Assessment of the mode of action underlying development of liver lesions in mice following oral exposure to HFPO-DA and relevance to humans

Abstract HFPO-DA (ammonium, 2,3,3,3-tetrafluoro-2-(heptafluoropropoxy)propanoate) is a short-chain polyfluorinated alkyl substance (PFAS) used in the manufacture of some types of fluorinated polymers. Like many PFAS, toxicity studies with HFPO-DA indicate the liver is the primary target of toxicity in rodents following oral exposure. Due to the structural diversity of PFAS, the mode of action (MOA) can differ between PFAS for the same target tissue. There is significant evidence for involvement of peroxisome proliferator-activated receptor alpha (PPARα) activation based on molecular and histopathological responses in the liver following HFPO-DA exposure, but other MOAs have also been hypothesized based on limited evidence. The MOA underlying the liver effects in mice exposed to HFPO-DA was assessed in the context of the Key Events (KEs) outlined in the MOA framework for PPARα activator-induced rodent hepatocarcinogenesis. The first 3 KEs (ie, PPARα activation, alteration of cell growth pathways, and perturbation of cell growth/survival) are supported by several lines of evidence from both in vitro and in vivo data available for HFPO-DA. In contrast, alternate MOAs, including cytotoxicity, PPARγ and mitochondrial dysfunction are generally not supported by the scientific literature. HFPO-DA-mediated liver effects in mice are not expected in humans as only KE 1, PPARα activation, is shared across species. PPARα-mediated gene expression in humans produces only a subset (ie, lipid modulating effects) of the responses observed in rodents. As such, the adverse effects observed in rodent livers should not be used as the basis of toxicity values for HFPO-DA for purposes of human health risk assessment.

Per-and polyfluoroalkyl substances (PFAS) are a large anthropogenic family of organic fluorinated compounds that have been used for decades in a broad range of industrial and consumer applications or products. The multiplicity of uses for these compounds stems from their physical and chemical properties. These same properties that impart functionality in products/applications also make PFAS resistant to biodegradation, photooxidation, direct photolysis, and hydrolysis, resulting in persistence and detection in the environment (Lau et al., 2007). Toxicity studies in rodents examining effects following oral exposure to these compounds indicate that the liver is the primary target of toxicity (Costello et al., 2022). However, considering the structural diversity of PFAS (eg, carbon chain length, interchain linkages, and functional groups), the mode of action (MOA) may differ between PFAS. Recently, an expert review panel, convened to provide guidance on key questions in PFAS risk assessment, concluded that "all PFAS" should not be grouped together for the purposes of assessing human health risk, as environmental persistence alone is not an adequate reason for grouping. The expert panelists also generally agreed that assuming similar toxicity or equal potency across the diverse class of PFAS is inappropriate (Anderson et al., 2022).
HFPO-DA (ammonium, 2,3,3,3-tetrafluoro-2-(heptafluoropropoxy)-propanoate; CASRN 62037-80-3) is a short-chain PFAS used as a polymerization aid in the manufacture of some types of fluorinated polymers. Although HFPO-DA is broadly considered the replacement for perfluorooctanoate acid (PFOA) (See Guo et al. (2021) for structural comparisons of HFPO-DA and other PFAS.), different fluoropolymer manufacturers have developed their own individual PFOA replacement polymerization aid chemistries (Wang et al., 2013). HFPO-DA is not used or found in the myriad of applications that PFOA has been historically used in, such as firefighting foam, carpets, paper, textiles, and other industrial and consumer products. In contrast, the limited number of fluoropolymers manufactured using HFPO-DA are utilized in important technologies including semiconductor fluid handling, high purity chemical processing, aerospace and telecommunications cabling, renewable hydrogen production, and lithium-ion batteries in transportation (https://www.teflon.com/en/industries-and-solutions/industries; https://www.nafion.com/en/industries).
In comparison to longer chain counterparts (eg, PFOA) which are often bioaccumulative in tissues, HFPO-DA is rapidly eliminated and does not bioaccumulate in mammalian tissues (Gannon et al., 2016). The rapid elimination of HFPO-DA is consistent across species, with similar clearance rates in monkeys, rats, and mice, except for an approximate 10-fold increase in clearance in female rats (Gannon et al., 2016). Further, recent ecological bioaccumulation studies in plants and aquatic organisms also demonstrate that HFPO-DA is not bioaccumulative in these organisms either (Wang et al., 2022b).
The toxicity database for HFPO-DA has been reviewed previously (Thompson et al., 2019). Therein, it was shown that HFPO-DA is not genotoxic, but induced cancers in Sprague Dawley rats in a pattern consistent with the PPARa tumor triad of liver adenomas/carcinomas, testicular Leydig cell tumors and pancreatic acinar cell tumors Felter et al., 2018;Klaunig et al., 2003). Specifically, female rats developed liver tumors at 500 mg/kg-day HFPO-DA, whereas male rats, exposed up to 50 mg/kg-day, exhibited a small but statistically significant increase in pancreatic acinar cell adenomas/carcinomas and a slight elevation in Leydig cell tumors-both at 50 mg/kg-day (Caverly Rae et al., 2015). Thus, the carcinogenicity profile for HFPO-DA is consistent with other PPARa activators, inducing changes in the 3 tissues in the rat associated with the tumor triad. Although the carcinogenicity of HFPO-DA has not been evaluated in mice, the histopathological responses to HFPO-DA observed in mouse livers from subchronic studies are consistent with other PPARa activators and have been recently reviewed by Thompson et al. (forthcoming).
The MOA for liver tumors in rodents induced by PPARa activators is well-established in the scientific literature (Corton et al., 2014;Klaunig et al., 2003), and has also been proposed as the MOA for liver tumors in rats exposed to PFOA (Klaunig et al., 2012). The human relevance of each Key Event (KE) within the PPARa MOA was thoroughly evaluated by Corton et al. (2018), with authors concluding that only KE 1, PPARa activation, is shared across humans and rodents. Furthermore, within KE 1, only receptor activation and induction of lipid metabolism genes occur in both species and downstream cell proliferation signaling, a key and required event in the formation of hepatic tumors, occurs specifically in rodents . Based on this evaluation, these authors concluded that tumors induced via the PPARa MOA are not relevant to humans .
In a 2021 toxicity assessment of HFPO-DA, the United States Environmental Protection Agency (USEPA) developed chronic and subchronic reference doses (RfDs) based on liver effects in mice from a subchronic oral toxicity study. Despite the significant evidence for involvement of peroxisome proliferator-activated receptor alpha (PPARa) activation according to both molecular and histopathological responses in the liver following HFPO-DA exposure in both rats and mice, alternate MOAs have been hypothesized with little supporting evidence. These hypothesized MOAs include those that involve cytotoxicity, participation of other PPAR subtypes (eg, PPARc), or mitochondrial dysfunction (USEPA, 2021). Importantly, the liver effects used to develop USEPA's RfDs for HFPO-DA are part of the early KEs in the PPARa MOA. Herein, we reviewed toxicity data for HFPO-DA in mice from the primary peer-reviewed literature and USEPA's Health and Environmental Research Online (HERO) database and assessed these data in the context of the first 3 KEs outlined in the PPARa MOA framework as described in Corton et al. (2014). Since the time of USEPA's assessment, additional data important for informing the MOA for HFPO-DA has been generated, including caspase-3 immunohistochemical and transcriptomic analyses of mouse liver from the reproductive/developmental study used to develop USEPA's RfDs for HFPO-DA (Heintz et al., 2022;Thompson et al., forthcoming). This new information, along with existing information, including published hepatic transcriptomic results in mice for HFPO-DA from a 90-day subchronic study (Chappell et al., 2020) were analyzed together and integrated herein to evaluate transcriptomic support for the early KEs of the PPARa MOA and alternate MOAs. The concordance of timing and dose-response of observed liver effects for HFPO-DA, biological plausibility, and human relevance are evaluated within the PPARa MOA framework.

