Pulsed electromagnetic fields activate a peripheral interoceptive pathway to suppress sympathetic Npy for osteogenesis

Abstract

Chronic stress-induced osteoporosis is a prevalent yet therapeutically challenging condition, with sympathetic overactivation as a key contributor. The dominant pathological mediator and strategies for its precise, non-invasive modulation remain unclear. Here, we identify sympathetic neuropeptide Y as a critical driver of this pathology. Using an unpredictable chronic mild stress model in male mice, we show that neuropeptide Y directly induces cellular senescence in bone marrow adipocytes, thereby impairing osteogenesis. Pulsed electromagnetic field treatment rescued bone loss in wild-type mice, but not in mice with sympathetic neuron–specific deletion of neuropeptide Y (Th^Cre;NPY^fl/fl), demonstrating that neuropeptide Y suppression is required for pulsed electromagnetic field efficacy. Mechanistically, pulsed electromagnetic field action depends on intact sensory innervation and operates through the adipocyte neuropeptide Y receptor Y1R. We further delineate a novel peripheral circuit: pulsed electromagnetic field activates sensory nerves to trigger local release of semaphorin 3 A, which in turn directly suppresses sympathetic neuropeptide Y secretion within the bone marrow microenvironment—a mechanism validated by in vitro co-culture and semaphorin 3 A blockade. This sensory–sympathetic feedback loop functions autonomously, independent of central pathways. Our study reveals a local interoceptive mechanism by which pulsed electromagnetic field counteracts stress-induced osteoporosis, positioning peripheral neuropeptide Y –Y1R signaling as a precise target for neuromodulatory therapy.

Here, the authors show that neuropeptide Y drives stress-induced osteoporosis by promoting bone marrow adipocyte senescence and impairing bone formation. Pulsed electromagnetic field treatment reverses this via sensory nerve-derived SEMA3A, which suppresses neuropeptide Y through a local circuit.

Introduction

Chronic stress is a major and growing public health concern, contributing to the pathogenesis of a wide spectrum of disorders, including metabolic diseases, immune dysregulation, and neurodegeneration^1–4. Beyond its psychiatric impact, chronic stress is increasingly recognized as a significant risk factor for somatic diseases, with the skeleton being a prominent target⁵. Meta-analyses have firmly established depression as a major risk factor for osteoporosis. Indeed, individuals with depression exhibit substantial reductions (7–15%) in bone mineral density at clinically critical sites such as the spine and hip, and over 40% of these patients develop osteoporosis^6,7. Unlike postmenopausal osteoporosis, stress-induced bone loss disproportionately affects premenopausal women and adolescents, potentially leading to lifelong skeletal fragility—a challenge compounded by the recurrent nature of depressive episodes^8–10. Despite growing appreciation of the link between chronic stress and osteoporosis, the mechanisms, particularly the specific neural circuits and key efferent molecules involved, remain incompletely understood^11,12. Recent studies have identified discrete central neural circuits that mediate stress-induced bone loss¹³, and the emerging concept of ‘skeletal interoception’ provides a novel perspective on brain-bone communication. This dynamic, bidirectional crosstalk between the brain and the skeleton, essential for maintaining bone homeostasis, is increasingly conceptualized within the framework of ‘skeletal interoception’^14–16.

The afferent (sensory) limb of this circuit is primarily composed of sensory nerves, which detect mechanical and chemical cues within the bone microenvironment and relay this information to the central nervous system. Landmark studies have demonstrated that osteoblasts release prostaglandin E2 (PGE2), which activates sensory nerves to signal the hypothalamus, thereby modulating neuropeptide Y (NPY) expression and controlling systemic bone and fat metabolism^17–19. The efferent limb involves sympathetic outflow, which classically exerts catabolic effects on bone via norepinephrine signaling^17–19. Within this framework, sensory nerves also release local signaling molecules that directly regulate bone remodeling^20,21. Sema3A, a class 3 semaphorin, is of particular relevance: beyond its classic role in axon guidance, it is a potent osteoanabolic factor secreted by sensory nerves, promoting osteoblastogenesis and inhibiting osteoclastogenesis^22–24. We previously demonstrated that pulsed electromagnetic fields (PEMFs) therapy upregulates Sema3A expression in sensory nerves, shifting mesenchymal stem cell fate toward osteogenesis and attenuating cellular senescence in aged bone²⁵. Disruption of this neural balance underlies various bone pathologies. Chronic stress, for instance, drives sympathetic overactivity and leads to bone loss^11,26. While recent work has begun mapping central circuits mediating stress-induced osteoporosis^13,27, whether analogous peripheral, local neural circuits exist within bone—and if they can be therapeutically harnessed—remains unknown. This knowledge gap forms the rationale for our present study.

Emerging evidence suggests that sensory and sympathetic nerves are often anatomically juxtaposed in peripheral tissues, implying potential direct coupling and crosstalk in both physiological and pathological responses^28,29. These nerves may communicate directly through local signaling molecules, forming ‘peripheral neural circuits’ that operate independently of central relay. Such local circuits enable rapid, tissue-specific responses to external stimuli, representing a form of ‘peripheral interoception’ that maintains tissue homeostasis without requiring central integration. However, whether such a peripheral neural circuit exists in bone, and how it might be therapeutically harnessed, remains unknown.

The cellular and molecular mechanisms by which stress impairs skeletal integrity are multifaceted. Previous research has primarily focused on the shift in bone marrow mesenchymal stem cell (BMSC) fate away from osteogenesis and toward adipogenesis, and the lipotoxic bone marrow environment resulting from the accumulation of bone marrow adipocytes (BMAds)^30,31. Recently, a novel mechanism has emerged: the induction of cellular senescence in BMAds has been shown to damage osteogenesis and angiogenesis, contributing to glucocorticoid-induced osteoporosis and aging-induced osteoporosis^32–34. However, whether BMAds senescence plays a causative role in chronic stress-induced bone loss, and how it might be linked to the dominant pathological driver of stress—sympathetic overactivation—remains largely unexplored.

PEMFs represent a promising non-invasive therapeutic modality known to promote osteogenesis^25,35,36. The skeletal benefits of PEMFs are mediated through multiple mechanisms, including the promotion of osteogenesis, and potentially, the clearance of senescent cells and suppression of senescence-associated secretory phenotype (SASP)^25,35,37,38. We have previously demonstrated that in aged bone, PEMFs act through sensory nerves to induce the secretion of Sema3A, which shifts leptin receptor-positive (LepR⁺) mesenchymal stem cells (MSCs) toward osteogenesis and attenuates cellular senescence²⁵. This finding positioned sensory nerves as a key sensor for PEMFs. Having identified sensory nerves as a key sensor for PEMFs, and given their established role as the afferent limb of skeletal interoception^16–19,39, we hypothesized that PEMFs might exert their therapeutic effect by targeting this specific interoceptive circuit within the bone microenvironment.

Here, we investigate the novel hypothesis that PEMFs alleviate chronic stress-induced osteoporosis by activating a peripheral interoceptive-like mechanism in bone. Specifically, we test whether sensory nerve-derived Sema3A directly suppresses sympathetic NPY release within the local bone microenvironment, thereby mitigating NPY-driven BMAd senescence and promoting bone formation.

Results

PEMF ameliorates osteoporosis by modulating sympathetic activity in a sensory nerve-dependent manner

In our previous study, we demonstrated that PEMFs prevent bone loss in aging models by modulating sensory nerves²⁵. To investigate the impact of PEMFs on the sympathetic nervous system in aged bone, we first assessed local sympathetic innervation via tyrosine hydroxylase (TH) immunohistochemistry (Fig. 1a). In aged (20-month-old) mice, PEMF treatment significantly preserved TH⁺ nerve fiber density in the femur compared to sham-treated controls (Fig. 1b, d). In contrast, young (4-month-old) mice displayed a higher baseline density of TH⁺ fibers, which was not further increased by PEMF under these non-stressed conditions (Fig. 1b, d). Consistent with prior reports of sympathetic overactivation in aging, we directly measured tissue norepinephrine (NE) levels in young and aged mice. Aged mice exhibited significantly elevated bone NE content relative to young controls (Fig. 1e), confirming sympathetic hyperactivity in the aged bone microenvironment despite the observed reduction in TH⁺ fiber density (Fig. 1b, d). Notably, PEMF treatment did not significantly alter NE levels in aged bone, despite eliciting a clear osteoprotective effect. This dissociation indicates that the therapeutic action of PEMFs is not mediated through modulation of global tissue NE levels. Crucially, to test if sensory nerves mediate this PEMFs-dependent modulation of sympathetic function, we employed 20-month-old iDTR;Advillin^Cre mice. In sensory nerve-ablated mice, the effects of PEMFs on preserving TH⁺ fiber density (Fig. 1c, f) were completely abolished. Collectively, these findings delineate a novel interoceptive pathway whereby PEMFs, acting via sensory nerves, regulate sympathetic innervation to preserve its density in aged bone.

Fig. 1. PEMFs prevent bone loss by regulating sympathetic innervation through sensory nerves.

a Schematic drawing of the PEMF device. Illustration created using Blender (version 4.3, Blender Foundation, Amsterdam). b Representative immunofluorescence images of femur sections from 4-month and 20-month mice stained for tyrosine hydroxylase (TH, green, a marker for sympathetic nerves) and DAPI (blue, nuclei). c Representative immunofluorescence images of femur sections from sensory nerve-ablated (iDTR;Advillin^Cre; termed iDTR_Avifl/-) mice stained for TH (green) and DAPI (blue). d Quantitative analysis of the TH⁺ sympathetic nerve fibers. e Quantitative analysis of the NE level. f Quantitative analysis of the TH⁺ sensory nerves. n = 5 biological replicates for each group for all quantifications. Data are represented as mean ± SD, and statistical analyses were conducted with a 95% confidence interval. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by two-way ANOVA with Tukey’s multiple comparisons for (d–f). P value: d P = 0.0019 (4mon Sham vs 20mon Sham), P = 0.0466 (20mon Sham vs 20mon PEMFs). e P < 0.0001 (4mon Sham vs 20mon Sham). f P < 0.0001 (iDTR^wt Sham vs iDTR^wt PEMFs), P < 0.0001 (iDTR^wt PEMFs vs iDTR_Avifl/- PEMFs). Source data are provided as a Source Data file.

