The strains used in this study were S. mutans UA159, C180-2 and their derived fluoride-resistant strains UA159-FR (kindly provided by Prof. Zhimin Zhang from Jilin University, Changchun, China) and C180-2FR (Van Loveren et al., 1989). Both fluoride-resistant strains were obtained through a step-wise procedure in laboratory (Van Loveren et al., 1989; Zhu et al., 2012). Briefly, the wild-type strains (S. mutans UA159 and C180-2) were incubated on agar plates containing increasing concentrations of NaF (maximum of 52.6 mM). Colonies which could grow with high levels of fluoride were isolated and characterized. S. mutans strains were cultured anaerobically (80% N₂, 10% CO₂, and 10% H₂) at 37°C using either Brain Heart Infusion (BHI) broth or agar.

The growth of S. mutans UA159-FR and its parental strain UA159 was characterized on BHI agar and BHI broth with different levels of sodium fluoride (NaF). For growth assay on BHI agar, 5 μL of 10¹ to 10⁵-fold serially diluted overnight cultures was spotted on BHI agar plates supplemented with 0, 2, 4, 6, 12, and 25 mM NaF. All plates were incubated anaerobically at 37°C for 3 days before recording images. For growth assay in BHI broth, overnight cultures of S. mutans UA159 and UA159-FR were diluted 1:20 in fresh BHI broth and grown until early exponential phase (OD₆₀₀ ∼0.3). NaF was then supplemented to reach final concentration of 0, 5, 10, 20, 40, and 60 mM. All cultures were incubated at 37°C and the OD₆₀₀ value was recorded every 30 min for a total duration of 20 h with SpectraMax Plus 384 (Molecular Devices, CA, United States). Both growth assay experiments were performed with triplicates.

We compared the genome sequence of S. mutans UA159 (NCBI accession: NC_004350.2) to that of the fluoride-resistant strain S. mutans UA159-FR (NCBI accession: NZ_CP007016.1) to identify SNPs. Firstly, the genomes were aligned using progressive Mauve (version 2.4.0, snapshot February 26, 2015) (Darling et al., 2004, 2010). All SNPs provided by Mauve were manually examined to confirm whether they locate in a coding sequence (CDS) or an intergenic region. The identified non-synonymous and intergenic SNPs in S. mutans UA159-FR were compared to the published SNPs in S. mutans C180-2FR (Liao et al., 2015). Figure 1A shows the scheme for this comparison. The loci with mutations in both fluoride-resistant strains were selected for further investigation. Where relevant, the effect of the consequential amino-acid substitutions on the function of corresponding proteins was predicted with SIFT¹ using default settings (Ng and Henikoff, 2003).

Expression of the genes associated with mutations was examined in S. mutans UA159 and UA159-FR with real-time PCR. Both strains were grown in BHI broth until early exponential phase (OD₆₀₀ = 0.2), late exponential phase (OD₆₀₀ = 0.8) and stationary phase (average OD₆₀₀ = 1.2). Samples were taken from each growth phase and RNA was extracted and processed as described previously (Liao et al., 2015). Primers used for tested genes are listed in Supplementary Table 1. gyrA and recA were used as reference genes (Smoot et al., 2001; Brenot et al., 2005). The 2^-ΔΔCt method was used for the log-transformed data to calculate relative gene expression (Liao et al., 2015). The experiment was performed in triplicate.

To evaluate the promoter activity alone, reporter strains were constructed for the selected promoter regions (mutp) obtained from the various S. mutans strains. Three mutp regions, including one wild-type mutp and two mutated mutp were used for the construction of the reporter strains. Only one wild-type mutp reporter construct was made because the mutp regions from S. mutans C180-2 and UA159 are identical. The mutp regions from S. mutans UA159, UA159-FR and C180-2FR were PCR-amplified with primer set mutp_gfp (Supplementary Table 1). The fragments were ligated into a shuttle vector containing a promoterless green fluorescent protein (GFP)-coding gene (pDM45) (Cormack et al., 1996). All constructs were separately transformed into S. mutans strain UA159. Later, the fluorescence intensities of overnight cultures of these reporter strains were quantified using the green fluorescence assay. The optical density of the cultures from the different strains was first adjusted to OD₆₀₀ 1.0 by diluting with BHI broth. They were then diluted in PBS and measured with BD Accuri C6 Cytometer (BD Biosciences, CA, United States) using 488 nm as excitation wavelength and 530 nm as emission wavelength. Ten thousand events were collected in each sample. UA159 containing pDM45 was used as negative control. Data were acquired and analyzed with CFlow software (BD Biosciences, CA, United States). The experiment was performed in triplicate.

