Cholinergic regulation of osteocyte mechanobiology: A paradigm for bone adaptation

Abstract

Recent evidence suggests that acetylcholine has a positive influence on bone mechanotransduction. Osteocytes express components for nicotinic acetylcholine receptors (nAChRs), which are known for mediating calcium signaling and may affect mechanosensitivity. Here, we use fluorescent imaging approaches to provide evidence of close spatial association between osteocytes and cholinergic nerve fibers in cortical bone in vivo. Moreover, we show that osteocytes are responsive to cholinergic signaling, influencing bone mechanoadaptation. We report sexually dimorphic patterns in bone structure and mechanobiology based on nAChR function. In females, osteocyte mechanosensitivity was decreased at small force magnitudes and tissue level deficits were recovered with anabolic loading. In males, osteocyte mechanosensitivity was increased in some groups and anabolic loading had very little effect on overall tissue architecture. This work establishes a previously unidentified paradigm wherein osteocytes interface with cholinergic nerves and bone mechanotransduction is regulated by osteocyte cholinergic signaling in a sexually dimorphic way.

Cholinergic signaling alters bone metabolism differently in males versus females when osteocyte ACh receptors are disrupted.

INTRODUCTION

Bone is an exquisitely mechanically sensitive tissue capable of adapting its architecture and mass to maintain a structurally sound but energetically efficient form. Harold Frost proposed the idea that bone cells actively detect their strain environment to shape and maintain bone mass accordingly (1–3). He posited that if bone experienced prolonged periods of mechanical deformation below or above a set-point value, the mineralized matrix would be resorbed or deposited, respectively. Peak healthy in vivo deformations for a broad range of species are tightly conserved between 150 and 3000 με (4), and aberrant loading results in anabolic bone growth in the area of maximal bending (5). In some bone metabolic diseases, the relationship between tissue deformation and bone mass becomes dysregulated. One such disease, osteoporosis, is characterized by the rapid loss of bone mass and increased risk of fracture. It affects more than 200 million people worldwide, with one in three women and one in five men projected to have a bone fragility fracture in their lifetimes (6, 7). Consequently, understanding cellular sensing mechanisms and the systems-level components that regulate bone mechanotransduction is an area of critical study.

A key biological player in bone mechanotransduction is the osteocyte. These terminally differentiated bone cells express proteins responsible for regulating the behavior of bone-resorbing osteoclasts and bone-building osteoblasts (8–10). They reside encased in bone matrix and are functionally connected by connexin-43 gap junctions near the ends of long dendritic processes, creating a functional syncytium (11, 12). Their location throughout bone and ability to control the behavior of tissue-modifying cells make them ideally designed for their role as resident bone mechanosensors. Osteocytes bear an uncanny resemblance to nerve cells. Both are highly dendritic cells that exist in interconnected networks. Like how nerves sense various stimuli (e.g., heat and pain) in different tissues, osteocytes sense force in bone. Osteocyte mechanotransduction is a complex cell signaling event, and unraveling the critical features that serve as control mechanisms is the subject of ongoing investigation.

Calcium (Ca²⁺) signaling is a ubiquitous second messenger system with pronounced biological influence over nearly every aspect of cell function, including protein conformations, signal transduction, excitability, exocytosis, cell motility, mitochondrial integrity, apoptosis, and gene transcription (13, 14). In osteocytes, Ca²⁺ signaling is a key second messenger in mechanotransduction pathways. Activation of Ca²⁺ signaling in osteocytes in vitro occurs in response to fluid flow shear stress of similar magnitude to that experienced by endothelial cells (15–21) as first predicted by Weinbaum and Cowin (22). Osteocytes also exhibit increases in cytoplasmic Ca²⁺ in response to substrate deformation (23–26) and hydrostatic pressure (27–29). Our group has shown that osteocytes use Ca²⁺ signaling to biologically encode tissue level force magnitude in vivo (30). The variety of components and downstream effects of Ca²⁺ signaling support the theory that it is an important control mechanism for bone mechanoregulation. Hence, the mechanisms that regulate Ca²⁺ signaling dynamics in osteocytes represent important features in bone mechanotransduction. One such mechanism recently implicated is acetylcholine (ACh) receptors.

