Piezo1 activation suppresses bone marrow adipogenesis to prevent osteoporosis by inhibiting a mechanoinflammatory autocrine loop

Our initial aim was to explore the potential role of Piezo1 in the development and function of adipose tissues. To this end, we intended to use the Platelet-Derived Growth Factor Receptor Alpha (PDGFRα)-Cre mice to genetically invalidate Piezo1 in adipocyte progenitor cells.²⁵ In addition to adipocyte progenitor cells, previous studies have shown that PDGFRα is also expressed in BMMSCs, brain oligodendrocyte progenitor cells (OPCs), and muscle fibro-adipogenic progenitors (FAPs).^26–28 Therefore, we first generated PDGFRα-Cre tdTomato reporter mice to demonstrate the tissue distribution pattern of tdTomato driven by the PDGFRα promoter by crossing PDGFRα-Cre mice with tdTomato fluorescent reporter (Ai14) mice (supplementary Fig. 1a). Examination of the tdTomato fluorescence intensity across different tissues of the reporter mice revealed its most abundant enrichment in the bone marrow, followed by subcutaneous white adipose tissue (scWAT), OPC-enriched brain regions (subventricular zone [SVZ], white matter, and corpus callosum), and skeletal muscles (gastrocnemius, quadriceps, and soleus) (supplementary Fig. 1b). Consistently, fluorescence microscopy revealed that tdTomato-expressing (PDGFRα⁺) cells were most abundant in BMMSCs, moderately present in stromal vascular fractions (SVFs) and brain areas containing OPCs, scattered in skeletal muscle, but completely absent in non-OPC-containing brain area such as the hypothalamus (supplementary Fig. 1c, d).

Next, we generated mice with specific depletion of Piezo1 in PDGFRα⁺ cells (named PDGFRα-Piezo1 KO mice) by crossing Piezo1^flox/flox mice (WT) with PDGFRα-Cre mice. As expected, both qPCR and immunofluorescence staining confirmed the ablation of Piezo1 in SVFs and BMMSCs of PDGFRα-Piezo1 KO mice, with no compensatory expression of Piezo2 in Piezo1 KO SVFs and BMMSCs (supplementary Fig. 2a, b, d). On the other hand, there was no difference in Piezo1 expression in the brain or skeletal muscle between WT and KO mice (supplementary Fig. 2c, d), which could be explained by the undetectable Piezo1 expression in OPC-enriched brain areas and the absence of colocalization between Piezo1-expressing cells and tdTomato-expressing PDGFRα⁺ cells in skeletal muscle (supplementary Fig. 1d). Taken together, these observations indicate that the PDGFRα-Cre driver enables the selective ablation of Piezo1 in BMMSCs and SVFs.

Notably, microcomputed tomography (micro-CT) analysis revealed a significant reduction in bone volume and bone density in PDGFRα-Piezo1 KO mice compared with WT and PDGFRα-Cre control mice, as evidenced by markedly lower trabecular bone volume and number, increased trabecular separation, and reduced thickness of cortical bone (Fig. 1b–e). However, the lengths of major long bones (femurs and tibias) were similar between WT and KO mice (supplementary Fig. 4f), suggesting that skeletal development is unaltered in PDGFRα-Piezo1 KO mice. Notably, the reduction in bone volume in PDGFRα-Piezo1 KO mice was accompanied by a significantly lower circulating level of procollagen type I N-propeptide (PINP, a bone formation marker²⁹), whereas the serum level of type I collagen cross-linked C-telopeptide (CTX-1, a bone resorption marker³⁰) remained unchanged (Fig. 1f, g). Histological analysis revealed that PDGFRα-Piezo1 KO mice had significantly fewer alkaline phosphatase (ALP)-stained osteoblasts, but a similar number of tartrate-resistant acid phosphatase (TRAP)-stained osteoclasts compared with their WT littermates and PDGFRα-Cre controls (supplementary Fig. 5a–d). Dynamic histomorphometry via in vivo double labeling with calcein and alizarin red³¹ showed significantly reduced bone formation in PDGFRα-Piezo1 KO mice, as evidenced by decreased bone formation rate (BFR) and mineral apposition rate (MAR) (Fig. 1h, i). Conversely, PDGFRα-Piezo1 KO mice exhibited a significantly increased volume of bone marrow adipocytes (BMAs) compared to their controls, as determined by osmium tetroxide staining of decalcified tibias followed by micro-CT analysis³² (Fig. 1j, k) and histological analysis via hematoxylin and eosin (H&E) staining (supplementary Fig. 5e–g). Similarly, female PDGFRα-Piezo1 KO mice also displayed reduced bone volume and increased bone marrow adiposity compared to their WT littermates (supplementary Fig. 6a–k). Since PDGFRα-Cre control mice and Piezo1^flox/flox WT mice displayed similar phenotypes, only Piezo1^flox/flox WT mice (littermates of PDGFRα-Piezo1 KO mice) were included in all subsequent experiments. Collectively, these data suggest that ablation of Piezo1 in PDGFRα⁺ cells impairs osteogenesis but increases bone marrow adiposity without apparent changes in peripheral adipose tissues.

