Drug repurposing for glucosyltransferase inhibition for targeted oral biofilm disruption Authors: Fernanda S. de Lucena, Matthew G. Logan, Steven H. Lewis, Hua Zhang, Matthew Barbisan, Marcelo Y. Icimoto, Hui Wu, Dustin Higashi, Justin L. Merritt, Jonathon L. Baker, Carmem S. Pfeifer Published: 2026 ABSTRACT Microbial dysbiosis is the root cause for many diseases. The acidogenic Streptococcus mutans (S.m.) forms a exopolysaccharide-rich biofilm (EPS), mediated by species-specific glucosyltransferases (Gtf), which synthesizes both soluble and insoluble glucans directly on the bacterial surface. GtfC was selected as the primary druggable target since its activity is essential for biofilm initiation and cohesion. Commercial drugs were screened in silico against GtfC, followed by experimental biofilm assays. Two lead compounds, Radotinib and Pranlukast, significantly inhibited GtfC activity and reduced biofilm mass by up to 80. GtfC knock-out models confirmed biofilm disruption specifically via enzyme inhibition. Importantly, Radotinib selectively inhibited S.m. growth while preserving commensal species. This study identified at least two compounds capable of specifically inactivating a primary virulence factor of S.m. without inhibiting its growth, with a much lower selective pressure for drug resistance development, while simultaneously providing a growth advantage to commensal species that promote oral health. Introduction Tooth decay (dental caries) is among the most prevalent infectious diseases, affecting approximately 2.5 billion people worldwide1. Loss of mineralized structure results from an imbalance between commensal and acidogenic bacterial species, due in large part to host-dependent factors such as diet and oral hygiene2. Streptococcus mutans (S.m.), is a predominant etiological microorganism in caries development, with virulence mediated by a group of extracellular enzymes called glucosyltransferases (Gtfs), which directly catalyze both the hydrolysis of complex dietary carbohydrates into monosaccharides, as well as the production of the principal water-soluble and insoluble exopolysaccharides found within its biofilm matrix (glucans)3,4. Glucans provide coalescence and mechanical integrity to the biofilm structure, supporting its persistence in harsh environments5,6. As such, Gtfs, particularly Gtf-C, are important mediators of virulence for S.m., as demonstrated in animal models of dental caries by the profoundly reduced cariogenicity of S.m. mutants lacking the Gtf gene7. Gtfs are conserved in most streptococci, with selective specificity. For example, while S.m. predominantly expresses GtfC, other commensal bacteria produce either GtfB or -D, making this enzyme a potential druggable target for reduction of cariogenic biofilms8. This targeted approach has many advantages over more traditional broad-spectrum antibiotics such as chlorhexidine and quaternary ammonium compounds (QACs). Since the anti-biofilm mechanism does not rely upon growth inhibition, the risk of induction of antibiotic resistance is severely decreased9. Several studies have demonstrated that using, for example, QACs in skin and disinfectant products may lead to antibiotic resistance10, with potential disastrous selection of superbugs11–13. Instead, by preserving the commensal bacteria, ecology-based targeted approaches bolster endogenous anti-S.m. mechanisms14. Therefore, in the context of oral biofilms, homeostasis depends on the balance of complex ecological relationships. For example, in a reductionist model of S.m. with S. gordonii (S.g.), each of the bacteria has a by-product that impedes the proliferation of the other: S.m. is acidogenic, which is harmful to S.g. S.g., in turn, produces hydrogen peroxide, which is harmful to S.m.15. The development of disease then depends on the relative abundance of each of those species16, highlighting the necessity of considering multispecies biofilms in the development of effective therapeutic strategies against oral pathogens. Specific small molecules targeting Gtf activity have been developed de novo, with efficacy in the 25–50 micromolar range8,17. However, the process to bring drugs to market against a specific condition is burdensome, both in terms cost and timeline18,19. Therefore, the repurposing of existing drugs through computational screening (in silico docking) has been considered a novel and effective approach to identify new candidates with a much more streamlined pathway for clinical translation. Most drugs have some level of pleiotropic effects, which can result in unintended deleterious side-effects, but also potentially additional utility towards novel targets20,21. Successful drug repurposing (or reprofiling or repositioning) has already been achieved for Alzheimer’s, cancer, cardiovascular, and infectious diseases like tuberculosis and malaria22. Evaluating anti-Gtf drugs within a multi-species biofilm model is of significant interest due to the critical role of glucans in S.m. biofilm formation and cariogenesis. The complex interspecies interactions between S.m. and S.g. within the biofilm provide a more physiologically relevant context for assessing drug efficacy. Such a model could elucidate whether the presence of non-cariogenic species on its own like S.g. modulates the activity of anti-Gtf agents, either by attenuating or potentiating their effects on S.m. and the overall biofilm architecture. Insights gained from this approach could inform the development of more targeted and robust therapeutic strategies for caries prevention within the multifaceted environment of the oral microbiome. The ideal drug would fulfill the following requisites: 1. Disruption of biofilm formation, either by altered morphology or deficient attachment to surfaces; 2. Maintenance of planktonic cell viability to reduce the likelihood of antibiotic resistance (as already mentioned); 3. Absence of cytotoxicity to mammalian cells; and 4. Compatibility with the aqueous environment for easy targeting in situ23,24. The last two requisites are important for the eventual application of those drugs in the oral environment. In the present study, an in silico docking approach, which uses computational methods to estimate the interaction of a target protein with likely targets, was used to identify potential commercially available/regulatory-approved drugs that could inhibit the active sites of GtfC and selectively reduce the virulence of S.m. After initial in silico screening, biological assays were performed to validate the computational profiling and identify compounds that can disrupt biofilm formation without affecting cell viability. We tested the hypotheses that: 1. The drugs identified through in silico docking can effectively bind to the active site of GtfC, inhibit its enzymatic activity, and attenuate S.m. virulence, 2. The selected drugs will disrupt biofilm formation by S.m. without affecting the growth and viability of planktonic cells, thereby minimizing the risk of inducing antibiotic resistance, 3. The inhibition of GtfC by repurposed drugs will result in a significant reduction in insoluble exopolysaccharides (EPS), leading to compromised biofilm integrity and persistence in the oral environment, and 4. Within the reductionist dual-species model, inhibition of GtfC will alter competitive dynamics between S.m. and S.g., favoring the commensal species and suppressing S.m. virulence. Results Identification of candidate glucosyltransferase inhibitors through structure-guided drug repurposing Structure-guided screening identified small molecules targeting S.m. glucosyltransferases (Gtfs), key enzymes in EPS synthesis that underpin early biofilm development and tooth demineralization. GtfB and GtfC generate distinct glucans from sucrose, differing in the ratio of α-1,6 and α-1,3 glycosidic linkages that shape biofilm architecture25. Studies have demonstrated that inhibition of GtfC in mono-species S.m. biofilms leads to altered α-1,6/α-1,3 ratios in the glucans found within its biofilm matrix26. By targeting Gtfs, particularly GtfC due to its multifunctional role in pellicle binding and EPS production, our strategy aims to impair the architectural integrity and cariogenic potential of S.m. biofilms at multiple stages of their development (Fig. 1a). One additional advantage of this approach is that, while disruptive of the biofilm structure, it does not affect bacterial viability.Fig. 1Structure-guided virtual screening pipeline identifies small-molecule inhibitors of