5 mg/kg (17 % and 26 % reductions at 62.5 and 125 mg/kg, respectively; Table 2, S4). Newborn pup liver glycogen concentration was significantly reduced in all dose groups (Table 2). Pup survival was significantly reduced at 125 mg/kg on both PND1 and PND2 based on number of live born pups and number of uterine implants (Table 2, S4). Similar to birthweight, pup bodyweights on PND1 and PND2 were significantly reduced at ≥ 62.5 mg/kg (Table 2, S4). PND2 unadjusted pup liver weight and ANCOVA bodyweight-adjusted liver weight displayed an apparent high dose non-monotonic dose response being significantly elevated at 30 and 62.5 mg/kg but lower or similar to control at 125 mg/kg, likely due to severe pup toxicity (Table 2). Relative liver weight was elevated in all dose groups but and significant at ≥ 30 mg/kg (Table S4). Livers of PND0 pups displayed a significant trend in decreasing hepatocyte vacuolization assumed to be associated with the reduced glycogen concentration (Table S5).

PND0 pup serum glucose was reduced, while triglycerides, BUN, and urea were higher than control at all dose levels but only significant at 125 mg/kg (Table S4). PND0 pup serum total cholesterol was significantly elevated and GLDH was reduced at all dose levels. PND0 pup serum albumin, total protein, and globulin were reduced and total bilirubin and BUN:creatinine were elevated at ≥ 62.5 mg/kg. PND2 pup serum was similar to PND0 with reduced glucose, albumin, and total protein and elevated cholesterol, triglycerides, BUN, BUN:creatinine. Further, PND2 pups had significantly elevated total bile acids and AST: ALT at 30 mg/kg (Table S4).

Among the clinical chemistry variables, triglyceride was the only measure to change dramatically with pup age, increasing by about 5-fold in the control pups from PND0-2. Several serum markers of kidney function in the pups were more severely affected at PND2 than PND0 including serum BUN, urea, creatinine, and BUN:creatinine, indicating a possible disruption of the neonatal ontogeny of nephrogenesis, which proceeds at a rapid pace between birth and 11–14 days of age (Cullen-McEwen et al. 2016).

Similar to dams, PND2 pup serum total T3 and T4 were significantly reduced in all dose groups (free T3 and T4 were not measured due to serum volume requirements) (Table 2, S4). Reverse T3 was significantly lower than control at 10 and 30 mg/kg, but elevated at 125 mg/kg (163 % of control).

3.2. PFOS – Maternal and F1 endpoints

Dams tolerated all dose levels but there was a significant reduction in dam bodyweight and cumulative gestational weight gain at 5 mg/kg (Table 3, S6). Gestational weight gain adjusted for litter size with ANCOVA was significantly reduced at ≥ 1 mg/kg. Maternal bodyweight and bodyweight gain on PND2 were lower in all dose groups and significant at 5 mg/kg (Table 3). ANCOVA bodyweight-adjusted liver weight was significantly increased at ≥ 0.3 mg/kg while relative liver weight was significantly increased at 5 mg/kg (Table 3, S6) Kidney weights and ANCOVA bodyweight-adjusted kidney weights were unaffected but relative kidney weights were significantly elevated at ≥ 2 mg/kg (Table 3, S6). Maternal livers on PND2 displayed hepatocyte necrosis (1/5 at 0.1 mg/kg, 2/5 at 5 mg/kg) and a slight trend of increased liver lipid content (Table S3). PND2 maternal kidneys presented a few lesions (tubule degeneration, medullary cyst, mononuclear cell infiltration), but these did not display a clear dose response (Table S3).

Table 3.

Maternal and neonatal endpoints (mean ± standard error) from GD8-PND2 maternal oral exposure to PFOS.

	Control	0.1 mg/kg	0.3 mg/kg	1 mg/kg	2 mg/kg	5 mg/kg
Maternal endpoints	n = 5	n = 4	n =5	n = 5	n = 5	n =5
GD8 bodyweight (g)	270.6 ± 7.3	261.3 ± 3.3	271.1 ± 12.4	261.2 ± 10.0	257.4 ± 8.9	270.1 ± 12.2
GD22 bodyweight (g)	422.1 ± 21.2	412.5 ± 12.0	423.5 ± 23.3	396.4 ± 18.6	393.8 ± 18.2	371.2 ± 21.2 **
GD8-GD22 bodyweight gain (g)	151.5 ± 17.0	151.2 ± 9.3	152.4 ± 11.5	135.1 ± 10.0	136.3 ± 9.5	101.1 ± 14.0 **
PND2 bodyweight (g)	326.4 ± 12.3	305.9 ± 11.8	312.4 ± 15.9	299.6 ± 17.4	296.8 ± 19.2	292.2 ± 14.1 **
GD8-PND2 bodyweight gain (g)	55.9 ± 6.8	44.7 ± 8.7	41.3 ± 5.4	38.3 ± 8.5	39.3 ± 10.9	22.1 ± 2.5 **
Uterine implants (#)	11.4 ± 1.5	12.5 ± 0.6	14.2 ± 1.1	12.4 ± 0.9	11.8 ± 0.2	9.2 ± 1.8
Liver weight (g)	13.0 ± 0.9	12.6 ± 0.8	13.0 ± 0.9	12.8 ± 0.9	12.4 ± 1.2	14.3 ± 1.0
Adjusted liver weight (g) a	11.8 ± 0.3	12.5 ± 0.3	12.6 ± 0.3	13.2 ± 0.3 **	12.9 ± 0.3 **	15.1 ± 0.3 **
Paired kidney weight (g)	1.89 ± 0.10	1.89 ± 0.03	1.91 ± 0.10	1.85 ± 0.12	1.90 ± 0.12	1.86 ± 0.04
Adjusted kidney weight (g) a	1.78 ± 0.05	1.89 ± 0.06	1.88 ± 0.05	1.88 ± 0.05	1.95 ± 0.05	1.93 ± 0.05
Free T3 (pg/mL)	0.62 ± 0.14	0.72 ± 0.16	0.61 ± 0.16	0.66 ± 0.11	0.66 ± 0.30	0.58 ± 0.09
Free T4 (pg/mL)	50.9 ± 1.5	48.1 ± 6.8	33.2 ± 4.6 *	40.7 ± 5.6	24.8 ± 5.4 **	23.3 ± 4.8 **
Total T3 (ng/mL)	0.64 ± 0.02	0.57 ± 0.02	0.52 ± 0.04 **	0.53 ± 0.02 **	0.42 ± 0.03 **	0.43 ± 0.02 **
Total T4 (ng/mL)	32.5 ± 2.5	25.8 ± 2.3 *	23.1 ± 2.5 **	18.7 ± 1.1 **	10.5 ± 0.3 **	10.5 ± 0.4 **
Neonatal endpoints
Delivery time b	17.1 ± 1.6	16.6 ± 1.2	19.1 ± 3.4	17.8 ± 2.5	15.6 ± 0.6	18.0 ± 1.2
Litter size (# pups)	10.4 ± 1.6	12.0 ± 0.4	12.8 ± 1.2	11.6 ± 0.7	11.4 ± 0.4	7.4 ± 1.9
Birthweight (g)	6.64 ± 0.12	6.48 ± 0.13	6.54 ± 0.12	6.41 ± 0.27	6.27 ± 0.26	6.37 ± 0.26
Adjusted birthweight (g) c	6.62 ± 0.15	6.60 ± 0.17	6.56 ± 0.16	6.43 ± 0.15	6.41 ± 0.16	6.10 ± 0.18 *
PND2 bodyweight (g)	8.34 ± 0.15	8.26 ± 0.19	8.09 ± 0.13	7.98 ± 0.33	7.23 ± 0.51 *	na ± na
Survival (% live born)	100.0 ± 0.0	97.5 ± 2.5	100.0 ± 0.0	96.0 ± 2.5	95.6 ± 4.4	0.0 ± 0.0 **
Survival (% implants)	87.0 ± 4.7	93.3 ± 4.1	88.1 ± 4.0	89.4 ± 5.2	91.8 ± 5.9	0.0 ± 0.0 **
Liver weight (mg)	370 ± 8	340 ± 10	340 ± 8	360 ± 10	340 ± 10 *	na ± na
Adjusted liver weight (mg) a	360 ± 10	330 ± 10 **	340 ± 10 *	360 ± 10	360 ± 10	na ± na
Pup liver glycogen (ug/mg)	39.6 ± 2.7	48.0 ± 5.0	35.8 ± 5.8	35.3 ± 2.8	33.5 ± 4.6	24.1 ± 3.7 **
Total T3 (pg/mL)	130 ± 8d	140 ± 8	120 ± 10	110 ± 7 *	80 ± 9 ** d	na ± na
Total reverseT3 (pg/ mL)	60 ± 3d	60 ± 6	50 ± 4	30 ± 4 **	30 ± 8 ** d	na ± na
Total T4 (ng/mL)	6.38 ± 0.28d	6.37 ± 0.53	4.92 ± 0.59 *	3.00 ± 0.36 **	2.37 ± 0.31 ** d	na ± na

