Dastinib (HY-10181, Purity: 99.83%), Quercetin (HY-18085, Purity: 99.80%), Streptozotocin (HY-13753), PEG400 (HY-Y0873A), and Tween-80 (HY-Y1891) were supplied by MedChemExpress (Shanghai, China). Paraformaldehyde (PFA, BL539A), hematoxylin and eosin staining kit (H&E staining kit, C0105S) was obtained from Beyotime Biotechnology Company (Shanghai, China). Tartrate Resistant Acid Phosphatase (TRAcP) staining kit (BB-4421) was obtained from Bestbio Company (Nanjing, China). Ethylenediaminetetraacetic acid (EDTA, E8008) was purchased from Sigma–Aldrich (Sydney, Australia). Primary antibodies against LC3A/B (D3U4C) was obtained from Cell Signaling Technology (Massachusetts, USA). Primary antibodies against P62 (ab109012) and BECLIN1 (ab302670) were received from Abcam (Cambridge, UK).

All animal experiments were conducted in accordance with the guidelines and approval of the Animal Committee of the Laboratory Animal Center of Guangzhou University of Chinese Medicine (no. 20230427002). Male C57BL/6J mice, weighing 20 ± 5 g, obtained from the Laboratory Animal Center of Guangzhou University of Chinese Medicine (Guangzhou, China), were used in this study. The mice were randomly divided into three groups (n = 6 per group): Control (CON), T2DOP model (MOD), and D + Q. Mice in MOD and D + Q groups were adaptively fed for 1 week and then fed on a high-fat diet (HFD) for 4 weeks. All mice, except those in CON group, were intraperitoneally injected with STZ (35 mg/kg) for 3 days to induce T2DM. The D + Q group was then treated with D (50 mg/kg) + Q (50 mg/kg) via intragastric gavage for 7 weeks. After establishing the T2DOP model, body weight and fasting blood glucose (FBG) levels were measured weekly for each group. Mice were subjected to persistent hyperglycemia, and those with blood glucose levels ≥ 12.3 mg/dL were included in MOD and D + Q group.

The left femurs of the mice were fixed in 4% PFA at room temperature for 24 h and scanned using the Skyscan 1,172 model (Bruker Corporation, Belgium). The evaluation indices included bone volume over total volume (BV/TV, %), trabecular bone thickness (Tb.Th, mm), trabecular bone separation (Tb.Sp, mm), trabecular bone number (Tb.N, 1/mm), and structure model index (SMI). Before scanning, the parameters were set as follows: 80 kV voltage, 100 μA current, 0.4-degree rotation step, 0.5 mm aluminum filter, and a slice thickness of 5 μm. The index data were analyzed using the CTAn software (Skyscan, Bruker Corporation).

After fixation, the femurs were decalcified in 10% EDTA for 4 weeks and then embedded in paraffin blocks. The liver tissues were directly embedded in paraffin blocks. All samples were cut into 5 μm sagittal slices and stained using H&E staining solution. A Panoramic Midi Digital Slide Scanner (3D HISTECH Ltd, Hungary) was used for image acquisition. The images were analyzed using SlideViewer 2.6 software (3D HISTECH Ltd, Hungary) and ImageJ software.

The slides were incubated in TRAcP incubation solution at 37°C for 50 min. Osteoclasts were subsequently observed under a microscope. Osteoclasts were considered TRAcP-positive if they turned burgundy. The sections were then stained with Harris Hematoxylin for 3–8 min after rinsing with distilled water. The slides were then soaked in 1% hydrochloric acid and alcohol for a few seconds for differentiation, followed by rinsing with running water. The slices were then treated with 0.6% ammonia to restore their blue color and rinsed again with tap water. Finally, the slides were examined using a panoramic Midi digital biopsy scanner (3D HISTECH Ltd, Hungary), and the figures were analyzed using SlideViewer software (version 2.6; 3D HISTECH Ltd, Hungary) and the ImageJ software.

After antigen retrieval, the sections were incubated with primary antibodies against P21, LC3, P62, Beclin 1, and mTOR, followed by treatment with biotinylated goat anti-rabbit secondary antibodies. Subsequently, the sections were incubated with DAB for 10 min. Image J software was utilized to analyze the images of IHC staining.

