Octamethylcyclotetrasiloxane (D4) lacks endocrine disruptive potential via estrogen pathways

Abstract

Octamethylcyclotetrasiloxane (D4) is a volatile, highly lipophilic monomer used to produce silicone polymers found in many consumer products and used widely in industrial applications and processes. Many reviews of the toxicology of D4 conclude that its adverse effects on endocrine-sensitive endpoints occur by a MoA dependent on systemic toxicity rather than one mediated via endocrine activity, but others identify D4 as an estrogenic endocrine disruptive chemical (EDC) based on results of screening-level assays indicating that D4 interacts with ERα and at high doses, affects estrogen-sensitive endpoints in rodents. To resolve these divergent interpretations, we tested two specific hypotheses related to the interaction of D4 with estrogen receptor–alpha subtype (ERα) at the biochemical and molecular levels of biological organization and a third specific hypothesis related to estrogenic and anti-estrogenic pathways at the physiological level. At the physiological level, we used an established WoE methodology to evaluate all data relevant to estrogen agonist and antagonist activity of D4 by examining its effects on ERα-relevant endpoints in rodent toxicology studies. At the biochemical level, we calculated whether D4 could produce a functionally significant change in the ERα occupancy by 17β-estradiol (E2) using equations well-established in pharmacology. For these calculations, we used data on the potency and kinetics of D4 from studies in rats as well as published potency and affinity data on endogenous estrogens and their circulating concentrations in humans. At the molecular level, we used established molecular docking techniques to evaluate the potential for D4 and related chemicals to fit within and to activate or block the binding pocket of ERα. Our analyses indicate that the estrogenic effect of D4 is molecularly, biochemically, and physiologically implausible, which corroborates previous evaluations of D4 that concluded it is not an estrogenic endocrine disruptor. The claim that D4 exhibits estrogenic endocrine disruptive properties based on a presumed link between the results of screening-level assays (RUA and ERTA) and adverse effects is not supported by the data and relies on deficient evaluative and interpretative methods. Instead, a plausible mechanistic explanation for the various adverse effects of D4 observed in rodent studies, including its effects in reproduction studies, is that these are secondary to high-dose-dependent, physico-chemical effects that perturb cell membrane function and produce rodent-specific sensory irritation.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00204-024-03896-y.

Introduction

Octamethylcyclotetrasiloxane (D4) is a volatile, highly lipophilic monomer used to produce silicone polymers found in many consumer products and used widely in industrial applications and processes. Detailed reviews of the toxicology of D4, its kinetic behavior and potential modes of action have been published (Campbell, 2017; Dekant et al. 2017a, 2017b; Franzen et al. 2017; Gentry et al. 2017; Jean and Plotzke 2017; Jean et al. 2017) and recent mechanistic evaluations concluded that D4 should not be classified as an EDC (endocrine disruptive chemical) or a CMR (carcinogen, mutagen or reproductive toxicant) (Andersen 2022; Matthews 2021). The toxicological effects of D4 appear to be due to high-dose-dependent mechanisms, including physico-chemical interactions with neural membranes and sensory irritant responses that lack relevance for human risk assessment, and which may occur secondary to kinetic overload (Meeks et al., 2024; Andersen 2022; Borgert et al. 2021; Dekant et al. 2017a, b; Matthews 2021; Pauluhn 2021; Slikker et al. 2004). The latter hypothesis has been tested (Borgert et al. 2024a) using kinetic data probed for the existence of a kinetic maximum dose (KMD). These analyses indicate that the effects of D4 do not satisfy the definition of an endocrine disruptive chemical (EDC) because the definition requires a causal link between an endocrine MoA and an adverse effect in an intact organism, its progeny or subpopulations (WHO/IPCS, 2002). In contrast, the effects of D4 on endocrine-sensitive endpoints occur secondary to kinetic overload at high doses above its KMD (Borgert et al. 2018; 2024a).

Notwithstanding those analyses, some agencies have misinterpreted the results of screening-level assays for estrogenic activity as indicative of a link between D4’s interaction with the alpha subtype of the estrogen receptor (ERα) and the effects of D4 on estrogen-sensitive endpoints in rodents. Endocrine MoAs other than via ERα are not supported and are not controversial (Andersen 2022; Borgert et al. 2018; He et al. 2003; Lee et al. 2015; Matthews 2021; McKim et al. 2001; Quinn et al. 2007a). Based on the misinterpretation of estrogenic screening assays, some claim that D4 is an estrogenic EDC (Christiansen et al. 2022). The misinterpretation appears to arise from the presumption that the results of mechanistic, screening level assays can be interpreted as either “positive” or “negative” whereby all statistically significant results—so-called “positives”—provide evidence of an endocrine mode of action (MoA) for observed adverse effects. By conceptual and empirical analyses, we (Borgert et al. 2013; 2018, 2024b) and others (Pande et al. 2019) have shown it is invalid to assume that “positive results” in screening-level mechanistic assays portend physiological effects via an endocrine MoA. Alone, screening level assays merely probe the ability of a chemical to interact with components of the endocrine system. Without a consideration of the mechanistic potency of a chemical in those screening level assays and in the context of endogenous physiological environment, such assays do not provide interpretable information as to whether an interaction is physiologically relevant in an organism with a functionally intact endocrine system, or whether that interaction is likely to cause adverse effects (Borgert et al. 2011; 2014; 2024b; Marty et al. 2011; 2018).

Thus, to satisfy the definition of an EDC (WHO/IPCS, 2002), it is necessary, but insufficient, to consider the results of screening-level assays that probe potential mechanistic interactions with the endocrine system and the results of in vivo toxicology studies that identify adverse effects. It is also insufficient and incorrect to arbitrarily assume that results of screening-level assays portend physiological effects via an endocrine MoA. To establish a biologically plausible causal connection between a putative endocrine mode of action and an adverse effect, a minimum of two requirements must be met. The strength of the interaction of the chemical with the endocrine system must be shown to be sufficient to elicit effects under physiologically relevant conditions, i.e., the interaction must exceed a physiologically relevant mechanistic threshold (Borgert et al. 2013; 2018, 2024b; Pande et al. 2019; Zhang et al. 2014). In addition, the pattern of effects produced by the chemical under physiologically relevant conditions should be consistent with those of known effectors of the specific endocrine pathway, i.e., the effects of the chemical must resemble those of a known positive control (Borgert et al. 2011; 2014; Marty et al. 2011; 2018).