Given the abundant expression of PDGFRα in BMMSCs (supplementary Fig. 1c),^26,33 the decreased bone volume and increased marrow adiposity in PDGFRα-Piezo1 KO mice are likely attributed to Piezo1 deficiency in BMMSCs. To explore this possibility, we further used the cell-attached patch clamp configuration to measure Piezo1 currents in response to fast negative pressure pulses to confirm successful invalidation of Piezo1 channel activity in BMMSCs isolated from PDGFRα-Piezo1 KO mice. For electrophysiological recording, we used PDGFRα-Cre tdTomato reporter mice to identify tdTomato-positive BMMSCs in culture. The pressure-induced inward currents (holding potential of −80 mV) were characterized by an absence of inactivation and a remarkably slow deactivation when pressure stimulation was resumed. In contrast, the pressure-induced current was dramatically reduced in Piezo1 KO BMMSCs (supplementary Fig. 7a), indicating that the mechanogated ion channel recorded in mouse BMMSCs is due to Piezo1.

Considering the significant role of Piezo1 in mouse BMMSCs, we further investigated whether Piezo1 plays a similar role in human BMMSCs (hBMMSCs) by treating hBMMSCs with the Piezo1 activator Yoda1 (5 μM) or the Piezo1 inhibitor GsMTx4 (5 μM). We found that Piezo1 activation by Yoda1 significantly suppressed adipogenesis but promoted osteogenesis. Conversely, inhibition of Piezo1 via GsMTx4 increased adipogenesis but inhibited osteogenesis, as evidenced by qPCR analysis of adipogenic and osteogenic genes, together with Oil Red O staining of BMAs and Alizarin Red S staining of osteoblasts (supplementary Fig.). 9a–d

To identify secretory factors involved in the downstream effects of Piezo1 in modulating osteogenesis or adipogenesis of BMMSCs, we performed Olink-based proteomics for the quantitative analysis of 92 soluble factors in the conditioned media of Piezo1 KO and WT BMMSCs. Additionally, we also used multiplex assay and enzyme-linked immunosorbent assay (ELISA) to measure a panel of cytokines, chemokines, and growth factors, including C-C motif chemokine ligand 7 (Ccl7), C-X-C motif chemokine ligand 12 (Cxcl12), insulin-like growth factor 1 (Igf-1), fatty acid binding protein 4 (Fabp4) and Lcn2, which have been implicated in adipogenesis and/or bone homeostasis^42–46 but are not included in the Olink Mouse Exploratory panel. Among the 97 soluble factors analyzed, we identified 17 differentially secreted proteins (15 upregulated and 2 downregulated) between Piezo1 KO and WT BMMSCs (Fig. 5f, supplementary Fig. 12a–f, and supplementary Table 1). The top five significantly altered factors with a fold change >2 were Lcn2, C-C motif chemokine ligand 2 (Ccl2), peroxiredoxin 5 (Prdx5), follistatin (Fst), and matrilin 2 (Matn2) (Fig. 5g, h). Consistent with the changes in protein levels in the conditioned medium, the mRNA abundance of these five genes was also significantly increased in Piezo1 KO BMMSCs, with both Lcn2 and Ccl2 showing greater increases in KO cells (Fig. 5i). Similarly, a modest but significant elevation in the serum Lcn2 and Ccl2 levels was also observed in Piezo1 KO mice (supplementary Fig. 12g, h).

To address how Ccl2-CCR2 signaling modulates Lcn2 levels, we first used in silico analysis to predict the transcription factors that regulate Lcn2 gene expression by taking advantage of four different databases, including hTFtarget, GTRD, ChIP_Atlas, and TRRUST (supplementary Fig. 15a). Nuclear factor kappa B (NF-κB) was identified as the only target in the intersection among the four databases, suggesting its possible involvement in regulating Lcn2 expression. Next, we confirmed the NF-κB binding site within the Lcn2 promoter region between −261 and −252 (supplementary Fig. 15b). DNA sequence alignment analysis demonstrated that the NF-κB binding motif was highly conserved among several mammalian species examined (supplementary Fig. 15c). Chromatin immunoprecipitation (ChIP) was subsequently performed, and the binding of p65 to the Lcn2 promoter region was validated (Fig. 7h, i). We also conducted a promoter activity assay by co-transfecting BMMSCs with pGL3-based luciferase reporters containing either the WT Lcn2 promoter (Lcn2-WT) or the mutated Lcn2 promoter with mutations in the NF-κB binding motif (Lcn2-Mut; supplementary Fig. 15c), together with a Renilla luciferase plasmid for normalization. This analysis revealed that Lcn2 promoter activity in KO BMMSCs was robustly enhanced compared with that in WT BMMSCs, whereas it was significantly attenuated when the NF-κB binding motif was mutated (Fig. 7j).

