Probing the Biological Plausibility of Fluoride as an Endocrine Disruptor

ABSTRACT

Background

Fluoride has come under recent scrutiny regarding concerns over potential neurodevelopmental and endocrine‐related toxicities, with recent reviews by the National Toxicology Program (NTP) and European Food Safety Authority (EFSA) concluding with moderate or reasonable confidence, respectively, that exposure to drinking water having greater than 1.5 mg fluoride/L is associated with lower IQ in children. A key outcome of these reviews is the uncertainty regarding the biological plausibility of these findings. However, it has been hypothesized that endocrine disruption could be a potential factor.

Methods

To determine if sodium fluoride exerts direct biological activity on molecular and cellular targets related to endocrine disruption, receptor binding and activity assays of thyroid and other hormone‐related targets, H295R steroidogenesis, and sodium‐iodide symporter (NIS) assays were carried out using exposures comparable to or in excess of those that have been reported to be associated with neurodevelopmental outcomes and other effects.

Results

Sodium fluoride at up to 316 μM NaF did not affect synthesis of estrogen or testosterone. Sodium fluoride at up to 10 μM NaF did not interact with aromatase, steroid 5 alpha‐reductase, estrogen receptors, androgen receptors, thyroid hormone receptors, nor did it inhibit thyroid peroxidase. Furthermore, there were no changes in iodide uptake via symporter transport (up to 300 μM NaF). Other endocrine targets were also evaluated at 10 μM NaF, including PR, PPARα, PPARγ, AhR, CAR, PXR, RARα, or GR, and no binding was observed either.

Conclusions

Together, the results from this series of experiments demonstrate an absence of effects of fluoride on endocrine disruption targets at concentrations comparable to or in excess of exposures reported in the literature to be associated with neurodevelopmental outcomes and other effects.

1. Introduction

Fluoridation of water is one of the greatest public health interventions of the 20th century (CDC ¹⁹⁹⁹), effectively reducing dental caries and improving the oral health of adults and children. In contrast to intentional fluoride exposure, in arid or semi‐arid fluoride belts such as areas in Turkey, China, and India, exposures can be much higher due to endemic groundwater contamination with high levels of fluoride and exposures from coal combustion, particularly, for food contaminated during the cooking process (Ando et al. 2001; Wang et al. ²⁰²¹). Much of the research on fluoride toxicity has been conducted in these regions given the concern for public health and the unknowns regarding the effects of fluoride at extremely elevated doses over extended periods of time. In China, fluoride reduction initiatives over the past few decades have effectively reduced the drinking water fluoride content resulting in reduced prevalence of dental fluorosis in children living in fluorosis‐endemic areas of the country (Wang et al. 2021; Yang, et al. ²⁰²⁵). Dental fluorosis has long been considered to be the most sensitive endpoint for fluoride toxicity. At exposures above this amount, the more debilitating skeletal fluorosis can occur. These two endpoints provide a conservative reference from which to derive limits and to evaluate the specificity of purported effects of fluoride (WHO ²⁰¹⁹).

In this study, we investigated the potential effects of NaF on endocrine‐mediated pathways, particularly those relevant to neurodevelopment and reproduction. These endpoints have been of particular focus over the last decade since in addition to the more traditional dental and skeletal fluorosis endpoints, there has been interest in exploring more generalized systemic toxicity, particularly, at elevated exposures (Guth et al. 2020). Of the long list of reported effects, particular interest has been on the potential reproductive (Luo et al. 2023; Cheng et al. ²⁰²⁴; Li et al. ²⁰²⁴), thyroid (Guan et al. 1988), and cognitive effects (National Toxicology Program ²⁰¹⁶) that have been reported in non‐GLP studies.

In addition to the above studies, several high‐quality and GLP‐compliant studies have been conducted including a multigenerational reproductive toxicity conducted at the US FDA CFSAN (Collins et al. 2001), a neurodevelopmental study from the US NTP (McPherson et al. ²⁰¹⁸), and several subacute and chronic repeat dose studies (ECHA ²⁰²⁴), where only dental, skeletal, and GI (mild hyperkeratosis and acanthosis of the nonglandular portion of the stomach, renal effects, and chronic inflammation of the gastric mucosa) toxicities were noted along with decrements in body weight. These marked differences between the results attained in the higher quality regulatory and GLP studies and those from the non‐GLP studies could potentially be due to differing experimental conditions, high exposure scenarios, and inadequately controlled sources of additive exposures, as mentioned by McPherson et al. (2018).

Given the conflicting results in the literature and potential public health implications, several comprehensive reviews on fluoride safety have been conducted by scientific agencies, including the US NTP in 2016 (National Toxicology Program ²⁰¹⁶) and 2024 (National Toxicology Program ²⁰²⁴) and EFSA in 2025 (EFSA ²⁰²⁵). The most recent reviews have focused on associations reported in epidemiological studies linking elevated fluoride intake to decreased IQ in children.