To resolve the remaining controversy regarding whether D4 can produce effects in humans via a MoA involving ERα, either estrogenic or anti-estrogenic, we evaluated three specific hypotheses:

1. The effects of D4 on estrogen-sensitive endpoints in rodent studies resemble those of known estrogen-receptor agonists or antagonists. Consistency of D4’s effects on estrogen-sensitive endpoints with patterns of effects expected of ERα agonists and antagonists would support the contention that D4 may have an (anti)estrogenic MoA in humans, whereas an inconsistency between D4’s effects on estrogen-sensitive endpoints with patterns of effects expected of ERα agonists and antagonists would contradict it.

2. D4 can occupy a fraction of estrogen receptors in humans sufficient to elicit a functionally significant change in ERα-mediated estrogenic pathways (agonist or antagonist). Borgert et al. (2024b) recently demonstrated a biochemical and physiological basis for potency thresholds based on classical receptor and enzyme kinetics. Accordingly, the ability of a chemical to interact via a particular molecular MoA, such as receptor agonism/antagonism or enzyme inhibition, can be evaluated through fractional occupancy calculations that include a consideration of the concentrations and affinities of endogenous ligands or substrates. This enables potency information derived from screening-level in vitro and in vivo endocrine assays to be interpreted within a relevant physiological context, and thus, to be used in evaluating whether a chemical can produce effects via a particular endocrine MoA. The ability of D4 to occupy a functionally significant fraction of ERα would support the contention that D4 may have an (anti)estrogenic MoA in humans, whereas the inability to occupy a functionally significant fraction of ERα would refute that claim.

3. D4 or its metabolites will fit the binding pocket of ERα with a sufficient degree of specificity to activate the receptor or block access to it. The ability of D4 or its metabolites to fit the binding pocket of ERα would support the contention that D4 may have an (anti)estrogenic MoA in humans, whereas an inability to fit the binding pocket of ERα would refute it. This provides an independent check of Hypothesis 2, since a molecule must have a close molecular fit to a receptor binding site in order for it to be a strong receptor ligand with high affinity for the receptor (reviewed in Borgert et al. 2024b).

Hypotheses 1 and 2 have been partially addressed in previous publications (Andersen 2022; Borgert et al. 2018; 2024b; Matthews 2021). Borgert et al. (2018) showed that D4’s potency via ERα is below a clinically observable potency threshold for ERα agonism in humans, and the results of Borgert et al. (2024b) suggest that D4 is insufficiently potent via ERα to be an effective agonist or antagonist of this receptor pathway. Hypothesis 3 was suggested as an as-yet-untested possibility in a recent review of D4’s potential to exhibit endocrine activity (Matthews 2021). If D4 or its metabolites show a structurally specific fit to the binding pocket of ERα, this could suggest affinity sufficient to occupy the receptor. If a metabolite shows a better fit than D4, this could provide support for the hypothesis that D4 acts via an estrogenic MoA indirectly, and would suggest that this MoA may be influenced by kinetics. Molecular docking analysis should also serve as an independent test of the receptor occupancy calculations to evaluate Hypothesis 2. Threshold network concepts are explained in Zhang et al. (2014) and background on pharmacological principles, mechanistic potency, and potency thresholds is found in Borgert et al. (2013; 2018; 2024b) and in Matthews (2021).

To more fully and rigorously address the three hypotheses, we examined aspects of the interaction of D4 with estrogenic pathways at three levels of biological organization. At the physiological level (Hypothesis 1), we applied a well-established weight-of-evidence (WoE) framework (Borgert et al. 2011; 2014) to evaluate whether D4 elicits effects in endocrine-sensitive screening assays and apical tests consistent with an anti(estrogenic) MoA. At the biochemical level, (Hypothesis 2), we evaluated whether the mechanistic potency of D4 via ERα is sufficient to produce a physiologically-significant change in ERα receptor occupancy in the presence of normal physiological levels of endogenous ERα ligands. At the molecular level (Hypothesis 3), we conducted receptor docking analyses for D4 and its metabolites heptamethylcyclotetrasiloxanol, dimethylsilanediol (DMSD), and DMSD-dimer diol (hydroxy-[hydroxy(dimethyl)silyl]oxy-dimethylsilane) to determine whether these structures fit the ligand binding pocket of ERα sufficiently well to activate it or to block access to it by endogenous estrogens. Since molecular docking analyses provide an independent test of Hypothesis 2 (receptor occupancy), and since Hypothesis 1 compares the pattern of D4’s effects on estrogen-responsive endpoints with those expected of known ERα agonists that fit the binding pocket of the receptor with sufficient affinity to achieve high receptor occupancy at physiological concentrations, a consistency and coherence of findings across these three levels of biological organization would either support or refute the hypothesis that D4 has the potential to act as an anti(estrogenic) endocrine disruptor, whereas an inconsistency or incoherence of findings could leave the question unresolved.

Methods

Hypothesis 1: weight-of-evidence evaluation

The methodology used for this WoE analysis is based on a broadly applicable, general WoE framework (Borgert et al. 2011). Briefly, the approach defines specific hypotheses related to interactions with specific MoAs—in this case, estrogen and anti-estrogen pathways – and weights the importance of endpoints as Rank 1, 2, or 3 according to their mechanistic relevance for each hypothesized MoA. It then applies an interpretive algorithm that sequentially considers responses produced by a chemical in Rank 1, 2, and 3 endpoints, in order of their importance for evaluating the hypothesis. Principal components of that sequential analysis are the pattern of response produced by the test chemical compared to endogenous effectors of the endpoint—in this case, the endogenous hormone E2—as well as a comparison of the strength with which the chemical affects those endpoints relative to the endogenous effector, which we refer to as “mechanistic potency.” Mechanistic potency distinguishes the strength by which a chemical may act via a particular MoA from its potency for producing adverse effects, which encompass all contributory MoAs (Borgert et al. 2018; Marty et al. 2018). In this step, it is important to consider the dose level required to affect the endpoint as well as the magnitude of the maximal achievable response for both the test chemical and the endogenous effector.