MOA analysis
To evaluate the evidence concerning the PPARa MOA for HFPO-DA in mice, we have characterized the evidence base for the first 3 KEs of the PPARa MOA framework described in Corton et al. (2014) (Figure 1A; Table 1).

KE 1: PPARa activation
Broadly, evidence streams for PPARa activation (KE 1) include PPARa receptor binding and/or activation, increased expression of genes/proteins involved in fatty acid b-oxidation, increased palmitoyl-CoA oxidase activity, and morphological evidence of peroxisome proliferation . In addition, analysis of mRNA or transcriptomic responses to PPARa activation, or the loss of any of the aforementioned effects in knockout studies, also provides evidence of PPARa activation. To date, we are not aware of any published studies on HFPO-DA in genetically modified rodent models.
HFPO-DA-mediated PPARa activation is supported by several lines of evidence (Table 1), including the activation of both mouse and rat PPARa receptors by HFPO-DA in in vitro reporter assays (Chappell et al., 2020;Evans et al., 2022). Activation of PPARa by HFPO-DA was measured by luciferase activity in reporter gene constructs, with a lower EC 50 , equal to 40 mM, observed for mouse PPARa, compared to an EC 50 equal to 114 mM for rat PPARa (Chappell et al., 2020). Investigations on the potency of PFAS activating rat PPARa found that HFPO-DA and HFPO-DA ammonium salt were the most potent among 16 PFAS studied, with EC 20 values equal to 83.18 mM and 75.86 mM, respectively. Rat PPARa activation levels for all other PFAS tested (including PFOA) were too low compared to the positive control to calculate EC 20 values. HFPO-DA was also a more potent activator of rat PPARa than the endogenous fatty acids, oleic acid (EC 20 ¼ 794.33 mM) and clofibric acid (EC 20 ¼ 562.34 mM) (Evans et al., 2022). Triglycerides of oleic acid are the primary component of some vegetable oils, most notably olive oil (Boskou et al., 2006).
Exposure of HFPO-DA for 28 days also increased hepatic peroxisome b-oxidation activity in both mice and rats (Thompson et al., 2019). Although a significant increase in peroxisome b-oxidation activity was observed at HFPO-DA dose levels ! 3 mg/kg in mice dosed with 0, 0.1, 3, or 30 mg/kg for 28 days (Thompson et al., 2019), no enzyme activity data are available at doses between 0.1 and 3 mg/kg. However, analysis of hepatic transcriptomic data in mice from a 90-day subchronic study by Chappell et al. (2020) and from a reproductive/developmental study by Heintz et al. (2022) demonstrate significant enrichment of the specific gene sets: KEGG peroxisome, REACTOME peroxisomal lipid metabolism, and REACTOME b-oxidation of very long chain fatty acids, at both 0.5 and 5 mg/kg HFPO-DA (Supplementary Table 1), with median benchmark doses (BMDs) for the latter REACTOME gene sets ranging between 0.39-0.67 and 0.24-0.34 mg/kg HFPO-DA, respectively (Heintz et al., 2022). These results indicate induction of peroxisome b-oxidation by HFPO-DA at dose levels lower than 0.5 mg/kg in male and female mice.
Additional evidence for PPARa activation in the livers of HFPO-DA-exposed mice is also provided by the transcriptomic findings from Chappell et al. (2020) and Heintz et al. (2022). Hepatic transcriptomic data from these studies show significant enrichment of gene sets related to fatty acid metabolism, peroxisome proliferation, and PPAR signaling (ie, "KEGG PPAR Signaling Pathway" and "WP PPAR Signaling"). Although general PPAR signaling gene sets were among the most significantly upregulated in both Figure 1. Transcriptomic concordance and support for the mode of action (MOA) underlying development of liver lesions in mice following oral exposure to HFPO-DA. A, Key and associative events of the PPARa MOA in rodents. Data supporting the key and associative events within the dashed box are evaluated herein for HFPO-DA. Key Events are empirical and observable causal precursor steps to the formation of liver tumors. Each sequential event is necessary, but not sufficient by itself in the absence of other Key Events, for the formation of liver tumors. Associative Events are biological endpoints that can be used as indicators or biomarkers of Key Events in the PPARa MOA. For the purposes of this current evaluation Key and Associative Events are not discerned. Adapted from Corton et al. (2014). B, Range plots (median BMDL, median BMD, and median BMDU) for select pathways containing dose-responsive genes in livers of HFPO-DA exposed male and female mice from Heintz et al. (2022) and Chappell et al. (2020). See Heintz et al. (2022) for additional benchmark dose modeling details. C, Heatmap of hepatic gene expression of selected genes included in PPARamediated lipid transport and metabolism (KEGG PPAR Signaling Pathway) or cytotoxicity (Corton et al., 2020;Glaab et al., 2021) gene sets, as well as genes encoding liver enzymes (Alpl (ALP), Got1 (AST), and Gpt (ALT)) and bile acid transporters. Significant (FDR < 0.1) differentially expressed genes (rows) are indicated by green (up-regulated genes) or red (down-regulated genes) colors. Intensity of colors is based on the log(fold change) value of each gene for each study, sex, and dose group. White cells indicate gene was not significantly altered by HFPO-DA exposure compared to respective controls. Transcriptomic data included in this heatmap are from the OECD 421 reproductive/developmental toxicity and the OECD 408 90-day toxicity study for male and female mice. HFPO-DA dose levels are indicated by right triangles at the top of the heatmap for each study, increasing from 0.1 to 5 mg/kg for each study and sex. Methods for transcriptomic analyses are described in Chappell et al. (2020) and Heintz et al. (2022). D, Summary of the role of PPARa in the observed liver changes in mice exposed to HFPO-DA. Upon PPARa activation, solid arrows represent either subsequential PPARamediated or downstream events, and dashed arrows indicate positive feedback mechanisms.
GD, gestation day. a þ indicates positive findings and À indicates negative findings in HFPO-DA-exposed mouse or rat models. b Putative marker of hepatocellular proliferation and PPARa-mediated bile acid perturbation in rodents. studies, the specificity for PPARa activation by HFPO-DA was demonstrated by the differentially expressed genes that are specific to the a subtype and a general lack of changes in expression levels of most of the genes specific to other PPAR subtypes (Supplementary Figure 1). Importantly, although some publicly curated PPAR signaling gene sets (eg, KEGG) denote that gluconeogenesis genes are specifically regulated by PPARc, which is likely the case for adipose tissue, several studies have determined PPARa as a key regulator of glucose metabolism in the liver (Kersten, 2014;Peeters and Baes, 2010). Thus, differential gene expression of Pck1 and Aqp7 in high dose groups across mouse studies in Supplementary Figure 1 likely indicates PPARa-mediated regulation of gluconeogenesis in the liver. Gene sets specific to the PPARa subtype were also significantly enriched in both male and female mice at 0.5 and 5 mg/kg HFPO-DA (Supplementary Table 1). In addition to the upregulation of genes that affect fatty acid uptake, activation and oxidation, HFPO-DA upregulated several Pex genes, involved in peroxisome biogenesis, in male and female mice at 0.5 and 5 mg/kg. Similarly, other research groups (Guo et al.,2021(Guo et al., , 2022Wang et al., 2017) also determined that KEGG PPAR signaling and lipid metabolism gene sets were significantly enriched in male mice treated for 28 days with 0.4-10 mg/kg or 1 mg/kg HFPO-DA, respectively. A recent review of rodent liver histopathology in response to PFAS noted that "PFAS-induced liver injury and steatosis may not depend on PPARa alone" (Costello et al., 2022). However, a separate recent review of multiple HFPO-DA studies in mice found that most studies did not report steatosis and those that did were not convincing based on the published images (Thompson et al., forthcoming). Morphological evidence of peroxisome proliferation (eg, hepatocellular hypertrophy/cytoplasmic alteration) from HFPO-DA exposure was observed in H&E stained livers of mice from subchronic and reproductive/developmental toxicity studies (DuPont, 2010a,b; NTP, 2019). In addition, Blake et al. (2020) also reported an increased number of peroxisomes with increasing HFPO-DA dose in mice using transmission electron microscopy.
Evidence for KE 2 following HFPO-DA exposure is supported by benchmark dose modeling (BMD) results of gene expression data from Chappell et al. (2020) and Heintz et al. (2022) in addition to liver histopathology data (Table 1) (NTP 2019). Functional classification of significant dose-responsive genes showed lower median BMDs are associated with fatty acid metabolism-related gene sets, whereas enriched gene sets with higher median BMDs are related to mitotic cell cycle ( Figure 1B). These transcriptomic results are phenotypically anchored by H&E staining of liver sections, as mitotic figures were increased at the highest dose levels in both studies (Chappell et al., 2020; NTP, 2019).

