Cold exposure-induced loss of beneficial gut bacterial sEV is associated with the pathogenesis of bone loss through pyroptosis pathway Authors: Li-Min Lei, Xiao Lin, Feng Xu, Fu-Xing-Zi Li, Su-Kang Shan, Ming-Hui Zheng, Bei Guo, Ye-Chi Cao, Zhi-Ang Zhou, Si-Yang He, Yan-Lin Wu, Xi Chen, Fei Zhang, Wen-Hui Zhang, Jia-Qi Zhu, Ya-Rong Wang, Qi-Rong Deng, Zi-Ping He, Feng Wu, Ling-Qing Yuan, Zhi-Hong Li Published: 2026 ABSTRACT AbstractRecent studies suggest that cold temperature (CT) exposure contributes to osteoporosis. While the gut microbiota (GM) is a key regulator of host physiology, CT conditions significantly alter its composition. In this study, we found that co-housing with room temperature (RT) mice, as well as colonization with RT-donor GM, conferred protection against CT-induced bone loss. Using 16 S rRNA gene sequencing, we demonstrated that the CT-triggered decline in intestinal Bifidobacterium pseudolongum (B. p) abundance was rescued by co-housing with RT mice. Furthermore, oral administration of B. p effectively alleviated CT-induced bone loss. Mechanistically, we identified that small extracellular vesicles derived from Bifidobacterium. pseudolongum (B. p-sEVs) were taken up by bone tissue, where they enhanced osteogenesis and suppressed pyroptosis in bone marrow stromal cells (BMSCs). Collectively, our findings suggest that the GM helps preserve bone mass via bacterial sEVs, and that the reduction of B. p and its derived sEVs drives CT-induced bone loss.Graphical Abstract Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12951-026-04719-1.Keywords: Cold exposure, Gut microbiota, Osteoporosis, Bifidobacterium pseudolongum, Pyroptosis Introduction As a prevalent metabolic bone disorder, osteoporosis stems from a disrupted equilibrium between bone formation and resorption. This cellular dysregulation impairs the remodeling process, leading to a progressive loss of bone density and a subsequent rise in the risk of skeletal fractures [1]. Advanced age and declining estrogen levels are the primary drivers of osteoporosis. Consequently, the elderly population, especially women in the postmenopausal phase, remains the group most severely affected by this skeletal condition [2]. Recent evidence identifies exposure to cold temperature as a novel risk factor for bone loss. By analyzing nationwide clinical data from Norway (2008–2018), Dahl et al. demonstrated that lower ambient temperatures were associated with a significantly increased risk of forearm and hip fractures, as well as elevated mortality rates following a hip fracture [3]. As reported by Wee and colleagues, neuropeptide Y (NPY) acts as a pivotal modulator that balances energy distribution and bone density in response to cold environments [4]. Research has demonstrated that osteocyte apoptosis mediates the cold exposure-induced bone loss [5]. Our prior work found that cold exposure can also induce osteoporosis through plasma derived exosomes [6].Notably, exposure to cold temperature induces a suite of systemic effects, and these collateral adaptations may contribute to the onset and progression of osteoporosis. The gut microbiota (GM) acts as an indispensable microbial community that orchestrates systemic stability within the host, thereby underpinning overall physiological health [7].Thriving within the intestinal niche, GM symbionts maintain homeostatic control by regulating enteric tissue differentiation, substrate catabolism, immune programming, and neural circuit modulation [8, 9]. Emerging evidence implicates the GM in the modulation of bone homeostasis. For instance, Sjögren et al. demonstrated that germ-free (GF) mice exhibited a marked decrease in osteoclastogenesis and an elevation in bone mass relative to specific pathogen-free (SPF) controls [10], suggesting that GM appears to drive deleterious bone alterations. Research by Yan et al. established that reconstituting GF mice with a complex SPF microbiota promotes longitudinal bone growth and optimizes the microarchitectural properties of trabecular bone [11], indicating a beneficial role of GM in bone homeostasis. The regulatory influence of gut microbes on bone homeostasis varies by species and can be either protective or detrimental; consequently, the evolution of the microbial landscape serves as a primary driver of divergent skeletal health. Recent studies indicate that microbiota derived from children helps maintain skeletal mass and mechanical integrity in an ovariectomy (OVX) model of osteoporosis [2]. In an OVX mouse model, Chevalier et al. revealed that the protection against skeletal deterioration afforded by warmth is fundamentally driven by alterations in intestinal flora [12]. While previous studies have shown that warmth protects bone mass via the gut microbiota, the mechanisms underlying cold-induced bone loss are distinct and cannot simply be assumed as the exact reverse process. Therefore, exploring the specific role of the microbiota under cold exposure remains a critical scientific gap. The GM functions as a principal modulator of bone homeostasis, making it imperative to delineate the signaling mechanisms linking the GM and host bone. Small extracellular vesicle (sEV)-mediated signal transduction represents an evolutionarily conserved pathway that facilitates crosstalk between bacterial and eukaryotic cells [2]. GM–derived small extracellular vesicle (sEV) serves as carrier that deliver diverse effector cargos to distal target cells, thereby modulating recipient-cell functions [13, 14]. GM derived sEV trigger insulin signaling dysfunction, consequently disrupting glycolytic flux in myocytes [15]. Chen et al. identified corticosteroid-triggered depletion of probiotic sEV as a key mechanism in osteonecrosis development [16]. Despite this, the potential role of GM-derived sEVs in the pathogenesis of cold-induced bone loss remains completely unexplored. Pyroptosis functions as a caspase-mediated, programmed necrotic process. It is uniquely defined by gasdermin-induced membrane perforation, a step that allows for the explosive liberation of potent inflammatory cytokines from the dying cell [17]. Emerging evidence identifies pyroptosis as a fundamental driver across a diverse range of pathologies. This inflammatory death process plays a central role in the progression of neuroinflammation, oncogenesis, and metabolic dysfunction [18–20]. Accumulating data suggest a mechanistic link between pyroptotic pathways and osteoporotic pathologies. Specifically, Tang et al. established that METTL14 directs the m6A modification of HOXA5 to upregulate WNK1 expression; thereby suppressing NLRP3 inflammasome-dependent pyroptosis in macrophages and offering protection against osteoporosis [21].Inhibition of BMSC pyroptosis has been identified as the key mechanism through which melatonin rescues the osteoporotic phenotype triggered by cadmium toxicity [22]. Nevertheless, the potential involvement of pyroptotic pathways in bone loss triggered by cold exposure has yet to be fully elucidated. Here, we investigated whether the GM is a key determinant of CT-induced bone loss in mice. High-throughput 16 S rRNA gene sequencing was further implemented to characterize the taxonomic shifts and isolate key bacterial species potentially driving the CT-induced bone loss. We then determined whether transplanting these candidate strains could mitigate the observed skeletal deterioration. Finally, we assessed the capacity of sEVs derived from the identified gut bacterium to distribute to bone tissue and modulate bone microarchitecture, osteogenesis, and pyroptosis in CT-treated mice. Materials and methods Cell isolation and culture Following established protocols, we isolated bone marrow stromal cells (BMSCs) from C57BL/6 mice [23]. The initial step involved aseptic isolation of femurs and tibias, which were then transferred to low-glucose DMEM + 1% penicillin/streptomycin (Gibco, Invitrogen, NY, USA). We next flushed the bone marrow cavity with low-glucose DMEM + 1% penicillin/streptomycin. The initial culture medium consisted of low-glucose DMEM, 15% FBS (Bovogen, NZ), and 1% penicillin/streptomycin. Following 72 h in culture, nonadherent cells were eliminated by renewing the complete low-glucose DMEM. Adherent cells reaching 90% confluence post 3–5-day culture were passaged. Following BMSC seeding for experimental use, osteogenic induction medium (Cyagen, China) was introduced when cultures attained 80% confluence. For bone marrow macrophages (BMMs) preparation, bone marrow cells were harvested and lysed to remove red blood cells, followed by a 72-h culture in α-MEM with 10% FBS and 1% penicillin–streptomycin. Centrifugation was then employed to isolate the cell pellet, which was subsequently stimulated with 30 ng/mL M-CSF (PeproTech, USA) in α-MEM to obtain mature BMMs. To induce osteoclastogenesis, adherent BMMs obtained after a 3-day culture period were maintained in α-MEM complete medium containing M-CSF (30 ng/mL) and stimulated with RANKL (50 ng/mL; PeproTech). Fresh α-MEM was provided to the cultures every two days. We observed the onset of osteoclast syncytia formation via microscopy at the 5-to-7-day mark, which occurred prior to detectable TRAP enzyme activity. Standard incubation conditions (37 °C, 5% CO₂, and high humidity) were applied to all cell groups. Collection of gut microbiota To investigate the effectt of fecal microbiota transplantation (FMT) on bone mineral density (BMD), fecal samples were obtained from 4 weeks post-RT/CT treated mice. Fresh feces from the 10 mice in each group were collected twice a week, between 9:00 AM and 10:00 AM, pooled, and immediately processed into a bacterial suspension. To ensure the optimal viability of the transplanted microbiota, this suspension was administered via oral gavage to the recipient mice on the exact same day of collection. Gut microbiota (GM) suspensions from RT/CT-treated groups (designated RGM/CGM) were prepared by homogenizing feces in PBS (100 mg/ml) followed by centrifugation (500 ×g, 1 min). Supernatants containing GM were collected, and bacterial loads were quantified via colony-forming unit (CFU) assays. Culture of Bacteria, sEV isolation