S.m. glucosyltransferases.a Schematic overview of biofilm development on the tooth surface. Phase 1: Formation of the acquired enamel pellicle (AEP) by salivary proteins such as amylase, statherin, proline-rich proteins - PRPs, histatins, and mucins. Phase 2: Early colonizers (Streptococcus sanguinis, S.g., S. oralis, Actinomyces spp.) adhere to AEP-coated enamel. Phase 3: Secondary colonizers, including S.m., attach and secrete Gtfs, which synthesize soluble and insoluble glucans (EPS) from dietary sucrose. GtfC binds both AEP and bacterial surfaces, promoting co-adhesion and matrix integration. GtfB primarily generates α-1,3-insoluble glucans that anchor bacteria to enamel. Phase 4: Biofilm maturation occurs in an EPS-rich, acidic environment that promotes enamel demineralization and caries progression. b Crystal structure of S.m. GtfC (PDB: 3AIC), selected as the primary docking template due to its dual functional role in early biofilm formation. c Grid box parameters (center and size) used for docking simulations targeting the active site and blind docking (blind) of GtfC. d, Crystal structure of GtfB (PDB: 8FG8) used in parallel docking to capture conserved catalytic architecture across glucosyltransferases. e Docking box coordinates and dimensions for GtfB active and allosteric sites. f Schematic representation of the screening pipeline. From an initial library of 4,574 FDA-approved compounds, 56 were selected based on docking scores, physicochemical filters, and commercial availability. These were further screened for Gtf inhibition, resulting in three prioritized hits. Biological characterization of PC and top compounds, respectively: (g) NBC, (h) Radotinib, and (i) Pranlukast, assessed for inhibition of planktonic growth, biofilm formation, morphological disruption, and bacterial viability, along with Gtf enzymatic inhibition (μM) and minimum inhibitory concentration (MIC, μM) (j). Structure-guided virtual screening pipeline identifies small-molecule inhibitors of S.m. glucosyltransferases. a Schematic overview of biofilm development on the tooth surface. Phase 1: Formation of the acquired enamel pellicle (AEP) by salivary proteins such as amylase, statherin, proline-rich proteins - PRPs, histatins, and mucins. Phase 2: Early colonizers (Streptococcus sanguinis, S.g., S. oralis, Actinomyces spp.) adhere to AEP-coated enamel. Phase 3: Secondary colonizers, including S.m., attach and secrete Gtfs, which synthesize soluble and insoluble glucans (EPS) from dietary sucrose. GtfC binds both AEP and bacterial surfaces, promoting co-adhesion and matrix integration. GtfB primarily generates α-1,3-insoluble glucans that anchor bacteria to enamel. Phase 4: Biofilm maturation occurs in an EPS-rich, acidic environment that promotes enamel demineralization and caries progression. b Crystal structure of S.m. GtfC (PDB: 3AIC), selected as the primary docking template due to its dual functional role in early biofilm formation. c Grid box parameters (center and size) used for docking simulations targeting the active site and blind docking (blind) of GtfC. d, Crystal structure of GtfB (PDB: 8FG8) used in parallel docking to capture conserved catalytic architecture across glucosyltransferases. e Docking box coordinates and dimensions for GtfB active and allosteric sites. f Schematic representation of the screening pipeline. From an initial library of 4,574 FDA-approved compounds, 56 were selected based on docking scores, physicochemical filters, and commercial availability. These were further screened for Gtf inhibition, resulting in three prioritized hits. Biological characterization of PC and top compounds, respectively: (g) NBC, (h) Radotinib, and (i) Pranlukast, assessed for inhibition of planktonic growth, biofilm formation, morphological disruption, and bacterial viability, along with Gtf enzymatic inhibition (μM) and minimum inhibitory concentration (MIC, μM) (j). The in silico docking model used here focused on the GtfB and -C active sites and surrounding allosteric pockets (Fig. 1b–e). GtfC was selected as the primary docking target due to its central role in early biofilm development and extracellular polysaccharide synthesis27. A curated library of 4,574 regulatory-approved compounds was docked against catalytically relevant sites of GtfC (PDB: 3AIC28) and GtfB (PDB: 8FG829) using AutoDock Vina via an automated virtual screening server, MTiOpenScreen30. Of the compounds screened, 56 met predefined thresholds for binding energy (<–10 kcal mol⁻¹), LogP (<6.0), and commercial. These candidates were prioritized for phenotypic screening to assess their capacity to impair Gtf-dependent biofilm formation while maintaining S.m. viability, as assessed by the survival of planktonic bacteria. Using a tiered screening pipeline (Fig. 1f), the 56 selected compounds were first evaluated for effects on S.m. planktonic growth, biofilm formation, and viability. Compounds that selectively disrupted biofilms without impairing growth were then evaluated for Gtf specificity using a colorimetric assay (pNPG) and biolayer interferometry, confirming whether biofilm inhibition reflected direct interference with glucosyltransferase activity. Only compounds exhibiting selective, non-bactericidal activity, disrupting biofilm development without compromising bacterial survival, were advanced (Fig. 1f–j). Among the top-performing compounds, Radotinib, a tyrosine kinase inhibitor approved for chronic myeloid leukemia, and Pranlukast, a leukotriene receptor antagonist used in asthma and allergic rhinitis, both induced marked biofilm disorganization and reduced biomass, and were advanced for mechanistic characterization. Nitrotetrazolium Blue chloride (NBC), a redox dye previously applied in cell viability assays, was included as a positive control given its ability to completely abolish biofilm formation. This stepwise strategy enabled the identification of virulence-targeting agents with potential to attenuate cariogenic biofilm architecture while preserving microbial homeostasis. Validation of glucosyltransferase targeting by lead compounds To validate glucosyltransferase targeting as the mechanism underlying biofilm inhibition, compound binding affinity and enzymatic inhibition were assessed using Microscale Thermophoresis (MST, Monolith), chromogenic hydrolysis assays (pNPG), and biolayer interferometry (Octet-red). MST analyses confirmed direct, dose-dependent binding of NBC and Radotinib to GtfC-mCherry, while Pranlukast showed no measurable interaction under identical conditions (Fig. 2a–c). Each compound was titrated against GtfC (1 µM) across a concentration range of 100 µM–781 nM, and normalized fluorescence changes (FNORM) were fitted to determine equilibrium dissociation constants (KD). NBC exhibited a clear sigmoidal binding curve with a KD = 2.6 × 10⁻⁷ M (95% CI: 1.6 × 10⁻⁷–4.3 × 10⁻⁷ M) and a strong signal-to-noise ratio (65.7), indicating specific sub-micromolar binding. Radotinib displayed comparable affinity (KD = 2.5 × 10⁻⁷ M, 95% CI: 1.6 × 10⁻⁸–3.9 × 10⁻⁶ M) with a moderate signal-to-noise ratio (11.8). In contrast, Pranlukast produced a flat thermophoretic response without a definable inflection point, suggesting either very weak or no detectable binding to GtfC under the tested conditions (25 °C, excitation = 100%, MST power = 40%). To assess functional consequences of binding, glucosyltransferase hydrolytic activity was quantified using the p-nitrophenyl-β-D-glucopyranoside (pNPG) assay (Fig. 2d). This colorimetric assay quantifies pNPG hydrolysis to p-nitrophenol (pNP) as a readout of early Gtf activity. All compounds reduced enzymatic activity in a concentration-dependent manner with IC₅₀ < 50 μM (NBC 4.94 μM; Radotinib 14.86 μM; Pranlukast 37.77 μM; Fig. 2e). Octet-Red analysis revealed dose-dependent binding of the compounds to GtfC. Binding affinities (KD) for GtfC were lowest for Radotinib (7.2 ± 2.2 μM) and Pranlukast (8.4 ± 3.9 μM), while NBC showed markedly weaker interaction (270 ± 40 μM) (Fig. 2e). Despite weak binding to Gtfs by BLI, NBC showed the greatest inhibition in the pNPG assay, indicating a mechanism not explained by high-affinity binding and therefore used here strictly as a positive control. Radotinib and Pranlukast displayed low-micromolar KD values for both GtfC and GtfD, consistent with direct target engagement, with particularly robust binding to GtfC, a key producer of soluble and insoluble glucans. The divergence between NBC’s weak BLI binding and strong pNPG inhibition may reflect