A cholinergic neuroskeletal interface promotes bone formation during postnatal growth and exercise

Summary The autonomic nervous system is a master regulator of homeostatic processes and stress responses. Sympathetic noradrenergic nerve fibers decrease bone mass, but the role of cholinergic signaling in bone has remained largely unknown. Here, we describe that early postnatally, a subset of sympathetic nerve fibers undergoes an interleukin-6 (IL-6)-induced cholinergic switch upon contacting the bone. A neurotrophic dependency mediated through GDNF-family receptor-α2 (GFRα2) and its ligand, neurturin (NRTN), is established between sympathetic cholinergic fibers and bone-embedded osteocytes, which require cholinergic innervation for their survival and connectivity. Bone-lining osteoprogenitors amplify and propagate cholinergic signals in the bone marrow (BM). Moderate exercise augments trabecular bone partly through an IL-6-dependent expansion of sympathetic cholinergic nerve fibers. Consequently, loss of cholinergic skeletal innervation reduces osteocyte survival and function, causing osteopenia and impaired skeletal adaptation to moderate exercise. These results uncover a cholinergic neuro-osteocyte interface that regulates skeletogenesis and skeletal turnover through bone-anabolic effects.


Correspondence
In brief Gadomski et al. describe a neuroosteocyte interface whereby sympathetic cholinergic neurons support boneembedded osteocytes through the GFRa2 neurotrophic pathway. Developmentally, these sympathetic neurons undergo a neurotransmitter switch from adrenergic to cholinergic-a process that is induced by interleukin-6 and is dynamically enhanced by physical activity to increase bone mass.

INTRODUCTION
The two branches of the autonomic nervous system, sympathetic and parasympathetic, normally use the postsynaptic neurotransmitters norepinephrine (noradrenergic) and acetylcholine (ACh) (cholinergic), respectively. However, some embryonic sympathetic neurons exhibit cholinergic features, but their frequency gradually diminishes to $4% of sympathetic neurons by birth (Huang et al., 2013;Sch€ afer et al., 1997). It is unclear whether these early target-independent sympathetic cholinergic neurons overlap with sympathetic neurons that become cholinergic postnatally (Sch€ utz et al., 2015). This ''cholinergic switch'' (Wolinsky and Patterson, 1983) of sympathetic neurons occurs during the first postnatal weeks in rodents (Guidry and Landis, 1998) and was characterized in vivo in the sweat glands and the periosteum (Asmus et al., 2000;Hohmann et al., 1986). In bone, the cholinergic switch resembles the neurotransmitter change in sweat glands (Habecker and Landis, 1994), since it requires initially noradrenergic activity, ensuing secretion of yet unidentified cholinergic differentiation factors, ACh release, and maturation of both the target organ and its cholinergic innervation. Further, the role of skeletal cholinergic fibers in bone development and remodeling remains largely unexplored. One study suggested that cholinergic fibers innervate bone and transmit anabolic signals from the brain (Bajayo et al., 2012). Alternatively, parasympathetic signals can promote bone formation by antagonizing bone-catabolic sympathetic noradrenergic signals in the brain via muscarinic ACh receptors (Shi et al., 2010). Therefore, we sought to identify the factor driving the cholinergic switch in vivo and examine the functional significance of skeletal sympathetic cholinergic fibers.
Here, we have identified IL-6 as a driver of the cholinergic switch of bone-associated sympathetic neurons during postnatal development and a promoter of cholinergic signaling in response to physical activity during adolescence. A neurotrophic dependency is established between cholinergic nerve fibers and osteocytes relying on the GFRa2-neurturin (NRTN) axis. Bonelining osteoprogenitors connected with the osteocyte network transmit and amplify the cholinergic signals in the bone marrow (BM). Lack of skeletal cholinergic nerve fibers causes osteocyte atrophy and osteopenia due to reduced bone formation, while increased IL-6 during exercise drives expansion of boneanabolic cholinergic fibers. These results uncover a dynamic bone-anabolic function of sympathetic cholinergic fibers coupled with the osteocyte network.

