Computational and preliminary screening of anti-caries active compounds of Caesalpinia sappan Authors: Yongliang Jia, Dongdong Zhang, Hongxing Lu, Yanhong Li, Yuehu wang Published: 2026 ABSTRACT Dental caries is a significant global oral health problem, primarily caused by Streptococcus mutans (S. mutans). Although previous studies have indicated that Caesalpinia sappan (CS) exhibits anti-caries properties, its specific anti-caries active compounds remain largely unidentified. This study aimed to identify the potential anti-caries components of CS through an integrated strategy involving bioassay-guided isolation targeting S. mutans, UPLC-MS/MS, network pharmacology, molecular docking, and molecular dynamics simulation. Bioassay-guided isolation led to the identification of 33 compounds in fraction 5 from CS, among which brazilin showed the highest content and effectively inhibited biofilm formation and acid production by S. mutans. Network pharmacology analysis predicted 11 active ingredients in CS, including brazilin, targeting 7 potential targets related to dental caries: AKT1, EGFR, BCL2, PTGS2, MMP9, ERBB2, and HSP90AA1, as well as two pathways—nitrogen metabolism and calcium signaling. Molecular docking revealed strong binding affinities of brazilin to PTGS2, EGFR, MMP9, GtfC, GtfD, and GbpC. Further, molecular dynamics simulation revealed strong stability between brazilin and PTGS2, GtfC. Together, these findings not only clarify the anti-caries chemical constituents of CS and their potential mechanisms of action, but also provide new insights into the bioactive components and mechanisms of natural plant-derived medicines for the prevention and treatment of dental caries.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-47281-4. Introduction According to the Global Burden of Disease Study 2021, dental caries ranks as one of the most prevalent oral diseases globally1. The sustained progression of dental caries not only severely compromises oral health but also poses risks to overall systemic health2,3. Modern etiological research indicates that dental caries is a chronic oral disease resulting from dysbiosis of the oral microbiome. Currently, Streptococcus mutans (S. mutans) is still regarded as the primary cariogenic microorganism. It demonstrates strong acidogenicity and acid tolerance, facilitating enamel demineralization and driving the initiation and progression of dental caries4,5. Moreover, S. mutans can synthesize substantial amounts of extracellular polysaccharides (EPS), which assemble into structurally complex biofilms that enhance bacterial adhesion and provide protection for embedded microorganisms5,6. Although fluoride remains the most effective agent for the prevention of dental caries in global public health, its long-term use entails risks such as dental fluorosis, microbial resistance, and ecological imbalance7–10. Chlorhexidine, the main alternative, is associated with adverse effects including tooth staining, taste disturbance, and disruption of the commensal microbiota11. These limitations highlight the need for safer, next-generation anti-caries agents. In recent years, research on natural herbal medicines has garnered significant attention from the scientific community. Numerous plant extracts have demonstrated inhibitory effects against common oral pathogens. For instance, Galla chinensis, Rhodiola, and Arnebia root have been applied in the prevention of dental caries12–14. Their primary mechanisms include: (1) inhibiting the growth, acid production, and biofilm formation of S. mutans; and (2) preventing demineralization or promoting remineralization of dental hard tissues. In a previous study by our group, the inhibitory effects of six common natural plant extracts from Yunnan Province, China, on dominant oral bacteria were evaluated. It was found that the extract of Caesalpinia sappan (CSE) exhibited strong inhibitory activity against S. mutans15. However, the specific bioactive constituents responsible for these effects in CS remain to be fully elucidated. Based on the inhibitory activity against S. mutans, this study further screened potential anti-caries components through bioassay-guided isolation. By integrating network pharmacology and molecular docking, the underlying molecular mechanisms were elucidated. In conclusion, this study further explored the effective anti-caries active components in CS and the possible related mechanisms compared to the previous research conducted by our group. These findings provide a basis for research on caries prevention and contribute to public health efforts in the management and treatment of dental caries. Materials and methods Bioassay-guided isolation of compounds from CS against S. mutans Chemicals and reagents The heartwood of CS was purchased from Luosiwan market (Kunming, China). Ethyl alcohol (EtOH), petroleum ether (PE), ethyl acetate (EtOAc), n-butanol was purchased from Chuandong Chemical Co., LTD (Chongqing, China). S. mutans UA159 was provided by the Yunnan Key Laboratory of Stomatology (China). Crystal violet (CV) and sucrose were purchased from Yuanye (Shanghai, China) and Solarbio (Beijing, China). Brain heart infusion (BHI) broth and agar were from Bacton, Dickinson and Company. Plant materials and extracts preparation The CS powders were ultrasonically extracted for 30 min with different concertration EtOH (0–90%) at 60 °C. Gradient EtOH crude extract were obtained by concentrated the solvent. The 70% EtOH extract were suspended in water and then successively extracted with PE, EtOAc, and n-butanol to yield the PE-soluble, EtOAc-soluble, and n-butanol-soluble extracts, and the remaining water phase part. Then, the EtOAc extract were subjected to silica gel column chromatography (PE-EtOAc, 1:0→0:1) to yield fractions 1–10 (Fr. 1–Fr. 10). Bacterial strains culture conditions S. mutans was recovered and inoculated on BHI agar plates, and single bacterial colony was added into BHI broth grown in anaerobic chamber (10% H2, 5% CO2, and 85% N2) at 37 °C. BHI broth was replaced with BHI supplemented with 1% (w/v) sucrose (BHIS) in biofilm assays and glycolytic pH drop assay. Determination of the minimum