The relevance rankings are described as follows (Borgert et al. 2014):

• Rank 1 endpoints are specific and sensitive to the hypothesis being evaluated. These can be interpreted without clarification from other endpoints and are rarely confounded by non-specific activity. Rank 1 endpoints are in vivo measurements only because in vitro responses rarely identify a relevant biological effect.

• Rank 2 endpoints are also specific and sensitive for the hypothesis being evaluated but are less informative than Rank 1 as these are often subject to confounding influences or other modes of action. Rank 2 endpoints include both in vitro and in vivo data.

• Rank 3 endpoints are relevant for the hypothesis being evaluated but only when corroborative of Rank 1 and 2 endpoints. Rank 3 endpoints are not specific for a particular hypothesis and include some in vitro and many apical in vivo endpoints.

The application of this WoE approach is well-established for evaluating the endocrine disruptive potential of chemicals based on results of screening-level assays and apical endpoints from in vivo toxicology studies, as described in greater detail elsewhere (e.g., Borgert 2023; Marty et al. 2015; Mihaich & Borgert 2018; Mihaich et al. 2017; Neal et al. 2017). The WoE evaluation conducted here considered only estrogen agonist and antagonist pathways and androgenic endpoints that would be affected by a chemical that exhibits anti(estrogenic) activity. Quinn et al. (2007a) have shown that D4 does not interact with ERβ or with the androgen receptor, nor does it affect any endpoint in the Hershberger assay for effects on males. Data for this evaluation was taken from a comprehensive set of publications and unpublished reports detailing the effects of D4 on estrogen-sensitive endpoints observed in controlled laboratory studies. These publications and reports are numbered as shown in Supplemental Materials Table 1, with brief information on animals and exposure concentrations.

Hypothesis 2: receptor occupancy calculations

Using laws of mass action and published potency data applied to the anti(estrogenic) MoA as a test case, we recently demonstrated that the endogenous human metabolome provides a physiological and biochemical basis for the human-relevant potency threshold proposed previously (Borgert et al. 2024b). Here we used the same methodology to calculate the fraction of human ERα that would be occupied by the primary endogenous estrogen, 17β-estradiol (E2) in the presence of endogenous concentrations of ERα ligands and published potency data for D4 (Borgert et al. 2018 and citations therein). Briefly, a series of receptor occupancy calculations based on the Gaddum equation was used to test the specific hypothesis that D4 could attain sufficient occupancy of ERα receptors to alter the estrogenic tone of an intact animal amidst normal levels of even a subset of endogenous natural ERα-ligands. The subset of endogenous ERα ligands used for these calculations includes DHEA, DHEA-sulfate, androstenediol, E2, estrone (E1), and estriol (E3).

Supplemental Materials Table 2 lists K_d values and endogenous concentrations used to calculate receptor occupancies. Justification and literature citations for those affinity values and concentrations were published recently (Borgert et al. 2024b). The affinity value used for D4 was based on the potency estimated previously (Borgert et al. 2018), which is consistent with that of McKim et al. (2001). Endogenous concentrations of D4 in rats were obtained from a whole-body inhalation study in rats exposed to D4 in chambers under dynamic airflow conditions (Marty et al. 2019). Groups (n = 7–25) of estrous-staged, non-pregnant female Crl:CD(SD) rats were exposed on three consecutive days for 6 h each day to D4 at concentrations of 25, 75, 150,300, 500, or 700 ppm. Blood collections were taken each day at 0, 2, 6, 12, and 24 h, and samples were analyzed for blood levels of D4 parent compound by a modification of the method of Varaprath et al. (2000). Ovaries were sectioned and examined for corpora lutea and ovarian follicles for all animals in all dose groups following the final exposure period. Two D4 concentrations were used in the receptor occupancy calculations, shown in Supplemental Materials Table 2, Row G, one from the animal with the highest blood concentration among the 700-ppm group, and other from the animal with the highest blood concentration among the 150-ppm exposure group, which was the highest no-effect concentration in this study. The highest blood concentration occurred at time zero for every animal in every exposure group. An assumed human concentration of D4 in blood (Row H) was set at 100 × lower than rat blood concentrations measured in the kinetic study, which corresponds to a value likely to be more than 200-times higher than the highest blood concentration possible based on a human health risk assessment of D4 (Gentry et al. (2017).

Hypothesis 3: molecular docking analysis

OECD (2018) contends that output QSAR models may be applied cautiously for interpreting potential endocrine mechanisms underlying in vivo results in vertebrate species and that QSAR methods may help to identify groups of chemicals and structural alerts that are linked to in vivo effects. They note that QSAR models are now available for several receptor types, including the estrogen receptor (ER). Molecular interaction with ERα is historically the most developed model. Websites maintained by the Danish Environmental Protection Agency (DK EPA), the United States Environmental Protection Agency (US EPA), and the OECD provide links to some of these models and link to databases that provide outputs for thousands of chemicals, allowing for the combined interrogation of many types of properties, from bioaccumulation to reproductive toxicity.

Manual docking was performed within the estrogen receptor alpha-binding site using JMol (v. 14.31.53). For docking, we used RCSB PDB structure 1A52 (https://doi.org/10.2210/pdb1A52/pdb), the estrogen receptor alpha crystal structure with estradiol-bound (Tanenbaum et al. 1998). The binding site was identified as those parts of the protein surrounding estradiol. Each query chemical was docked in 3 different poses. Pose 1 aligned the/an oxygen atom from the query chemical with the hydroxyl oxygen atom in the A-ring of estradiol to minimize steric hindrance, and maximize interactions with polar side-chains (i.e., to achieve the same hydrogen bonding as the hydroxyl group of the A-ring would achieve). Pose 2 was the same as Pose 1, except that we aligned to the hydroxyl oxygen atom in the D-ring of estradiol. In Pose 3, the query chemical was moved into a pose that minimized steric hindrance within the binding site.