In its 2016 review of the available data on the effects of fluoride on learning and memory in animal studies, the US NTP identified 68 studies using mice or rats and testing drinking water or dietary concentrations of 0.45–272 ppm fluoride (0.12–40 mg/kg/day). They concluded that at exposure levels greater than 0.7 ppm, there was a moderate level of evidence that fluoride had an impact on learning and memory in animals exposed as adults and weaker (low level‐of‐evidence) evidence in animals exposed during development. The report highlighted that there were very few studies that assessed learning and memory effects in experimental animals at exposure levels near 0.7 ppm, the recommended level for community water fluoridation in the United States and that confidence in these findings, as highlighted by the NTP, was reduced primarily based on potential confounding of the learning and memory assessments by deficits in motor function or fear‐related behaviors and risk of bias limitations (National Toxicology Program ²⁰¹⁶). In response to these deficiencies, the NTP designed and executed a developmental neurotoxicity study (McPherson et al. ²⁰¹⁸). This well‐conducted study failed to show any effects of fluoride at levels relevant for drinking water supplementation on neurobehavioral development.

In 2024, the US NTP published an updated monograph on fluoride (National Toxicology Program ²⁰²⁴). NTP concluded, with moderate confidence, that higher levels of fluoride exposure, such as drinking water containing more than 1.5 mg of fluoride per liter, were associated with lower IQ in children. They also highlighted that there were insufficient data to determine if the low fluoride level of 0.7 mg/L currently recommended for US community water has a negative effect on children's IQ and no evidence that fluoride exposure had adverse effects on adult cognition. Furthermore, the US NTP states that their review does not establish a causal link between water fluoridation and lower IQs.

In 2025, EFSA updated their consumer risk assessment of fluoride in food and drinking water, and included contributions from fluoride salts and dental products (EFSA ²⁰²⁵). Their assessment reviewed hazard data related to the central nervous system (CNS), thyroid function, and bone health. EFSA also reviewed the potential for reproductive toxicity but did not consider this a priority endpoint for fluoride in its review. EFSA concluded that dental fluorosis remains the most sensitive adverse health effect associated with fluoride exposure in children < 8 years. Based on dental fluorosis, EFSA established tolerable upper intake levels (ULs) of 1, 1.6, and 2 mg/day for children < 1, 1–3, and 4–8 years old, respectively. The ULs were established for aggregate fluoride exposure from multiple sources.

The evidence of adverse effects in the CNS and thyroid function at fluoride concentrations in drinking water below 1.5 mg/L was determined to be inconclusive, with the evidence of effects on the thyroid highlighted as being less robust (EFSA ²⁰²⁵). At concentrations of drinking water above 1.5 mg/L, it was concluded that there was evidence that there are adverse CNS effects. CNS effects were used to establish a safe level of intake of 3.3 mg/day for adults and children > 8 years of age. The safe level of intake was established for aggregate fluoride exposure from multiple sources.

One of the recommendations that the EFSA scientific committee made was for “Additional evidence to be generated to determine the biological activity of fluoride on molecular and cellular targets in order to support mode(s) of action of fluoride relevant to the endpoints assessed and assist in the interpretation of the biological relevance of the effects reported in human and animal studies.”

These reviews highlight the uncertainty regarding whether the reported findings on lower IQ and thyroid function are biologically plausible and the relevance of such findings at typical exposure levels. This lack of certainty regarding the reported IQ findings has been further highlighted in recent publications such as in a study by Warren et al., where they utilized data from the nationally representative High School and Beyond cohort in the United States to see if fluoride exposure from drinking water across adolescence impacted cognitive test performance in secondary school and later in life. The authors found that children exposed to fluoride in drinking water at the recommended levels exhibited modestly better cognition in secondary school, which became no longer statistically significant in later life around the age of 60 (Warren et al. 2025). Similar positive or lack of associations between fluoride exposure and scores in cognitive domains was also highlighted previously in a study by Ibarluzea et al., using data from the Infancia y Medio Ambiente (Childhood and Environment, INMA) birth cohort project (Ibarluzea et al. 2022). Therefore, in order to address this uncertainty and better understand if fluoride exerts direct biological activity on molecular and cellular targets related to the endocrine disruption and thus neurodevelopmental toxicity, an approach consisting of receptor binding and activity assays of thyroid and other EATS (estrogen, androgen, thyroid, steroidogenic) and non‐EATS hormone related targets, such as H295R steroidogenesis and sodium‐iodide symporter (NIS) assays were carried out using exposures comparable to and in excess to those associated with the reported effects on neurodevelopment. These studies were meant to provide context and greater mechanistic depth to the high quality regulatory, GLP and guideline studies that have previously been conducted.