Cholinergic nerve fibers and bone-lining cells in bone and BM
We performed immunofluorescence studies to map nerve fibers in bone and BM. 3D imaging of wild-type (WT) and Nes-gfp transgenic mice-in which a subset of GFP-labeled cells marks skeletal stem cells (SSCs) (Mé ndez-Ferrer et al., 2010)-showed protein gene product 9.5 (PGP9.5) + nerve fibers in cortical bone, near the growth plate, and throughout the skull (Figures S1A and S1B). Unexpectedly, the pan-neural marker b-III tubulin (TUJ1) did not only label nerve fibers ( Figures 1A-1C, arrowheads) but also osteolineage cells expressing runt-related transcription factor 2 (RUNX2) or osteolectin (Yue et al., 2016) (Figures 1A-1D, arrows). Choline acteyltransferase (ChAT)-IRES-cre mice (Rossi et al., 2011) were intercrossed with Ai35D reporter mice (Madisen et al., 2012) to genetically label cholinergic neurons. Resembling TUJ1, ChAT-IRES-Cre tracing did not only mark nerve fibers in cra-nial ( Figure S1C) and femoral (Figures 1E, S1D, and S1E, arrowheads) bones but also appeared to label bone-lining cells near the osteochondral junction of the growth plate ( Figure 1F, arrows). Expression of vesicular ACh transporter (VAChT), which loads ACh into secretory organelles of cholinergic nerve terminals (Weihe et al., 1996), co-localized with ChAT-labeled bone-lining cells and cholinergic fibers associated with blood vessels (Figures 1G,  1H, S1F, and S1G). Vasoactive intestinal peptide (VIP), marking sympathetic cholinergic fibers in the periosteum (Asmus et al., 2000;Francis et al., 1997;Hohmann et al., 1986), followed a similar periosteal and perivascular staining pattern in cortical bone (Figure 1I). These data confirm the presence of cholinergic innervation in the periosteum and extend these findings to bone matrix and BM, where non-neural cholinergic osteolineage cells were additionally detected and characterized below (see Figure 4).

Sympathetic cholinergic nerve fibers in bone
A previous study suggested that cholinergic fibers innervating bone are parasympathetic based on retrograde tracing to thoracic and sacral spinal cord segments (Bajayo et al., 2012). However, a sympathetic origin has been proposed for sacral autonomic outflow (Espinosa-Medina et al., 2016). To clarify the origin of cholinergic fibers, neonatal mice were treated with 6-hydroxydopamine (6-OHDA) to ablate sympathetic fibers before the cholinergic switch during postnatal development (Figure 2A). At adulthood, similar reductions of noradrenergic (TH + ) fibers and cholinergic (GFRa2 + or VAChT + ) fibers (Hiltunen and Airaksinen, 2004) were observed in the femurs and skull bones of 6-OHDA-treated mice ( Figures 2B, 2C, S2A, and S2C), suggesting a sympathetic origin of skeletal cholinergic fibers. For confirmation, we intercrossed TH-cre mice with Ai14D reporter mice and found that VAChT + staining frequently co-localized with genetically traced sympathetic fibers near bone ( Figure 2D). Furthermore, VAChT + and GFRa2 + cholinergic axons traveled in the same nerve bundles as TH + noradrenergic fibers, showing separation with successive branching (Figures S2D and S2E). Overall, these results support a sympathetic origin for skeletal cholinergic fibers.
Interleukin-6 triggers a cholinergic switch in sympathetic neurons IL-6 was an interesting candidate because-similar to CNTF, CT-1, and LIF-its signaling requires gp130 (Ip et al., 1992;Taga et al., 1989), which is essential for the cholinergic switch (Stanke et al., 2006) but also entails a unique co-receptor, IL-6R. Primary superior cervical ganglion (SCG) sympathetic neurons were treated with recombinant mouse (rm) IL-6 alone or in combination with inactivating antibodies against the mouse soluble IL-6 ligand (anti-mIL-6-IgG) or the human IL-6 receptor (tocilizumab). The expression of cholinergic and noradrenergic markers was measured after 14 days in culture ( Figure 2E). rmIL-6 caused selective induction of cholinergic markers (Figure 2F) and downregulation of noradrenergic markers (Figure 2G). These effects were reversed by IL-6 inhibitors ( Figures  2F and 2G), demonstrating specificity. Confocal analyses of rmIL-6-treated SCG cultures confirmed increased GFRa2 + (cholinergic) and reduced TH + (noradrenergic) staining, while co-treatment with tocilizumab abrogated the cholinergic switch (Figures 2H and 2I).