PEMFs treatment rescues chronic stress-induced bone loss by inhibiting senescence of BMAds and modulating sympathetic activity

Having established that PEMFs regulate bone sympathetic innervation via sensory nerves in aging (Fig. 1), we sought to determine whether this newly identified interoceptive pathway also underpins the therapeutic effects of PEMFs in other conditions characterized by sympathetic overactivity. Chronic stress is a well-established inducer of sympathetic excitation and osteoporosis. Therefore, we next investigated the therapeutic potential of PEMFs against chronic stress-induced bone loss and examined whether their efficacy involves modulation of the sensory-sympathetic axis. We subjected mice to an unpredictable chronic mild stress protocol to establish the model. Chronic stress successfully induced anxiety-like behaviors, as evidenced by reduced central exploration in the open field test and fewer open-arm entries in the elevated plus maze (Supplementary Fig. 1a–d). Micro-CT analysis confirmed that chronic stress induced significant trabecular bone loss in the femur, characterized by decreased bone volume fraction (BV/TV), reduced trabecular number (Tb.N), and increased trabecular separation (Tb.Sp) in the Sham-Stress group (Fig. 2a, c–f). Strikingly, daily PEMF treatment largely prevented this bone loss, as shown by improved bone mass and microarchitecture in the PEMF-Stress group (Fig. 2a, c–f). In parallel, we assessed the impact on cortical bone. Chronic stress led to a significant deterioration of cortical bone quality, as indicated by reduced cortical bone area (Ct.Ar),cortical thickness (Ct.Th) and Ct.Ar/Tt.Ar (Supplementary Fig. 1k–o). Notably, PEMF treatment effectively prevented this cortical bone loss, restoring parameters to levels comparable to non-stressed controls. This demonstrates that the osteoprotective effect of PEMF extends to the cortical compartment, underscoring its comprehensive efficacy in preserving overall bone integrity under chronic stress.

Fig. 2. PEMFs prevent chronic stress-induced bone loss and its mechanisms.

a A representative micro-CT image of the distal femoral cancellous bone. b Representative immunofluorescence images of femoral sections for tyrosine hydroxylase (TH, sympathetic marker, green) and DAPI (blue, nuclei). c–f Quantification of trabecular bone volume fraction (BV/TV), trabecular cancellous bone number (Tb.N), trabecular cancellous bone thickness (Tb.Th), and trabecular cancellous separation (Tb.Sp). g Quantification of TH⁺ nerves. h Representative calcein double-labeling images of femoral sections. i Quantification of MAR. j Immunofluorescence images of femoral sections for Perilipin (green) and Opn (red). k, m Quantification of the number of Perilipin-positive cells per mm² tissue area and the relative areas of Opn (N. Perilipin⁺ cells/Ar, Quantitative relative area). l Immunofluorescence images of femoral sections for Perilipin (green) and γH2AX (red). n Quantification of the number of Perilipin - and γH2AX-double positive cells per mm² tissue area (No. Perilipin⁺γH2AX⁺ cells). o Representative SA-βGal and oil red staining of femoral sections. p Percentage of SA-βGal⁺ cells among oil red⁺ BMAds. n = 6 biological replicates per group for all quantifications. Data are represented as mean ± SD, and statistical analyses were conducted with a 95% confidence interval. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by two-way ANOVA with Tukey’s multiple comparisons for (c–g, i, k, m, n, p). P value: c P < 0.0001 (Conn sham vs Stress sham), P < 0.0001 (Stress Sham vs Stress PEMFs). d P < 0.0001 (Conn sham vs Stress sham), P = 0.0011 (Stress sham vs Stress PEMFs). e P = 0.001 (Conn sham vs Stress sham), P = 0.0083 (Stress sham vs Stress PEMFs). f P < 0.0001 (Conn sham vs Stress sham), P < 0.0001 (Stress sham vs Stress PEMFs). g P < 0.0001 (Conn sham vs Stress sham), P < 0.0001 (Stress sham vs Stress PEMFs). i P = 0.0018 (Conn sham vs Stress sham), P = 0.0224 (Stress sham vs Stress PEMFs). k P = 0.0007 (Conn sham vs Stress sham), P = 0.0173 (Stress sham vs Stress PEMFs). m P < 0.0001 (Conn sham vs Stress sham), P = 0.0174 (Stress sham vs Stress PEMFs). n P < 0.0001(Conn sham vs Stress sham), P = 0.0085 (Stress sham vs Stress PEMFs). p P < 0.0001 (Conn sham vs Stress sham), P < 0.0001 (Stress sham vs Stress PEMFs). Source data are provided as a Source Data file.

Given the link between sympathetic tone and bone mass, we assessed sympathetic nerve density. TH immunostaining revealed that chronic stress increased TH⁺ nerve fibers within the bone marrow, and PEMF treatment normalized this increase (Fig. 2b, g), suggesting a modulatory effect on sympathetic activity. We next explored the cellular mechanisms underlying this protection. Dynamic histomorphometry revealed a severe suppression of bone formation (mineral apposition rate, MAR) in the Sham-Stress group, which was significantly rescued by PEMF treatment (Fig. 2h, i). The bone marrow environment in stressed mice shifted toward adipogenesis at the expense of osteogenesis, manifested by a significant increase in Perilipin⁺ adipocytes and decreased osteopontin (Opn) expression. PEMF treatment effectively reversed this pathological shift and reduced marrow fat accumulation (Fig. 2j, k, m). Finally, we investigated the role of cellular senescence in BMAds. Immunofluorescence revealed that the DNA damage marker γH2AX was significantly elevated and co-localized with Perilipin⁺ BMAds in stress-treated mice (Fig. 2l). Quantification confirmed a marked increase in γH2AX⁺ area, which was effectively reduced by PEMF treatment (Fig. 2n). Furthermore, co-staining showed a significant increase in SA-βGal⁺/Oil Red O⁺ double-positive senescent BMAds following stress, and PEMF treatment effectively counteracted this senescence (Fig. 2o, p). In contrast, and consistent with previous studies, the number of TRAP-positive osteoclasts was not significantly altered by PEMF treatment in either stressed or control mice (Supplementary Fig. 1e, f). While chronic stress elevated NE levels, PEMF rescued bone loss without normalizing this catecholamine surge (Supplementary Fig. 1g). Moreover, PEMF treatment did not reverse the elevated levels of corticosterone or pro-inflammatory cytokines (e.g., TNF-α, IL-6) induced by chronic stress (Supplementary Fig. 1h–j). This dissociation prompted us to investigate downstream effectors. Collectively, these data establish PEMF as an effective countermeasure against stress-induced osteoporosis, with effects linked to restored bone formation, modulated sympathetic overactivity, and alleviated BMAd senescence. Importantly, these findings provide a key mechanistic insight: the osteoprotective effect of PEMF is dissociated from global NE modulation.

Sensory innervation is required for the protective effects of PEMF against stress-induced bone loss

To determine whether intact bone sensory innervation is necessary for the anti-osteoporotic effects of PEMF against chronic stress, we subjected sensory neuron-ablated mice (iDTR;Advillin^Cre; termed iDTR_Avifl/-) to unpredictable chronic mild stress, followed by either PEMF or sham treatment. The absence of sensory neurons completely abolished the bone-protective efficacy of PEMF. Micro-CT analysis revealed that PEMF treatment conferred no significant improvement in trabecular bone parameters—including BV/TV, Tb.N, and Tb.Th—in iDTR_Avifl/- mice compared to their sham-treated counterparts (Fig. 3a, b). Furthermore, in these mice, PEMF failed to modulate the severe anxiety-like behaviors induced by chronic stress, as assessed by the elevated plus maze and open field tests (Fig. 3c–f). At the mechanistic level, sensory neuron ablation prevented PEMF from normalizing the pathological bone microenvironment. First, PEMF lost its ability to suppress the elevated sympathetic tone (as indicated by TH⁺ nerve fiber density) in the absence of sensory nerves (Fig. 3g, h). Second, the inhibitory effect of PEMF on BMAd senescence was also rendered ineffective; the percentage of senescent (γH2AX⁺) BMAds remained equally elevated in iDTR_Avifl/- mice regardless of PEMF treatment (Fig. 3i, j). We then interrogated the molecular link underpinning this strict sensory nerve dependency by measuring Sema3A, a key neurotrophic factor. In wild-type (WT) mice, chronic stress significantly reduced Sema3A protein levels specifically in the bone marrow microenvironment, an effect that was robustly reversed by PEMF treatment (Fig. 3k). In contrast, serum Sema3A levels remained unaltered by either chronic stress or PEMF intervention (Fig. 3l), indicating a local rather than systemic mode of action. Strikingly, this local PEMF-induced upregulation of bone marrow Sema3A was completely abolished in iDTR_Avifl/- mice (Fig. 3m). In these sensory nerve-ablated mice, bone marrow Sema3A levels remained low and were unaffected by PEMF treatment (Fig. 3m). Consistent with the observation in WT mice, serum Sema3A concentration was also unchanged by PEMF in the iDTR_Avifl/- background (Fig. 3n). These data unequivocally demonstrate that PEMF stimulates the local production of Sema3A in bone in a manner absolutely dependent on intact sensory innervation. Finally, to establish the sufficiency of Sema3A in mediating osteoprotection downstream of sensory nerves, we performed a unilateral intraosseous rescue experiment in iDTR_Avifl/- mice under continued chronic stress. Local injection of recombinant Sema3A protein into the right femur, but not PBS vehicle into the contralateral femur, significantly rescued trabecular bone mass and microarchitecture (Supplementary Fig. 2a–e). Moreover, Sema3A treatment markedly reduced the burden of senescent (γH2AX⁺/Perilipin⁺) BMAds (Supplementary Fig. 2f, g). In summary, these phenotypic, cellular, and molecular data converge to demonstrate that the integrity of sensory innervation is an indispensable prerequisite for PEMF to counteract chronic stress-induced osteoporosis. The therapeutic signal of PEMF must be transduced by intraosseous sensory nerves to elevate local Sema3A, which is both necessary and sufficient for alleviating sympathetic overactivity, BMAd senescence, and consequent bone loss.

