Silica nanoparticles (NPs) were synthesized using a modified Stöber method (Bai et al., 2019). Briefly, 6.28 mL of ammonium hydroxide solution (25%–28%; Aladdin, China), 42.8 mL of absolute ethanol (Xilong Scientific, China), and 20 mL of deionized water were mixed for 15 min under constant stirring, while 12 mL of tetraethyl orthosilicate (TEOS; Sigma, United States) was dissolved in 100 mL of ethanol. The latter solution was poured into the former solution under continuous stirring for 2 h (Mutlu et al., 2021). The SiO₂ NPs were collected by centrifugation, deionized water was used to wash the sediment one time, and then the NPs were washed three times with ethanol. Subsequently, the SiO₂ NPs was dried in an oven at 60°C overnight.

The SiO₂ NPs were transformed into HMSNs using a self-templating strategy (Fang et al., 2013; Yu et al., 2021a). Five hundred milligrams of the SiO₂ NPs were dispersed ultrasonically in 250 mL of deionized water. Then, 750 mg of cetyltrimethylammonium bromide (CTAB; Sigma, United States of America), 150 mL of ethanol, and 2.75 mL of ammonium hydroxide (25%–28%) were added, and the mixture was stirred continuously for 30 min at ambient temperature, after which TEOS (6 mL) was added and stirring continued for 6 h. After centrifugation, the sediment was redispersed in 100 mL of deionized water. Subsequently, 2.12 g of sodium bicarbonate (Na₂CO₃; Aladdin, China) was added, and the mixture was continuously stirred at 65°C for 24 h to, etch the hollow core. After centrifugation, the precipitate was washed twice with deionized water and ethanol, and the collected particles were calcined at 550°C for 6 h.

One hundred milligrams of HMSNs was dispersed ultrasonically in 10 mL of ethanol. Then, 20 mg of quercetin (Sigma, United States) was added to the HMSNs solution with constant stirring at 400 rpm for 24 h. The mixture was then shaken at 150 rpm for an additional 24 h in the dark to achieve maximum quercetin loading. The suspension was washed with deionized water twice (Sapino et al., 2015; Saputra et al., 2022).

The morphologies of the HMSNs and Q@HMSNs were observed using transmission electron microscopy (TEM; JEM-F200, JEOL, Japan). Scanning TEM equipped with energy dispersive spectroscopy (EDS) was employed to examine the mapping patterns of elements. The crystalline phase was determined by applying X-ray diffraction (XRD; Smartlab 9KW, Rigaku, Japan). Small-angle and wide-angle XRD patterns were scanned from 0.5° to 10° and 10°–70°, respectively, and recorded in the 2θ range with a scanning speed of 1°/min. The functional groups in the sample were detected by Fourier transform infrared (FTIR) spectroscopy (iN10, Thermo Scientific, United States of America). The nitrogen adsorption–desorption isotherm was studied by an ASAP 2460M gas adsorption analyser (Micromeritics, United States of America) at 77 K. The pore size distribution, pore volume and surface area were calculated by the Brunauer–Emmett–Teller and Barrett–Joyner–Halenda (BET/BJH) methods. A Zetasizer Nano ZSP system (ZS90, Malvern Instruments, United Kingdom) was used to determine the size distribution. The effective amount of quercetin loaded into the HMSNs was determined using thermogravimetric analysis (TGA; TG 209 F3 Tarsus, NETZSCH, Germany) from 30°C to 1,000°C with a heating speed of 15°C/min.

One hundred milligrams of Q@HMSNs was dispersed in 10 mL of phosphate-buffered saline (PBS; Xilong Scientific, China), and then the mixture was shaken at 120 rpm in the dark. The release of quercetin was measured at 0.5, 1, 2, 4, 8, 12, 24, 48, 72, 120 and 168 h. At each time point, the Q@HMSNs solution was centrifuged at 4,200 rpm for 4 min, and 1 mL of the liquid supernatant was collected for UV–Vis measurement (SpectraMax iD3, Molecular Devices, United States of America). The solution was then replenished with 1 mL of fresh PBS. The content of released quercetin was determined by measuring the absorbance at 377 nm (Yu et al., 2021b; Saputra et al., 2022).

