β-Carotene liposome-based responsive hydrogel inhibits palmitoylation to restore intracellular calcium homeostasis for effective periodontitis treatment

Abstract

Endoplasmic reticulum (ER) stress is a key pathological mechanism in periodontitis. Our study demonstrates that the periodontitis microenvironment induces ER stress, leading to ER Ca²⁺ efflux via inositol 1,4,5-trisphosphate receptors (IP3R) and subsequent mitochondrial Ca²⁺ overload through the IP3R-GRP75-VDAC1-MCU axis. This disrupts intracellular Ca²⁺ homeostasis and ER-mitochondrial function, resulting in excessive generation of reactive oxygen species (ROS). ROS accumulation induces pathological activation of S-palmitoylation, which amplifies the cascade by enhancing Ca²⁺ transporter activity and establishing a vicious Ca²⁺/ROS cycle. Disrupting this cycle thus presents a promising therapeutic approach. We identify DHHC6 as the specific acyltransferase for the ER Ca²⁺ channel IP3R and β-carotene (β-C) as its natural inhibitor. Therapeutically, β-C suppresses pathological S-palmitoylation, impairs Ca²⁺ transporter activity and restores cellular calcium homeostasis. To enhance its therapeutic delivery, we developed a reactive oxygen species (ROS)-responsive hydrogel (PBTGL) that incorporates β-C-loaded liposomes (L@β-C) and graphene oxide/silver nanoparticles (GO/Ag⁺). PBTGL exerts synergistic antibacterial, anti-inflammatory, and osteo-regenerative effects, resulting in marked amelioration of periodontitis. Thus, PBTGL offers a promising alternative therapeutic strategy for clinical periodontitis management.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-025-03973-z.

Introduction

Periodontitis, a chronic inflammatory disease initiated by dental plaque biofilm, progressively destroys periodontal tissues, leading to tooth loss and a diminished quality of life [1–4]. It also constitutes a modifiable risk factor for systemic disorders, posing substantial global public health challenges [5–7]. Recent advances have introduced novel therapies, including anti-inflammatory procedures and localized delivery of biologics. However, their efficacy is limited by inadequate plaque control, high recurrence rates, and high costs [8–10].

Recently, numerous studies have demonstrated that chronic inflammation and bacterial toxins induce endoplasmic reticulum (ER) stress, thereby impairing the function of local periodontal cells [11, 12].Concurrently, disruption of intracellular calcium ion (Ca²⁺) homeostasis exerts a pivotal role in the early stages of ER stress [13]. Among these, the inositol 1,4,5-trisphosphate receptor (IP3R) serves as a critical ER Ca²⁺ channel [14], interacting with glucose-regulated protein 75 (GRP75), voltage-dependent anion-selective channel 1 (VDAC1), and mitochondrial calcium uniporter (MCU) to form a Ca²⁺ transport complex on the mitochondria-associated membranes (MAMs) [15–17]. This complex mediates ER Ca²⁺ efflux and mitochondrial uptake, leading to mitochondrial Ca²⁺ overload and oxidative damage [18, 19]. A pathological cycle is thus formed, linking ER stress, Ca²⁺ efflux, cytoplasmic/mitochondrial Ca²⁺ overload, and ER-mitochondrial dysfunction, which is considered an essential mechanism exacerbating the pathological process. Previous studies have suggested that S-palmitoylation is a reversible post-translational modification involving the attachment of palmitic acid to cysteine residues via a thioester bond. It regulates the expression, transport, membrane binding, and stability of calcium transporters [20, 21]. Therefore, we hypothesize that moderate inhibition of S-palmitoylation may interrupt this pathological cascade by attenuating calcium transporter activity.

Hence, we screened through computational analysis (protein structure prediction [22], toxicity analysis [23–25], kinetic simulation [26–28], and molecular docking [29–31], etc.) and found that β-carotene (β-C) could selectively bind to the catalytic domain of DHHC-type zinc finger acyltransferase 6 (DHHC6), which is a specific palmitoyl transferase for IP3R. This interaction results in a significant inhibition of DHHC6 activity. Therefore, theoretically, β-C might inhibit pathological palmitoylation of IP3R via DHHC6, thereby restoring intracellular calcium homeostasis. Thus, it could serve as a potential therapeutic candidate for periodontitis. However, the critical challenge lies in local β-C delivery, which requires overcoming challenges such as gingival crevicular fluid flushing, ROS-induced degradation, and biofilm interference [32, 33]. To address these issues, we engineered a ROS-responsive hydrogel (PBTGL) based on polyvinyl alcohol (PVA), borax, and tannic acid (TA). This hydrogel allowed the delivery of graphene oxide/silver nanoparticles (GO/Ag⁺) and liposome-encapsulated β-carotene (L@β-C) (Scheme 1). By inhibiting pathological palmitoylation, β-C efficiently restores Ca²⁺ homeostasis, alleviates endoplasmic reticulum/mitochondrial stress, thereby reducing inflammation, enhancing osteogenic differentiation of bone marrow mesenchymal stem cells (BMSCs), and inhibiting osteoclast generation. Thus, this system alleviates periodontal inflammation and promotes soft and hard tissue regeneration.

Fig. 1.

Construction process of PBTGL and its triple actions: antibacterial, anti-inflammatory, and bone repair. The mechanism involves palmitoylation inhibition to restrict calcium transport activity, remodel calcium homeostasis, and restore cellular function.(IP3R: inositol 1,4,5-trisphosphate receptor, GRP75:interacting with glucose-regulated protein 75, VDAC1:voltage-dependent anion-selective channel 1, MCU: mitochondrial calcium uniporter, STIM1:Stromal Interaction Molecule 1,Orai1:Calcium Release-Activated Calcium Channel Protein 1, mPTP: mitochondrial permeability transition pore, mtDNA: Mitochondrial DNA.)

Materials and methods

Identification of protein interactions via co-immunoprecipitation

RAW 264.7 macrophages were seeded at a density of 6 × 10⁵ cells/well in 6-well plates. when cells reach confluence, the collected cells were lysed with RIPA, the collected proteins were subjected to co-immunoprecipitation (Co-IP) using an anti-IP3R antibody according to the manufacturer’s instructions for the rProtein A/G Magnetic IP/Co-IP Kit. The immunoprecipitated complexes were then separated by routine electrophoresis. The gel bands were excised and sent for protein identification to determine proteins interacting with IP3R.