2. Methods

2.1. Dose Selection

Before carrying out the experiments, we first ensured that the fluoride exposures were comparable to or in excess of levels associated with adverse effects. Much work has been done to determine the kinetics of fluorides. Soluble fluorides such as sodium fluoride (NaF) are known to almost completely be absorbed from the gastrointestinal tract into the bloodstream, with peak plasma levels occurring within 20–60 min after oral ingestion (EFSA ²⁰⁰⁵). Once absorbed, fluoride does not accumulate in most soft tissues, with a typical ratio of tissue vs. plasma fluoride level ratio of 0.4–0.9 (Whitford et al. 1979; ATSDR ²⁰⁰³). Therefore, blood levels provide a useful reference for estimating internal fluoride exposure concentrations. According to literature from the Mayo Clinical Laboratory associated with evaluation of patients exposed to fluoride from common sources, including the diet, beverages, drinking water, and the normal use of fluoride toothpastes, indicates that plasma fluoride is consistently within the range of 1–4 μM (Wermers et al. 2011). This range is used clinically to assess accidental fluoride ingestion and for monitoring patients receiving NaF for bone disease or patients receiving voriconazole (a triazole antifungal medication) therapy. Further, it is considered to be representative of the peak plasma fluoride levels in children as well, even in preschool age children after ingestion of fluoride containing toothpaste, which were shown to be 3.63 ± 0.45 μM (Ekstrand et al. 1983). In addition, serum fluoride of 0.081 mg/L (approximately 4.3 μM) was reported to be associated with lower IQ in children (Xiang et al. 2011). Therefore, the upper end of the range, 4 μM fluoride (equivalent to roughly 8.8 μM NaF), provides a helpful reference point to evaluate fluoride toxicity and in contextualizing the results discussed below.

All assays were conducted by external third‐party labs/contract research organizations using their standard protocols. USP/EP grade NaF (Supelco, Sigma Aldrich) was used in the studies.

2.2. NIS Assay

Prior to conducting the NIS assay, the solubility, cytotoxicity, and the ability of NaF to interfere with the Sandell‐Kolthoff reaction (colorimetric detection method) were assessed. Briefly, the kinetic solubility of NaF in the uptake assay buffer, Hanks'‐Balanced Salt Solution (HBSS) was determined via nephelometry. The ability of NaF to induce cytotoxic effects was investigated in HEK293‐NIS‐LV and HEK293‐Mock‐LV cells up to 300 μM under conditions mirroring those of the uptake assay. Viability was assessed using a resazurin‐based cytotoxicity assay.

HEK293‐NIS‐LV and HEK293‐Mock‐LV cells were cultured in DMEM 4.5 g/L glucose at 37°C and 5% CO₂ and were plated onto standard 96‐well tissue culture plates at 1 × 10⁵ cells per well. Before treatment, the medium was removed, and the cells were washed with prewarmed assay buffer, HBSS. The cells were then preincubated at 37°C for 30 min in assay buffer containing 100 nM–300 μM NaF. Results were compared to solvent control (SC) or 24 nM–100 μM potassium perchlorate as a reference inhibitor. Following preincubation, solutions were removed and the uptake experiments were performed in a fresh dosing solution of 100 μL of HBSS containing the probe substrate, sodium iodide, and either NaF, potassium perchlorate, or SC for 3 min. After incubation, cells were washed twice with 100 μL of cold HBSS to stop NIS activity. Cells were then lysed with 50 μL/well purified water and incubated at 4°C for 5 min. Then the assay plates were centrifuged at 4°C for 20 min with 2250g. Following centrifugation, 10 μL samples were transferred from the assay plates to clear flat‐bottom plates containing 40 μL purified water per well. The Sandell–Kolthoff reaction was carried out on this plate after 1 min via addition of sodium arsenite and acidic ammonium cerium (IV) sulfate solutions. The measurement of resulting iodide concentration was conducted as follows. First, 40 μL/well purified water was added per well to which 10 μL/well samples (and 50 μL purified water for Abs. ctrl and 10 μL iodide solution for total in triplicates) were transferred to the detection plate. Afterward, 50 μL/well 25 mM sodium arsenite solution and 50 μL/well 25 mM acidic ammonium cerium (IV) sulfate solution was added to each well and mixed. The detection plate was incubated for 1 min at room temperature and absorbance was measured at 415 nm. IC50 is defined as the concentration of a substance required to inhibit maximal activity by 50%.

2.3. Steroidogenesis Assay

The H295R steroidogenesis assay was performed as described in OECD guideline TG 456 (OECD ²⁰²³). Before carrying out the test, the solubility of NaF in the assay medium was determined and the maximum soluble concentration was used as the maximum treatment concentration for the studies. Briefly, H295R cells were cultured in Dulbecco's modified Eagle's medium with nutrient mixture F‐12 Ham supplemented with 1% ITS + Premix, 2.5% Nu‐serum and 0.1% penicillin/streptomycin with 5.0% ± 0.5% CO₂ in air at 37.0°C ± 1.0°C. Twenty‐four hours after seeding, when the cells reached 50%–60% confluency, they were treated, in triplicate, with: medium alone, SC, forskolin (1–10 μM), prochloraz (100 nM–1 μM), or NaF (1 nM–316 μM). After 48 h, the experiment was terminated and the media were removed and split into two equal amounts and stored at −80°C for hormone analysis. Estradiol concentrations were determined using an ELISA kit (Enzo Life Sciences BVBA, Brussels, Belgium). Testosterone concentrations in the samples were determined by UPLC‐MS/MS. Cell viability for the treated cells was determined using 3‐[4,5‐dimethylthiazol‐2‐yl]‐2,5‐diphenyl tetrazolium bromide (MTT). No hormone analysis was performed in medium obtained from wells that showed viability less than 80%. For each experiment, a quality control plate was prepared.