Given that activation of the Ccl2-CCR2 axis has been shown to stimulate the NF-κB signaling pathway,^49,50 we therefore investigated whether activation or inhibition of Ccl2-CCR2 signaling controls Lcn2 expression by modulating NF-κB activity (Fig. 7k). We found that WT BMMSCs treated with rmCcl2 could markedly enhance the binding of p65 to the Lcn2 promoter region, while KO BMMSCs treated with the CCR2 antagonist INCB3344 significantly reversed the increase in p65 binding to the Lcn2 promoter (Fig. 7h, i). ELISA analysis for cytosolic and nuclear p65 and immunofluorescence staining of p65 collectively revealed that p65 translocation from the cytosol to the nucleus was significantly increased in KO BMMSCs, indicating enhanced NF-κB activity in KO BMMSCs (Fig. 7l–n). Notably, treatment with rmCcl2 potently induced the nuclear translocation of p65 in WT BMMSCs, whereas the pharmacological inhibition of CCR2 reversed the augmented nuclear translocation of p65 in KO BMMSCs (Fig. 7l-n). Additionally, treatment with the NF-κB inhibitor QNZ markedly suppressed the increase in Lcn2 gene expression and secretion in Piezo1 KO BMMSCs and WT BMMSCs treated with rmCcl2 (Fig. 7o, p).

To decipher how Piezo1 regulates Ccl2 expression, we also used the aforementioned in silico analysis to predict the potential transcription factors implicated in regulating Ccl2 gene expression (supplementary Fig. 16a) and then validated these predictions by ChIP assay. Among the five predicted transcription factors, namely, jun proto-oncogene (c-Jun), NF-κB, Spi-1 proto-oncogene (SPI1), signal transducer and activator of transcription 1 (STAT1), and signal transducer and activator of transcription 3 (STAT3), only the binding of c-Jun with the Ccl2 promoter region was significantly enhanced in Piezo1 KO BMMSCs (supplementary Fig. 16b, c and Fig. 7r, s). DNA sequence alignment analysis revealed a highly conserved c-Jun binding motif within the Ccl2 promoter region among different mammalian species (supplementary Fig. 16d). Given that c-Jun is a key component of the transcription factor activator protein-1 (AP-1) complex,⁵¹ we used the AP-1 inhibitor T-5224 to further confirm that Piezo1 regulates Ccl2 expression through c-Jun⁵² (Fig. 7q). This analysis revealed that the increase in c-Jun binding to the Ccl2 promoter region, as well as the augmented Ccl2 expression and secretion in Piezo1 KO BMMSCs, were largely reversed by preincubation with the AP-1 inhibitor (Fig. 7r–u). Additionally, a promoter activity assay using luciferase reporters containing the WT Ccl2 promoter (Ccl2-WT) or mutated Ccl2 promoter with mutations in the c-Jun binding motif (Ccl2-Mut; supplementary Fig. 16d) revealed significantly enhanced Ccl2 promoter activity in KO BMMSCs than in WT BMMSCs, whereas this augmented Ccl2 promoter activity in KO BMMSCs was markedly abrogated by mutations in the c-Jun binding motif (Fig. 7v). These data collectively suggest that Piezo1 suppresses Ccl2 gene expression via c-Jun.

Interestingly, in epithelial and endothelial cells, Piezo1 activation was previously reported to induce the transcription factor Kruppel-like factor 2 (Klf2).^53,54 Moreover, Klf2 was shown to inhibit the transcriptional activity of both NF-κB and AP-1, thus reducing the secretion of proinflammatory cytokines such as Ccl2.^53,55 We therefore investigated whether Piezo1 may inhibit the Ccl2/Lcn2 pathway through Klf2, possibly by competing with NF-κB and AP-1. Notably, we observed significantly lower expression of Klf2 in Piezo1 KO BMMSCs than in WT BMMSCs (supplementary Fig. 17a). Next, we performed lentivirus-mediated overexpression of Klf2 (supplementary Fig. 17a) and observed that Klf2 overexpression reversed enhanced adipogenesis and reduced osteogenesis in KO BMMSCs (supplementary Fig. 17b–e). These findings indicate that Klf2 acts as a key downstream mediator of Piezo1 in BMMSCs, influencing their fate determination. Finally, we explored whether Piezo1 opening influences Klf2 expression through the calcium/calmodulin-dependent protein kinase II (CaMKII) pathway.⁵⁴ We observed that WT BMMSCs treated with the CaMKII inhibitor KN-93 showed significantly lower Klf2 levels (supplementary Fig. 17f). While the Piezo1 activator Yoda1 dramatically upregulated Klf2 expression in WT BMMSCs, this effect was markedly reversed by treatment with KN-93 (supplementary Fig. 17f), thus further highlighting the contribution of CaMKII downstream of Piezo1, which influences Klf2 expression and consequently downregulates Ccl2 and Lcn2 in BMMSCs.