The SMILES codes for D4, DMSD, and DMSD-dimer diol were obtained from PubChem. The SMILES code for heptamethylcylcotetrasiloxanol was obtained by drawing the structure in PubChem. The SMILES codes for each are listed in Supplemental Materials Table 3. We obtained 3-dimensional PDB structure files by converting the smiles codes using Open Babel (v. 3.1.1).

Results

Hypothesis 1: weight-of-evidence results

Supplemental Materials Table 4 lists endpoints sensitive to estrogen agonists, arranged according to their relevance for evaluating the hypothesis that a chemical acts via an estrogen agonist mode of action as described in WoE framework (Borgert et al. 2011). This table allows an evaluation of whether the responses elicited by D4 resemble a pattern of effects expected for known estrogen agonists, e.g., positive controls. The first column on the left gives a categorical relevance ranking according to a scheme described by Borgert et al. (2014). The second column lists the type of study in which the endpoints, listed in column 3, were measured. The fourth column from the left shows the expected response based on data for known estrogen agonists on each endpoint. The two columns on the right, columns 5 and 6, show numbered citations to studies reporting responses or non-responses to D4, respectively. A comparison of these columns with column 4 enables an evaluation of the consistency of endpoint responses to D4 against patterns of responses expected for effectors of this endocrine pathway.

Supplemental Materials Table 4 shows that among Rank 1 endpoints for the estrogen agonist pathway, D4 has not been evaluated for vitellogenin production in the fish short-term reproduction assay (FSTRA), but has been evaluated extensively in the rodent uterotrophic assay (RUA). D4 increased uterus weight in five RUAs conducted in rats [4, 7, 8, 9, 10] and in one RUA conducted in mice [11]. D4 produced no change in uterus weight in a sixth RUA [14], however, this was the only RUA conducted with D4 by the subcutaneous route of exposure, which may have produced lower circulating levels of D4 than inhalation and oral routes. The definitive endpoint in RUA assays is an increase in uterine weight, which was not observed in that study, but D4 did alter estrogen-receptor-related gene expression [14]. Although not shown in Supplemental Materials Table 4, the estrogen receptor antagonist ICI 182,780 significantly reduced D4’s activity in the RUA [4, 7, 8, 11, 14]. The strength of D4’s effect in these RUAs is on the order of 5 × 10^–6 relative to E2, and much less potent than common botanical estrogens as determined within (McKim et al. 2001) and across studies (Borgert et al. 2018). Thus, although D4 consistently affects the primary Rank 1 endpoint for the estrogen agonist MoA, the RUA, it does so with lower potency than any chemical capable of producing observable estrogenic effects in humans (Borgert et al. 2018).

Rank 2 comprises in vitro assays specific for the estrogen agonist MoA as well as apical endpoints from in vivo assays that can be affected by many different endocrine and non-endocrine MoAs. The latter include short-term assays such as the Fish Short-Term Reproduction Assay (FSTRA) and the male and female pubertal assays, as well as sub-chronic and chronic assays such as repeat dose toxicity studies, developmental toxicity studies, and reproductive toxicity tests. A lack of response to the test chemical in such endpoints is more compelling evidence of a lack of activity than a response is evidence of activity (Borgert et al. 2014; Mihaich & Borgert 2018; Borgert 2023). This is because a response can be elicited via MoAs other than the endocrine MoA under evaluation, whereas a lack of response indicates a lack of activity via any MoA that affects the endpoint, regardless of its endocrinological basis.

Supplemental Materials Table 4 shows that responses to D4 were measured in all but 13 of the 53 Rank 2 endpoints for the estrogen agonist MoA. Among the 40 Rank 2 endpoints measured, 26 were unaltered by D4 at any dose or concentration level. Conspicuous among them is a lack of response in male offspring exposed in utero to D4 in reproductive toxicology and developmental toxicity studies, as those endpoints are sensitive to the effects of estrogen and would be expected to respond to a chemical that operates via an estrogen agonist MoA. Among the 14 Rank 2 endpoints that responded to D4, only the in vitro estrogen receptor transactivation assay (ERTA) did so consistently. The rest responded inconsistently across studies, including various endpoints evaluated in repeat dose, developmental, and reproductive toxicity studies. The apical Rank 2 endpoints that responded inconsistently to D4 relative to the direction expected for a chemical with estrogen agonist activity include ovary histopathology, ovary weight, uterus histopathology, uterus weight, vagina histopathology, corpora lutea, estrous cyclicity, fertility, number of implantations, litter size, mammary histopathology, and mating index. Testis weight was unaltered in five repeat dose toxicity studies [1, 2, 3, 16, 17] but was increased in one repeat dose toxicity study [12]. However, a chemical with estrogen agonist activity would decrease, not increase testis weight, and testis histopathology was unaltered in the same study, as well as in five other repeat dose toxicity studies [1, 2, 3, 16, 17]. Since testis histopathology and weight would be expected to both be affected by an estrogen agonist, the increase in testis weight with unaltered testis histopathology in that study [12] is evidence against an estrogen agonist effect on the testis. Overall, where responses were observed in Rank 2 endpoints, they were constrained to the high doses at or above the NOAELs reported for D4 (Gentry et al. 2017). Rank 3 endpoints are interpretable only as corroboration of responses in Rank 1 and Rank 2 endpoints. Only 2 of 15 Rank 3 endpoints were measured in response to D4, one that showed no effect (gross pathology) and the other (displacement of estradiol in the estrogen receptor binding assay) produced inconsistent results.

