IPA (cat. no. SJ-MX4391) was purchased from Shandong Sparkjade Scientific Instruments Co., Ltd. H₂O₂ was purchased from MilliporeSigma. Anti-GAPDH antibody was obtained from Cell Signalling Technology, Inc. Antibodies against COL1A1 (cat. no. HA722517), RUNX2 (cat. no. ET1612-47), SP7 (cat. no. HA722817) and OPN (cat. no. HA723082) were obtained from HUABIO. Anti-PPARγ antibody (cat. no. 14N93N79) was purchased from Epizyme, Inc.

Mouse BMSCs (mBMSCs) were isolated as previously described (37,38). Briefly, mBMSCs were obtained from the femora and tibiae of C57BL/6 mice by flushing the bone marrow cavity. The C57BL/6 mice were sourced from same mice used in the animal experiments. The isolated cells were cultured in α-MEM (Procell Life Science & Technology Co., Ltd.) supplemented with 10% fetal bovine serum (FBS; cat. no. SA101.02; Cellmax; https://www.cellmaxcell.com/) and 1% penicillin/streptomycin (MedChemExpress). To establish an oxidative stress microenvironment in vitro, cells were treated with H₂O₂ (100 µM) for 2 h (7). For osteogenic differentiation, mBMSCs were cultured in osteogenic induction medium consisting of low-glucose Dulbecco's Modified Eagle Medium (Procell Life Science & Technology Co., Ltd.) supplemented with 10% FBS, β-glycerophosphate (10 mM), ascorbic acid (200 µM) and dexamethasone (100 nM) (MedChemExpress).

mBMSCs were seeded in 24-well plates and cultured in osteogenic induction medium. At the indicated time points, ALP staining was performed using a BCIP/NBT Alkaline Phosphatase Chromogenic kit (cat. no. C3206; Beyotime Institute of Biotechnology) according to the manufacturer's instructions. ARS staining was performed using an Alizarin Red kit (cat. no. ALIR-10001; Cyagen Biosciences, Inc.) following the manufacturer's protocol.

mBMSCs were seeded in 24-well plates and cultured in adipogenic induction medium ((Procell Life Science & Technology Co., Ltd.) according to the manufacturer's instructions. Cells were subsequently stained with Oil Red O staining solution (Procell Life Science & Technology Co., Ltd.).

To evaluate the effect of IPA on cell proliferation, 2,000 cells were seeded into 96-well plates and treated with different concentrations of IPA (0, 10, 25, 50, 75 and 100 µM) for the indicated durations. Subsequently, 10 µl of CCK-8 solution (Beijing Boxbio Science & Technology Co., Ltd.) was added to each well and incubated at 37°C for 2 h. Absorbance at 450 nm was measured using a microplate reader (BioTek; Agilent Technologies, Inc.).

MMP was assessed using a JC-1 assay kit (MedChemExpress). Briefly, cells were pretreated with H₂O₂ (100 µM) and subsequently exposed to different concentrations of IPA for 12, 24, or 48 h. After treatment, cells were incubated with JC-1 working solution for 20 min. Fluorescence intensities were measured at 590 nm (aggregates, red) and 520 nm (monomers, green). The ratio of red-to-green fluorescence was calculated to evaluate changes in MMP.

Cells were harvested and lysed in RIPA buffer containing protease and phosphatase inhibitors (Beijing LABLEAD Inc.). Lysates were centrifuged at 12,000 × g for 15 min at 4°C, and the supernatants were collected and mixed with loading buffer. Total protein concentration was determined using a BCA protein assay kit (cat. no. P0012; Beyotime Institute of Biotechnology). Equal amounts of protein (20 µg per lane) were separated by 10% sodium dodecyl sulphate polyacrylamide gel electrophoresis and transferred onto polyvinylidene fluoride membranes (MilliporeSigma). Membranes were blocked at room temperature with 5% non-fat milk for 1 h and incubated with primary antibodies (1:1,000) overnight at 4°C. The membranes were then incubated with HRP-conjugated secondary antibodies (1:5,000) for 1 h. Protein bands were visualised using ECL Luminescent Solution (cat. no. KGC4601; Nanjing KeyGen Biotech Co., Ltd.) using a chemiluminescent detection system (Thermo Fisher Scientific, Inc.), and band intensities were analysed using the 'Gel Analysis' function in ImageJ software (V1.53; National Institutes of Health) as previously described (39). GAPDH was used as the internal loading control.