KE 3: perturbation of cell growth and survival
Evidence for perturbation of cell growth and survival (KE 3) includes hepatocyte proliferation (increased cell number) and altered apoptotic rates, resulting in hepatocyte hypertrophy and subsequent liver enlargement. At high doses, potent PPARa activators exhibit sustained or chronic increases in cell proliferation Ward et al., 1988). Support for KE 3 is well established based on transcriptomic, histopathology, and liver weight data (Table 1). Relative liver weight significantly increased in male and female mice from subchronic studies for HFPO-DA (Blake et al., 2020;DuPont, 2010a,b;Guo et al., 2021Guo et al., , 2022. Mitotic figures have been observed in H&E stained mouse liver sections at 5 mg/kg HFPO-DA (Chappell et al., 2020;DuPont, 2010a,b;NTP, 2019), as well as in the livers of mouse dams exposed to 2 and 10 mg/kg (Blake et al., 2020). In addition, increased incidence of "single cell necrosis" has been reported in mice from 90-day and reproductive/developmental toxicity studies (DuPont, 2010a,b; NTP, 2019); however, a recent reevaluation of liver sections from these studies demonstrated that some hepatocytes putatively considered to be necrotic also stained with the apoptotic marker, activated caspase-3 (Thompson et al., forthcoming). As such, apoptotic cell death, as opposed to necrotic cell death, may be the predominant form of cell death in the livers of mice exposed to HFPO-DA.
Pro-apoptotic gene sets were also significantly enriched at higher dose levels in livers from HFPO-DA-exposed mice ( Figure 1B) (Chappell et al., 2020;Heintz et al., 2022). This induction of pro-apoptotic gene expression is anchored to phenotypic evidence for hepatocellular apoptosis via H&E staining and caspase-3 immunostaining (Chappell et al., 2020;Thompson et al., forthcoming). While PPARa activators are reported to suppress apoptosis under acute exposure scenarios, PPARa activators have also been reported to increase apoptosis in the livers of mice undergoing cell proliferation in repeat dose studies . In addition, exposure to the PPARa activator, WY-14643, has been shown to induce some pro-apoptotic gene expression and repress some anti-apoptotic genes in wild type but not PPARa-null mice (Xiao et al., 2006). These data suggest that PPARa might promote apoptotic signaling directly or as a homeostatic response to PPARa-related growth signals.
Temporal and dose-response concordance of HFPO-DA liver effects with PPARa MOA The U.S. EPA Guidelines for Carcinogen Risk Assessment and the IPCS advocated the adoption of the Bradford Hill criteria for assessing causality in epidemiological studies for application in judging the strength of data in supporting MOA analyses (Hill 1965;Sonich-Mullin et al., 2001;USEPA 2005). More recently, inconsistencies in the use of these frameworks have been identified and clarified by development of so-called evolved Bradford Hill criteria by asking specific questions related to each criterion . In addition to temporal concordance, each of these modified Hill criteria (ie, biological concordance, essentiality, concordance of empirical observations among KEs, consistency, and analogy) are evaluated below as they relate to the MOA for HFPO-DA-induced liver changes in mice.
The first evolved Bradford Hill consideration, biological concordance, determines whether the hypothesized MOA conflicts with broader scientific knowledge and additionally whether a MOA is well established . In the case of the PPARa MOA for HFPO-DA presented herein, the KEs are consistent with established PPARa activators including other PFAS. As for how well the MOA is established, the PPARa MOA is well established for rodent liver tumors (Corton et al., 2014;Felter et al., 2018;Klaunig et al., 2003). As previously noted, the analyses herein focus on the first 3 KEs because the non-neoplastic liver changes used to develop USEPA's RfDs for HFPO-DA are a part of the early KEs in the PPARa MOA. Although chronic bioassays with HFPO-DA are not currently available in mice, chronic exposures in mice are expected to yield similar results as to what has been observed in rats (ie, liver tumors) (Caverly Rae et al., 2015). As described in the MOA analysis above, HFPO-DA has been shown to cause sustained PPARa activation, altered cell growth and increased incidence of non-neoplastic lesions in the livers of mice. The PPARa MOA for HFPO-DA is supported by data consistent with the molecular biology of carcinogenesis (Hanahan and Weinberg, 2011) as well as data for other biological endpoints associated events not related to hepatocarcinogenesis, but related to PPARa-specific induction, such as peroxisome proliferation and the induction of lipid metabolism genes. Furthermore, increased hepatocyte and liver growth are fundamental features of tumor growth (Corton et al., 2014). Therefore, the available data for HFPO-DA are consistent with the early KEs in the PPARa MOA for rodent hepatocarcinogenesis including hepatocellular hypertrophy (cytoplasmic alteration), apoptosis, and focal necrosis.
The second evolved Bradford Hill consideration, essentiality, examines whether KEs are reversible if dosing ceases or if a KE is prevented . Although studies in PPARa-null mice are not yet available for HFPO-DA, the KEs are prevented/ mitigated in studies conducted with classic PPARa activators or other PFAS in PPARa-null mice.
The third evolved Bradford Hill consideration, concordance of empirical observations, assesses the dose-response and temporality of KEs in a hypothesized MOA . It should be appreciated that most MOA frameworks focus on carcinogenicity and thus dose and temporal concordance can be more easily assessed due to the usual delay in the onset of most tumors. However, because tissues constantly maintain homeostasis one KE does not necessarily occur to completion before the next KE begins. As such, early KEs tend to occur concomitantly. For example, chemical induced intestinal cytotoxicity does not necessarily result in obvious atrophy before regenerative processed begin (Bhat et al., 2020;Cohen et al., 2010;Cullen et al., 2016). Table 2 shows the dose and temporal concordance of the KEs in the proposed MOA for liver changes following exposure to HFPO-DA. Regarding dose-response concordance, there is little evidence for transcriptomic, enzymatic (eg, b-oxidation), or histopathological changes in the mouse liver at 0.1 mg/kg after either 28 or 90 days of exposure, suggesting a potential threshold in response. Beginning at 0.5 mg/kg, there are transcriptomic responses related to PPAR signaling and fatty acid metabolism. At the tissue level, there is also evidence of hepatocellular hypertrophy in increased liver weight at !0.5 mg/kg. There is also evidence for PPARa activation, hepatocellular hypertrophy and hepatomegaly at 1 mg/kg. At 3 mg/kg, there is a large and statistically significant increase in b-oxidation activity. At 5 mg/kg, there is transcriptomic evidence for altered cell growth pathways related to mitotic and apoptotic signaling. At 5 mg/kg, there is also histopathological evidence for both mitosis and apoptosis.
The fourth evolved Bradford Hill consideration, consistency, asks whether the pattern of observations across species/strain/ organs/test systems is what would be expected based on the hypothesized MOA (Meek et al., 2014). In other MOA frameworks (eg, USEPA, 2005;Sonich-Mullin et al., 2001) consistency has typically referred to the reproducibility of results across studies. Regarding the former, the tumor triad present in rats chronically exposed to HFPO-DA is similar to those observed with other PPARa activators. Regarding the latter definitions of consistency, multiple in vivo studies have demonstrated PPARa activation at the transcriptomic level (Chappell et al., 2020;Guo et al., 2022;Heintz et al., 2022;Wang et al., 2017), and multiple studies have reported similar histopathology (see Thompson et al., forthcoming).
The final evolved Bradford Hill consideration, analogy, questions whether the MOA would be anticipated based on broader chemical specific knowledge or information on chemically similar substances . Indeed, legacy PFAS such as PFOA are known to activate PPARa along with mitigation of many effects in PPARa-null mice. As such, one would anticipate that many PFAS might have activity toward PPARa. The data herein clearly demonstrate that HFPO-DA shares PPARa activation with many other PFAS.
Regarding temporal concordance, there is molecular evidence for PPARa signaling at 28 days (the earliest timepoint with transcriptomic data yet investigated; Table 2). There is also evidence for hepatocellular hypertrophy, hepatomegaly, and mitosis at 28 days of exposure. One potential inconsistency in Table 2 is evidence for single cell necrosis/apoptosis (see discussion in Cytotoxicity section below) at 3 mg/kg HFPO-DA without reported evidence of mitotic figures at that dose at day 28. However, enrichment of cell cycle and DNA replication pathways was reported in mice exposed to various PFAS including HFPO-DA (2 mg/kg) for 28 days.