Fig. 3. Sensory innervation is required for the protective effects of PEMF against chronic stress-induced bone loss.

a Representative 3D micro-CT images of distal femoral trabeculae from sensory neuron-ablated (iDTR;Advillin^Cre; termed iDTR_Avifl/-) mice subjected to unpredictable chronic mild stress with sham or pulsed electromagnetic fields (PEMF) treatment. b Quantification of bone volume fraction (BV/TV), trabecular number (Tb.N), trabecular thickness (Tb.Th), and trabecular separation (Tb.Sp). c–f Anxiety-like behaviors assessed by elevated plus maze (entries and time in open arms) and open field test (central area entries and time). g Immunofluorescence images of tyrosine hydroxylase (TH⁺) sympathetic nerves (green) and DAPI (blue) in bone marrow. h Quantitative analysis of the TH⁺ nerves. i Immunofluorescence co-localization of phosphorylated histone H2AX (γH2AX, red) with Perilipin⁺ bone marrow adipocytes (green). j Number of Perilipin⁺γH2AX⁺ cells/mm². k, l Semaphorin 3A (Sema3A) protein levels in bone marrow (k) and serum (l) of WT mice under non-stress (Conn) or unpredictable chronic mild stress with sham/PEMF. m, n Sema3A levels in bone marrow (m) and serum (n) of iDTR_Avifl/- mice under chronic stress with sham or PEMF. n = 6 mice per group for all panels. Data are presented as mean ± SD. Statistical significance was determined by two-tailed unpaired Student’s t-test for (b, d, f, h, j, m, n), and by two-way ANOVA with Tukey’s multiple comparisons for (k, l). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. P value: b BV/TV: P = 0.9779 (iDTR_Avifl/- sham vs iDTR_Avifl/- PEMFs). Tb.Th: P = 0.7619 (iDTR_Avifl/- sham vs iDTR_Avifl/- PEMFs). Tb.N: P = 0.6884 (iDTR_Avifl/- sham vs iDTR_Avifl/- PEMFs). Tb.Sp: P = 0.3693 (iDTR_Avifl/- sham vs iDTR_Avifl/- PEMFs). d Open-arm entries: P = 0.7847 (iDTR_Avifl/- sham vs iDTR_Avifl/- PEMFs). Open-arm time: P = 0.8175 (iDTR_Avifl/- sham vs iDTR_Avifl/- PEMFs). f Central entries: P = 0.689 (iDTR_Avifl/- sham vs iDTR_Avifl/- PEMFs). Central time: P = 0.5239 (iDTR_Avifl/- sham vs iDTR_Avifl/- PEMFs). h P = 0.9575 (iDTR_Avifl/- sham vs iDTR_Avifl/- PEMFs). (j) P = 0.9697 (iDTR_Avifl/- sham vs iDTR_Avifl/- PEMFs). k P < 0.0001 (Conn sham vs Stress sham), P = 0.0001 (Stress sham vs Stress PEMFs). l P = 0.996 (Conn sham vs Stress sham), 0.9762 (Stress sham vs Stress PEMFs). m P = 0.7333 (iDTR_Avifl/- sham vs iDTR_Avifl/- PEMFs). n P = 0.6723 (iDTR_Avifl/- sham vs iDTR_Avifl/- PEMFs). Source data are provided as a Source Data file.

PEMFs alleviate BMAd senescence by suppressing sympathetic nerve-derived NPY

We next sought to elucidate the mechanism by which PEMFs alleviate BMAd senescence. BMAds were isolated from mice subjected to chronic stress followed by either sham or PEMF treatment. Quantitative PCR revealed that PEMF treatment significantly downregulated the expression of key senescence markers, including Cdkn1a and Cdkn2a, in BMAds from stressed mice compared to the sham-treated controls (Fig. 4a, b). This confirmed that BMAds are a cellular target of the anti-senescence effect of PEMFs in vivo. However, direct in vitro exposure of chronic stress-derived BMAds to PEMF stimulation failed to alter their senescence marker expression (Fig. 4c), indicating that the effect is indirect and requires an intermediary cell type. We first considered the role of sensory nerve-derived Sema3A, but recombinant Sema3A treatment of chronic stress-derived BMAds did not affect their senescence status (Fig. 4d). We therefore hypothesized that the sympathetic nervous system mediates this effect. Conditioned medium (CM) collected from sympathetic neurons exposed to PEMF (PEMF-CM) exhibited a significantly attenuated capacity to induce senescence in naïve BMAds in vitro, compared to CM from control neurons (Ctrl-CM) (Fig. 4e). To identify the responsible factor, we performed a comprehensive proteomic analysis (Astral mass spectrometry with DIA) on the supernatant of cultured sympathetic neurons from stressed-mice, with or without PEMF treatment (Fig. 4f). Gene Ontology enrichment analysis of differentially expressed proteins highlighted pathways related to synaptic signaling and neuronal projection (Fig. 4g). Among the most significantly downregulated candidates was NPY, a canonical sympathetic neurotransmitter (Fig. 4h). This reduction was validated at both the protein level by ELISA (Fig. 4l) and the transcriptional level by qPCR (Fig. 4i). To establish causality, we performed a rescue experiment. The addition of recombinant NPY (rNPY) to PEMFs-CM restored its pro-senescent activity on BMAds (Fig. 4j, k). Conversely, direct treatment of BMAds with rNPY alone was sufficient to upregulate Cdkn1a and Cdkn2a expression, recapitulating the senescence phenotype (Fig. 4m, n). Finally, to definitively establish the pathological relevance of NPY-induced senescent BMAds in vivo, we performed a BMAd transplantation assay. Transplantation of BMAds from rNPY-treated donor mice (NPY-WT), but not from PBS-treated controls (PBS-WT), induced significant bone loss in healthy young recipient mice, as evidenced by decreased BV/TV and Tb.N, and increased Tb.Sp (Supplementary Fig. 3a–c). This bone loss was accompanied by a marked expansion of the marrow adipocyte area and a severe suppression of bone formation, as evidenced by reduced calcein labeling and MAR (Supplementary Fig. 3d–g). These results demonstrate that reduced secretion of NPY from sympathetic neurons is both necessary and sufficient for PEMFs to alleviate BMAd senescence.

Fig. 4. PEMFs indirectly inhibit the senescence of BMAds through NPY secreted by the sympathetic nervous system.

a Schematic of bone marrow adipocyte (BMAd) isolation from four groups: Control (Conn), Chronic Stress (Stress), Conn+pulsed electromagnetic fields (PEMF), and Stress+PEMF. Created in BioRender. Chen, S. (2026) https://BioRender.com/v4yoa87b Quantitative PCR (qPCR) analysis of senescence markers Cdkn1a and Cdkn2a in BMAds isolated in vivo. c, d qPCR of Cdkn1a and Cdkn2a in BMAds from stressed mice treated ± PEMF or semaphorin 3A (Sema3A) in vitro. e qPCR of Cdkn1a and Cdkn2a in BMAds in vitro. f A schematic timeline of sympathetic neuron isolation and culture from mice with unpredictable chronic mild stress ± PEMF. Created in BioRender. Chen, S. (2026) https://BioRender.com/402rktqg GO enrichment analysis of the mouse sympathetic secretome indicates that differentially expressed proteins primarily associate with synaptic and axonal functions. h Heatmap of differentially expressed proteins in supernatants of Non-PEMF vs PEMF sympathetic neurons. i, l qPCR and ELISA analysis of neuropeptide Y (Npy). j, k qPCR of Cdkn1a and Cdkn2a in BMAds treated with conditioned medium (CM) from control (Conn) or PEMF sympathetic neurons, ± recombinant neuropeptide Y (rNPY). m, n qPCR of Cdkn1a and Cdkn2a in BMAds treated with bovine serum albumin (BSA) or rNPY. All panels: n = 3 per group from three independent experiments, except (g, h). Data are represented as mean ± SD, and statistical analyses were conducted with a 95% confidence interval. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by two-way ANOVA with Tukey’s multiple comparisons for (b)(e)(j)(k) and two-tailed unpaired Student’s t-test for (c, d, i, l–n). In vitro experiments were repeated 3 times. P value: b Cdkn1a: P < 0.0001 (Conn vs Stress), P < 0.0001 (Stress vs Stress+PEMFs). Cdkn2a: P < 0.0001 (Conn vs Stress), P < 0.0001 (Stress vs Stress+PEMFs). c Cdkn1a: P = 0.231 (Conn vs PEMFs). Cdkn2a: P = 0.6781 (Conn vs PEMFs). d Cdkn1a: P = 0.8712 (Conn vs Sema3a). Cdkn2a: P = 0.7618 (Conn vs Sema3a). e Cdkn1a: P < 0.0001 (Conn vs Stress), P < 0.0001 (Stress vs Stress+PEMFs). Cdkn2a: P < 0.0001 (Conn vs Stress), P < 0.0001 (Stress vs Stress+PEMFs). i P = 0.0058 (Stress vs Stress+PEMFs). j P = 0.0053 (Conn vs PEMFs), P = 0.0007 (PEMFs vs PEMFs+rNPY). k P < 0.0001 (Conn vs PEMFs), P < 0.0001 (PEMFs vs PEMFs+rNPY). l P < 0.0001 (Stress vs Stress+PEMFs). m P = 0.0006 (BSA vs rNPY). n P < 0.0001 (BSA vs rNPY). Source data are provided as a Source Data file.

PEMFs decrease NPY secretion from sympathetic nerves via sensory nerve-derived Sema3A

Our previous data established that PEMFs act via sensory nerves to produce local Sema3A and suppress BMAd senescence. We next investigated whether sensory nerves transduce the PEMF signal to regulate local sympathetic nerve activity directly within the bone microenvironment. Consistent with the suppression of sympathetic tone (Figs. 2g and 3g, h), PEMF treatment significantly reduced the local concentration of the sympathetic neurotransmitter NPY in the bone marrow of WT mice subjected to unpredictable chronic mild stress (Supplementary Fig. 4a). Strikingly, this effect was completely abolished in sensory nerve-ablated (iDTR;Advillin^Cre) mice, where NPY levels remained high and indistinguishable from sham-treated controls (Supplementary Fig. 4c). In contrast, serum NPY levels were unaffected by PEMF treatment in either genotype (Supplementary Fig. 4b, d). These in vivo findings demonstrate that the PEMF-mediated reduction of bone marrow NPY is locally restricted and strictly dependent on intact sensory innervation. Since both Sema3A (Fig. 3k, m) and NPY (Supplementary Fig. 4a, c) were specifically modulated in the bone marrow but not in the serum, we hypothesized that sensory nerves might directly regulate sympathetic nerve terminals within the local bone niche. To test this, we first assessed the direct effect of PEMFs on sympathetic neurons in vitro. Cultured sympathetic neurons extracted from stressed mice showed no change in NPY secretion upon direct PEMF exposure (Supplementary Fig. 4e, f), confirming that PEMFs require an intermediary cell type. We therefore posited that sensory neurons are this essential intermediary. Indeed, co-culturing sympathetic neurons with conditioned medium from dorsal root ganglia (DRGs) that had been exposed to PEMFs (PEMF-DRG-CM) significantly reduced sympathetic neuronal NPY secretion compared to co-culture with control DRG medium (Ctrl-DRG-CM) (Supplementary Fig. 4e, f). Crucially, the addition of a neutralizing anti-Sema3A antibody to the co-culture system completely abolished this effect. Similarly, using DRG neurons isolated from Sema3a-knockout mice (Advillecre:Sema3aflox) for co-culture with sympathetic neurons also negated the beneficial effect of PEMFs (Supplementary Fig. 4e, f). This establishes that Sema3A, secreted by PEMF-stimulated sensory neurons, is necessary and sufficient to suppress NPY production in sympathetic neurons.