and characterization Bifidobacterium pseudolongum (ATCC 25526) was propagated in PYG medium (Mingzhou Biotechnology, China) under anaerobic conditions at 37 °C. E. coli ATCC 25,922 was grown concurrently in LB broth (MB2454; Meilunbio, China) at 37 °C with orbital agitation at 170 rpm in aerobic culture. Bacterial concentrations were determined by optical density measurement at 600 nm or colony-forming unit (CFU) enumeration on supplemented PYG/LB agar plates. Bacterial-derived sEVs were isolated according to previously established methods [24]. Upon completion of centrifugation (8,000 × g; 30 min), the supernatant underwent sequential membrane filtration using 0.45-µm followed by 0.22-µm pores. A Centricon® Plus-70 centrifugal concentrator (Merck; 100-kDa molecular-weight cutoff) was used to concentrate the filtrate. The supernatant was processed by ultracentrifugation at 100,000 × g (1 h, 4 °C), resulting in membrane vesicle sedimentation. Following centrifugation, the top phase was eliminated, the sEV pellet resuspended in 1 mL 0.22-µm–filtered DPBS, and the suspension stored at − 80 °C. The ultrastructure of sEVs was evaluated by TEM (Hitachi, Japan), and their size distribution was analyzed using a ZetaView PMX 110 instrument (Particle Metrix, Germany). sEV uptake by BMSCs and BMMs SEV were fluorescently tagged with DiO (Solarbio D5840) and co-cultured with BMSCs or BMMs to verify cellular internalization. For fluorescent labeling, 100 µg sEVs were incubated with 6 µL DiO (5 mM) in 1 mL sterile PBS (37 °C, 30 min). Samples were subjected to a 70-min ultracentrifugation at 100,000 × g and 4 °C to deplete unbound dye. Following 12-hour co-culture with DiO-labeled sEVs at 37 °C, BMSCs or BMMs underwent PBS washes and fixation in 4% paraformaldehyde (30 min, RT). BMSCs underwent permeabilization with 0.5% Triton X-100 for 15 min, followed by three PBS washes, and incubation with red phalloidin (YP0052S, UElandy; 200 µL PBS) for 40 min at RT. After three PBS washes, BMSCs or BMMs underwent nuclear staining using DAPI (Beyotime Biotechnology, China; 5 min, RT). Triple fluorescence signals (green/red/blue) were then imaged on a confocal microscope (Andor B43) following final PBS rinses. Protein Extraction from Bone Tissue The bone tissues were placed in protein lysis buffer, supplemented with steel beads, and mechanically pulverized using a tissue grinder. Following thorough homogenization, the samples were centrifuged, and the supernatant containing total bone protein was collected. Protein concentrations were then determined and normalized using a BCA protein assay kit prior to Western blot analysis. Western blot analysis Protein levels were analyzed via Western blot using established methods [25]. After SDS–PAGE separation of 30 µg protein, membranes of PVDF (Millipore, USA) were used for transfer. After blocking in 5% non-fat milk for 1 h, the membranes were incubated with: anti-RUNX2 (1:2000; Abcam, ab23981), anti-GAPDH (1:4000; Proteintech, 10494-1-AP), anti-COL-I (1:1000; Proteintech, 14695-1-AP), anti-BMP2 (1:2000; Abcam, ab214821), anti-GSDMD (1:1000, Abmart, TA4012F), anti-NLRP3 (1:1000; Proteintech, 30109-1-AP), anti-Caspase 1/20 (1:1000; Proteintech, 22915-1-AP). HRP-secondary antibody incubation (1 h, RT) followed. Immunoblot signals were captured using an Amersham Imager 600 (GE Healthcare, USA) with enhanced chemiluminescence detection reagents (Millipore). Alizarin red staining and ALP staining In 24-well cultures, BMSCs underwent treatment with the intervention factors for 7 or 14 d; following PBS washing, fixation was performed with 4% paraformaldehyde (30 min, RT). Following three PBS rinses, cultures were subjected to either: ALP staining (Servicebio, China) at day 7, or Alizarin Red staining (Servicebio, China) at day 14 (5 min, RT). After final PBS washes, microscopic imaging was performed. Enzyme-linked immunosorbent assay (ELISA) Serum and supernatant levels of pro-inflammatory cytokines (IL-1β, IL-18) and bone turnover markers (Osteocalcin, CTX-1) were determined using specialized ELISA kits (Elabscience, Wuhan, China) following the manufacturer’s protocols. To isolate serum, murine blood was left to coagulate at ambient temperature before being centrifuged (2000 × g, 15 min). Similarly, culture supernatants were cleared of debris via centrifugation at 1000 × g for 20 min. Measurement of absorbance was performed at 450 nm with a microplate reader. Quantification was achieved by referencing standard curves, with results expressed in pg/mL or ng/mL. The specific assay parameters are detailed as follows: Osteocalcin (E-EL-M0864; 0.39-25 ng/mL), CTX-1 (E-EL-M3023; 31.25–2000 pg/mL), IL-1β (E-EL-M0037; 7.81–500 pg/mL), and IL-18 (E-EL-M0730; 15.63–1000 pg/mL). Quantitative real-time PCR (qRT-PCR) analysis Extraction of total RNA was conducted with TRIzol reagent (Invitrogen, USA), followed by quantitative real-time PCR (qRT-PCR) analysis. Each reaction was carried out using TaKaRa SYBR Green Master Mix (Japan) according to standard protocols. The 2^−ΔΔCt algorithm was applied to determine relative expression levels, normalized to GAPDH. In parallel, the abundance of Bifidobacterium was evaluated by normalizing specific gene copies to the total 16 S rRNA gene pool. All primer sequences utilized in this study are documented in Supplementary Table 1. 16 S rRNA gene sequencing After 4-week interventions (RT/CT/RT-CH/CT-CH), fecal DNA from co-housed/non-co-housed mice was sequenced (V3-V4 16 S rRNA, Illumina platform) by Genesky Biotech. Microbiota profiling encompassed both relative abundance and absolute quantitation. Post-QC DNA samples received synthetic spike-ins and were PCR-amplified (16 S V3-V4). After ligating indices to the pooled amplicons, libraries underwent Qubit quantification, normalization, and Bioanalyzer quality assessment, then were sequenced (250-bp PE) on an Illumina MiSeq (San Diego, CA). Raw sequencing data underwent quality filtering via a TrimGalore-Mothur-Usearch pipeline to eliminate: (1) Low-quality reads (Q-score < 20), (2) Adapter/primer contaminants, (3) Short fragments (< 100 bp), (4) High-error sequences (> 2 erroneous bases). This pipeline generated high-integrity clean reads. Sequences were then clustered into OTUs (97% identity cutoff) via UPARSE. Reference spike-in DNA was excluded, and its OTU copy counts per sample were quantified. A quantification curve correlating OTU read counts with spike-in DNA copies was established. Absolute OTU copy numbers were derived from the quantification curve and normalized using rRNA operon copy numbers (rrnDB v5.7). Mothur-based taxonomic assignment classified OTUs at 80% confidence. Microbial α-diversity was assessed via Chao1 and ACE richness indices. Animals Following random group allocation, 12-week-old C57BL/6 males (obtained from the Second Xiangya Hospital, Changsha) were used for all experiments. Specifically, the mice were nourished with a routine maintenance diet purchased from Hunan Slaike Jingda Laboratory Animal Co., Ltd. (Catalog No. hnslkjd00038), which was fully sterilized prior to use. Furthermore, the animal housing facility was strictly maintained under a standard 12-hour light/12-hour dark cycle。We deliberately restricted our cohort to male mice to mitigate potential data noise arising from the female estrous cycle and its subsequent impact on skeletal and microbial homeostasis. The mice were group-housed with their original cage-mates to minimize the stress of establishing new social hierarchies. Furthermore, adequate environmental enrichment, such as nesting materials and cardboard tubes, was provided in each cage to reduce aggressive behaviors. The animals were monitored daily by our researchers and the animal care staff for any signs of fighting, stress, or bite wounds. No severe aggression-related injuries were observed during the experimental period. The mice were group-housed at a density of 4–5 mice per cage. Animals were kept on a routine maintenance diet and housed on sterile corncob-based bedding within a controlled facility. CT mice were housed at 4–6 °C whereas those in the RT group were maintained at approximately 22 °C. To assess the impact of GM exchange on bone homeostasis, we used a co-housing paradigm. CT mice (pretreated for 4 weeks) were housed together with age-matched, RT mice (pretreated for 4 weeks) under RT conditions for an additional 4 weeks. To confirm GM involvement in the effects of cold exposure, both RT and CT cohorts received broad-spectrum antibiotics (ABX) in their drinking water to deplete the GM. Based on previously described protocols [16, 26], the broad-spectrum antibiotic mixture delivered in the drinking water contained vancomycin (50 mg/ml), ciprofloxacin (125 mg/ml), bacitracin (1 mg/ml), ceftazidime (100 mg/ml), gentamycin (170 mg/ml), penicillin (100 U/ml), metronidazole (100 mg/ml), streptomycin (50 mg/ml), and neomycin (100 mg/ml),. To assess the mitigation of cold-induced bone loss, CT mice received direct gut microbiota colonization from RT donors (RGM) via twice-weekly intragastric gavage over 8 weeks. To evaluate the Bifidobacterium pseudolongum against cold-induced osteoporosis, CT mice received twice-weekly intragastric supplementation of Bifidobacterium pseudolongum (B. p) or E. coli (both at 3 × 10⁸ CFUs) for 8 weeks, comparing their osteoprotective efficacy. To test whether extracellular vesicle (EV) secretion is essential for B. p–induced modulation of bone homeostasis in CT miceby delivering 3 × 10^8 CFUs of B. p pre-exposed to GW4869 or DMSO via oral gavage. The bacterial pretreatments were GW4869 (10 µM; S7609, Selleck) or DMSO vehicle. To compare the osteoprotective efficacy of bacterial sEVs, CT mice received twice-weekly administrations for 8 weeks of either B. p-sEVs (1.2 × 10¹⁰ vesicles, ≈ 200 µg protein), or E. coli-sEVs (equivalent dose). To evaluate pyroptosis involvement in cold-induced bone loss, CT mice received every-other-day intraperitoneal injections of MCC950 (10 mg/kg) throughout the 8-week period. Tracing Bacterial-derived of sEV in vivo sEVs were labeled with DiR (Thermo Fisher Scientific, USA) per manufacturer’s protocol and delivered