Molecular docking

Molecular docking was performed using HADDOCK 2.4 with default parameters configured as follows: Protein structures (DHHC6 and calcium-overload-related DHHC isoforms) retrieved from UniProt were preprocessed by removing crystallographic waters, adding hydrogen atoms at pH 7.4 using PDB2PQR, and defining the DHHC domain as the active residues with automatic passive residue selection. Ligands (compounds and 2BP) derived from PubChem underwent energy minimization with the MMFF94 force field, with Gasteiger charges assigned and rotatable bonds automatically detected.

Molecular dynamics and binding free energy calculation

AmberTools22 was used for molecular docking studies, obtained from the official AMBER website (https://ambermd.org). Hydrogen atoms were added, and missing residues were reconstructed using LEaP, a tool from the AmberTools suite. The protein was then parameterized using the ff14SB force field. The ligand molecule was prepared by assigning atomic charges using the AM1-BCC method, and parameterization was conducted using the GAFF (General Amber Force Field). Once the ligand and protein structures were prepared, molecular docking was performed using the SANDER module in Amber, which was employed for energy minimization and conformation sampling. The ligand was placed in the binding site of the protein based on initial predictions by molecular docking. A two-step energy minimization was performed: the first with constraints on the protein backbone, and the second without any constraints to allow for full relaxation of the system. The binding poses were evaluated by calculating the binding free energies using both the Molecular Mechanics/Generalized Born Surface Area (MM/GBSA) and Molecular Mechanics/Poisson-Boltzmann Surface Area (MM/PBSA) methods.

Preparation of L@β-C

L@β-C liposomes loaded with β-carotene (β-C) were prepared using a modified ethanol injection method. 80 mg of egg yolk lecithin, 16 mg of cholesterol, and 0.48 mL of β-C dichloromethane solution (1 mg/mL) were dissolved in 4 mL of anhydrous ethanol and mixed thoroughly. The solution was injected dropwise with a syringe into 20 mL of phosphate buffer (20 mM, pH 6.8) at 42 °C and magnetically stirred for 20 min. The resulting crude liposomes were rotary evaporated at 42 °C to remove the dichloromethane and ethanol. After adjusting the sample volume to 20 mL with phosphate buffer, the sample was sonicated for 10 min at 100 W with 2 s on/2 s off. Filtration was performed using a 0.25 μm nitrocellulose membrane to obtain homogenized liposomes with impurities removed.

Preparation of PBTGL

Polyvinyl alcohol (PVA), borax, and tannic acid were completely dissolved in distilled water to final concentrations of 8 wt%, 2 wt%, and 8 wt%, respectively. A blank gel PVA/Borax (PB) was obtained by adding the borax solution to the PVA solution. Equal volumes of borax and tannic acid solutions were mixed and added to equal volumes of PVA solution to promote viscous PVA/Borax/TA (PBT) gel formation. To prepare PVA/Borax/TA/GO-Ag⁺/L@β-C hydrogels (PBTGL), 10µL GO-Ag⁺ and 100 µL L@β-C were uniformly distributed in the 95 µL PVA solution and further crosslinked by equal volumes of borax and tannic acid solutions to construct PBTGL hydrogels.

Physicochemical characterization of PBTGL

The microscopic morphology of the three hydrogels, PB, PBT, and PBTGL, was observed and characterized by scanning electron microscopy, in which energy spectroscopy was used to verify the successful encapsulation of the actives. The FTIR spectra of the three hydrogels were tested in ATR mode using an infrared spectrometer (Thermo Scientific Nicolet iS20 USA) in the wave number range of 600–4000 cm⁻¹, followed by the analysis and characterization of the molecular structures and chemical compositions of the hydrogels using OMNIC software. Strain amplitude scanning tests of hydrogels were performed on a rheometer (Haake Mars 60 Germany) using an oscillatory mode, where the fixed strain range was 1–1000% and the constant frequency was 1 HZ to determine the critical strain and viscoelastic region of the hydrogels. To demonstrate the self-healing ability of the hydrogels, time-scan tests were performed with 1000% high strain and 1% low strain continuous step changes with a single measurement of 100 s and a total duration of 1000 s for a total of 5 cycles to measure the changes in the hydrogel energy storage modulus G′ and loss modulus G″. The adhesive properties of PBTGL were confirmed using teeth. Before testing, 100 µL of hydrogel was injected into the crown void, and another tooth was attached in an end-to-end fashion, and after curing for 2 h, 200 g weights were suspended to test the adhesive capacity of the hydrogel.

Fluorescence staining

RAW 264.7 macrophages were seeded in 24-well plates at a density of 2 × 10⁶ cells/well. To establish an inflammatory model, cells were stimulated with 100 ng/mL LPS for 12 h. For Mouse bone marrow mesenchymal stem cells (mBMSCs),, a seeding density of 5 × 10⁴ cells/well was employed, followed by 24 h stimulation with 1 µg/mL LPS. Prior to LPS exposure, both cell types were pretreated for 12 h with L@β-C (40 µM),2-BP (40 µM), ML348 (10 µM), ML349 (10 µM), or hydrogel extracts (50 µL/mL), and subsequently co-cultured with LPS for an additional 12 h (RAW 264.7) or 24 h (mBMSCs). Following stimulation, mitochondrial parameters were evaluated using fluorescent probes targeting mitochondrial calcium (mitoCa²⁺), cytosolic calcium (cCa²⁺), mitochondrial reactive oxygen species (mROS), cytosolic ROS (cROS), mitochondrial permeability transition pore (mPTP), and mitochondrial membrane potential (JC-1). Fluorescence imaging was performed via confocal laser scanning microscopy (CLSM) or fluorescence microscopy (Leica, Germany), with quantitative analysis of fluorescence intensity conducted using ImageJ software.

Immunoprecipitation and acyl-biotinyl exchange (ABE) experiments

RAW 264.7 macrophages were seeded at a density of 6 × 10⁵ cells/well in 6-well plates, and other stimulation conditions were the same as fluorescent staining samples. The collected cells were lysed with RIPA, The collected proteins were subjected to co-immunoprecipitation (Co-IP) using an anti-IP3R antibody according to the manufacturer’s instructions for the rProtein A/G Magnetic IP/Co-IP Kit. The immunoprecipitated IP3R protein was subjected to ABE conversion using the IP-ABE Palmitoylation Kit for WB. Following routine electrophoresis and membrane transfer, the membrane was blocked with 5% BSA at room temperature for 2 h, then incubated overnight at 4 °C with Streptavidin-HRP. After three washes with TBST, the membrane was developed to detect the palmitoylation level of the IP3R protein.