Since glycolytic enzymes were found to be associated with fluoride resistance of both strain UA159-FR and C180-2FR (see Results), the activities of pyruvate kinase and enolase were evaluated for both S. mutans pairs.

To examine the activity of pyruvate kinase, late exponential bacterial cells (OD₆₀₀ = 0.8) were harvested with centrifugation, washed with 300 mM HEPES pH 7.0 (Sigma-Aldrich, St. Louis, MO, United States), and resuspended in the same buffer. The cells were then challenged with various concentrations (0, 2.5, 5, and 10 mM) of NaF in presence of 0.22 mM glucose. Milli-Q was added to the negative control group. After incubating at 37°C for 30 min, the cells were pelleted with centrifugation at 16,100 × g for 2 min. Cell-free extracts were prepared by suspending the pellets in lysis buffer (20 mM Tris–HCl pH 7.5, 0.04% Triton X-100 and complete protease inhibitor) and vigorously beating with 0.1 mm glass beads. The protein concentration in the cell-free extract was quantified using a Bradford protein assay kit (Life Science, California, CA, United States). Next, the cell-free extract was used for the pyruvate kinase assay (Yamada and Carlsson, 1975; Zoraghi et al., 2010). The reaction mix contained 60 mM Na⁺-HEPES pH 7.5, 100 mM KCl, 10 mM MgCl₂, 2 mM ADP (Sigma-Aldrich), 0.24 mM NADH (Sigma-Aldrich), 0.5 mM glucose-6-phosphate (Sigma-Aldrich) and 12 μg/ml LDH (Sigma-Aldrich). Ten mM phosphoenolpyruvate (PEP) (Sigma-Aldrich) was added to the experimental group and the same volume of Milli-Q was added to the control group. The reaction was started by adding the 10-fold diluted cell-free extract to the reaction mix and followed by recording absorbance at 340 nm at 25°C for 40 min using the SPECTRAmax PLUS 384 (Molecular Devices Corp., Sunnyvale, CA, United States). The activity of pyruvate kinase was calculated as the amount of pyruvate produced per min per mg total protein. The experiment was repeated four times.

The activity of enolase was examined using a previously established protocol (van Loveren et al., 2008). Briefly, cells from mid-exponential phase (OD₆₀₀ = 0.6) were permeabilized, incubated with 20 mM potassium phosphate buffer pH 7.0 containing 0, 0.5, 1, or 5 mM NaF for 30 min, and then used for the enolase assay. Enolase activity was measured by monitoring the formation of PEP at 240 nm (van Loveren et al., 2008). The experiments were repeated four times.

Data was analyzed with GraphPad Prism (version 5.00, GraphPad Software, San Diego California, CA, United States). Student’s t-test was performed for comparing the gene expression between two strains in each growth phase. The p-values from the comparisons were corrected for multiple testing using false discovery rate (FDR). One-way ANOVA and Turkey’s multiple comparisons test was performed to compare fluorescence intensities of three reporter strains. Two-way ANOVA was performed to compare the activities of enolase and pyruvate kinase under different fluoride concentrations in different strains. Differences were considered significant for p < 0.05.

In the absence of NaF, S. mutans UA159 had a significantly shorter doubling time (54.4 ± 1.7 min) than UA159-FR (65.4 ± 0.3 min) in BHI broth (p < 0.05, Supplementary Figure 1). The lag phase of UA159-FR was seemingly longer than that of UA159. On BHI agar, UA159-FR grew in the presence of higher fluoride concentration (25 mM) whereas the wild type UA159 stopped growing at 4 mM NaF, which confirms the fluoride resistance phenotype of UA159-FR (Figure 2). Similarly, in BHI broth, UA159-FR showed full growth with as much as 20 mM NaF while UA159 was obviously inhibited by the adding of only 5 mM NaF (Supplementary Figure 2). S. mutans C180-2FR, characterized in the previous study (Liao et al., 2015), also had a longer doubling time (in absence of NaF) and stronger ability to grow with NaF than its wild-type strain. It is worth mentioning that we tested growth inhibition instead of cell killing to evaluate fluoride resistance in S. mutans.

The comparison of the genomes of UA159 and UA159-FR identified 24 SNPs (Supplementary Table 2). This list was compared to that from C180-2FR, which was reported previously (Liao et al., 2015). Shared loci containing mutations in both fluoride-resistant strains were identified (Figure 1).