ACh is a neurotransmitter associated with Ca²⁺ signaling channels that has recently been connected to bone mechanobiology. Neurological disorders like Alzheimer’s disease are correlated with the dysregulation of cholinergic activity. Lower bone mineral density (BMD) (31), higher fracture risk (32), and osteoporosis (33) are all linked to higher rates of dementia. Patients with Alzheimer’s treated with ACh esterase (AChE) inhibitors, which increase ACh half-life, have the added benefit of 20 to 30% reduction in hip fracture, improved hip fracture healing, and improved bone quality (34, 35). In mice, AChE inhibitors increase trabecular bone volume and reduce bone resorption (36). Parasympathetic (36) and sympathetic derived ACh affect skeletal mass and development (37). All three major bone cells (i.e., osteocytes, osteoblasts, and osteoclasts) express proteins associated with ACh metabolism (36–39). Despite convincing evidence of the impactful role of ACh in bone biology, the details regarding cholinergic signaling in bone mechanotransduction remain largely unexplored. What’s more, there are very few data describing ACh signaling in osteocyte mechanotransduction. This is largely due to the technical challenges associated with studying osteocytes embedded in mineralized matrix. Our group specializes in studying osteocyte cell biology in vivo and has worked to shed light on this subject in the work presented here.

Nicotinic ACh receptors (nAChRs) are ACh-gated ion channels that have high permeability to Ca²⁺ (40, 41). Chrna1 encodes for nAChR subunit α1 of the nicotinic receptor, which binds ACh and regulates cellular activation through direct modulation of membrane potential (42, 43). The differentiation of osteoblasts into osteocytes is associated with the up-regulation of several neurotransmitter-associated genes, including Chrna1 (38, 44). In muscle (i.e., neuromuscular junctions on mature neuromuscular synapses), nAChR α1 subunits are involved in binding ACh and channel gating to control contraction through Ca²⁺ signaling (45, 46). In rat bone, Chrna1 is down-regulated after mechanical loading, indicating potential mechanical regulation of nAChRs in bone cells (44). Hence, ACh may have an influence over osteocyte Ca²⁺ dynamics at the cellular level and consequently bone mechanoadaptation at the tissue level. Rapsyn, also expressed by osteocytes, is an intracellular scaffolding protein that tightly regulates nAChR mobility on the membrane and clusters nAChRs into functional postsynaptic groups, encoded by the gene Rapsn (47, 48). This cytosolic protein is critical for proper ACh signaling and is highly expressed in osteocytes (38); however, very little is currently known about its role in osteocyte function. While compelling evidence links ACh to bone biology, the precise mechanisms underlying ACh-induced tissue-level bone architecture changes—specifically whether they result from direct osteocyte signaling or other mediated pathways—remain unclear. Hence, there is a need for a deeper analysis of the impacts of altered cholinergic signaling in osteocytes on bone mechanotransduction.

We believe that the osteocyte serves as a functional neurotransmitter signaling target of the central nervous system in bone. The goal of this study was to determine whether cholinergic signaling components nAChR subunit α1 and rapsyn in osteocytes are relevant for bone development and mechanoadaptation. We hypothesized that disrupted osteocyte cholinergic signaling negatively affects bone tissue organization and responses to mechanical loading. To investigate this question, we generated osteocyte-targeted conditional knockouts (cKOs) for either Chrna1 or Rapsn in mice. We collected skeletal morphometric measurements throughout postnatal development and after anabolic mechanical loading in vivo. We also performed an in vivo live cell functional assay to determine changes to osteocyte mechanosensitivity in their native environment, with nervous system interactions intact. Our results showed sexually dimorphic differences in bone structure and formation rates between Cre-negative controls and cKO mice. In females, reductions to bone geometry were rescued with anabolic loading but not in males. Female Chrna1 cKO osteocytes were less mechanosensitive at lower forces, perhaps providing a mechanistic explanation for our other results. Our findings provide evidence supporting the role of osteocytes as responsive elements in cholinergic signaling, affecting bone mechanoadaptation and may assist in defining therapeutic targets for bone metabolic diseases (e.g., osteoporosis).