Alternative MOAs for HFPO-DA-mediated liver effects in mice
The data show strong support for HFPO-DA functioning through the PPARa MOA in the mouse liver as described above. However, alternative MOAs have been hypothesized (USEPA, 2021), including cytotoxicity, mitochondrial dysfunction, and involvement of other PPARa subtypes. Each of these is addressed below.

Cytotoxicity
Based on the evidence of PPARa activation and hepatomegaly observed in subchronic mouse studies, it is likely that mice would develop liver tumors under chronic HFPO-DA exposure scenarios. In a 2016 Toxicology Forum workshop on the human relevance of rodent liver tumors, 3 nongenotoxic MOAs (PPARa, constitutive androstane receptor [CAR], cytotoxicity) were discussed for their relevance for human health risk assessment (Felter et al., 2018). The consensus, although not unanimous, was that PPARa-and CAR-mediated liver tumors were not relevant to humans, whereas cytotoxicity and regenerative hyperplasia were considered relevant to humans (Felter et al., 2018). Among the six criteria listed for establishing a cytotoxic MOA were "clear evidence of cytotoxicity by histopathology, such as presence of necrosis and/ or increased apoptosis", "evidence of toxicity by increased serum enzymes that are relevant to humans", and "presence of increased cell proliferation as evidenced by increased labeling index and/or increased number of hepatocytes" (Felter et al., 2018). A fourth criteria, "the chemical is not DNA reactive", explicitly considers evidence for an alternative MOA (ie, genotoxicity). The latter criterion, "not DNA reactive" comes from a data stream that has little in common with the other criteria that are markers of cytotoxicity. Clearly, other data streams are used for informing a cytotoxic MOA, and we believe that evidence streams supporting PPARa (as described above) should similarly factor into cytotoxic MOA determinations. As such, any toxicity that occurs as a result of PPARa mediated effects should not be interpreted as evidence for a cytotoxic MOA.
Regarding necrosis, the NTP Nonneoplastic Lesion Atlas states that various forms of necrosis (centrilobular, coagulation, focal, etc.) should not be subclassified with the exception of single cell necrosis. As such, it is unclear if single cell necrosis truly represents "necrosis" that is indicative of a cytotoxic MOA. As already discussed, there is evidence that some of the hepatocytes that might be considered necrotic stain positive for activated caspase-3 (Thompson et al., forthcoming). Notably, the only other form of necrosis that has been diagnosed in liver from HFPO-DA exposed mice is focal necrosis, which was not significantly elevated in exposed mice. Focal necrosis can occur in enlarged livers due to compression against the capsule or adjacent organs, resulting in focal hypoxia and cell death because the blood supply is already limited just below the capsule (Thoolen et al., 2010). As discussed above, Felter et al. (2018) also include apoptosis and increased cell proliferation as evidence of a cytotoxic MOA; however, we have already discussed how these are known to occur in the PPARa MOA.
Recently, a gene expression signature indicative of liver cytotoxicity has been developed from short-term rat toxicity studies (Glaab et al., 2021). Using published transcriptomic data (Chappell et al., 2020;Heintz et al., 2022), the expression of the 10 genes in the cytotoxicity gene set (CGS) were assessed in livers of mice exposed to HFPO-DA. Compared to PPARa-regulated genes involved in lipid transport and metabolism, which were largely induced at dose levels as low as 0.1 and 0.5 mg/kg HFPO-DA, these 10 genes associated/predictive of hepatic cytotoxicity were not significantly differentially expressed in either study, dose group or sex, with the exception of 4 genes (Anxa2, Gpnmb, Timp1, and Tnfrsf12a) in the highest dose group (5 mg/kg HFPO-DA) in parental male mice from the reproductive/developmental toxicity study ( Figure 1C). That these 4 genes were elevated after the activation of PPARa pathways suggest they might be elevated as a result of PPARa activation. Moreover, the consistent evidence for increased PPARa-related gene expression in the absence of any change in the CGS in the female mice from the reproductive/ developmental toxicity and male and female mice from the 90day study support a PPARa MOA for the liver changes. The role(s) of each of these genes in cell physiology and toxicity are not fully known, and some of the gene changes might be downstream of PPARa activation. For example, glycoprotein nonmetastatic melanoma protein B (Gpnmb) appears to be a hepatokine that is secreted by the liver to promote lipogenesis in white adipose tissue (Gong et al., 2019) and therefore