Ethical approval and animal care
All experimental procedures complied with the ethical guidelines of the Institutional Ethics Committee for the Use of Animals in Research (CEUA–FORP/USP, protocol no. 2022.1.7.58.0) and followed the Guide for the Care and Use of Laboratory Animals issued by the National Council for the Control of Animal Experimentation.
Twelve Wistar rats (8 females and 4 males; 7 weeks old; 150–170 g) were obtained from the institutional breeding facility and, after a brief acclimatization period, randomly assigned at the onset of gestation to four experimental groups (two females and one male per group): Control (filtered water), F (50 ppm fluoride as fluosilicic acid, H₂SiF₆), Pb (30 ppm lead as lead acetate, Pb(CH₃COO)₂·3H₂O), and Pb + F (50 ppm fluoride plus 30 ppm lead).
The animals were housed under controlled conditions (12-h light/dark cycle, 25°C) with ad libitum access to standard chow and water, and body weight was monitored weekly (Fig. S1).
The offspring were born 3–5 weeks after exposure onset and, after weaning, continued receiving the same treatments as their dams. At 30 days of age, rats were euthanized under deep anesthesia, mandibles were collected and frozen for analysis, and bone samples were obtained for Pb and fluoride quantification, with results presented in the supplementary material (Figs. S2–S3).
Tooth examination and fluorosis score assessment
After euthanasia, molars were carefully extracted from the mandibles, cleaned, air-dried, and visually inspected under a stereomicroscope. Enamel opacities and surface defects were identified, particularly on the proximal surfaces of mandibular third molars in the most affected groups. The mesial surfaces were photographed at FORP-USP using a Canon EOS Rebel T6i equipped with a 100 mm macro lens and extension tube, and ten lower third molars per group were analyzed.
To objectively assess the severity of fluoride-induced enamel alterations, a fluorosis scoring system was applied. Because the macroscopic features resembled human dental fluorosis described by Thylstrup and Fejerskov [22], a lesion score modified from the Thylstrup–Fejerskov (TF) index [22] was developed for rat third molars, considering the specific characteristics observed in this model (Table 1).
This scoring system categorized teeth according to fluorosis severity, ranging from mild opacities to marked enamel loss. Representative examples of each score are shown in Fig. 1, illustrating the gradient of opacity and structural disruption across experimental groups. All teeth were evaluated independently by two blinded examiners.
Table 1
Fluorosis Score (Modified Thylstrup–Fejerskov (TF) index).
| Score | Description |
| 0 | Normal enamel without alterations |
| 1 | Opacity affecting less than 50% of the surface |
| 2 | Opacity affecting more than 50% of the surface |
| 3 | Opacity and focal enamel loss |
| 4 | Opacity and enamel loss in bands |
| 5 | Opacity and enamel loss affecting more than 50% of the surface |
Preparation of ground sections
Undemineralized, unfixed longitudinal ground sections (80–100 µm thick) were prepared from each tooth following established protocols [23–25]. Dental slices (~ 300 µm thick) were sectioned under continuous water irrigation using a diamond disc and thinned to the final thickness using a precision grinding device and silicon carbide papers.
Final section thickness was verified at the histological sites of interest by positioning the specimens edge-on under a polarizing microscope equipped with a 20× objective and an eyepiece reticle (0.7 µm resolution). All sections were stored in 0.02% aqueous sodium azide (NaN₃) until analysis.
Mineral volume quantification
Quantitative microradiographic analysis was performed using a digital X-ray camera coupled to a high-resolution micro-computed tomography system (Skyscan 1172, Bruker, Belgium), operated at 60 kV (peak energy 10 keV), with flat-field correction, no additional filters, and a pixel size of 0.94 µm.
Each section was scanned together with an aluminum step-wedge consisting of ten high-purity foils (99.9%; ESPI Chemicals, USA), each 20 µm thick, providing a calibration range of 20–200 µm. Based on the X-ray energy, aluminum density (2.7 g/cm³), and the empirical formula and density of enamel mineral [26], linear attenuation coefficients were calculated for aluminum (70.740 cm⁻¹) and enamel mineral (134.017 cm⁻¹). Calibration curves were obtained by non-linear regression between aluminum thickness and grayscale values.
To capture spatial variation across the enamel thickness, six standardized histological sites were selected along a line parallel to the enamel prisms. Measurements were performed at 7, 15, 40, 60, 80, and 100 µm from the enamel surface, using a fixed area of 10 × 10 µm at each site (Fig. 2). Grayscale values obtained at each histological site were converted into mineral volume percentages using the Angmar equation [27], following the approach described by Gan et al. [28].