RESULTS

Cholinergic nerve fibers are proximate to cortical bone osteocytes in situ

Work from others has reported the presence of cholinergic fibers in bone using histological sections, with a particular emphasis on bone directly beneath articular cartilage at the ends of bones (36–39). Our group has developed an approach for whole-mount three-dimensional (3D) imaging of mouse bones to assess the presence of cholinergic nerve fibers. We used a modified BABB (benzoic acid benzyl benzoate) clearing approach (49, 50) to achieve excellent clarity and image quality in intact tissues compared to previous protocols (50, 51). We combined this clearing approach with bulk immunolabeling of cholinergic fibers (green) and adrenergic fibers (red) in mouse third metatarsals (MT3s). The cleared and labeled bones were imaged via light-sheet microscopy. In a representative image, it can be seen that cholinergic fibers penetrated the MT3 throughout the length of the bone, with numerous transcortical events at the midshaft (Fig. 1). An extensive network of labeled structures can be seen in the marrow space. Adrenergic fibers can also be seen with density and arrangement at the ends of the bones that replicate data reported by other groups (36, 37, 52). These tissue level imaging data provide strong evidence for the presence of cholinergic nerve fibers in cortical bone and, thus, supports the possibility for functional and/or direct contact between cholinergic nerve fibers and osteocytes.

Fig. 1. Fluorescent light-sheet imaging of a cleared MT3.

Representative 3D projection light-sheet microscopy images of a cleared bulk immunohistochemistry-stained MT3 with anti–vesicular ACh transporter (green) and tyrosine hydroxylase (red) representing cholinergic and adrenergic nerve fibers, respectively. Clear, distinct nerve labeling can be seen at the distal and proximal ends of the bone. At the midshaft, transcortical innervation can be observed, with extensive branching throughout the distal half of the bone. White arrows indicate transcortical cholinergic fibers.

We used an intravital imaging approach to interrogate the prospect of direct contact between cholinergic nerve fibers and osteocytes. Mice expressing GCaMP6f in a CHaT-dependent manner (Strain# 028865, 006410; Jackson Laboratory) were injected with Cy5 fluorescent nanoparticles that are readily taken up by osteocytes in vivo (53). The nanoparticles were injected subcutaneously at the volar aspect of the hind paw and incubated for 45 min. The bones of anesthetized mice were surgically exposed and imaged using two-photon microscopy. Very fine green cholinergic nerve fibers were seen throughout the cortex (Fig. 2A). These fibers occasionally interfaced directly with red-labeled osteocytes, as indicated by yellow overlap signal (Fig. 2, B and C). When the representative z-stack is rotated 90°, the interaction between some osteocytes and nerves becomes readily apparent (fig. S2). We evaluated the distribution of osteocytes within 100 μm of cholinergic nerve fibers in metatarsal cortical bone in mice in vivo (Fig. 2, D to F). Female mice exhibited a more skewed distribution compared to males (Fig. 2, E and F, and fig. S3). These data validate the presence of cholinergic nerves in the osteocyte microenvironment and suggest a close spatial association between nerves and osteocytes, indicating a potential for intercellular communication. Moreover, they show that embedded nerve fibers are present within mineralized matrix tissue in cortical bone.

Fig. 2. Osteocytes are positioned close to cholinergic nerves in vivo.