To validate the Klf2-Ccl2-Lcn2 axis as a downstream signaling effector of Piezo1 in modulating BMMSC fate in vivo, we performed site-specific genetic manipulation of this signaling axis via intra-tibial injection of lentiviruses encoding eGFP together with scrambled shRNA (Control), shRNA against Ccl2 (shCcl2) or Lcn2 (shLcn2), or overexpression of the mouse Klf2 gene (oe Klf2) under the control of the validated PDGFRα promoter sequence (−2000/+100).⁵⁶ We first confirmed in vitro that lentivirus-mediated Klf2 overexpression or knockdown of Ccl2 or Lcn2 in KO BMMSCs restored their expression to levels comparable to those in WT controls (supplementary Fig. 18a-b). Furthermore, intra-tibial lentiviral delivery achieved highly efficient and specific genetic manipulation of these genes in PDGFRα⁺ BMMSCs in the bone marrow (supplementary Fig. 18c–i) without significant off-target effects in peripheral adipose SVFs (supplementary Fig. 18j), confirming the efficiency and site-specificity of this strategy.

Piezo1^flox/flox mice were generated as previously described^89–91 and were then bred with mice expressing the Cre transgene under the control of PDGFRα (PDGFRα-Cre mice, The Jackson Laboratory). Heterozygous Piezo1^flox/wt mice with or without PDGFRα-Cre transgene expression were then bred to generate homozygous Piezo1^flox/flox mice with (PDGFRα-Piezo1 KO) or without Cre (WT control) expression. These homozygous mice were backcrossed onto a C57BL/6J genetic background for at least ten generations. PDGFRα-Cre tdTomato reporter mice were generated by crossing tdTomato fluorescent reporter (Ai14) mice with PDGFRα-Cre driver mice. Mice of the same genotype were randomly assigned to each group. All the mice were housed in a controlled facility (22 ± 1 °C, 12 h light/dark cycle, 60–70% humidity) with ad libitum access to water and either a standard chow diet (LabDiet 5053) or a high-fat diet (45 kcal% fat, D12451, Research Diet, USA). Fat percentage and lean percentage were determined by the Bruker Minispec LF90 body composition analyzer. Oxygen consumption and the respiratory exchange ratio (RER) were measured via metabolic chambers (Columbus Instruments). The animals were euthanized via cardiac puncture after anesthesia at the end of the experiments. All the animal experiments were approved by the Committee on the Use of Live Animals in Teaching and Research (CULATR, #5184-19, #23-482, #23-479) at the University of Hong Kong. Sample size determination is based on previous experience to obtain significance and reproducibility,^92,93 as well as minimizing the number of animals used as required by the animal ethics committee. The sample size followed common standards and employed four or more biological replicates. All sample sizes are listed in each figure legend. Animals that failed to reach the experimental endpoints due to premature mortality or veterinary-mandated euthanasia were excluded prospectively per predefined criteria.

Mouse tibias were excised and fixed with 4% paraformaldehyde solution (PFA) and then subjected to micro-CT scan using the Skyscan1076 micro-CT system (Bruker, Kartuizersweg, Belgium). The region of interest was 100 to 200 slices below the growth plate, and the following parameters were analyzed using CTAn software (Bruker): bone volume fraction (BV/TV), trabecular number (Tb.N), trabecular spacing (Tb.Sp), trabecular thickness (Tb.Th), and bone mineral density (BMD). Three-dimensional images were reconstructed by stacking the two-dimensional slices from the selected regions using CTVox software (Bruker).

Osmium tetroxide, a radiopaque chemical compound which stains unsaturated lipids and lipoproteins, was used to visualize and quantify bone marrow fat using micro-CT.³² Tibias were decalcified in 0.5 M EDTA (pH = 7.4) for 21 days after micro-CT scanning, and then washed twice with distilled water and once with Sorensen's phosphate buffer (pH = 7.4) for 5 min each. The decalcified tibias were transferred to 1.5 mL tubes containing 300 μL of Sorensen's buffer. Another 300 μL of 2% osmium tetroxide solution was added to each tube to make a 1% solution. The specimens were stained in the solution for 48 h at room temperature on a shaking platform. The stained tibias were washed for three times with 1 mL of Sorensen's buffer for 3 h each at room temperature on a shaking platform. The stained tibias were embedded in 1% low-melting agarose (Invitrogen, #16520100) and placed in a fresh set of 1.5 mL microtubes. The content of BMAs was measured using a Skyscan1076 micro-CT system. The region of interest was the entire tibia. The percentage of BMAs was calculated by dividing the osmium-stained adipose volume by the total marrow volume of the whole tibia (AV/TV) using CTAn software.