inhibitory concentration (MIC) The antimicrobial activity of different extracts was evaluated by MIC16. Different extracts mother solutions were diluted with BHI medium to different concentrations by the doubled dilution method. S. mutans was cultured to mid-log phase and diluted with BHI medium according to the volume ratio of 1:100, and then anaerobically cocultured with different extracts at 37 °C for 24 h, the lowest concentration with no bacterial growth was defined as the MIC. The concentrations of 1/2 (1/2 MIC) and 1/4 (1/4 MIC) as the research subjects. Biofilm formation and mature biofilm eradication assays The effects of different extracts on S. mutans was analyzed using crystal violet (CV) staining17. Briefly, S. mutans was diluted using BHIS medium and co-cultured anaerobically with different extracts at 37 °C for 24 h or after cultured 24 h to add compound at 37 °C for 24 h. The biofilm was rinsed three times with sterile water. The bacterial cells were fixed with 4% (w/v) paraformaldehyde for 10 min and then stained with 0.1% (w/v) CV solution for 10 min in each well. The solution was removed and rinsed repeatedly and gently with sterile water. The dye was solubilized by adding 33% (v/v) acetic acid solvent to the well plates for 10 min. Finally, the 100 µL of solution per well was pipetted into a new 96-well plate and the optical density (OD) was read at 575 nm using a Microplate reader. Glycolytic pH drop assay The effect of different extracts on acid production for S. mutans biofilm was assessed by determining the pH of the medium18. S. mutans was co-cultured with different extracts in 24-well plates, the initial pH was adjusted to 7.0 and anaerobically incubated at 37 °C for 24 h. The decrease in pH of each well was measured using a pH meter (Mettler Toledo, Switzerland). UPLC-MS/MS analysis of the main extracts In order to further explore the composition of active components, Fr. 5 was analyzed by ultra-performance liquid chromatography coupled with tandem mass spectrometry (UPLC-MS/MS)19. The Fr. 5 powder was redissolved in methanol with a concentration of 4 mg/mL by ultrasonication, centrifuge at 13,000 rpm/min for 10 min and take the supernatant for analysis, storage time should not exceed 24 h. UltiMate 3000 UHPLC and Thremo Hypersil gold C18 (2.1 mm×100 mm, 1.9 μm) were used in liquid phase separation, the liquid phase was 0.1% formic acid/methanol (B)-0.1% formic acid/water (A), the gradient elution procedure of the solution, as detailed in Table 1. Chromatographic condition: 0–20 min, flow rate was 0.3 mL/min. Table 1Mobile phase gradient elution procedure.Time (min)AB090101901013010017010017.19010209010 Table 1Mobile phase gradient elution procedure.Time (min)AB090101901013010017010017.19010209010 Mobile phase gradient elution procedure. The mass spectrometer was a Q-Exactive (Thermo Fisher Scientific, CA, USA) HESI source with an ion source temperature of 310 °C, a capillary temperature of 320 °C, a sheath gas flow rate of 30 units, an auxiliary gas flow rate of 10 units, and a spray voltage of 2.8 kV for the negative ion mode. The analysis was performed using data-dependent scanning analysis (DDA) with the loop count set to 10, and the HCD energies were step-wise normalized collision energies set to 10, 28, and 35 eV. The primary mass spectrometry was scanned over a range of 100–1500 m/z, with the resolution set to 70,000, the AGC target set to 3E6, and the injection time set to 200 ms. The resolution of the secondary mass spectrum was set to 17,500, the AGC target was set to 1E5, and the injection time was set to 50 ms. Cells viability assay Brazilin was purchased from Aladdin Co.,Ltd. was dissolved in dimethyl sulfoxide (DMSO, the final concentration of DMSO < 0.1% v/v). Gingival fibroblasts (Prof. Peng Yi’s donation from Kunming medical university) were seeded in 96-well plates at a density of 1 × 104 cells/well with various concentrations of brazilin for 24 h. A 10% cell-counting kit-8 (CCK-8, Meilune, Dalian, China) solution was added in each well for 2 h, and then the OD value was detected at 450 nm using a Microplate Reader. Cell viability (100%) = (OD brazilin/DMSO – OD blank control) / (OD negative control – OD blank control)⋅100. Network pharmacology analysis Data acquisition We obtained the compounds for CS from TCMSP databases (https://www.tcmsp-e.com/tcmsp.php) and HERB databases (http://herb.ac.cn/), obtained the compounds’s Simplified Molecular Input Line Entry Specification (SMILES) from Pubchem database (https://pubchem.ncbi.nlm.nih.gov/), and obtained the related targets from Swiss Target Prediction database (http://www.swisstargetprediction.ch/)20,21. We limited our search to the “Homo sapiens” species and probability “0” to find the targets of compounds. We obtained the target genes for dental caries from the GeneCards (https://www.genecards.org/), TTD (https://db.idrblab.net/ttd/), and OMIM (https://omim.org/) database using “dental caries” and “caries” as search keywords. We used the online Venn platform (http://bioinformatics.psb.ugent.be/webtools/Venn/) to visualize the results. PPI network construction In order to explore the core targets from coincident targets of CS and dental caries, the coincident targets were input into the STRING 12.0 database (https://cn.string-db.org/). The biological was set to “Homo sapience”, the meaning of network edges was set to “confidence”, the minimum required interaction score was set to “medium confidence (0.400)”. Then, the Cytoscape v3.10.0 (https://cytoscape.org/) was used to construct the Protein-Protein Interaction (PPI) network among targets. In this network analysis, the degree value is used to reflect the interaction intensity between targets. GO and pathway enrichment analyses The targets were input into Metascape database (https://metascape.org/) to identify the term enrichment and analyze the share genes. The species was set as “Homo sapience” and the statistical significance of enrichment was set as “P value cut off = 0.01”. We limited the minimum number of pathways of 3, showing only the top 20. Then, the online Weishengxin platform (http://www.bioinformatics.com.cn/) was used to visualize the