Intravital z-stack image (40 μm thick) of metatarsal bone osteocytes in mice expressing GCaMP6f in ACh transferase–expressing cells taken using two-photon microscopy at about 20 μm below the periosteal surface (A). The red fluorescence signal is driven by a 15-μl volume of ultrasmall RGD-functionalized integrin targeting nanoparticles (5 to 7 nm) at 10 μM concentration. This static 3D image indicates the colocalization of osteocytes, labeled in red, with cholinergic nerve fibers, labeled in green. (B and C) Yellow and blue insets depict single 2D frames where the nerve can be seen directly interacting with osteocytes. (D) Representative images for nerve-osteocyte proximity analysis showing the input image with osteocytes in red and ChAT signal containing cholinergic nerve fibers in green. The masks were drawn over the nerve fibers of the input image, and proximity analysis of the shortest distance between each osteocyte and a nerve was calculated and represented in the output image. (E) Histogram with an overlaying Gaussian curve showing the distances of osteocyte distribution around the nerve in both sexes. (F) Both males and females fail the normality test (P value < 0.0001), and a significant difference (P value = 0.001) exists in the distribution of osteocytes between the sexes (fig. S3). Kolmogorov-Smirnov test was used for normality assessment. Mann-Whitney nonparametric test was used for comparing between sexes.

Osteocyte-targeted deletion of ACh receptor components alters load-induced Ca²⁺ dynamics

We observed osteocyte Ca²⁺ signaling responses to mechanical loading in vivo to test the functional impact of cholinergic signaling on osteocyte mechanotransduction. Osteocyte-targeted GCaMP6f-expressing mice were generated with concurrent conditional deletion (cKO) of Chrna1 or Rapsn. Osteocytes were imaged intravitally for each mouse group at the dorsal mid-diaphysis of the MT3 during active mechanical loading. We investigated the strain dose response by subjecting bones to strain values between 250 and 3000 με. For all groups, the percentage of responding osteocytes as indicated by Ca²⁺ signaling increased with increasing strain levels as expected on the basis of previous studies (Fig. 3) (30). In female mice, Chrna1 cKO mice had fewer responsive osteocytes at 250 and 500 με loading. In addition, Rapsn cKO mice had a lower number of responding cells at 250 με (Fig. 3C). These data indicate a decreased mechanosensitivity among osteocytes at lower force levels in vivo. There were no differences to the change in fluorescent intensity for the responding cells in female mice.

Fig. 3. In vivo osteocyte mechanotransduction is affected by Chrna1 and Rapsn cKO in a sexually dimorphic way.

Intravital imaging of osteocyte Ca²⁺ signaling responses to mechanical loading was used to assess molecular level functional impacts of either Chrna1 or Rapsn cKO. (A) In males, the percentage of responding osteocytes decreased with Chrna1 cKO at very low strains but increased at higher strains for both Chrna1 and Rapsn cKO. (B) Rapsn cKO also increased response intensity among responding cells at three of the five strain levels tested. These results indicate dysregulation of mechanosensation in male osteocytes with ACh signaling disruption. (C and D) While female response intensities were unaltered, the number of responding cells was decreased at lower strain values tested for Chrna1 cKO mice compared to controls. t tests were performed across groups within each strain level (*P < 0.05; **P < 0.01). Sample sizes are as follows: % Cells responding: (GCaMP: 7 to 9, Chrna1: 9 to 10, Rapsn: 12 to 14). Delta F: (GCaMP: 3 to 9, Chrna1: 2 to 10, Rapsn: 4 to 14).

The results in male mice were categorically different. The number of responsive cells at lower magnitude values was largely unchanged, except for a slight but significant reduction in Chrna1 cKO mice at 250 με compared to controls. At 1000 and 2000 με, Rapsn cKO males had a higher number of responsive osteocytes compared to controls (Fig. 3A). Likewise, Chrna1 deletion led to an increase in responding cells at 1000 με (Fig. 3A). Moreover, the intensity among responding cells was increased at 500, 2000, and 3000 με for Rapsn cKO mice (Fig. 3B). These results indicate a general trend for increased mechanosensitivity in male mice with constitutively dysfunctional ACh signaling. This trend appears to be centered on the mid-high range of physiological loading magnitudes.

Endogenously expressed GCaMP6f signal in osteocytes was used to evaluate 2D cellular morphology in vivo. There were no differences in the number of cells between any of the groups (Fig. 4A). Female Chrna1 cKO osteocytes had a slightly higher cell area when compared to GCaMPf controls (Fig. 4B). For both sexes, Chrna1 cKO cells were more elongated and less round compared to controls (Fig. 4, C to F). Rapsn cKO did not alter 2D osteocyte morphology in females (Fig. 4, A to F). In Rapsn cKO males, again, cells were more elongated (Fig. 4, H, J, and K). These results suggest that Ach signaling disruption results in altered osteocyte shape, which could explain the impacts on in vivo osteocyte mechanotransduction.