In total, three loci contained mutations in both fluoride-resistant strains. Two of them (mutp and glpFp) are intergenic regions and one (pyk) is a gene from the glycolytic pathway. The nucleotide or amino-acid sequences of these loci are shown in Figures 1B–D. Mutations in mutp from C180-2FR and UA159-FR were both located in the putative -35 element. Three downstream genes, encoding a putative mutase (mut) and two fluoride antiporters (perA and perB), are regulated by mutp (Figure 1B). The other shared mutation located in the intergenic region (glpFp), which is shared by two genes encoded in opposite directions. One gene encodes glycerol uptake facilitator protein (glpF) and the other gene encodes an x-prolyl-dipeptidyl aminopeptidase (pepX) (Figure 1C). C180-2FR had 2 SNPs in pyk which led to two amino-acid substitutions (M290V and Y419D), while UA159-FR had one SNP in pyk leading to one amino-acid substitution (E369G) (Figure 1D). Except the abovementioned loci, a SNP was identified in the enolase-coding gene (eno) of UA159-FR (Supplementary Table 2) This mutation resulted in an amino-acid substitution (T287I). Enolase catalyzes the reaction upstream of pyruvate kinase, namely the transformation from 2-phosphoglycerate to PEP. Like pyruvate kinase, it also plays an important role in glycolysis, and therefore was also examined.

SIFT prediction was performed for the amino-acid substitutions in pyruvate kinase and enolase. The Y419D substitution in pyruvate kinase of C180-2FR was predicted to be deleterious to the protein function (score 0.01). The other mutation, namely the E369G substitution, in pyruvate kinase of C180-2FR was predicted to be tolerated. The substitution in enolase of UA159-FR (T287I) was predicted to affect protein function (score 0.00) but with low confidence, as the reference sequences were reported to be not diverse enough.

The bioinformatics analysis uncovered three shared loci with mutations in the two fluoride-resistant strains. Thus, the expression of 7 genes closely related to the shared loci, mut, perA, perB, glpF, pepX, pyk, and eno, was examined in UA159 and UA159-FR. Mutations were identified either in the coding region or intergenic region of these selected genes. Figure 3 shows the relative gene expression of samples taken from late exponential phase. All three genes downstream of mutp, namely mut, perA, and perB, showed a significantly higher expression in UA159-FR than in UA159. The expression of mut in UA159-FR was 6-fold higher than in UA159 (p < 0.0005), while perA and perB showed approximately 2-fold up-regulation (p < 0.05). One of the downstream genes of glpFp, glpF, had a significantly lower expression (about 30-fold) in UA159-FR than in UA159 (p < 0.0005). The other downstream gene, pepX, showed a 3-fold higher expression in UA159-FR (p < 0.005). The other two tested genes, pyk and eno, did not show differential expression. The results of samples from early exponential and stationary phases were similar to those from late exponential phase (data not shown). The expression of these genes (except for eno) in C180-2 and C180-2FR was examined and reported previously (Liao et al., 2015). Briefly, the expression of mut, perA and perB in C180-2FR was approximately 10-fold higher than that in C180-2. The expression of glpF in C180-2FR was significantly lower (about 3-fold) than that in C180-2 in early exponential phase. The expression of pepX and pyk did not differ between C180-2 and C180-2FR.

The fluorescence intensities (FI) were quantified for the three mutp reporter strains (Supplementary Figure 3). UA159 containing mutp from either C180-2FR or UA159-FR showed significantly higher fluorescence than the reporter strain containing wild-type mutp (p < 0.0005). No statistically significant difference was found between the FI of the two fluoride-resistant mutp reporter strains.

Since both in the literature (Belfiore et al., 1989) and in our pilot study, pyruvate kinase activity was found to be severely inhibited by high glucose concentrations, we used limited glucose (0.22 mM) in our experiment. UA159 and its fluoride-resistant derivative UA159-FR showed similar pyruvate kinase activities, with all tested concentrations of NaF (Figure 4). While C180-2 exhibited detectable pyruvate kinase activity under all conditions, C180-2FR showed no activity at all, independent of the presence of NaF (Figure 4). Enolase from both UA159 and UA159-FR was inhibited by NaF. The inhibition was stronger in UA159 than in UA159-FR (Figure 5). After pre-incubation with 0.5 mM NaF, only 35% of the original enolase activity remained in UA159 (p < 0.001), while approximately 70% activity was retained in UA159-FR (p < 0.01). Increased levels of NaF led to more inhibition in enolase from UA159, with only 18% residual activity after incubation with 5 mM NaF (p < 0.001). For UA159-FR, the inhibitory effect of 5 mM NaF is similar to that of 0.5 mM NaF. The enolase activity of C180-2 and C180-2FR has been evaluated in a previous study (van Loveren et al., 2008). Enolase from these two strains was similarly inhibited by NaF.