analytical results. This analysis included Gene Ontology (GO) enrichment for Biological Process (BP), Molecular Function (MF), Cellular Component (CC), and KEGG metabolic pathways. Construction of the target-drug-pathways network The Cytoscape version 3.10.0 was used to constructed the interaction network of target-drug-pathways. Prediction of pharmacokinetic properties and drug-likeness In order to determine whether the compounds can be developed as drugs and applied in clinical research, online drugability predictions were conducted for the compounds using Swiss ADME (http://www.swissadme.ch/) and Molinspiration tool (https://molinspiration.com/). Prediction of compound toxicity ProTox-II (http://tox.charite.de/protox_II) was used to estimates median lethal doses (LD50 values), categorises compounds into six classes ranging from Class I to Class VI (Table 2) and. analyses various toxicological pathways. Table 2Toxicological classification for LD50 values.ClassInferenceClass IFatal if swallowed (LD50 ≤ 5 mg/kg)Class IIFatal if swallowed (5 < LD50 ≤ 50 mg/kg)Class IIIToxic if swallowed (50 < LD50 ≤ 300 mg/kg)Class IVHarmful if swallowed (300 < LD50 ≤ 2000 mg/kg)Class VMay be harmful if swallowed (2000 < LD50 ≤ 5000 mg/kg)Class VINon-toxic (LD50 > 5000 mg/kg) Table 2Toxicological classification for LD50 values.ClassInferenceClass IFatal if swallowed (LD50 ≤ 5 mg/kg)Class IIFatal if swallowed (5 < LD50 ≤ 50 mg/kg)Class IIIToxic if swallowed (50 < LD50 ≤ 300 mg/kg)Class IVHarmful if swallowed (300 < LD50 ≤ 2000 mg/kg)Class VMay be harmful if swallowed (2000 < LD50 ≤ 5000 mg/kg)Class VINon-toxic (LD50 > 5000 mg/kg) Toxicological classification for LD50 values. In silico studies Molecular docking analysis We used molecular docking to test the reliability of potential targets for dental caries treatment. The PDB files of the targets were obtained from PDB database (https://www.rcsb.org/) and the SDF files of the compound 2D structure were obtained from Pubchem database (https://pubchem.ncbi.nlm.nih.gov/). We used the online platform CB-Dock2 (http://183.56.231.194:8001/cb-dock2/php/index.php) to perform the molecular docking. Molecular dynamics simulation Molecular dynamics simulations were conducted using the GROningen MAchine for Chemical Simulations (GROMACS) v2022.3 (https://www.gromacs.org/) software. After the simulation was completed, analysis was carried out using software tools and Python. Visualization was performed using the Visual Molecular Dynamics (VMD) v1.9.4 (https://www.ks.uiuc.edu/Research/vmd/) and Python Molecular viewer (PyMOL) v3 (https://www.pymol.org/pymol.html) software. Data curves were plotted using Grace and matplotlib, mainly including the root mean square deviation (RMSD), root mean square rise and fall (RMSF), hydrogen bonds (HBond), solvent-accessible surface area (SASA), protein radius of gyration (Rg), and the free binding energy (MM-PBSA) of each amino acid trajectory were calculated. Statistical analysis Statistical analysis was performed using GraphPad Prism v9 (https://www.graphpad.com/scientific-software/prism/). T-test was used for comparisons between two groups. One-way ANOVA was performed to assess variance among multiple groups, and Tukey’s multiple comparison test was used for multiple comparisons. All significant differences were identified at 0.05. Results Bioassay-guided isolation of active ingredients from CS against S. mutans Effects of ethanol extracts on biofilm formation and acid production in S. mutans According to the results, the MIC of the 0% ethanol extracts was 1.25 mg/mL, while extracts prepared with 30–90% ethanol exhibited a lower MIC of 0.625 mg/mL. The effect of ethanol extracts on S. mutans biofilm formation is shown in Fig. 1A. The crystal violet staining experiment revealed that compared to the control group, the biomass of S. mutans biofilms could be effectively decreased by 70–90% ethanol extracts at 1/2 MIC concentration (89–93% reduction). Furthermore, as shown in Fig. 1B, all ethanol extracts at 1/2 MIC concentration could inhibit acid production by S. mutans compared to the control. Effect of EtOAc, n-butanol, and water extracts on biofilm formation and acid production in S. mutans EtOAc, n-butanol, and water extracts were obtained by further fractionating the 70% ethanol extract with various solvents. A PE fraction was not obtained in significant yield. The MIC of the EtOAc extract was 0.625 mg/mL, the MIC of the n-butanol extract was 0.3125 mg/mL, and the MIC of the water extract was more than 5 mg/mL. CV staining revealed that only the EtOAc extract significantly inhibited S. mutans biofilm formation (Fig. 2A, B). Furthermore, assessment of acid production inhibition showed that only the EtOAc extract significantly inhibited S. mutans acid production (Fig. 2C, D), consistent with the biofilm formation results. Effect of different fractions on biofilm formation and acid production in S. mutans The EtOAc extract was further fractionated to yield 10 fractions (Fr. 1-Fr. 10), of which the MIC of Fr. 5 was 0.625 mg/mL, Fr. 8, Fr. 9 and Fr. 10 were 2.5 mg/mL, and the other fractions were more than 5 mg/mL. The inhibition of biofilm formation of S. mutans by different fractions was investigated by CV staining and it was found that Fr. 5, Fr. 8, and Fr. 10 were effective in inhibiting the biofilm formation at the 1/2 MIC and 1/4 MIC, and Fr. 9 could significantly inhibit the biofilm formation at only 1/2 MIC (Fig. 3A). However, in acid production, Fr. 5 and Fr. 10 could effectively inhibit the acid production by S. mutans, Fr. 8 and Fr. 9 could significantly inhibit the acid production at only 1/2 MIC (Fig. 3C). Interestingly, Fr. 2 and Fr. 6 inhibit the acid production at 2.5 mg/mL (Fig. 3D). These results suggest that different components may influence the cariogenicity of S. mutans through different mechanisms. Fig. 1Different ethanol extracts on biofilm formation and acid production in S. mutans. (A) Quantitative analysis by crystal violet staining. (B) S. mutans medium pH measurement. Different ethanol extracts on biofilm formation and acid production in S. mutans. (A) Quantitative analysis by crystal violet staining. (B) S. mutans