^aOrgan weight adjusted for body weight using ANCOVA

^bDelivery time = hrs after 12:00AM PND0

^cBirthweight adjusted for delivery time and littersize using ANCOVA

^dn = 4 litters due to limited serum volume; na = no value due to pup mortality prior to PND2; GD = gestation day; PND = postnatal day; T3 = triiodothyronine; T4 = thyroxine

^*p < 0.05

^**p < 0.01.

PND2 maternal serum free T3 was unaffected, while free T4 was reduced in all doses and significant at 0.3, 2, and 5 mg/kg (Table 3, S6). Total T4 was significantly reduced at all dose levels and total T3 was significantly reduced at ≥ 0.3 mg/kg. PND2 maternal clinical chemistry identified reduced glucose (≥0.3 mg/kg), reduced cholesterol (≥1 mg/kg), elevated triglycerides (5 mg/kg), reduced AST (2 mg/kg), reduced creatinine (≥1 mg/kg), reduced globulin (≥1 mg/kg), and elevated albumin:globulin (≥1 mg/kg) (Table S6).

Pup birthweight was lower at all doses but only significant at 5 mg/kg when adjusted for litter size and birthtime using ANCOVA; whereas, litter size at birth was not affected by PFOS (Table 3, S7). Complete litter loss occurred at 5 mg/kg by PND1 but survival was unaffected at other doses (Table S7). Pup bodyweight on PND1 and PND2 was reduced at 2 mg/kg (Table S7). PND2 absolute liver weight was lower in all dose groups and significant at 2 mg/kg, while ANCOVA bodyweight-adjusted liver weight was significantly reduced at 0.1 and 0.3 mg/kg and relative liver weight was significantly reduced at 0.1 mg/kg (Table 3, S7). PND0 pup liver glycogen was significantly reduced at 5 mg/kg (Table 3, S7). PND0 liver histopathology displayed a trend in increasing lipid accumulation with dose (Table S5). PND2 pup serum had significantly reduced total T3 (≥1 mg/kg), total rT3 (≥1 mg/kg), and total T4 (≥0.3 mg/kg) (Table S7). PND0 and PND2 pup serum clinical chemistry had few significant findings and included elevated cholesterol and glucose on PND0 and elevated total bile acids on PND2 (Table S7).

3.3. PFOA + PFOS mixture – Maternal and F1 endpoints

Dams were exposed to a mixture containing increasing dose levels of PFOA (3–80 mg/kg) each combined with a fixed dose of PFOS (2 mg/kg). Dams tolerated all dose levels and there was no significant effect of treatment on dam absolute bodyweight or cumulative gestational weight gain, albeit slightly reduced (Table 4, S8). Mean dam bodyweight on PND2 and bodyweight change from GD8-PND2 were significantly reduced at 80 mg/kg (Table 4). Maternal absolute liver weight, relative liver weight, and ANCOVA bodyweight-adjusted liver weight were higher in all dose groups and significant at ≥ 62.5 mg/kg, ≥40 mg/kg, and ≥ 30 mg/kg, respectively (Table 4, S8). Absolute, relative, and ANCOVA bodyweight-adjusted kidney weights were higher in all doses and significant at 80 mg/kg, 62.5 mg/kg, and ≥ 40 mg/kg, respectively (Table 4, S8). Maternal livers on PND2 displayed hepatocyte necrosis (2/4 at 40 mg/kg and 1/4 at 80 mg/kg), hepatocyte hypertrophy (2/4 at 62.5 mg/kg and 4/4 at 80 mg/kg) and a trend in decreasing glycogen but no effect on liver lipid accumulation (Table S3). Maternal kidneys displayed a range of lesions not identified in controls primarily at ≥ 30 mg/kg including tubule mineralization, medullary cyst, retrograde nephropathy, and cortex fibrosis (Table S3).

Table 4.

Maternal and neonatal endpoints (mean ± standard error) from GD8-PND2 maternal oral exposure to mixture of PFOA + PFOS.