To evaluate dynamic periosteal bone formation, 6-week-old male WT, PDGFRα-Piezo1 KO, and PDGFRα-Cre mice received intraperitoneal injections of calcein (20 mg/kg, Sigma-Aldrich #C0875) and Alizarin Red S (40 mg/kg, Sigma-Aldrich #A5533) with a 7-day interval. The mice were euthanized 48 h after the second injection, and tibias were dissected, fixed in 4% PFA, dehydrated in graded ethanol, and embedded in polymethylmethacrylate. Transverse sections of the mid-diaphysis were imaged using a confocal laser scanning microscope (Carl Zeiss LSM 880) with sequential fluorescence excitation to distinguish calcein (488 nm) and alizarin (561 nm) signals. The mineral apposition rate (MAR) was determined by measuring the average distance between fluorescent labels, while the bone formation rate (BFR) was calculated as MAR × MS (mineral surface)/BS (bone surface).

The protocol for isolation and culture of BMMSCs were adhered strictly to standard protocols.⁹⁴ Briefly, 10-week-old male WT and PDGFRα-Piezo1 KO mice were deeply anesthetized with 5% pentobarbital sodium (Dorminal, #013003) and sacrificed by cervical dislocation, followed by immersing in 75% ethanol for 3 min and isolation of femurs and tibias under sterile condition. The bones were then cut at both ends, and the bone marrow was flushed out using a 5 mL syringe equipped with a 25 gauge needle filled with alpha modified Eagle's minimum essential medium (α-MEM) (Gibco, #11900024). The cells were then filtered through a 70 μm cell strainer (Corning, #352350). After 3 h, remove the non-adherent cells that accumulate on the surface of the dish by changing the medium and replacing with fresh complete medium. The MycoBlue Mycoplasma Detector (Vazyme, #D101) was used to detect mycoplasma in BMMSCs according to the manufacturer's instructions. If the reaction solution remains purple, it is determined to be mycoplasma negative; if the reaction solution changes to sky blue, it is determined to be mycoplasma positive.

Mycoplasma negative BMMSCs were grown in the α-MEM medium containing 10% fetal bovine serum (FBS) (Gibco, #10270106) and 1% Penicillin/Streptomycin/Amphotericin B solution (Sigma, #P3032, #S1277, #A9538). When the cells reached 65–70% confluence, subculture the adherent BMMSCs. The differentiation of BMMSCs into osteoblasts was initiated once the subcultured BMMSCs reached 80% confluence using the osteogenic induction medium containing 100 nM dexamethasone (Sigma, #D2915), 50 μg/mL ascorbic acid (Aladdin, #S105026), and 10 mM β-glycerophosphate (Sigma, #G9422). The medium was then replaced every 4 days until day 21. For the adipogenic differentiation, BMMSCs with 100% confluency were treated with the adipogenic induction medium containing 1 μM dexamethasone, 0.5 mM 3-Isobutyl-1-methylxanthine (IBMX) (Sigma, #I5879), 1 μM rosiglitazone (Supelco, # PHR2932), and 1.8 μM insulin (Actrapid, #23103007). After 48 h, the medium was changed to a maintenance medium, i.e., α-MEM containing 1 μM dexamethasone, 1 μM rosiglitazone, and 1.8 μM insulin. The maintenance medium was replaced every 2 days until day 8.

To explore the effects of the Ccl2/CCR2 axis on BMMSC differentiation, 100 ng/mL of recombinant mouse Ccl2 protein (PeproTech, #250-10) or 10 nM CCR2 antagonist INCB3344 (MCE, #HY-50674) was supplemented into the adipogenic and osteogenic induction medium of WT or Piezo1 KO BMMSCs, respectively. To investigate the roles of BMMSC-secreted Lcn2, BMMSCs were either infected with 40 multiplicity of infection (MOI) of lentivirus encoding short hairpin RNA (shRNA) against Lcn2 (Lenti-shLcn2, target sequence: CAGGCAATGCGGTCCAGAAAAA) or scrambled shRNA (Lenti-sc, target sequence: CCTAAGGTTAAGTCGCCCTCG) (OBiO Technology) along with 5 μg/mL of polybrene (OBiO Technology, used for enhancing viral transduction efficiency) for 72 h, or treated with 200 ng/mL of rabbit anti-mouse Lcn2 antibody (raised using full-length mouse Lcn2 as an immunogen⁶⁴) (ImmunoDiagnostics Limited, #12050) or non-immune rabbit IgG in the adipogenic and osteogenic induction medium. To investigate the regulatory effects of AP-1 inhibitor on Ccl2, 10 μM AP-1 inhibitor T-5224 (MCE, #HY-12270) was supplemented into the medium of WT or Piezo1 KO BMMSCs for 24 h. To investigate the regulatory role of NF-κB on Lcn2 expression, WT and PDGFRα-Piezo1 KO BMMSCs were first treated with recombinant mouse Ccl2 protein (rmCcl2, 100 ng/mL) and CCR2 antagonist INCB3344 (10 nM) for 24 h, followed by supplementation with the NF-κB-specific inhibitor QNZ (10 nM, MCE, #HY-13812) for additional 2 h.