medium pH measurement. Fig. 2EtoAc, n-butanol, and water extracts on biofilm formation and acid production in S. mutans. (A,B) Quantitative analysis by crystal violet staining. (C,D) S. mutans medium pH measurement. EtoAc, n-butanol, and water extracts on biofilm formation and acid production in S. mutans. (A,B) Quantitative analysis by crystal violet staining. (C,D) S. mutans medium pH measurement. Fig. 3Different fractions on biofilm formation and acid production in S. mutans. (A) Quantitative analysis by crystal violet staining. (B) S. mutans medium pH measurement. Different fractions on biofilm formation and acid production in S. mutans. (A) Quantitative analysis by crystal violet staining. (B) S. mutans medium pH measurement. The ingredients analysis of Fr.5 by UPLC-MS/MS Due to its superior inhibitory activity against S. mutans biofilm formation and acid production at lower concentrations compared to other fractions, Fr.5 was selected for compositional analysis. According to the retention time, the high-resolution accurate molecular weight and MS/MS fragment information of each chemical component were obtained by UPLC-MS detection. A total of 33 constituents were tentatively identified in Fr.5. The total ion chromatogram (TIC) in negative ion mode is presented in Fig. 4. Based on relative peak abundance, the major compounds were tentatively identified as brazilin (Table 3). The molecular structure of brazilin was shown in Fig. 5 (Table 4). Fig. 4The total ion current map of Fr. 5 in the negative ion mode. The total ion current map of Fr. 5 in the negative ion mode. Fig. 5The chemical structural formula of brazilin. The chemical structural formula of brazilin. Table 3List of ingredients detected from Fr. 5.NoRT (min)Molecular formulaMeasured mass [M − H] ⁻ (m/z)Putative IdentificationPeak area (%)16.594C16H14O5285.07718Brazilin23.3123.361C16H16O6303.0878Sappanol20.7334.02C16H14O5285.077123-Deoxysappanone B18.2645.604C16H12O5283.06155Brazilein9.0355.805C16H14O6301.07199Hematoxylin7.8065.709C15H12O5271.06149Protosappanin A1.9077.47C15H12O5271.061493’,4’,7-Trihydroxyflavanone1.3887.006C15H10O7301.03577Quercetin0.7793.388C7H6O3137.023482,5-Dihydroxybenzaldehyde0.74107.256C16H14O4269.0822110,11-Dihydroxydracaenone C0.73111.929C8H8O4167.034366-Methoxysalicylic acid0.49128.798C17H14O7329.06702Ombuin0.321312.04C18H32O4293.212659-HpODE0.30147.242C16H12O7315.05148Isorhamnetin0.29155.682C15H12O7303.05133Taxifolin0.28164.703C8H8O2135.044242-Methylbenzoic acid0.14178.513C16H12O6299.05634Tectorigenin0.10185.187C9H8O3163.039322-Hydroxycinnamic acid0.07196.903C7H6O3137.023383-Hydroxybenzoic acid0.06203.837C8H8O4167.03421Isovanillic acid0.06212.24C8H10O3153.054893,4-Dihydroxyphenylethanol0.062215.953C14H30O4293.17972Myristyl sulfate0.052318.608C30H48O3455.35391Ursolic acid0.05243.877C10H8O4191.034237,8-Dihydroxy-4-methylcoumarin0.05255.807C8H8O3151.039082-Hydroxyphenylacetic acid0.042611.255C16H22O4277.14481Dibutyl phthalate0.02275.307C9H10O5197.04507Ethyl gallate0.02286.145C11H12O4207.0659Sinapoyl aldehyde0.02293.728C9H6O4177.018627,8-Dihydroxycoumarin0.02301.103C16H12O6299.05649Hematein0.02313.567C9H10O3165.054933-Hydroxy-1-(4-hydroxyphenyl) propan-1-one0.01329.279C15H20O3247.13405Atractylenolide III0.01339.477C14H22O2221.154512,5-Di-tert-butylhydroquinone0.01 Table 3List of ingredients detected from Fr. 5.NoRT (min)Molecular formulaMeasured mass [M − H] ⁻ (m/z)Putative IdentificationPeak area (%)16.594C16H14O5285.07718Brazilin23.3123.361C16H16O6303.0878Sappanol20.7334.02C16H14O5285.077123-Deoxysappanone B18.2645.604C16H12O5283.06155Brazilein9.0355.805C16H14O6301.07199Hematoxylin7.8065.709C15H12O5271.06149Protosappanin A1.9077.47C15H12O5271.061493’,4’,7-Trihydroxyflavanone1.3887.006C15H10O7301.03577Quercetin0.7793.388C7H6O3137.023482,5-Dihydroxybenzaldehyde0.74107.256C16H14O4269.0822110,11-Dihydroxydracaenone C0.73111.929C8H8O4167.034366-Methoxysalicylic acid0.49128.798C17H14O7329.06702Ombuin0.321312.04C18H32O4293.212659-HpODE0.30147.242C16H12O7315.05148Isorhamnetin0.29155.682C15H12O7303.05133Taxifolin0.28164.703C8H8O2135.044242-Methylbenzoic acid0.14178.513C16H12O6299.05634Tectorigenin0.10185.187C9H8O3163.039322-Hydroxycinnamic acid0.07196.903C7H6O3137.023383-Hydroxybenzoic acid0.06203.837C8H8O4167.03421Isovanillic acid0.06212.24C8H10O3153.054893,4-Dihydroxyphenylethanol0.062215.953C14H30O4293.17972Myristyl sulfate0.052318.608C30H48O3455.35391Ursolic acid0.05243.877C10H8O4191.034237,8-Dihydroxy-4-methylcoumarin0.05255.807C8H8O3151.039082-Hydroxyphenylacetic acid0.042611.255C16H22O4277.14481Dibutyl phthalate0.02275.307C9H10O5197.04507Ethyl gallate0.02286.145C11H12O4207.0659Sinapoyl aldehyde0.02293.728C9H6O4177.018627,8-Dihydroxycoumarin0.02301.103C16H12O6299.05649Hematein0.02313.567C9H10O3165.054933-Hydroxy-1-(4-hydroxyphenyl) propan-1-one0.01329.279C15H20O3247.13405Atractylenolide III0.01339.477C14H22O2221.154512,5-Di-tert-butylhydroquinone0.01 List of ingredients detected from Fr. 5. Table 4Information on the main active ingredients of the CS.Code numberIngredient IDIngredient nameIngredient formulaC1HBIN0010201,3,4,5-Tetracaffeoylquinic acidC43H36O18C2HBIN0028751-OctacosanolC28H58OC3HBIN0078453,9-Dihydroxy-8-methoxydibenzo [b, d] pyran-6-oneC14H10O5C4HBIN0084353-DeoxysappanchalconeC16H14O4C5HBIN0084363’-DeoxysappanolC16H16O5C6HBIN0084373-Deoxysappanone bC16H14O5C7HBIN0087923’-Methoxy-4’,5,7-trihydroxyflavoneC16H12O6C8HBIN0129037,3′,4′-Trihydroxy-3-benzyl-2H-chromeneC16H14O4C9HBIN0132247-Hydroxy-3-(4’-hydroxybenzylidene)-chroman-4-oneC16H12O4C10HBIN018719BonducellinC17H14O4C11HBIN018818BrazileinC16H12O5C12HBIN018819BrazilinC16H14O5C13HBIN024649(e)-2-NonenalC9H16OC14HBIN025418EpisappanolC16H16O6C15HBIN027030Gallic acidC7H6O5C16HBIN028989HemateinC16H12O6C17HBIN034745MentholC10H20OC18HBIN035418(−)-Methyl selina-3,11-dien-14-oateC16H24O2C19HBIN036718NeoprotosappaninC32H26O10C20HBIN038026Oleic acidC18H34O2C21HBIN038089OmbuinC17H14O7C22HBIN039794PhytolC20H40OC23HBIN040939Protosappanin AC15H12O5C24HBIN040941Protosappanin A dimethyl acetalC17H18O6C25HBIN040942Protosappanin BC16H16O6C26HBIN040948ProtostemonineC23H31NO6C27HBIN041495QuercetinC15H10O7C28HBIN042158RhamnetinC16H12O7C29HBIN043129SappanchalconeC16H14O5C30HBIN043131SappanolC16H16O6C31HBIN043133Sappanone