	Vehicle Control	3 mg/kg PFOA + 2 mg/kg PFOS	10 mg/kg PFOA + 2 mg/kg PFOS	30 mg/kg PFOA + 2 mg/kg PFOS	40 mg/kg PFOA + 2 mg/kg PFOS	62.5 mg/kg PFOA + 2 mg/kg PFOS	80 mg/kg PFOA + 2 m^kg PFOS
Maternal endpoints	n = 5	n = 4	n = 4	n = 5	n = 4	n = 4	n = 4
GD8 bodyweight (g)	273.6 ± 10.5	265.9 ± 10.3	273.7 ± 11.6	265.6 ± 6.6	270.5 ± 12.3	268.4 ± 6.5	271.8 ± 8.7
GD22 bodyweight (g)	414.1 ± 14.2	408.5 ± 13.4	429.8 ± 18.4	401.1 ± 10.0	393.1 ± 31.0	388.5 ± 15.7	394.2 ± 11.8
GD8-GD22 bodyweight gain (g)	140.6 ± 4.3	142.6 ± 13.9	156.1 ± 10.1	135.5 ± 6.0	122.6 ± 21.3	120.1 ± 11.0	122.4 ± 11.8
PND2 bodyweight (g)	311.9 ± 10.4	308.7 ± 6.7	324.6 ± 15.6	296.4 ± 7.2	295.0 ± 14.5	287.9 ± 13.8	278.3 ± 6.7 *
GD8-PND2 bodyweight gain (g)	38.4 ± 0.7	42.8 ± 5.8	50.8 ± 7.5	30.8 ± 6.4	24.5 ± 6.3	19.5 ± 11.8	6.5 ± 7.7 *
Uterine implants (#)	12.8 ± 0.4	12.0 ± 1.4	12.8 ± 0.5	12.4 ± 0.5	10.5 ± 2.6	13.8 ± 0.8	14.0 ± 1.1
Liver weight (g)	12.9 ± 0.7	13.2 ± 0.3	14.4 ± 0.8	13.3 ± 0.4	13.7 ± 1.2	15.5 ± 1.0 *	15.3 ± 0.9 *
Adjusted liver weight (g) a	12.2 ± 0.4	12.8 ± 0.5	13.0 ± 0.5	13.5 ± 0.4 *	14.0 ± 0.4 **	16.3 ± 0.5 **	16.6 ± 0.5 **
Paired kidney weight (g)	1.82 ± 0.04	1.90 ± 0.08	2.00 ± 0.11	1.90 ± 0.06	1.96 ± 0.07	2.06 ± 0.11	2.27 ± 0.24 *
Adjusted kidney weight (g) a	1.75 ± 0.08	1.85 ± 0.09	1.84 ± 0.10	1.93 ± 0.08	2.00 ± 0.09 *	2.14 ± 0.09 **	2.42 ± 0.10 **
Free T3 (pg/mL)	0.65 ± 0.21	0.25 ± 0d	0.25 ± 0d	0.25 ± 0d	0.65 ± 0.15	0.25 ± 0d	0.44 ± 0.19
Free T4 (pg/mL)	47.0 ± 4.1	21.2 ± 2.0 **	12.8 ± 3.5 **	19.9 ± 4.9 **	19.9 ± 4.2 **	14.9 ± 1.4 **	13.9 ± 4.5 **
Total T3 (ng/mL)	0.52 ± 0.05	0.38 ± 0.04 *	0.27 ± 0.04 **	0.33 ± 0.02 **	0.39 ± 0.04 *	0.38 ± 0.04 *	0.36 ± 0.01 **
Total T4 (ng/mL)	24.7 ± 1.3	9.3 ± 1.3 **	6.5 ± 0.9 **	8.4 ± 0.9 **	8.4 ± 0.5 **	7.7 ± 0.8 **	6.9 ± 0.4 **
Neonatal endpoints
Delivery time b	27.2 ± 4.1	14.7 ± 0.2 **	16.6 ± 1.1 *	15.6 ± 1.0 **	21.5 ± 6.1	15.6 ± 1.4 **	17.3 ± 1.1 *
Litter size (# pups)	12.2 ± 0.6	11.3 ± 1.8	12.8 ± 0.3	12.4 ± 0.5	10.5 ± 2.6	12.0 ± 1.1	13.3 ± 1.0
Birthweight (g)	7.00 ± 0.16	6.31 ± 0.31	6.17 ± 0.17 *	6.10 ± 0.17 *	6.14 ± 0.39 **	5.43 ± 0.14 **	5.36 ± 0.13 **
Adjusted birthweight (g) c	6.81 ± 0.18	6.33 ± 0.17	6.26 ± 0.17 *	6.21 ± 0.15 *	5.93 ± 0.17 **	5.50 ± 0.17 **	5.51 ± 0.17 **
PND2 bodyweight (g)	8.44 ± 0.25	7.33 ± 0.41 * e	7.11 ± 0.30 **	7.15 ± 0.26 **	6.57 ± 0.18 **	5.26 ± 0.29 ** e	5.70 ± 0.17 ** f
Survival (% live born)	100.0 ± 0.0	65.7 ± 22.0 **	95.5 ± 4.6	89.9 ± 4.8	95.2 ± 2.9	39.9 ± 18.4 **	27.3 ± 24.3 **
Survival (% implants)	94.4 ± 3.7	63.8 ± 21.7 *	95.4 ± 2.7	89.9 ± 4.8	95.1 ± 2.9	34.4 ± 15.0 **	24.8 ± 22.4 **
Liver weight (mg)	330 ± 8	350 ± 10e	360 ± 10	400 ± 20 **	370 ± 20	320 ± 20e	330 ± 1f
Adjusted liver weight (mg) a	260 ± 20	330 ± 10 ** e	360 ± 10 *	390 ± 10 **	400 ± 10 **	400 ± 20 ** e	420 ± 20 ** f
Liver glycogen (ug/mg)	30.7 ± 3.6	27.5 ± 3.1	20.7 ± 3.8 **	14.7 ± 1.8 **	12.3 ± 2.3 **	13.5 ± 2.2 **	8.7 ± 2.5 **
Total T3 (pg/mL)	110 ± 7	80 ± 3 ** e	70 ± 3 **	50 ± 2 **	40 ± 4 ** h	30 ± . ** g	50 ± . ** g
Total reverseT3 (pg/mL)	70 ± 6	30 ± 6 ** e	30 ± 6 **	40 ± 5 **	50 ± 10 * h	70 ± .g	60 ± .g
Total T4 (ng/mL)	6.42 ± 0.40	2.08 ± 0.11 ** e	2.01 ± 0.16 **	2.11 ± 0.20 **	1.87 ± 0.45 ** h	1.32 ± . ** g	2.28 ± . ** g

^aOrgan weight adjusted for body weight using ANCOVA

^bDelivery time = hrs after 12:00AM PND0

^cBirthweight adjusted for delivery time and littersize using ANCOVA

^dAll values below LOD of 0.5 pg/mL and assigned LOD/2

^en = 3 and

^fn = 2 due to pup mortality

^gn = 1 and

^hn = 3 due to limited serum volume; GD = gestation day; PND = postnatal day; T3 = triiodothyronine; T4 = thyroxine

^*p < 0.05

^**p < 0.01.

Maternal serum on PND2 had significantly reduced total T4, total T3 and free T4 at all dose levels (Table 4, S8). Free T3 was below limit of detection in all but 3 treated animals but was detectable in 3/5 controls. Serum clinical chemistry indicated slightly elevated high dose non-monotonic effects on glucose (significant at 30 and 40 mg/kg), albumin (significant at 30–62.5 mg/kg), and total protein (significant at 40 and 62.5 mg/kg) and dose related increases in BUN (significant ≥ 40 mg/kg), BUN:creatinine (significant at 3 and ≥ 62.5 mg/kg), and albumin:globulin (significant ≥ 30 mg/kg) (Table S8).