Supplemental Materials Table 5 lists endpoints sensitive to estrogen antagonists, arranged according to their relevance for evaluating the hypothesis that a chemical acts via an estrogen antagonist mode of action as described in the WoE framework (Borgert et al. 2011). This table allows an evaluation of whether the responses elicited by D4 resemble a pattern of effects expected for known estrogen antagonists, e.g., positive controls. The first column on the left gives a categorical relevance ranking according to the scheme mentioned previously (Borgert et al. (2014). The second column lists the type of study in which the endpoints (column 3) were measured. The fourth column shows the expected response based on data for known estrogen agonists on each endpoint, and the last two columns (right) show numbered citations to studies reporting responses or non-responses to D4, which enable a comparison of consistency with patterns of responses expected for effectors of this endocrine pathway.

Supplemental Materials Table 5 shows that D4 reduced the uterotrophic response to E2 in two RUA assays, but not in two others, which constitutes an equivocal Rank 1 response for the estrogen antagonist MoA. A response to D4 was measured in 26 of the possible 29 Rank 2 endpoints for this MoA, 14 of which were unaltered by D4 administration at any doses or concentration tested. Of the seven endpoints that responded to D4 administration, none responded consistently across studies, including displacement of bound estradiol in vitro (ERBA), ovary histopathology, testis weight, corpora lutea (reproduction study), estrous cyclicity, fertility, and litter size. Testis weight was unaltered in five repeat dose toxicity studies [1, 2, 3, 16, 17] but was altered (increased) in the opposite direction expected of a chemical with estrogen agonist activity in one repeat dose toxicity study [12]. Testis histopathology was unaltered in the same study [12]. Since testis histopathology and weight would be expected to change together, the observed result [12] is evidence against estrogen antagonist effects on the testis. Gross pathology was unaltered in five studies, the only Rank 3 endpoint measured following D4 administration. Taken together, the lack of a pattern of responses expected of an estrogen antagonist and the infrequency and inconsistency of responses across studies indicate a lack of activity of D4 via this pathway.

Hypothesis 2: receptor occupancy results

The ability of a chemical to produce effects via a particular hormone receptor can be evaluated through fractional occupancy calculations that include a consideration of the concentrations and affinities of endogenous receptor ligands that all compete for receptors under physiological conditions (Borgert et al. 2024b). Fractional receptor occupancy calculations allow potency information derived from screening-level in vitro and in vivo endocrine assays to be interpreted within a relevant physiological context, and thus, to be used to evaluate the hypothesis that a chemical can produce physiological effects via a particular endocrine MoA. Supplemental Materials Table 6 shows the results of fractional receptor occupancy calculations for D4 amidst a physiologically relevant background of endogenous ERα ligands. The fraction of ERα occupied by E2 alone is shown in row Q at its mid-point and minimal plasma concentrations. The mid-point plasma concentrations of E2 and endogenous ERα ligands correspond to values typical for pre-menopausal women and minimal concentrations correspond to values typical for pre-pubescent males, who have the lowest E2 levels of any human life stage. Row R shows the fraction of receptors occupied by E2 in the presence of the subset of five endogenous ERα ligands (row R) at their mid-point and minimal plasma concentrations in humans. Rows S and T show the fraction of ERα occupied by E2 in the presence of the endogenous ligand at mid-point and minimal plasma concentrations, with the addition of D4 at the highest blood concentration measured in rats following exposure to 700 ppm and 150 ppm, respectively. Rows U, V, and W show the total fraction of ERα occupied by the mixture of all ligands considered in Rows R, S, and T, respectively. Rows X, X₁, Y, and Y₁ show the percent change in receptor occupancy by E2 caused by the addition of D4 to the endogenous metabolic milieu at the plasma levels of the endogenous steroids used for these calculations.

The results of receptor occupancy calculations shown in Rows Q, R, and U were used previously to establish that a 5% change in the ERα occupancy by E2 is within the normal physiological range (Borgert et al. 2024b). Thus, to be capable of causing effects through ERα, an exogenous ligand would need to have affinity for the receptor sufficient to alter E2 occupancy by more than 5% under physiological conditions, i.e., when in competition with the endogenous metabolic milieu. Row X shows that if D4 were present in human blood at a concentration equal to the highest concentration observed in rats exposed to 700 pm D4, total ERα occupancy would be altered by less than 0.01 percent at mid-point plasma concentrations of the subset of 6 endogenous ligands considered here. At minimal concentrations of those six endogenous ligands, total receptor occupancy would change by less than 4%, which is within the range of normal physiological variation (> 5%). Row X₁ shows that the corresponding effect on the fraction of receptors occupied by E2 would be less than 0.2 percent, which is more than 25 times below the range of normal physiological variation (> 5%). If D4 were present in human blood at a concentration equal to the concentration observed in rats exposed to 150 ppm D4, which is the highest concentration that showed no effects on corpora lutea in rats, the total occupancy of ERα would change by only 0.001% at mid-point levels of the endogenous ligands and by less than 1.3% at their minimal concentrations (Row Y). Row Y₁ shows that the corresponding changes in receptor occupancy by E2 are also minuscule; D4 would reduce receptor occupancy by E2 less than 0.002% at the mid-point of the plasma concentrations of the endogenous steroids, and by less than 0.01% at minimal plasma concentrations of the endogenous steroids. Thus, due to overwhelming concentrations of endogenous ERα ligands, D4 would be unable to occupy biologically meaningful fractions of estrogen receptors under any foreseeable physiological condition in humans.

Hypothesis 3: molecular docking results

Molecular interaction with ERα is historically the most well-developed QSAR model among endocrine receptors and can be used cautiously for interpreting a chemical’s potential for producing effects via an estrogenic MoA. We have used a type of QSAR, manual molecular docking, to test the hypothesis that D4 or its metabolites may interact with the ligand binding site of ERα with sufficient specificity to provide a plausible MoA for in vivo effects. Since the relationship between docking in the binding site of a hormone receptor should parallel its affinity for the receptor and thus, its ability to occupy receptors amidst competition with other receptor ligands, this molecular docking analysis serves as an independent test of Hypothesis 2.