To investigate the effects of Klf2 on BMMSC differentiation, 40 MOI of lentivirus encoding eGFP or eGFP together with Klf2 (WZ Biosciences Inc) was used to infect WT or Piezo1 KO BMMSCs for 72 h with the presence of 5 μg/mL polybrene before differentiation. To explore the role of CaMKII in regulating Piezo1-dependent Klf2 expression, WT BMMSCs were treated with 10 μM CaMKII inhibitor KN-93 (MCE, #HY-15465) and/or 5 μM Piezo1 activator Yoda1 (MCE, #HY-18723) for 2 h.

After 21 days of osteogenic induction, cells were washed once with phosphate-buffered saline (PBS), fixed with 10% PFA for 30 min at room temperature, and were subsequently stained with 2% Alizarin Red S (ARS) solution (Sigma, #A5533, adjusted to pH = 4.1–4.3) for 15 min at room temperature. The excess stain was removed by washing with PBS. Calcium deposits were visualized as red staining under a light microscope (Olympus, #ckx53). For quantification, the attached ARS were then destained with 10% cetylpyridinium chloride (Sigma, #C0732) as previously described.⁹⁵ The absorbance was measured at 405 nm using a spectrophotometer.

To visualize the lipid content in adipocytes, cells were fixed with 4% PFA for 15 min at room temperature, washed for three times with PBS (5 min each) and then stained with Oil Red O solution (Sigma, #O0625) for 15 min at room temperature. After staining, the Oil Red O dye was extracted from the stained lipids using 100% isopropanol (500 µL), and the absorbance of this solution was measured at 520 nm with a spectrophotometer to quantify the lipid content.

Ten-week-old C57BL/6J male recipient mice were pretreated with enrofloxacin (Baytril) (0.17 mg/mL) in drinking water for 2 weeks and were then subjected to 9 Gy total body irradiation (split into 2 doses separated by 4 h to minimize gastrointestinal toxicity) using MDS Nordion Gammacell 3000 Elan irradiator to deplete bone marrow. In the following day, bone marrow cells were isolated from femurs and tibias of 6-week-old male donor mice (PDGFRα-Piezo1 KO mice and their WT littermates, or PDGFRα-Cre tdTomato reporter mice) by flushing out the bone marrow with a solution of 10 mL of Hanks' balanced salt solution (HBSS) (Gibco, #14025092) containing collagenase type IV (10 mg, Gibco, #17104019) and dispase (20 mg, Gibco, #17105041) and were incubated at 37 °C for 10 min.⁹⁶ Cells were then filtered through a 70 μm cell strainer and resuspended within HBSS at a concentration of approximately 1 × 10⁷ cells/0.2 mL for tail vein injection into the irradiated recipient mice. After bone marrow transplantation, the recipient mice were treated with enrofloxacin (0.17 mg/mL) in drinking water for another 2 weeks.

Eight-week-old male PDGFRα-Piezo1 KO mice and their WT littermates were acclimated to treadmill (LE8710M, Panlab, Spain) running at a low speed for 2 days as previously described⁹⁷ and then subjected to treadmill exercise at a speed of 15 m/min for 30 min per day. The mice were trained 5 days per week for 6 weeks. Another two age-matched groups without exercise were used as sedentary controls.

BMMSCs isolated from femurs and tibias of 10-week-old male PDGFRα-Piezo1 KO or WT mice were cultured in α-MEM until they reached 80–100% confluence. Conditioned medium was collected from the cells cultured in serum-free medium for 24 h. BMMSCs isolated from femurs and tibias of 10-week-old male C57BL/6J WT mice were treated with half of the adipogenic or osteogenic induction medium and half of the conditioned medium from Piezo1 KO or WT BMMSCs during the induction period. The adipogenic induction medium was changed every 2 days for 8 days, while the osteogenic induction medium was changed every 3 days for 21 days.

After 21 days of osteogenic induction, cells were washed with PBS, fixed with 10% PFA for 30 min at room temperature. Subsequently, cells were stained using an alkaline phosphatase (ALP) staining kit (Sigma, #86R) for 30 min at 37 °C. The staining reaction was terminated by washing with distilled water, and the stained cells were observed under a light microscope.

The Olink Target 96 Mouse Exploratory Probe Kit (Olink, Sweden, #95380A) was employed to assess the protein secretion profile of mouse BMMSCs. Once the cells reached 80% confluence, the culture medium was replaced with serum-free α-MEM and grown for another 24 h. The conditioned medium was collected for analysis using the Olink Signature Q100 system (Olink, Sweden). The Olink data was presented by normalized protein expression (NPX) values on a log2 scale, which was then transformed to linear NPX by using the formula 2^NPX. Data with a missing frequency larger than 75% were excluded from the analysis.⁹⁸