Human dental pulp stem cells (HDPSCs) and human gingival fibroblasts (HGFs) were taken from the pulp and peridental membranes of healthy adults who had given informed consent for third molar extraction, as approved by the Biomedical Research Ethics Committee of Local University (FMUSS 2018–20). α-Modified essential medium (α-MEM; HyClone, United States of America) with 1% penicillin/streptomycin (PS; Beyotime, China) and 10% foetal bovine serum (FBS; Gibco, Australia) were applied to culture the HDPSCs and HGFs in a humidified atmosphere of 5% CO₂ at 37 °C. Subculturing was performed when the cells reached confluence. The HDPSCs and HGFs were collected at third passage and seeded in 96-well plates (2000 cells/well). After 24 h, the cells were cultured with complete medium containing Q@HMSNs extract solutions at concentrations of 1.0, 2.5, and 5.0 mg/mL. A control group without Q@HMSNs was included in the experiment. After an additional 1, 4, and 7 days of incubation, cell counting kit-8 (CCK-8; Dojindo Japan) (10 μL) was added to each well. Incubation was continued at 37°C for 2 h. Then, the absorbance at 450 nm was measured (SpectraMax iD3, Molecular Devices, United States). Each test was performed in triplicate, the relative viability rates of the HDPSCs and HGFs were calculated using the following equation (Xing et al., 2022). Relative cell viability  % = O D S a m p l e − O D M e d i u m O D C o n t r o l − O D M e d i u m × 100   %

In addition, calcein/propidium iodide (PI) live/dead fluorescent dye (Beyotime, China) was utilized to stain the cells for 30 min on day 7. Images of cell morphology and proliferation were recorded by an inverted fluorescence microscope (Axio Observer A1, Carl Zeiss, Germany) (Huang et al., 2020).

The study protocol was approved by the Institutional Review Board of the local university (2020QH2045), and all teeth were obtained after donors provided signed informed consent.

Seventy intact fresh human third molars were collected. All molars were immersed in a 0.5% thymol solution and stored at 4°C within 1 month of extraction prior to use. A low-speed saw (Buehler; Lake Bluff, IL, United States of America) was used to slice the molar to 145 dentine specimens (each 2 mm × 2 mm × 2 mm, the mid-crown portions beneath the enamel–dentinal junction) under water cooling. The dentine blocks were distributed for various analyses: 50 were taken for surface profile measurements (n = 10); 20 for surface morphology observations (n = 4); 25 for rhodamine B fluorescein evaluations (n = 5); and 50 for type I collagen telopeptide (ICTP) assessment (n = 10).

The dentine blocks were coupled to a customed silicone mould, embedded with acrylic resin (New century, China) and wet polished using silicon carbide sandpapers (#320, #600, #800, #1000, #1200, and #2000; Matador, Germany) for 60 s each. Subsequently, ultrasound was used to clean the surface of each specimen. The final dimensions of the specimens were as follows: 4 mm × 4 mm for the top surface, 5 mm × 5 mm for the bottom surface, and a thickness of 3 mm. Each specimen was covered on both sides with nail varnish (Top Speed; Revlon, United States), leaving a 1 mm × 1 mm area exposed as the erosion region (Jiang et al., 2020).

All samples were subjected to erosive and abrasive challenges (4 cycles/d) for 7 days to simulate an eroded tooth (Wiegand and Attin, 2011). In each erosion cycle, specimens were first immersed in artificial saliva (containing 0.795 g/L CaCl₂·H₂O, 0.4 g/L NaCl, 0.005 g/L Na₂S·9H₂O, 0.4 g/L KCl, 0.3 g/L KSCN, 1 g/L urea, and 0.69 g/L NaH₂PO₄·H₂O, pH = 6.8) for 1 h (O Toole et al., 2015). Afterwards, the specimens were soaked in citric acid (pH = 2.45; Sigma, United States of America) for 5 min (Shellis et al., 2011), followed by a 10 s water rinse. Subsequently, the specimens were stored in renewed artificial saliva for 2 h before the next erosive challenge. Additionally, the specimens were brushed with a toothbrush for 2 min after the first and final daily erosive attacks (a total of 2 cycles per day) over a 7 days period, as previously described (Hong et al., 2020).

After the initial 7 days of erosive and abrasive challenges, the specimens were randomly divided into 5 groups (n = 26): DW (deionized water, negative control), NaF (12.3 mg/mL sodium fluoride, positive control), Q (300 μg/mL quercetin), HMSN (5.0 mg/mL HMSNs), and Q@HMSN (5.0 mg/mL Q@HMSNs).