To definitively establish the peripheral autonomy of this circuit, we performed complete spinal cord transection (SCT) at the T10 level, physically severing all supraspinal connections to the hindlimb. SCT itself induced a peripheral pathology akin to chronic stress, characterized by sympathetic hyperactivation (↑NPY, ↑TH⁺ fibers) and sensory impairment (↓Sema3A, ↓CGRP⁺ fibers) (Supplementary Fig. 5a–f). Most critically, PEMF treatment in SCT mice remained fully effective, robustly suppressing sympathetic tone, activating sensory nerves, and reducing the burden of senescent bone marrow adipocytes (Supplementary Fig. 5a–g). Supporting the feasibility of this direct local cross-talk, immunohistochemical analysis revealed close anatomical proximity between sensory (CGRP⁺, red) and sympathetic (TH⁺, green) nerve fibers within the femoral bone marrow (Supplementary Fig. 4g, white arrows). 3D imaging of cleared bone further demonstrated this close spatial relationship between the two nerve populations (Supplementary Fig. 4h and Supplementary Movie 1). Together, these data delineate a local neuro-to-neuro regulatory circuit: PEMF stimulation is sensed by intraosseous sensory nerves, prompting them to secrete Sema3A. Sema3A then acts directly on nearby sympathetic nerve terminals to suppress their release of NPY, thereby completing the pathway that alleviates BMAd senescence and promotes bone formation.

Genetic ablation of sympathetic NPY confers resistance to stress-induced osteoporosis and is a prerequisite for PEMFs efficacy

To establish a direct causal role for sympathetic neuron-derived NPY in stress-induced osteoporosis, we utilized Th^Cre; NPY^flox/flox(Npy ^th) mice, which lack NPY specifically in tyrosine hydroxylase-expressing sympathetic neurons. Strikingly, unlike Npy^wt mice, Npy^th mice were protected from chronic stress-induced bone loss, exhibiting significantly higher bone mass parameters (Fig. 5a–f). This genetic evidence unequivocally identifies sympathetic NPY as a critical driver of pathological bone loss.

Fig. 5. NPY in the sympathetic nervous system is a key driver of stress-induced osteoporosis and a necessary target for the action of PEMFs.

a Schematic illustration of the experimental procedure is shown in the panel. It was created in BioRender. Created in BioRender. Chen, S. (2026) https://BioRender.com/vqbljmg. b Representative micro-CT images of distal femoral cancellous bones. c–f Quantification of the trabecular cancellous bone volume fraction (BV/TV), trabecular cancellous bone number (Tb.N), trabecular cancellous bone thickness (Tb.Th), and trabecular cancellous separation (Tb.Sp). g Representative images of immunofluorescence staining of femur sections using antibodies against Perilipin (green) and Opn (red), representative images of calcein double-labeling and representative images of immunofluorescence staining of femur sections using antibodies against Perilipin (green) and γH2AX (red). h, i Quantitative analysis of the number of Perilipin⁺ cells per mm² of primary trabecular tissue and the relative areas of Opn (N. Perilipin⁺ cells/Ar, Quantitative relative area). j Quantification of mineral apposition rate (MAR). k Quantified numbers of Perilipin- and γH2AX-double positive cells per mm² tissue area (No. Perilipin⁺γH2AX⁺ cells) are shown. n = 6 biological replicates per group for all quantifications. Data are represented as mean ± SD, statistical analyses were conducted with a 95% confidence interval. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by two-way ANOVA with Tukey’s multiple comparisons for (c–f, h–k). P value: c P < 0.0001 (Npy^wt Sham vs Npy^wt PEMFs), P < 0.0001 (Npy^wt Sham vs Npy^th-/- Sham). d P = 0.0015 (Npy^wt Sham vs Npy^wt PEMFs), P = 0.0002 (Npy^wt Sham vs Npy^th-/- Sham). e P = 0.0007 (Npy^wt Sham vs Npy^wt PEMFs), P = 0.0021 (Npy^wt Sham vs Npy^th-/- Sham). f P < 0.0001 (Npy^wt Sham vs Npy^wt PEMFs), P < 0.0001 (Npy^wt Sham vs Npy^th-/- Sham). h P = 0.0143 (Npy^wt Sham vs Npy^wt PEMFs), P = 0.03 (Npy^wt Sham vs Npy^th-/- Sham). i P = 0.0132 (Npy^wt Sham vs Npy^wt PEMFs), P = 0.0446 (Npy^wt Sham vs Npy^th-/- Sham). j P = 0.0192 (Npy^wt Sham vs Npy^wt PEMFs), P = 0.0333 (Npy^wt Sham vs Npy^th-/- Sham). k P = 0.0024 (Npy^wt Sham vs Npy^wt PEMFs), P = 0.005 (Npy^wt Sham vs Npy^th-/- Sham). Source data are provided as a Source Data file.

We next investigated whether the osteoprotective effect of PEMFs requires the suppression of sympathetic NPY. In chronically stressed Npy^wt mice, PEMF treatment effectively rescued bone mass. In contrast, PEMFs provided no significant additional benefit to the already protected Npy^th (Fig. 5b–f). Histomorphometric analysis of the bone marrow microenvironment revealed that the intrinsic protection in Npy^th mice was associated with a shift toward osteogenesis, including a higher mineral apposition rate (MAR), a higher expression of Opn and a lower adiposity, as indicated by reduced number of Perilipin⁺ cells, compared to stressed Npy^wt mice (Fig. 5g–k). PEMFs treatment in Npy^wt mice recapitulated this healthy microenvironment by further enhancing osteogenesis and suppressing adipogenesis, but it had no significant effect in Npy^th mice (Fig. 5g–k). Furthermore, PEMFs treatment reversed BMAd senescence in chronic stress-treated Npy^wt mice, evidenced by reduced co-localization of γH2AX with Perilipin⁺ BMAds (Fig. 5g, k). These data demonstrate that the presence and pathological activity of sympathetic NPY are required for PEMFs to exert their therapeutic effect.

To dissect the NPY pathway from the canonical norepinephrine branch of the sympathetic response, we clamped β-adrenergic signaling using chronic isoproterenol (ISO) infusion. A critical finding emerged: Npy^th mice were fully resistant to ISO-induced bone loss, whereas wild-type mice developed severe osteopenia (Supplementary Fig. 6a–e). Consequently, PEMFs treatment rescued bone mass in ISO-infused wild-type mice but, as expected, provided no additional benefit to the resistant Npy^th mice (Supplementary Fig. 6a–e). Immunofluorescence analysis confirmed that ISO-induced accumulation of Perilipin⁺ marrow adipocytes and suppression of Opn expression in wild-type mice were mitigated by PEMFs, but these parameters remained unchanged in Npy^th mice regardless of PEMFs exposure (Supplementary Fig. 6f–j).

Collectively, these data demonstrate that sympathetic NPY is the dominant pathological effector in stress-induced osteoporosis. PEMFs require the presence of this NPY pathway to execute its bone-protective function, likely by specifically suppressing its activity.

Adipocyte YIR signaling is required for PEMFs-induced bone anabolism under chronic stress

Our previous findings identified NPY as a key pathological factor released by sympathetic nerves under chronic stress. Given that NPY exerts its biological effects predominantly through the Y1 receptor (Y1R), we hypothesized that Y1R expression might be modulated in the bone marrow microenvironment under stress conditions. To precisely identify the cellular source, we performed double immunofluorescence staining for Y1R together with the adipocyte marker Perilipin on bone sections. As shown in Supplementary Fig. 7a, the Y1R signal was clearly localized to Perilipin-positive BMAds after stress, confirming BMAds as the relevant cell type. To establish a causal role of adipocyte Y1R signaling in stress-induced bone pathology and to test whether it constitutes a non-redundant pathway for PEMF efficacy, we utilized Adipoq^Cre;Y1R^fl/fl mice (Y1R-Adipoq cKO), which lack Y1R specifically in adipocytes, and their Y1R^fl/fl littermates as controls. Mice were subjected to chronic stress, with or without PEMFs treatment. Micro-CT analysis revealed that adipocyte Y1R is essential for the skeletal response to chronic stress. In control (Y1R^fl/fl) mice, chronic stress induced significant trabecular bone loss, as evidenced by decreased BV/TV and Tb.N (Fig. 6a, b). Strikingly, adipocyte-specific ablation of Y1R rendered mice fully resistant to chronic stress-induced bone loss, with bone parameters indistinguishable from non-stressed controls (Fig. 6a, b). This complete phenotypic rescue establishes Y1R signaling in BMAds as the principal mediator of stress-induced skeletal deterioration. In control mice, PEMFs treatment effectively rescued chronic stress-induced bone loss (Fig. 6a, b). However, in Y1R-Adipoq cKO mice, which were already protected from stress-induced bone loss, PEMFs treatment did not result in any further increase in bone mass (Fig. 6a, b). Dynamic histomorphometry using calcein double-labeling demonstrated that chronic stress suppressed the MAR in control mice, and this was restored by PEMFs treatment (Fig. 6c, g). In Y1R-Adipoq cKO mice, MAR was maintained at control levels even under stress, and PEMFs treatment did not further increase it (Fig. 6c, g). Similarly, histological analysis of the bone marrow revealed that Y1R-Adipoq cKO mice were resistant to stress-induced expansion of Perilipin⁺ marrow adipocytes and exhibited normalized Opn expression (Fig. 6d–f). PEMFs treatment reduced adiposity and increased Opn expression in stressed control mice, but did not produce additional changes in the already-protected Y1R-Adipoq cKO mice (Fig. 6d–f). Furthermore, the stress-induced increase in γH2AX co-localization with Perilipin⁺ BMAds—a marker of cellular senescence—was abrogated in Y1R-Adipoq cKO mice, and PEMFs treatment did not further reduce it (Fig. 6i, h). We measured Y1R fluorescence intensity in individual BMAds. Chronic stress significantly increased the Y1R signal, while PEMF treatment did not significantly alter the Y1R intensity compared to the stress group (Supplementary Fig. 7b). Then, we isolated BMAds and analyzed Y1R mRNA. Results were consistent: Chronic stress upregulated Y1R, and PEMF did not reduce its expression (Supplementary Fig. 7c). Taken together, these results indicate that chronic stress upregulates Y1R expression in BMAds, but PEMF does not significantly reverse Y1R expression within BMAds themselves. To further exclude a direct effect of PEMF on adipocyte Y1R expression, we performed an additional in vitro experiment using isolated BMAds from stressed mice. Consistent with our in vivo findings, direct PEMF exposure did not alter Y1R expression (Supplementary Fig. 7d), further supporting that the effect is not cell-autonomous. Taken together, our genetic and pharmacological data converge on a single mechanism: chronic stress-induced bone loss is mediated by NPY-Y1R signaling in BMAds, and PEMF therapy functions by interrupting this specific pathway, likely via suppression of sympathetic NPY release rather than direct modulation of Y1R expression.