RAW 264.7 macrophages were seeded at a density of 6 × 10⁵ cells/well in 6-well plates, mBMSC were seeded at a density of 2.5 × 10⁵ cells/well in 6-well plates, and other stimulation conditions were the same as fluorescent staining samples. The collected cells were lysed with RIPA, The collected proteins underwent ABE conversion using the Total protein-ABE palmitoylation Kit for WB. Subsequently, they were subjected to routine electrophoresis and membrane transfer. The membrane was then blocked with 5% BSA at room temperature for 2 h, followed by overnight incubation at 4 °C with Streptavidin-HRP. After washing three times with TBST, protein palmitoylation levels were detected by developing the membrane.

Transmission electron microscopy (TEM) for RAW 264.7

After the same stimulation conditions as fluorescent staining samples, 2 × 10⁶ cells were fixed in 4% glutaraldehyde for 24 h at 4 ℃. The samples were then incubated with 1% osmium tetroxide, followed by dehydration in an ethanol series and embedding in Araldite resin. Thin sections of 85 nm were cut and stained with uranyl acetate and lead citrate. Ultrastructural analysis was performed by transmission electron microscopy (TEM) (JEM-1400FLASH, JEOL, Japan) at 80 kV.

Modeling of ligature-induced periodontitis

The ligature-induced periodontitis (LIP) model was constructed according to a previous report [34]. Briefly, 6-week-old male C57BL/6 mice (Laboratory Animal Center, Chongqing Medical University) were adaptively fed for at least 3 days. Under chloral hydrate anesthesia, the right maxillary second molars were tied with 5.0 silk and knotted on the palatal side. Unligated mice were used as controls. 10 days after ligation, Saline/PERIO/different groups of hydrogels were administered 5 µL per side of the second molar, once a day from day 10 to day 24. Mice were housed with 12 h dark/light cycles at 21 °C to 23 °C and 45% to 55% relative humidity. All mice were euthanized by cervical dislocation after 21 days, the gingival crevicular fluid of the right maxillary second molar, pericardial blood, organs, and right maxillary bone were collected rapidly. The animal experiments were approved by the Ethics Committee (Stomatological Hospital of Chongqing Medical University, 2024(145)).

Proteome sequencing and bioinformatics analysis

RAW 264.7 macrophages were seeded in 6-well plates at 6 × 10⁵ cells/well and divided into Control, LPS, and LPS + PBTGL groups. For the LPS + PBTGL group, 400 µL PBTGL was added to the upper chamber of Transwell inserts. After LPS stimulation, Cells were collected, flash-frozen in liquid nitrogen, and stored at −80 °C before proteomic sequencing performed by Shanghai Applied Protein Technology. Protein samples were tryptically digested and analyzed via liquid chromatography-tandem mass spectrometry (LC-MS/MS) in data-independent acquisition (DIA) mode. DIA-NN 1.8.1 was employed for data processing, with differentially expressed proteins (DEPs) defined by |fold change (FC)| > 1.5 (upregulated: FC > 1.5; downregulated: FC < 0.67) and P < 0.05. Subsequent bioinformatic analyses included hierarchical clustering analysis, Gene Ontology (GO) functional enrichment, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway mapping, Gene Set Enrichment Analysis (GSEA), and protein-protein interaction (PPI) network construction.

Statistical analysis

All data are expressed as the mean ± SD with a minimum of three independent replicates. The unpaired Student’s t test (two groups) and one-way ANOVA (more than two groups) were performed for statistical analyses, and post hoc analysis was performed using Tukey’s test. Differences were considered statistically significant at p < 0.05 with two-sided testing (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001) and ns indicated non-significant. All statistical analyses and the statistical charts were performed with GraphPad Prism 10.0 or OriginPro 2021.

Results

Screening for natural inhibitors targeting DHHC6, the IP3R-specific acyltransferase

To simulate the periodontitis microenvironment, cells were stimulated with lipopolysaccharide (LPS). Initial observations showed that LPS stimulation induced cytosolic Ca²⁺ overload in both RAW264.7 cells and mouse bone marrow stromal cells (mBMSCs) (Fig. S1A, B). Concurrently, LPS upregulated IP3R expression in periodontal tissues (Fig. S1C). Given its role as an endoplasmic reticulum (ER) calcium release channel, IP3R was subsequently investigated. Following LPS exposure, co-immunoprecipitation (Co-IP) analysis of IP3R in RAW264.7 cells confirmed an interaction between the acyltransferase DHHC6 and IP3R (Fig. 2A). In parallel, LPS stimulation significantly enhanced global protein S-palmitoylation in RAW264.7 cells, with notably intensified bands observed within the 270–320 kDa molecular weight range corresponding to IP3R (Fig. S1D). Given existing evidence that palmitoylation regulates IP3R expression, stability, and membrane localization^23,24, we propose that inhibition of DHHC6 activity may serve as a critical therapeutic strategy to suppress pathological ER Ca²⁺ release in the periodontitis microenvironment. To develop DHHC6-targeted therapeutics, compounds were systematically screened from PubChem via substructure queries for hydrophobic chains (12–20 carbons, permitting unsaturation or branching) terminating in electrophilic groups, carboxylic acids, or thioesters. The initial retrieval (n = 143) was refined using molecular weight (< 600 Da) and rotatable bond (≤ 10) filters, followed by toxicity screening through ProTox-II and ADMETlab 2.0. Fifteen compounds with predicted LD₅₀ values > 1500 mg/kg and synthetic accessibility were prioritized. Among them, carotenoid derivatives displayed optimal polarity for penetrating the hydrophobic DHHC6 binding pocket due to structural compatibility (Fig. 2B, Table S1). Molecular docking using HADDOCK 2.4 identified β-carotene (β-C) as the top-performing compound, with the highest docking score (–50.5 ± 1; Table S2), outperforming the reference inhibitor 2-bromopalmitate (2BP) (Fig. 2C), and exhibiting the lowest cytotoxicity in its chemical class (Fig. 2D). Cross-docking against Ca²⁺-overload-associated DHHC isoforms (DHHC2/3/5/6/7/9) confirmed that β-C had the highest binding selectivity for DHHC6 (Fig. 2E). Binding conformations of β-C and 2BP with DHHC6 showed that both compounds impair DHHC6 activity by occupying its catalytic site at Cys156 (Fig. 2F). Further analysis via MMPBSA/MMGBSA (AMBER 22) and GROMACS simulations revealed that β-C had comparable binding free energy (ΔG = − 31.87 ± 5.67 kcal/mol) to 2BP (ΔG = − 37.45 ± 4.94 kcal/mol) (Fig. 2G), and exhibited superior binding stability of DHHC6-bound β-C, as demonstrated by root mean square fluctuation (RMSF), reduced backbone flexibility, Gibbs free energy landscapes, and secondary structure retention (Fig. S1E-J).