might be related to PPARainduced changes in lipid metabolism. Tumor necrosis factor receptor (Tnfrsf12a) codes for Fn14 that binds to tumor necrosis factor-like weak inducer of apoptosis (TWEAK) (Ratajczak et al., 2022), which is potentially related to increased apoptosis present at 5 mg/kg. However, apoptosis was also observed at 5 mg/kg in other mice exposed to HFPO-DA that did not exhibit an increase in this gene ( Figure 1C). Increased Anxa2 expression has been observed in wild type but not humanized-PPARa mice treated with the PPARa-specific ligand WY-14643 (Yang et al., 2008), suggesting species-specific PPARa-mediated gene expression of Anxa2. In addition, both Timp1 and Anxa2 may have roles in fibrosis (Wang et al., 2022a;Zisser et al., 2021); however, no histopathological evidence of fibrosis has been observed in tissue sections from the same mice the transcriptomic analyses were conducted. Overall, there is little/no transcriptomic evidence that HFPO-DA is acting via a cytotoxic MOA. Moreover, among various forms of programmed cell death (necroptosis, ferroptosis, pyroptosis, autophagy, and apoptosis), only apoptosis and autophagy gene sets were enriched in these transcriptomic studies (see Table 4 in Thompson et al., forthcoming)-both of which have been linked to PPARa (Byun et al., 2020;Corton et al., 2014;Xiao et al., 2006). In addition, mice exposed to 5 mg/kg HFPO-DA exhibit increased serum liver enzymes including alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP), and total bile acids. These changes in enzyme levels have been suggested to be indicative of a cytotoxic MOA (USEPA, 2021) based, in part, on criteria described in Hall et al., (2012). Importantly in their 2012 review, Hall and colleagues concluded that rodent liver hypertrophy might be considered nonadverse for human health risk assessment when there is mechanistic evidence for nuclear receptor activation (eg, CAR, PXR, and PPARa) in the absence of other histological effects considered adverse such as necrosis, steatosis, fibrosis and cholestasis or large increases in serum liver enzymes such as ALT, AST, ALP, bile acids, and others. Serum liver enzyme levels in HFPO-DA exposed mice from the highest dose group (5 mg/kg) were increased greater than 2-fold compared to controls, exceeding what is considered nonadverse metabolic induction (Hall et al., 2012), and suggesting potential for cytotoxicity. However, hepatic PPARa and the farnesoid X receptor (FXR) are key regulators of bile acid homeostasis, and PPARa activators and PFAS have been found to disrupt bile acid homeostasis in mice by altering gene expression of hepatic bile acid transporters and concurrently increasing serum liver enzyme levels in a PPARa-dependent manner (Cheng and Klaassen 2008;Gyamfi and Wan 2009;Liu et al., 2014;Xie et al., 2019). Perturbation of hepatic bile acid transporter gene expression, including the downregulation of the basolateral transporters, Slc10a (Na þ -dependent bile acid transporter, Ntcp) and Slco1a (organic anion transporting polypeptide 1a1, Oatp1a1), also occurred in mice exposed to HFPO-DA ( Figure 1C) (Chappell et al., 2020;Guo et al., 2022;Heintz et al., 2022). It is conceivable that these PPARa-mediated changes in bile acids might have contributed to the gene changes within the CGS described above . As a result, these effects are most likely part of a PPARa-mediated response in rodents and not indicative of non-PPARa mechanism that might be considered adverse in the context of human health risk assessment. These findings suggest that the criteria for assessing human relevant adverse effects in Hall et al. (2012) requires revisiting.
In addition to bile acid-associated increases in serum liver enzymes, it has also been demonstrated by partial hepatectomyinduced liver regeneration in rats, that liver cell proliferation also increases serum levels of liver enzymes, and these elevated serum enzyme levels are unrelated to hepatocellular necrosis and mitochondrial dysfunction (Contreras-Zentella and Hern andez-Muñoz 2016;D ıaz-Ju arez et al., 2006). Thus, liver enzyme release in mice treated with HFPO-DA may occur as a result of PPARa-related increases in enzyme expression and/or hepatic proliferation and not only as a consequence of necrotic cell death.
Toxicity studies for other PFAS, including PFNA and PFDA, observed increased serum liver enzymes in wild-type but not PPARa-null mice (Luo et al., 2017;Zhang et al., 2018). Administration of PPARa-specific ligands such as WY-14643, pemafibrate or fenofibrate has also been shown to increase serum liver enzymes in wild-type mice (Araki et al., 2018;Gyamfi and Wan 2009). Therefore, the observed increase in ALT, AST, and ALP in mice exposed to HFPO-DA is likely PPARa-dependent.