assay-specific effects (including potential redox/optical interference), transient/low-affinity interactions, or indirect enzyme inactivation.Fig. 2Biophysical and enzymatic validation of glucosyltransferase-binding compounds.a–c Microscale thermophoresis (MST) binding curves showing normalized fluorescence changes (FNORM) of GtfC-mCherry (1 µM) upon titration with NBC (a), Radotinib (b), or Pranlukast (c) across a concentration range of 781 nM–100 µM. Data were fitted using a standard binding isotherm to derive equilibrium dissociation constants (KD). NBC and Radotinib exhibited dose-dependent binding to GtfC, whereas Pranlukast showed no detectable interaction under the tested conditions. d Representative p-nitrophenyl-β-D-glucopyranoside (pNPG) hydrolysis curves showing concentration-dependent inhibition of glucosyltransferase activity by NBC, Radotinib, and Pranlukast. e Summary of quantitative binding and inhibition parameters derived from MST (KD), biolayer interferometry (Octet-Red; KD), and pNPG assays (IC₅₀). Values represent mean ± s.d. from independent experiments. Biophysical and enzymatic validation of glucosyltransferase-binding compounds. a–c Microscale thermophoresis (MST) binding curves showing normalized fluorescence changes (FNORM) of GtfC-mCherry (1 µM) upon titration with NBC (a), Radotinib (b), or Pranlukast (c) across a concentration range of 781 nM–100 µM. Data were fitted using a standard binding isotherm to derive equilibrium dissociation constants (KD). NBC and Radotinib exhibited dose-dependent binding to GtfC, whereas Pranlukast showed no detectable interaction under the tested conditions. d Representative p-nitrophenyl-β-D-glucopyranoside (pNPG) hydrolysis curves showing concentration-dependent inhibition of glucosyltransferase activity by NBC, Radotinib, and Pranlukast. e Summary of quantitative binding and inhibition parameters derived from MST (KD), biolayer interferometry (Octet-Red; KD), and pNPG assays (IC₅₀). Values represent mean ± s.d. from independent experiments. Lead Gtf-targeting compounds reduces S.m. biofilm development without compromising viability The next set of experiments examined whether glucosyltransferase (Gtf) inhibition translated into suppression of S.m. biofilm formation and whether this effect was driven by impaired attachment, viability loss, or disruption of extracellular matrix cohesion. The drugs were added to the growth medium as indicated in the methods section as to evaluate the inhibition of biofilm formation (preventative effect). Planktonic growth was quantified by OD₆₀₀ of culture supernatants (with compound concentrations ranging from 0.1 to 50 μM), and biofilm biomass was quantified by crystal violet staining followed by acetic acid solubilization. For each drug concentration, absorbance values were normalized to the corresponding DMSO control, and media/drug-only wells (blanks) were subtracted to correct for background. Dose-dependent quantification of planktonic (Fig. 3a) and biofilm-associated (Fig. 3b) populations revealed that NBC and Radotinib significantly reduced biomass across concentrations ≥12.5 μM, with NBC consistently outperforming the others. Pranlukast exhibited more modest reductions, with only partial biofilm suppression/disorganization. To evaluate whether reductions in biomass were due to loss of cell viability, luminescence from a luciferase-expressing S.m. strain was used. NBC induced a sharp decline in reporter activity across all concentrations tested (Fig. 3c), indicating potent bactericidal effects. In contrast, Radotinib preserved viability at all concentrations (Fig. 3d). Pranlukast caused intermediate decreases in viability, particularly at 50 μM and in planktonic-associated populations (Fig. 3e). These findings were corroborated by MIC determination using optical density measurements, which revealed an MIC of 3.12 μM for compound NBC (Fig. 3f), whereas compounds Radotinib and Pranlukast displayed MICs >50 μM (Fig. 3g, h), consistent with matrix-targeting effects.Fig. 3Radotinib and Pranlukast selectively disrupt S.m. biofilm development without bactericidal activity. Quantification of planktonic (a) and biofilm-associated (b) S.m. populations following 24 h exposure to NBC, Radotinib, or Pranlukast (6.25–50 µM) relative to vehicle control. NBC reduced both planktonic and surface-attached populations in a dose-dependent manner, whereas Radotinib primarily reduced biofilm-associated biomass while preserving planktonic growth. Pranlukast exerted intermediate effects with partial reductions at higher concentrations. Luminescence from a luciferase-expressing S.m. strain exposed to NBC (c), Radotinib (d), or Pranlukast (e) showed that NBC markedly suppressed viability in both fractions, Pranlukast moderately impaired biofilm-associated viability, and Radotinib exerted negligible effects on either compartment. Minimum inhibitory concentration (MIC) assays based on OD₆₀₀ confirmed potent bactericidal activity of NBC (MIC = 3.12 µM, f), whereas Radotinib (g) and Pranlukast (h) showed no measurable inhibition up to 50 µM, consistent with matrix-targeted rather than bactericidal activity. Statistical significance determined by unpaired two-tailed Student’s t-test:P < 0.05 (*), P < 0.01 (**), P < 0.001 (***), NS: not significant. Bars and symbols represent mean ± s.d. from three biological replicates. Radotinib and Pranlukast selectively disrupt S.m. biofilm development without bactericidal activity. Quantification of planktonic (a) and biofilm-associated (b) S.m. populations following 24 h exposure to NBC, Radotinib, or Pranlukast (6.25–50 µM) relative to vehicle control. NBC reduced both planktonic and surface-attached populations in a dose-dependent manner, whereas Radotinib primarily reduced biofilm-associated biomass while preserving planktonic growth. Pranlukast exerted intermediate effects with partial reductions at higher concentrations. Luminescence from a luciferase-expressing S.m. strain exposed to NBC (c), Radotinib (d), or Pranlukast (e) showed that NBC markedly suppressed viability in both fractions, Pranlukast moderately impaired biofilm-associated viability, and Radotinib exerted negligible effects on either compartment. Minimum inhibitory concentration (MIC) assays based on OD₆₀₀ confirmed potent bactericidal activity of NBC (MIC = 3.12 µM, f), whereas Radotinib (g) and Pranlukast (h) showed no measurable inhibition up to 50 µM, consistent with matrix-targeted rather than bactericidal activity. Statistical significance determined by unpaired two-tailed Student’s t-test:P < 0.05 (*), P < 0.01 (**), P < 0.001 (***), NS: not significant. Bars and symbols represent mean ± s.d. from three biological replicates. Mechanical resilience and structural cohesion of S.m. biofilms are altered by Gtf-targeting compounds To assess the effects of Gtf-targeting compounds on biofilm structural integrity, surface adhesion, and cell viability, complementary assays were performed, again with the drugs added to the culture medium prior to biofilm formation. Crystal violet staining revealed marked differences in biofilm morphology (Fig. 4a). NBC eliminated detectable biomass at all concentrations, whereas Radotinib and Pranlukast induced concentration-dependent disorganization of surface coverage, leaving patchy and irregular remnants compared to dense, uniform DMSO controls. Mechanical robustness was next evaluated using a hydrodynamic impingement assay (at 45 psi, for 5 s). DMSO-treated biofilms resisted detachment, maintaining cohesive coverage across wells (Fig. 4b). In contrast, Radotinib treatment produced distinct clearance halos at ≥25 μM, indicative of weakened surface adhesion, while Pranlukast caused partial detachment with less pronounced effects. Quantification of detached biofilm fractions (Fig. 4c) confirmed significant increases in removal for Radotinib and Pranlukast, supporting a mode of action that destabilizes biofilm cohesion without directly killing cells. Fluorescence imaging with Live/Dead staining provided spatial resolution of viability and structural disruption (Fig. 4d). DMSO controls were uniformly green, reflecting dense and viable biofilms. NBC treatment yielded diffuse red fluorescence and cellular debris, consistent with complete loss of viability. Radotinib preserved viability but induced conspicuous voids in the biofilm (“black spots”), which expanded with increasing concentration, consistent with impaired extracellular matrix cohesion. e-h, Quantitative analysis of S.m. biofilm viability