bC16H14O6C32HBIN043148SarcosineC3H7NO2C33HBIN044730Stearic acidC18H36O2C34HBIN045541TaraxerolC30H50OC35HBIN046012TetraacetylbrazilinC24H22O9 Table 4Information on the main active ingredients of the CS.Code numberIngredient IDIngredient nameIngredient formulaC1HBIN0010201,3,4,5-Tetracaffeoylquinic acidC43H36O18C2HBIN0028751-OctacosanolC28H58OC3HBIN0078453,9-Dihydroxy-8-methoxydibenzo [b, d] pyran-6-oneC14H10O5C4HBIN0084353-DeoxysappanchalconeC16H14O4C5HBIN0084363’-DeoxysappanolC16H16O5C6HBIN0084373-Deoxysappanone bC16H14O5C7HBIN0087923’-Methoxy-4’,5,7-trihydroxyflavoneC16H12O6C8HBIN0129037,3′,4′-Trihydroxy-3-benzyl-2H-chromeneC16H14O4C9HBIN0132247-Hydroxy-3-(4’-hydroxybenzylidene)-chroman-4-oneC16H12O4C10HBIN018719BonducellinC17H14O4C11HBIN018818BrazileinC16H12O5C12HBIN018819BrazilinC16H14O5C13HBIN024649(e)-2-NonenalC9H16OC14HBIN025418EpisappanolC16H16O6C15HBIN027030Gallic acidC7H6O5C16HBIN028989HemateinC16H12O6C17HBIN034745MentholC10H20OC18HBIN035418(−)-Methyl selina-3,11-dien-14-oateC16H24O2C19HBIN036718NeoprotosappaninC32H26O10C20HBIN038026Oleic acidC18H34O2C21HBIN038089OmbuinC17H14O7C22HBIN039794PhytolC20H40OC23HBIN040939Protosappanin AC15H12O5C24HBIN040941Protosappanin A dimethyl acetalC17H18O6C25HBIN040942Protosappanin BC16H16O6C26HBIN040948ProtostemonineC23H31NO6C27HBIN041495QuercetinC15H10O7C28HBIN042158RhamnetinC16H12O7C29HBIN043129SappanchalconeC16H14O5C30HBIN043131SappanolC16H16O6C31HBIN043133Sappanone bC16H14O6C32HBIN043148SarcosineC3H7NO2C33HBIN044730Stearic acidC18H36O2C34HBIN045541TaraxerolC30H50OC35HBIN046012TetraacetylbrazilinC24H22O9 Information on the main active ingredients of the CS. Effect of brazilin on biofilm formation and acid production in S. mutans The aim of this study was to assess the effect of brazilin on S. mutans biofilm formation and acid production. The MIC of brazilin was 37.5 µg/mL, sodium fluoride (NaF) was set as control drug at 18.75 µg/mL, it was consistent with brazilin at 1/2 MIC. The results demonstrated that brazilin at 1/2 MIC significantly inhibited biofilm formation better than NaF, reducing total biomass by approximately 45% compared to the control (Fig. 6A, B). Furthermore, brazilin effectively reduced acid production, as evidenced by a higher final pH compared to the control (Fig. 6C). Fig. 6Effect of brazilin on biofilm formation and acid production in S. mutans. (A) Crystal violet staining. (B) Quantitative analysis by crystal violet staining. (C) S. mutans medium pH measurement. Effect of brazilin on biofilm formation and acid production in S. mutans. (A) Crystal violet staining. (B) Quantitative analysis by crystal violet staining. (C) S. mutans medium pH measurement. Effect of brazilin on mature biofilm eradication in S. mutans As Fig. 7 showed that after the biofilm matured, adding different drugs did not effectively eliminate the biofilm except for the 1/2 MIC brazilin group. In brazilin at 1/2 MIC, biofilm decreased compared with control group, which might be due to the fact that brazilin removed some of the surface bacteria. Effects of brazilin on cells viability CCK-8 is a commonly used method for evaluating cytotoxicity. As Fig. 8 shows that when the concentration of brazilin is higher than 37.5 µg/mL, the cell survival rate is relatively low, and it approaches 60% at 18.75 µg/mL. Fig. 7Effect of brazilin on mature biofilm eradication in S. mutans. (A) Crystal violet staining. (B) Quantitative analysis by crystal violet staining. Effect of brazilin on mature biofilm eradication in S. mutans. (A) Crystal violet staining. (B) Quantitative analysis by crystal violet staining. Fig. 8Effect of brazilin on cells viability. Cells were cultured with brazilin at different concentrations for 24 h. Effect of brazilin on cells viability. Cells were cultured with brazilin at different concentrations for 24 h. Network pharmacology analysis Active ingredients of CS After excluding active ingredients without predicted targets, a total of 35 active ingredients were obtained (Table 3). Potential targets prediction Using a probability threshold > 0 as the selection criterion, 348 putative targets were obtained for the 35 active compounds, and 2568 dental caries targets were retrieved. Through Venn mapping, 79 common targets of CS and dental caries were obtained (Fig. 9A). PPI network construction and analysis The interaction among 79 overlapping targets were analyzed by Protein-protein interaction (PPI) network. The number of edges in PPI network is 444, and the degree value of average nodes is 11.53. In addition, the degree values of 34 targets were exceeded the average value (Fig. 9B), identifying AKT1, EGFR, BCL2, PTGS2, MMP9, ERBB2, and HSP90AA1 as topologically significant hub nodes. These high-degree targets represent potential key mediators of CS’s related to dental caries. GO function and KEGG pathway enrichment Functional enrichment analysis identified 881 significant GO terms and 113 KEGG pathways. The results showed that the most significant biological processes were associated with cellular response to nitrogen compound and cellular response to hormone stimulus, the most significant cellular component was associated with extracellular matrix, and the most significant molecular function was related to protein kinase activity (Fig. 10A). Moreover, KEGG pathway enrichment analysis demonstrated that the overlapped targets were mainly enriched in the nitrogen metabolism, pathways in cancer, prostate cancer, microRNAs in cancer, calcium signaling pathway (Fig. 10B). Target-drug-pathways (T-D-P) network in dental caries construction and analysis The final interaction network was constructed using the previously obtained information of 79 overlapping targets, 17 KEGG pathways involved in the anti-caries mechanism of CS. The results of the T-D-P network revealed five hub pathways with degree centrality exceeding the network average (12.53): prostate cancer, pathways in cancer, calcium signaling pathway, microRNAs in cancer, and nitrogen metabolism. Concurrently, the 11 active compounds were prioritized based on the degree value of average nodes (7.86), including C7, C4, C3, C1, C6, C15, C8, C27, C35, C19, and