To evaluate the relative increase or decrease in chemically altered hormone concentration, results were normalized to the mean SC value for each assay, and results were expressed as changes relative to SC in each exposure plate. Before conducting statistical analysis for each assay, the assumptions of normality and variance homogeneity were evaluated. Normality was evaluated using a Shapiro–Wilk test. Variance homogeneity was evaluated using a Levene's test. In case of variance homogeneity, differences between test material treatments and SCs were analyzed using the Dunnett's test. Differences were considered significant when p ≤ 0.05. Statistical evaluation was performed based on average values for each well that represented independent replicate data points.

Interpretation of the results followed the OECD guideline 456, which includes a prediction model based on a minimum of two and maximum of three valid runs/experiments. Briefly, a test material was judged to be positive if the fold induction was statistically different (p ≤ 0.05) and above the 1.5‐fold threshold from the SC at two adjacent concentrations in at least two independent runs. A test material was judged to be weak positive if the fold induction was statistically different (p ≤ 0.05) and above the 1.5‐fold threshold from the SC at one concentration in at least two independent runs. A test material was judged to be negative following two independent negative runs, or in three runs, comprising two negative runs and one equivocal or positive run. If the data generated in three independent experiments did not meet the decision criteria described above, the experimental results were deemed to not be interpretable.

2.4. Receptor Binding and Activity Assays

The ability of NaF to interact with estrogen receptors (ERα, ERβ), thyroid peroxidase (TPO), thyroid hormone receptors (TRα, TRβ), aromatase, steroid 5α reductase, progesterone receptors (PRs), and androgen receptors (ARs), PPARs (PPARα, PPARγ), constitutive androstane receptor (CAR), pregnane X receptor (PXR), aryl hydrocarbon receptor (AhR), retinoic acid receptor alpha (RARα), and glucocorticoid receptors was evaluated.

The solubility in each of the three assay buffers (50 mM Tris–HCl (pH 7.4)/1% DMSO, 8 mM MOPS pH 7.0, 0.2 mM EDTA, 10 mM MnCl₂, 10 mM magnesium acetate/2% DMSO, or 50 mM Tris pH 7.5, 0.1 mM EGTA, 0.1 mM Na₃VO₄, 10 mM magnesium acetate/2% DMSO) in 96‐well plates was confirmed using laser‐based nephelometric detection of light‐scattering at 635 nm. Initially, NaF was tested at a single concentration (10 μM) in duplicate, along with the appropriate positive controls for each receptor. If positive during the screening, multiple concentrations would then be tested. For the single‐concentration screening, NaF was tested at 10 μM in duplicate, along with respective positive reference chemicals.

Results from the assays were interpreted according to the contract research organization's protocols as follows: results showing either an inhibition or stimulation higher than 50% were considered to represent significant effects and were deemed to be a “hit,” which were followed up with multi‐concentration assays. Inhibition (or stimulation) between 25% and 50% is indicative of weak to moderate effects, and since this is a range where more inter‐experimental variability can occur, no follow‐up was undertaken. Results for chemicals showing an inhibition (or stimulation) lower than 25% were not considered significant and are attributable to variability of the signal around the control level. It should also be noted that high negative values (≥ 50%) can sometimes be obtained with high concentrations of test compounds and would be generally attributable to non‐specific effects of the test compounds in the assays.

2.5. Data Analysis

GraphPad Prism 10.5 (GraphPad Software Inc., Boston, MA) was used for data analysis and the generation of figures presented in this manuscript.

3. Results

3.1. NaF Does Not Alter Uptake of Iodide via the Sodium Iodide Symporter

To determine if NaF interferes with iodide uptake, and by extension thyroid hormone formation, a NIS assay was conducted, the results of which are shown in Figure 1. Results for the viability assay (data not shown) showed that NaF did not affect the viability of the HEK293‐NIS‐LV and HEK293‐Mock‐LV cells up to maximum soluble concentration of 300 μM. Furthermore, NaF was also shown not to interfere with the Sandell‐Kolthoff reaction up to the maximum tested dose of 300 μM (results not shown).

FIGURE 1.

NaF does not impact NIS‐mediated transport of NaI in uptake transporter assay. HEK293‐NIS‐LV were treated with SC, reference inhibitor potassium perchlorate at 0.024–100 μM or NaF at 0.14–300 μM, respectively. Relative NaI levels were then measured and plotted for reference inhibitor, potassium perchlorate (A), and NaF (B). Potassium perchlorate IC50 value = 0.9091 μM.