Fig. 2.

Screening for natural inhibitors of DHHC6, the specific acyltransferase for IP3R. (A) Co-IP confirming DHHC6 as the specific acyltransferase for IP3R. (B) Molecular Weight (MW) and LogP values of the 13 candidate compounds. (C) Comparison of docking metrics between the 13 candidate compounds, 2BP, and DHHC6. (D) Toxicity analysis (of the candidates. (E) Assessment of β-carotene’s specific binding to DHHC6. (F) 3D models depicting the docking poses of 2BP and β-carotene with DHHC6. (G) MMGBSA calculations

L@β-C restores cellular function via inhibition of palmitoylation

Liposomes loaded with β-C (L@β-C) were synthesized via ethanol injection (Fig. S2A). The encapsulation efficiency of L@β-C was 70.88% ± 3.2%, which indicates good reproducibility between batches (Fig. S2F). Transmission electron microscopy (TEM) images confirmed successful synthesis with an average particle size of 290 nm (Fig. S2B, E). Liposome encapsulation significantly improved the storage stability and antioxidant stability of β-C (Fig. S2C, D). For the final therapeutic formulation, 100 µL of L@β-C was incorporated into a 200 µL PBTG hydrogel matrix to yield a composite PBTGL with a final β-carotene concentration of 22.23 µg/mL. Continuous fluorescence monitoring showed that Ca²⁺ overload occurred as early as 2 h post-LPS induction, whereas ROS accumulation was detected at 9 h. Both L@β-C and the reference inhibitor 2BP effectively normalized these pathological changes (Fig. 3A, B), suggesting that Ca²⁺ overload is an early event in LPS-induced stress, and suppression of S-palmitoylation mitigates this cascade. Next, S-palmitoylation was artificially enhanced using the APT1 inhibitor ML348 and the APT2 inhibitor ML349. The optimal concentrations (10 µM) were determined via CCK-8 assay (Fig. 3C). Enhanced palmitoylation worsened LPS-induced cytoplasmic Ca²⁺ overload and ROS accumulation in macrophages, effects that were reversed by L@β-C (Fig. 3Da, b). Similar outcomes were observed in mBMSCs (Fig. 3Dc, d).

Fig. 3.

L@β-C regulates calcium homeostasis by inhibiting palmitoylation and limiting calcium transporter function. (A) Continuous detection of Ca²+. (B) Continuous detection of ROS. (C) Cell viability of RAW264.7 cells assessed by the CCK-8 assay to evaluate the cytotoxicity of ML348 (APT1 inhibitor) and ML349 (APT2 inhibitor) at different concentrations. (D) Fluorescence detection of cellular Ca²⁺and ROS levels under different treatment conditions, Simultaneous use of ML348(APT1 inhibitor) and ML349(APT2 inhibitor) is expected to increase cellular palmitoylation levels by inhibiting depalmitoylation: (a) Cytosolic Ca²⁺in RAW264.7 cells detected by Fluo-4 (green); (b) Cytosolic ROS in RAW264.7 cells detected by DCFH-DA (green); (c) Cytosolic Ca²⁺in mBMSCs detected by Fluo-4 (green); (d) Cytosolic ROS in mBMSCs detected by DCFH-DA (green). (E) Expression levels of calcium transporter proteins and inflammation-related proteins in RAW264.7 cells under different treatments. (F) Expression levels of calcium transporter proteins and osteogenic differentiation-related proteins in mBMSCs under different treatments. (G-J) Expression levels of ER stress-related genes, mitochondrial dynamics-related genes, calcium transporter genes, and inflammation-related genes in RAW264.7 cells under different treatments. (K-N) Expression levels of ER stress-related genes, mitochondrial dynamics-related genes, calcium transporter genes, and osteogenic differentiation-related genes in mBMSCs under different treatments. (O) Expression of IP3R and palmitoylation level per unit mass of IP3R in RAW264.7 cells under different treatments detected by Co-IP and ABE assays. (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.)

Real-time quantitative polymerase chain reaction (RT-qPCR) analysis showed that enhanced S-palmitoylation intensified LPS-induced ER stress, mitochondrial dysregulation, calcium transporter overexpression, and inflammatory activation in macrophages—all of which were attenuated by L@β-C. Likewise, in mBMSCs, L@β-C reversed enhanced S-palmitoylation–driven ER stress, mitochondrial dysfunction, calcium dysregulation, and osteogenic suppression (Fig. 3K-N). Western blot (WB) analysis further validated that enhanced palmitoylation exacerbated calcium transporter overexpression and pro-inflammatory protein levels in macrophages (Fig. 3E) while suppressing osteogenic markers in mBMSCs (Fig. 3F); L@β-C counteracted these effects through palmitoylation inhibition.

To confirm the mechanism, global protein palmitoylation levels were measured post-LPS induction in both macrophages and BMSCs. L@β-C significantly reduced LPS-induced palmitoylation Fig. S2G, H). IP3R was immunoprecipitated from macrophages, followed by an acyl-biotin exchange (ABE) assay, which showed that LPS enhanced both IP3R expression and palmitoylation levels, LPS stimulation significantly increased the IP3R S-palmitoylation ratio from 0.742 ± 0.043 in the Control group to 0.982 ± 0.023. This trend was markedly reversed by treatment with L@β-C, which reduced the ratio to 0.410 ± 0.145. (Fig. 3O).

Evaluation of the anti-inflammatory properties of L@β-C

We next assessed the anti-inflammatory effects of L@β-C in macrophages. First, its biocompatibility was evaluated using the CCK-8 assay. RAW264.7 cells exposed to 80 µM L@β-C for 7 days exhibited viability comparable to the control group, indicating no significant cytotoxicity (Fig. 4B). The ability of L@β-C to mitigate Ca²⁺ overload and ROS accumulation was then examined. Fluorescence imaging revealed that LPS stimulation induced marked cytosolic Ca²⁺ overload and ROS accumulation in RAW264.7 cells (Fig. 4Aab). This Ca²⁺ dysregulation also triggered sustained opening of the mPTP, compromising mitochondrial membrane integrity (Fig. 4Ac) and increasing mitochondrial Ca²⁺ and ROS levels (Fig. 4Ade). These effects were effectively reversed by both L@β-c and the palmitoylation inhibitor 2BP.