C12 (Fig. 11). Prediction of pharmacokinetic properties and drug-likeness SwissADME and Molinspiration were utilized to predict pharmacokinetic properties and drug-likeness for active compounds. For orally available small molecules, a TPSA < 140 offers an optimal human utilization rate20. As S1 showed that 9 compounds pass the Lipinski, with high GI absorption, and with TPSA < 140. Toxicity analysis The toxicity prediction of active compounds was performed by Protox- II web server. The result of toxicity analysis revealed only 4 compounds belong to Class V and Class VI. Brazilin was found to have respiratory toxicity and cardiotoxicity. The detailed toxicity analysis is reported in S2. Validation of key targets using molecular docking Molecular docking analysis was performed to brazilin with seven key potential targets (AKT1, EGFR, BCL2, PTGS2, MMP9, ERBB2, and HSP90AA1) and virulence factors of S. mutans, including glucosyltransferase B (GtfB), glucosyltransferase C (GtfC), glucosyltransferase D (GtfD), surface protein antigen P (SpaP) and glucan-binding protein C (GbpC). Generally, lower docking scores indicate stronger predicted binding affinity, with docking scores < -7 kcal/mol typically considered indicative of strong binding. The interaction of brazilin and keys targets was shown in Fig. 12, the results showed docking scores were − 5.8, -8.1, -7.8, -8.7, -8.1, -6.7, -7.6 kcal/mol respectively, indicating that brazilin demonstrated stronger predicted binding affinity towards EGFR, BCL2, PTGS2, MMP9, and HSP90AA1 (Table 5). The interaction of brazilin and virulence factors of S. mutans was shown in Figure S1, indicating that brazilin demonstrated stronger predicted binding affinity towards GtfB, GtfC, GtfD, and GbpC (Table 5). The interaction of apigenin and Gtfs of S. mutans was shown in Figure S2, indicating that apigenin demonstrated stronger predicted binding affinity towards GtfB, GtfC, and GtfD (Table 6). Fig. 9Venn diagram and PPI network. (A) Venn diagram of the intersection of the CS targets and dental caries targets. (B) PPI network of the 79 potential therapeutic targets. Each node represents a relevant gene, and the edge thickness indicates the strength of the data support. Venn diagram and PPI network. (A) Venn diagram of the intersection of the CS targets and dental caries targets. (B) PPI network of the 79 potential therapeutic targets. Each node represents a relevant gene, and the edge thickness indicates the strength of the data support. Fig. 10GO enrichment and KEGG pathway enrichment analysis. (A) GO enrichment analysis: the top 10 Biological Process (BP), Cellular Component (CC) and Molecular Function (MF); (B) bubble chart of the KEGG pathways. GO enrichment and KEGG pathway enrichment analysis. (A) GO enrichment analysis: the top 10 Biological Process (BP), Cellular Component (CC) and Molecular Function (MF); (B) bubble chart of the KEGG pathways. Fig. 11T-D-P network. The diamonds nodes indicate metabolic pathways, the triangle nodes indicate active ingredients of CS, the blue oval nodes indicate targets. The lines indicate interactions. T-D-P network. The diamonds nodes indicate metabolic pathways, the triangle nodes indicate active ingredients of CS, the blue oval nodes indicate targets. The lines indicate interactions. Fig. 12Schematic diagram of the interaction of brazilin and keys targets. (A) Brazilin-AKT1. (B) Brazilin-EGFR. (C) Brazilin-BCL2. (D) Brazilin-PTGS2. (E) Brazilin-MMP9. (F) Brazilin-ERBB2. (G) Brazilin-HSP90AA1. Performed by CB-Dock2 (http://183.56.231.194:8001/cb-dock2/php/index.php). Schematic diagram of the interaction of brazilin and keys targets. (A) Brazilin-AKT1. (B) Brazilin-EGFR. (C) Brazilin-BCL2. (D) Brazilin-PTGS2. (E) Brazilin-MMP9. (F) Brazilin-ERBB2. (G) Brazilin-HSP90AA1. Performed by CB-Dock2 (http://183.56.231.194:8001/cb-dock2/php/index.php). Table 5The binding energy of brazilin with targets.No.TargetsBinding energy1AKT1-5.83EGFR-8.14BCL2-7.85PTGS2-8.76MMP9-8.17ERBB2-6.78HSP90AA1-7.69GtfB-7.210GtfC-8.511GtfD-8.312SpaP-5.113GbpC-7.8 Table 5The binding energy of brazilin with targets.No.TargetsBinding energy1AKT1-5.83EGFR-8.14BCL2-7.85PTGS2-8.76MMP9-8.17ERBB2-6.78HSP90AA1-7.69GtfB-7.210GtfC-8.511GtfD-8.312SpaP-5.113GbpC-7.8 The binding energy of brazilin with targets. Table 6The binding energy of apigenin with Gtfs.No.TargetsBinding energy1GtfB-8.52GtfC-8.13GtfD-8.5 Table 6The binding energy of apigenin with Gtfs.No.TargetsBinding energy1GtfB-8.52GtfC-8.13GtfD-8.5 The binding energy of apigenin with Gtfs. Molecular dynamics simulation To verify the docking results, we selected the two pairs with the best binding ability and calculated their RMSD, RMSF, HBond, SASA and Rg respectively of brazilin-PTGS2 and brazilin-GtfC. Meanwhile, the method of MM-PBSA was used to calculate the binding free energy. RMSD and RMSF were analyzed to determine the stability of the system and the flexibility of amino acids. Figure 13A show that the value fluctuated slightly within the range of 40–100 ns and 0.23–0.25 nm, indicating that the system has reached stable state while brazilin-GtfC not yet reached a strictly stable state (Fig. 14A). Subsequently, we analyzed the RMSF values to see the movement of each amino acid. Figures 13B and 14B show that the RMSF values of the two protein-complexes are relatively low. HBond analysis (Figs. 13C and 14C) revealed that the number of hydrogen bonds formed between brazilin and PTGS2 and GtfC fluctuates dynamically within the range of 4 to 10, which is crucial for maintaining a highly specific binding. Rg and SASA jointly verified the integrity of the protein structure during the binding process. Rg and SASA remained relatively stable, indicating that the overall shape and compactness of the protein were well maintained in the simulation (Figs. 13D and E and 14D and E). Additionally, we calculated binding free energy using the MM-PBSA. As shown in Table 7, the total binding free energies (ΔTOTAL) of brazilin-PTGS2 and brazilin-GtfC complexes were − 38.7 and − 25.45 kcal/mol respectively, indicating that brazilin has good affinity for PTGS2 and GtfC. Fig. 13Brazilin-PTGS2 in conventional