The in vitro inhibition potential of NaF and potassium perchlorate (reference inhibitor) with the human NIS was tested at concentrations of 0.14–300 μM NaF or 0.024–100 μM potassium perchlorate. As shown in Figure 1B, NaF did not inhibit the NIS‐mediated sodium iodide accumulation up to concentrations of 300 μM NaF. The reference inhibitor, potassium perchlorate, however, inhibited the NIS‐mediated sodium iodide accumulation (Figure 1A) by 92% at 100 μM, with a calculated IC50 value of 0.9091 μM potassium perchlorate. Thus, NaF is not a NIS inhibitor in vitro up to 300 μM.

3.2. NaF Does Not Alter Thyroid Peroxidase (TPO) Activity or Bind to TRα or TRβ

To determine if NaF interferes with TPO activity or binding to TRα or TRβ, activity and binding assays were undertaken. Results from the assays, shown in Table 1, show that at 10 μM NaF, no interference with TPO activity (measured via the conversion of fluorogenic peroxidase substrate Amplex UltraRed to Amplex UltroxRed) was observed. However, for the positive control, Methimazole, a 50% reduction was observed at 25 nM. Similarly, no binding was observed for NaF at 10 μM with TRα or TRβ. The positive controls, triiodothyronine, Liothyronine, and NH‐3 yielded IC and EC50 values of 0.034 nM, 0.15 nM, 0.12 nM, and 0.16 μM, respectively.

TABLE 1.

Impact of NaF exposure on thyroid pathway targets TPO, TRα, and TRβ.

Target	% inhibition (10 μM NaF)	Result	Reference EC50 or IC50 values
TPO	−3.67	Negative	Methimazole: 25 nM
TR (nonselective) rat thyroid hormone NHR binding	−2.36	Negative	Triiodothyronine: 0.034 nM
TRβ human thyroid hormone NHR binding agonist	0.04	Negative	Triiodothyronine: 0.15 nM
TRα human thyroid hormone NHR functional agonist coactivator	0.02	Negative	Liothyronine: 0.12 nM
TRα human thyroid hormone NHR functional antagonist coactivator	−2.52	Negative	NH‐3: 0.16 μM

3.3. NaF Does Not Interfere With the Steroidogenesis Pathway

The effects of NaF on pooled H295R steroidogenesis assay results are shown in Figure 2. As shown in Figure 2A, no significant changes were observed in the measured levels of estradiol following administration of NaF between the concentrations of 1 nM and 316 μM, with clear significant changes in the positive controls forskolin and prochloraz, tested at 1 and 10 μM and 0.1 and 1 μM, respectively. Some variability was observed between experiments, with estradiol showing a minor but statistically significant decrease at 10 nM to 316 μM in one experiment, and a minor but statistically significant increase at 1 μM in another experiment. It should be noted that the perturbation was not dose‐responsive and had opposite effects from experiment to experiment. In both cases, the minor statistically significant changes remained below the 1.5‐fold threshold when compared to the solvent control. Per the OECD guideline decision matrix, the results of both experiments were deemed ’negative.’

FIGURE 2.

NaF does not impact levels of estradiol or testosterone in the steroidogenesis assay. H295R cells were treated with solvent control, positive controls forskolin or prochloraz, or NaF at 0.001–316 μM. After 48 h, estradiol (A) and testosterone (B) were measured. Results from Experiments 1 and 2 are shown as fold change relative to the solvent control. *p ≤ 0.05, **p ≤ 0.01, and ****p ≤ 0.0001.

Similar to the lack of effects demonstrated for estradiol, treatment with NaF did not induce changes in testosterone levels between the concentrations of 1 nM and 316 μM NaF. Statistically significant changes in the positive controls forskolin and prochloraz were achieved in both experiments (Figure 2B). A minor but statistically significant decrease in testosterone was observed at 316 μM in one experiment, but not the other. No change in testosterone was observed that met the 1.5‐fold threshold when compared to the SC. Per the OECD guideline decision matrix, the results of both experiments were deemed “negative.”

Overall, NaF did not affect testosterone or estradiol synthesis, with a No Observed Effect Concentration (NOEC) of 316 μM NaF, which was the highest soluble, non‐cytotoxic concentration tested.

3.4. NaF Exposure Does Not Alter Aromatase or Steroid 5α Reductase Activity or Bind to the ER, PR, AR

In order to determine if NaF interferes with aromatase, steroid 5α reductase activity or binding to ERα, ERβ, PR, or AR, activity and binding assays were undertaken. Results from the assays, presented in Table 2, show that at 10 μM NaF, no interference with aromatase activity (as a measure of the formation of estradiol from testosterone) was observed; however, for the positive control, Letrozole, a 50% reduction was observed at 0.94 nM. Similarly, no interference with 5α reductase activity (as a measure of the disappearance of testosterone) was observed following exposure to 10 μM NaF. For the positive control, dutasteride, a 50% reduction was observed at 4.8 nM. Similarly, no binding was observed for NaF at 10 μM, with ERα or ERβ. The positive control, Diethylstilbestrol, showed a 50% reduction in binding at 0.26 and 1.1 nM, respectively. In addition to the negative results for ER, no specific binding was observed for NaF at 10 μM with PR or AR, while a 50% reduction was observed for the control compounds Promegestone at 0.67 nM or testosterone and 17β‐estradiol at 1 and 21 nM, respectively.