Other PPAR subtypes
Findings from 2 in vitro studies have been suggested to indicate a role for PPARc in the MOA for HFPO-DA (USEPA, 2021). Li et al., (2019) measured the transcriptional activation of mouse PPARc by HFPO-DA in a reporter gene assay and determined that HFPO-DA weakly activated mouse PPARc, with only a 1.2-fold increase in activity at the highest concentration tested. These authors also examined the binding affinity of HFPO-DA to the PPARc ligand binding domain (LBD) and showed that HFPO-DA had little to no binding affinity to the mouse PPARc LBD, with the IC 50 value for HFPO-DA being indeterminate due to the minimal binding in the concentration range tested (up to 1 mM). Additionally, HFPO-DA caused minimal changes in PPARc genes involved in adipogenesis in mouse adipocytes (3T3-L1 cells) . Evans et al. (2022) measured the transcriptional activation of rat PPARc by HFPO-DA in a reporter gene assay and determined the lowest observed effect concentration (LOEC) producing a 2-fold increase over vehicle control for HFPO-DA and the ammonium salt of HFPO-DA was equal to 300 mM, the second highest concentration tested. It should be appreciated that the N-terminal binding domain of the PPARc in this assay is substituted with the N-terminal binding domain of yeast GAL4, and that the luciferase is not being activated by a PPAR response element but rather the GAL4 upstream activator sequence (INDIGO Biosciences, State College, Pennsylvania). As such, these reporter gene assays should be interpreted with caution.
In vivo evidence for other PPAR subtypes is also limited and specific to the PPARc subtype. Conley et al., (2019;2021) reported upregulation of several genes related to glucose metabolism in fetal and neonatal mouse livers exposed to HFPO-DA in utero (eg, Pck1, Gk, and Aqp7), that are associated with PPARc signaling in adipose tissue (Rodr ıguez et al., 2006); however, these gluconeogenesis genes are also regulated by PPARa in the liver (Kersten, 2014;Peeters and Baes, 2010). Therefore, it is likely that these genes were induced by PPARa rather than PPARc based on increased expression of numerous other PPARa-regulated genes in maternal, fetal and neonatal livers (Conley et al., 2019(Conley et al., , 2021. As described in the MOA analysis above (see Supplementary  Figure 1; Supplementary Table 1), HFPO-DA has little to no effect on the regulation of downstream genes or pathways associated with PPARc (Chappell et al., 2020;Heintz et al., 2022). It should also be appreciated that there is very little expression of PPARc in the rodent liver (Corrales et al., 2018). As such, it is highly unlikely that PPARc plays an important role in the MOA for HFPO-DAinduced liver changes in the mouse liver.
In addition to PPARc, gene expression and pathway enrichment analysis of hepatic transcriptomic data from 90-day subchronic and reproductive/developmental toxicity studies for HFPO-DA in mice indicate little to no evidence for PPARd activation (Supplementary Figure 1) (Chappell et al., 2020;Heintz et al., 2022). Although gene sets specific to PPARd are not currently available within the canonical pathway subcollection used for gene set enrichment analyses in Chappell et al., (2020) and Heintz et al., (2022), individual genes associated with PPARd signaling were investigated (Supplementary Figure 1). In addition, PPARd and PPARc are predominately expressed in muscle (Holst et al., 2003) and adipose tissues (Chawla et al., 1994), respectively, whereas PPARa is predominantly expressed in the liver (Corrales et al., 2018). Overall, these data demonstrate that the weight of evidence for involvement of other PPAR subtypes in the MOA for HFPO-DA is poorly supported.
In contrast to HFPO-DA, studies on other PFAS examining the activation of PPARa compared to other nuclear receptors have reported mixed findings, with both PPARa-dependent and -independent effects. For example, Rosen et al., (2017) reported upregulation of genes related to the activation of PPARc and CAR, as well as PPARa in the livers of mice exposed to PFHxS or PFNA. These study authors also compared mouse liver gene expression data for PFOA and PFOS and concluded that over $75% of all genes regulated by these PFAS in wild-type mice are in fact PPARa-dependent (Rosen et al., 2017). However, the activation of other nuclear receptors besides PPARa, and the extent to which they are activated, appears to be specific to individual PFAS and the model species being tested. As described in previous sections for rodents and in the sections below for humans, data for HFPO-DA clearly indicate that the compound is a PPARa-specific agonist.