Fig. 6. Adipocyte-specific Y1R is an absolutely essential target for the bone-protective and anti-adipogenic effects of PEMFs.

a A representative micro-CT image of the distal femoral cancellous bones. b Quantification of the trabecular cancellous bone volume fraction (BV/TV), trabecular cancellous bone number (Tb.N), trabecular cancellous bone thickness (Tb.Th), and trabecular cancellous separation (Tb.Sp). c Representative images of calcein double-labeling. d Representative images of immunofluorescence staining of femur sections using antibodies against Perilipin (green) and Opn (red). DAPI stains nuclei blue. e, f Quantitative analysis of the number of Perilipin-positive cells per mm² tissue area and the relative areas of Opn (N. Perilipin⁺ cells/Ar, Quantitative relative area). g Quantification of mineral apposition rate (MAR). h Quantified numbers of Perilipin - and γH2AX-double positive cells per mm² tissue area (No. Perilipin⁺γH2AX⁺ cells) are shown. i Representative images of immunofluorescence staining of femur sections using antibodies against Perilipin (green) and γH2AX(red). Data are represented as mean ± SD with n = 6 biological replicates per group for all quantifications, and statistical analyses were conducted with a 95% confidence interval. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 by two-way ANOVA with Tukey’s multiple comparisons for (b, e–h). P value: b BV/TV: P < 0.0001 (Y1R^wt sham vs Y1R^wt PEMFs), P < 0.0001 (Y1R^wt sham vs Y1R^adipoq sham). Tb.N: P = 0.0053 (Y1R^wt sham vs Y1R^wt PEMFs), P = 0.0004 (Y1R^wt sham vs Y1R^adipoq sham). Tb.Th: P = 0.0123 (Y1R^wt sham vs Y1R^wt PEMFs), P = 0.0407 (Y1R^wt sham vs Y1R^adipoq sham). Tb.Sp: P < 0.0001 (Y1R^wt sham vs Y1R^wt PEMFs), P < 0.0001 (Y1R^wt sham vs Y1R^adipoq sham). e P = 0.0327 (Y1R^wt sham vs Y1R^wt PEMFs), P = 0.0389 (Y1R^wt sham vs Y1R^adipoq sham). f P < 0.0001 (Y1R^wt sham vs Y1R^wt PEMFs), P < 0.0001 (Y1R^wt sham vs Y1R^adipoq sham). g P = 0.0477 (Y1R^wt sham vs Y1R^wt PEMFs), P = 0.0148 (Y1R^wt sham vs Y1R^adipoq sham). h P = 0.0015 (Y1R^wt sham vs Y1R^wt PEMFs), P = 0.0023 (Y1R^wt sham vs Y1R^adipoq sham). Source data are provided as a Source Data file.

To directly test whether Y1R signaling is required for PEMF’s osteoprotective effect, we performed a pharmacological gain-of-function experiment using the selective Y1R agonist [Leu³¹, Pro³⁴]-NPY. This approach avoids the baseline protection issue inherent to the cKO model. As shown in the new Supplementary Fig. 8a–g: Y1R agonist alone exacerbated chronic stress-induced bone loss, confirming that the agonist effectively activates the catabolic Y1R pathway. PEMF alone rescued chronic stress-induced bone loss. Co-administration of the Y1R agonist completely abrogated PEMF’s osteoprotective effect, resulting in bone parameters and senescent BMAd burden comparable to the stress+agonist group. Critically, PEMF failed to improve bone mass in the presence of the agonist, indicating that PEMF cannot overcome direct Y1R activation. This pharmacological epistasis experiment demonstrates that constitutive activation of Y1R is sufficient to abrogate PEMF’s osteoprotective effect, positioning Y1R as an obligatory downstream effector in the PEMF signaling pathway. Collectively, these genetic evidences demonstrate that Y1R signaling in adipocytes is not merely involved but is absolutely required for PEMFs to exert their bone anabolic and anti-adipogenic effects in the context of chronic stress.

To determine whether NPY acts directly on BMAds to induce senescence and to explore the downstream signaling mechanism, we treated primary BMAds with recombinant NPY. As NPY can signal through multiple receptors, we used the selective Y1R antagonist BIBO3304 to assess the specificity of the response. Treatment of BMAds with NPY (100 nM, 48 h) significantly increased the senescent cells compared to vehicle control, demonstrated by p21 immunostaining (Supplementary Fig. 9a). This pro-senescent effect was completely abolished by pre-treatment with the Y1R antagonist BIBO3304 (1 μM) (Supplementary Fig. 9a), demonstrating that NPY induces senescence specifically through Y1R activation. Given our in vivo observation of increased γH2AX⁺ BMAds in stressed mice (Fig. 2l–n), we investigated whether NPY triggers DNA damage. Immunofluorescence analysis revealed that NPY treatment induced pronounced formation of γH2AX foci in BMAds, a hallmark of DNA double-strand breaks (Supplementary Fig. 9b). Again, this effect was blocked by Y1R antagonism (Supplementary Fig. 9b), confirming that DNA damage is a specific downstream event of Y1R activation. At the transcriptional level, NPY treatment markedly upregulated the mRNA expression of the cell cycle inhibitors Cdkn1a and Cdkn2a (Supplementary Fig. 9c, d). This upregulation was prevented by BIBO3304 (Supplementary Fig. 9c, d), linking Y1R activation to the core genetic program of cellular senescence. Collectively, these in vitro data define a direct and sequential pathogenic axis: NPY, released from sympathetic nerves, acts directly on BMAds via Y1R to trigger DNA damage (γH2AX), which in turn drives the expression of p16/p21 and executes the senescent program. This mechanism provides a cellular and molecular basis for the accumulation of senescent BMAds observed in our chronic stress model in vivo.

To explore the translational relevance of our findings, we examined SEMA3A-NPY-Y1R signaling in human osteoporotic bone tissue. Analysis of the publicly available transcriptomic dataset GSE230665, which contains gene expression profiles of femoral bone tissue from 12 postmenopausal osteoporosis patients and 3 healthy postmenopausal controls⁴⁰, revealed that Y1R expression was significantly elevated in osteoporotic bone compared to controls (P = 0.0028). NPY also showed a trend toward upregulation (median expression: 6.95 in controls vs 13.2 in osteoporosis), while SEMA3A exhibited a trend toward downregulation (median expression: 6.99 in controls vs 5.25 in osteoporosis). Together, these human data support the pathological relevance of the SEMA3A-NPY-Y1R signaling axis in skeletal disease and suggest that the molecular mechanism identified in our chronic stress model is conserved in human osteoporotic bone (Supplementary Fig. 9e).

Discussion

This work establishes a novel peripheral neural circuit within bone, revealing a previously unrecognized local mechanism for PEMF therapy. We demonstrate that PEMFs do not act directly on bone cells but orchestrate osteoprotection by activating sensory nerves, which in turn directly suppress sympathetic NPY release within the bone marrow microenvironment. This local, nerve-mediated mechanism alleviates BMAd senescence via the Y1R, thereby rescuing bone formation (Fig. 7). Our work identifies sympathetic NPY as a paramount pathological effector in stress-induced bone loss and positions the adipocyte NPY-Y1R axis as a precise, druggable target for neuromodulatory therapy.

Fig. 7. Schematic illustration of the neural circuitry underlying chronic stress-induced osteoporosis and its modulation by PEMFs.

Chronic stress leads to overactivation of the sympathetic nervous system, resulting in elevated release of NPY from sympathetic nerve terminals within the bone marrow. NPY binds to its receptor Y1R on BMAds, promoting their cellular senescence. This senescent BMAd phenotype contributes to a pathological bone marrow microenvironment and impaired osteogenesis. PEMF treatment targets skeletal sensory nerves. Upon PEMF stimulation, sensory nerves are activated and locally secrete Sema3A. This Sema3A signaling suppresses sympathetic NPY release, thereby alleviating BMAd senescence and restoring osteogenic potential. Key symbols are defined in the figure legend (left panel). Created in BioRender. Chen, S. (2026) https://BioRender.com/f1zouuq.

The most salient finding of our study is the identification of NPY—distinct from the classical neurotransmitter norepinephrine (NE)—as the principal sympathetic mediator of bone catabolism under chronic stress. Although sympathetic regulation of bone remodeling via β2‑adrenergic receptor signaling is well‑established^41,42, its role in sustained pathological states appears more complex. Our data introduce a paradigm shift: mice with sympathetic neuron–specific deletion of Npy (Npy^th) were completely protected from osteoporosis induced by chronic β‑adrenergic stimulation (isoproterenol infusion) (Supplementary Fig. 6). This apparent functional dichotomy—where central NPY may exert protective effects on bone⁴³, whereas peripheral NPY drives bone loss—highlights a critical spatial context‑dependence of NPY action^31,44. Our work demonstrates that under pathological conditions, elevated skeletal NPY operates predominantly through Y1 receptors to promote bone loss. Thus, our work nominates the peripheral NPY–Y1R axis as a novel and precise therapeutic target for osteoporosis, offering a strategy that may circumvent the central side effects associated with broad sympathetic blockade.

The spatial and functional segregation of NPY systems provides a coherent framework to reconcile our skeletal findings with the established roles of NPY in affective behavior. The Npy^th model used here achieves selective deletion in peripheral sympathetic neurons, leaving central NPY circuits—which are primary regulators of anxiety and depression—largely intact⁴⁵. It is therefore plausible that the profound osteoprotection observed in these mice could be dissociated from baseline alterations in classic anxiety‑like behaviors, as these are governed by distinct, brain‑derived NPY pools^46,47. Although the present study does not directly test affective behaviors, this functional dissociation suggests that the peripheral sympathetic NPY–Y1R axis constitutes a promising therapeutic target for bone loss, one that could be modulated without directly perturbing central emotional homeostasis. Future studies that integrate circuit‑specific manipulations with comprehensive behavioral and autonomic phenotyping will be essential to definitively map the crosstalk between peripheral sympathetic tone, bone integrity, and affective state.