Organic and water volume quantification
The quantification of non-mineral components (organic and water fractions) was performed at the same histological sites previously used for mineral volume assessment. Measurements were carried out under water immersion using a polarizing microscope (Axioskop 40, Carl Zeiss, Germany) equipped with a 0–5 order Berek compensator and a 550 nm interference filter (10 nm bandwidth; Edmund Optics, USA). At each site, phase retardation was measured five times and averaged by a single trained examiner.
Birefringence sign was determined using a Red I retardation filter, and birefringence values were calculated from mean phase retardation adjusted for section thickness. Combined with mineral volume data, organic and water volumes were calculated according to the optical model described by Sousa, Vianna, and Magalhães [29] and subsequently validated by De Medeiros, Soares, and De Sousa [23] and Dantas et al. [25].
In addition, enamel permeability was quantified at each histological site as previously described (De Sousa et al., 2013), using the ratio of squared water volume to non-mineral volume. This approach enabled differentiation of mineral, organic, and water components based on enamel birefringence behavior under polarized light.
Statistical analyses
All analyses were performed in Rstudio software (version 4.5.2).
Sample size calculation
Sample size calculation was based on a previously published effect size (Cohen’s d of 1.4) in fluoride groups in a similar study [18]. Along with a 2-tailed significance level of 5%, a power of 80%, and a sample loss estimate of 10%, the sample size per group was 10, as calculated with the function pwr.t.test (package pwr).
Examiner reliability
Two examiners evaluated visual surface features of the samples using a scoring system. Forty samples were analyzed twice, with a time interval of 15 days. The inter- and intra-reliabilities were tested using the function cohen.kappa (pacote psych).
Descriptive and inferential statistics on component enamel volumes
Descriptive statistics of component volumes per group as a function of the distance from the enamel surface were calculated using the function describe (psych package). Considering that the solid component volumes are the main ones in the pathogenesis of hypomineralized developmental defects of enamel, the remaining inferential analyses were focused on two continuous outcomes (mineral and organic volumes). Because the distances from the enamel surface represent ordered spatial locations within the same experimental unit rather than independent observations, they were not analyzed as an isolated factor. To capture spatial variation while avoiding pseudo-replication, mineral and organic volume profiles were separately integrated across predefined enamel regions using area under the curve (ΔZ, vol%xµm) metrics, calculated by a trapezoidal rule. Enamel regions were defined from the enamel surface as: superficial (7–15 µm), central (40–60 µm), and close to dentin (80–100 µm), as well as composite regions corresponding to the outer enamel half (7, 15, and 40 µm), inner enamel half (60, 80, and 100 µm), and the whole enamel layer (7–100 µm).
The aim of the inferential analyses was the interaction between treatment and enamel region, which was tested to determine whether treatment effects varied across the enamel regions, while the mains effect of treatment and region were not interpreted in isolation. This was done using the function lmer (package lme4), with the syntax “lmer(DZ_outcome ~ treatment * region + (1 | ID))”, where the term ”1| ID”(ID = sample single identifier) avoids that different regions from the same sample contribute to multiple comparisons. The effect of the interaction between treatment and enamel region was calculated for each outcome. Then, pairwise comparisons between treatments within each enamel region were performed using model-based t-tests derived from the mixed-effects model, without adjustment for multiple comparisons [30]. Model-based estimated marginal means were obtained using the emmeans() function (emmeans package).
Pairwise contrasts between treatment groups were computed using the contrast() function (emmeans package) with the "pairwise" method, generating differences in regional ΔZ values between treatment groups while preserving the variance–covariance structure specified in the mixed-effects model. Following the recommendation for post hoc pairwise analyses planned during study design [30]. Statistical inference for these contrasts was based on t statistics derived from the fitted mixed-effects models, with degrees of freedom estimated using the Satterthwaite approximation as implemented in the lmerTest package. Cohen’s d effect size for each pairwise contrast was calculated by dividing the model-estimated difference between treatment means by the residual standard deviation of the corresponding mixed-effects model, obtained via the sigma() function (lme4 package). Confidence intervals for Cohen’s d were derived by scaling the confidence limits of the model-based contrasts. The one-tailed significance level of 5% was used in all analyses.
Inferential statistics on visual aspects of dental enamel
The effect of treatment on the visual aspect of dental enamel surface (quantified by a scoring system) was tested using the functions kruskal.test (package stats; for p value) and kruskal_effsize (package rstatix; for the effect size and its 95% confidence interval). Pairwise post-host analyses were performed using the functions pairwise_wilcox_test (package rstatix) and wilconxonR (package rcompanion), for p value, effect size, and its 95% confidence interval, respectively.