via intragastric administration. For labeling, A 200 µg/mL working solution of DiR was made in ethanol.We added 5 µL of this solution to 100 µg sEV, followed by a 1-hour light-protected incubation at RT. Unbound molecules were eliminated through a high-speed spin (110,000 × g, 70 min, 4 °C). The pellet of sEVs was brought to 100 µg/100 µL in PBS. Mice received 100 µL intragastric administration of either: (a) DiR-labeled sEV in PBS, or (b) DiR solution alone.In vivo imaging of whole animals and isolated femurs/tibiae was conducted at 12 h post- intragastric administration. The imaging was performed using the IVIS Spectrum (PerkinElmer, USA). For bone-tissue sEV tracking, DiO-labeled sEV (Solarbio D5840) were administered via intragastric gavage. Femurs were harvested 24 h post-administration for ex vivo fluorescence imaging.Femoral bones were fixed (4% PFA, 24 h), decalcified (0.5 M EDTA, 4 d), and prepared as cryosections. After triplicate PBS rinses, sections received DAPI counterstain (Beyotime C1002; 5 min, RT). Post-PBS washing, sections underwent confocal imaging (Andor B43) to record green and blue fluorescenceDual-channel fluorescence (DiO/DAPI) was captured post-wash using an Andor B43 confocal system. µCT analysis Femurs were initially fixed for 48 h and then imaged via µCTVivaCT 80) at 11.4 μm resolution, with the X-ray source set to 70 kV and 400 µA. High-resolution reconstructions were generated via NRecon, and skeletal microarchitecture was quantified using CTAn. The trabecular ROI (5% total length) was localized 0.3 mm proximal to the distal growth plate, yielding metrics such as Tb.BV/TV and Tb.Sp. Cortical bone integrity was assessed within a segment covering 10% of the femur, initiated at a 40% distal-to-proximal offset. Parameters including Ct.Th and Ct.Ar/Tt.Ar were calculated to characterize the diaphyseal properties. Tissue staining assays Following a 48-hour fixation in 4% PFA, bone specimens were demineralized using 0.5 M EDTA for a 14-day period and subsequently processed for paraffin embedding. Longitudinal Sect. (5 μm) were then prepared for tartrate-resistant acid phosphatase (TRAP) and osteocalcin (OCN) evaluations. OCN detection was achieved via the Universal Two-Step IHC kit (zsbio, China) in strict adherence to the provided manual, utilizing a specific anti-OCN antibody (23418-1-AP, Proteintech). For pyroptosis and osteoclast assessment, sections were probed with anti-GSDMD (1:300, Abmart, TA4012F), anti-IL-1β (1:300, Abmart, PK563103S) or processed using a TRAP staining kit (G1492, Solarbio). Dynamic bone formation Bone-forming activity was labeled via two intraperitoneal calcein doses (10 mg/kg) administered 9 and 3 days before the end of the study. Fixed and ethanol-dehydrated femurs were embedded in methyl methacrylate, from which 10 μm undecalcified sections were prepared. Calcein-positive double-labeling was observed under fluorescent microscopy, allowing for the calculation of the mineral apposition rate (MAR) and bone formation rate (BFR/BS) within the ROI using Image-Pro Plus 6. Statistical analysis Mice were randomly assigned to experimental groups using a random number table. To eliminate observer bias, all histological quantifications and Micro-CT analyses were performed by two independent researchers who were blinded to the group assignments during the entire assessment process. Statistical significance was determined using GraphPad Prism (v8.0), with a threshold of P < 0.05 considered indicative of a significant difference. Quantitative results are presented as mean ± SD. We applied Student’s t-tests to evaluate differences between two groups, while one-way ANOVA with Dunnett’s multiple-comparison test was employed for multi-group data. Experimental findings were validated across a minimum of three biological replicates. Results CT pre-treatment of mice promotes bone loss in co-housed (CH) RT mice under room temperature conditions To investigate the potential adverse effects of cold exposure on skeletal health, we extracted data regarding the disease burden attributable to low bone mineral density (LBMD) from the Global Burden of Disease (GBD) 2019 database (Table S2). We compiled the mean latitude and daytime temperatures across these 204 countries and territories, and evaluated their association with LBMD fracture-related DALYs (Table S2). Specifically, we employed Pearson correlation analysis to evaluate the relationship between latitude and fracture DALYs (Fig. S1A), as well as between mean temperature and fracture DALYs (Fig. S1B). As shown in Fig. S1A, we observed a positive correlation between latitude and DALYs attributable to LBMD fractures across 204 nations and territories. Conversely, mean country-level temperature was negatively correlated with this disease burden (Fig. S1B). These data indicate that higher latitude and lower average temperature are associated with an increased burden of LBMD-related fractures. Although variables like sunlight and vitamin D were not excluded in this correlational observation, our previous findings have already verified the direct capacity of cold exposure to drive bone loss in mice [6]. Next, we sought to further elucidate the mechanisms responsible for cold-induced bone loss. Mice exhibit coprophagic behavior, characterized by the consumption of feces from cage mates, which facilitates the horizontal transfer of gut microbiota. This natural behavior can result in the transmission of specific phenotypes between individuals within the same environment [27–29]. To evaluate the impact of GM exchange on bone homeostasis, 12-week-old mice were initially assigned to two environmental conditions for four weeks: cold temperature (CT; 4–6 °C) or room temperature (RT; 22–24 °C). In the second 4-week stage, RT-pretreated mice were subdivided into: (1) an RT group (continued RT housing without co-housing) and (2) an RT-CH group (co-housed with CT-pretreated mice at RT). Simultaneously, CT-pretreated mice were allocated into three subgroups: (1) a CT group (continued CT housing), (2) a CT-CH group (co-housed with RT-pretreated mice at RT), and (3) a CT-RT group (transferred to RT without co-housing) (Fig. 1A). Subsequently, micro-computed tomography (µCT) was employed to evaluate femoral microarchitecture. Compared with their RT counterparts, CT mice exhibited significantly lower bone mineral density (BMD) and pronounced deterioration of skeletal microarchitecture (Fig. 1B). These results corroborated the findings from our prior investigations [6].Since cohousing facilitates the coprophagy-mediated exchange of gut microbiota (GM) between cagemates, we hypothesized that this cohabitation enables bidirectional microbial transfer: specifically, the translocation of osteoprotective commensals from RT mice to their CT-CH counterparts, and conversely, osteodestructive signals from CT mice to the RT-CH group. Compared to RT control mice, RT-CH group demonstrated reduced bone mass and deteriorated bone microarchitecture, indicting by the lower BMD, Tb. BV/TV, Tb. N, Tb. Th, Ct. Ar/Tt. Ar and Ct. Th, as well as a trend of increased Tb. Sp (Fig. 1C-I). To isolate the effects of co-housing from those of ambient temperature recovery, a subset of CT mice was similarly returned to the RT environment but housed individually (CT-RT group). Notably, CT-CH mice exhibited increased bone mass and mitigated microarchitectural deterioration compared with CT-RT counterparts (Fig. 1B). Compared with CT-RT controls, CT-CH mice exhibited significant increases in BMD, Tb.BV/TV, Tb.N, Tb.Th, Ct.Ar/Tt.Ar, and Ct.Th, alongside a tendency for lower Tb.Sp (Fig. 1C-I). Dynamic bone morphometry utilizing Calcein double labeling quantified significant deficits in osteogenic activity among RT-CH mice relative to RT mice, manifested by suppression of BFR/BS and MAR, respectively (Fig. 1J-L). Calcein double labeling revealed significantly higher BFR and MAR in CT-CH mice compared with CT-RT controls (Fig. 1J-L). OCN immunostaining demonstrated a significant decrease in trabecular surface osteoblasts in RT-CH mice versus RT controls (Fig. 1M-N). Compared with the CT-RT cohort, CT-CH mice displayed an elevation in osteoblast numbers on trabecular bone surfaces (Fig. 1M-N). Enzymelinked immunosorbent assay (ELISA) showed that a significant decreased osteocalcin (OCN) in RT-CH mice versus RT control (Fig. 1O). Compared with the CT-RT cohort, CT-CH mice displayed an elevation in OCN levle (Fig. 1O). Compared with the RT group, RT-CH mice exhibited markedly higher osteoclast counts by TRAP staining (Fig. 1P, Q). CT-CH animals showed a lower osteoclast count on trabecular bone compared with CT-RT mice (Fig. 1P, Q). ELISA showed a significant increased serum level of C-terminal telopeptides of type I collagen (CTX-I) in RT-CH mice versus RT control (Fig. 1R). However, compared with the CT-RT group, CT-CH mice exhibited markedly lower CTX-I (Fig. 1R). These combined observations indicate that alterations in GM may facilitate bone loss resulting from cold exposure. CT pre-treatment of mice promotes bone loss in co-housed (CH) RT mice under room temperature conditions. A, In vivo treatment protocols for the RT, CT, RT-CH, CT-CH, and CT-RT groups are illustrated in the schematic flow diagram (n = 5 per group). B, Representative µCT reconstructions of mouse femora illustrate trabecular (topr) and cortical ((bottom) compartments. Scale bars: 500 μm (trabecular), 1 mm (cortical) C–I, µCT-derived microarchitectural metrics included BMD, Tb.BV/TV, Tb.Th, Tb.N, Tb.Sp, Ct.Ar/Tt.Ar, and Ct.Th (n = 5 per group). J, Fluorescence microscopy images showing calcein double labelling on the mineralizing surfaces of femoral sections. Scale bar: 50 μm. K, L, Dynamic bone formation indices: mineral apposition rate (MAR) and bone formation rate per bone surface (BFR/BS) (n = 5 per group). M, Representative OCN immunostaining of distal femoral trabeculae and the associated quantification (N) of osteoblast number normalized to bone surface (N.OBs/BS) for RT, CT, RT-CH, CT-CH, and CT-RT cohorts. Scale bar: 50 μm (n = 5 per group). O, ELISA for serum OCN (n = 5 per group) P, Representative TRAP staining in distal femoral sections highlighting osteoclasts, with accompanying quantification (Q) of osteoclast number per bone surface (N.OCs/BS). Scale bar: 50 μm (n = 5 per group). R, ELISA for serum CTX-I (n = 5 per group). Statistical significance was determined using one-way ANOVA with Dunnett’s multiple-comparison test for multiple groups. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 Microbiota depletion affected bone metabolic processes in both RT and CT mice To further determine whether CT-induced bone loss is driven by gut microbiota-dependent mechanisms, we depleted the commensal microbiota in both RT and CT mice using an antibiotic cocktail (ABX) administered via drinking water (Fig. S2A). The efficacy of gut microbiota depletion was confirmed in RT and CT mice, by PCR analysis of total bacterial 16 S rRNA gene from fecal DNA (Fig. S2B). As presented in Fig. S2C–J, RT mice receiving ABX showed BMD loss, decreases in Tb.BV/TV, Tb.N, Tb.Th, Ct.Ar/Tt.Ar, and Ct.Th, and an increase in Tb.Sp when compared to RT controls. However, ABX treatment could alleviate CT-induced bone loss (Fig.S2C-J). Calcein double labeling revealed that ABX treatment suppressed bone formation and mineral deposition in RT mice, whereas it partially restored these activities in CT mice (Fig.S2K-M). Compared with RT mice, RT + ABX mice showed fewer osteoblasts along trabecular surfaces by OCN immunohistochemistry (Fig.S2N, O). Conversely, depletion of the microbiota by ABX rescued the decline in trabecular surface osteoblasts in CT mice (Fig. S2N, O). ELISA assessments revealed that the administration of ABX effectively attenuated circulating OCN concentrations in RT mice, whereas an opposing trend was observed in the CT cohort, where serum OCN levels significantly rose (Fig. S2P). TRAP staining indicated that ABX elevated osteoclast counts under RT conditions (Fig. S2Q, R). Depletion of the gut microbiota reduced osteoclast numbers in CT mice (Fig. S2Q, R). ELISA measurements revealed that ABX administration significantly elevated circulating CTX-I concentrations in the RT mice, whereas it notably suppressed these levels in the CT group (Fig. S2S). These findings provide robust evidence that gut microbes may play an important role in regulating bone homeostasis under both RT and CT conditions. Colonization with GM from RT mice attenuates CT-induced bone loss in the microbiota-depleted mice We evaluated whether colonizing CT mice with gut microbiota from RT donors (RGM) via oral gavage twice weekly for eight weeks could alleviate CT-induced bone loss (Fig. 2A). As expected, RGM attenuated CT-induced bone loss, with higher BMD; increased Tb.BV/TV, Tb.N, Tb.Th, Ct.Th, and Ct.Ar/Tt.Ar; and reduced Tb.Sp (Fig. 2B-I). Calcein double-labeling indicated that CT + RGM mice had augmented new bone formation and mineral apposition compared with CT mice (Fig. 2J-L). Relative to CT mice, CT + RGM mice exhibited more OCN-positive osteoblasts lining trabecular surfaces (Fig. 2M, N). ELISA showed that RGM treatment increased serum OCN levels in CT mice (Fig. 2O). TRAP staining showed that RGM decreased osteoclast numbers in CT mice (Fig. 2P, Q). ELISA showed that RGM treatment decreased serum CTX-I levels in CT mice (Fig. 2R). By contrast, RT mice colonized with CT-derived GM (CGM) exhibited lower bone mass and compromised microarchitecture versus RT controls, with decreases in BMD, Tb.BV/TV, Tb.N, Tb.Th, Ct.Th, and Ct.Ar/Tt.Ar, alongside an increase in Tb.Sp (Fig. 2B-I). Moreover, CGM compromised new bone formation and mineralization in RT mice, as shown by calcein double labeling (Fig. 2J-L). OCN immunostaining showed decreased trabecular osteoblast counts in RT + CGM mice relative to RT controls (Fig. 2M, N). ELISA results showed that CGM treatment decreased serum OCN levels in RT mice (Fig. 2O). TRAP staining documented a significant upsurge of osteoclasts in RT + CGM mice versus RT counterparts (Fig. 2P, Q). An increase in serum CTX-I levels was observed in RT mice following CGM treatment, as determined by ELISA (Fig. 2R). Collectively, these data indicate that the RT-derived GM harbors protective commensals that mitigate CT-induced bone loss; conversely, the CT-derived GM contains gut microbiota dysbiosis elements that exacerbate bone loss and microarchitectural deterioration in RT mice. Colonization with GM from RT mice attenuates CT-induced bone loss in the microbiota-depleted mice. A, A flowchart summarizes in vivo treatment schedules for RT, RT+CGM, CT, and CT+RGM (n=5 per group). The blue box indicates mice housed under cold-temperature conditions, and the yellow box indicates mice housed at room temperature. B, Representative μCT reconstructions of mouse femora illustrate trabecular (topr) and cortical ((bottom) compartments. Scale bars: 500 μm (trabecular), 1 mm (cortical) C–I, μCT-derived microarchitectural metrics included BMD, Tb.BV/TV, Tb.Th, Tb.N, Tb.Sp, Ct.Ar/Tt.Ar, and Ct.Th (n=5 per group). J, Fluorescence microscopy images showing calcein double labelling on the mineralizing surfaces of femoral sections. Scale bar: 50 μm. K, L, Dynamic bone formation indices: mineral apposition rate (MAR) and bone formation rate per bone surface (BFR/BS) (n=5 per group). M, Representative OCN immunostaining of distal femoral trabeculae and the associated quantification (N) of osteoblast number normalized to bone surface (N.OBs/BS) for RT, RT+CGM, CT, and CT+RGMcohorts. Scale bar: 50 μm (n=5 per group). O, ELISA for serum OCN (n=5 per group) P, Representative TRAP staining in distal femoral sections highlighting osteoclasts, with accompanying quantification (Q) of osteoclast number per bone surface (N.OCs/BS). Scale bar: 50 μm (n=5 per group). R, ELISA for serum CTX-I (n=5 per group). Statistical significance was determined using one-way ANOVA with Dunnett’s multiple-comparison test for multiple groups. * P<0.05, ** P<0.01, *** P<0.001, **** P<0.0001 CT induced the loss of Bifidobacterium pseudolongum in mice These data establish that the gut microbiota underpins CT-induced bone loss. To characterize the microbial shifts associated with this skeletal deterioration, we performed high-throughput 16 S rRNA profiling on fecal pellets. As demonstrated by the plateaued rarefaction curves (Fig. S3A), the sequencing depth was comprehensive enough to represent the overall microbial community structures. Rank-Abundance analysis revealed that CT exposure markedly diminished alpha diversity, as evidenced by a significant contraction in OTU-level richness relative to the RT group (Fig. 3A). And RT- CH mice, CT-CH mice had the higher species abundance compared with RT and CT group (Fig. 3A). PCA analysis revealed a distinct separation in gut microbial β-diversity between the RT and CT groups. In contrast, the RT-CH and CT-CH cohorts clustered closely together, suggesting that co-housing intervention drives a robust microbial convergence between these two groups (Fig. 3B). The Ace, Chao1, and Sobs indices indicated significant loss of OTU richness following cold exposure, a phenotype that was partially reversed through co-housing-mediated microbial exchange (Fig. 3C-E). The analysis of dysbiosis index revealed that cold temperature exposure increased the dysbiosis index, but when the CT mice cohoused with RT mice, the dysbiosis index were decreased (Fig. 3F). We analyzed the detected microbiota across multiple taxonomic levels (phylum, family, genus, and species) (Fig. 3G, Fig. S3B-D). At the genus level, cold exposure significantly suppressed Bifidobacterium abundance in mice intestinal microbiota (Fig. 3G). RT-CH mice also decreased the Bifidobacterium abundance compared with RT mice, but CT-CH mice increased the Bifidobacterium abundance compared with CT mice (Fig. 3G). In addition, the abundance of the Bifidobacterium genus was further validated using qPCR with groEL-targeted degenerate primers, with the results normalized to total 16 S rRNA levels (Fig. S3E). At the species level, the relative abundance of Bifidobacterium pseudolongum (B. p) was lower in CT mice than in RT mice (Fig. 3H, I, Fig. S3B). RT-CH mice showed reduced B. p abundance relative to RT controls, whereas CT-CH mice showed increased abundance versus CT, consistent with co-housing–mediated transfer of B. p from RT-CH to CT-CH mice (Fig. 3H, I, Fig. S3B). These findings suggest that B. p may be the principal protective microbe mediating the bone-protective effect of RGM against CT-induced bone loss. CT induced the loss of Bifidobacterium pseudolongum in mice. A, The Rank-Abundance curves in fecal microbiota from RT, CT, RT-CH, CT-CH, and CT-RT groups in OUT level (n = 3 per group). The blue box indicates mice housed under cold-temperature conditions, and the yellow box indicates mice housed at room temperature. B, The PCA analysis in fecal microbiota from RT, CT, RT-CH, CT-CH, and CT-RT groups in OUT level (n = 3 per group). C-E, The analysis the ace index, chao index and sobs index in fecal microbiota from RT, CT, RT-CH, CT-CH, and CT-RT groups in OUT level (n = 3 per group). F, The analysis of microbial dysbiosis index in RT, CT, RT-CH, CT-CH, and CT-RT groups (n = 3 per group). G,The microbiota composition was examined at the genus level (n = 3 per group). H, I, The richness of Bifidobacterium pseudolongum in fecal microbiota from RT, CT, RT-CH, CT-CH, and CT-RT groups (n = 3 per group). Statistical significance was determined using one-way ANOVA with Dunnett’s multiple-comparison test for multiple groups. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 Supplementation of B. p attenuated CT- induced bone loss To establish a causal link, mice were supplemented with B. p twice weekly for eight weeks to evaluate its preventive efficacy against CT-induced osteoporosis. E. coli, a common intestinal commensal in humans and warm-blooded animals, was used as a negative control [2, 30] (Fig. 4A). µCT assessment demonstrated that oral gavage of live B. p counteracted CT-induced osteopenia and improved microarchitecture, reflected by higher BMD, incerased Tb.BV/TV, Tb.N, Tb.Th, Ct.Ar/Tt.Ar, Ct.Th and decerased Tb.Sp relative to CT mice (Fig. 4B–I). However, E. coli supplementation did not modify bone mass or microarchitectural indices in CT mice (Fig. 4B–I). Notably, quantitative µCT analysis revealed no significant statistical difference between the RT and CT + B. p groups, indicating that B. p supplementation effectively rescued cold-induced bone loss to baseline room temperature levels (Fig. 4B–I). We next examined whether B. p could recapitulate the effects of RGM by promoting osteogenesis and inhibiting osteoclastogenesis. Calcein double-label analysis showed improved bone formation and mineral apposition in CT + B. p mice compared with CT mice (Fig. 4J–L). However, calcein double-labeling analysis revealed that the bone mineral apposition rate in the CT + B. p group was lower than that in the RT mice (Fig. 4J–L). OCN immunohistochemistry revealed more osteoblasts along trabecular surfaces in B. p–treated CT controls (Fig. 4M, N). However, the number of osteoblasts along the trabecular surfaces in the CT + B. p group was lower than that in the RT mice (Fig. 4M, N). ELISA showed that B. p treatment increased serum OCN levels in CT mice (Fig. 4O). However, the serum OCN level in the CT + B. p group was lower than that in the RT group (Fig. 4O). TRAP staining indicated that B. p reduced osteoclast numbers in CT mice (Fig. 4P, Q). There is no significant statistical difference between the RT and CT + B. p groups in osteoclast numbers (Fig. 4P, Q). ELISA showed that B. p treatment reduced serum CTX-I levels in CT mice (Fig. 4R). However, the serum CTX-I level in the CT + B. p mice was higher than that in the RT mice (Fig. 4R). However, E. coli had no discernible impact on osteogenesis or osteoclastogenesis in CT mice (Fig. 4J-R). Together, these results indicate that B. p supplementation can partially attenuate CT-induced bone loss in mice. Furthermore, our data suggest that the reduction of B. p may be one of the contributing factors to the skeletal changes observed during cold exposure. Supplementation of B. p attenuated CT- induced bone loss. A, the schematic figure details the in vivo treatment designs for RT, CT, CT + E. coli, and CT + B. p groups (n = 5 per group). The blue box indicates mice housed under cold-temperature conditions, and the yellow box indicates mice housed at room temperature. B, Representative µCT reconstructions of mouse femora illustrate trabecular (topr) and cortical ((bottom) compartments. Scale bars: 500 μm (trabecular), 1 mm (cortical) C–I, µCT-derived microarchitectural metrics included BMD, Tb.BV/TV, Tb.Th, Tb.N, Tb.Sp, Ct.Ar/Tt.Ar, and Ct.Th (n = 5 per group). J, Fluorescence microscopy images showing calcein double labelling on the mineralizing surfaces of femoral sections. Scale bar: 50 μm. K, L, Dynamic bone formation indices: mineral apposition rate (MAR) and bone formation rate per bone surface (BFR/BS) (n = 5 per group). M, Representative OCN immunostaining of distal femoral trabeculae and the associated quantification (N) of osteoblast number normalized to bone surface (N.OBs/BS) for RT, CT, CT + E. coli, and CT + B. p cohorts. Scale bar: 50 μm (n = 5 per group). O, ELISA for serum OCN (n = 5 per group) P, Representative TRAP staining in distal femoral sections highlighting osteoclasts, with accompanying quantification (Q) of osteoclast number per bone surface (N.OCs/BS). Scale bar: 50 μm (n = 5 per group). R, ELISA for serum CTX-I (n = 5 per group). Statistical significance was determined using one-way ANOVA with Dunnett’s multiple-comparison test for multiple groups. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 sEV secretion is necessary for the Bone-Sparing Effect of B. p We next evaluated whether sEV production is essential for the osteoprotective effects of B. p. To this end, we utilized GW4869, a pharmacological inhibitor of neutral sphingomyelinase (nSMase), to abrogate sEV secretion of GM [2] (Fig. S4A). PYG agar plate counts demonstrated that B. p viability was unaffected by GW4869 (Fig. S4B). As shown in Fig. S2C, three days of GW4869 treatment markedly suppressed sEV production in B. p as evidenced by a significant drop in the total protein of isolated sEVs (Fig. S4C). Notably, even after GW4869 washout followed by an additional three days of culture, the total sEV protein yield from GW4869-pretreated B. p remained significantly lower than that of the DMSO-pretreated control (Fig. S4C). In addition, sEV particle concentration was measured by NTA for B. p treated with GW4869 versus DMSO. Consistent with decreased sEV protein, particle numbers from B. p were significantly lower after 3 days of GW4869 and continued to be reduced 3 days post-removal (Fig. S4D). These findings indicate that GW4869 exerts a sustained inhibitory effect on sEV secretion. Then we used DMSO-pretreated B. p and GW4869-pretreated B. p to treat CT mice (Fig. S4A). µCT indicated that DMSO-pretreated B. p likewise benefited skeletal outcomes in CT mice, elevating BMD, Tb.BV/TV, Tb.N, Tb.Th, Ct.Ar/Tt.Ar, Ct.Th and reducing Tb.Sp versus CT (Fig. 5A-H). However, GW4869 pretreatment largely abolished the skeletal benefits of B. p, with bone mass and microarchitectural indices in GW4869-pretreated B. p mice comparable to DMSO-pretreated B. p CT mice (Fig. 5A-H). Calcein double labeling showed that GW4869 pretreatment nearly abolished the gains in new bone formation and mineralization seen in B. p–treated CT mice (Fig. 5I–K). OCN immunohistochemistry showed that GW4869 pretreatment abolished B. p ability to increase osteoblast numbers on trabecular surfaces in CT mice (Fig. 5L, M). ELISA showed that inhibition of sEV release attenuated the effect of B. p in increasing serum OCN levels in CT mice (Fig. 5N). However, inhibiting sEV secretion did not affect B. p’s ability to reduce osteoclast numbers on trabecular surfaces in CT mice (Fig. 5O, P). ELISA revealed that inhibition of sEV release did not alter the effect of B. p in decreasing serum CTX-I levels in CT mice (Fig. 5Q). These findings indicate that sEV act as positive regulators of bone mass in CT mice under B. p treatment. After inhibiting the release of sEV, the promoting effect of B. p on the number of osteoblasts was suppressed. However, no significant difference was observed in its effect on osteoclasts between CT + B. p−preDMSO mice andCT + B. p−preGW4869 mice. These findings suggest that sEV secretion is a prerequisite for the osteoprotective role of B. p, which primarily exerts its skeletal benefits by potentiating osteoblast function in CT mice. sEV secretion is necessary for the Bone-Sparing Effect of B. p. A, Representative micro-CT reconstructions depict trabecular (top) and cortical (bottom) bone architecture in femora from mice subjected to RT, CT, B. p-preDMSO, and CT + B. p-preGW4869. Scale bars: 500 μm (trabecular), 1 mm (cortical). B–H, µCT-derived microarchitectural metrics included BMD, Tb.BV/TV, Tb.Th, Tb.N, Tb.Sp, Ct.Ar/Tt.Ar, and Ct.Th (n = 5 per group). I, Fluorescence microscopy images showing calcein double labelling on the mineralizing surfaces of femoral sections. Scale bar: 50 μm. J, K, Dynamic bone formation indices: mineral apposition rate (MAR) and bone formation rate per bone surface (BFR/BS) (n = 5 per group). L, Representative OCN immunostaining of distal femoral trabeculae and the associated quantification (M) of osteoblast number normalized to bone surface (N.OBs/BS) for RT, CT, B. p-preDMSO, and CT + B. p-preGW4869 cohorts. Scale bar: 50 μm (n = 5 per group). N, ELISA for serum OCN (n = 5 per group) O, Representative TRAP staining in distal femoral sections highlighting osteoclasts, with accompanying quantification (P) of osteoclast number per bone surface (N.OCs/BS). Scale bar: 50 μm (n = 5 per group). Q, ELISA for serum CTX-I (n = 5 per group). Statistical significance was determined using one-way ANOVA with Dunnett’s multiple-comparison test for multiple groups. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 B. p-sEV promote osteogensis of BMSCs in vitro We next isolated and purified sEVs from B. p or E. coli culture media, with their physicochemical properties validated via TEM and NTA. Both B. p-derived sEVs (B.p-sEV) and E. coli-derived sEVs (E. coli-sEV) exhibited typical cup-shaped or spherical morphologies, with mean diameters of 113.8 ± 58.6 nm and 117.8 ± 40.1 nm, respectively (Fig. 6A, B). DiO-labeled B.p-sEV (green) were co-incubated with BMSCs or BMMs for 12 h, followed by cytoskeletal staining with phalloidin (red) and nuclear counterstaining with DAPI. Confocal microscopy revealed that B.p-sEV were efficiently internalized by both BMSCs and BMMs (Fig. 6C). Furthermore, B.p-sEV treatment induced a dose-dependent enhancement of BMSC osteogenesis within the range of 10 to 50 µg/mL, as evidenced by the upregulated expression of RUNX2 (Fig. 6D, Fig.S5A). Accordingly, we used 50 µg/mL as the working dose for subsequent experiments. ARS and ALP staining illustrated that B. p-sEV enhances the osteogenic differentiation of BMSCs compared to PBS, E. coli-sEV group (Fig. 6E). We also assessed additional osteogenic markers, which confirmed that B.p-sEVs promote osteogenic differentiation, as indicated by increased COL1, RUNX2, and BMP2 expression (Fig. 6F, Fig.S5B). By TRAP staining, osteoclastogenesis of BMMs was unaffected by sEVs from B. p (Fig. 6G, H). By measuring mRNA levels of osteoclast differentiation markers (Ctsk, Trap, Atp6vd2, Nfatc1), we similarly observed that B. p-sEV exhibited no significant effect on osteoclast differentiation of BMMs compared to the PPS, E. coli-sEVs group (Fig. 6I). These results demonstrate that B. p-sEV promote the osteogenic differentiation of BMSCs while exerting no significant effect on BMMs. B. p-sEV promote osteogenes of BMSCs in vitro. A, Representative TEM visualization of the ultrastructural features of sEVs. Scale bar: 200 nm. B, Average diameter distribution profile of sEVs. C, Representative fluorescence micrograph showing DiO-labeled sEVs (green) internalized by primary BMSCs and BMMs; nuclei are blue and the cytoskeleton is stained with red phalloidin. Labeled sEVs were co-incubated with BMSCs for 12 h. Scale bar: 20 μm. D, Representative western blot image showing the effect of deferent dosage of B. p-sEV on the protein levels of RUNX2 in BMSCs after 48 h co-incubation. E, Representative ARS and ALP staining of BMSCs after B. p-sEV treatment, presented at the microscopic level and as entire-well images. Scale bars: 500 μm. F, Representative Western blot showing the effect of 50 µg/mL B. p-sEV on COL1, RUNX2, and BMP2 protein levels in BMSCs after 48 h co-incubation. G, Representative images of TRAP staining in BMSCs post–B. p-sEV treatment. Scale bars: 500 μm. H,Quantitative assessment of osteoclasts per well (48-well plate) (n = 3 per group). I, qRT-PCR quantitative results of mRNA (ctsk, trap, atp6vd2, nfatc1 level) in PBS, E. coli-sEV, B. p-sEV treated BMMs. (n = 3 per group) Statistical significance was determined using one-way ANOVA with Dunnett’s multiple-comparison test for multiple groups. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 B. p-sEV could mitigate CT induced bone loss in mice DiR-labeled B.p-sEV and E. coli-sEV were administered via oral gavage, followed by whole-body fluorescence imaging to assess skeletal uptake. While primary signals were sequestered in the liver and spleen (Fig. S6A, B), the high-intensity fluorescence from these organs likely shielded the detection of skeletal signals. To circumvent this, we performed ex vivo imaging of excised bones. The tibiae and femora from sacrificed mice displayed robust fluorescence (Fig. S6C), confirming that orally delivered sEVs successfully translocated to and localized within the bone tissue. To visualize the microdistribution of B.p-sEV and E. coli–sEV, we labeled the vesicles with the green fluorophore DiO. The DiO signal observed in bone (Fig. S6D) confirms sEV presence and points to potential targeting of bone cells by B.p-sEV and E. coli-sEV, with consequent metabolic effects. We next tested whether B. p–sEV was sufficient to preserve bone mass in CT mice, and compared the outcome with E. coli–sEV. We administered 1.2 × 1010 vesicles B. p–sEV twice weekly for 8 weeks to CT mice (Fig. 7A). µCT analysis showed that oral B. p–sEVs prevented CT-induced bone mass loss and improved microarchitecture, with higher BMD, Tb.BV/TV, Tb.N, Tb.Th, Ct.Ar/Tt.Ar, and Ct.Th, and a trend toward lower Tb.Sp versus CT mice and CT + E. coli-sEV mice (Fig. 7B–I). By contrast, E. coli–sEV failed to alter bone mass or microstructural parameters in CT mice (Fig. 7B–I). B. p-sEV could mitigate CT induced bone loss in mice. A, The schematic illustrates in vivo protocols for RT, CT, CT + E. coli– sEVs, and CT + B. p–sEVs (n = 5 per group). The blue box indicates mice housed under cold-temperature conditions, and the yellow box indicates mice housed at room temperature.B, Representative µCT reconstructions of mouse femora illustrate trabecular (topr) and cortical ((bottom) compartments. Scale bars: 500 μm (trabecular), 1 mm (cortical). C–I, µCT-derived microarchitectural metrics included BMD, Tb.BV/TV, Tb.Th, Tb.N, Tb.Sp, Ct.Ar/Tt.Ar, and Ct.Th (n = 5 per group). J, Fluorescence microscopy images showing calcein double labelling on the mineralizing surfaces of femoral sections. Scale bar: 50 μm. K, L, Dynamic bone formation indices: mineral apposition rate (MAR) and bone formation rate per bone surface (BFR/BS) (n = 5 per group). M, Representative OCN immunostaining of distal femoral trabeculae and the associated quantification (N) of osteoblast number normalized to bone surface (N.OBs/BS) for RT, CT, CT + E. coli– sEVs, and CT + B. p–sEVs cohorts. Scale bar: 50 μm (n = 5 per group). O, ELISA for serum OCN (n = 5 per group) P, Representative TRAP staining in distal femoral sections highlighting osteoclasts, with accompanying quantification (Q) of osteoclast number per bone surface (N.OCs/BS). Scale bar: 50 μm (n = 5 per group). R, ELISA for serum CTX-I (n = 5 per group). Statistical significance was determined using one-way ANOVA with Dunnett’s multiple-comparison test for multiple groups. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 Calcein double labeling indicated that B.p-sEV treatment may enhance new bone formation and mineralization in CT mice, relative to both CT controls and E. coli–sEV–treated CT mice (Fig. 7J–L). OCN immunostaining confirmed more osteoblasts on trabecular surfaces in B.p-sEV–treated CT mice than in either CT controls or E. coli–sEV–treated CT mice (Fig. 7M, N). ELISA showed that B.p-sEV treatment increased serum OCN levels in CT mice (Fig. 7O). By TRAP staining, B.p-sEV did not reduce the CT-driven elevation in osteoclast counts (Fig. 7P, Q). ELISA showed that B. p-sEVs failed to reduce serum CTX-I levels in CT mice (Fig. 7R). These findings suggest that B. p-sEV alleviate CT-induced bone loss by promoting osteoblastic bone formation, while having no significant impact on osteoclast numbers in CT mice. B. p-sEV mitigate CT-triggered pyroptosis in bone tissue Growing evidence indicates that pyroptosis serves as a critical regulator of bone metabolic processes [31]. We hypothesized that CT-induced pyroptosis within bone tissue serves as a mechanistic driver of osteoporosis progression. To test this, we quantified the expression of pyroptosis-related proteins in the bones of RT and CT mice. Western blot analysis revealed that CT-exposed mice exhibited elevated levels of key pyroptotic markers, including NLRP3, cleaved-CASP1, and cleaved-GSDMD, relative to the RT group (Fig. 8A, Fig. S7A). Cold stress triggered a pro-inflammatory systemic response, evidenced by the marked augmentation of serum IL-1β and IL-18. These elevations, confirmed by ELISA, suggest that the pyroptotic machinery is actively recruited in the CT mice (Fig. 8B, C). The above results demonstrate that cold exposure promotes pyroptosis levels in bone tissue. B. p-sEV mitigate CT-triggered pyroptosis in bone tissue. A, Representative Western blot illustrating the effects of cold exposure on NLRP3, cleaved CASP1, and cleaved GSDMD protein levels in bone tissue. B, ELISA for serum IL-1β (n = 7 per group). C, ELISA for serum IL-18 (n = 7 per group). D, Representative immunohistochemical staining for GSDMD and IL-1β in distal femoral trabecular regions. E, Representative Western blot of CT mouse bone demonstrating B. p-sEV–induced changes in NLRP3, cleaved CASP1, and cleaved GSDMD. F, ELISA for serum IL-1β (n = 5 per group). G, ELISA for serum IL-18 (n = 5 per group). H, Representative Western blot showing the effects of B. p-sEV on NLRP3, cleaved CASP1, and cleaved GSDMD protein levels in BMSCs. I, ELISA for cell supernatant IL-1β in BMSCs (n = 3 per group). J, ELISA for cell supernatant IL-18 in BMSCs (n = 3 per group). K, Representative microscopy images showing TUNEL-positive staining in BMSCs treated with B. p-sEV. Scale bars: 100 μm. Statistical significance was determined using one-way ANOVA with Dunnett’s multiple-comparison test for multiple groups. P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 Our previous findings demonstrated that B. p-sEV rescued cold exposure-induced bone loss (Fig. 7A-R), and their regulation of bone metabolism may proceed through modulation of pyroptosis in bone. Then we evaluated pyroptosis levels in bone tissue via immunohistochemistry (IHC), revealing that CT mice exhibited markedly enhanced expression of GSDMD and IL-1β compared to the RT group (Fig. 8D). Notably, B. p-sEV treatment significantly suppressed the expression of both GSDMD and IL-1β in CT mice (Fig. 8D). We measured bone-tissue expression of the pyroptosis markers NLRP3, cleaved CASP1, and cleaved GSDMD. Levels were increaed in CT mice compared to RT mice, but were strongly suppressed by B. p-sEV treatment (Fig. 8E, Fig. S7B). Additionally, ELISA quantification of serum IL-1β and IL-18 revealed significant elevations in CT mice compared with RT mice, which were notably reduced following B. p-sEV treatmen (Fig. 8F, G). Our previous findings have demonstrated that B. p-sEV promote the osteogenic differentiation of BMSCs. Given this, we subsequently sought to investigate whether B. p-sEV modulate pyroptosis levels in BMSCs. ELISA revealed that B. p-sEV markedly suppressed IL-1β and IL-18 in BMSCs relative to both PBS and E. coli sEV control (Fig. 8H, I). We evaluated expression of proteins implicated in pyroptosis (NLRP3, Cleaved-CASP1 and Cleaved-GSDMD) in BMSCs, demonstrating B. p-sEV treatment markedly suppressed their expression, compared to PBS and E. coli-sEV group (Fig. 8J, Fig. S7C). TUNEL assay demonstrated that B. p-sEV reduced DNA-damaged BMSCs (Fig. 8K). Collectively, these data indicate that CT promotes pyroptosis in bone, whereas B. p-sEV treatment suppresses CT-induced bone pyroptosis and reduces pyroptotic activity in BMSCs. B. p-sEV restored nigericin-impaired osteogenic differentiation by inhibiting pyroptosis in BMSCs To verify if B. p-sEV could rescue BMSCs from pyroptosis-induced osteogenic failure, we employed nigericin (5 µM) as a pro-pyroptotic stimulus. We then evaluated the efficacy of B. p-sEV in attenuating pyroptosis levels and maintaining osteogenic differentiation under these stressed conditions. TUNEL staining revealed that Nigericin significantly increased the level of DNA-damaged BMSCs; however, this increase was markedly reduced when the cells were treated with B. p-sEV (Fig. S8A). SEM imaging demonstrated that BMSCs treated with nigericin exhibited cell swelling and plasma-membrane pore formation consistent with pyroptosis, while B. p-sEV markedly reduced these alterations (Fig. S8B). We evaluated expression of proteins implicated in pyroptosis (NLRP3, Cleaved-CASP1 and Cleaved-GSDMD) in BMSCs, demonstrating B. p-sEV significantly reversed the Nigericin-induced elevation in pyroptosis protein levels (Fig. S8C, D). ELISA of BMSCs culture supernatants showed that nigericin significantly increased IL-1β and IL-18, whereas B. p-sEV treatment reduced these inflammatory cytokines in nigericin-treated cells (Fig. S8E, F). Nigericin reduced BMSC osteogenesis, as evidenced by decreased RUNX2 and BMP2, while B. p-sEV restored these markers (Fig. S8G, H). ALP and Alizarin Red Staining further confirmed that B. p-sEV mitigated the nigericin-induced impairment of BMSC osteogenic differentiation (Fig. S8I). These results collectively indicate that B. p-sEV rescue the osteogenic differentiation capacity