D4 and its metabolites DMSD, DMSD-dimer diol, and heptamethylcylcotetrasiloxanol all show steric hindrance in Poses 1 and 2 (Supplemental Materials Figs. 1–4). All but DMSD show steric hindrance in Pose 3. This is not surprising as DMSD is much smaller than all the other chemicals queried. The steric hindrance is evidenced by overlapping van der Waals radii between the query chemicals and the amino acids lining the binding pocket. This demonstrates that even in the poses where these chemicals would have the best chance of coordinating activation of the estrogen receptor, activation simply cannot occur due to steric hindrance. Poses 1, 2 and 3 for D4 (Supplemental Materials Fig. 1 E/F) shows that there is no room for D4 to coordinate with two water molecules (D4-1,5-dihydrate). This provides further evidence that the D4-1,5-dihydrate proffered by Matthews (2021), although an intriguing hypothesis, is simply not possible due to steric hindrance.

Discussion

Many authors have reviewed the toxicology of D4 and have concluded that its adverse on endocrine-sensitive endpoints generally occur at doses that exceed saturation of D4 metabolism (Gentry et al. 2017; Sarangapani et al. 2002), suggesting a MoA dependent on systemic toxicity rather than one mediated via endocrine activity (Andersen 2022; Dekant et al. 2017a, b; Matthews 2021; Pauluhn 2021). Nonetheless, some lists of purported Endocrine Disruptors include D4 on the premise that it exhibits estrogenic endocrine disruptive properties.

To resolve whether D4 has the potential to produce adverse effects consequent to an anti(estrogenic) MoA, we examined aspects of the interaction of D4 with estrogenic pathways at three levels of biological organization. At the physiological level, we applied an established weight-of-evidence (WoE) framework (Borgert et al. 2011; 2014) to evaluate whether D4 elicits a pattern of effects consistent with an anti(estrogenic) MoA. At the biochemical level, we used receptor occupancy calculations and published data on endogenous estrogens in humans to determine whether D4 can occupy a sufficient fraction of estrogen receptors to produce effects by this pathway under physiological conditions. At the molecular level, we evaluated whether D4 exhibits ligand-binding properties consistent with agonism or antagonism of ERα.

WoE methodologies are constructed specifically to enable transparent evaluations of data according to clearly specified criteria when no gold-standard methodology exists to unequivocally determine the correct interpretation. Since no gold-standard assay exists for any mode of endocrine action, WoE methods cannot be formally “validated,” but in lieu of a gold standard, WoE methods are appropriate for determining whether the mechanistic and toxicological data on a chemical satisfy the definition of EDC (WHO/IPCS, 2002). The U.S. EPA and the OECD both recommend the use of WoE for this purpose. According to the guidance for applying the OECD toolbox (OECD 2018):

All existing relevant data should be maximally used (e.g. structural; physico-chemical information; in vivo and in vitro guideline and non-guideline testing; QSAR models; computational and other non-testing assays; toxicokinetic, pharmacokinetic, and toxicodynamic information; category and read-across assessment methodologies) in a WOE approach before entering any other level of the CF.

Hypothesis 1 was tested by our WoE evaluation of D4 via the estrogen agonist and antagonist pathways at the physiological level (Supplemental Materials Tables 4 and 5) and indicates that D4 does not produce a pattern of effects consistent with either of these MoAs. Most of the apical endpoints measured in repeat dose, developmental, and reproductive toxicology studies that are responsive to estrogen agonists were unaffected by D4, as was true of the endpoints responsive to estrogen antagonists. Those endpoints are categorized as Rank 2 because they can be affected by a variety of the MoAs and are not as specific for estrogen agonism or estrogen antagonism as Rank 1 endpoints. However, many of these Rank 2 endpoints are equally as sensitive as Rank 1 endpoints and more of those would be expected to have responded if D4 were acting via an estrogen agonist or antagonist MoA. The endpoints that did respond to D4 in some studies failed to respond in other studies, and for most of those endpoints, more studies showed a lack of response than showed a response. D4 has been shown not to interact with ERβ, but even if D4 were postulated to stimulate ERα in a unique way that affected only the endpoints that responded in some studies, a greater consistency of response among those endpoints would be expected across studies. As explained previously, a consistent lack of response in endocrine-sensitive apical endpoints is a strong indication of the lack of an endocrine MoA, whereas a response may be due to either an endocrine or non-endocrine MoA.

Our WoE analysis reveals that the only consistency of response to D4 is among screening level assays for the estrogen agonist pathway, specifically the RUA, a Rank 1 endpoint, and the ERTA, a Rank 2 endpoint. Among these, it is noteworthy that although the EPA’s ToxCast^® algorithm for the estrogenic MoA finds that D4 lacks this bioactivity, the ToxCast^® data are unusable due to quality control problems (Supplemental Materials Appendix A). The interaction of D4 with ERα as evidenced by the RUA, ERTA, and ERBA is both consistent and extremely weak, with a mechanistic potency well below the threshold (1E-04 relative to E2) necessary to affect human physiology (Borgert et al. 2018; Borgert 2023). That the mechanistic potency of D4 is insufficient to elicit physiological responses via either the estrogen agonist or antagonist pathways is consistent with its lack of effect on most estrogen-sensitive endpoints, even at high doses, and the inconsistency of D4’s effects on other endpoints (Supplemental Materials Tables 4 and 5). Nonetheless, we further investigated the possibility that D4 could affect human physiology via estrogenic pathways by conducting an extensive receptor occupancy evaluation of D4 in physiologically relevant contexts using large overestimates of potential D4 blood levels.