TABLE 2.

Impact of NaF exposure on aromatase, steroid 5α‐reductase activity and ER, PR, or AR binding.

Target	% inhibition (10 μM NaF)	Result	Reference EC50 or IC50 values
Aromatase	5.23	Negative	Letrozole: 0.94 nM
Steroid 5α reductase	2.87	Negative	Dutasteride: 4.8 nM
ERα	−18.1	Negative	Diethylstilbestrol: 0.26 nM
ERβ	11.7	Negative	Diethylstilbestrol: 1.1 nM
PR	4.6	Negative	Promegestone: 0.67 nM
AR	−6.09	Negative	Testosterone: 1 nM 17‐β‐estradiol: 21 nM

3.5. NaF Does Not Bind to PPARα, PPARγ, AhR, CAR, PXR, RARα, PR, or GR Receptors

We also conducted binding and coactivator studies for non‐EATS or indirectly EATS linked targets including PPARα, PPARγ, AhR, CAR, PXR, RARα, or GR. Results from these assays, depicted in Table 3, show that at 10 μM NaF, no hits were observed, while all the positive control substances had the expected effects.

TABLE 3.

Impact of NaF exposure on non‐EATS or indirectly EATS linked targets.

Target	% inhibition (10 μM)	Result	Reference EC50 or IC50 values
CAR coactivator assay (antagonist)	−0.5	Negative	Clotrimazole: 1.1 μM
CAR coactivator assay (agonist)	1.0	Negative	CITCO: 44 nM Clotrimazole: 0.14 μM
PXR	−4.2	Negative	T0901317: 12.0 nM
AhR	0	Negative	MeBIO: 4.8 nM
PPARα	−1.9	Negative	GW 7647: 120 nM
PPARγ	−6.0	Negative	Rosiglitazone: 12.0 nM
RARα	−4.9	Negative	AM580: 3.0 nM
GR	2.5	Negative	Dexamethasone: 4.3 nM

4. Discussion

We investigated the effects of NaF on key EATS and non‐EATS targets including the NIS, TPO, thyroid hormone receptors, steroidogenesis pathway, aromatase, steroid 5α reductase, estrogen, progesterone, and androgen receptors, PPARα, PPARγ, CAR, PXR, AhR, RARα, and glucocorticoid receptors.

Although historically much of the focus on excessive exposure to fluoride has been on dental and skeletal fluorosis, concerns have been raised recently about potential neurodevelopmental effects at exposures that exceed global drinking water limits of 1.5 mg/L, especially in regions with endemic fluorosis such as China, India, and Iraq. One potential endocrine‐mediated MOA for fluoride neurodevelopmental toxicity is impaired thyroid function (Hall et al. 2023) since maternal thyroid disruption can lead to neurodevelopmental toxicity (Crofton et al. 2019; Demeneix ²⁰¹⁹). The epidemiological literature has yielded inconsistent findings from study to study where no change, decreases or increases in T3, T4, and TSH have been reported (Michael et al. 1996; Hosur et al. ²⁰¹²; Barberio et al. ²⁰¹⁷; Hall et al. ²⁰²³). Inconsistent findings have been reported in animal studies as well, particularly, in ones at very high exposures (Wang et al. 2009; McPherson et al. ²⁰¹⁸), where exposures from the food and elsewhere are poorly accounted for (McPherson et al. ²⁰¹⁸).

To determine whether fluoride has the ability to affect thyroid hormone synthesis or function, we evaluated the function of the NIS, biosynthesis of T3 and T4, and interference with thyroid hormone receptor function. This work was conducted to mechanistically bolster well conducted neurodevelopmental studies (McPherson et al. ²⁰¹⁸) and probe whether a biologically plausible link can be made between the positive findings in the peer‐reviewed literature and an endocrine‐mediated mode of action.

It has been hypothesized that fluoride may compete with iodide for transport via the NIS. Iodine deficiency disorders (IDDs) can lead to severe and irreversible cognitive and physical impairments. Several chemicals such as perchlorate, thiocyanate, and nitrate have been found to interfere with NIS (Serrano‐Nascimento and Nunes ²⁰²²). Only one study had been conducted previously on the effects of F on NIS, where the substrate specificity of NIS (Eskandari et al. 1997) was compared in NIS cRNA‐injected oocytes for a number of anions including fluoride and iodide. Fluoride had a nominal effect but much less than iodine or any other anions tested and was not examined further. No experimental data have since been identified in the published literature to support a competitive interaction between these two ions, as confirmed in the most recent 2025 EFSA review (EFSA ²⁰²⁵). In this study, we investigated whether fluoride has the potential to interfere with iodide uptake via the NIS, using a NIS uptake transporter inhibition assay. Results showed that treatment with NaF did not alter iodide uptake up to the maximum soluble concentration of NaF, 300 μM. This concentration is well in excess of levels that would induce extreme skeletal fluorosis in areas with endemic fluoride ground water contamination. Therefore, it was concluded that it is highly improbable that NIS‐mediated thyroid disruption would occur when exposed to F.