Beyond the identity of the sympathetic signal, we sought to define its pathological impact on the bone marrow microenvironment. Cellular senescence is a key driver of osteoporosis^48–52. Paradoxically, although BMAds constitute a substantial portion of the bone marrow microenvironment⁵³, the specific role of senescent BMAds in bone homeostasis has remained elusive. Recent studies posit that senescent BMAds can initiate bone loss under certain pathologies, mirroring the phenomenon in peripheral fat where senescent adipocytes emerge early with aging and propagate senescence signals^54,55. Building on this, our study identifies chronic stress as a potent inducer of BMAd senescence, which subsequently disrupts bone remodeling. This work expands the functional repertoire of BMAds beyond well-established roles in adipogenic differentiation and lipotoxicity^48,56,57 by delineating a novel role for BMAd senescence in stress-induced osteoporosis and its reversal by PEMF therapy.

A pivotal mechanistic advance offered by our study is the elucidation of NPY as a direct inducer of BMAd senescence, creating a critical link between neural dysregulation and the deterioration of the bone marrow microenvironment. We substantiate this causality through a multi-tiered experimental approach: direct application of recombinant NPY was sufficient to trigger a senescent phenotype in BMAds in vitro (Fig. 4m, n); conditioned medium from PEMF-treated sympathetic neurons, which contained reduced levels of NPY, lost its pro-senescent activity (Fig. 4e, j, k); and most decisively, the transplantation of BMAds that had been pre-exposed to NPY in vivo was capable of inducing an osteoporotic phenotype in healthy young recipient mice (Supplementary Fig. 3). These results firmly establish that NPY does not merely correlate with but actively drives the pathological process through the induction of a senescent, dysfunctional adipose compartment. By directly linking NPY to cellular senescence in BMAds, our work bridges two major paradigms in bone biology—neuro-osteogenic signaling and the aging microenvironment—providing a unified mechanism for stress-induced bone loss. The accumulation of senescent BMAds, with their associated SASP, likely creates an inflammatory and anti-osteogenic milieu that is detrimental to bone health. Genetic ablation of Y1R specifically in adipocytes prevents stress-induced bone loss, demonstrating that adipocytic Y1R signaling is required for the development of osteoporosis under chronic stress. Crucially, we have delineated the direct pathogenic axis through which NPY acts: sympathetic-derived NPY signals through the adipocyte Y1R to trigger DNA damage (γH2AX), which in turn drives the expression of p16/p21, executing the senescence program in BMAds.

To directly test whether suppression of Y1R signaling is required for PEMF’s osteoprotective effect, we performed a pharmacological gain-of-function experiment using the selective Y1R agonist, [Leu³¹, Pro³⁴]-NPY. This approach bypasses the confounding effects of baseline protection inherent to knockout models. Co-administration of the Y1R agonist completely abrogated the osteoprotective effect of PEMF, resulting in bone parameters indistinguishable from those in the chronic stress plus agonist group. Crucially, PEMF failed to improve bone mass in the presence of sustained Y1R activation, indicating that its therapeutic action cannot overcome direct receptor stimulation. These data provide definitive evidence that functional antagonism of Y1R signaling is a necessary downstream event for PEMF efficacy.

Thus, PEMFs promote osteogenesis indirectly, by normalizing the bone marrow niche—specifically, by preventing NPY-Y1R-driven BMAd senescence. This paradigm aligns with the growing recognition of bone marrow adipose tissue as a dynamic endocrine regulator of bone metabolism. The specificity of this mechanism—wherein PEMFs require the presence of Y1R on the target cell but act by suppressing its NPY—illustrates the precision of neuromodulation compared to conventional receptor-targeted pharmacology.

A central and novel insight from this study is the discovery of a peripheral sensory-sympathetic circuit that operates autonomously within the bone microenvironment. We demonstrate that PEMF activation of sensory nerves leads to the local secretion of Sema3A, which directly suppresses NPY release from sympathetic nerve terminals. This mechanism is supported by four lines of evidence: (1) genetic ablation of sensory neurons abolished PEMF’s effect on NPY and bone mass; (2) in vitro co-culture experiments confirmed that Sema3A from PEMF-stimulated sensory neurons directly inhibits sympathetic NPY secretion; and neutralization of Sema3A blocked this effect; (3) In mice with SCT—which physically severs all supraspinal input—PEMF treatment remained fully effective, still suppressing bone marrow sympathetic tone (↓TH) and activating sensory nerves (↑CGRP, Sema3A); (4) PEMF’s osteoprotection occurred without normalizing the elevated systemic stress markers (serum corticosterone) or improving central anxiety-like behaviors, dissociating its effect from global stress alleviation. This circuit constitutes a form of ‘peripheral interoception’, in which sensory nerves directly perceive the PEMF stimulus and, through a local signaling molecule (Sema3A), fine-tune the activity of neighboring sympathetic nerves to maintain tissue homeostasis—independent of central nervous system integration. This finding challenges the conventional view that communication between sensory and sympathetic systems requires central relay and reveals a rapid, tissue-autonomous regulatory mechanism. While genetic interrogation of central pathways remains a future direction, our combined use of SCT, unchanged systemic markers, and efficacy in peripheral models provides compelling, multi-faceted evidence for a CNS-independent, autonomous peripheral circuit.

A critical question arising from our study is whether the neuro-osteogenic circuit identified in mice is conserved and pathologically relevant in human skeletal disease. Our analysis of human transcriptomic data, combined with established clinical evidence, strongly suggests that it is. First, re-analysis of a public dataset (GSE230665) from human osteoporotic bone revealed a significant upregulation of Y1R (P = 0.0028), with concordant trends for increased NPY and decreased SEMA3A expression. This genetic signature directly mirrors the dysregulation we observed in our murine chronic stress model. The upregulation of the NPY-Y1R axis in human disease is further corroborated by histopathological studies showing elevated NPY and Y1R protein in osteoporotic bone, where their levels correlate negatively with bone mass⁵⁸. Second, and conversely, the protective arm of this circuit—Sema3A—is impaired in human osteoporosis. Clinical studies consistently report that circulating levels of Sema3A are reduced in patients with postmenopausal osteoporosis, and its concentration positively correlates with bone mineral density^23,59. Although serum levels may not directly reflect the bone marrow microenvironment, this clinical association positions Sema3A as a key osteoprotective factor in humans, aligning with its role in our model. The convergence of these human observations—heightened pathological signaling (NPY/Y1R↑) and diminished protective signaling (Sema3A↓)—provides compelling translational validation for the sensory-sympathetic balance we have delineated. It suggests that the disruption of this local neural homeostasis is not an artifact of the mouse model but a conserved pathophysiological feature of osteoporosis.

While PEMFs ameliorate various forms of osteoporosis, their mechanism of action has remained elusive^25,36,38. Here, we define the essential components of a peripheral interoceptive circuit and propose that PEMF acts as a physiological rheostat, recalibrating local neural balance within the bone microenvironment. Strikingly, PEMF exhibits context-dependent, bidirectional regulation of sympathetic tone. In aged bone, where sympathetic activity is pathologically diminished, PEMF treatment restores it²⁵. Conversely, in chronic stress, where sympathetic tone is pathologically elevated, PEMF normalizes it (this study). This bidirectional modulation is exemplified by its effects on TH⁺ sympathetic fibers: PEMF prevents their age-related decline and suppresses their stress-induced hyperinnervation. In both contexts, PEMF restores sympathetic nerve density toward a physiological setpoint. Thus, PEMF is not a simple stimulator or inhibitor, but a homeostatic modulator that harnesses innate neural circuitry for state-dependent therapy, aligning with its ability to restore physiological equilibria^60–62. The biophysical trigger for this neuromodulation likely involves direct sensory neuron activation. Studies on DRG neurons show that PEMF exposure rapidly increases intracellular calcium ([Ca²⁺]i) via L-type voltage-gated calcium channels (L-VGCCs), a primary trigger for excitation^63,64. Consistently, we previously identified CGRP⁺ peptidergic sensory fibers as the key PEMF-responsive subtype in bone²⁵. Activation of these sensory neurons leads to activity-dependent Sema3A release, completing the circuit. Seminal work has established that Sema3A secretion is tightly coupled to neuronal activity, involving calcium influx and SNARE-dependent exocytosis^65,66. Therefore, we propose that PEMF acts by depolarizing sensory neurons (presumably through L-VGCCs), which induces Sema3A secretion into the bone marrow. Secreted Sema3A then suppresses sympathetic NPY release, ultimately inhibiting NPY/Y1R-driven BMAd senescence and rescuing bone mass.

While our data establish a novel peripheral neural circuit and a promising neuromodulatory strategy, several important questions remain. First, the precise molecular “sensor” within sensory nerves that transduces PEMF into neuronal activation awaits identification. Uncovering this initial biophysical step is essential to connect external stimuli to intracellular neurobiological signaling. Second, the specific receptor(s) on sympathetic neurons that mediate Sema3A’s suppression of NPY release warrant definitive characterization. Elucidating these upstream (PEMF sensor) and downstream (Sema3A receptor) molecular nodes will refine our model and may reveal more specific druggable targets. Third, and critically, our study was conducted exclusively in male mice to control for the profound confounding effects of sex hormones on both stress responses and bone metabolism. Significant sex differences in neuroendocrine pathways and skeletal physiology are well-documented. Therefore, future investigations must include both sexes to determine the generalizability of this circuit and the potential for sex-specific therapeutic outcomes, a crucial step for translational relevance.

In conclusion, our findings transform the perception of PEMF therapy from a generic osteogenic stimulus to a precise neuromodulatory strategy that harnesses a local peripheral neural circuit within bone. We have mapped a complete pathway from a non-invasive physical stimulus to a specific neural interaction and its molecular effector, resulting in the reversal of a complex pathology. By identifying the sympathetic NPY-Y1R axis on BMAds as a critical leverage point and revealing Sema3A as the key local mediator, this work opens new avenues for treating not only stress-related osteoporosis but potentially other conditions characterized by sympathetic dysregulation and cellular senescence. It underscores the immense potential of targeting local neural circuits to achieve highly specific therapeutic outcomes with minimal invasiveness and side effects.

Methods

Animals and drug treatment

Advillin^Cre, iDTR, Th^cre, Npy^fl/fl, Adipoq^Cre, Y1R^fl/fl mouse lines were purchased from Cyagen Biosciences (Guangzhou, China) and Shanghai Model Organisms Center, Inc. C57BL/6 mice were purchased from BEIJING HFK BIOSCIENCE CO, LTD Experimental Animals (Beijing, China). All these mice were housed in specific pathogen-free conditions and analyzed with a mixed background. The genotypes of the mice were measured by PCR analyses of genomic DNA, which was extracted from mouse tails. The mice were housed in a clean room with free access to water, in a humidity-controlled environment, and subjected to a 12-h light/dark cycle (8 a.m. to 8 p.m.). They were maintained on a standard chow diet (25043213) at a constant temperature of 22–26 °C. 2-month-old male transgenic and C57BL/6 mice were subjected to either chronic stress and PEMF treatment or a control condition.