following treatment with DMSO (e), NBC (f), Radotinib (g), or Pranlukast (h). Biofilm viability was quantified from Live/Dead fluorescence images using BiofilmQ (MATLAB) and is presented as the relative proportions of live (green bars) and dead (red bars) cells for each condition and concentration. DMSO-treated biofilms showed a predominance of live cells, whereas NBC and Pranlukast treatments resulted in increased fractions of dead cells. Radotinib-treated biofilms maintained a higher proportion of live cells despite pronounced structural disruption. Data represents the mean from independent experiments.Fig. 4Structural fragility and loss of cohesion define compound-specific biofilm responses.a Biofilm morphology after 24 h exposure to NBC, Radotinib, Pranlukast, or DMSO vehicle, visualized by crystal violet staining. DMSO controls formed dense, uniform biofilms, whereas NBC eradicated biomass at all concentrations. Radotinib and Pranlukast induced progressive surface disorganization, leaving patchy and irregular remnants. b Hydrodynamic impingement assay (45 psi for 5 s) revealed robust adherence of DMSO-treated biofilms, while Radotinib and Pranlukast increased susceptibility to detachment in a concentration-dependent manner. c Quantification of biofilm detachment following impingement, expressed as the fraction of biomass removed (“halo”) relative to total biofilm. d Representative Live/Dead fluorescence images of treated biofilms, with live cells shown in green and dead cells in red. Quantitative analysis of S.m. biofilm viability following treatment with DMSO (e), NBC (f), Radotinib (g), or Pranlukast (h). Viability was quantified from Live/Dead fluorescence images using BiofilmQ (MATLAB) and is presented as the relative proportions of live (green bars) and dead (red bars) cells for each condition and concentration. Data represents mean values from independent experiments. Structural fragility and loss of cohesion define compound-specific biofilm responses. a Biofilm morphology after 24 h exposure to NBC, Radotinib, Pranlukast, or DMSO vehicle, visualized by crystal violet staining. DMSO controls formed dense, uniform biofilms, whereas NBC eradicated biomass at all concentrations. Radotinib and Pranlukast induced progressive surface disorganization, leaving patchy and irregular remnants. b Hydrodynamic impingement assay (45 psi for 5 s) revealed robust adherence of DMSO-treated biofilms, while Radotinib and Pranlukast increased susceptibility to detachment in a concentration-dependent manner. c Quantification of biofilm detachment following impingement, expressed as the fraction of biomass removed (“halo”) relative to total biofilm. d Representative Live/Dead fluorescence images of treated biofilms, with live cells shown in green and dead cells in red. Quantitative analysis of S.m. biofilm viability following treatment with DMSO (e), NBC (f), Radotinib (g), or Pranlukast (h). Viability was quantified from Live/Dead fluorescence images using BiofilmQ (MATLAB) and is presented as the relative proportions of live (green bars) and dead (red bars) cells for each condition and concentration. Data represents mean values from independent experiments. Radotinib and Pranlukast impair biofilm architecture and inhibit EPS accumulation in a concentration-dependent manner High-resolution confocal imaging after 24 h of biofilm growth in the presence of increasing concentrations (6.25–50 µM) of Radotinib and Pranlukast was performed to investigate how glucosyltransferase-targeting compounds affect the structural and biochemical integrity of S.m. biofilms. Biofilms were co-stained for bacterial biomass (SYTO9; green) and exopolysaccharides (EPS; Cascade Blue-conjugated dextran; blue). In the DMSO control, S.m. formed robust, multilayered communities with dense EPS coverage across the surface and vertical dimension (Fig. 5a–d). In contrast, Radotinib induced a progressive architectural collapse, with decreased biofilm thickness, disrupted cohesion, and the emergence of spatial gaps (Fig. 5e–h). Pranlukast led to a significantly thinner, sparsely populated biofilm with marked depletion of EPS signal and minimal vertical structuring (Fig. 5i–l). Two-dimensional distribution maps of bacterial and EPS signals (Fig. 5m, n, o, q, r, s) showed homogeneous, high-density coverage in control groups, while compound-treated biofilms displayed disrupted, heterogeneous patterns with attenuated signal intensities and EPS clustering. Quantitative surface analysis corroborated the confocal observations: S.m. surface coverage was significantly reduced at concentrations ≥12.5 µM for both compounds (P < 0.01), with Pranlukast inducing the most substantial decrease (Fig. 5u). Interestingly, EPS surface coverage remained relatively stable for Radotinib but dropped significantly in the presence of Pranlukast, particularly at the highest concentration (p < 0.01) (Fig. 5v). Biochemical quantification of water-insoluble EPS supported these findings, with both compounds inducing a concentration-dependent reduction in EPS mass at 25 and 50 µM, and Pranlukast achieving the strongest inhibition (p < 0.05) (Fig. 5x). Volumetric analysis further confirmed a significant decrease in total biofilm biomass for Radotinib at ≥25 µM (p < 0.001), without affecting planktonic viability in prior assays.Fig. 5Lead glucosyltransferase inhibitors disrupt S.m. biofilm structure and reduce exopolysaccharide production.a–l 3D renderings of S.m. biofilms imaged by confocal laser scanning microscopy (CLSM) after 24 h growth in the presence of increasing concentrations (6.25, 12.5, 25, and 50 µM) of the drugs dissolved in DMSO: DMSO controls at the same volume used to dissolve the drugs (a–d), Radotinib (e–h), or Pranlukast (i–l). Bacterial cells are shown in green and exopolysaccharides (EPS) in blue. Radotinib treatment led to progressive microcolony disaggregation and EPS disruption in a dose-dependent manner, while Pranlukast induced a more diffuse and sparse biofilm structure across all tested concentrations. m-o, Local thickness maps (in µm) of S.m. biovolume for untreated control (m), DMSO (n), and Radotinib-treated (o) groups. Yellow regions indicate increased biofilm thickness. q–s Corresponding EPS thickness maps for control (q), DMSO (r), and Radotinib (s), showing reduced matrix accumulation in treated groups. u Quantification of S.m. surface coverage across concentrations of DMSO and Radotinib. Bacterial coverage decreased from ~30% to <15% with Radotinib in a concentration-dependent manner. v Quantification of EPS surface coverage under the same conditions. EPS coverage was significantly reduced at higher concentrations of Radotinib supporting glucosyltransferase inhibition. x Water-insoluble EPS quantification by phenol-sulfuric acid method for DMSO, Radotinib, and Pranlukast treatments. Radotinib reduced EPS biomass significantly at ≥12.5 µM, with near-complete inhibition at 50 µM. Pranlukast also reduced EPS, but with less consistency. Bars and symbols represent mean ± s.d. from three biological replicates (five CLSM stacks per replicate). Statistical significance determined by one-way ANOVA with Tukey’s post hoc test. Lead glucosyltransferase inhibitors disrupt S.m. biofilm structure and reduce exopolysaccharide production. a–l 3D renderings of S.m. biofilms imaged by confocal laser scanning microscopy (CLSM) after 24 h growth in the presence of increasing concentrations (6.25, 12.5, 25, and 50 µM) of the drugs dissolved in DMSO: DMSO controls at the same volume used to dissolve the drugs (a–d), Radotinib (e–h), or Pranlukast (i–l). Bacterial cells are shown in green and exopolysaccharides (EPS) in blue. Radotinib treatment led to progressive microcolony disaggregation and EPS disruption in a dose-dependent manner, while Pranlukast induced a more diffuse and sparse biofilm structure across all tested concentrations. m-o, Local thickness maps (in µm) of S.m. biovolume for untreated control (m), DMSO (n), and Radotinib-treated (o) groups. Yellow regions indicate increased biofilm thickness. q–s Corresponding EPS thickness maps for control (q), DMSO (r), and Radotinib (s), showing reduced matrix accumulation in treated groups. u Quantification of S.m. surface coverage across concentrations of DMSO and Radotinib. Bacterial coverage decreased from ~30% to <15% with Radotinib in a concentration-dependent manner. v Quantification of EPS surface coverage under the same conditions. EPS coverage was significantly reduced at higher concentrations