WT SCG cultures showed endogenous IL-6 expression (Figures S3A and S3B) and spontaneous induction of cholinergic markers; in contrast, the cholinergic switch was nearly abrogated in Il6 À/À cultures ( Figures 2J and S3C-S3E). These results demonstrate that IL-6 can induce a neuronal cholinergic switch in vitro. Contrarily, noradrenergic gene expression increased over time in untreated Gfra2 À/À SCG cultures ( Figure 2J); however, Il6 mRNA expression was normal ( Figures S3A and S3B), suggesting an altered response to IL-6. Indeed, instead of inducing a cholinergic switch, rmIL-6 increased noradrenergic marker expression in Gfra2 À/À neurons ( Figures S3F and S3G), likely due to different cis/trans IL-6 signaling: cis-signaling involves the natural binding of IL-6 to its receptor and subsequent gp130 activation, while trans-signaling results from the cleavage of IL-6R, producing a soluble IL-6R that can bind IL-6 and acti-vate gp130 in other cells, leading to distinct differences in signal specificity, timing, amplification, and overall cellular phenotypes (Rose-John et al., 2017). Gfra2 À/À SCG neurons showed increased expression of TNFa-converting enzyme (TACE), one of the proteases responsible for the cleavage of membranebound IL-6R (Solomon et al., 2007) (Figure S3H), suggesting that resistance to cholinergic induction may be mediated through trans-IL-6-signaling. Supporting this possibility, TACE inhibition during the 1 st two postnatal weeks normalized cholinergic and noradrenergic nerve fibers in the bones of Gfra2 À/À mice ( Figures S3I and S3J).
Interleukin-6 promotes a sympathetic cholinergic switch in bone Because IL-6 enhances the cholinergic phenotype in developing sympathetic neurons in vitro, we examined IL-6's source and potential to drive the cholinergic switch in vivo. A proximity ligation assay of postnatal day 3 developing limbs showed high IL-6 near the periosteum, mainly in adjacent skeletal muscle (Figures 3A and 3B). Therefore, we treated mice with IL-6 inhibitors or control IgG weekly during the first 6 postnatal weeks ( Figure 3C). Femurs and skulls of mice treated with IL-6 inhibitors showed a normal presence of TH + noradrenergic nerve fibers ( Figures S3K and  S3L), but a 3-to 4-fold reduction in VAChT + or GFRa2 + cholinergic fibers ( Figures 3D-3F). Similarly, Il6 À/À mice exhibited $3-fold-reduced cholinergic fiber density in femurs and skulls ( Figures 3G-3I), suggesting that IL-6 can promote the cholinergic phenotype in vivo.
Given that cholinergic innervation is enriched at periosteal and cortical sites but regulates hematopoietic cells deeper in the BM (Fielding et al., 2022;García-García et al., 2019), we hypothesized that osteolineage cells containing ACh could transmit and amplify the cholinergic signal in BM. CD51 + osteolineage cells showed higher expression of nicotinic ACh receptors compared with CD51 À cells ( Figures S4H and S4I). Therefore, we treated CD51 + cells with cholinergic agonists, antagonists, or control medium ( Figure 4M). ACh or nicotine doubled ChAT and VAChT mRNA expression and increased ACh content in cultured CD51 + cells; these effects were attenuated with the nicotinic antagonist, hexamethonium ( Figures 4N and  4O). In a separate study, we found that cholinergic signals increase after myeloablation or irradiation and preserve hematopoietic stem cell quiescence after transplantation via an a7 nicotinic receptor in niche cells (Fielding et al., 2022). Four weeks after lethal irradiation and transplantation of BM cells, ACh content was reduced in the endosteal (not central) BM of recipient mice lacking an a7 nicotinic receptor in Leptin-receptor-Cre-targeted niche cells (Ding et al., 2012), which largely overlap with Nes-GFP + SSC-enriched cells (Mende et al., 2019; Mé ndez-Ferrer, 2019) ( Figure 4P). Therefore, osteolineage cells may transmit and amplify cholinergic neural signals in bone and BM. Osteopenia and reduced bone formation in Gfra2 -/mice Since cholinergic activity promotes bone mass accrual (Bajayo et al., 2012;Shi et al., 2010), we asked whether cholinergic neural deficiency might compromise skeletogenesis or skeletal turnover. Cortical morphometry of tibias from Gfra2 À/À females showed reduced cortical bone size, volume, volume fraction, cortical bone thickness, and trabecular thickness, while trabecular separation and number remained unchanged ( Figures 5A-5C). Gfra2 À/À male mice exhibited a milder phenotype with a trend toward a decrease in bone volume and thickness, which inversely correlated with cholinergic nerve fibers in cortical bone ( Figures S5A-S5C). This suggests gender-specific bone phenotypes, as shown for CNTF À/À mice lacking another gp130 ligand (McGregor et al., 2010). Cranial sutures were markedly expanded and skulls appeared flatter in Gfra2 À/À males ( Figure S5D). Skeletal parameters reduced in Gfra2 À/À mice did not correlate with the expectedly reduced body weight of these mice (data not shown), uncoupling nutrition defects (McDonagh et al., 2007;Rossi et al., 1999Rossi et al., , 2003 from the skeletal phenotypes. Notably, three-point bend tests showed reduced stiffness and strength in both female and male Gfra2 À/À tibias ( Figures 5D and S5E).