PFAS treatment accelerated the onset of parturition in all doses (significant at all doses except 40 mg/kg) (Table 4, S9). Pup birthweight and birthweight adjusted for birthtime and litter size by ANCOVA was reduced in all dose levels (significant ≥ 10 mg/kg for both) (Table 4, S9). Pup liver glycogen at birth was lower in all dose groups and significant at ≥ 10 mg/kg (Table 4, S9). Pup bodyweights on PND1 and PND2 were significantly lower than control in all dose levels (Table 4, S9). Pup bodyweight gain from PND1-2 was significantly reduced at ≥ 40 mg/kg (Table S9). Pup survival as both percent of live born pups and percent of uterine implants was significantly reduced at 3, 62.5, and 80 mg/kg (Table 4, S9). Pup absolute liver weight displayed a high dose non-monotonic response with significantly higher weights only at 30 mg/kg (Table 4, Figure S1), in contrast relative and ANCOVA bodyweight-adjusted liver weight were significantly elevated at all dose levels (Table 4, S9). PND0 pup liver histopathology indicated reduced hepatocyte vacuolization assumed to be associated with the reduced glycogen concentration and elevated lipid accumulation in the middle but not higher doses (Table S5). PND2 pup liver histopathology identified significantly increased trends in hepatocyte necrosis (p = 0.02, Fig. 5) and increased peribiliary polymorphonuclear cells (p < 0.01), and a significant (p < 0.01) trend in reduced liver lipid accumulation (decrease grade 2 incidence) (Table S5).

Fig. 5.

Photomicrographs of PND2 pup liver stained with hematoxylin and eosin. Control liver at 40x (A) and 100x (B) magnification compared to PND2 pup liver exposed to 30 mg/kg PFOA + 2 mg/kg PFOS displaying mild hepatocyte necrosis (grade 2, see Table S5) at 40x (C) and 100x (D) magnification.

PND2 pup serum displayed significantly reduced total T3 and T4 at all dose levels, and significantly reduced total rT3 at 3–40 mg/kg but not 62.5 or 80 mg/kg (only n = 1 at top two doses due to inadequate serum volume for thyroid hormone analyses because of complete or extensive pup mortality) (Table 4, S9). PND0 and PND2 serum clinical chemistry both displayed reduced glucose and creatinine, elevated cholesterol, BUN, BUN:creatinine, and total bilirubin (Table S9). Similar to the PFOA results, the kidney serum biomarkers were more affected at PND2 than PND0, again indicating a significant alteration in neonatal nephrogenesis. PND0 pups had elevated triglycerides but not at PND2 (Table S9). PND0 pups had reduced globulin and GLDH and PND2 pups were elevated. PND0 pups had elevated albumin:globulin which was reduced on PND2, while PND2 pups had elevated AST and AST:ALT that was not affected at PND0 (Table S9).

3.4. Clinical chemistry Factor analysis

Factor analysis of the PND2 pup clinical chemistry data indicated that Factor 1 was loaded with decreased serum glucose and increased serum cholesterol, AST, BUN, bilirubin, and total bile acids. This Factor displayed dose-related alterations (p < 0.0001) and displayed a shift in the dose response curve when PFOS was combined with PFOA (p = 0.012) (Fig. 6A). PFOA Factor 1 scores were significantly increased only at 125 mg/kg/d (p < 0.02), whereas scores were significantly increased at ≥ 40 mg/kg/d for the PFOA + PFOS mixture (p < 0.0003).

Fig. 6.

Linear regressions of Factor analysis scores for serum clinical chemistry parameters. Factor analysis combines clusters of correlated data together into “latent” variables in the data set. Analysis of the PND2 pup clinical chemistry data indicated that Factor 1 (loaded variables: glucose, cholesterol, AST, BUN, bilirubin, and total bile acids) displayed treatment-related alterations (p < 0.0001) and displayed a shift in the dose response curve when PFOS was combined with PFOA (panel A). Comparison of the PFOA + PFOS mixture pup and maternal clinical chemistry data on PND2 also produced a single Factor (loaded variables: glucose, cholesterol, AST, BUN, globulin, total bile acids and GLDH) that indicated the pup clinical chemistry was significantly more disrupted than the dam (Panel B). Data points represent mean ± standard error.

Factor analysis of the maternal clinical chemistry data on PND2 for PFOA and PFOA + PFOS indicated that there were minimal effects in the dam with only a single Factor loaded with decreased serum albumin, total protein, and globulin and increased serum BUN. This Factor displayed small, dose related changes (p < 0.03). Maternal Factor 1 scores increased at 80 mg/kg/d (p < 0.02) for the mixture, but mixture results did not differ from the PFOA only exposure. The clinical chemistry variables loaded on this Factor included decreased serum albumin, total protein, and globulin and increased serum BUN.

Comparison of the PND2 pup and maternal clinical chemistry data from the PFOA + PFOS mixture also produced a single Factor that was loaded with decreased serum glucose, and increased serum cholesterol AST, BUN, globulin, total bile acids and GLDH. Analysis of this Factor indicated the pup clinical chemistry was significantly more disrupted than the dam. Pup Factor 1 scores significantly increased at ≥ 40 mg/kg/d (p < 0.0001), whereas the Factor scores were not affected at any dose in the dams (Fig. 6B). Taken together, Factor analyses of the clinical chemistry data indicate that combined exposure with PFOS significantly shifted the PFOA dose response curve and that the F1 pups were more severely affected than the dams.

3.5. Maternal serum and liver PFAS concentrations

In the PFOA-only study there was no difference in the maternal serum and liver concentrations on a parts per billion (ppb) basis within a given oral dose level (range ~ 20–200 ppb, Fig. 7A, Table S10). In contrast, in the PFOS-only study maternal liver concentrations (range ~ 10–400 ppb) were significantly greater than serum concentrations (range ~ 2–200 ppb, Fig. 7B, Table S10) within a given dose level, however the fold difference decreased with increasing oral dose. Maternal serum and liver concentrations of PFOA increased log-linearly as a function of oral dose with no significant difference in the average concentrations at common doses between the PFOA-only and PFOA + PFOS mixture studies and no significant difference in the linear regression function between studies (Fig. 7E,F, Table S10). These results indicate that co-exposure of PFOA with 2 mg/kg PFOS did not alter the toxicokinetics of PFOA in the mixture study. Similarly maternal serum and liver concentrations of PFOS increased log-linearly as a function of oral dose in the individual chemical study (Fig. 7E,F, Table S10). In the mixture study PFOS was held constant at 2 mg/kg and there was no significant difference in the serum and liver concentrations of PFOS across increasing doses of PFOA, indicating co-exposure with PFOA did not alter PFOS toxicokinetics (Fig. 7C,D, Table S10). When comparing the PFOA and PFOS individual studies, despite a ~ 2 order of magnitude difference in oral dosing range, the more rapid elimination of PFOA compared to PFOS for the female rat (Dzierlenga et al. 2020; Huang et al. 2019) resulted in similar serum and liver exposure ranges. Similar to prior PFAS rodent studies, control animals had some detections of PFOA and/or PFOS in serum and/or liver slightly above LOD and were ~ 2 orders of magnitude lower than concentrations detected in the lowest dose groups (Table S10).

Fig. 7.