Another potential mechanism for disruption in thyroid hormone synthesis is inhibition of TPO, which catalyzes the iodination of the tyrosine residues of thyroglobulin to form T3 and T4. Compounds that interfere with this process include 6‐propyl‐2‐thiouracil (PTU) and isoflavones such as genistein (Diamanti‐Kandarakis et al. ²⁰⁰⁹). Clinical manifestations of TPO inhibition can include a reduction in circulating levels of T3 and T4, and a corresponding increase in the circulating levels of TSH. In addition to TPO, alterations in binding to the thyroid hormone receptor, TRβ in the pituitary, can influence the negative feedback regulation of the HPT axis controlling TSH production and thus T4 and potentially T3 levels. Results from the activity and binding assays showed that at exposures of NaF that are above normal fluoride plasma or serum concentrations, no alterations in TPO activity or TRα or TRβ binding were observed. In summary, there are no plausible thyroid‐mediated mechanisms that are affected by relevant concentrations of fluoride.

Reports of effects on steroidogenesis and potential for ER, PR, and AR‐mediated effects can also be found in the literature, suggesting that levels that approximate or exceed the levels of fluoride found in areas with endemic ground water contamination may lead to perturbations in levels of the steroid hormones estradiol and testosterone (Narayana and Chinoy 1994; Niu et al. ²⁰¹⁵; Dhurvey and Thakare ²⁰¹⁶; Li et al. ²⁰²³). There have also been studies showing no effects, such as a study conducted by Sprando et al. (1997), which showed no effect on testosterone levels in male rats or their offspring when exposed to 16 mg fluoride/kg/day as NaF in the drinking water for 14 weeks or during gestation, lactation and for 14 weeks after weaning, respectively (Sprando et al. 1997). In addition to the above studies, other high‐quality studies have been conducted on NaF. Collins et al. (2001) reported the results of a 3‐generation study in which Sprague–Dawley rats were exposed to NaF via the drinking water at concentrations up to 250 ppm. This corresponds to an actual dosage of 28.4 mg NaF/kg bw/day. The study reported no adverse effects on reproduction throughout the three generations. Mating, fertility and survival indices were not affected. The NOAEL for fertility was thus set at 28.4 mg NaF/kg bw/day (Collins et al. 2001). To directly probe whether NaF exposure impacts steroidogenesis, we conducted an OECD TG 456 compliant H295R steroidogenesis assay and determined if NaF exposure alters steroid 5α reductase and aromatase activity. Results from the guideline‐compliant steroidogenesis assay showed that exposure to NaF up to 316 μM (the solubility limit) and at a dose that is well in excess of exposures that would elicit dental and skeletal fluorosis did not alter the biosynthesis of estradiol or testosterone in H295R cells. These negative results were further supported by the negative findings in the steroid 5α reductase and aromatase activity assays. Based upon the above findings, exposure to NaF did not directly target or disrupt the steroidogenesis biosynthetic pathway, even at very high exposures.

In addition to probing the steroidogenesis pathway, we also investigated the potential for NaF to bind to ERα, ERβ, PR, and AR. Although not directly mentioned as a potential direct target of fluoride in the literature, the ability of NaF to bind to these hormone receptors was investigated since such activity can lead to a broad range of effects, such as negative feedback of steroidogenesis, which has been mentioned in the literature for NaF as the potential for decreased testosterone levels at high exposures. Results from the binding assays showed that NaF does not bind to the ERα, ERβ, PR, or AR receptors. This is not surprising given the structural differences between fluoride and the known ligands for these receptors. However, it provides additional evidence that NaF exposure does not directly impact testosterone or estrogen function.

The ability for NaF to bind or alter binding to additional endocrine targets was investigated. These additional targets included: PXR, CAR, AhR, PPARα, PPARγ, RARα, and GR. Although not previously referenced in the literature as a potential cause for fluoride toxicity, these targets are key regulators of the hormone and xenobiotic sensing machinery impacting metabolism, inflammation, immune response, hormone and energy homeostasis, which have indirect effects on the endpoints of concern ie thyroid, estrogenic, androgenic, and steroidogenesis and they more broadly fall into non‐EATS targets of interest.