GFRa2 signaling maintains osteocyte connectivity and survival
The persistent osteopenia and reduced bone formation in Gfra2 À/À mice despite the increased osteoprogenitors suggested a defect in the orchestration of surface bone formation, which is normally achieved through the fine control of OB activity and recruitment by the network of mineral-embedded osteocytes (retired OBs); the osteocyte syncytium fulfills this role in controlling surface activity via connected dendrites networked within billions of fine canaliculi (Robling and Bonewald, 2020). In Gfra2 À/À mice, we observed grossly abnormal osteocyte morphology, showing large spherical or flattened cell bodies and reduced dendrites ( Figures 5H-5K, S6A, and S6B). Transmission electron microscopy (TEM) confirmed reduced branching in Gfra2 À/À osteocytes and revealed membrane blebbing, abundant autophagosomes, and reduced lacunar space (Figures 5L, S6C, and S6D), suggesting osteocyte degeneration and impaired collagen cleavage. Osteocyte-like cells (OLCs) differentiated from Gfra2 À/À pOBs showed decreased survival, explaining $30% reduced osteocytes in Gfra2 À/À femurs .
To investigate GFRa2 signaling, we profiled the GDNF family of ligands and receptors in pOBs, OLCs, and primary osteolineage cells. Gfra2 and related ligands and receptors were expressed in WT osteolineage cells, while Gfra2 mRNA expression increased following osteogenic differentiation, and Gfra2 À/À osteocytes showed decreased expression of Mt1-Mmp-a membrane-anchored proteinase required for collagen cleavage and osteocyte branching (Holmbeck et al., 2005) and increased mRNA expression of sclerostin (Sost)-an inhibitor of Wnt signaling and bone formation secreted by osteocytes (Holdsworth et al., 2019) ( Figures 6D and 6E). Furthermore, Gfra2 À/À osteocytes showed high sclerostin protein levels, which were resistant to their normal repression by mechanical loading (treadmill exercise) ( Figure S6E). Mechanistically, sclerostin inhibition in exercised Gfra2 À/À mice normalized BFR, strength, and trabecular thickness ( Figures 6F-6H), highlighting the relevance of sclerostin in the osteopenia of Gfra2 À/À mice. Skeletal responses of WT mice to sclerostin blockade were as expected (Holdsworth et al., 2019).

Cholinergic fibers in bone maintain osteocyte survival and connectivity
Proximity ligation assay showed the highest NRTN expression among cholinergic fibers in bone ( Figure 6K), suggesting that these fibers can activate NRTN co-receptor RET signaling in osteocytes and their lack may cause osteocyte degeneration in GFRa2-expressing osteocytes in vivo. Indeed, neonatal sympathectomy (to ablate adult peripheral skeletal cholinergic fibers) similarly reduced adult osteocyte number and dendritic branching ( Figures 6L-6N), which was phenocopied in Nrtn À/À mice ( Figures 6O-6Q), suggesting that NRTN-GFRa2 signaling promotes osteocyte survival. Since 6-OHDA treatment in neonates (before the cholinergic switch) ablates adult sympathetic cholinergic and noradrenergic fibers, for comparison we administered 6-OHDA in adult mice, selectively ablating noradrenergic (but not cholinergic) fibers ( Figure S6H). Contrasting neonatal treatment, adult 6-OHDA treatment did not affect osteocyte morphology, branching, or numbers ( Figures S6I-S6K). Therefore, lack of peripheral sympathetic cholinergic fibers, or NRTN, in mice with GFRa2-competent osteocytes phenocopies the osteocyte defects associated with global GFRa2 deficiency, suggesting that this neuro-osteocyte interface preserves the osteocyte network.