Fig. 4.

Validation of L@β-C anti-inflammatory activity. (A) Fluorescence detection of cellular Ca²⁺and ROS levels under different treatment conditions: (a) Cytosolic Ca²⁺detected by Fluo-4 (green); (b) Cytosolic ROS detected by DCFH-DA (green); (c) mPTP opening detected by calcein-AM (green) loading/CoCl₂ quenching; (d) Mitochondrial Ca²⁺detected by Rhod-2 (red); (e) Mitochondrial ROS detected by MitoSOX (red). (B) The CCK-8 assay assessed cell viability to observe the cytotoxicity of L@β-C at different concentrations. (C-F) Expression levels of ER stress-related genes, mitochondrial dynamics-related genes, calcium transporter genes, and inflammation-related genes in RAW264.7 cells under different treatments. (G) Expression levels of calcium transporter proteins and inflammation-related proteins in RAW264.7 cells under different treatments. (H) TEM images showing the morphology of the ER and mitochondria in RAW264.7 cells under different treatments. Red arrows indicate ER-mitochondria contact sites in the LPS group; blue arrows indicate mitochondrial swelling and rupture. (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.)

Expression of ER stress-related genes, mitochondrial dynamics regulators, calcium transporters, and inflammation-related genes was quantified via RT-qPCR. LPS significantly upregulated markers of ER stress (Fig. 4C), mitochondrial dysfunction (Fig. 4D), calcium transporter overexpression (Fig. 4E), and inflammatory activation (Fig. 4F). These alterations were reversed by L@β-C and 2BP, which also promoted M2 macrophage polarization (Fig. S3A, B). WB analysis showed that LPS increased the protein levels of inflammatory markers and calcium transporters, while L@β-C and 2BP restored their expression to near-baseline levels (Fig. 4G).

Finally, transmission electron microscopy (TEM) was used to observe the morphology of the ER and mitochondria in RAW264.7 cells (Fig. 4H). LPS stimulation caused noticeable ER swelling and increased ER-mitochondria contact sites (red arrows), likely facilitating mitochondrial Ca²⁺ transfer. This led to mitochondrial membrane swelling and rupture (blue arrows). Although mild mitochondrial membrane swelling persisted in the LPS + L@β-C and LPS + 2BP groups, both treatments substantially preserved the structural integrity of the ER and mitochondria. These findings indicate that the pathological cascade—comprising ER stress, Ca²⁺ release and translocation, mitochondrial dysfunction, and ROS accumulation—strongly activates inflammatory responses. L@β-C effectively blocks this cascade by inhibiting palmitoylation.

Evaluation of the bone homeostasis-promoting properties of L@β-C

We investigated the effects of L@β-C on mouse bone marrow stromal cells (mBMSCs) and osteoclasts. The biocompatibility of L@β-C was first assessed using the CCK-8 assay. mBMSCs exposed to 40 µM L@β-C for 7 days exhibited comparable viability to control cells, indicating minimal cytotoxicity (Fig. 5F). We next evaluated the ability of L@β-C to mitigate Ca²⁺ overload and ROS accumulation. Fluorescence imaging revealed that LPS stimulation caused marked cytosolic calcium overload and ROS generation in mBMSCs (Fig. 5Aab). This overload led to sustained opening of the mPTP, disrupting mitochondrial membrane integrity (Fig. 5Ac), and resulted in elevated mitochondrial Ca²⁺ and ROS levels (Fig. 5Ade). These pathological changes were effectively reversed by both L@β-C and the palmitoylation inhibitor 2BP.

Fig. 5.

Validation of the osteogenic differentiation-promoting and osteoclast formation-inhibiting properties of L@β-C. (A) Fluorescence detection of cellular Ca²⁺and ROS levels under different treatment conditions: (a) Cytosolic Ca²⁺detected by Fluo-4 (green); (b) Cytosolic ROS detected by 2’,7’-DCFH-DA (green); (c) mPTP opening detected by calcein-AM (green) loading/CoCl₂ quenching; (d) Mitochondrial Ca²⁺detected by Rhod-2 (red); (e) Mitochondrial ROS detected by MitoSOX (red). (B-E) Expression levels of ER stress-related genes, mitochondrial dynamics-related genes, calcium transporter genes, and osteogenic differentiation-related genes in mBMSCs under different treatments. (F) Cell viability of mBMSCs assessed by the CCK-8 assay to observe the cytotoxicity of L@β-C at different concentrations. (G, H) ALP staining and ARS staining showing the osteogenic differentiation capacity of mBMSCs under different treatments. (I) TRAP staining showing osteoclast formation under different treatments. (J) Immunofluorescence staining for CTSK (green) and phalloidin (red) showing osteoclast formation and CTSK expression under different treatments. (K) Expression levels of calcium transporter proteins and osteogenic differentiation-related proteins in mBMSCs under different treatments. (L) Quantification of osteoclast number. (M) Quantification of osteoclast-positive area. (N) Expression levels of osteoclast-related genes under different treatments. (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.)

RT-qPCR analysis showed that LPS induced upregulation of ER stress-related genes (Fig. 5B), dysregulation of mitochondrial dynamics (Fig. 5C), overexpression of calcium transporters (Fig. 5D), and inhibition of osteogenic markers (Fig. 5E). These effects were reversed by L@β-C and 2BP, which also promoted osteogenesis. Alkaline phosphatase (ALP) and Alizarin Red S (ARS) staining further supported that LPS suppressed osteogenic differentiation, while L@β-C and 2BP effectively rescued it (Fig. 5G, H). WB analysis confirmed that LPS increased calcium transporter expression and reduced osteogenesis-associated protein levels, both of which were corrected by L@β-C and 2BP (Fig. 5K).

Finally, the inhibitory effects of L@β-C on osteoclast formation were assessed using TRAP staining, Cathepsin K(CTSK) immunofluorescence, and RT-qPCR. L@β-C and 2BP markedly reduced osteoclast formation (Fig. 5L). and reduced TRAP-positive areas (Fig. 5M). Additionally, cytoplasmic CTSK expression (Fig. 5J) and the mRNA levels of CTSK and TRAP (Fig. 5N) were significantly decreased.