Maternal serum and liver concentrations of PFOA and PFOS at necropsy on PND2 following GD8-PND2 oral exposure. Concentrations reported for individual PFOA (panel A) and PFOS (panel B) experiments and from the PFOA + PFOS mixture experiment (panels C, D). Log-log linear regressions (panels E, F) of both PFOA and PFOS concentrations in individual and mixture experiments indicate no effect of combined exposure on chemical disposition of PFOA or PFOS. Mean concentrations, including controls reported in Table S10. Values represent mean ± standard error.

3.6. Mixture evaluation

3.6.1. Dose response shift analyses

The first mixture evaluation approach was to determine if combined exposure to PFOA + PFOS significantly shifted the dose response curve compared to PFOA-only exposure (Fig. 8, S2, S3, S4). We hypothesized that the PFOA + PFOS dose response curve would display a parallel shift compared to PFOA-only for those variables that PFOS also affected due to the contribution of a fixed 2 mg/kg PFOS dose. If the dose response curve for the mixture was significantly shifted toward a similar response for a shared endpoint at lower PFOA doses and/or greater response at similar PFOA doses than PFOA exposure alone, this clearly indicated PFOS co-exposure produced cumulative toxicity with PFOA. We determined significant dose response shifts for the following F1 endpoints: pup survival (% live born and % uterine implants; Fig. 8A, S2), pup bodyweight (PND1 and PND2; Fig. 8B, S2), pup bodyweight gain (PND1-2; Figure S2), pup liver weight (relative and ANCOVA adjusted; Fig. 8C, S1), and pup total T3 (Figure S2). Further, PND2 pup serum clinical chemistry parameters with significant shifts in dose response included: creatinine, globulin, glucose, bile acids, total bilirubin, BUN, total protein, and triglycerides (Figure S3). Similarly, there were significant dose response shifts for the following maternal endpoints: bodyweight (GD22, Figure S2), weight gain (GD8-22, Fig. 8D), liver weight (absolute, relative, and ANCOVA adjusted; Fig. 8F, S2), and kidney weight (absolute, relative, and ANCOVA adjusted; Fig. 8E, S2). Finally, maternal serum clinical chemistry parameters with significant dose response shifts included: BUN, albumin, total protein, globulin and GLDH (Figure S4).

Fig. 8.

Selected dose response shift regression curves for the PFOA-only (blue squares) and PFOA + PFOS mixture (green circles) experiments. Data were fit with four parameter logistic regressions and determined to have similar slopes (p > 0.05), then baseline (top or bottom) and potency parameters were evaluated for statistical difference between individual and mixture experiments. Endpoints with a significant shift indicate combined effect of PFOA and PFOS, while those endpoints with no shift demonstrate the effect was largely determined by the PFOA dose. Additional dose response shift graphs for maternal and F1 endpoints, including clinical chemistry, in Figures S1–S3. Data points represent mean ± standard error.

For all endpoints listed above the mixture dose response shifted in-parallel towards effects at lower PFOA doses except for maternal bodyweight (GD22), maternal weight gain (GD8-22) and pup triglycerides, which displayed less of an effect in the mixture than the PFOA-only exposure (Fig. 8, S2, S3). Further, some measures of clinical chemistry, particularly maternal and pup serum proteins (e.g., albumin, globulin), had downward trends in the mixture, similar to PFOA-only, however the baseline was shifted up in the mixture, likely from the contribution of PFOS (Figure S3, S4). The pup parameters of total bile acids and bilirubin appeared to have a dramatically increased effect in the mixture compared to PFOA-only, however due to differences in slopes it was not possible to further evaluate mixture additivity for these effects (Figure S3).

In contrast, for several endpoints there were no significant shifts in the PFOA dose responses for endpoints which 2 mg/kg PFOS had minimal to no effect. Pup endpoints with no significant dose response shifts included: birthweight (absolute and ANCOVA adjusted, Fig. 8G, S2) and liver glycogen (Fig. 8H). Pup PND2 clinical chemistry parameters with no significant dose response shifts included: cholesterol, GLDH, and AST (Figure S3). Maternal endpoints with no shifts included: bodyweight (PND2; Fig. 8I) and weight gain (GD8-PND2; Figure S2). Maternal clinical chemistry parameters with no shifts included: triglycerides and ALT (Figure S4).

3.6.2. DA and RA models

The second mixture evaluation approach involved calculation of DA and RA mixture model predictions to determine whether DA or RA better predicted the observed mixture effects and if observed data displayed mixture interactions that deviated from additivity (Fig. 9 and S5). First, we analyzed the dose response for PFOA-only and PFOS-only endpoints to determine those that had congruent slopes and could produce unambiguous estimates of ED₅₀ to use in the mixture model calculations. The endpoints suitable for evaluating mixture models included: pup survival (% live born and % uterine implants), pup bodyweight (PND1 and PND2), pup total T3, pup liver glycogen concentration, maternal relative liver weight, and maternal weight gain (GD8-22).

Fig. 9.

Comparison of observed PFOA + PFOS mixture (green circles, black dotted line) dose response curves (with 95 % confidence intervals in grey) with dose addition (DA, red line) and response addition (RA, blue line) mixture model predictions. Dose addition more consistently fell within the observed 95 % CI, while RA tended to under- or over-predict some effects, while both mixture models overpredicted effects on maternal weight gain (GD8-22). Analogous graphs depicting 95 % prediction intervals found in Figure S4. Data points represent mean ± standard error.

After analyzing the PFOA-only and PFOS-only dose response data and calculating DA and RA mixture model predictions, we calculated AICc scores for the observed data fit with DA or RA constrained regression models and calculate the ΔAICc between the two models. Further, the ΔAICc was calculated between each of the DA and RA models and the observed best fit model and these values were used to calculate Akaike weights (Wm) for each of the DA and RA models, which represent probabilities that one model is preferred versus the other (Table 5). DA models had significantly lower AICc scores (i.e., >95 % probabilities) for the following endpoints: pup survival (% live born and % of uterine implants), pup total T3, pup liver glycogen, and maternal relative liver weight. For the remaining variables, the ΔAICc between models resulted in Akaike weights that did not differ significantly between DA and RA predictions including pup bodyweight (PND1 and PND2) and maternal weight gain (GD8-22) (Table 5).

Table 5.

Mixture model fits for DA and RA models using corrected Akaike’s Information Criteria (AICc).

Endpoint	Best Fit AICc	DA AICc	RA AICc	ΔAICc (RA-DA)	DA Wm	RA Wm
Pup total T3	61.7	74.2	92.4	18.2	1.000	0.000
Pup survival (% live born)	170.7	180.0	190.5	10.5	0.995	0.005
Maternal relative liver weight	94.8	100.5	109.6	9.1	0.990	0.010
Pup survival (% implants)	167.9	181.0	189.5	8.5	0.986	0.014
Pup liver glycogen	140.1	147	155.5	8.5	0.986	0.014
Pup bodyweight (PND2)	80.8	83.1	83.6	0.5	0.562	0.438
Pup bodyweight (PND1)	87.1	90.4	90.2	−0.2	0.480	0.520
Maternal weight gain (GD8-22)	143.2	169.8	168.9	−0.9	0.389	0.611

Wm = Akaike Weighted mean values are model likelihoods and represent the probability that a given model (DA or RA) is the best approximating model. DA = dose addition, RA = response addition, PND = postnatal day.