The samples in the DW, NaF, and Q groups were immersed in their respective solutions for 2 min each day, while the samples in the HMSN and Q@HMSN groups were treated as follows. First, 0.2 mL of the appropriate suspension was gently applied to the surface of the specimen, and then the surface was blown dry using a dental air‒water syringe for 30 s. Following these treatments, the specimens were placed in renewed artificial saliva for 2 h before undergoing daily cycles of erosion (4 cycles/d) and abrasion (2 cycles/d). The abrasion procedure, as described previously, was performed after the first and last erosive attacks over 7 days (2 cycles/d). However, unlike the DW, NaF and Q groups, the HMSN and Q@HMSN groups did not received additional treatment with HMSNs or Q@HMSNs during the subsequent 6 days of erosive and abrasive challenges (Figure 1).

Scanning electron microscopy (SEM; EM8000, KYKY, China) was employed to observe the morphology of the dentine surface after 7 and 14 days of erosive and abrasive challenges. Twenty specimens were placed in the freeze dryer for desiccation for 24 h and coated with gold via a sputter-coating process. Observations were made with both surface and cross-sectional views, with central site of each segment being examined at a ×5,000 magnification (Yu et al., 2017).

The surface profiles of the specimens were measured after the 14 days experimental period using a contact profilometer (SEF680, Kosaka Laboratory, Japan) (Capalbo et al., 2022). Before evaluation, a scalpel blade was used to carefully remove the nail varnish. The EDL was calculated by determining the difference in height between the eroded surface and the reference surface. The surface of each sample was scanned with a stylus, and the mean value of 5 height difference measurements was recorded (in μm) as the EDL (Hannas et al., 2016).

Ten dentine specimens from each group (n = 2) were selected. Rhodamine B (0.1% (w/v)) was applied to the dentine surfaces of the specimens using a microbrush for 1 min. Afterwards, the specimens were washed with PBS three times. Subsequently, the specimens were longitudinally sectioned into slices, and confocal laser scanning microscopy (CLSM; FV3000, Olympus, Japan) was applied to observe the penetration depth of the specimens using laser excitation at 523 nm (Park et al., 2019).

To evaluate the effect of the Q@HMSNs on the DOM, the release of ICTP was measured. The remaining 50 dentine specimens were demineralized with 10% phosphoric acid (pH = 1.0) (Aladdin, China) for 12 h at 25°C. Subsequently, the samples were rinsed in deionized water with continuous stirring at 4°C for 72 h, followed by drying in 60°C oven for 8 h. The specimens were then immersed in tubes contained 500 μL of artificial saliva at 37°C for 1 week. After centrifugation, the supernatants from each tube were collected, and an ELISA kit (MBbio, China) was used to analyse the release of ICTP (Osorio et al., 2011).

The data were analysed using the SPSS statistical software package (IBM, SPSS 19, United States of America). The equal variances and normality of the data were confirmed using the Levene test and the Kolmogorov–Smirnov test, respectively. For the CCK-8, EDL, and ICTP release assays, one-way analysis of variance (ANOVA) with post hoc Tukey’s test was used for data analysis. A value of p < 0.05 was considered to indicate statistical significance.

The TEM image of the HMSNs in Figures 2A, B revealed spherical NPs with an approximate diameter of 400 nm and a well-ordered channel framework shell structure with a cavity. After loading quercetin, the mesoporous structure remained intact, albeit slightly obscured, indicating that quercetin had been successfully incorporated into the HMSNs with maintenance of the overall morphology (Figure 2C). The mapping pattern of scanning TEM/EDS was used to identify the element distribution. The presence of Si and O signals was demonstrated in both the HMSNs and Q@HMSNs, with the intensity of C in Q@HMSNs being further highlighted (Supplementary Figure S1).

The mesoporous crystalline phase and structure were analysed using XRD. For the small-angle pattern (Figure 3A), the XRD analysis of the HMSNs displayed a characteristic diffraction peak at 2θ = 1.9° (100), consistent with a mesoporous structure. Notably, the intensity of this diffraction peak decreased after quercetin loading. In the wide-angle pattern (Supplementary Figure S2), both the HMSNs and Q@HMSNs presented a broad band of non-crystalline scattering around 2θ = 22°. The Q@HMSNs did not exhibit any crystalline peaks in the observed 2-theta range. This could be explained by the quercetin changing into a non-crystalline state (Yu et al., 2023).

Additionally, in the FTIR spectrum of the HMSNs, distinctive peaks at 968 cm⁻¹ (Si–OH bending vibration), 1,096 cm⁻¹ (Si–O–Si vibration), 800 cm⁻¹, 470 cm⁻¹ (Si–O stretching and bending vibration), and 1,632 cm⁻¹ (H–O–H bending vibration) were observed. When compared to the HMSNs spectrum, the spectrum of the Q@HMSNs exhibited changes due to quercetin loading at 1,635 cm⁻¹ (C=O stretching vibration), 1,559 cm⁻¹, 1,384 cm⁻¹ (C=C stretching vibrations) and 3,442 cm⁻¹ (OH stretching) (Elmowafy et al., 2022). The presence of these functional groups confirmed the successful loading of quercetin (Figure 3B).