Hypothesis 2 was tested by our receptor occupancy calculations (Supplemental Materials Tables 2, 6). These are based on more than 100 years of established pharmacological receptor and enzyme kinetic theory and empirical data on endogenous ERα ligands, as reviewed more fully in Borgert et al. (2024b) and applied to experimental data on D4. Our calculations reflect the highest conceivable effect of D4 on total ERα receptor occupancy and occupancy by E2 for at least three reasons. First, instead of group mean concentrations, the D4 concentrations used were from the animal with the highest blood concentration in the highest exposure group (700 ppm) and from the animal with the highest blood concentration in the group that showed no biological changes (150 ppm). Second, although those D4 concentrations were obtained in rats exposed to concentrations many thousands of times higher than are possible in the general human population (Gentry et al. 2017), those concentrations were evaluated in the context of mid-point and minimal levels of endogenous ERα ligands in humans, respectively, representing the endogenous metabolome of non-pregnant human females and pre-pubescent males. Third, the greater the pool of endogenous ligands competing with a low-affinity exogenous ligand for occupancy at a receptor, the less will be the proportional effect of an additional weak exogenous ligand on occupancy by a high-affinity ligand such as a hormone. The receptor occupancy calculations shown here represent an excessively conservative estimate of the effects of D4 on ERα occupancy by E2 because only a subset of the endogenous ERα ligands present in humans was included in the calculations; six endogenous ligands were considered here, whereas the normal human metabolome contains additional ERα ligands (Borgert et al. 2024b).

The receptor occupancy calculations shown here reveal that D4 would be unable to occupy a biologically meaningful fraction of estrogen receptors in humans, even under physiological conditions where E2 and other ERα ligands are at their lowest levels, as might occur in prepubertal males, which is modeled by our “minimal plasma concentrations” of endogenous ligands. Thus, our results are inconsistent with the proposition that D4 could produce effects in humans via an estrogenic MoA. It is difficult to envision a mechanism whereby a chemical with the minuscule affinity and potency of D4 could nonetheless produce effects via the estrogen pathway. If D4 acted downstream of ERα, one would expect a much higher potency in vivo than in vitro. This is clearly not borne out by the in vivo data. D4 has generally shown lower potency in in vivo versus in vitro screening assays (Borgert et al. 2018; He et al., 2019; Matthews 2021; McKim et al. 2001; Quinn et al 2007a).

Chemicals with high mechanistic potency for a receptor-mediated MoA are expected to exhibit strong ligand-binding characteristics at that receptor, including size, shape, and molecular surface characteristics. Conversely, chemicals with low mechanistic potency would be expected to exhibit poor ligand-binding properties. Matthews (2021) recently evaluated the plausibility of an estrogenic MoA for D4 and finding no evidence for that proposition, proposed the only conceivable molecular alternatives that might explain the anomalous results of the RUA and ERTA with D4. Although he considered them unlikely, he proposed the formation of a D4-dihydrate that might possess the requisite molecular character to activate the estrogen receptor and displacement of E2 from plasma-binding proteins. We tested the former hypothesis through molecular docking analysis and found excessive steric hindrance when D4, its dihydrate coordinate, or related siloxanes were force-fitted into the ligand binding pocket of ERα, rendering an agonist or antagonist MoA via the estrogen receptor highly implausible. Thus, the premise that D4 exhibits estrogenic endocrine disruptive properties is not based on a thorough or rigorous analysis of the data, but on the faulty presumption of a causal link between the results of short-term screening assays with D4, such as the RUA and ERTA, and apical effects in reproduction studies without a rigorous consideration of the plausibility of that causal link. That presumption is deficient for at least four reasons.

The first deficiency is a failure to rigorously evaluate mechanistic potency at the biochemical level. Results of short-term screening assays cannot be interpreted properly without considering the mechanistic potency of a chemical to act via the postulated MoA, as we have shown (Borgert et al. 2018). To make valid and reliable interpretations, such assays cannot be considered purely qualitatively, as either “positive versus negative,” or “weak versus strong,” but must be considered by quantitative comparisons to the endogenous hormones and effectors of the hormonal pathway under consideration. When the mechanistic potency of D4 was properly considered, it was shown to be well below the Human-Relevant Potency Threshold (HRPT) for estrogen agonism through ERα (Andersen 2022; Borgert et al. 2018; Matthews 2021).

The second deficiency is the failure to evaluate the presumed MoA at the physiological level. This requires an appreciation that short-term mechanistic screening assays such as the RUA, ERTA and ERBA, are optimized for sensitivity so that fine distinctions can be measured between the ability of different chemicals to act via a specific MoA. For example, RUAs are conducted for a very short window of time either just following ovariectomy, when E2 levels are low but sensitivity to estrogens has not been down-regulated, or at a very specific time prior to puberty when E2 levels are still very low but the pre-pubescent animal is fully sensitive to estrogen. Similarly, in vitro conditions for assays such as the ERTA and ERBA are purposely conducted under non-physiological conditions so that no competing estrogens are present, e.g., culture media are serum-free and all phenolic additives that can interact with estrogen receptors are avoided. To make physiologically relevant interpretations, the results of such assays must be considered and interpreted in the context of physiological conditions rather than in the vacuum of the artificial conditions under which they are conducted (Borgert et al. 2024b).

The third deficiency is the failure to consider the consistency of the data and instead, to focus only on the data that support the presumed link between an estrogenic MoA and adverse effects. This is a critical step because apical endpoints measured in endocrine-sensitive tissues in repeat dose toxicity, developmental toxicity, and reproductive toxicology studies can be affected by many different MoAs, not only endocrine MoAs (Marty et al. 2018). To determine whether D4 produces its adverse effects via the estrogen MoA, it is critical to compare the full spectrum of D4’s effects on estrogen-sensitive endpoints with the pattern of effects produced in those endpoints by known estrogens (Borgert et al. 2011; 2014). The fourth deficiency is the failure to consider the plausibility of the MoA at the molecular level and its consistency with data on mechanistic potency. Considering the widespread availability of this technology and the fact that in situ methods are Level 1 assays in the European Chemical Hazard Assessment’s Endocrine Disruptor Toolbox, molecular docking could and should have been done before leaping to the conclusion that D4 has estrogenic potential.