of Radotinib supporting glucosyltransferase inhibition. x Water-insoluble EPS quantification by phenol-sulfuric acid method for DMSO, Radotinib, and Pranlukast treatments. Radotinib reduced EPS biomass significantly at ≥12.5 µM, with near-complete inhibition at 50 µM. Pranlukast also reduced EPS, but with less consistency. Bars and symbols represent mean ± s.d. from three biological replicates (five CLSM stacks per replicate). Statistical significance determined by one-way ANOVA with Tukey’s post hoc test. Radotinib disrupts S.m. biomass in a GtfC-dependent manner in both mono- and dual-species biofilms Isogenic glucosyltransferase mutants lacking either GtfB or GtfC (ΔGtfB and ΔGtfC knockouts) were used to further validate the molecular targets involved in the antibiofilm activity of Radotinib. Biofilm formation was assessed under both monospecies and dual-species conditions with Streptococcus gordonii DL-1 (S.g.). ΔgtfB biofilms, which have the intact GtfC gene, behaved the same way as the wild-type, with marked biomass reduction and morphological disintegration upon treatment with Radotinib at 25 μM (Fig. 6a, b). Remarkably, the largest biomass reduction (41%) occurred in the ΔgtfB background, where the absence of GtfB-derived insoluble glucans increases reliance on GtfC for matrix assembly. This sensitized state renders the strain more vulnerable to Radotinib-mediated GtfC inhibition, further supporting the conclusion that GtfC is the principal target of the compound. These findings suggest that Radotinib exerts its antibiofilm effects primarily through interference with GtfC activity. Interestingly, treatment of the ΔgtfC strain with Radotinib also resulted in biofilm disruption, despite the absence of GtfC (Fig. 6c). While biomass quantification revealed no statistically significant difference between DMSO- and compound-treated ΔgtfC samples, the altered morphology observed suggests a partial loss of matrix cohesion, possibly reflecting off-target effects on GtfB and/or GtfD, which remain intact in this background.Fig. 6Radotinib reduces S.m. biomass in mono- and dual-species biofilms in a GtfC-dependent manner.Biomass quantification (OD₅₆₂) of monospecies biofilms formed by wild-type S.m. UA159 (a) and isogenic knockout strains lacking gtfB (ΔgtfB, b) or gtfC (ΔgtfC, c) following 24 h treatment with DMSO (blue) or Radotinib (green, 25 µM). Biomass quantification of dual-species biofilms formed by S.m. (UA159) (d), ΔgtfB (e), or ΔgtfC (f) co-cultured with S.g. under identical treatment conditions. Violin plots depict median (white dashed line) and interquartile ranges (straight light grey line); percent biomass reduction relative to DMSO is indicated. Representative crystal violet-stained wells are shown for each condition. Treatment with Radotinib at 25 µM, selected for its ability to markedly alter biofilm morphology and reduce biomass, significantly inhibited biofilm formation in wild-type and ΔgtfB strains, but not in ΔgtfC backgrounds, indicating that GtfC is essential for the compound’s antibiofilm activity. g Time-course of biomass reduction in wild-type and ΔgtfB dual-species biofilms relative to DMSO, compared with ΔgtfC and ΔgtfBCD strains. h Proportional contributions of S.m. and S.g. to total biomass in dual-species biofilms following Radotinib treatment, showing selective loss of S.m. populations. Treatment with Radotinib at 25 µM significantly inhibited biofilm formation in wild-type and ΔgtfB strains, but not in ΔtfC backgrounds, indicating that GtfC is essential for the compound’s antibiofilm activity. *P < 0.05, **P < 0.01, ***P < 0.001, NS: not significant; unpaired two-tailed t-test. Data represents n = 3 biologically independent experiments. Radotinib reduces S.m. biomass in mono- and dual-species biofilms in a GtfC-dependent manner. Biomass quantification (OD₅₆₂) of monospecies biofilms formed by wild-type S.m. UA159 (a) and isogenic knockout strains lacking gtfB (ΔgtfB, b) or gtfC (ΔgtfC, c) following 24 h treatment with DMSO (blue) or Radotinib (green, 25 µM). Biomass quantification of dual-species biofilms formed by S.m. (UA159) (d), ΔgtfB (e), or ΔgtfC (f) co-cultured with S.g. under identical treatment conditions. Violin plots depict median (white dashed line) and interquartile ranges (straight light grey line); percent biomass reduction relative to DMSO is indicated. Representative crystal violet-stained wells are shown for each condition. Treatment with Radotinib at 25 µM, selected for its ability to markedly alter biofilm morphology and reduce biomass, significantly inhibited biofilm formation in wild-type and ΔgtfB strains, but not in ΔgtfC backgrounds, indicating that GtfC is essential for the compound’s antibiofilm activity. g Time-course of biomass reduction in wild-type and ΔgtfB dual-species biofilms relative to DMSO, compared with ΔgtfC and ΔgtfBCD strains. h Proportional contributions of S.m. and S.g. to total biomass in dual-species biofilms following Radotinib treatment, showing selective loss of S.m. populations. Treatment with Radotinib at 25 µM significantly inhibited biofilm formation in wild-type and ΔgtfB strains, but not in ΔtfC backgrounds, indicating that GtfC is essential for the compound’s antibiofilm activity. *P < 0.05, **P < 0.01, ***P < 0.001, NS: not significant; unpaired two-tailed t-test. Data represents n = 3 biologically independent experiments. In dual-species settings, similar trends were observed. Co-cultures of wild-type or ΔgtfB strains with S.g. exhibited significant biomass reduction upon Radotinib treatment (Fig. 6d, e), while S.g./ΔgtfC biofilms showed no significant differences in total biomass between control and treatment (Fig. 6f). This further supports the notion that GtfC is the principal mediator of compound susceptibility under these conditions. Notably, S.g. lacks homologs of gtfB or gtfC, confirming that biofilm changes arise solely from effects on S.m. Discussion Drug repurposing represents a transformative approach in modern medicine, offering a cost- and time-effective strategy to address critical unmet clinical needs31,32. Identifying new therapeutic applications from existing drugs circumvents the lengthy and expensive process of de novo drug development, which often spans over a decade and involves substantial financial investment3,33. In our work, we define drug repurposing as the exploration of FDA-approved drugs, originally developed for non-dental indications, for their potential to address key challenges in oral health, such as biofilm control and prevention of secondary caries. This strategy leverages the existing safety, pharmacokinetic, and manufacturing data of approved therapeutics to accelerate their application in new contexts. Our screening approach evaluated a library of structurally and mechanistically diverse drugs with established clinical use, identifying candidates that modulate S.m. behavior through reducing the production of EPS by targeting the Gtfs, under biofilm-forming conditions. Among the 56 compounds screened, nitrotetrazolium blue chloride (NBC) served as a positive control due to its robust bactericidal effects. It exhibited complete biofilm eradication at concentrations as low as 6.25 μM and had a minimum inhibitory concentration (MIC) of 3.12 μM, with effects both against biofilm and planktonic S.m. In contrast, Radotinib emerged as uniquely capable of increasing planktonic S.m. populations while simultaneously reducing surface-adherent biofilm biomass. We have demonstrated that, unlike most antibiofilm agents that rely on bactericidal activity, Radotinib preserved the viability of S.m. in both planktonic and biofilm states, suggesting its primary effect is to interfere with bacterial adhesion and biofilm maturation rather than cell death (contact-kill). This is the first time this drug is being proposed for antibiofilm applications, and in this case, specifically against oral plaque-forming bacteria. However, the antifouling of biomedical materials surfaces can be envisioned much more broadly. In the oral applications, since the antifouling mechanism relies on inactivation of the activity of extra-cellular and cell-wall bound enzymes, the clinical translational utilization of this approach may take the form or direct materials surface functionalization, as well as in the formulation of preventive mouthwashes. Moreover, and importantly, by maintaining microbial viability, Radotinib does not contribute to antibiotic resistance development, such as is the case for broad-spectrum