Total genomic DNA was extracted from the samples and used as a template for PCR amplification of bacterial 16S rRNA genes. Amplification was performed using barcoded primers and Ex Taq (Takara). For bacterial diversity analysis, V3–V4 (or V4–V5) variable regions of the 16S rRNA genes were amplified using universal primers 343F (5′-TACGGRAGGCAGCAG-3′) and 798R (5′-AGGGTATCTAATCCT-3′) for V3–V4 regions or 515F (5′-GTGCCAGCMGCCGCGG-3′) and 907R (5′-CCGTCAATTCMTTTRAGTTT-3′) for V4–V5 regions. The quality of the amplicon was assessed using agarose gel electrophoresis. PCR products were purified using AMPure XP beads before undergoing another round of PCR. Sequencing was performed using an Illumina NovaSeq 6,000 with 250 bp paired-end reads (Illumina Inc., San Diego, CA).

Alpha and beta diversity analyses were conducted using QIIME 2 software. Alpha diversity, including the Ace, Chao 1, Shannon, and Simpson indices, was used to estimate the microbial diversity in the samples. To assess beta diversity, the unweighted pair-group method with arithmetic means (UPGMA) was applied using the unweighted UniFrac distance matrix. Significant differences between groups were analyzed using the analysis of variance (ANOVA). The linear discriminant analysis effect size (LEfSe) method was utilized to compare the taxonomic abundance spectra.

Serum metabolic profiles in electrospray ionization (ESI) positive and ESI negative ion modes were analyzed using an ESI source (Thermo Fisher Scientific, Waltham, MA, USA). An ACQUITY UPLC HSS T3 column (100 mm × 2.1 mm, 1.8 μm) was used for both positive and negative modes. The original LC-MS/MS data were processed using Progenesis QI V2.3 software (Non-linear Dynamics, Newcastle, UK).

The raw data matrix was imported into the R software for partial least-squares discriminant analysis (PLS-DA) to identify metabolites that differed between the groups. The variable importance in projection (VIP) values obtained from the PLS-DA model were used to rank the overall contribution of each variable to the discrimination between groups. One-way analysis of variance (ANOVA) test was further conducted to verify the significance of the differential metabolites between groups. The threshold values were set as VIP values > 1.0 and p-value < 0.05.

All experiments were conducted in triplicate. The data were analyzed using one-way ANOVA and expressed as mean ± standard deviation (SD). Changes in body weight and FBG over time were analyzed using repeated-measures (RM) ANOVA with Group as the between-subjects factor and time as the within-subjects factor, followed by Tukey's HSD where significant effects were detected. The assumptions of RM ANOVA, including normality of residuals (assessed via the Shapiro-Wilk test), sphericity (assessed via Mauchly's test, with Greenhouse-Geisser correction applied where violated), and homogeneity of variances between groups at each time point (assessed via Levene's test), were evaluated. The data met these assumptions sufficiently to proceed with RM ANOVA. Differences with p < 0.05 were considered statistically significant. Statistical analyses were performed using GraphPad Prism software (version 9.4.1).

T2DOP mice were successfully induced after 4 weeks of HFD followed by 3 days of STZ intraperitoneal injection (Figure 1A). During the 7-week detection period, T2DOP mice consistently gained weight (Figure 1B). However, D + Q treatment significantly reduced the rate of weight gain (Figure 1B). In T2DOP mice, FBG levels increased rapidly and remained at ~27.43 mmol/L (Figure 1C). After D + Q treatment, FBG levels decreased to ~21.29 mmol/L and reached a plateau (Figure 1C).

Three-dimensional reconstruction of the trabecular bone revealed a significant decrease in trabecular bone mass in mice with T2DOP (Figure 2A). After 6 weeks of D + Q treatment, the bone mass was restored (Figure 2A). BV/TV data indicated that D + Q protected the bone volume from the detrimental effects of hyperglycemia (Figure 2B). Moreover, D + Q treatment significantly increased the trabecular bone thickness and decreased trabecular bone number while decreasing the trabecular bone separation (Figures 2C–F). H&E staining revealed an increased number of lipid droplets and a decrease in bone mass in the MOD group, whereas D + Q treatment significantly alleviated bone loss (Figure 2G). In H&E-stained sections of the liver, more lipid droplets were observed in the MOD group, while D + Q treatment significantly reduced the number of lipid droplets (Supplementary Figure 1). No evidence of drug-induced hepatotoxicity was observed across D + Q group. Hepatic architecture remained intact, with no signs of necrosis, inflammation, or degenerative changes (Supplementary Figure 1). The number of TRAcP-positive cells was significantly higher in the MOD group, while D + Q treatment effectively reduced the number of osteoclasts (Figure 2H). To detect the senescence phenotype, the P21 expression level in bone tissue was assessed. The results revealed that D + Q restores p21 expression increased by T2DOP (Figure 2I).