Synthesis, characterization, and in vitro evaluation of PBTGL hydrogel

PBTGL hydrogel was synthesized by incorporating L@β-C and (GO/Ag⁺) into a base mixture of PVA, borax, and TA. For comparison, PBTG hydrogel was prepared using GO/Ag⁺ without L@β-C, while PBT hydrogel was prepared without either additive. The successful synthesis of GO/Ag⁺ was confirmed (Fig. S4A, B), with a silver content of 0.29 mg/mL. Upon incorporation into the hydrogel, the final silver concentration reached 9.67 µg/mL. Scanning electron microscopy (SEM) verified the encapsulation of L@β-C and GO/Ag⁺ without compromising the hydrogel structure (Fig. 6B). The PBTGL hydrogel exhibited excellent plasticity, injectability, viscosity, and self-healing capability (Fig. 6C). Rheological analysis revealed pronounced shear-thinning behavior and efficient self-recovery (Fig. S4C). FTIR spectroscopy demonstrated that the incorporation of L@β-C and GO/Ag⁺ did not interfere with borate ester bond formation or alter the phenolic hydroxyl groups of TA (Fig. 6D). Furthermore, the hydrogel exhibited ROS-responsive, concentration-dependent degradation (Fig. 6E) and triggered drug release under oxidative conditions (Fig. 6F). Cytocompatibility assays showed negligible toxicity toward RAW264.7 cells and mBMSCs after 72 h of coculture (Fig. S5A, B). Compared to the group treated with the same concentration of L@β-C alone, the PBTGL conditioned release extract exerted consistent biological effects, effectively suppressing LPS-induced cytosolic calcium overload and ROS accumulation in macrophages (Fig. S6 A, B), while also reversing LPS-induced endoplasmic reticulum stress, mitochondrial dynamic disturbances, overexpression of calcium transport proteins, and inflammatory activation (Fig. S6 C-F).

Fig. 6.

Synthesis of PBTGL and its in vitro antibacterial validation. (A) Schematic diagram of the PBTGL synthesis process. (B) SEM images of hydrogels from different groups and the EDS spectrum of PBTGL. (C) Images demonstrating the plasticity, injectability, adhesiveness, and self-healing properties of PBTGL. (D) FTIR spectra of hydrogels from different groups. (E) Degradation profile of PBTGL in H₂O₂ solutions at different concentrations. (F) Drug release profile of PBTGL in H₂O₂ solutions at different concentrations. (G) Live/Dead staining of E. coli, S. aureus, and P. gingivalis under different treatments. (H-J) Absorbance at 600 nm of bacterial suspensions (E. coli, S. aureus, P. gingivalis) under different treatments. (K-M) Absorbance at 590 nm of eluted crystal violet dye from biofilms (E. coli, S. aureus, P. gingivalis) under different treatments. (n = 3, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.)

Antibacterial activity was tested against Escherichia coli, Staphylococcus aureus, and Porphyromonas gingivalis. Hydrogels were incubated with bacterial suspensions for 2 h. PBTGL showed significantly stronger antibacterial effects than the commercial product PERIO (minocycline hydrochloride ointment) (Fig. 6H-J, Fig. S7A). Live/Dead staining confirmed this antibacterial efficacy (Fig. 6G). Given that periodontal pathogens often exist in biofilms, biofilm eradication was assessed via crystal violet staining. Biofilms extensively covered pore surfaces in the saline and PBT control groups, while PBTGL treatment markedly reduced biofilm coverage, outperforming PERIO (Fig. 6K-L, Fig. S7B).

In vivo evaluation of the PBTGL hydrogel in alleviating periodontitis

To establish a mouse model of periodontitis, a 0.1-mm ligature was placed around the right maxillary second molar for 10 days. Subsequently, hydrogel formulations were injected daily into the periodontal pockets for two weeks to evaluate alveolar bone resorption and therapeutic efficacy (Fig. 7A). Micro-CT reconstructions revealed more pronounced alveolar ridge resorption in the ligature and PBT groups compared to other groups. Notably, microCT analysis revealed an intermediate phenotype in the PBTG group. This group exhibited less alveolar bone loss than the ligature and PBT groups, but more than the PBTGL group. This suggests that the GO/Ag⁺ component alone provides some therapeutic benefit through its antimicrobial activity, but does not have the same comprehensive regenerative capacity as the full PBTGL formulation. Cross-sectional and sagittal views showed the largest hypodense areas in the ligature and PBT groups and the smallest in the PBTGL group (Fig. 7B).

Fig. 7.

Evaluation of the in vivo performance of PBTGL in a mouse periodontitis model. (A) Schematic diagram of the periodontitis model establishment and treatment procedure in C57BL/6 mice. (B) 3D reconstruction, sagittal, and cross-sectional micro-CT images of the alveolar bone from different treatment groups. (C) Quantitative analysis of the CEJ-ABC distance. (D-G) Bone volume analysis for the alveolar bone surrounding the maxillary second molar in the mesio-distal dimension. (H) Culture plates of GCF samples collected from the sulcus of the second molar in different treatment groups. (I) H&E staining of alveolar bone sections. (J) Masson’s trichrome staining of alveolar bone sections. (K) TRAP staining of alveolar bone sections. (n = 6, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.)

Quantitative analysis demonstrated significantly shorter cementoenamel junction to alveolar bone crest (CEJ-ABC) distances at the second molars in the PBTGL group (Fig. 7C). Additionally, trabecular bone parameters indicated that PBTGL treatment led to significantly higher bone volume/total volume (BV/TV) ratios compared to the ligature group (Fig. 7D), along with superior trabecular number (Tb.N), thickness (Tb.Th), and spacing (Tb.Sp) (Fig. 7E-G).

Cultures of gingival crevicular fluid from treated periodontal pockets confirmed PBTGL’s potent antibacterial efficacy in vivo (Fig. 7H). Supporting the mechanism, the PBTG group also exhibited a significant reduction in bacterial viability compared to the ligature and PBT groups, directly attributing the observed partial efficacy to the localized release of GO/Ag⁺. Histological staining with hematoxylin and eosin (H&E) and Masson’s trichrome revealed reduced bone loss and preservation of the junctional epithelium in the PBTGL group (Fig. 7I, J). Histological evaluation of the PBTG group corroborated these findings, showing a modest attenuation of inflammatory infiltration and tissue destruction compared to the disease controls. TRAP staining further indicated a marked reduction in osteoclastogenesis following PBTGL treatment (Fig. 7K). Immunofluorescence analysis demonstrated enhanced M1-to-M2 macrophage polarization (Fig. S8A, C) and improved osteogenic differentiation (Fig. S8B, D). Finally, a comprehensive biosafety evaluation—including hemolysis assay (Fig. S9A), hepatic and renal function tests (Fig. S9B-I), and organ histopathology (Fig. S10), performing ELISA and flow cytometry analyses of blood samples (Fig. S11A, B), and spleen PCR (Fig. S11C)—confirmed the excellent biocompatibility of PBTGL, no observable systemic toxicity, organ damage, or detectable systemic inflammatory response in vivo.

