S. mutans UA159 (ATCC 700610) was routinely cultivated in brain heart infusion (BHI; Difco, USA) at 37 °C under anaerobic conditions (85% N₂, 10% H₂, 5% CO₂). Epigallocatechin gallate (EGCG; MedChemExpress, China) was dissolved in 10% DMSO to generate stock solutions, followed by dilution in sterile water. The final DMSO concentration was maintained at 1% in all experimental and control groups. Sodium fluoride (NaF; Beijing Pufei Biotechnology Co., Ltd., China) was used as the fluoride source, with concentrations reported based on fluoride content. For initial inhibitory screening, EGCG and fluoride were evaluated at final concentrations ranging from 0.03–2 mg/mL and 3.9–500 ppm, respectively.

The inhibitory effects of fluoride or EGCG alone were assessed using a modified crystal violet staining assay to determine effective concentration ranges for subsequent synergy testing (Zhang et al., 2019; Hirasawa et al., 2006). Overnight cultures of S. mutans were diluted 1:1000 in BHI supplemented with 2% sucrose to yield approximately 2 × 10⁶ CFU/mL. Equal volumes of fluoride and EGCG were added at the indicated concentrations and incubated anaerobically for 24 h. Biofilms were washed three times with sterile water, fixed in absolute methanol, stained with 0.1% crystal violet for ≥15 min, rinsed, and solubilized in 95% ethanol. Biomass was quantified at OD₅₉₅ (BioTek, USA). All measurements were performed in triplicate. Relative inhibition was calculated as previously described (Dzoyem et al., 2025):

The combined effects of fluoride with EGCG were assessed by the checkerboard microdilution assay reported previously with some modifications (Zheng et al., 2015), and experiments were carried out in triplicate. Biofilm formation was nearly abolished when EGCG was applied at 2 mg/mL or fluoride at 250 ppm, whereas lower concentrations resulted in progressive biomass recovery (Figures 1A,B). Therefore, EGCG concentrations of 0.125–2 mg/mL and fluoride concentrations of 15.6–250 ppm was selected for synergy testing. Briefly, two-fold serial dilutions of NaF and EGCG were prepared along the ordinate and abscissa of 96-well microtiter plates, respectively. This configuration generated a concentration matrix allowing assessment of combined treatments across a range of dose combinations. Importantly, the marginal rows and columns of the checkerboard matrix contained wells with NaF alone or EGCG alone, respectively, thereby serving as internal single-agent controls within the same experimental setup. Equal volumes (100 μL each) of EGCG, fluoride, and bacterial inoculum were combined in 96-well plates and incubated for 24 h. Biomass was quantified by crystal violet staining. Fractional inhibitory concentration index (FIC) values were interpreted as synergy (≤0.5), indifference (0.5–4.0), or antagonism (>4.0). Based on these analyses, two treatment conditions were selected:

High-concentration (HC) group: 62.5 ppm fluoride + 0.5 mg/mL EGCG (biofilm inhibition >90%, FIC = 0.5).

Low-concentration (LC) group: 15.6 ppm fluoride + 0.125 mg/mL EGCG (minimal inhibition).

Bactericidal effects under HC or LC conditions were quantified by colony-forming unit (CFU) enumeration (Chen et al., 2021). After 24 h treatment, cultures were serially diluted (10⁻¹–10⁻⁶), plated onto BHI agar, and incubated anaerobically for 36 h at 37 °C. CFU counts were obtained from three independent biological replicates.

Biofilm morphology was evaluated using scanning electron microscopy (SEM; Zeiss Sigma 300) following established procedures (Chen et al., 2021). Biofilms grown on sterile coverslips for 24 h were fixed with 2.5% glutaraldehyde for ≥3 h, dehydrated through a graded ethanol series (30–100%), freeze-dried, sputter-coated with gold, and imaged at 1000×, 5,000×, and 10,000 × magnifications in three random fields per sample.

Three-dimensional architecture was examined using confocal laser scanning microscopy (CLSM; Zeiss LSM880) with a 63 × oil-immersion objective (Chen et al., 2021). SYTO9 and PI were excited at 488 nm and 543 nm, respectively. Z-stacks were collected at 3 μm intervals under constant gain and offset settings. Three biological samples per group were analyzed, with triplicate fields acquired from each.