Nitrogen absorption-desorption isotherms were used to characterize the mesoporous characteristics of both the undoped HMSNs and Q@HMSNs. Both the HMSNs and Q@HMSNs presented type IV isotherms, and their mesoporous structure curves were in accordance with the TEM images and XRD analysis. The specific surface area (SBET), pore volume (V_P), and pore diameter (D_P) of the HMSNs, calculated using the BET/BJH method, were found to be 597.5 m²/g, 0.567 cm³/g, and 3.93 nm, respectively. After quercetin loading, the SBET and V_P decreased to 391.8.9 m²/g and 0.242 cm³/g, respectively. This reduction indicated the successful loading of quercetin into the mesopores of the silica NPs (Figure 3C; Supplementary Figures S3, S4; Supplementary Table S1).

TGA and derivative thermogravimetric analysis (DTG) were employed to assess the quantity and thermal stability of the quercetin loaded into the HMSNs. TGA revealed a weight loss of 5.62 wt% for the HMSNs and 16.13 wt% for the Q@HMSNs (Figure 3D). The prominent downwards peak at 66.8°C in the DTG curve corresponds to the evaporation of physically adsorbed water, while another significant downwards peak at 373.1°C indicates the decomposition of the organic substance quercetin. These results confirmed the encapsulation of quercetin into the HMSNs, with a calculated loading amount of 10.51 wt% (Figure 3E; Supplementary Figure S5).

Figure 3F; Supplementary Figure S6 illustrates the drug release profile from the Q@HMSNs, which indicated an initial burst release during the first 8 h, followed by a gradual slowing of the release rate, finally resulting in sustained release over 7 days.

Figures 4A, B depict the relative viabilities of HGFs and HDPSCs after exposure to different concentrations of Q@HMSNs (1.0, 2.5, and 5.0 mg/mL) for 1, 4, and 7 days. These figures illustrate that even at the highest concentration of 5.0 mg/mL, the viabilities of HDPSCs and HGFs exceeded 80% after 7 days of incubation. The live/dead fluorescence staining results of HGFs and HDPSCs in Figures 4C, D further illustrate the substantial cell proliferation at each concentration, with minimal induction of cell death by Q@HMSNs compared to the other groups. These results provide strong evidence that Q@HMSNs induced relatively negligible toxicity to HGFs and HDPSCs.

The effects of different treatments on tubule occlusion and DOM preservation are presented in Figures 5, 6. After the application of HMSNs and Q@HMSNs, the dentinal tubules in the HMSN and Q@HMSN groups were completely obstructed. The cross-sectional images confirmed that these tubes were effectively sealed by the NPs. In contrast, the tubules in the DW, NaF, and Q groups were exposed due to acid attack (Figure 5). After 14 days of erosion and abrasion, the DW and NaF groups showed mostly exposed tubules. Moreover, narrower tubules were observed in group Q, while the tubules were blocked in the HMSN group. In contrast, the specimens treated with Q@HMSNs demonstrated well-defined dentine tubule occlusion with integration of the DOM and microparticles. Moreover, the cross-sectional view shows a notably distinct and thick DOM in the Q@HMSN group compared with the other groups (Figure 6).

The CLSM images revealed the penetration depth of rhodamine B in each group. The DW, NaF, and Q groups showed extensive rhodamine B penetration into the dentinal tubules (Figures 7A–C). In contrast, superficial fluorescent bands were observed in the HMSN and Q@HMSN groups (Figures 7D, E).

Figure 8 shows the EDL of the specimens treated with different solutions. Significantly less EDL was observed in the NaF, Q, HMSN and Q@HMSN groups than in the DW group (p < 0.05). The Q@HMSN group had the lowest EDL among the tested groups (p < 0.05).

The results from ICTP analysis revealed that 44.34 ± 3.37, 40.13 ± 4.38, and 43.35 ± 3.43 g/L ICTP were released in the DW, NaF, and HMSN groups, respectively, with no significant differences among these values (p > 0.05). In contrast, ICTP release was significantly reduced in the Q and Q@HMSN groups (19.37 ± 4.38 and 12.36 ± 2.49 g/L, respectively) (p < 0.05).