Visual observations of plots of the DA and RA mixture model regressions with the observed data indicate that the DA model regression line was either mostly or entirely within the observed 95 % confidence interval, while RA either over- or under-estimated some responses. Both models overestimated effects on maternal gestational weight gain (Fig. 9), which indicates a less-than-additive mixture effect for this endpoint. When using 95 % prediction intervals (as opposed to 95 % confidence intervals) as described by Gennings (1995), the DA models were within the observed 95 % prediction intervals for all effects indicating no significant mixture interactions (i.e., no departures from additivity) (Figure S5). In contrast, the RA model was lower than the 95 % prediction interval for pup total T3 in the lower dose range (≤10 mg/kg PFOA) (Figure S5).

We also determined effective dose level (EDx) potency estimates from the observed data and compared the accuracy of DA or RA predictions (Table 6). A lower effect level (i.e., effective dose 15 % or 20 % (ED₁₅, ED₂₀) was used when the data allowed for 95 % confidence intervals to be estimated and ED₅₀ was used otherwise. Further, we calculated Index of Prediction Quality values as described by (Martin et al. 2021) which represents the % deviation of the model prediction from the observed value. As described above, DA and RA both overestimated effects on maternal gestational weight gain and had predictions ~ 2400 % and ~ 1500 % more potent than observed, respectively. For all remaining endpoints DA produced potency estimates that only deviated 10–58 % from observed and within or nearly within the observed 95 % CI. In contrast, RA model estimates varied from observed by −105 – 205 % and were only within the observed 95 % CI for two endpoints (PND1 and PND2 pup bodyweight) (Table 6).

Table 6.

Observed effective dose levels (EDx) for PFOA + PFOS mixture compared to DA and RA model predictions.

Endpoint	ED level	Observed EDx (as PFOA mg/kg)	Observed (95 % CI)	DA EDx (mg/kg)	DA IPQ (%)	RA EDx (mg/kg)	RA IPQ (%)
Pup total T3	50	17.8	11.7–24.4	15.1*	17.5	5.8	205.9
Pup liver glycogen (PND0)	50	28.4	13.8–44	19.0*	49.3	11.4	147.8
Pup bodyweight (PND2)	20	34.2	20.6–47.4	25.2*	35.6	37.9*	−10.9
Pup bodyweight (PND1)	20	42.1	22.3–60	33.6*	25.4	36.4*	15.8
Maternal relative liver weight	20	52.4	42.9–60.5	47.6*	10.1	69.6	−32.7
Pup survival (% implants)	50	59.4	44.3–72.9	37.6	44.9	121.7	−104.5
Pup survival (% live born)	50	61.7	45.1–79.1	42.6	57.9	126.0	−104.9
Maternal weight gain (GD8-22)	15	72.6	27.1–118.1	2.9	2418.9	4.4	1551.7

Observed EDx level reported as dose of PFOA plus 2 mg/kg PFOS

^*Model estimate within 95 % CI of observed value.

IPQ = Index of Prediction Quality as reported by Martin et al. (2021), represents % deviation of model value from. observed. BW = body weight, ED20 = effective dose 20 %, ED50 = effective dose 50 %, PND = postnatal day, DA = dose addition, RA = response addition, GWG = gestational weight gain (GD8-22).

3.6.3. RPF approach

Finally, we calculated RPFs for all endpoints with congruent slopes between PFOA-only and PFOS-only based on the available data. Using PFOA as the index chemical, the PFOS RPFs ranged over an order of magnitude from 1.8 to 43.7 with no consistency within maternal (RPF range 7.8 – 26.3) or pup (RPF range 1.8 – 43.7) endpoints (Table 7, Fig. 10). Similar to DA model predictions above, the dose response for all endpoints except maternal weight gain (GD8-22) and maternal GD22 bodyweight were statistically similar between all three experiments using RPF-adjusted PFOA equivalent doses (Figs. 11, 12). The effect on maternal weight gain and bodyweight was significantly less than the PFOA-only and PFOS-only effects, indicating a less-than-additive response. Maternal endpoints consistent with the RPF approach included: bodyweight (PND2, Fig. 11C), liver weight (absolute and relative, Fig. 11D,E), relative kidney weight (Fig. 11F), total T4 (Fig. 11G), total T3 (Fig. 11H), and free T4 (Fig. 11I). Neonatal endpoints consistent with the RPF approach included: bodyweight (adjusted birth, PND1, PND2, Fig. 12A–C), survival (% implants and % live born, Fig. 12D,E), liver glycogen concentration (Fig. 12F), total T3 (Fig. 12G), and total T4 (Fig. 12H). Overall, for nearly all maternal and neonatal endpoints modeled here the RPF approach produced highly accurate estimates of dose additive effects from combined exposure to PFOA and PFOS.

Table 7.

PFOA and PFOS dose response slope, effect dose levels (EDx), and RPF values (PFOA = index chemical).

Endpoint	Slope	ED level	PFOA EDx	PFOS EDx	PFOA RPF	PFOS RPF
Maternal Total T3	−0.39	20	5.2(1.9–14.2)	0.7(0.3–1.4)	1(0.5–2.1)	7.8(2.9–21.5)
Maternal bodyweight (GD22)	−0.61	10	40.3(17.5–92.8)	3.7(1.4–9.8)	1(0.4–2.7)	11.0(4.8–25.3)
Maternal Total T4	−0.61	50	14.8(10.0–21.8)	1.1(0.8–1.6)	1(0.7–1.4)	12.9(8.8–19.1)
Maternal Free T4	−0.55	50	45.4(18.9–109.3)	3.3(1.3–8.4)	1(0.4–2.6)	13.9(5.8–33.4)
Maternal weight gain (GD8-22)	−1.20	50	141.4(84.4–237)	9.2(4.9–17.6)	1(0.5–1.9)	15.3(9.1–25.6)
Maternal relative kidney wt	1.17	10	54.8(33.5–89.7)	3.5(2.2–5.7)	1(0.6–1.6)	15.6(9.5–25.5)
Maternal relative liver wt	1.74	50	189.2(143.3–249.8)	9.8(7.1–13.5)	1(0.7–1.4)	19.3(14.6–25.5)
Maternal absolute liver wt	5.89	10	117.9(56.0–247.9)	5.0(3.9–6.3)	1(0.8–1.3)	23.7(11.3–49.9)
Maternal bodyweight (PND2)	−0.38	10	74.6(14.8–375.5)	2.8(0.7–12)	1(0.2–4.2)	26.3(5.2–132.2)
Pup liver glycogen	−0.76	50	22.5(12.5–40.8)	12.2(3.9–38.5)	1(0.3–3.2)	1.8(1.0–3.3)
Pup adjusted birthweight	−0.92	10	37.4(26.2–53.5)	6.5(3.6–11.9)	1(0.5–1.8)	5.7(4.0–8.2)
Pup total T3	−0.78	50	30.2(19.9–45.9)	4.1(1.9–8.8)	1(0.5–2.2)	7.4(4.9–11.2)
Pup bodyweight (PND1)	−1.16	50	233.2(158.7–342.8)	12.5(7.2–22)	1(0.6–1.8)	18.6(12.7–27.3)
Pup bodyweight (PND2)	−1.48	50	180.1(141.1–229.9)	7.7(5.5–10.8)	1(0.7–1.4)	23.4(18.3–29.8)
Pup total T4	−10.27	20	9(3.4–25.2)	0.3(0.1–0.8)	1(0.4–2.9)	31.0(11.6–87)
Pup survival (live born)	−6.27	50	129.1(119.5–139.4)	3(2.7–3.4)	1(0.9 –1.1)	42.7(39.6–46.2)
Pup survival (implants)	−6.21	50	125.9(113.9–139.3)	2.9(2.2–3.8)	1(0.8–1.3)	43.7(39.5–48.4)