Water-insoluble EPS was measured using the anthrone method (Cao et al., 2020). Biofilms grown in 12-well plates were harvested into PBS, centrifuged at 6000 × g for 10 min, and washed three times to remove soluble EPS. Pellets were incubated in 0.4 M NaOH for 2 h at 37 °C to extract alkali-soluble polysaccharides. Anthrone–sulfuric acid reagent was added, heated at 95 °C for 5 min, and absorbance was recorded at OD₆₂₅. CLSM visualization of EPS was performed using Alexa Fluor 647–conjugated Concanavalin A (Wu et al., 2020), followed by SYTO9 staining and imaging on a Leica Stellaris 8 confocal microscope using a 63 × oil objective.

Biofilms were grown for 24 h in 6-well plates under control, LC, or HC conditions. After washing with PBS, biofilms were scraped and collected under aseptic conditions. Samples were divided into aliquots for transcriptomic RNA extraction, proteomic protein isolation, and metabolite preparation.

Total RNA (n = 4 biological replicates per group) was extracted, converted to cDNA, and sequenced (details in Supplementary Material). Gene abundance was normalized as FPKM. Differential gene expression analysis was performed using DESeq2 (padj ≤ 0.05, |log₂FC| ≥ 0) or edgeR (padj ≤ 0.005, |log₂FC| ≥ 1). GO (Gene Ontology) and KEGG (Kyoto Encyclopedia of Genes and Genomes) enrichment were conducted using clusterProfiler.

Protein extraction and TMT labeling were performed on four biological replicates per group (details in Supplementary Material). Mass spectrometry–based quantification was conducted using a database search against the S. mutans UA159 protein sequence dataset retrieved from the NCBI database. Differentially expressed proteins (DEPs) were identified based on thresholds of p < 0.05 and |fold change| > 1.2 or <0.83. Enrichment analyses (GO, KEGG, p < 0.05) were performed for functional interpretation.

Six replicates for each group of biofilm samples were harvested and metabolites were extracted (details in Supplementary Material). Metabolites were identified by spectral matching within 10 ppm tolerance against KEGG, HMDB (Human Metabolome Database), and LIPIDMAPS (Lipid Metabolites and Pathways Strategy) databases. Quality control–based normalization (CV < 30%) was applied prior to statistical analysis. Multivariate analyses included principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) via the metaX platform, alongside univariate t-tests to find differential metabolites (VIP > 1, p < 0.05, fold change ≥2 or ≤0.5). Differential metabolites were visualized using volcano plots and z-score normalized heatmaps. Pearson correlation and pathway enrichment analysis (x/n > y/N, p < 0.05) were performed to explore significant metabolic pathways and network alterations.

Transcriptomic, proteomic, and metabolomic datasets were integrated through pairwise Pearson correlation (|r| > 0.7, p < 0.05). Significant features were mapped onto KEGG pathways and enriched based on hypergeometric testing (p < 0.05).

Analysis was conducted using the GraphPad Software (San Diego, USA). Comparison of continuous outcomes between two groups was based on Student’s two-sample t-test. Comparison of multiple groups was based on the analysis of variance (ANOVA). Post-hoc comparisons after ANOVA were based on Dunnett’s multiple comparison procedure. Statistical significance was set at the 0.05 level.

The inhibitory activity of fluoride and EGCG against S. mutans biofilms was first evaluated using crystal violet staining. Both agents exhibited concentration-dependent suppression of biomass (Figures 1A,B). Fluoride produced pronounced inhibition at 250–500 ppm, while concentrations ≤62.5 ppm had negligible effects. EGCG nearly abolished biofilm formation at 2 mg/mL and showed a partial inhibitory effect at 1 mg/mL, whereas lower concentrations (≤0.5 mg/mL) resulted in minimal reduction of biomass, consistent with previous reports (Aragao et al., 2022; Han et al., 2021; Xu et al., 2011; Schneider-Rayman et al., 2021; Hairul Islam et al., 2020).

The checkerboard microdilution assay was subsequently employed to assess combined effects. A clear synergistic interaction was observed between fluoride and EGCG (Figures 1C,D). Notably, the combination of 62.5 ppm fluoride with 0.5 mg/mL EGCG yielded >90% inhibition and an FIC index of 0.5, confirming a synergistic relationship. Increasing either compound above this threshold did not further enhance inhibition, indicating a plateau effect. These findings guided the selection of two conditions for mechanistic studies: High-concentration (HC): 62.5 ppm fluoride + 0.5 mg/mL EGCG, Low-concentration (LC): 15.6 ppm fluoride + 0.125 mg/mL EGCG.

To further characterize the biological response to the combined treatments, planktonic growth, biofilm architecture, and EPS formation were examined. CFU enumeration revealed significant differences among the control, LC, and HC groups, with marked viability loss under HC conditions (Figure 2A).