Conditioned medium was harvested from WT and KO BMMSCs after 24 h of serum-free culture. Fabp4 (ImmunoDiagnostics Limited, #32030), Lcn2 (ImmunoDiagnostics Limited, #32050), and Ccl2 (R&D Systems, #MJE00B) levels in the conditioned medium were detected by ELISA according to the manufacturer's instructions and were normalized using protein concentrations determined by bicinchoninic acid (BCA) assay (Thermo Scientific, #23225). Serum levels of Lcn2 and Ccl2 were also quantified by ELISA in PDGFRα-Piezo1 KO mice or WT mice, and mice injected with either rabbit non-immune IgG or anti-Lcn2 neutralizing antibody. For the measurement of adiponectin and leptin levels in the conditioned medium of scWAT explants, serum-free conditioned medium after 24-h culture was collected and subjected to ELISA analysis of adiponectin (ImmunoDiagnostics Limited, #32010) and leptin (R&D Systems, #MOB00B) according to the manufacturer's instructions and were normalized to tissue weight. For ELISA analysis to quantify p65 subcellular localization, BMMSCs were harvested and subjected to separation into the nuclear and cytoplasmic proteins according to the manufacturer's instructions (Beyotime, #P0028). The nuclear and cytoplasmic fractions were then assayed for NF-κB p65 using ELISA kits (Abcam, #ab176648) following the manufacturer's instructions.

BMMSCs were crosslinked using 37% PFA (final concentration 1.42%) in culture medium for 15 min at room temperature, followed by quenching with 1 M glycine solution (final concentration 125 mM) for 5 min. The cells were subsequently collected and washed thrice with cold PBS. Chromatin shearing was achieved through sonication at 25% amplitude for 5 min in ChIP buffer according to a previous established protocol,⁹⁹ alternating between 10-s on and 10-s off sonication frequency. After centrifugation, the supernatant was subjected to DNA isolation as input or immunoprecipitation at 4 °C for overnight with rabbit anti-NF-κB p65 monoclonal antibody (Cell Signaling Technology, #8242, 1:100), c-Jun (Cell Signaling Technology, #9165, 1:50), STAT1 (Cell Signaling Technology, #9172, 1:50), STAT3 (Cell Signaling Technology, #9139, 1:100), SPI1 (Santa Cruz Biotechnology, sc-390405, 1:100), or rabbit non-immune IgG (50 μg/mL). Protein A/G agarose beads (YEASEN, #36404ES60) were then added to incubate with samples for 2 h at 4 °C with rotation. The samples were then washed and eluted using ChIP buffer (excluding NP-40 and Triton-X) and an elution buffer containing 1% SDS and 100 mM NaHCO₃ in distilled water. The protein-DNA complexes in the elution fraction were reverse crosslinked by incubation with 5 M NaCl for 4 h at 65 °C. DNA was then precipitated using absolute ethanol and dissolved in distilled water. To quantify the enrichment efficiency of the chromatin immunoprecipitation, 1 μL DNA sample was used to amplify the predicted promoter region by qPCR. The enrichment of the transcription factor on the targeting region was accessed by calculating the value of "% of input" using the formula: % of Input = 2^(CT^Input−CT^IP)/dilution factor × 100%. The input served as a positive control, while IgG was used as a negative control. The amplified products were collected for agarose gel electrophoresis. The band intensities were quantified using ImageJ software (v1.53). Uncropped films of agarose gel are provided in Supplementary Information.

The luciferase reporter assay was performed by transfecting 0.2 μg of firefly and Renilla luciferase reporter vectors (Addgene, #128046) driven by either the wild-type mouse Lcn2 or Ccl2 promoter (−1999/+1) or their respective mutants (in which the NF-κB or c-Jun DNA binding site was mutated) (Azenta) into BMMSCs seeded in 96-well plates using Lipofectamine 3000 transfection reagent (Thermo Fisher Scientific, #L3000001). After 48-h incubation post-transfection, cells were lysed with Passive Lysis Buffer (Promega, #E1500) and luciferase activities were quantified using the CLARIOstar microplate reader (BMG LABTECH, #0430-101) with dual-luciferase detection capabilities. Firefly luciferase activity values were normalized against Renilla luciferase readings to account for transfection efficiency variations, and relative promoter activities were expressed as the ratio of firefly to Renilla luminescence signals.

Cryosections of the brain, quadriceps, gastrocnemius, and soleus from 8-week-old male PDGFRα-Piezo1 KO mice or WT mice, or PDGFRα-Cre tdTomato reporter mice, along with cultured BMMSCs and scWAT-derived SVFs, were fixed in 4% PFA for 15 min. After PBS washing, samples were blocked with 3% bovine serum albumin (BSA, Sigma, #A7906) in PBS for 1 h, then incubated with rabbit anti-Piezo1 antibody (Proteintech, #15939-1-AP, 1:300) or anti-NF-κB p65 rabbit monoclonal antibody (Cell Signaling Technology, #8242, 1:250) for overnight at 4 °C. Following PBS washing for three times, sections and cells were incubated with Alexa Fluor 488-conjugated goat anti-rabbit IgG (Invitrogen, #A-11008, 1:500) for 1 h at room temperature. Nuclei were counterstained with 4',6-diamidino-2-phenylindole (DAPI) (Thermo Fisher, #62248, 1:500) for 5 min. After final PBS washing, slides were mounted with antifade medium (Abcam, #ab104135). Imaging was performed on a ZEISS LSM 880 confocal microscope. Piezo1-positive signal intensity (green channel) was quantified in ≥5 fields/sample using ZEN Lite software (v3.12). P65 nuclear translocation was quantified using Pearson's correlation coefficient (PCC), calculated in ImageJ (v1.53) between p65 immunofluorescence and DAPI signals across segmented nuclear regions.