Serum samples were obtained by centrifugation at 3000 × g for 15 min and subsequently stored at −80 °C until further analysis. Additionally, femurs, tibias, heart, liver, spleen, kidney tissues, and urine were harvested from the mice.

The drugs and compounds used in this study are as follows: diphtheria toxin (DTX, Sigma-Aldrich, D0564); Recombinant mouse Sema3A Fc chimera protein (R&D Systems, 5926-S3). Isoproterenol (Sigma) was injected intraperitoneally.

Unpredictable chronic mild stress model

Our chronic stress model was established using an unpredictable chronic mild stress procedure. Eight-week-old male C57BL/6J mice were subjected to an 8-week unpredictable chronic mild stress model protocol, as adapted with modifications from²⁷. The regimen consisted of a randomized daily sequence of three distinct environmental stressors: (a) restraint stress for 2 h in a confined tube (3 cm × 5 cm × 7 cm); (b) cage dampening for 6 h, where bedding was moistened to create a humid environment without standing water; and (c) elevated platform stress for 2 h on a 100-cm high platform. The order of stressors was counterbalanced weekly to minimize habituation and to ensure each mouse experienced an equal number of exposures to each stressor type over the entire study period. The successful induction of a chronic stress phenotype, characterized by anxiety-like behaviors, was validated by a battery of established behavioral tests.

Behavioral analysis

Prior to behavioral testing, all mice were handled for 15–30 min per day over three consecutive days to minimize stress associated with human contact. Anxiety-like behaviors were assessed using two well-established paradigms: the elevated plus maze and the open field test. The elevated plus maze apparatus was constructed of white plastic and elevated 65 cm above the floor. It consisted of a central platform (5 × 5 cm) from which two open arms (25 × 5 cm) and two enclosed arms (25 × 5 cm, with 15-cm-high walls) extended opposite to each other, forming a plus shape. Each mouse was individually placed in the center of the maze facing a closed arm. Behavior was recorded for 5 min using a camera mounted directly above the apparatus. The number of entries into and the time spent in the open and closed arms were quantified. For the open field test, mice were placed in a white plastic square arena (50 × 50 cm) for a 5-min session. The arena was conceptually divided into a central zone (25 × 25 cm) and a peripheral zone for analysis. Each mouse was introduced to the arena at a point within the peripheral zone. All trials were video-recorded, and behavioral parameters were automatically analyzed using the ANY-maze video tracking system (Stoelting, USA).

PEMFs treatment

The magnetic field device used in this study was a custom-designed PEMFs exposure system, as previously described^25,38. The system comprised a signal generator and a Helmholtz coil assembly with a three-coil array. Mice in the PEMFs group received daily 1-h exposures for 4 weeks, with parameters set at 8 Hz frequency, 3.8 mT intensity, and a square-shaped pulsed burst waveform (burst width: 62.5 ms; pulse width: 62.5 ms; duty cycle: 50%). These parameters were selected based on our prior research demonstrating their osteogenic efficacy^25,38. Ambient temperature was maintained at 24 ± 2 °C using an infrared thermometer, and a Gauss meter confirmed stable magnetic field intensity throughout all experiments. Sham group mice were placed in the exposure chamber without a magnetic field output for the same duration. For in vitro studies, cells were exposed to identical parameters (8 Hz, 3.8 mT) for 3 h daily within a cell incubator.

Micro-computed tomography (micro-CT)

Micro-CT analysis of left distal femurs was performed using an NMC-100 scanner (PINGSHENG, Shanghai, China) with the following acquisition parameters: 70 kV, 114 mA, and 700 ms integration time. A total of 100 consecutive micro-CT images were acquired with a 10-μm layer spacing. The region of interest (ROI) was defined as the trabecular bone region extending from 0.6 mm to 1.5 mm distal to the growth plate. For 3D reconstruction and analysis, Avatar software was used to reconstruct the complete femoral metaphyseal structure. The trabecular bone ROI was delineated starting from 5% of the femur length proximal to the distal metaphyseal growth plate and extending proximally for an additional 5% of the total femur length. The following trabecular bone parameters were quantified: bone volume fraction (BV/TV), trabecular thickness (Tb.Th), trabecular number (Tb.N), and trabecular separation (Tb.Sp).

Immunohistochemistry and immunofluorescence

Femoral bones were dissected and immediately fixed in 4% paraformaldehyde overnight, followed by decalcification in 0.5 M EDTA (pH 7.4) for 14 days. The samples were then dehydrated in 20% sucrose plus 2% polyvinylpyrrolidone solution for 24 h and embedded in OCT compound (Sakura Finetek, Torrance, CA, USA). Ten μm-thick coronal sections of the femurs were prepared for SA-βGal staining. Senescent cells were detected using a senescence βGal staining kit according to the manufacturer’s instructions (Cell Signaling Technology, Danvers, MA). For Oil Red O staining following SA-βGal staining, the slides were rinsed in PBS three times, placed in 100% propylene glycol for 2 min, and stained in filtered 0.5% Oil Red O solution in propylene glycol for 30 min. The slides were then transferred into an 85% propylene glycol solution for 1 min, rinsed in PBS three times, mounted, and sealed with nail polish. Representative images were acquired using a BX51 microscope (Olympus, Tokyo, Japan).

Tissue sections for immunohistochemistry were prepared as follows. Twenty-μm-thick coronal sections of femurs were cut on a cryostat microtome. All sections were incubated overnight at 4 °C with the following primary antibodies: Opn (R&D, AF808; 1:200), Perilipin (MilliporeSigma, P1873; 1:500; CST, 9349; 1:200), γH2AX (Cell Signaling, 20E3; 1:200), TH (Millipore Sigma, ab152; 1:100), TH (Abcam, ab137869; 1:100), CGRP (Abcam, ab81887, 1:50; ab36001, 1:100), YR1(sc-393192, 1:100). The samples were dehydrated in ethanol, embedded in paraffin, and sectioned into 5 μm-thick coronal-oriented slices using a rotary microtome. TRAP staining (Sigma-Aldrich, 387 A) was performed on selected slides from each sample following the manufacturer’s guidelines. Briefly, the selected slides were pre-warmed to 37 °C in a water bath and then incubated in TRAP Staining Solution Mix at 37 °C for 15 min. The nuclei were counterstained with Methyl Green (Millipore Sigma, M8884). The following secondary antibodies were used: Alexa 647-conjugated goat anti-rabbit antibody (Invitrogen, A32733, 1:500), Alexa 488-conjugated donkey anti-rabbit antibody (Invitrogen, A21206, 1:500), Alexa 555-conjugated donkey anti-goat antibody (Invitrogen, A21432, 1:500), Alexa 647-conjugated donkey anti-goat antibody (Invitrogen, A32849, 1:500), Alexa 488-conjugated goat anti-mouse antibody (Invitrogen, A32723, 1:500), and Alexa 647-conjugated donkey anti-mouse antibody (Invitrogen, A31571, 1:500). The nuclei were counterstained with Hoechst 33324 (Thermo Fisher Scientific). Images were captured using a Nikon N-STORM & A1 confocal microscope, Zeiss LSM880 and a polarized light microscope (Nikon Eclipse VL100POL, Tokyo, Japan). Quantitative histomorphometric analysis was performed using ImageJ software (v.1.5, National Institutes of Health, Bethesda, MD, USA).

Whole-mount tissue clearing, immunolabeling, and 3D imaging

Whole tissue immunolabeling and clearing were performed using the iDISCO+ protocol^67,68 with modifications. Briefly, harvested femoral bones were washed in 100% methanol for 1 h and subsequently incubated overnight in 66% dichloromethane/33% methanol at room temperature. Samples were then washed twice in 100% methanol (30 min each) and bleached in fresh chilled 3% H₂O₂ in methanol overnight at 4 °C. Following bleaching, samples were rehydrated in a graded methanol series (80%, 60%, 40%, 20% methanol in PBS containing 0.2% Triton X-100, 1 h per step) at room temperature, followed by two 1-h washes in PBS with 0.2% Triton X-100. Tissues were then permeabilized in PBS containing 0.2% Triton X-100, 20% (v/v) DMSO, and 2.3% (w/v) glycine for 24 h at 37 °C. Non-specific binding was blocked in blocking solution (PBS with 0.2% Triton X-100, 10% DMSO, and 6% donkey serum) for 24 h at 37 °C. Samples were incubated with primary antibodies diluted in PTwH (PBS with 0.2% Tween-20 and 0.1% heparin) containing 5% DMSO and 3% donkey serum for 7 days at 37 °C. The following primary antibodies were used: rabbit anti-tyrosine hydroxylase (TH; 1:200, Abcam, ab152) and mouse anti-CGRP (1:200, Abcam, ab81887). After primary antibody incubation, samples were washed in PTwH (4 × 2 h, then overnight). Secondary antibody incubation was performed in PTwH with 3% donkey serum for 7 days at 37 °C using donkey anti-rabbit Alexa Fluor 488 and donkey anti-mouse Alexa Fluor 647 (both 1:500, Jackson ImmunoResearch). Following secondary staining, samples were washed in PTwH (4 × 2 h, then overnight). For optical clearing, tissues were dehydrated in a graded methanol series (20%, 40%, 60%, 80%, 100% methanol, 1 h per step), incubated in 66% dichloromethane/33% methanol for 3 h, followed by two 15-min washes in 100% dichloromethane. Finally, samples were transferred to dibenzyl ether and stored in light-protected glass vials until imaging. 3D imaging was performed on a Zeiss Lightsheet Z.7 microscope equipped with a 5×/0.16 NA objective.

Dynamic bone formation analysis by calcein double-injection

To assess dynamic bone formation, mice received intraperitoneal injections of calcein (20 mg/kg; Sigma-Aldrich, C0875) at 9 days and 3 days prior to euthanasia, creating a 6-day labeling interval. Undecalcified bone samples were fixed in 70% ethanol, dehydrated through a graded ethanol series, and embedded in methyl methacrylate. Longitudinal sections (10-μm thick) were prepared using a microtome and visualized under a fluorescence microscope. The Mineral Apposition Rate (MAR), defined as the distance between the two calcein labels in the newly formed bone divided by the 6-day interval, was quantified using OsteoMeasureXP software (Osteo-Metrics, Inc.).