Mitochondrial dysfunction
Mitochondrial dysfunction as an alternative MOA for HFPO-DAinduced liver changes has been proposed (USEPA, 2021) based on reports of increased numbers of mitochondria in livers of mice exposed to HFPO-DA (Blake et al., 2020) and increased hepatic expression of genes associated with mitochondrial b-oxidation in livers of mice and rats exposed to HFPO-DA (Chappell et al., 2020;Conley et al., 2021;Conley et al., 2019;Heintz et al., 2022). However, both peroxisomes and mitochondria play a critical role in lipid catabolism via b-oxidation of fatty acids, with each organelle metabolizing long-chain fatty acids or very long-chain fatty acids, respectively (Demarquoy and Le Borgne 2015;Tahri-Joutey et al., 2021). Furthermore, some genes involved in mitochondrial b-oxidation are regulated by PPARa activators (Kersten and Stienstra 2017;Tahri-Joutey et al., 2021). Aoyama et al., (1998) showed that PPARa modulates the expression of genes involved in mitochondrial b-oxidation, as both peroxisomal and mitochondrial enzymes were induced following treatment with WY-14643 in wild type but not PPARa-null mice . Similar findings have also been observed in mice treated with other PPARa agonists such as ciprofibrate (Cook et al., 2000). In addition to PPARa, it is recognized the activity and abundance (ie, biogenesis) of peroxisomes and mitochondria are co-regulated in a PPARa-and PPARc coactivator 1-alpha (PGC1a)-dependent manner (Fransen et al., 2017). Transcriptomic analyses of livers from HFPO-DA-exposed mice indicate induction of both mitochondrial and peroxisomal fatty acid metabolism at similarly low median BMDs (ie, between 0.2-0.3 mg/kg for male mice and 0.5-0.9 mg/kg for female mice), and enrichment of gene sets related to mitochondrial biogenesis generally at higher BMDs (ie, median BMDs between 1.5 and 2.5 mg/kg; Figure 1B) (Chappell et al., 2020;Heintz et al., 2022). These data support PPARa's role in maintaining systemic and cellular energy homeostasis by modulating the expression of genes involved in fatty acid b-oxidation and biogenesis for both peroxisomes and mitochondria Fransen et al., 2017).

Data gaps
A current data gap in the PPARa MOA is the absence of any in vivo studies examining the effects of HFPO-DA in PPARa-null mice. As indicated elsewhere in this article, the KEs described herein for HFPO-DA are often not observed when PPARa-null mice are exposed to classic PPARa activators such as WY or fibrates. We are currently in the process of conducting short-term assays in wild type and PPARa-null mice exposed to HFPO-DA, using exposure paradigms similar to those proposed for screening PFAS compounds (Gwinn et al., 2020). Similarly, in vitro assays in wild type and PPARa-null mouse hepatocytes cells as well as human hepatocytes exposed to HFPO-DA can inform both the MOA and human relevance.

Human relevance
It is widely accepted that rodent liver tumors resulting from exposure to PPARa activators are not relevant to humans . While PPARa is expressed in many species and plays a role in lipid metabolism across species, the downstream cell proliferation signaling occurs specifically in rodents. Increased cell proliferation is a key and required event in the formation of rodent-specific hepatic tumors (Corton et al., 2014). Yet, a remaining critical question is whether the non-neoplastic changes in the liver that occur during KE 2 and KE 3, as seen with PPARa activators like HFPO-DA, are unique to rodents.
The human relevance of the first 3 KEs underlying the PPARa MOA is summarized in Table 3. Experimental data for HFPO-DA exposure using human in vitro models are only available for KE 1; however, data are available for other PPARa activators that address the human relevance of KE 2 and KE 3. As shown in Table 3, only KE 1, PPARa activation, is shared across humans and rodents. Furthermore, within KE 1, only receptor activation and induction of lipid metabolism genes occur in both species. Nielsen et al. (2022) determined that HFPO-DA acted as full human PPARa agonist using a full length human PPARa construct in a transactivation assay, with a potency (50% effective concentration) equal to 2.1 mM and an efficacy (maximum PPARa activity) of 134% compared to positive control levels. Evans et al. (2022) and Behr et al. (2020) observed similar results for HFPO-DA compared to other PFAS, as HFPO-DA had the greatest potency for human PPARa activation using a reporter assay construct consisting of human PPARa ligand binding domain fused with a Gal4 DNA-binding domain. Activation of PPARa by other PPARa-specific ligands and hyperlipidemic agents was also investigated by Evans et al. (2022), who reported similar 20% effective concentration (EC 20 ) and area under the curve (AUC) values for HFPO-DA and clofibric acid (the metabolically active form of clofibrate) (Evans et al., 2022). Consequently, at the same internal dose, clofibric acid and HFPO-DA may be expected to generate a similar level of PPARa activation in humans. However, the human effective clinical dose of fibrates, for example, TRICOR fenofibrate, is 0.69-2.1 mg/kg-day (assuming 70 kg adult) (USFDA, 2018), whereas RfD values for HFPO-DA are 0.01 mg/kg-day (Thompson et al., 2019;USEPA, 2021).
In vitro studies in HepG2 or primary human hepatocytes treated with HFPO-DA or PPARa-specific ligands (ie, WY-14643 and GW7647), respectively, found induction of gene expression related to peroxisomal and mitochondrial fatty acid b-oxidation, lipid transport, and lipoprotein metabolism (Behr et al., 2020;McMullen et al., 2020;Rakhshandehroo et al., 2009). The upregulation of hepatic lipid metabolism functional categories in primary hepatocytes was well-conserved between humans and mice (Rakhshandehroo et al., 2009) or rats (McMullen et al., 2020). However, at similar treatment concentrations, these conserved gene expression changes between species occurred at lower expression levels in humans compared to rodents, resulting in an overall lower transcriptional response to PPARa activation in humans (Bjork et al., 2011;McMullen et al., 2020;Rakhshandehroo et al., 2009). In addition, induction of hepatic peroxisome proliferation by PPARa activators is not conserved across species. Bentley et al. (1993) reviewed the significance of hepatic peroxisome proliferation in humans and concluded that the available data from hypolipidemic patient biopsies demonstrated an overall absence of increased peroxisome proliferation in the livers of patients treated with hypolipidemic drugs (eg, fenofibrate and gemfibrozil). Goll et al., (1999) also confirmed a lack of peroxisome proliferation-associated parameters in human hepatocyte cultures treated with various hypolipidemic agents (Table 3). Evidence for KE 2, alteration of cell growth pathways, is absent in primary human hepatocytes treated with PPARa-specific ligands or hypolipidemic drugs, with no indication of increased DNA synthesis (Elcombe et al., 1996;Goll et al., 1999;Perrone et al., 1998), enrichment of cellular pathways related to cell proliferation, or induction of pro-apoptotic gene expression (Table 3) (McMullen et al., 2020;McMullen et al., 2014). Based on hepatic gene expression results in vitro for PPARa-specific ligand WY-14643, Rakhshandehroo et al., (2009) concluded that "PPARa regulates a mostly (ie, besides lipid modulating effects) divergent set of genes in mouse and human liver". In addition, the human data available for KE 3, perturbation of cell growth and survival, show that liver size and serum liver enzyme levels (a putative marker of hepatocellular proliferation and PPARa-mediated bile acid perturbation in rodents), do not increase in patients treated with hypolipidemic drugs (Table 3) (Black et al., 2014;Gariot et al., 1987;Liss and Finck 2017). Therefore, the human evidence base for HFPO-DA and other potent PPARa agonists demonstrates that the non-neoplastic changes (ie, KE 2 and 3) observed in the livers of mice treated with HFPO-DA do not occur in humans.