Skeletal muscle-derived IL-6 regulates bone remodeling during exercise (Chowdhury et al., 2020). Since IL-6 can drive the cholinergic switch (see Figures 2, 3, S2, and S3), we wondered whether IL-6 could boost cholinergic activity to facilitate skeletal adaptation to exercise in young mice. Cholinergic fiber density was doubled in skulls of exercised mice, but not after IL-6 blockade ( Figures 7A-7C), suggesting a role for circulating IL-6. Similar results were obtained in ChAT-IREScre;Ai14D;Nes-gfp mice; moderate exercise increased cholinergic fiber density near perivascular Nes-GFP + SSC-enriched cells, but not after IL-6 blockade ( Figures 7D and 7E).
Since our in vitro studies showed that ACh stimulation can increase ACh content in osteolineage cells (see Figures 4N and  4O), we asked whether exercise-induced sympathetic cholinergic activity propagates to osteolineage cells in vivo. Supporting this concept, ChAT-traced osteoprogenitors expanded  and ACh concentration ( Figures 7I and 7J) increased in osteoprogenitors and osteocytes from exercised mice, but not after IL-6 blockade, matching the cholinergic neural response (see Figures 7B-7E) and further suggesting impaired cholinergic propagation in bone-forming cells. Importantly, consistent with results in the rat model, moderate exercise increased trabecular thickness in WT mice, but not in Gfra2 À/À mice or WT mice with IL-6 blockade ( Figures 7K and 7L). These results suggest that IL-6 not only drives the cholinergic switch during postnatal development, but also serves to strengthen the cholinergic regulation of the skeleton in response to physical activity during adolescence.

DISCUSSION
Here, we characterized the neuronal and non-neuronal cholinergic system in bone. We found that skeletal sympathetic cholinergic nerve fibers, which are induced by IL-6, preserve osteocyte survival and function through the NRTN-GFRa2 neurotrophic axis during postnatal development and physical activity in adolescence. These conclusions are supported by: (1) cholinergic nerve fibers being the main source of NRTN near bone (Figure 6K); (2) NRTN directly promoting osteocyte survival ( Figures  6I and 6J) in GFRa2-and RET-expressing osteocytes ( Figures  6D and 6E); (3) treatment with GDNF or NRTN improving growth and survival in WT pOBs, while NRTN's trophic effect is reduced in Gfra2 À/À pOBs ( Figures 6I and 6J); (4) MLO-Y4 OLCs (Kato et al., 1997) similarly exhibiting reduced apoptosis upon GFR treatment ( Figures S6F and S6G); (5) osteocyte numbers being reduced and atrophic in Nrtn KO mice ( Figures 6O-6Q) or after neonatal sympathectomy of cholinergic fibers ( Figures 6L-6N), but not after adult sympathectomy of noradrenergic fibers (Figures S6H-S6K); (6) bone adaptation to moderate exercise being impaired in cholinergic-neural-deficient mice ( Figure 7L) or in rats after neonatal sympathectomy of cholinergic fibers (Figures S7F-S7I); (7) deficient bone-anabolic responses in cholinergicneural-deficient mice, explained by the incapacity of osteocytes to repress sclerostin, which is a key inhibitor of bone-anabolic Wnt signaling ( Figures 6D, 6E, and S6E); and (8) the key role of deregulated sclerostin in the absence of sympathetic cholinergic fibers, which is demonstrated by the rescue of osteopenia and bone strength in GFRa2 KO mice treated with sclerostin-blocking antibody (Figures 6F-6H).
Our study confirms the presence of cholinergic innervation in the periosteum and extends these findings to bone matrix and BM near the epiphyseal growth plate. Furthermore, osteolineage cells emerge as an additional component of the non-neuronal cholinergic system in bone. Treatment with ACh increases cholinergic markers and ACh content in CD51 + osteolineage cells, but not after nicotinic receptor blockade, suggesting that cholinergic neural signals are relayed to osteolineage cholinergic cells. Supporting this possibility, ACh levels were higher in the endosteal BM of chimeric mice and were specifically reduced in the endosteal BM upon a7 nicotinic deletion in LepR-Cre-targeted cells. These results suggest that cholinergic neural signals are relayed to bone-forming cells.