molecular dynamics simulations. (A) RMSD; (B) RMSF; (C) HBond; (D) SASA; (E) Rg. Brazilin-PTGS2 in conventional molecular dynamics simulations. (A) RMSD; (B) RMSF; (C) HBond; (D) SASA; (E) Rg. Fig. 14Brazilin-GtfC in conventional molecular dynamics simulations. (A) RMSD; (B) RMSF; (C) HBond; (D) SASA; (E) Rg. Brazilin-GtfC in conventional molecular dynamics simulations. (A) RMSD; (B) RMSF; (C) HBond; (D) SASA; (E) Rg. Table 7The binding free energy of brazilin with two core targets predicted by MM-PBSA (kcal/mol).Energy componentBrazilin-PTGS2Brazilin-GtfCΔVDWAALS-44.24-29.81ΔEEL0.000.00ΔEGB10.187.75ΔESURF-4.01-3.39ΔGGAS-44.24-29.81ΔGSOLV6.174.36ΔTOTAL-38.7-25.45ΔVDWAALS van der Waals energy; ΔEEL electrostatic energy; ΔEGB Polar solvation energy; ΔESURF Nonpolar solvation energy; ΔGGAS gas molecular energy = ΔVDWAALS + ΔEEL; ΔGSOLV Energy of solvation = ΔEGB + ΔESURF; ΔTOTAL total binding free energy = ΔGGAS + ΔGSOLV. Table 7The binding free energy of brazilin with two core targets predicted by MM-PBSA (kcal/mol).Energy componentBrazilin-PTGS2Brazilin-GtfCΔVDWAALS-44.24-29.81ΔEEL0.000.00ΔEGB10.187.75ΔESURF-4.01-3.39ΔGGAS-44.24-29.81ΔGSOLV6.174.36ΔTOTAL-38.7-25.45ΔVDWAALS van der Waals energy; ΔEEL electrostatic energy; ΔEGB Polar solvation energy; ΔESURF Nonpolar solvation energy; ΔGGAS gas molecular energy = ΔVDWAALS + ΔEEL; ΔGSOLV Energy of solvation = ΔEGB + ΔESURF; ΔTOTAL total binding free energy = ΔGGAS + ΔGSOLV. The binding free energy of brazilin with two core targets predicted by MM-PBSA (kcal/mol). ΔVDWAALS van der Waals energy; ΔEEL electrostatic energy; ΔEGB Polar solvation energy; ΔESURF Nonpolar solvation energy; ΔGGAS gas molecular energy = ΔVDWAALS + ΔEEL; ΔGSOLV Energy of solvation = ΔEGB + ΔESURF; ΔTOTAL total binding free energy = ΔGGAS + ΔGSOLV. Discussion Dental caries remains a global public health challenge, adversely affecting both oral and systemic health and imposing substantial economic burdens. Natural plant-based medicines show considerable potential for the management of dental caries. Conventional separation methods often encounter difficulties in efficiently isolating compounds with specific biological activities. Bioassay-guided isolation overcomes this limitation by integrating activity evaluation throughout the separation process, thereby increasing the likelihood of obtaining bioactive compounds and improving their practical applicability21. The cariogenic potential of S. mutans is mainly attributed to its ability to form dental plaque biofilms and produce acidic metabolites. Accordingly, during the isolation process, the inhibitory effects of different extracts on S. mutans biofilm formation and acid production were assessed. Initial extraction was performed using ethanol as the solvent, and the results revealed that 70–90% ethanol extracts exhibited significant inhibition of S. mutans. Since lower solvent concentrations are known to better preserve compound activity22, the 70% ethanol extract was selected for further fractionation. Subsequent UPLC-MS/MS analysis of Fr.5 identified brazilin as the predominant compound. Brazilin, a main compound derived from the heartwood of CS23, was confirmed as the key compound responsible for the anti-caries activity of CS. The MIC of brazilin is 37.5 µg/mL, which is lower than the concentrations of chlorhexidine (0.12% − 2%) and fluoride (0.1%-2.26%) commonly used in clinical, moreover, its demonstrated efficacy in inhibiting S. mutans biofilm formation and acid production compared with the same concentration of fluoride further supports its potential as an anti-caries agent. In addition, the potential anti-caries compounds, associated targets, and molecular mechanisms were investigated using a network pharmacology approach. This analysis predicted 11 active compounds including brazilin and 7 relevant targets. Bioavailability and toxicity are important factors to consider in drug development. The Total Polar Surface Area (TPSA), a key parameter related to passive molecular diffusion, can be used to predict bioavailability24. Brazilin exhibits a reasonable TPSA value, indicated its potential for further development as a drug candidate. In toxicity prediction, brazilin may pose risks of cardiotoxicity and respiratory toxicity; however, the MIC required for antibacterial activity is considerably lower than the LD50, suggesting a reduced likelihood of severe toxicity at effective doses. In vitro cytotoxicity assays revealed significant cytotoxicity at the MIC, which may be attributed to the high concentration or prolonged exposure time. It should be noted that anti-caries agents are typically applied topically and are not maintained at high concentrations continuously in clinical use. Gene ontology enrichment analysis indicated that the molecular function of protein kinase activity is related to CS against dental caries. The activities of numerous protein kinases are critically involved in caries pathogenesis, primarily by modulating S. mutans virulence factors. Specifically, serine/threonine protein kinases (STPKs) and histidine kinase (HK) are involved in two-component system (TCS) signaling of S. mutans and can regulate the expression of metabolism-related genes in the cell wall by sensing environmental signals and activating response regulators25,26. KEGG pathway analysis further identified nitrogen metabolism and calcium signaling as enriched pathways related to dental caries. Nitrogen metabolism is essential for bacterial growth and survival, as bacteria utilize nitrogen-containing compounds to synthesize vital biomolecules27,28. In S. mutans, the GlnR regulon orchestrates nitrogen assimilation; its dual control by PmrA (activator at neutral pH) and GlnR (repressor at acidic pH) acts as an ATP-saving switch that reallocates energy from amino-acid biosynthesis to proton extrusion, thereby enhancing acid tolerance and cariogenic fitness29. Furthermore, the first gene of the ciaRH operon in S. mutans is a calcium sensing signal peptide that allows the CiaRH system to regulate the cariogenicity of S. mutans by modulating the expression of its own operon in response to calcium ions in the oral environment30. CS contains