CLSM analysis demonstrated substantial structural disruption in HC-treated biofilms (Figure 2B). Biofilms in the control group displayed dense, multilayered microcolonies, whereas HC treatment resulted in sparse, disorganized bacterial clusters with visibly reduced biomass. The LC group exhibited only minor structural alterations.

SEM imaging corroborated these observations, with untreated biofilms showing extensive extracellular matrix and tightly packed cells (Figure 2C). HC treatment nearly eliminated biofilm formation, leaving only scattered bacterial remnants and markedly diminished matrix material. LC-treated biofilms more closely resembled controls, retaining aggregated structures and intact cellular morphology. Importantly, no overt signs of bacterial lysis were detected in any group.

Consistent with these structural disruptions, HC treatment significantly reduced water-insoluble EPS as quantified by the anthrone assay (Figure 2D). CLSM visualization of EPS further confirmed these findings, showing a near-complete absence of Concanavalin A–stained EPS in the HC group (Figure 2E). Together, these results demonstrated that fluoride and EGCG acted cooperatively to impair S. mutans viability, disrupt biofilm organization, and suppress EPS production.

To elucidate molecular responses underlying these phenotypic changes, transcriptomic profiling was conducted. Differential expression analysis revealed extensive transcriptional remodeling in both treatment groups. In the HC vs. control comparison, 1,350 genes were differentially expressed (674 downregulated, 676 upregulated, Figures 3A,B). The LC vs. control comparison yielded 1,135 DEGs, including 546 downregulated and 589 upregulated genes. The heat map distinctively showed the transcriptional response of fluoride and EGCG (Figure 3C). In Venn diagram analysis, 846 DEGs were identified between the two comparisons (Figure 3D), indicating substantial overlap in transcriptional responses.

GO enrichment summarized the top 30 terms in both HC vs. control and LC vs. control comparisons (Supplementary Tables S1, S2, Supplementary Figures S1A,B). One significant GO term oxidation–reduction processes, which is relative to carbohydrate metabolism was found in LC vs. control, while no terms were significant in HC vs. control (Table 1). KEGG pathway analysis (Supplementary Figures S1C,D) revealed butanoate metabolism (smu00650) significantly enriched in HC vs. control comparison, a pathway implicated in metabolic flexibility, redox balance, and the provision of acetyl-CoA for EPS biosynthesis. Within this pathway, 13 genes up-regulated and 2 were down-regulated (Table 2). No significant pathways were observed in LC vs. control comparison.

To complement transcriptomic data and identify protein-level responses, TMT-based quantitative proteomics was performed. Principal component analysis revealed partial separation of control, LC and HC groups along PC1 (36.41% of variance) and PC2 (14.38%), although substantial overlap remained (Figure 4A), suggesting moderate but consistent proteomic reprogramming.

In total, 1,628 proteins were detected in HC vs. control comparison, with 181 significantly expressed upregulated and 95 downregulated proteins (Figure 4B). In LC vs. control comparison, 1,632 proteins were identified, including 134 significantly upregulated and 69 downregulated proteins (Figure 4C). Significant GO terms and KEGG terms were presented (Supplementary Figures S2A–D, Supplementary Tables S1–S4). It could be visualized that most DEPs were localized to the cell membrane protein in both comparisons (Figures 4D,E). Integration of DEPs with GO and KEGG enrichment analyses identified 8 proteins in HC vs. control comparison that were significantly enriched in both annotation systems (Table 3). These DEPs were classified into four functional categories: extracellular polysaccharide synthesis–related proteins, nutrient sensing and regulatory proteins, metabolic enzymes, and stress response–associated proteins. Proteins involved in extracellular polysaccharide synthesis were predominantly downregulated. Three glucosyltransferases (AAN58619.1, AAN58705.1, AAN58706.1) showed significant reduction in abundance and were annotated to glucan biosynthetic processes and glycosyltransferase activity. Levansucrase (AAN59631.1), a β-D-fructosyltransferase, was also markedly downregulated. A nutrient-sensing regulatory protein, nitrogen regulatory protein PII (AAN59296.1), annotated to cellular responses to nutrient availability, was significantly downregulated and mapped to the two-component system pathway. Two DEPs were classified as metabolic enzymes. Phosphoribosylaminoimidazolecarboxamide formyltransferase/IMP cyclohydrolase (AAN57826.1), involved in the one-carbon pool by folate pathway, exhibited a significant reduction in abundance. A conserved hypothetical protein (AAN57803.1) was also markedly downregulated and annotated to β-lactam–related catalytic activity and the two-component system. In addition, one stress response–associated protein with antioxidant activity (putative manganese-type superoxide dismutase, FeMnSOD, AAN58363.1) was significantly downregulated. In LC vs. control comparison (Table 4), DEPs including AAN58619.1, AAN58705.1, AAN58706.1, AAN59631.1, AAN59296.1, AAN58363.1 and AAN57803.1 were consistently downregulated and enriched, whereas a putative CitG protein (AAN58712.1), annotated with phosphotransferase activity for substituted phosphate groups, was significantly upregulated.