Total RNA from BMSCs was extracted using TRIzol reagent (Shandong Sparkjade Scientific Instruments Co., Ltd.), and cDNA was synthesised using the SPARKscript II RT Plus Kit (cat. no. AG0304; Shandong Sparkjade Scientific Instruments Co., Ltd.). RT-qPCR was performed using SYBR Green Master Mix (cat. no. AH0104; Shandong Sparkjade Scientific Instruments Co., Ltd.). Thermal cycling conditions were set according to the manufacturer's instructions. The mRNA levels of the target genes were normalized to those of the housekeeping gene GAPDH. The relative gene expression levels were calculated using the 2^−ΔΔCq method. Primer sequences used in the present study are listed in Table I.

For IF staining, mBMSCs were fixed with 4% paraformaldehyde and permeabilised with 0.1% Triton X-100 (Beyotime Institute of Biotechnology). Cells were then blocked with 5% bovine serum albumin (BSA; MilliporeSigma) and incubated at 4°C with primary antibodies [PPARγ, COL1A1, RUNX2 (cat. no. K003506P) Beijing Solarbio Science & Technology Co., Ltd.] overnight. The following day, cells were incubated at room temperature with fluorophore-conjugated secondary antibodies. Nuclei were counterstained with DAPI with 5 µg/ml (Beyotime Institute of Biotechnology). Fluorescence signals were detected using a confocal microscope (Leica Microsystems GmbH) at ×400 magnification. Fluorescence intensity was analysed using ImageJ software.

The molecular docking analysis of IPA and PPARγ was performed using the CB-Dock2 platform (https://cadd.labshare.cn/cb-dock2/index.php), as previously described (40). The structure of IPA (PubChem compound ID: 3744) was obtained from PubChem (https://pubchem.ncbi.nlm.nih.gov/compound/3744), whereas the crystal structure of PPARγ (8BF1) was downloaded from the Protein Data Bank (https://www.rcsb.org/structure/8BF1). The binding sites and binding affinity between IPA and PPARγ were subsequently predicted.

Total RNA from BMSCs was extracted using TRIzol reagent according to the manufacturer's instructions. RNA quality and integrity were assessed using 5300 Bioanalyser (Agilent Technologies, Inc.), and samples with RNA integrity number OD260/280=1.8~2.2 were used for library construction. Sequencing libraries were generated using the NEBNext Ultra II RNA Library Prep Kit (cat. no. E7770; New England Biolabs) according to the manufacturer's instructions. The final library concentration was determined using a Qubit fluorometer (Thermo Fisher Scientific, Inc.) and quantitative PCR, and libraries were loaded at a concentration of 4 nM for sequencing on the Illumina NovaSeq platform, yielding ~50 million paired-end reads per sample. Raw data processing, including adaptor trimming and quality filtering, was performed using FastQC (v. 0.11.9; https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). High-quality reads were aligned to the mouse reference genome using HISAT2 (v. 2.2.1; https://daehwankimlab.github.io/hisat2/). Transcript abundance was quantified and expressed as transcripts per million. Gene Ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed using the Majorbio online analysis platform (https://www.majorbio.com/).

Microbial DNA was extracted from faecal samples collected from sham and ovariectomy (OVX) mice. The 338F/806R primers (Table I) were used to amplify the V3-V4 region of the bacterial 16S rRNA gene, and PCR products were purified using a PCR Clean-Up Kit (YuHua Biotechnology). Sequencing libraries were prepared and sequenced according to the manufacturer's instructions to generate paired-end 300-bp reads. All sequencing and bioinformatics analyses were performed by Majorbio Bio-Pharm Technology.

Serum samples were collected from Sham and OVX mice. Serum IPA concentrations were measured using gas chromatography-mass spectrometry according to the manufacturer's instructions.