of BMSCs by reversing Nigericin-induced pyroptosis. MCC950 ameliorated CT-induced bone loss by suppressing pyroptosis To determine whether pyroptosis is the driving mechanism behind CT-induced bone loss, CT mice were treated with MCC950 (10 mg/kg, i.p.) every other day for eight weeks (Fig. 9A). To monitor the activation of the pyroptotic cascade, we quantified the expression of NLRP3, cleaved CASP1, and cleaved GSDMD within bone tissues. Although cold stress triggered a robust upregulation of these key mediators, their protein levels were significantly dampened following the administration of the NLRP3 inhibitor MCC950 (Fig. S9A, B). Our ELISA data indicated a substantial rise in circulating IL-1β and IL-18 concentrations among CT mice relative to those maintained at RT (Fig. S9C, D). Notably, the systemic release of both pro-inflammatory cytokines was markedly attenuated following MCC950 treatment (Fig. S9C, D). Immunohistochemical (IHC) analysis of bone sections revealed a marked increase in the immuno-reactivity of IL-1β and GSDMD in CT mice compared to RT controls (Fig. S9E). Conversely, treatment with MCC950 significantly abolished the induction of both pyroptotic markers (Fig. S9E). The above results demonstrate that MCC950 significantly suppressed the pyroptosis level in CT mice. MCC950 ameliorated cold exposure-induced bone loss by suppressing pyroptosis. A,he schematic illustrates the in vivo protocols applied to RT, CT, and CT+MCC950 cohorts (n = 7 per group). The blue box indicates mice housed under cold-temperature conditions, and the yellow box indicates mice housed at room temperature. B, Representative µCT reconstructions of mouse femora illustrate trabecular (top) and cortical ((bottom) compartments. Scale bars: 500 μm (trabecular), 1 mm (cortical).C–I, µCT-derived microarchitectural metrics included BMD, Tb.BV/TV, Tb.Th, Tb.N, Tb.Sp, Ct.Ar/Tt.Ar, and Ct.Th (n = 7 per group). J, Fluorescence microscopy images showing calcein double labelling on the mineralizing surfaces of femoral sections. Scale bar: 50 μm. K, L, Dynamic bone formation indices: mineral apposition rate (MAR) and bone formation rate per bone surface (BFR/BS) (n = 7 per group). M, Representative OCN immunostaining of distal femoral trabeculae and the associated quantification (N) of osteoblast number normalized to bone surface (N.OBs/BS) for RT, CT, and CT+MCC950 cohorts. Scale bar: 50 μm (n = 7 per group). O, ELISA for serum OCN (n = 7 per group) P, Representative TRAP staining in distal femoral sections highlighting osteoclasts, with accompanying quantification (Q) of osteoclast number per bone surface (N.OCs/BS). Scale bar: 50 μm (n = 7 per group). R, ELISA for serum CTX-I (n = 7 per group). Statistical significance was determined using one-way ANOVA with Dunnett’s multiple-comparison test for multiple groups. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 We next assessed the functional impact on bone metabolism following the pharmacological inhibition of pyroptosis in CT mice. µCT demonstrated that MCC950 mitigated CT-induced bone loss and improved bone microarchitecture, as evidenced by higher BMD, Tb.BV/TV, Tb.N, Tb.Th, Ct.Ar/Tt.Ar, and Ct.Th, with a trend toward reduced Tb.Sp versus CT mice (Fig. 9B–I). Calcein double labeling confirmed that MCC950 promoted new bone formation and mineralization in the CT mice (Fig. 9J-L). Osteocalcin (OCN) IHC revealed a rise in osteoblast numbers on trabecular surfaces following MCC950 treatment in CT mice (Fig. 9M, N). ELISA showed that MCC950 increased serum OCN levels in CT mice (Fig. 9O). However, TRAP staining results showed that MCC950 did not alleviate the CT-induced increase in osteoclast numbers (Fig. 9P, Q). ELISA showed that MCC950 failed to reduce serum CTX-I levels in CT mice (Fig. 9R). Collectively, these findings demonstrate that pharmacological inhibition of pyroptosis effectively mitigates CT-induced bone loss, establishing pyroptotic cell death as a pivotal driver in the pathogenesis of cold-induced skeletal deterioration. Discussion The association between the gut and bone has long been recognized, primarily centered on the intestinal absorption of calcium for skeletal mineralization. However, emerging evidence highlights that additional gut-derived regulators—most notably the microbiota, their diverse metabolites, incretins, cytokines, and immune cells—collectively modulate bone homeostasis [32]. Research conducted to date has provided substantial evidence demonstrating the existence of GM-bone axis [2, 10, 16, 33–35]. Our study identifies gut microbiota as a primary mediator of CT-induced bone loss. We demonstrated that RGM transplantation significantly alleviates this skeletal deterioration. Specifically, B. p, a strain preferentially enriched in RGM, functions as a key driver of these anti-osteoporotic benefits. Crucially, our findings reveal that B. p secretes osteoprotective sEVs, which ultimately preserve bone mass in CT mice by suppressing the pyroptotic signaling cascade within bone tissue. Temperature, as a crucial parameter of the human living environment, influences the physiological state of the body. Within the thermoneutral zone, oxygen consumption is minimized as the organism maintains core temperature and homeostasis without metabolic heat production or active cooling [36]. The physiological impact of sustained cold temperature exposure includes significant changes to glucose, lipid, and bone metabolism [4, 37, 38]. Environmental cold exposure also plays a significant role in disease pathogenesis. Substantial epidemiological evidence links cold stress to the development of hypertension [39–41]. CT exposure promotes both the development of atherosclerosis and plaque instability [42]. CT exposure could protect against medial arterial calcification development via autophagy [43]. CT exposure can constrain limb growth in homeotherms through the inhibition of cartilage formation [44]. Our prior study has demonstrated that CT promotes bone loss in mice by inhibiting autophagy [6]. Evidence indicates that maintaining warmth mitigates bone loss by modulating the GM [12]. In this study, we demonstrate that the GM is a major contributor to CT-induced bone loss. When pre-treated RT and CT mice were co-housed at room temperature, RT-CH mice displayed markedly lower bone mass than RT controls, implicating cold-conditioned gut microbiota as drivers of bone loss. However, compared to the CT-RT group, CT-CH mice showed a significantly higher in bone mass. This indirectly suggests that gut microbiota from RT mice may exert a bone mass-promoting effect. Although co-housing suggests microbiota-driven bone loss, potential confounding factors such as social stress and altered behavioral patterns must be considered. Consequently, these non-microbial influences cannot be entirely excluded as contributors to the observed skeletal phenotype. To decouple the effects of the microbiota from environmental temperature, we employed an antibiotic-mediated depletion model. Interestingly, microbial depletion exerted divergent effects on skeletal integrity depending on the thermal environment. This causality was further substantiated by FMT experiments, which showed that RGM mitigated CT-induced bone loss, while CGM transferred the bone-depletion phenotype to RT recipients. Our data thus establish the gut microbiota as a central mediator of bone mass homeostasis under cold stress. Bifidobacterium pseudolongum is a predominant bifidobacterial species across diverse mammalian gut ecosystems [45].The health benefits conferred by different Bifidobacterium species on their hosts have been extensively studied [46–50]. It is reported that Bifidobacterium may ameliorate osteoporosis. Bifidobacterium bifidum mitigates postmenopausal symptoms by inhibiting NF-κB signaling and correcting gut dysbiosis [51]. Bifidobacterium longum ameliorates osteoporosis by elevating serum vitamin D metabolites and remodeling gut microbial composition [52]. Bone metabolism in postmenopausal osteoporosis is improved by Bifidobacterium lactis Probio-M8 via gut microbiome remodeling [53]. Bifidobacterium pseudolongum also functions as a probiotic. Alzheimer’s disease cognitive deficits are ameliorated by time-restricted feeding through a Bifidobacterium pseudolongum-derived propionate/FFAR3 signaling axis [54]. In male mice bone cancer pain models, Bifidobacterium pseudolongum counters morphine tolerance through TLR4 signaling modulation [55]. Nevertheless, no studies to date have established a regulatory relationship between Bifidobacterium pseudolongum and bone metabolism. In this study, we detected a significant decline in the abundance of Bifidobacterium pseudolongum in CT mice. Furthermore, Bifidobacterium pseudolongum was also markedly decreased in RT-CH mice (which were co-housed with CT mice) compared with RT mice. Furthermore, supplementation with Bifidobacterium pseudolongum significantly increased bone mass in CT mice. In the future, it will be crucial to clarify whether Bifidobacterium pseudolongum can serve as a probiotic to prevent osteoporosis of diverse aetiologie. The gut microbiota exerts a regulatory influence on bone homeostasis via diverse signaling axes. Certain commensal taxa augment bone accretion by optimizing the bioavailability of minerals and producing essential metabolites, including short-chain fatty acids (SCFAs) and vitamins B and K [32, 56]. Furthermore, gut bacteria modulate bone metabolism by regulating the profile of immune cells and the secretion of pro-inflammatory cytokines [57, 58]. Bacterial sEVs have emerged as pivotal mediators of molecular exchange, providing a conduit for long-range communication between the gut microbiome and various host tissues [58, 59]. Moreover, bacterial sEV have been shown to modulate bone metabolism [2, 16]. Our study revealed that suppression of sEV release from Bifidobacterium pseudolongum reversed its protective effect on bone mass in CT mice. After oral administration to CT-induced mice, B. p-sEV could enter to bone tissues. And the direct oral administration of B. p-sEV could markedly augment bone mass in the CT model. We further demonstrated that