antimicrobial agent. In this way, Radotinib offers a dual benefit: it reduces biofilm-associated pathogenicity while preserving the ecological balance of the oral microbiome. Radotinib is an orally bioavailable tyrosine kinase inhibitor (TKI) approved by the FDA for the treatment of chronic myeloid leukemia (CML), where it acts through selective inhibition of BCR-ABL and c-KIT kinases, thereby blocking downstream proliferative signaling in leukemic cells34. Clinically, Radotinib is administered at daily oral doses of 400-800 mg, achieving peak plasma concentrations in the low micromolar range (C_max ≈ 3.3–4.2 μM)35, with reported adverse effects, including rash, gastrointestinal discomfort, and elevated liver enzymes, being dose-dependent and generally manageable within the approved therapeutic window. In contrast to its systemic oncologic use, the antibiofilm activity described here relies on localized, extracellular inhibition of glucosyltransferase activity within the biofilm matrix, without cellular uptake or bactericidal effects. Although the concentrations evaluated in vitro (6.25–50 μM) overlap with or exceed reported systemic plasma levels, topical oral delivery strategies (e.g., mouthrinses, varnishes, or surface-functionalized materials) are expected to result in minimal systemic absorption and transient exposure restricted to the oral cavity. Consequently, the risk of off-target kinase inhibition or systemic side effects is anticipated to be substantially lower than with chronic systemic administration, supporting a favorable therapeutic index for localized oral applications aimed at attenuating biofilm virulence rather than eradicating the resident microbiota. From a translational standpoint, Radotinib’s antibiofilm activity may be best exploited through topical formulations, such as varnishes, rinses, or incorporation into dental materials. Because the mechanism relies on inhibition of extracellular enzymes (e.g., GtfC), rather than intracellular targets or cytotoxicity, biofilm disruption can be achieved without inducing bactericidal stress, a key advantage in preserving the commensal microbiota and minimizing the development of resistance. Moreover, its effect on biofilm detachment, rather than killing, supports its use in combination with existing physical or chemical debridement strategies, or as a prophylactic in high-risk patients. Phenotypically, Radotinib induced significant reductions in biofilm biomass, accompanied by striking alterations in biofilm architecture. Live/Dead imaging revealed that cells remained viable across all concentrations tested, but the biofilm matrix exhibited a discontinuous, heterogeneous morphology, with black voids and irregular microcolony spacing visible from 6.25 μM upward. This pattern indicates disruption of cohesive matrix formation and microcolony aggregation, hallmarks of GtfC inhibition, rather than a bactericidal effect. These observations were further supported by the luciferase viability assay, which showed no loss of luminescent signal, and by the MIC assay, which confirmed unaltered planktonic growth at all tested concentrations. Biofilm mechanical stability was also compromised by Radotinib, as demonstrated by the impingement assay. Application of shear stress via controlled water flow revealed a dose-dependent increase in biomass detachment, with up to 80% biofilm removal at the highest concentration. This weakening of adhesion is consistent with a reduction in insoluble glucan production and EPS-mediated matrix cohesion. Confocal imaging further confirmed a reduction in local biofilm thickness and total volume, along with decreased EPS coverage. Quantitative analysis revealed that Radotinib selectively reduced the water-insoluble EPS fraction, which is synthesized primarily by GtfC and is critical for biofilm structural integrity. One additional point to consider is the potential disruption of later additions to the consortia of microorganisms to the oral biofilm, including Candida albicans (Ca). Even though this was not evaluated in this study, future work should focus on the effects that the lack of glucans will exert on Ca attachment to tooth structures. Ca is not an early colonizer, and it depends on Sm to join the consortia at later stages of biofilm formation36. On that token, one could argue that by eliminating Sm from the biofilm, and more importantly, by eliminating the glucans Ca needs to attach to tooth surfaces, that the likelihood of Ca colonization also decreases. Instead, we hypothesize that the absence of Sm will favor the proliferation of commensal bacteria (S. gordonii and S. sanguis), which produce hydrogen peroxide, in turn creating additional inhospitable conditions for Sm37. Considered in concert, these findings position Radotinib as an ideal candidate for biofilm-targeted therapy. It satisfies key criteria for selective modulation of pathogenic biofilms: strong and specific binding to a validated virulence-associated enzyme (GtfC), inhibition of exopolysaccharide synthesis, disruption of aggregation and cohesion, and complete preservation of planktonic viability. This decoupling of antimicrobial and antibiofilm effects is critical for developing non-bactericidal strategies that attenuate virulence without applying selective pressure on the microbial community. The mechanism of inhibition of the biofilm in this case was demonstrated to be inhibition of glucosyl transferases, which have dual function in caries formation: 1 – hydrolysis of the complex polysaccharides from the diet into building blocks for glucans, and 2 – direct glucan synthesis. Without glucans, especially the insoluble and “sticky” kind made by S. mutans specifically, the structure of the biofilm is severely compromised, as demonstrated here by the addition of the GTF inhibitor in the culture medium, as well as by the complete absence of biofilm formation on the mutant GTF-knockout bacteria. The most innovative finding of this manuscript is exactly that the presence of the inhibitors in the wild-type culture medium led to the same biofilm morphology as the GTF-C knockout. The fact that this was accomplished with an FDA-approved, repurposed drug, makes these results highly translatable. These features make it an attractive candidate for caries prevention or adjunctive treatment, particularly in formulations aimed at rebalancing the oral microbiome rather than eradicating it. Methods Modeling and small molecule selection Structure-based virtual screening was performed using the crystallographic structure of Streptococcus mutans GtfB (PDB ID: 8FG8) and GtfC (PDB ID: 3AIC). Molecular modeling and visualization were carried out using PyMOL (PyMOL Molecular Graphics System, Version 3.0 Schrödinger, LLC.), while ligand docking was conducted through the MTiOpenScreen (MTiOS) server. An MTiOS library of regulatory-approved compounds, Drugs-lib, was screened to identify potential inhibitors. Candidate selection was based on docking affinity to the active site and predicted physicochemical properties, particularly LogP. From the initial screen, 56 compounds were prioritized for in vitro testing based on favorable docking scores and acceptable aqueous solubility profiles. Screening criteria The criteria for narrowing down the compounds obtained through virtual screening: not affecting the planktonic bacteria in the surrounding media, reducing the biofilm biomass and the pNPG activity, and changing the morphology of the biofilm, without compromising the viability of the S. mutans. This approach aims to minimize the risk of dysbiosis. The compounds were randomly assigned to the series according to the date of arrival and labeled as numbers (#.