Values represent mean (95% confidence interval).

Fig. 10.

Distribution of PFOA and PFOS relative potency factors (RPFs) across maternal and neonatal endpoints. PFOA was used as the index chemical (RPF = 1) and data points display mean value, whiskers represent 95 % CI. Note log scale of x-axis. RPFs ranged > 20-fold from 1.8 to 43.7 (values reported in Table 7).

Fig. 11.

Relative potency factor (RPF) evaluation of PFOA, PFOS, and PFOA + PFOS mixture dose response data for maternal endpoints. The dose of PFOS was adjusted based on the RPFs reported in Table 7 and plotted as PFOA equivalents. Adjusted data were fit with four parameter logistic regression to statistically compare dose response functions and determine if a single curve could fit all data sets. Only for maternal weight gain (GD8-22) and maternal bodyweight (GD22) did the mixture data deviate from RPF-adjusted PFOS and PFOA data, indicating less than dose additive effects for these two endpoints. Data points represent mean ± standard error.

Fig. 12.

Relative potency factor (RPF) evaluation of PFOA, PFOS, and PFOA + PFOS mixture dose response data for neonatal (pup) endpoints. The dose of PFOS was adjusted based on the RPFs reported in Table 7 and plotted as PFOA equivalents. Adjusted data were fit with four parameter logistic regression to statistically compare dose response functions and determine if a single curve could fit all data sets. Data points represent mean ± standard error.

4. Discussion

The mixture of PFOA + PFOS clearly demonstrated cumulative effects on both maternal and F1 animals from combined exposure during pregnancy and early lactation. The total spectra of effects were not identical between PFOA-only and PFOS-only exposures across the dose ranges utilized, but there were multiple shared endpoints significantly altered with the same direction of effect by both compounds. The cumulative effects of PFOA + PFOS mixture exposure were clearly demonstrated by the significant dose response shifts compared to PFOA-only exposure for multiple endpoints. For all endpoints that were shared between compounds and could be modeled to generate appropriate dose response parameters, there were statistically equivalent or better predictions of mixture effects using a DA model as opposed to a RA model. Importantly, PFOA and PFOS RPFs produced accurate estimates of mixture effects, but spanned more than an order of magnitude across endpoints. Finally, multiple lines of evaluation indicated that mixture interactions deviating from dose additivity were rare, relatively minor, and were less-than-additive when observed, with no indication of greater-than-additive mixture effects.

Currently, there is broad support for the scientific basis of cumulative effects from combined exposure to two or more compounds that have shared responses regardless of mode of action (Kortenkamp 2022; NRC 2008; Van Der Ven et al. 2022). This is particularly supported for shared apical endpoints, whereby multiple pathways can converge on a given adverse outcome. Dose addition is widely viewed as the default approach for estimation of cumulative effects from combined exposure to toxicologically similar compounds, unless empirical evidence indicates the use of a different approach (i.e., RA), and the present results further support this assumption (USEPA 2000). Recent systematic reviews of published mixture studies have found that statistical support for greater-than-additive or less-than-additive mixture interactions are uncommon and rarely deviate >2-fold from dose additive predictions (Boobis et al. 2011; Martin et al. 2021). There are multiple model equations that can be used to estimate mixture effects for both DA and RA (Altenburger et al. 2000). Generalized concentration addition (GCA) was recently demonstrated to be the most accurate for estimating in vitro nuclear receptor activation of mixtures of partial and full PPARα agonists (Nielsen et al. 2021); however, it has not been demonstrated that this is a suitable model for prediction of in vivo effects. Further, the GCA model assumes a slope of 1 and some endpoints modeled here deviated significantly from this default slope constant. The simple DA model used here has been shown to produce more accurate results to RA in studies of different receptor-based pathways and effect endpoints (Conley et al. 2018; Conley et al. 2021a; Rider et al. 2008; Rider et al. 2010). The RPF approach is broadly used in risk assessment and assumes statistically similar slope parameters such that mixture chemicals act as a dilution of an index chemical. Here, most endpoints displayed statistically similar slopes and were suitable for RPF evaluation; however, the PFOA and PFOS RPFs varied depending on the specific endpoints modeled and one RPF was not suitable for use across all endpoints for these chemicals (Fig. 10, Table 7; RPFs ranged > 20-fold from 1.8 to 43.7). Thus, data here support the use of dose addition approaches for estimating mixture effects, but the maternal and developmental effects of PFOA and PFOS studied here indicate RPFs are variable and endpoint specific.

The most sensitive effect measured in the present studies was reduced serum thyroid hormone concentrations, which occurred at all or nearly all doses for both PFOA and PFOS depending on the specific hormone (total or free T3 or T4) and lifestage (maternal or neonatal). Interestingly, PFOS did not reduce maternal free T3. It is important to reiterate the pleiotropic effects of thyroid hormones, particularly during human and rodent fetal and neonatal development, including impacts on metabolic homeostasis and growth (Eng and Lam 2020). Maternal, fetal and neonatal hypothyroxinemia (reduced T3 and/or T4 in the absence of an effect on TSH) has adverse consequences for pregnancy maintenance and child development (Kooistra et al. 2006; Yang et al. 2020). Thus, the reductions in serum thyroid hormones described here were potentially involved in the observed reductions in pup bodyweights and growth; however, this requires additional investigation. We observed that the 2 mg/kg PFOS dose used in the mixture experiment produced near maximal reductions in thyroid hormones, except for pup total T3, and thus in the mixture experiment both maternal and F1 thyroid hormone levels were essentially at the lower plateau in all mixture doses and aligned well with RPF-based estimates of mixture effects. Future studies should employ designs and dosing structures with lower doses of PFOS to more fully model mixture effects on thyroid hormone levels and test hypotheses of dose additivity for these effects.