a variety of active compounds with obvious antibacterial effects31. However, dental caries involves multifactorial mechanisms, and the pathways identified here are primarily linked to S. mutans. Whether CS affects dental caries through these specific pathways requires further experimental validation. Brazilin is a homoisoflavonoid compound that exhibits various pharmacological activities such as anti-inflammatory effects and promotion of bone formation32. In this study, we found that brazilin has a significant inhibitory effect on S. mutans and based on the network pharmacology analysis of the predicted targets for dental caries, the binding ability was predicted through molecular docking. Lower docking scores indicate stronger predicted binding affinity. Notably, the docking scores for EGFR, MMP9, PTGS2 are lowest, suggesting strong binding potential. EGFR plays a critical role in proliferation, differentiation, development, and promoting downstream signaling, the expression of EGFR is up-regulated in dental caries, which may be involved in the extracellular matrix (ECM) degradation process33,34. Interestingly, brazilin has been shown to reduce ECM synthesis, suggesting a potential counteracting mechanism35. MMP9 is a protein of the matrix metalloproteinase family that plays an essential role in the proteolysis of the ECM, and may be involved in the destruction of the dentin organic matrix during caries36. A study showed that brazilin can suppress MMP937. PTGS2, also known as cyclooxygenase-2 (COX-2), is an inducible enzyme associated with prostaglandin synthesis and also plays an important role in inflammatory responses, cell proliferation, and apoptosis38,39. However, PTGS2 is mainly related to dental pulp inflammation in dental caries research, the AgI/II protein produced by S. mutans is capable of binding to receptors on the host cell surface, which is responsible for producing a variety of inflammatory mediators and causing the occurrence of dental pulp inflammation40,41. S. mutans possesses multiple virulence factors, however, the virulence proteins that have been extensively studied and can be retrieved from the PDB database mainly consist of GtfB, GtfC, GtfD, SpaP and GbpC. Furthermore, the virulence factors of S. mutans with brazilin also performed by molecular docking. The GtfC, GtfD, GbpC has the lowest docking scores. Gtf are one of the most important virulence factors of S. mutans. GtfC initiates plaque by forming an insoluble glucan scaffold on the tooth surface, whereas GtfD produces soluble glucans that prime GtfB-mediated insoluble glucan synthesis and serve as a fermentable carbon reserve, together driving the formation of cohesive and cariogenic biofilms42. Apigenin (4′,5,7-trihydroxyflavone), was identified as the most effective non-competitive inhibitor of Gtfs, wes selected as positive drug to conduct molecular docking verification43. Apigenin has a relatively low binding energy, indicating that it has a good binding ability with Gtfs. This also suggests that brazilin may be effective in targeting Gtfs. GbpC is a crucial multifunctional surface adhesion factor in S. mutans, featuring both structural specificity and functional diversity. It can simultaneously mediate sucrose-independent adhesion and sucrose-dependent adhesion. Moreover, it participates in the bacterial aggregation induced by glucan, promoting the early formation of biofilms44,45. Based on the stronger evidence linking their functions to caries pathogenesis and the direct experimental evidence of brazilin’s modulation, GtfC, GtfD, GbpC emerge as the most plausible primary targets for brazilin’s anti-caries effects. Meanwhile, although EGFR, and MMP9 were predicted and analyzed as potential targets, they are not directly related to dental caries or S. mutans. Further, brazilin-PTGS2 and brazilin-GtfC complexes with better binding ability were combined for molecular dynamics simulation, further confirming that the complexes of these two pairs were relatively stable. Among them, the MM-PBSA of PTGS2 was lower, and its binding stability was superior to that of GtfC. Finally, this study offers valuable insights into the potential of CS in the prevention and treatment of dental caries and proposes a possible mechanistic basis. However, it should be noted that the targets predicted herein are associated with dental caries indirectly rather than through direct interactions. Furthermore, while certain pathways have been reported to be linked to S. mutans, their precise functional roles remain unclear. Thus, our findings necessitate further experimental confirmation. Another limitation of this study is the focus solely on S. mutans for activity screening, albeit justified by its established role in the initiation and progression of dental caries. Despite this, brazilin was identified as a key active constituent of CS and exhibited considerable anti-caries potential against this principal cariogenic pathogen. Future research should aim to elucidate the effects of brazilin on both bacterial and host targets related to dental caries, as well as explore structural modification of brazilin or its combination with other agents to reduce toxicity and enhance anti-caries efficacy. Conclusion In conclusion, brazilin was identified as the primary active compound in CS with inhibitory activity against S. mutans. Network pharmacology analysis revealed that 11 core compounds, including brazilin, are associated with dental caries and interact with 7 targets (AKT1, EGFR, BCL2, PTGS2, MMP9, ERBB2, and HSP90AA1), implicating 2 potential pathways: nitrogen metabolism and the calcium signaling pathway. Molecular docking further identified EGFR and MMP9 as core targets related to dental caries, along with GtfC, GtfD, and GbpC as potential anti-caries targets, and molecular dynamics simulation analysis revealed that brazilin has a good affinity for PTGS2 and GtfC. These findings suggest that brazilin is a promising candidate for the prevention of dental caries, likely exerting its effects via multi-target and multi-pathway mechanisms. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2 Supplementary Material 3 Supplementary Material 3