#). Enzyme kinetics The DNA fragments encoding the catalytic domains of S. mutans GtfC (residues 295-1103) were PCR-amplified using specific primer sets and cloned into the pET-sumo vector. The recombinant constructs were transformed into E. coli BL21(DE3) and cultured in LB medium until OD600 reached 0.8. Protein expression was induced with 0.1 mM IPTG at 18 °C overnight. Cell lysates were subjected to Ni²⁺ affinity purification using a HiTrap™ column, followed by gel filtration chromatography for further purification. The activity GtfC was determined using the chromogenic molecule 4-nitrophenyl b-D-glucopyranoside (pNPG) as the substrate. A reaction mixture containing 10 mL of 20 mM pNPG and 200 nM of sumo-GtfC diluted at 1:10 in phosphate-buffered saline (PBS, pH 7.4) and 0-50 mM of either DMSO, Radotinib or Pranlukast, was preincubated at 28 °C, for 30 min. The enzyme kinetics were evaluated for 120 min and read at 405 nm wavelength (SpectraMaxâ iD3 Microplate, molecular devices). Dose-response curves were generated, and half-maximal inhibitory concentration (IC₅₀) values were calculated from nonlinear regression analysis. Biofilm growth and inhibition assay A frozen stock of a derivative of wild type UA159, the bioluminescent S. mutans strain IdhRenGSm38 was streaked onto agar plates to isolate single colonies, which were then grown for 16–18 h in a 5%CO2 incubator in Todd-Hewitt Broth (THB) medium to produce an overnight culture. After this time, 20 µL of this inoculum was placed in 1 mL of fresh TH medium on a cuvette, and the optical density was evaluated in absorbance mode in a spectrophotometer at 600 nm (OD600), and the quantity of bacteria was adjusted to an OD value of 0.1, which corresponds to approximately 5 × 107 CFU/mL16. This adjusted number of bacteria was subcultured into fresh THB medium, supplemented with 1% sucrose. Serial concentrations of the 56 selected small molecules solubilized in dimethyl sulfoxide (6.25, 12.5, 25, and 50 mM) were added to a 96-well microtiter plate for biofilm growth (n = 3 per concentration of each drug). For control, the same concentrations of only DMSO were used. Biofilm was grown for 24 h under shaking agitation, on a 5% CO2 incubator, at 37 °C. The 24 h period was determined through a pilot study, in which two strains of S. mutans, UA159 (wild type) and IdhRenGSm (renilla reporter), were tested for optical density, biofilm biomass, and viability at the time points of 3, 6, 12, and 24 h. After the 6 h growth, the supernatant of each well was transferred to a new plate and the optical density of the planktonic bacteria was evaluated at 600 nm wavelength on a spectrophotometer (Spectramax iD3 Microplate reader, Molecular Devices LLC., San Jose, CA, USA). This assay was performed to assess if the compounds were being released to the surrounding medium and affecting the non-attached bacteria (planktonic). A triplicate of wells with sterile media only were used as blank, and the average value was subtracted from the average of all drugs. Biofilm biomass and morphology assessment After checking the optical density, the supernatant was removed from the wells, the plates were washed three times with ultrapure water, then blotted vigorously to remove the excess liquid. After drying for 20 min, the wells were stained with a 0.01% Crystal Violet solution, and the plates were kept static for 15 min, at room temperature. Then, the staining solution was removed, and the plates were again rinsed with ultrapure water three times. After drying, each well was photographed with a stereomicroscope (DINO lite digital microscope) for morphological analysis of the biofilm. Total biofilm biomass was quantified by resuspension of the biofilm in 30% acetic acid solution on a spectrophotometer (SpectraMax iD3 Microplate reader, Molecular Devices LLC., San Jose, CA, USA) at 562 nm wavelength, as previously described39. The biomass values were normalized by subtracting the values of 30% acetic acid only. Reporter activity (luciferase assay) The molecules that presented any effect on the biofilm were also evaluated regarding bacteria viability, for planktonic and biofilm. For luciferase assay, a derivative of wild type UA159, the bioluminescent S. mutans strain IdhRenGSm was cultivated at 96-well white plates (Lumitrac, Greiner Bio-One, Kremsmunster, Austria) for 6 h. After this time, the planktonic bacteria were removed and placed on a new well plate, and 2 mL of coelenterazine-h ethanol solution was added to each well (for 200 mL total volume of bacteria). For the biofilm viability evaluation, fresh THB medium was added to the wells, and the plate was incubated at a 5% CO2 incubator at 37 °C for 30 min, the same amount of coelenterazine-h ethanol solution was added to the wells. The bioluminescence of both planktonic and biofilm viability was then immediately evaluated by a spectrophotometer (Spectramax iD3 Microplate reader, Molecular Devices LLC., San Jose, CA, USA) at all wavelengths. Fluorescent labeling and Microscale Thermophoresis (MST) The binding affinity between the GtfC protein and various ligands was measured using a Monolith instrument (NanoTemper Technologies). The GtfC protein was labelled with mCherry tag and adjusted to a final concentration of 200 nM in PBST buffer (PBS pH 7.4 supplemented with 0.01% Tween 20). Ligands were initially prepared as 10 mM stock solutions in DMSO and serially diluted (1:2) in PBST to yield 8 concentrations ranging from 200 μM to 91.4 nM. Each ligand dilution was then mixed with an equal volume of labeled GtfC, resulting in a final protein concentration of 100 nM and ligand concentrations from 100 μM to 45.7 nM. After a 10 min incubation, samples were centrifuged at 10,000 × g for 10 min and loaded into Monolith NT.115 capillaries (NanoTemper Technologies). MST measurements were performed at 25 °C, using 40% MST power. Dissociation constants (KD values) were determined using the Kd binding model available in the MO. Affinity Analysis Software (NanoTemper Technologies). Impingement test The biofilm growth was performed as described in the inhibition assay item, however, on 24-well white plates (Lumitrac, Greiner Bio-One, Kremsmunster, Austria) in the presence of serial concentrations of either only DMSO or Radotinib and Pranlukast (same concentrations used for the inhibition assays). After the 24 h growth for the formation of a consistent biofilm, the plates were washed three times with 1X PBS, and the biofilm on the wells was stained with a 0.01% crystal violet solution. Once stained, each well was photographed with a stereomicroscope (DINO lite digital microscope), and then the biofilm adhesion strength was quantified by the microjet impingement method (MJI). This method is based on shear force created by a microjet system, that can detach biofilm from a surface40,41. For this assay, a 0.20 mm ID nozzle perpendicularly oriented to the wells in which the biofilm was cultured, delivered a DI water jet at a 4 mm height, for 5 s at a 45-psi pressure. Afterward, the wells were blot-dried, and post-impingement pictures were taken. The amount of biofilm pre- and post-impingement was quantified through ImageJ software. Colony unit formation (CFU) assay The viable colonies of S. mutans after the inhibition assay for the serial concentrations of DMSO or Radotinib were assessed by six-fold serial dilution (1:10) in fresh THB and then seeding 100 mL of each dilution onto THB agar plates (n = 3) per concentration, per compound tested), followed by 48 h incubation at 37 °C at a 5% CO2 incubator. After incubation, a digital camera captured images of the plates, and Image J software was used to count the number of viable single colonies in each agar plate. The following equation was used to calculate the CFUs per well42.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${number}\,{of}\,{colonies}\,\times \,{dilution}\,{factor}\,\times \,{original}\,{volume}\,{of}\,{sample}\,\times \,10$$\end{document}numberofcolonies×dilutionfactor×originalvolumeofsample×10 Determination of water-insoluble polysaccharides For this assay, the IdhRenGSm S. mutans strain was cultured overnight (16 h), and the OD adjustment was made the same way described in the biofilm growth/inhibition assay topic. Then, the biofilm growth was made in 24-well plates in the presence of the four concentrations of either DMSO or Radotinib, with THB media supplemented with 1% sucrose solution, for 48 h, to obtain a thicker biofilm. Incubation conditions were also 5%CO2 and 37 °C, at lower shaking. After the incubation period, the well plates were washed twice with milli-Q water, to remove the non-adherent bacteria. Then, the biofilm was resuspended in 1 mL of 1X PBS by scraping the bottom of the well and vigorously pipetting the content up and down. The scraped content was then transferred to conical tubes and sonicated for 60 seconds. 400 mL of the sonicated biofilm solution were vortexed and centrifuged at 10,000 g at 4 °C, for 10 min. The supernatant was transferred to another tube, and 1.2 mL of cold 95% EtOH was added to each tube for 30 min, then these were centrifuged, and the pellet was washed with cold 70% EtOH. 200 mL of 1 M NaOH solution was used to resuspend the precipitated polysaccharides, and the total carbohydrate was calculated using glucose as standard43. 