Adverse liver effects are often reported in PFAS toxicity studies and were clearly observed here, particularly in neonatal pups exposed to PFOA-only or the mixture. Bodyweight-adjusted liver weight was increased in all mixture doses in PND2 pups and liver necrosis was identified in these animals at ≥ 10 mg/kg, whereas absolute pup liver weight displayed a high dose non-monotonic dose response (increasing at middle doses then decreasing at high dose, Figure S1) likely associated with severe pup toxicity at the top dose, similar to that seen with PFOA-only exposure. Pathological findings were also associated with increased serum total bile acids, total cholesterol, total bilirubin, GLDH, and AST in PND2 pups. Interestingly, there was a strong effect on PND2 pup serum triglycerides in the PFOA-only exposure that was abrogated in the PFOA + PFOS mixture exposure, indicating that PFOS dominated this effect. There were also clinical chemical indicators of adverse kidney effects in neonatal pups and histopathological lesions and clinical chemical indicators in maternal animals. Pup serum BUN dramatically increased from PND0 to PND2 in a dose responsive manner and was significant at all doses, along with the BUN:creatinine ratio. Similarly, BUN and BUN:creatinine were significantly elevated in all doses in maternal sera, coincident with findings of tubule mineralization, medulla cyst, retrograde nephropathy, and cortex fibrosis at ≥ 30 mg/kg in the mixture study.

The majority of maternal and offspring effects described here have previously been reported for PFOA and PFOS in either rat or mouse models, or both (Abbott et al. 2009; Abbott et al. 2007; Blake et al. 2020; Butenhoff et al. 2004; Lau et al. 2003; Luebker et al. 2005a; Luebker et al. 2005b; Thibodeaux et al. 2003). Further, similar maternal and F1 effects have been described for PFNA, PFHxS, HFPO-DA (GenX), and Nafion byproduct 2 (Blake et al. 2020; Butenhoff et al. 2009; Conley et al. 2021b; Conley et al. 2022; Conley et al. 2019; Das et al. 2015; Gilbert et al. 2021; Lieder et al. 2009). It appears that reduced pup survival, reduced pup bodyweight, decreased serum T4 and T3 and increased maternal liver weight are all common apical effects of exposure to straight chain and ether-linked carboxylate and sulfonate PFAS during pregnancy, among others. In our studies, the broader spectra of effects differ somewhat between carboxylates and sulfonates with carboxylates producing statistically significant effects for slightly more of the endpoints measured than sulfonates. Regardless of functional group, these effects appear to occur partially due to multiple key events associated with disrupted carbohydrate and lipid homeostasis between the mother, placenta, and fetus during pregnancy and further mechanistic investigation of these effects is needed.

Maternal and offspring carbohydrate and lipid dysregulation from PFAS exposure are plausibly associated with multiple molecular mechanisms, including receptor activation of peroxisome proliferator activated receptor alpha and/or gamma (PPARα and PPARγ). All sulfonate and carboxylate PFAS that have been evaluated for developmental toxicity (i.e., those listed above) have been shown in vitro to bind and activate both PPARα and PPARγ with variable potency (Behr et al. 2020; Evans et al. 2022; Houck et al. 2021; Nielsen et al. 2021; Takacs and Abbott 2007; Vanden Heuvel et al. 2006; Wolf et al. 2014). Further, short term in vivo exposures have demonstrated up-regulation of target genes associated with PPARα and PPARγ, as well as constitutive androstane receptor (CAR) and pregnane × receptor (PXR) (Bjork et al. 2008; Conley et al. 2022; Conley et al. 2019; Gilbert et al. 2021; Rosen et al. 2017; Rosen et al. 2007). It is important to note that PPARα knockout mouse studies of PFOA and PFOS have reported numerous maternal and developmental effects, underscoring the involvement of multiple nuclear receptors (Abbott et al. 2009; Abbott et al. 2007). It appears that regardless of carboxylate or sulfonate functional group, PFAS developmental toxicants target multiple nuclear receptors in multiple target tissues and that fatty acid oxidation is one major signaling pathway that is commonly upregulated. In this regard, the biological relevance of both PPARα and PPARγ to human health cannot be overstated (Dixit and Prabhu 2022). These two nuclear receptors have been and continue to be heavily exploited as pharmaceutical targets due to their myriad regulatory effects on carbohydrate and lipid metabolism (Fievet et al. 2006), which are essential physiological processes during pregnancy, among other lifestages (Ditzenberger 2018). Importantly, development of drugs that have dual PPARα/γ activity have universally failed various phases of clinical trials due to adverse effects (Hong et al. 2018). Apparent species-specific differences in particular key events downstream of receptor activation have been experimentally investigated (Berthou et al. 1996; Schlezinger et al. 2021) and discussed recently in the context of PFAS (Fragki et al. 2021) but the overall pathway of lipid metabolism regulated by PPARα, in particular, is highly conserved among species (Kersten 2014).

Combined exposure to PFOA and PFOS in the present mixture study, even at relatively high oral dose levels, did not impact serum or liver concentrations as compared to the individual chemical studies. Similar serum or liver data are not available in the literature for PFOA in the SD rat; however, comparison of PFOS concentrations with those reported by Thibodeaux et al. (2003) indicate very similar liver concentrations at the same oral dose, but higher serum levels from the postnatal maternal samples collected here as opposed to prenatal samples from Thibodeaux et al. (2003). PFAS, particularly PFOA and PFOS, are known to be eliminated and reabsorbed by renal organic anion transporters, which are potentially saturable processes (Loccisano et al. 2012; Weaver et al. 2010; Zhao et al. 2017). Combined exposure to multiple PFAS could potentially alter the chemical disposition and resulting serum and tissue concentrations of co-exposed PFAS, however our results provide no evidence to support this speculation up to part per million level concentrations in a SD rat model. If combined exposure to multiple PFAS did alter disposition, mixture studies could produce effects that deviated from DA predictions that are based on oral doses.

Combined exposure during pregnancy and early lactation to PFOA + PFOS produced cumulative effects on both the dams and their offspring. Models based on dose additivity, including the simple DA equation by Rider and LeBlanc (2005) and particularly the RPF approach (Van Der Ven et al. 2022) produced accurate estimates of mixture effects for nearly all endpoints evaluated. There was little indication of mixture interactions with maternal bodyweight and weight gain at term being the only endpoints significantly different between dose additive mixture models and observed data and these effects were less-than-additive. To our knowledge, this is the first study to empirically test the mixture-based effects of combined exposure to PFAS in a mammalian in vivo model. Prior mixture studies have been conducted and are informative but were not designed to statistically evaluate mixture effects using well established mixture models like DA and RPF (Crute et al. 2022; Marques et al. 2021; Roth et al. 2021). Further, it can be reasonably expected that combined exposure to multiple PFAS beyond the two studied here would produce cumulative effects on shared endpoints from combined exposure. The data presented here highlight the need for cumulative assessment of PFAS co-exposures and support recent efforts that utilize dose addition-based methods for estimating combined effects for toxicologically similar chemicals (RIVM 2018; USEPA 2021).

Abbreviations:

PFAS

Per- and polyfluoroalkyl substances

GD

Gestation day

PND

Postnatal day

RPF

Relative potency factor

T3

Triiodothyronine

T4

Thyroxine

DA

Dose addition

RA

Response addition

Data availability

Data will be made available on request.