Six-week-old male WT and PDGFRα-Piezo1 KO mice were anesthetized via intraperitoneal injection of ketamine/xylazine and administered with 1.25 × 10⁶ transducing units (TU) lentiviruses bearing the PDGFRα promoter-driven gene expression (WZ Biosciences Inc) through bilateral intra-tibial injection into the tibial bone marrow cavity using 1 mL insulin syringes (BD Biosciences). Four different lentiviruses were used in this study, including scrambled shRNA as control (Lenti-PDGFRα-eGFP-scramble), shRNA-mediated Ccl2 knockdown (Lenti-PDGFRα-eGFP-shCcl2), shRNA-mediated Lcn2 knockdown (Lenti-PDGFRα-eGFP-shLcn2), and Klf2 overexpression (Lenti-PDGFRα-eGFP-oe Klf2). Sequences of shRNAs were listed in supplementary Table 2. Under aseptic surgical conditions, 50 μL of viral suspension was delivered in each injection site. Seven days post-injection, in vivo transduction efficiency was assessed using the IVIS (In vivo imaging system) Spectrum imaging system (PerkinElmer) with 480/520 nm filters.

Deparaffinized and rehydrated bone sections (~5 µm thick) were subjected to antigen retrieval in citrate buffer (10 mM, pH 6.0) via microwave irradiation (95 °C, 15 min), followed by 30 min cooling at room temperature. Endogenous peroxidase activity was blocked with 3% H₂O₂ (30 min, dark). After PBS washing (5 min × 3 times), sections were incubated with 10% normal goat serum for 30 min at room temperature to block non-specific binding and then incubated with alkaline phosphatase recombinant rabbit monoclonal antibody (HUABIO, #ET1601-21, 1:1600) for overnight at 4 °C. Following PBS washing, sections were treated with HRP-conjugated goat anti-rabbit IgG polyclonal antibody (HUABIO, #HA1001, 1:500) for 1 h at room temperature. Signal detection employed ultra- diaminobenzidine (DAB) substrate kit (Maxim, #DAB-4033; 5 min at room temperature), with reaction termination in distilled water. Sections were counterstained with hematoxylin (Solarbio, #G1120; 30 s), dehydrated in ethanol, and mounted with clean and environmental mounting medium (Biosharp, #BL1519A).

Tibias were fixed in 4% PFA solution for 24 h and then decalcified in a 0.1 M ZnSO₄ and 0.5 M EDTA (pH 7.4) solution in a decalcification instrument (Milestone) for 48–72 h. The samples were subsequently dehydrated and embedded in paraffin wax using the EG1150 paraffin embedding station (Leica Biosystems). Adipose tissues, skeletal muscles, and decalcified bone sections were deparaffinized and rehydrated as described before.⁹² For H&E staining, the sections were stained with hematoxylin for 2–3 min. The sections were then differentiated in 1% acid alcohol (1% hydrochloric acid in 70% ethanol) for a few seconds to remove excess hematoxylin. After differentiation, the sections were rinsed in running tap water and subsequently stained with eosin for 2 min. The sections were then dehydrated through absolute ethanol solutions and cleared in xylene. Finally, the sections were mounted with coverslips and examined under a light microscope.

Total RNA was extracted from BMMSCs using RNAiso Plus (Takara, #9108) and reverse transcribed into complementary DNA using PrimeScript RT reagent kit (Takara, #RR037A). Quantitative real-time PCR was performed using TB Green® Premix Ex Taq (Takara, #CN830A) with specific primers as listed in supplementary Table 3 on a StepOne System (Applied Biosystems). The relative gene expression level was normalized to the β-actin gene.

All the data were statistically analyzed via GraphPad Prism 9.5 software (GraphPad, San Diego, CA) and are presented as the means ± standard errors of the means (SEMs). Statistical tests were chosen based on data distribution and variance. For normally distributed data with equal variances, an unpaired two-tailed Student's t test was used for pairwise comparisons, one-way analysis of variance (ANOVA) with Tukey's test for multiple groups with a single variable, or two-way ANOVA for comparisons with multiple variables. For normally distributed data with unequal variances, an unpaired t test with Welch's correction or Welch ANOVA was applied. Nonparametric tests, the Mann–Whitney test or the Kruskal-Wallis test, were used for nonnormally distributed data or unequal variances. A p value of <0.05 was considered statistically significant. P values and n values are indicated in the associated figure legends for each figure. Investigators were blinded to group allocation during data collection.