Since central or peripheral cholinergic activity promotes bone mass accrual (Bajayo et al., 2012;Shi et al., 2010), we asked whether the lack of skeletal sympathetic cholinergic fibers might compromise skeletogenesis or skeletal turnover. Long bones from Gfra2 À/À mice show normal bone-resorbing parameters but reduced SSCs and bone formation, leading to decreased bone mass and strength, enlarged cranial sutures, and flatter skulls. Impaired bone-anabolic response appears to result from structural and functional alterations in bone-embedded osteocytes. The osteocyte network plays a key role in orchestrating bone remodeling and IL-6-and Wnt-dependent bone-anabolic responses to mechanical loading during physical activity (Robling and Bonewald, 2020). Mechanistically, Gfra2 À/À osteocytes overproduce the Wnt inhibitor sclerostin, and sclerostin blockade rescues many of the histomorphometric defects. Peripheral sympathectomy at neonatal stage (ablating cholinergic (H-K) Phalloidin staining (H and J, green) and quantification (I and K) of osteocytes embedded in (H and I) cortical bone or (J and K) trabecular bone from WT or GFRa2 KO mice. Nuclei were counterstained with DAPI (blue). Scale bars, 50 mm. (L) Transmission electron micrographs of osteocytes (left) and surrounding collagen matrix (right) from WT or GFRa2 KO humeri. Arrowheads depict osteocyte cell processes. Scale bars, 500 nm. See also Figures S6C and S6D.  (A, B, D, F, G, I, and K) Data are mean ± SEM, *p < 0.05, **p < 0.01, ***p < 0.001, unpaired two-tailed t test. Article fibers)-but not at adult stage (ablating only noradrenergic fibers)-recapitulates the osteocyte defects of Gfra2 À/À mice. Therefore, we conclude that skeletal sympathetic cholinergic fibers have bone-anabolic effects complementary to those of central cholinergic inhibition of sympathetic tone (Shi et al, 2010). GFRa2 + fibers are also detected in the BM of Wistar rats and resemble cholinergic fibers recently reported in human bone (Courties et al., 2020), suggesting interspecies conservation.
Both in mice and Wistar rats, moderate exercise doubles skeletal cholinergic fibers, correlated with increased trabecular bone. However, cholinergic fiber induction and increased ACh concentration in osteoprogenitors are blunted by IL-6 blockade in mice. Furthermore, skeletal adaptation to moderate exercise is severely compromised by early postnatal sympathectomy in rats. Therefore, we conclude that IL-6-driven cholinergic signals are required for the skeletal adaptation to exercise. In humans, IL-6 gene variants have been associated with osteoporosis and osteopenia (Ota et al., 1999(Ota et al., , 2001. In our study, the boneanabolic effects of cholinergic signals appear to involve cis-IL6signaling (instead of trans-signaling, which may have opposite effects) (Rose-John et al., 2017). The conclusions are consistent with findings in menopause-related osteoporosis, where excessive trans-(not cis-) IL-6 signaling causes loss of trabecular bone (Lazzaro et al., 2018;Sims, 2021), mirroring the gain of trabecular bone through cis-IL6-induced cholinergic signals.
Genetic lineage tracing and early postnatal sympathectomy in rodents reveal a sympathetic origin of skeletal cholinergic nerve fibers. These axons appear to travel in the same nerve bundles as noradrenergic nerve fibers before branching, suggesting potential inhibitory feedback loops between these fibers as shown in other organs/tissues such as the pancreas (Benthem et al., 2001), eyelid smooth muscle (Beauregard and Smith, 1994), trachea (Pendry and Maclagan, 1991), and heart (Azevedo and Parker, 1999;Gavioli et al., 2014;Hasan and Smith, 2009;Miyashita et al., 1999;Smith-White et al., 1999). Moreover, Gfra2 À/À mice exhibit increased sympathetic noradrenergic innervation in the BM (García-García et al., 2019), similarly supporting putative inhibitory feedback loops. While noradrenergic fibers are found throughout the BM, cholinergic fibers are preferentially located in cortical bone with sprouting branches localized in trabecular BM. Although our data strongly argues for a spatial segregation of noradrenergic and cholinergic axons, we cannot exclude the possibility that some nerve fibers might have mixed and/or highly dynamic properties. The sympathetic SCG contains neurons with combined noradrenergic and cholinergic properties (Furshpan et al., 1986;Landis, 1976), and different neurotrophic factors can rapidly affect neurotransmitter synthesis, storage, release, and uptake (Luther and Birren, 2009;Yang et al., 2002).