PBTGL regulates inflammatory expression of macrophages through calcium signaling pathway

Proteomic analysis identified distinct protein expression profiles among the treatment groups. Specifically, 380 proteins were upregulated and 244 downregulated in LPS-treated compared with control groups (|log2FC| > 1.5, P < 0.05). PBTGL treatment reversed many of these alterations, with 42 proteins upregulated and 196 downregulated in the LPS + PBTGL compared with LPS groups (Fig. 8A-C). Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses of the differentially expressed proteins (DEPs) revealed strong associations with inflammatory processes and organelle-related functions (Fig. S12).

Fig. 8.

Bioinformatics analysis of the mechanism of action of PBTGL. (A) Volcano plot of DEPs between the Control and LPS groups. (B) Volcano plot of DEPs between the LPS and LPS + PBTGL groups. (C) DEPs common to or unique in pairwise comparisons among the three groups (Control, LPS, LPS + PBTGL). (D) Clustering analysis of DEPs across the three groups. (E) GO-BP enrichment analysis for Cluster 3. (F) GO-MF enrichment analysis for Cluster 3. (G-H) GSEA comparing the LPS group to the LPS + PBTGL group. (I-J) Immunofluorescence staining images and quantitative analysis of IP3R expression in periodontal tissues from the different groups

Hierarchical clustering analysis divided DEPs into five functional modules (Fig. 8D), and cluster 3 was particularly enriched in phosphoinositide biosynthesis (biological process), calmodulin binding, and phosphatidylinositol-3,4,5-trisphosphate binding (molecular functions) (Fig. 8E, F). Gene set enrichment analysis (GSEA) showed that PBTGL treatment significantly suppressed the interferon-γ response (NES = 2.108, FDR < 0.01) and calcium-dependent phospholipid signaling (NES = 2.058, FDR < 0.01) in RAW264.7 macrophages (Fig. 8G, H).

Immunofluorescence analysis revealed that PBTGL significantly suppressed IP3R expression in periodontal tissues (Fig. 8I, J), suggesting a key role in modulating calcium homeostasis by regulating both IP3R expression and activity. Overall, these results demonstrate that PBTGL restores calcium balance and reprograms cellular function by inhibiting palmitoylation, thereby attenuating calcium transporter dysregulation and suppressing pathological calcium signaling cascades (Scheme 1).

Discussion

Accumulating evidence suggests that the endoplasmic reticulum (ER), a central organelle involved in protein synthesis and calcium storage in eukaryotic cells [35], undergoes substantial stress when persistently exposed to inflammatory mediators or bacterial toxins (e.g., LPS) during periodontitis [11]. In studies of other inflammatory diseases, ER stress triggers rapid Ca²⁺ release via the ER, leading to cytosolic Ca²⁺ overload and subsequent cellular dysfunction. However, how ER Ca²⁺ release contributes to the pathological progression of periodontitis remains incompletely understood. Our findings demonstrate that calcium overload, primarily initiated by ER Ca²⁺ release, serves as an upstream trigger in the cellular responses underlying periodontitis. ER stress activates Ca²⁺ release through IP3R channels, while concurrent calcium depletion from the ER triggers store-operated calcium entry (SOCE) mediated by the STIM1–Orai1 pathway, further aggravating cytosolic Ca²⁺ overload. Notably, ER stress and pathological stimuli promote the formation of mitochondria-associated membranes (MAMs), enabling accelerated calcium transfer to mitochondria via the IP3R–GRP75–VDAC1–MCU axis. Mitochondrial calcium overload disrupts oxidative phosphorylation [36], impairs ATP synthesis, and leads to excessive ROS production. The synergistic effect of elevated Ca²⁺ and ROS induces the pathological opening of the mPTP [37, 38], resulting in depolarization of the mitochondrial membrane potential (ΔΨm) and the release of pro-apoptotic factors, such as cytochrome c and additional ROS. These events ultimately compromise ER-mitochondrial functional integrity. Mechanistically, this dual-damage process—characterized by calcium dysregulation and ER-mitochondrial dysfunction—not only impairs osteogenic differentiation (via downregulation of critical transcriptional regulators) but also promotes inflammatory cascades through NLRP3 inflammasome activation. This pathological mechanism represents a novel etiological framework for periodontitis and justifies the therapeutic strategies developed in this study.

S-palmitoylation, a reversible post-translational modification involving the covalent attachment of palmitate (a 16-carbon fatty acid) to cysteine residues via thioester bonds, facilitates membrane anchoring and stabilizes transmembrane protein expression. Excessive mitochondrial ROS can significantly upregulates global protein palmitoylation, promoting pyroptosis via enhanced membrane translocation of GSDMD [39, 40]. Recent evidence indicates that palmitoylation also reduces NLRP3 solubility, driving its phase separation and activation, thereby sustaining chronic inflammation [41]. Moreover, palmitoylation enhances calcium transporter function by stabilizing membrane anchoring, extending protein half-life, and increasing channel activity [20]. This may further amplifies pathological calcium flux between the ER and mitochondria, potentially sustaining the vicious cycle of Ca²⁺ overload, ROS generation, and ER–mitochondrial dysfunction. We hypothesize that dysregulated Ca²⁺–ROS crosstalk establishes a self-reinforcing feedforward loop: calcium overload → ER/mitochondrial damage → ROS accumulation → hyper-palmitoylation → exacerbated calcium dysregulation. Consequently, interrupting this loop by modulating pathological palmitoylation may not only help restore. Interrupting this loop by pharmacologically inhibiting pathological palmitoylation may not only restore intracellular calcium homeostasis but also reverse cellular dysfunction characteristic of periodontitis, thus representing a novel therapeutic approach.