CAR and PXR play critical roles in xenobiotic sensing via the regulation of Phase I through Phase III metabolism in the liver and intestine. Importantly, these orphan nuclear hormone receptors play a key role in the metabolism of thyroid and steroid hormones by inducing CYP2B and CYP3A family members which are responsible for the metabolism of thyroid and steroid hormones (Kretschmer and Baldwin 2005). Similarly, AhR is another key xenobiotic sensor classically known for its role regulating the expression of the CYP enzymes, particularly those in the CYP1 family. AhR has also been shown to regulate the expression of the steroidogenic acute regulatory (StAR) protein, which is a rate‐limiting factor in steroid hormone production (Sugawara et al. 2001). Alterations in the level of StAR protein have been referenced in several studies on fluoride toxicity with inconsistent results (Niu et al. 2015; Li et al. ²⁰²³). Results from the CAR functional antagonist and agonist coactivator assays, as well as cell‐based AhR nuclear receptor, and PXR binding agonist assays showed that NaF exposure at 10 μM did not bind and/or alter the function of any of the tested receptors. This provides additional support for the results seen in the steroidogenesis and thyroid assays, in that NaF is unlikely to exert either a direct or indirect effect on either thyroid or steroid metabolism. Additionally, we showed that NaF does not bind with RARα, which plays a critical role in germ cell development during spermatogenesis, along with other key biological processes (Akmal et al. 1998).

Another key regulator of metabolism, along with cellular differentiation and development, is the PPARs. PPARs such as PPARα and PPARγ are key players along the HPG axis, where they not only have critical roles in energy metabolism but also reproductive function. For instance, some studies have shown that PPARα agonists can cause reproductive and developmental toxicity by disrupting steroidogenesis. It is thought that activation of PPARα leads to downregulation of StAR protein and disruption of cholesterol transport in mitochondria, which, as mentioned previously, is the rate limiting step in steroidogenesis. An AOP is currently under review linking this disruption to male developmental toxicity (Grignard 2024). Similarly, PPARγ activation has been linked to alterations not only in lipid and glucose metabolism but also to the reduction in estradiol production, possibly by reducing the expression of aromatase by way of a PGE2/BRCA1‐mediated pathway, as seen for PPARγ agonists such as thiazolidinediones (Lebovic et al. 2013). In addition to their effects on hormone synthesis, PPARs have also been shown to crosstalk as reciprocal regulators with thyroid hormone receptors (Lu and Cheng 2010). Therefore, PPARα and PPARγ possess extremely diverse and complex biological roles, most of which are just beginning to be understood. Results from the binding assays showed that NaF does not bind to the PPARα and PPARγ receptors, even at exposures that exceed the reference limits for clinical monitoring.

Our results support the collective findings from the US FDA CFSAN three‐generation study (Collins et al. 2001), GLP chronic repeat dose toxicity study (ECHA ²⁰²⁴), and the NTP neurodevelopmental study by McPherson et al. (2018) where doses that elicited dental and skeletal fluorosis did not cause reproductive, thyroid, or neurodevelopmental toxicity. We observed no effects in the series of assays probing EATS and non‐EATS modalities. The results from this series of experiments provide key insights into the absence of effects on the studied targets at exposures comparable to or well in excess of typical exposure scenarios, thereby calling into question the biological plausibility for any endocrine disruption MOA.

5. Conclusion

In this study, we investigated the potential effects of NaF on endocrine‐mediated pathways, particularly those relevant to neurodevelopment and reproduction. These endpoints have been of particular focus over the last decade with a wide range of findings reported in the literature. However, EFSA reviewed the potential for reproductive toxicity but did not consider this a priority endpoint for fluoride in its review. Reviews by the US NTP and EFSA highlight a critical gap: the need for “Additional evidence to be generated to determine the biological activity of fluoride on molecular and cellular targets in order to support mode(s) of action of fluoride relevant to the endpoints assessed and assist in the interpretation of the biological relevance of the effects reported in human and animal studies” (EFSA ²⁰²⁵).

The primary focus of our research was to address this outage and better understand if fluoride exerts direct biological activity on molecular and cellular targets related to endocrine disruption and neurodevelopmental toxicity, via an approach consisting of receptor binding and activity assays of thyroid and other EATS and non‐EATS hormone related targets, including H295R Steroidogenesis and the NIS assay. The exposures in these assays were comparable to or well in excess of those incurred by the general public.

Results showed that no link could be made between the probed pathways and any endocrine mediated MOA, thereby calling into question whether a biologically plausible link between the adverse effects and the reported endocrine activity can be drawn, especially considering the absence of findings in several high‐quality in vivo studies either conducted by scientific agencies or under GLP conditions. The findings from this series of experiments provide key insights into the absence of effects, even at elevated exposure levels, thereby supporting the safety of fluoride under standard usage conditions for its beneficial oral effects.

Funding

The studies in this paper were funded by Procter & Gamble.

Conflicts of Interest

The authors are employed by Procter & Gamble, a company that manufactures and sells fluoride‐containing oral health products. This employment represents a financial interest that may be perceived as relevant to the subject matter of this manuscript and is disclosed as a potential source of conflicts of interest. However, our employment did not influence the results or conclusions presented in this study.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.