200 mL of a 5% phenol solution was added to the sample, followed by the rapid addition of 1 mL of concentrated sulfuric acid. The tubes were left to stand for 10 minutes, then vortexed and incubated in a 30 °C water bath for 20 min. Absorbance was measured at 490 nm using a microplate reader44,45. S. mutans biofilm and EPS coverage S. mutans biofilm was grown in 96-well square glass bottom Ibidi plates (IBIDI, Germany) by adding the amount correspondent to an optical density OD600 0.1 of optical density of an inoculum solution to THB media supplemented with 1% sucrose (200 mL final volume per well), in the presence of the four concentrations used for the previous assays of DMSO or Radotinib. The biofilm growth was carried out for 24 h, on a 5%CO2 incubator at 37 °C, under lower shaking. To the bacteria solution, 1 mM of Cascade Blue (Molecular Probes, Invitrogen), a fluorophore-conjugated with dextran (401 nm excitation/419 nm emission) was added to label the EPS, and after the biofilm growth, the plates were washed three times with 1X PBS, and stained with Syto9 green (483 nm excitation/503 nm emission) to label S. mutans nucleic acids for 20 min. The same volume of PBS was placed in each well, and the images were obtained at a x20 magnification on a Confocal Laser Scanning Microscope (Zeiss Axio Observer, Carl Zeiss US, White Plains, NY, USA). Z-stacks (n = 3) were obtained to track the biofilm thickness and the coverage of both S. mutans and EPS. Representative 3D renderings of each compound/concentration were made through iMaris software, the total S. mutans or EPS coverage was measured at ZEN blue software, and the thickness was measured at COMSTAT. 3D reconstruction and measurement of the biofilm properties Quantification and 3D rendering of the biofilms in the presence of the tested compounds was performed using MATLAB, BiofilmQ and Paraview46 Briefly, the workflow consisted of (1) separating the confocal z-stacks by channel (green for biofilm and blue for EPS), (2) object-specific segmentation and thresholding, (3) parameter calculation, including localized biofilm thickness, substrate area and global biofilm properties, (4) raw data export and 3D visualization in Paraview. Paraview software allowed for detailed visualization, including the spatial distribution of biofilm structures and EPS within the 3D environment. Live/Dead assay The biofilm growth was performed the same way as described for biofilm and EPS coverage measurements. After the 24 h growth, the biofilm was stained with a mixture of 3.34 mM Syto9 (live) and 20 mM propidium iodide (dead) dyes in PBS in a 1:1 ratio (3 µL of the staining solution per mL of media). The mixture was added for 20 min before imaging, and the plate was kept static in the dark. Then, the plates were washed three times with 1X PBS, and images were obtained at x20 magnification on a Confocal Laser Scanning Microscope (Zeiss Axio Observer, CarlZeiss US, White Plains, NY, USA). Z-stacks (n = 3) were obtained, and biofilm viability was quantified using ZEN Blue (Zeiss, version 3.2) according to Eq. 1, where Igreen and Ired are the arithmetic mean intensities of live and dead cells, respectively, across selected slices of the z-stack.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\boldsymbol{Viability}}\,{\boldsymbol{Ratio}}=\,\frac{{{\boldsymbol{I}}}_{{\boldsymbol{green}}}}{{{\boldsymbol{I}}}_{{\boldsymbol{green}}}+{{\boldsymbol{I}}}_{{\boldsymbol{red}}}}$$\end{document}ViabilityRatio=IgreenIgreen+Ired Dual-species biofilm models S. mutans (UA159) and S. gordonii (DL-1) were cultured in brain heart infusion (BHI) medium under aerobic conditions (5% CO₂) at 37 °C. Overnight cultures (16–18 h) were adjusted to the same optical density and combined in fresh BHI supplemented with 1% sucrose to promote extracellular polysaccharide synthesis. Biofilms were allowed to develop aerobically for 24 h in 96-well plates. Planktonic growth was assessed by OD₆₀₀, while biofilm biomass and viability were quantified using standard crystal violet staining, respectively. Mutant knockout strains To evaluate the specificity of Gtf inhibition by candidate compounds, S. mutans UA159 wild-type and isogenic glucosyltransferase-deficient strains ΔGtfB, ΔGtfC, and ΔGtfBCD were used. The ΔGtfB and ΔGtfC strains each lack a single glucosyltransferase gene, while the ΔgtfBCD triple knockout lacks all three major Gtf enzymes responsible for insoluble and soluble glucan synthesis. Strains were provided by Dr. Hui Wu (Oregon Health & Science University). For biofilm formation assays, overnight cultures of each strain were grown in BHI broth at 37 °C in a 5% CO₂ atmosphere. Cultures were diluted 1:100 in fresh BHI supplemented with 1% (w/v) sucrose and inoculated into sterile 96-well polystyrene plates. Biofilms were allowed to develop for 24 h in the presence or absence of selected inhibitors. Following incubation, biofilm biomass was quantified using 0.01% crystal violet staining, and absorbance was measured at 595 nm after ethanol solubilization. Data were normalized to wild-type control biofilms to assess the dependence of compound activity on specific Gtf enzymes. DNA extraction from biofilms Biofilms composed of S. mutans + S. gordonii, or S. mutans knockout strains (same as above) + S. gordonii were grown in 24-well plates (1 mL BHI + 1% sucrose) for 24 h under aerobic conditions. After incubation, supernatants were removed, and biofilms were resuspended in sterile deionized water and scraped using sterile pipette tips. Pellets were collected by centrifugation (10,000 × g, 10 min, 4 °C) and stored at -80 °C. DNA extraction was performed using the ZymoBIOMICS DNA Miniprep Kit (Zymo Research). Briefly, frozen pellets were resuspended in 200 μL of nuclease-free water and combined with 750 μL BashingBead™ buffer in lysis tubes. Samples were beaten for 5 min, centrifuged, and the supernatant was processed through a series of filtration, lysis, and wash steps per the manufacturer’s instructions. DNA was eluted in 100 μL of elution buffer and stored at -20 °C. 16S gene amplification and library preparation for nanopore sequencing Barcoded 16S libraries were prepared using the Oxford Nanopore 16S Barcoding Kit 24 V14 (Oxford Nanopore Technologies, Inc.) SQK-16S114.24) following the manufacturer’s protocol. Genomic DNA (10 ng) from each sample was diluted in nuclease-free water to a total volume of 15 μL and combined with 25 μL LongAmp™ Hot Start Taq 2X Master Mix (New England Biolabs) in 0.2 mL PCR tubes. Barcodes were thawed at room temperature, briefly centrifuged, and 10 μL of the appropriate 16S barcode was added to each reaction. PCR amplification was performed using the following conditions: 95 °C for 1 min; 25 cycles of 95 °C for 20 s, 55 °C for 30 s, and 65 °C for 2 min; followed by a final extension at 65 °C for 5 min. Following amplification, 4 μL of 0.5 M EDTA was added to each reaction to stop enzymatic activity, and samples were incubated at room temperature for 5 min. DNA concentration was quantified using a Qubit fluorometer (Thermo Fisher Scientific), and equimolar pooling of all barcoded samples was performed in a DNA LoBind 1.5 mL tube (Eppendorf). The pooled library was purified with a 0.6x volume of AMPure XP beads (Beckman Coulter), washed twice with 80% ethanol, and eluted in 15 μL of Elution Buffer (Zymo Research). The eluate containing the purified DNA library was quantified by Qubit. For adapter ligation, 50 fmol of the pooled library was brought to 11 μL with Elution Buffer and combined with 1 μL of diluted Rapid Adapter (prepared by mixing 1.5 μL Rapid Adapter with 3.5 μL Adapter Buffer). The reaction was gently mixed, incubated at room temperature for 5 min, and stored on ice until sequencing. Libraries were sequenced on a MinION Mk1B device using Flongle FLO-MIN106D (R9.4.1) flow cells and running MinKNOW 6.5.15 software. Basecalling was performed within MinKNOW using the High-accuracy-v4.3.0-400 bps model with a minimum qscore of 9. Microbiome analysis of the resulting FASTQ files was performed with EPI2ME (Oxford Nanopore Technologies, Inc.; v5.2.3) using the wf-16s workflow v1.4.0 with all parameters at kept at default except Taxonomic Rank, which was set at S (species-level). Statistical analysis Data were tested for normality (Shapiro-Wilk) and homoscedasticity (Bartlett and Levene), and analyzed with Student t test, one-way ANOVA or Kruskal-Wallis, and Tukey’s test for multiple comparisons, with a significance level of 95%. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.