As the principal ER calcium release channel, IP3R initiates calcium dysregulation cascades. Consistent with previous studies [21], co-immunoprecipitation assays confirmed DHHC6 as the IP3R-specific acyltransferase. We screened 13 candidate compounds targeting the DHHC6 catalytic domain, identifying β-carotene (β-C) as a promising candidate based on molecular docking and toxicity profiling. Dietary β-C intake has been associated with reduced periodontitis risk [42, 43], a benefit possibly attributable to its ability to modulate mitochondrial metabolism in macrophages via the JAK–STAT pathway [44]. Docking and molecular dynamics simulations suggested that β-C exhibits stronger binding affinity and greater selectivity for DHHC6 than 2-bromopalmitate (2BP), a non-specific palmitoylation inhibitor known for its hepatotoxicity and off-target effects [45, 46]. It is important to note that the observed modulation of IP3R palmitoylation by β-C, as indicated by our ABE assay, may be partially attributable to its well-established antioxidant activity, which reduces the oxidative stress known to stimulate palmitoylation. A direct and specific inhibition of DHHC6 enzymatic activity by β-C, while a compelling hypothesis based on our in silico findings, requires further validation through direct enzymatic assays or genetic (knockdown/overexpression) models.

To overcome β-C’s inherent limitations—instability, low solubility, and poor bioavailability—we developed a liposomal formulation (L@β-C) via ethanol injection. This delivery method significantly improved β-C stability and intracellular uptake. Mechanistically, L@β-C treatment was associated with a reduction in IP3R palmitoylation, thereby restoring calcium homeostasis and reducing LPS-induced cytosolic Ca²⁺ overload and ROS accumulation. Additionally, L@β-C promoted M2 macrophage polarization, enhanced osteogenesis in BMSCs, and inhibited osteoclastogenesis by attenuating ER stress, normalizing mitochondrial dynamics, and modulating immune–osteogenic crosstalk. These multifaceted effects enabled L@β-C to restore the pathological periodontal microenvironment and improve bone homeostasis.

However, clinical application of L@β-C is challenged by rapid clearance from periodontal pockets due to gingival crevicular fluid and saliva, as well as sustained toxin release from plaque biofilms. To address these limitations, we designed an intelligent hydrogel delivery system (PBTGL) composed of PVA, borax, tannic acid (TA), GO/Ag⁺, and L@β-C. TA enhances mucosal adhesion and provides antioxidant protection, thereby prolonging the local retention of the hydrogel. The dynamic borate ester bonds confer ROS-responsive characteristics, enabling triggered drug release in the inflammatory microenvironment. GO/Ag⁺ nanoparticles deliver antimicrobial action, and L@β-C modulates cellular responses. To mitigate potential safety concerns associated with excessive Ag⁺, we optimized the loading amount of GO/Ag⁺ to achieve the strongest antibacterial efficacy within a safe threshold. Reassuringly, Our comprehensive biosafety assessment, including systemic toxicity evaluation, histopathology of major organs, and analysis of local tissue response, confirmed that the low dose of Ag⁺ in PBTGL, coupled with its localized release profile, does not elicit detectable systemic toxicity or organ damage. Meanwhile, PBTGL exhibited potent in vitro and in vivo antibacterial effects, efficiently eliminating plaque biofilms. In a murine periodontitis model, we administered PBTGL via injection into the periodontal pockets adjacent to the second molars, mirroring the clinical application method of the commonly used antimicrobial gel PERIO. During the initial two days, the elevated local ROS levels characteristic of periodontitis triggered the rapid disintegration of PBTGL and the consequent release of therapeutic agents. As the treatment progressed, we observed that PBTGL could be effectively retained within the periodontal pocket for 2–3 days, which is sufficient to disrupt the pathogenic cycle during a critical therapeutic window. Ultimately, PBTGL significantly reduced LPS-induced alveolar bone loss and inflammation while promoting periodontal tissue regeneration. Despite these promising findings, we acknowledge the limitations of this study. Future work should focus on obtaining direct evidence for DHHC6 inhibition, designing site-specific inhibitors (e.g., peptides) targeting the palmitoylation sites on IP3R and on improving endoplasmic reticulum (ER)-targeted drug delivery to enhance therapeutic precision. Furthermore, to better align with clinical therapeutic requirements and extend the retention time, the proportional composition of the hydrogel components should be optimized according to specific clinical scenarios to ensure both safety and long-term effectiveness.

Conclusion

In summary, to address the pathological microenvironment of periodontitis, we developed PBTGL hydrogel—a platform exhibiting intelligent stimulus-responsive release, high biocompatibility, and robust biostability. PBTGL delivers potent antibacterial effects via its GO/Ag⁺ components, while the encapsulated β-carotene specifically inhibits palmitoylation. This palmitoylation blockade impairs calcium transporter function, thereby restoring periodontal calcium ion homeostasis. The resultant dual actions achieve effective anti-inflammatory and pro-osteogenic outcomes. Collectively, the engineered PBTGL hydrogel holds significant promise as a novel therapeutic strategy for periodontitis treatment.

Author contributions

Shun Huang : Conceptualization, Methodology, Investigation, Visualization, Formal analysis, Writing—original draft, Writing—review & editing. Menghan Li : Methodology, Investigation, Visualization, Formal analysis, Writing—original draft, Writing—review & editing. Zishuo Cheng: Methodology, Investigation, Visualization, Formal analysis, Writing—original draft. Minyu He : Investigation. Jiaju Deng : Investigation. Danlan Zhang : Investigation. Miao Tan : Investigation. Shijia Ding : Conceptualization, Supervision, Funding acquisition, Project administration, Writing—review & editing. Haiping Wu: Conceptualization, Supervision, Funding acquisition, Project administration, Writing—review & editing. Lan Huang: Conceptualization, Supervision, Funding acquisition, Project administration, Writing—review & editing.

Funding

This work was funded by the National Natural Science Foundation of China No. 82170989 (LH); National Natural Science Foundation of China No. 82272432 (SJD); National Natural Science Foundation of China No. 82402750 (HPW); The Natural Science Foundation of Chongqing, China CSTB2022NSCQ-MSX0794(LH); Chongqing medical scientific research project (Joint project of Chongqing Health Commission and Science and Technology Bureau) 2025GDRC001(LH); Sichuan Science and Technology Program 2024NSFSC1541(HPW);2025 Chongqing Medical University Elite Postgraduate Cultivation Program BJRC202504 (MHL).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The animal experiments were approved by the Ethics Committee (Stomatological Hospital of Chongqing Medical University, 2024(145)).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.