Mouse BMAd isolation and cell viability and purity assessments

Adipocytes were isolated from mouse bone marrow using established protocols³³. Briefly, bones were harvested, and the ends were trimmed. The marrow was extruded by brief centrifugation (1000 × g for 10 s) into a collection tube. Red blood cells were lysed using ACK lysing buffer (Thermo Fisher Scientific), and the remaining cells were washed with PBS. The cell suspension was then centrifuged at 300 × g for 5 min at 4 °C. The top layer containing mature adipocytes was carefully collected using wide-bore pipette tips (prepared by cutting standard tips to reduce shear stress). Adipocytes were washed three times with PBS, with each wash involving gentle centrifugation (300 × g, 5 min, 4 °C) and careful aspiration of the infranatant from beneath the floating adipocyte layer using a syringe. This protocol was optimized to maximize the yield of intact, mature adipocytes.

Cell transplantation

To evaluate the regulatory role of senescent BMAds in bone remodeling, BMAds isolated from donor mice were transplanted into healthy WT C57BL/6 mice via intrafemoral injection. After 3-day treatment with rNpy or PBS, BMAds were transplanted into 3-month-old recipient mice. Under anesthesia, a longitudinal incision was made on the anterior aspect of the right knee to expose the patellar tendon. A 27 G needle was carefully inserted through the patellar tendon, positioned between the femoral condyles, and gradually advanced with a twisting motion to a depth of 2–4 mm within the bone. Successful penetration was confirmed by slow but consistent bleeding. Cell suspensions containing BMAds (3 × 10³ cells) in 20 μL PBS were then slowly injected into the femoral medullary space. The injection site was immediately sealed with bone wax, and the skin was sutured to complete the procedure.

Sympathetic neuron extraction and culture

Male mice were euthanized by cervical dislocation and immersed in 75% ethanol for 3 min. The superior cervical ganglia (SCG) were carefully dissected under a microscope and transferred to a solution containing 0.1% collagenase and 0.05% trypsin in HBSS. The tissues were digested at 37 °C for 45–60 min. After digestion, the ganglia were triturated gently with a fire-polished Pasteur pipette to dissociate cells. The cell suspension was filtered through a 40 μm sieve and plated onto culture dishes pre-coated with poly-L-lysine and laminin. After incubation for 1–2 h to allow non-neuronal cells (fibroblasts) to attach, the non-adherent sympathetic neurons were collected and re-seeded onto new culture plates. The cultures were maintained in neuronal medium supplemented with B27 and NGF (50 ng/mL) for 4–5 days until axons extended.

DRG culture

For the isolation of DRG neurons, DRGs from the L2–L5 spinal levels were isolated in cold DMEM/F12 medium (Invitrogen, Thermo Fisher Scientific, 11320033) and then digested with 1 mg/mL collagenase type A (Roche, Basel, Switzerland, 05349907103) at 37 °C for 30–45 min. After trituration and centrifugation, cells were resuspended and seeded on glass coverslips coated with poly-D-lysine and laminin. The culture medium was replaced 6 h after seeding, and the adherent cells were further cultured at 37 °C with 5% CO₂ for 3 days before use. After 3 days of culture, the supernatant was collected following centrifugation at room temperature for 5 min at 200 × g.

Astral-DIA detection and data processing

The supernatant from sympathetic neurons in Non-PEMF and PEMF-treated groups (n = 3 biological replicates per group) was collected every 2 days using standard culture medium. Proteins in the supernatants were concentrated using Amicon Ultra filters (10 kDa cut-off, UFC9010, Millipore) and then quantified by a bicinchoninic acid (BCA) assay. Afterward, 20 μg proteins were reduced by dithiothreitol (10 mM) for 1 h at 37 °C, followed by the alkylation with iodoacetamide (20 mM) for 30 min at 25 °C in the dark. Trypsin was added at a protein-to-enzyme ratio of 50:1 (w/w) and incubated for 16 h at 37 °C. The peptide mixture was acidified with trifluoroacetic acid and desalted using C18 tips. The eluted peptides were dried in a vacuum centrifuge and then redissolved in 0.1% formic acid. All samples were detected by LC-MS/MS in the DIA mode using an Astral mass spectrometer coupled with a Vanquish Neo UHPLC system (Thermo Fisher Scientific). Mobile phases A (100% water, 0.1% formic acid) and B (80% acetonitrile, 0.1% formic acid) were prepared. The peptide sample was loaded and separated on a C18 column (2 μm, 150 μm × 150 mm) using an 8-min gradient at a flow rate of 2.5 μL/min. The gradient program was set as follows: 0–4 min, 4–25% B; 4–6.9 min, 25–35% B; 6.9–7.3 min, 35–99% B; and 7.3–8.0 min, 99% B. MS detection parameters were set as follows. MS1 was analyzed at an Orbitrap resolution of 240,000 with a scan range of 380–980 (m/z). The AGC target, RF lens and maximum injection time (IT) were 5.0E6, 40% and 5 ms, respectively. In DIA acquisition mode, each cycle consisted of one full MS1 scan followed by 300 isolation windows ranging from 380–980 (m/z) at 2 Th isolation width and 0 Th window overlap. MS2 fragment ion scan was acquired at the range of 150–2000 (m/z). The normalized collision energy, AGC target, and maximum IT were 25%, 5.0E4, and 3 ms, respectively. MS raw data were searched against protein sequences downloaded from the Uniprot database (Mus musculus; version 2026_01; 17,247 entries) with DIA-NN software (version 1.9.2) using a library-free approach. Trypsin/P was set as the enzyme, and two maximum missed cleavages were allowed for protein digestion. Carbamidomethylation (C) was set as the fixed modification, and oxidation (M) and acetylation (protein N-terminus) were used as variable modifications. Mass tolerances for precursor and fragment ions were set as 20 ppm and 0.05 Da, respectively. Minimum peptide length was set to 7. The minimum number of unique peptides for protein identification was set to 1 as the default. The results were filtered at 1% false discovery rates of the peptide spectrum matches, peptides, and protein groups.

Quantitative real-time PCR

Total RNA was isolated from cultured or sorted cells using the RNeasy Mini Kit (QIAGEN, Germany) according to the manufacturer’s instructions. Purified RNA was reverse-transcribed into cDNA with random primers using the SuperScript First-Strand Synthesis System (Invitrogen, USA). Quantitative reverse transcription PCR (qRT-PCR) was then performed with SYBR Green Master Mix (QIAGEN) in a thermal cycler, employing two independent primer sets for each target gene. Target gene expression was normalized to glyceraldehyde 3-phosphate dehydrogenase (Gapdh) mRNA, and relative expression levels were calculated using the 2^ − ΔΔCT method. The primers used for qRT-PCR were as follows: Cdkn2a (5′-GAAAGAGTTCGGGGCGTTG-3′ and 5′-GAGAGCCATCTGGAGCAGCAT-3′); Cdkn1a (5′- CCTGGTGATGTCCGACCTG-3′ and 5′-CCATGAGCGCATCGCAATC-3′); Npy (5′- ATGCTAGGTAACAAGCGAATGG-3′ and 5′-TGTCGCAGAGCGGAGTAGTAT-3′); Y1R (5′-CACGGCCATCCTATATGGTAA-3′ and 5′-GGGCCTGGTAGTTGTTGTGA-3′); Gapdh (5′-ACAACTTTGGCATTGTGGAA-3′ and 5′-GATGCAGGGATGATGTTCTG-3′).

ELISA assay

Sema3A concentrations in serum and cell culture supernatants were measured using a commercial ELISA kit (Biomatik, EKU07258) following the manufacturer’s protocol. Blood samples were collected from euthanized mice, and serum was obtained by centrifugation at 3000 × g for 15 min at 4 °C. Bone marrow supernatants were prepared by dissecting tibias, removing both epiphyses and soft tissues, and flushing the marrow with cold PBS, followed by centrifugation at 2000 × g for 15 min at 4 °C. All samples were aliquoted and stored at −80 °C until analysis to prevent repeated freeze-thaw cycles. NPY levels in serum and bone marrow supernatants were quantified with a commercial ELISA kit (Novus Biologicals) in accordance with the manufacturer’s instructions. Norepinephrine (NE) was extracted from homogenized tissues in 0.01 N HCl containing 0.15 mM EDTA and 4 mM sodium metabisulfite, and quantified using an ELISA kit (Eagle Biosciences). NE concentrations were normalized to total protein content determined by a BCA assay (ThermoFisher Scientific). Circulating levels of TNF-α and IL-6 were assessed with a high-sensitivity ELISA kit (R&D Systems). Serum corticosterone levels were measured as previously described. Blood samples were collected from the abdominal aorta between 8:00 a.m. and 12:00 p.m. to minimize circadian variation. Serum was separated by centrifugation at 3000 × g for 15 min at 4 °C and stored at −80 °C until analysis. For the ELISA assay (ADI-901-097, Enzo Life Sciences), serum samples were diluted 1:100 in assay buffer according to the manufacturer’s protocol.

Statistical analysis

Data are expressed as mean ± SD. For data that are normally distributed, comparisons between two groups were performed using two-tailed Student’s t-tests. For comparisons among more than two groups, one‑way or two‑way analysis of variance (ANOVA) was applied, followed by Tukey’s post‑hoc multiple comparison tests, with statistical significance assessed at a 95% confidence interval. All statistical analyses were conducted using GraphPad Prism 10. Statistical differences were considered significant at P < 0.05. Unless otherwise stated in the figure legends, all representative images of bones or cells are selected from at least three independent experiments that yielded similar results.

Ethical statement

All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Sichuan University (Protocol 2021215A, 20230207007), conducted in accordance with the guidelines of the National Institutes of Health Guide for the Care and Use of Laboratory Animals and complied with the principles of the Animal Welfare Act. All surgeries and terminal procedures were performed under isoflurane anesthesia, and every effort was made to minimize animal suffering. The reporting of all animal experiments in this study adheres to the ARRIVE guidelines.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

T.W. designed the experiments; T.W., Z.L., W.Z., J.C., Y.Z., H.W., L.T., Y.M., and S.C. carried out most of the experiments; S.C., C.W., and X.G. helped with figure draw; T.W., D.Z., W.Z., S.C., and Z.H. proofread the manuscript; T.W. and Z.H. supervised the experiments, analyzed results, and wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Andreas Nussler and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the iProX partner repository with the dataset identifier PXD080276 at public access level (https://www.iprox.cn//page/SCV017.html?query=PXD080276). The transcriptomic dataset of human osteoporotic bone tissue re-analyzed in this study is publicly available in the NCBI Gene Expression Omnibus (GEO) database under accession number GSE230665. Source data generated in this study are provided in Figshare (https://doi.org/10.6084/m9.figshare.32901515). Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77654-2.