Untargeted metabolomics was performed to capture functional metabolic changes underlying transcriptomic and proteomic responses. PCA confirmed stable data quality (Figures 5A,B). In HC vs. control comparison, treating with fluoride and EGCG significantly modulated the production of metabolites (Figures 5C₁, D₁) with 513 metabolites up-regulated at positive mode (ESI⁺, Electrospray Ionization), and 274 metabolites down-regulated at negative mode (ESI⁻). The clustering heatmap and Z-score plot distinctly illustrated the metabolic profile divergence between treatment groups exposed to high and low concentrations of fluoride and EGCG when compared with control group (Figures 5E, G₁, H₁). In LC vs. control comparison, 249 metabolites were up-regulated at positive mode and 302 metabolites were down-regulated at negative mode (Figures 5C₂, D₂). The clustering heatmap was also shown (Figures 5F,G₂, H₂).

Annotation of differential metabolites followed by pathway mapping and in-depth analysis enabled the identification of key metabolic pathways closely associated with these metabolites. In HC vs. control comparison, metabolomic interrogation disclosed pronounced rearrangements within 20 key metabolites and three biosynthetic super-pathways (Table 5), containing amino-acid biosynthesis with Ribose-5-phosphate, dihydroxyacetone phosphate, l-glutamic acid 5-phosphate, alpha-isopropylmalate, N-acetyl-l-glutamate, L-2-aminoadipic acid, Sedoheptulose 7-phosphate and O-succinylhomoserine, tyrosine catabolism with homogentisic acid, 4-hydroxyphenylpyruvic acid, 2,5-dihydroxybenzaldehyde, 3-methoxytyramine, succinic acid, indole-5,6-quinone and 4-hydroxycinnamic acid and aromatic amino-acid biosynthesis with shikimate, phosphoenolpyruvate, fosfructose, d-erythrose 4-phosphate and phenylpyruvate. In LC vs. control comparison, 12 key metabolites and three core KEGG pathways were listed (Table 6). Chemical carcinogenesis was populated by a five-member signature comprising the polycyclic aromatic hydrocarbon benzo[a]pyrene, the heterocyclic aromatic amine 2-amino-1-methyl-6-phenyl-imidazo[4,5-b]pyridine (PhIP), the nitrosoarene 4-nitrosobiphenyl, the tobacco-specific nitrosamine 4-(methylnitrosamino)-1-(3-pyridyl-N-oxide)-butan-1-ol, and the genotoxic N-acetoxy-MeIQx. The phosphonate/phosphinate module exhibited elevated phosphoenolpyruvate, α-D-ribose-5-phosphate, D-ribose-1,5-bisphosphate and phosphocholine. Within the vitamin B6 network, isopyridoxal, pyridoxamine and 4-pyridoxate were selectively enriched, reflecting an adaptive acceleration of pyridoxal-phosphate-dependent transamination reactions.

Additionally, 25 overlapped pathways were crossed based on HC vs. control group and LC vs. control group, and the same metabolites were also listed (Supplementary Table S5).

To characterize the molecular alterations induced by different concentration treatments, integrated transcriptomic, proteomic and metabolomic analyses were performed, with metabolomic data acquired in both negative and positive ionization modes. In HC vs. control comparison, amino acid- and carbohydrate metabolism- related pathways were found as the most pronounced pathways. Notably, at both positive and negative modes, histidine metabolism was significantly enriched at the transcriptional level (p < 0.05), whereas corresponding proteomic and metabolomic changes were less pronounced. Moreover, at negative mode, phenylalanine, tyrosine and tryptophan biosynthesis showed significant enrichment at the metabolomic level (p < 0.05) (Table 7). Several additional pathways were detected across multiple omics layers but did not reach statistical significance (Supplementary Table S6, Figures 6A₁, A₂, B₁, B₂). For LC vs. control comparison, joint pathway enrichment analyses identified that purine metabolism was the most frequently represented pathway across omics layers at positive and negative modes (Table 8). Notably, purine metabolism exhibited significant enrichment at the transcriptional and proteomic level (p < 0.05), whereas corresponding metabolomic changes were less pronounced. Additional enriched pathways above the significance threshold were detected and shown (Supplementary Table S7, Figures 6C₁, C₂, D₁, D₂).