All animal procedures were approved by Peking University Third Hospital (approval no. BCAA0292; Beijing, China) and conducted in accordance with ARRIVE guidelines. 8-week-old female C57BL/6 mice were acclimated for 1 week under standard laboratory conditions. The initial body weight of the mice was ~18-22 g at the start of the experiment. Animals were housed at 22±2°C with 50±10% humidity under a 12 h light/dark cycle. Food and water were available ad libitum throughout the study. A total of 24 mice were randomly assigned to 4 groups: Sham, OVX, OVX + IPA (10 mg/kg) and OVX + IPA (20 mg/kg). A postmenopausal OP model was established by bilateral OVX as previously described (41). Briefly, the mice were anaesthetised via inhalation of 3% isoflurane, with the anaesthesia maintained via 2% isoflurane inhalation during surgery, and a small dorsal midline incision was made to expose both ovaries, which were subsequently ligated and removed. Sham-operated mice underwent the same surgical procedure without ovary removal. Following surgery, mice received daily oral administration of IPA or an equal volume of normal saline for 10 weeks. At the end of the experiment, animals were euthanized by overdose of pentobarbital sodium administered via intraperitoneal injection (150 mg/kg). Death was confirmed by the absence of spontaneous respiration, heartbeat and pedal withdrawal reflex. The femurs and serum samples were collected for further analyses. A total of ~600 µl of whole blood was collected via orbital blood sampling following eyeball removal, and serum was subsequently isolated by centrifugation at 3,000 × g for 15 min at 4°C.

Femurs were scanned using a SkyScan micro-CT system (Bruker Corporation). The region of interest was defined as a 1-mm-high area distal to the growth plate. Three-dimensional reconstruction images were generated using CTvox software. Trabecular bone parameters, including bone volume fraction (BV/TV), trabecular number (Tb.N), trabecular thickness (Tb.Th) and trabecular separation (Tb.Sp), were analysed using CTAn software (V2.0; https://www.blue-scientific.com/bruker-micro-ct-software/). For haematoxylin and eosin (H&E) staining, decalcified bone tissues were processed and stained according to the manufacturer's protocols (Wuhan Servicebio Technology Co., Ltd.).

Undecalcified bone samples were subjected to Von Kossa staining according to the manufacturer's instructions (Wuhan Servicebio Technology Co., Ltd.). Images were captured using a light microscope, and mineralised areas were quantified using ImageJ software.

Decalcified bone tissues were embedded in paraffin. Tissue sections (5 µm) were deparaffinised in xylene and rehydrated through a graded ethanol series (100, 95, 85, and 75% ethanol) before antigen retrieval by trypsin digestion. Sections were then blocked with 3% hydrogen peroxide and incubated with 5% BSA for 1 h. Subsequently, the sections were incubated at 4°C with primary antibodies overnight. The following day, HRP-conjugated secondary antibodies were applied at room temperature for 1 h. Signals were developed using a DAB substrate. Images were acquired under a bright-field microscope, and staining intensity was quantified using ImageJ software.

Serum levels of P1NP and CTX-1 were measured using ELISA kits (cat. nos. E-EL-M3023 and E-EL-M0233; Elabscience Biotechnology, Inc.) according to according to the manufacturer. Absorbance at 450 nm was measured using a microplate reader (BioTek; Agilent Technologies, Inc.). All samples were analysed in triplicate.

Data were presented as the mean ± SD. All experiments were independently repeated at least three times. Comparisons between two groups were performed using unpaired Student's t-test, whereas comparisons among multiple groups were analysed using one-way or two-way analysis of variance followed by Tukey's post hoc test. P<0.05 was considered to indicate a statistically significant difference. Statistical analyses and data visualisation were performed using GraphPad Prism (v. 8.0; Dotmatics).

To investigate whether gut microbiota-derived metabolites contribute to OP, an OVX-induced mouse model was first established. The OVX model effectively mimicked postmenopausal OP (42). Micro-CT analysis confirmed a marked reduction in trabecular bone mass in OVX mice, validating the successful establishment of osteoporotic phenotypes (Fig. 1A and B). 16S rRNA sequencing was next performed to profile gut microbiota composition. Although α-diversity did not differ significantly between groups (Fig. 1C), β-diversity analysis revealed a clear separation between sham and OVX mice, indicating substantial alterations in microbial community structure (Fig. 1D). Consistently, the gut microbial health index was significantly reduced in the OVX group (Fig. S1A), whereas the microbial dysbiosis index was significantly increased (Fig. S1B).