B. p-sEV promote osteogenic differentiation of BMSCs, whereas they exert neutral effects on osteoclast differentiation of BMMs. Although the molecular profile of B. p-sEV remains to be fully elucidated, their bone-protective efficacy underscores their significance in inter-kingdom communication. These sEVs likely transport bioactive cargo—including proteins, metabolites, or small RNAs—that may inhibit osteoclastogenesis [60, 61] or directly modulate the NLRP3/Caspase-1 axis to suppress pyroptosis [62–64]. Future high-throughput omics analyses are warranted to identify the specific “active ingredients” responsible for regulating bone homeostasis under environmental stress. Despite successful in vivo tracing in long bones, the non-specific distribution and natural sequestration of B. p-sEVs in the liver and spleen remain a clinical hurdle. While a functional fraction of these vesicles clearly escapes this first-pass clearance to reach the bone microenvironment, future efforts to engineer these sEVs with bone-targeting ligands could further overcome these off-target effects, providing a more precise therapeutic framework for treating cold-induced or age-related osteoporosis. Pyroptosis is a highly regulated, pro-inflammatory, and lytic form of programmed cell death characterized by cell swelling and membrane rupture [65, 66]. Pyroptosis plays a critical contributor in multiple disease pathologies, including Parkinson’s disease, intestinal inflammation, tumor, and bone disease [67–71]. Exosomal miR-30a-3p from osteoclasts exacerbates Pb²⁺-associated osteoporosis through gasdermin-mediated pyroptosis in osteoblasts [70]. Rosmarinic acid reduces diabetic osteopenia by repressing NLRP3 activation and mitigating osteoblastic pyroptosis [72]. We specifically focused on IL-1β and IL-18 because they are the hallmark proinflammatory cytokines released during canonical Caspase-1-dependent pyroptosis. Unlike other inflammatory mediators, the maturation and secretion of IL-1β and IL-18 are directly triggered by the cleavage of their precursors by activated Caspase-1, which serves as a functional readout of the pyroptotic pathway in osteoclasts. Our study identified that CT triggers pyroptosis in bone tissue. Furthermore, we demonstrated that B. p-sEVs suppress pyroptosis in the bone tissue of CT mice. Our findings further indicate that B. p-sEVs promote the osteogenic differentiation of BMSCs by suppressing pyroptosis. In addition, MCC950, a specific pyroptosis inhibitor, significantly rescued CT-induced bone loss in mice. Interestingly, while the inhibition of pyroptosis effectively restored osteoblast function, it did not significantly attenuate the cold-induced elevation in osteoclast numbers. Therefore, rather than contradicting the classical theory, our data suggest that in the specific context of cold stress, the robust induction of osteoclastogenesis is likely driven by alternative upstream pathways that bypass the NLRP3/pyroptosis axis. Taken together, these findings demonstrate that B. p-sEVs exert their primary osteoprotective effects in CT mice by suppressing the pyroptosis pathway specifically within the osteoblast lineage, thereby highlighting the distinct mechanisms governing bone formation and resorption under cold stress. Moreover, targeted modulation of pyroptosis represents an emerging therapeutic strategy to specifically rescue osteogenesis against CT-induced bone loss. Despite the novel findings and robust mechanistic frameworks presented in this study, several limitations must be acknowledged. First, the epidemiological correlation linking cold temperatures to fracture burden cannot fully exclude secondary environmental and behavioral confounders, such as reduced outdoor physical activity, increased fall risk, and variations in sunlight exposure or dietary vitamin D intake. Similarly, in our animal co-housing models, the potential synergistic or independent contributions of non-microbial variables, such as social stress and behavioral thermoregulation, cannot be completely ruled out. Second, our in vivo models exclusively utilized young male mice; therefore, these findings cannot be directly extrapolated to female populations or specific pathological states like postmenopausal osteoporosis, necessitating future validation in female, aged, and ovariectomized (OVX) models. Third, while the B. p-sEVs effectively inhibit pyroptosis, the exact molecular effectors (specific miRNAs, proteins, or lipids) encapsulated within these vesicles remain unidentified. Furthermore, the precise spatial cascade mechanism linking the decreased abundance of B. p in the gut to the downstream functional action of sEVs in bone tissue remains to be fully elucidated. Specifically, it is still unclear how a decline in the absolute abundance of B. p within the intestinal lumen directly translates into reduced systemic sEV delivery to the skeletal microenvironment. It warrants further investigation whether this cross-organ communication involves other intermediate physiological mediators, such as active transport via circulating immune cells or dynamic alterations in intestinal barrier permeability. Additionally, while we identified the inhibition of pyroptosis as a primary pathway, the complex biomolecular cargo of B. p-sEVs suggests that their overall osteoprotective effects are undoubtedly driven by a synergistic regulation of multiple parallel cellular targets and pathways. Fourth, our current systemic delivery approaches lack exclusive targeting to specific cellular subpopulations like BMSCs, and potential off-target metabolic effects of the pharmacological inhibitor GW4869 cannot be entirely dismissed. Additionally, while MCC950 intervention primarily restored osteoblast function, secondary benefits derived from the suppression of systemic inflammation cannot be completely excluded. The relatively small sample size in certain experimental and sequencing cohorts may limit the statistical power for detecting more subtle physiological variations, highlighting the need for larger-scale investigations in the future. Finally, although the current evaluations were limited to the femur, further investigations into the microstructural changes of the lumbar spine will be valuable to fully elucidate the systemic skeletal impact of cold stress. Conclusion In conclusion, gut microbial dysbiosis serves as a fundamental mediator of CT-induced bone loss, revealing a pathological trajectory fundamentally distinct from previously established heat-adaptive frameworks. Advancing the temperature-microbiota-bone paradigm to the specific strain level, B. p is identified as a pivotal osteoprotective bacterium that is significantly depleted under cold stress. Crucially, a novel spatial and molecular communication mechanism is elucidated: B. p-sEVs act as physical mediators crossing the gut-bone axis to directly suppress BMSC pyroptosis, thereby preserving skeletal integrity (Fig. 10). Consequently, targeted intervention in this unique B. p-sEV-pyroptosis pathway, such as the direct oral administration of B. p or its derived sEVs, represents a potent and precise therapeutic strategy to counteract cold-induced bone mass deterioration. The mechanism of GM-mediated CT related bone loss Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements We extend profound gratitude to the veterinary technicians and animal care specialists at Central South University’s Second Xiangya Hospital Animal Center for their expert husbandry and dedicated oversight of the murine cohort.The scientific schematics were created using BioRender.com. Author contributions The experiments were conceived and designed by L-QY, Z-HL, and L-ML. L-ML conducted the experiments and assembled the figures. Data analysis was carried out by FZ, W-HZ, M-HZ, BG, and Q-RD. Technical support was provided by FX, XL, F-XZL, and S-KS. The manuscript was written by L-QY and L-ML, and revised by J-QZ, Y-RW, Z-PH, J-YD, Y-CC, Z-AZ, S-YH, Y-LW, XC, and FW. All authors read and approved the final manuscript. Funding This work was supported by funding from the National Natural Science Foundation of China (No. 82571030, 82370892, 82470927, 82100494 and 82200869), National Key Research and Development Program of China (No. 2023YFC2507600), National Clinical Key Specialties Main Research Projects (No. Z2023026). Key R&D Plan of Hunan Province (No. 2020SK2078), the Natural Science Foundation of Hunan Province (No. 2025JJ40084, 2025JJ40097, 2025JJ60687 and 2025JJ60687), Fundamental Research Funds for the Central Universities of Central South University (No. 2025ZZTS0807, 2025ZZTS0147 and 2026ZZTS0107). Data availability The data underlying the results presented in Supplementary Figure S1and Table S2 are available in the Global Burden of Disease 2019 (GBD 2019) repository at [http://ghdx.healthdata.org/gbd-2019/data-input-sources]. Declarations Ethics approval and consent to participate Ethical approval and informed consent were not required for this human study because the Global Burden of Disease 2019 (GBD 2019), reviewed and approved by the University of Washington Institutional Review Board (IRB), used de-identified, aggregate data to estimate low bone mineral density (LBMD) and falls, qualified for an informed-consent exemption, and posed no risk of personal identity disclosure. For more details about the original data sources, refer to the GBD 2019 data input sources tool site (http://ghdx.healthdata.org/gbd-2019/data-input-sources).Approval for animal experimentation was granted by the Ethics Committee of the Second Xiang-Ya Hospital, Central South University (No. 20220171). Procedures conformed to the NIH Guide for the Care and Use of Laboratory Animals (8th ed., 2011). Consent for publication All authors agree for publication. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Ling-Qing Yuan, Email: allenylq@csu.edu.cn. Zhi-Hong Li, Email: lizhihong@csu.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The data underlying the results presented in Supplementary Figure S1and Table S2 are available in the Global Burden of Disease 2019 (GBD 2019) repository at [http://ghdx.healthdata.org/gbd-2019/data-input-sources].