Cholinergic signals are propagated to the BM through bonelining osteoprogenitors, which transmit and amplify the cholinergic signal to the BM matrix, regulate the migration of hematopoietic stem cells and leukocytes , and preserve hematopoietic stem cell quiescence during hematopoietic regeneration (Fielding et al., 2022). These results add to the osteocyte network's regulatory role in propagating noradrenergic signals to BM (Asada et al., 2013). Finally, since increased IL-6 during moderate exercise expands boneanabolic cholinergic fibers, the achievement of peak bone mass, which is an important predictor of osteoporosis in late adulthood, may be mediated at least in part by the sympathetic cholinergic system and may represent a drug-able target for maintenance of peak bone mass.

Limitations of the study
Although the results show NRTN-GFRa2 in the maintenance of the neuro-osteocyte interface, other signals might also contribute. Similarly, while deregulated sclerostin expression in osteocytes explains many skeletal phenotypes, other mechanisms and cell types regulated by cholinergic signals might participate in the complex interplay identified here between the skeletal and peripheral neural systems.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
The authors declare no competing interests.

STAR+METHODS RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to the Lead Contact, Simon Mé ndez-Ferrer (sm2116@cam.ac.uk).

Materials availability
This study did not generate new unique reagents.
Data and code availability d Microscopy data reported in this paper will be shared by the lead contact upon request. d This paper does not report original code. d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS Animals
Age and sex-matched Gfra2 -/- (Rossi et al., 1999), Nes-gfp (Mignone et al., 2004) Table S1. Wistar rats (CEA, University of Seville) were used for exercise studies. Animals were housed in specific pathogen-free facilities. All animal experiments followed protocols approved by the Animal Welfare Ethical Committees, according to EU and United Kingdom Home Office regulations (PPL P0242B783).
Osteoblast and osteocyte-like cell cultures An illustration is provided in Figure 6A. For primary osteoblast isolation, calvaria of neonatal mice were removed on postnatal day 3 (P3) and incubated in 4mM EDTA/PBS solution for 10 minutes at 37 C with agitation. Digestion with EDTA (Sigma, Cat. No. E5134) was repeated. The supernatant was discarded and tissues were placed in 0.1% collagenase I/0.2% dispase solution (Sigma, Cat. No. C2674) for 10 minutes at 37 C with agitation. The supernatant was discarded, and enzymatic digestion was repeated four 0.5mg/kg anti-mIL-6-IgG (Invivogen, Cat. No. mabg-mil6-3) were injected subcutaneously once weekly on a rest day (see illustration in Figure 7A). Mice were sacrificed 24 hours after the final treadmill session. The same protocol was employed for sclerostin inhibition experiments in adult mice, with 25mg/kg sclerostin antibody (Scl-Ab, r13c7, UCB Pharma/Amgen Inc.) injected subcutaneously on rest days 1x/week and treadmill exercise performed 5x/week for 5 weeks.

QUANTIFICATION AND STATISTICAL ANALYSIS
Area measurements of confocal/Airyscan2 images were taken from at least 3 samples using ''Color Threshold'' in Fiji/Image J Software to quantify positive staining and dividing by total image area. In some cases, muscle outside the periosteum was cropped from images. For phalloidin area measurements inside bone, areas outside the periosteum and endosteum were cropped for cortical bone, and areas outside trabecular surface which contained hematopoietic cells were cropped for trabecular bone. Osteocytes were quantified by manually counting DAPI + phalloidin + cell bodies within bone from low-magnification Airyscan2 images. Distance analyses were performed using Arivis Vision 4D software (RRID:SCR_018000) with statistical significance determined by Kolmogorov-Smirnov analysis. Data shown in figures are expressed as mean ± standard error of the mean (SEM) and are representative of at least two trials with N values representing biological replicates (animals). One Way ANOVA and Bonferroni comparison were used for multiple group comparisons, and unpaired two-tailed t tests for two-group comparisons. Significant statistical differences between groups were indicated as: *p<0.05, **p<0.01, ***p<0.001. Statistical analyses and graphics were carried out with GraphPad Prism 8 software (RRID:SCR_002798) and Microsoft Excel.