The only remaining mechanistic possibility that might explain the RUA results for D4 was that at high concentrations in rodents, D4 displaces E2 from plasma carrier proteins and thereby increases circulating levels of free E2, which then produces estrogenic effects in rodents (Matthews 2021). Given the results of receptor occupancy calculations and molecular docking studies conducted here, and the fact that D4’s uterotrophic effect has been shown to be mediated via estrogen receptors as evidenced by antagonism by the estrogen receptor antagonist ICI 182,780, an indirect effect mediated by an increase in free E2 seems the most likely explanation for the uterotrophic effects of D4 in rodents. Although this MoA would explain the RUA results in rodents, it would also render the RUA results irrelevant to humans.

In rats, sex hormone binding globulin (SHBG) is not expressed postnatally in the liver and is found in only trace amounts in adult rats (Jänne et al. 1999; Hammond 2011, 2016); thereafter, circulating steroids are bound primarily to plasma albumin in rats. In contrast, although steroids are also bound to albumin in humans, SHBG levels increase during puberty and play important roles in modulating the free concentrations of primary endogenous sex hormones via selective binding of E2, testosterone, and dihydrotestosterone. SHBG has approximately four orders of magnitude greater affinity for E2 than does albumin (Rosner 2015). We have shown previously that affinity differences of 1E-4 render binding to ERα physiologically irrelevant in humans due to the overwhelming circulating concentrations of endogenous steroids (Borgert et al. 2024b), and herein, that D4 is incapable of a physiologically relevant displacement of E2 from ERα. Given the selectivity of SHBG for E2 relative to albumin, and the fact that E2 levels are hundreds of times lower in rodents than in humans (Witorsch 2002), this difference in affinity is also likely to be physiologically determinative for binding to SHBG versus albumin. Since competitive binding between E2 and D4 with SHBG is quantitatively similar to their competitive binding to ERα than to albumin, the difference in affinity likely renders displacement of E2 from SHBG physiologically irrelevant in humans, just as it is irrelevant for displacement of E2 from ERα Thus, although displacement of E2 and other endogenous estrogenic steroids from albumin by D4 might contribute to the uterotrophic activity of D4 in the rodent uterotrophic assay, in humans, this would not occur to any appreciable extent due to the much higher selectivity and affinity of E2 for SHBG relative to its affinity for albumin.

A final point should be made regarding the alleged potential for chemicals to produce non-monotonic dose responses (NMDR) via endocrine modes of action. These can be understood quite clearly from classical receptor theory, as used here in the evaluation of the potential for D4 to produce effects via ERα. The endogenous metabolic milieu helps to maintain the appropriate level of endocrine receptor stimulation for the life stage and sex of the organism, which is mediated primarily via receptor occupancy by the endogenous hormone. For ERα in humans, this level of receptor occupancy appears to be between 5 and 12% in males and in non-pregnant females but would increase considerably during pregnancy when E2 levels rise dramatically (reviewed in Borgert et al. 2024b). An exogenous chemical with a high affinity for ERα, but with low intrinsic efficacy—i.e., a partial agonist, examples of which include selective estrogen receptor modifiers (SERMs) such as Tamoxifen^® and Raloxifene^®—could conceivably exhibit a non-monotonic dose–response relationship. At low doses, a SERM with low intrinsic efficacy would reduce receptor occupancy by E2 without producing any agonist effect of its own due to low intrinsic efficacy. At high doses, it would displace an even greater fraction of E2, and could, with sufficient receptor occupancy, produce its own agonist effect. In a person, this would be observed as an apparent estrogen antagonism at low doses, but agonist effects at high doses. It must be emphasized that this potential mechanism carries the absolute requirement of sufficient affinity to successfully compete with the overwhelming endogenous milieu of ERα ligands, including E2. Tamoxifen®, for example, has an affinity for ERα within an order of magnitude the affinity of E2 (Kuiper et al. 1997), which is well above the HRPT for ERα (Borgert et al. 2018; 2024b). Chemicals with weak affinity for a hormone receptor could not produce non-monotonic dose responses via that receptor because they would not successfully compete with the endogenous metabolic milieu of receptor ligands, much less with the endogenous hormone. The affinity of D4 for ERα is well below the HRPT for ERα (Borgert et al. 2018; 2024b) and far too low to produce any physiological (anti)estrogenic response via that MoA, including an NMDR. If NMDRs are observed with weak receptor ligands, a non-receptor MoA would likely be involved, e.g., metabolism, kinetics or a secondary stress response.

Conclusions

To resolve differing interpretations of the potential for D4 to produce estrogenic endocrine disruptive effects, we examined aspects of the interaction of D4 with estrogenic pathways at the physiological, biochemical, and molecular levels of biological organization. We used an established WoE methodology to evaluate the consistency and strength of the data, established equations for calculating receptor occupancy based on published data on affinities and circulating concentrations of endogenous estrogens in humans and established molecular docking techniques. Our analyses produced concordant results indicating that an estrogenic effect of D4 is molecularly, biochemically, and physiologically implausible, consistent with previous evaluations that concluded the effects of D4 cannot be due to an estrogenic endocrine disruptive MoA. Claims that D4 exhibits estrogenic endocrine disruptive properties based on a presumed link between the results of screening-level assays (RUA and ERTA) and adverse effects rely on deficient evaluative and interpretative methods. Instead, a plausible mechanistic explanation for the various adverse effects of D4 observed in rodent studies, including its effects in reproduction studies, is that these are secondary to high-dose-dependent, physico-chemical effects that perturb cell membrane function and produce rodent-specific sensory irritation.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

Disclosure: this work was funded by the Silicones Environmental Health & Safety Center (SEHSC). SEHSC reviewed and commented on the initial draft of the manuscript.

Data availability

The underlying data have been provided in the Supplementary information (Supplemental Tables and Figures), or in other publications, as indicated in the manuscript.

Declarations

Conflict of interest

The methodologies, analyses, interpretations, content of the manuscript, and the decision to submit it for publication were made solely by the authors and did not depend on approval from SEHSC.