Alpha diversity was analyzed using Ace, Chao 1, Shannon, and Simpson methods. No changes were observed after treatment with D + Q (Supplementary Figures 2A–D). The UPGMA was used to determine beta diversity. The results demonstrated that the gut microbes from the D + Q group were clustered together (Supplementary Figure 2E). The ternary results at the phylum level revealed that Bacteroidota, Actinobacteriota, and Desulfobacterota exhibited higher relative abundances (Figure 3A). At the genus level, the top 15 abundant species among the three groups were selected based on the results of species annotations, and a columnar cumulative chart of species relative abundance was created to visualize the horizontal community structures across groups with higher relative abundance and their proportions. Bacteroides and Lachnospiraceae_NK4A136_group were decreased in the MOD group and restored in the D + Q group. Alistipes, Clostridia_cadinBB60_group, Mucispirillum, Helicobacter, and Butyrivibrio were increased in the MOD group and improved in the D + Q group (Figure 3B). The Random-Forest test revealed that Mucispirillum, Lactobacillus, Odoribacter, Helicobacter, and Ruminococcus were the top five key microbiomes affected by D + Q (Figure 3C).

PLS-DA is a common metabolomic classification method that addresses the limitation of Principal Component Analysis. The PLS-DA score was used to separate the metabolites of CON, MOD, and D + Q groups, indicating processing differences (Figure 4A). The top 12 of D + Q-induced down-regulated metabolites included PE[18:2(9Z, 12Z)/21:0], PE[17:1(9Z)/20:2(11Z, 14Z)], PC [P-18:1(11Z)/22:6(5Z, 7Z, 10Z, 13Z, 16Z, 19Z)-OH(4)], PS[20:1(11Z)/17:1(9Z)], D-Glucose, D-Psicose, Allodesmosine, 2-C-Methyl-D-Erythritol-4-Phosphate, Hypotaurine, 2,2-Dichloro-12-(4-Chlorophenyl)Dodecanoic Acid, 8,11-Eicosadiynoic Acid, and 6-(1,2,3,4-Tetrahydroxybutyl; Figure 4B, Supplementary Figures 3A–L). According to the heatmap of the correlation between serum metabolites, the top four most correlated metabolites were 2-C-Methyl-D-Erythritol-4-Phosphate and Hypotaurine, PS(P-20:0/16:0) and PC[16:0/18:2(11Z,13Z)], D-glucose and 2-C-Methyl-D-Erythritol-4-Phosphate, and D-glucose and Hypotaurine (Figure 4C). KEGG enrichment analysis was used to predict the differential metabolites-related pathways (Figure 4D). The results revealed that D + Q treatment influenced autophagy, glycerophospholipid metabolism, linoleic acid metabolism, and alpha-linoleic acid metabolism. Based on the Reactome terms, differential metabolites modulated by D + Q were primarily associated with LPC hydrolysis, glycerophospholipid catabolism, glycerophospholipid biosynthesis, and lipid metabolism (Figure 4E). To determine whether D + Q-induced metabolite changes were associated with the effects on gut microbiota, we performed Spearman's correlation analyses (Figures 4F–G). The results revealed that Rodentibacter was changed in both MOD and D + Q groups. Additionally, Rodentibacter was positively associated with serum PE[17:1(9Z)/20:2(11Z, 14Z)], and PS[20:1(11Z)/17:1(9Z)] levels.

We further performed IHC to determine the effects of D + Q. The cortical bone of the femur demonstrated that D + Q restored the expression levels of LC3, P62, Beclin 1, and mTOR, which were used to quantify autophagy level (Figures 5A–H).