Discussion
The weight of evidence from mechanistic and phenotypic data described herein strongly supports that the liver changes observed in mice exposed to HFPO-DA are occurring via a PPARa MOA. The central role of PPARa in the observed effects in mouse livers is summarized in Figure 1D. The relevance of PPARa-related liver toxicity in human health risk assessment has been a topic of great interest. In a review on the adversity of liver hypertrophy, experts concluded that "hepatomegaly as a consequence of hepatocellular hypertrophy without histologic or clinical pathology alterations indicative of liver toxicity was considered an adaptive and a nonadverse reaction" (Hall et al., 2012). Therein, evidence for PPARa and CAR activation in the absence of hepatoxicity was considered nonadverse. Being somewhat dated, the language on liver necrosis was somewhat vague and it was not considered that some changes in serum liver enzymes could be a direct consequence of PPARa-mediated gene changes as opposed to cytotoxicity. Nevertheless, there was a general view that PPARa and CAR mediated liver changes were "rodent-specific phenomenon." The USEPA (2021) risk assessment of HFPO-DA indicates that a PPARa MOA would potentially have limited relevance for humans: "The increases in relative liver weight, hepatocellular hypertrophy, and peroxisome activity (eg, peroxisomal beta-oxidation induction) can be associated with activation of cellular peroxisome proliferator-activated receptor alpha (PPARa) receptors, making it difficult to determine if this change is a reflection of PPARa activation or an indication of GenX chemical toxicity. This is important because the PPARa response could be more relevant to rodents than humans." Rather than directly addressing human relevance, the USEPA (2021) instead hypothesized several alternative MOAs for HFPO-DA-induced liver effects. For the reference dose (RfD) derivation, the USEPA modeled the combined incidence of several different liver lesions (hepatocellular hypertrophy/cytoplasmic alteration, single cell necrosis, apoptosis, and focal necrosis). Available scientific evidence strongly supports that these liver lesions in mice occur as part of the PPARa MOA and therefore are not relevant to humans. In contrast, the evidence for alternative MOAs for HFPO-DA-induced liver toxicity is not supported by the scientific literature. As such, these liver endpoints should not be used as the point of departure (POD) in developing a RfD for HFPO-DA. Additionally, given that data in the scientific literature demonstrate that other PFAS are PPARa agonists (Evans et al., 2022;Nielsen et al., 2022), candidate toxicity values or relative potency factors (RPFs) for other PFAS derived from liver endpoints in rodents should also be evaluated in the context of this MOA.
Additionally, in developing the proposed RfD, the USEPA applied a 3-fold interspecies uncertainty factor after accounting for potential interspecies pharmacokinetic differences via allometric scaling. This 3-fold adjustment is not necessary, as the considerable evidence that the liver lesions have limited human relevance indicate that humans are unlikely to be more sensitive to HFPO-DA than mice. In addition to a 10-fold uncertainty factor for database uncertainty and a 10-fold factor for human variability, USEPA also applied a full 10-fold uncertainty factor for use of a subchronic study. In totality, USEPA's maximum 3000-fold uncertainty factor was applied to a rodent specific endpoint, resulting in an RfD of 0.000003 mg/kg-day, one of the lowest RfD values in the IRIS database.
A series of recent papers provides a means of ground-truthing the RfD developed for HFPO-DA. Although the diversity of PFAS warrant chemical specific assessments, a recent article demonstrated the potential application of the concept of threshold for toxicological concern (TTC) to PFAS (Lea et al., 2022). Therein, data for 27 PFAS were classified via ToxTree (Patlewicz et al., 2008) as Cramer Class III structures (Cramer Class III chemical structures contain "elements other than carbon, hydrogen, oxygen, nitrogen or divalent sulfur" with structural features that "permit no strong initial presumption of safety or may even suggest significant toxicity" (Cramer et al., 1978), which have a current fifth percentile no-observed-adverse-effect level (NOAEL) value of 1.5 mg/kg/day. The expansion of the chemical space via the addition of PFAS resulted in a fifth percentile NOAEL value of 1.3 mg/kg-day (0.0013 mg/kg-day). An earlier evaluation demonstrated that, on average, Cramer Class III TTC values are $6-fold lower than corresponding IRIS values for the same chemical (Pham et al., 2020), indicating that TTC values, which are generally used in data-poor situations, are more conservative than detailed risk assessment toxicity values (eg, RfDs). For HFPO-DA, the TTC value is 433-fold higher than the RfD. As noted in Pham et al. (2020), a cursory examination of chemicals where the TTC value is significantly higher than corresponding RfD value indicated potential need for reevaluation of the RfD due to outdated methodology or overapplication/compounding of uncertainty factors. The fact that the RfD for HFPO-DA is significantly lower than the Cramer Class III TTC provides an additional line of evidence indicating that the RfD for HFPO-DA is overly conservative. As indicated above, the application of a maximum 3000-fold uncertainty factor to a rodent specific endpoint warrants reexamination.
In conclusion, the current weight of evidence indicates that the HFPO-DA induced effects in the mouse liver are the result of a PPARa MOA. These effects are widely considered to have limited relevance to humans in the context of tumor formation Felter et al., 2018). Hepatomegaly, and early precursor in the development of rodent liver tumors, is considered nonadverse when there is evidence for PPARa or CAR activation (Hall et al., 2012). Data indicate that some effects traditionally considered as markers of hepatotoxicity are PPARa mediated and not necessarily indicative of liver toxicity. As such, the liver changes in mice following exposure to HFPO-DA are not relevant for human health risk assessment.

Supplementary data
Supplementary data are available at Toxicological Sciences online.

Declaration of conflicting interests
The authors declared the following potential perceived conflicts of interest with respect to the research, authorship, and/or publication of this article: The authors include employees of ToxStrategies, LLC, a private consulting firm that provides services to private and public organizations for toxicology, epidemiology, and risk assessment issues. J.C. and J.K. served as independent consultants. The work reported in this article was conducted during the normal course of employment. The authors (C.T. and L.H.) have presented study findings in meetings with regulators, including public meetings, on behalf of The Chemours Company FC, LLC.

Funding
This work was supported by The Chemours Company FC, LLC. Chemours was given the opportunity to review the draft manuscript. The purpose of this review was for the authors to receive input on the clarity of the science presented but not on the interpretation of research results. The authors' scientific conclusions and professional judgments were not subject to the funder's control; the contents of this manuscript solely reflect the view of the authors.