Taxonomic profiling identified distinct alterations in bacterial genera between the two groups. OVX mice exhibited increased abundance of Muribaculaceae and reduced abundance of the Lactobacillus and Clostridia genera compared with sham controls (Fig. 1E). The abundance of Clostridia was significantly decreased in the OVX group, with statistical significance confirmed by Wilcoxon analysis (Fig. 1F and G). Given that Clostridia is a major bacterial genus involved in IPA production (26), serum IPA concentrations were next measured. Metabolite analysis demonstrated a significant reduction in circulating IPA levels in OVX mice (Fig. 1H). Importantly, correlation analyses revealed that serum IPA levels were positively associated with both BV/TV and Clostridium abundance (Fig. 1I and J). These findings suggested that depletion of microbiota-derived IPA is a characteristic feature of oestrogen deficiency-induced OP and may contribute functionally to bone loss.

It was next investigated whether IPA modulates BMSC survival under oxidative stress conditions. CCK-8 assays demonstrated that IPA alone (0-100 µM) neither promoted nor inhibited mBMSC proliferation (Fig. 2A). Exposure to H₂O₂ significantly reduced cell viability, whereas IPA treatment dose-dependently rescued this effect without altering basal proliferation (Fig. 2B). Live/dead staining further confirmed that IPA markedly attenuated oxidative stress-induced cell death (Fig. 2C and D).

Because loss of MMP is a hallmark of early apoptosis (43), JC-1 staining was performed to evaluate mitochondrial function. H₂O₂ treatment induced a significant reduction in MMP, whereas IPA administration effectively restored mitochondrial integrity (Fig. 2E and F). Collectively, these results demonstrated that IPA exerts cytoprotective effects on BMSCs by preserving mitochondrial function and attenuating oxidative stress-induced apoptosis.

It was next examined whether IPA directly regulates BMSC lineage commitment. Under physiological conditions, IPA did not significantly affect mineralised nodule formation (Fig. S2A and B) or ALP activity (Fig. S2C and D). Consistently, western blot analysis showed that IPA had no appreciable effect on the expression of osteogenic markers, including RUNX2, SP7, and OPN (Fig. S2E and F). These findings indicate that IPA does not function as a basal osteo-inductive factor.

By contrast, under oxidative stress conditions, H₂O₂ markedly impaired osteogenic differentiation, as evidenced by reduced ALP activity and decreased mineralised nodule formation (Fig. 3A and B). IPA treatment dose-dependently restored osteogenic capacity (Fig. 3D and E). In parallel, oxidative stress significantly promoted adipogenic differentiation (42), as reflected by increased lipid accumulation and upregulation of adipogenic markers (Fig. 3C and F). Further analysis of lineage-specific gene expression demonstrated that IPA prevented the H₂O₂-induced downregulation of osteogenic genes, including Col1a1, Runx2, Sp7 and Ocn (Fig. S3A), while reversing the upregulation of adipogenic genes, including Adiponectin, Cebpα, Fabp2 and Srebp-1 (Fig. S3B). Western blot analysis further confirmed that IPA rescued the H₂O₂-induced downregulation of osteogenic proteins (Fig. 3G and H). Similar findings were observed by IF staining (Fig. 3I; Fig. S4A and B), further supporting the protective effect of IPA on osteogenic differentiation under oxidative stress conditions. Together, these findings indicated that IPA functions as a context-dependent regulator that restores osteogenicadipogenic balance under pathological conditions rather than altering baseline differentiation.

To determine whether these protective effects also occur in vivo, an OVX mouse model was established and IPA was administered orally for 10 weeks. A schematic overview of the animal experimental design is presented in Fig. 4A. OVX mice exhibited significantly increased body weight and reduced uterine weight compared with the control group (Fig. S5A and B).

Micro-CT analysis revealed that IPA markedly alleviated OVX-induced bone loss (Fig. 4B). IPA supplementation significantly improved trabecular bone architecture, as indicated by increased BV/TV, Tb.N and Tb.Th, along with reduced Tb.Sp (Fig. 4E-H). Consistently, biochemical analyses demonstrated elevated levels of the bone formation marker PINP and decreased levels of the bone resorption marker CTX-1 following IPA treatment, indicating a shift toward anabolic bone remodelling (Fig. 4C and D). These results demonstrated that IPA supplementation effectively mitigates oestrogen deficiency-induced bone loss and promotes bone formation in vivo.

To investigate further the cellular basis of these effects, BMSCs isolated from experimental mice were analysed. Cells derived from OVX mice exhibited impaired mineralisation capacity, as reflected by significantly reduced ALP activity and ARS staining ability (Fig. 5A and B), along with increased intracellular lipid accumulation, confirming disruption of lineage balance (Fig. 5H). Importantly, IPA treatment restored osteogenic capacity (Fig. 5C and D) while suppressing adipogenesis (Fig. 5I) in BMSCs isolated from OVX mice. IPA administration significantly upregulated the expression of osteogenic genes and proteins (Fig. 5E-G), while suppressing adipogenic gene expression in OVX mice (Fig. 5J). These findings demonstrated that IPA alleviates OP by re-establishing bone marrow lineage equilibrium.

To elucidate the underlying mechanisms, transcriptomic sequencing was performed on mBMSCs treated with H₂O₂ in the presence or absence of IPA. RNA-seq analysis identified a total of 1,254 differentially expressed genes (DEGs; fold change ≥3.0), including 509 upregulated and 745 downregulated genes (Figs. 6A and S6A). GO functional annotation revealed that IPA-regulated genes were enriched in pathways associated with oxidative stress defence, cell proliferation and immune regulation (Fig. S6B). GO enrichment analysis further highlighted biological processes related to osteogenic differentiation, cell cycle regulation and oxidative phosphorylation (Fig. 6B). KEGG pathway analysis demonstrated significant enrichment in PPAR signalling, cholesterol biosynthesis, adipogenesis-related pathways, oxidative phosphorylation, and p53-mediated apoptosis (Fig. 6C). Gene Set Enrichment Analysis further showed that IPA markedly suppressed several adipogenesis-related pathways, including unsaturated fatty acid synthesis, cholesterol biosynthesis, and the PPARγ signalling pathway, while also inhibiting oxidative phosphorylation (Fig. 6D). These findings suggested that IPA preserves osteogenic function under oxidative stress, at least in part, by inhibiting adipogenic differentiation.

The three-dimensional structure of IPA is shown in Fig. 6E. Molecular docking analysis revealed a high predicted binding affinity (-6.6 kcal/mol) between IPA and PPARγ (Fig. 6F and G). Docking results demonstrated that IPA formed multiple hydrogen bonds and ionic interactions with key residues, including Arg280, Ile262 and Ile281, indicating a plausible structural basis for direct interaction (Fig. 6H). Consistently, H₂O₂-induced upregulation of PPARγ expression was significantly attenuated by IPA at both the mRNA and protein levels (Fig. 6I-K). IF analysis further confirmed reduced nuclear localisation of PPARγ following IPA treatment (Fig. 6L and N). Consistent with the in vitro findings, BMSCs isolated from OVX mice displayed elevated PPARγ expression, which was significantly reduced following IPA supplementation in vivo (Fig. 6M and O). These findings indicated that IPA regulates BMSC lineage commitment by inhibiting PPARγ signalling.

To validate this mechanism in vivo, histological and IHC analyses were performed. H&E staining revealed that IPA supplementation significantly improved BV/TV compared with the OVX group (Fig. 7A). Consistent with the H&E staining results, Von Kossa staining demonstrated reduced mineral deposition in OVX mice, which was effectively restored following IPA administration (Fig. 7B). IHC analysis further confirmed these observations. BMSCs derived from OVX-induced osteoporotic mice exhibited lower expression of osteogenic proteins, including COL1A1 and RUNX2, whereas IPA treatment restored their expression levels (Fig. 7C and D). By contrast, PPARγ expression was significantly downregulated by IPA supplementation compared with the OVX group (Fig. 7E). Furthermore, no obvious pathological injury was observed in the heart, liver, spleen, lung, or kidney following oral IPA supplementation (Fig. S7). Collectively, these data demonstrated that IPA alleviates OP by suppressing PPARγ signalling, thereby restoring the osteogenic-adipogenic balance and promoting bone formation.