Fig. 4. In vivo osteocyte morphology is affected by ACh signaling disruption in a sexually dimorphic way.

Endogenously expressed GCaMP6f signal in osteocytes was used to evaluate 2D cellular morphology in vivo. (A) There were no differences in the number of cells per area between the groups. (B) Female Chrna1 cKO osteocytes have a slightly lower cell area when compared to GCaMP controls. For both females (C to F) and males (G to L), Chrna1 cKO cells were more elongated and less round than controls. (A to F) Rapsn cKO did not alter 2D osteocyte morphology in females. In Rapsn cKO males, cells were again more elongated (J to L). These results suggest that ACh signaling disruption produces altered osteocyte shape, which could explain the impacts on in vivo osteocyte mechanotransduction. Two-way analysis of variance (ANOVA) with Tukey’s multiple comparison test was performed with cKO and GCaMP controls to determine significance (*P < 0.05; **P < 0.01, ***P < 0.001, ****P < 0.0001) for each metric. N = 4 to 69 cells per mouse, group sample size (GCaMP: 7 to 9, Chrna1: 9 to 10, Rapsn: 12 to 14).

Osteocyte-targeted Chrna1 conditional deletion leads to site-specific skeletal abnormalities

The α1 nicotinic subunit is responsible for ACh binding during receptor activation. To assess the impact of α1 nicotinic cholinergic receptor function in osteocytes, we achieved osteocyte-targeted cKO of Chrna1 using a 10-kb DMP-1 targeted Cre-Lox system (fig. S1). Dual-energy x-ray absorptiometry (DEXA) scans of the whole body, hindlimbs, and lumbar spine were performed at intermittent periods between the ages of 4 and 21 weeks. Comparisons were made between Chrna1^{∆Ocy/∆Ocy} mice and littermate aged-matched Chrna1^+/+ controls. We found that Chrna1^{∆Ocy/∆Ocy} mice generally formed skeletons with normal BMD compared to Chrna1^+/+ controls (Fig. 5, A to F). The only exception was in older female mice, where cKO mice had increased BMD in the lumbar spine at 17 weeks (+10.6%; Fig. 5B) and 21 weeks (+11.7%; Fig. 5B). Microcomputed tomography (μCT) analysis of adult tibiae indicated sexually dimorphic site-specific differences in tissue development (Fig. 6, A to J, nonloaded, and fig. S4). Chrna1^{∆Ocy/∆Ocy} females exhibited reduced bone mass at the metaphysis compared to Chrna1^+/+ mice. This included reductions in trabecular bone volume over total volume (−32.7%; Fig. 6B), the number of trabeculae (−10.6%; Fig. 6C), and the density of total volume (−20.4%; Fig. 6D). By contrast, Chrna1^{∆Ocy/∆Ocy} males exhibited a midshaft phenotype with reduced cortical bone parameters. They had midshafts with lower mean total area (−15.2%; Fig. 6G), resulting in lower minimum moment of inertia (−33.6%; Fig. 6H), lower maximum moment of inertia (−29.7%; Fig. 6I), and lower polar moment of inertia compared to Chrna1^+/+ males (−31.1%; Fig. 6J). These results collectively indicate that disrupting cholinergic signaling in osteocytes using constitutive deletion of Chrna1 or Rapsn leads to slight sexually dimorphic differences in skeletal development (tables S1 and S2). While the results are not uniform between the sexes, there is a clear role for osteocyte ACh signaling in bone development.

Fig. 5. Chrna1 cKO produces mostly normal skeletons during development.

Bone mineral content and density were measured using DEXA for the whole body, lumbar spine, and right hindlimb between 4 and 21 weeks of age. We found no differences between control and Chrna1 cKO groups in either female (A to C) or male (D to F) mice. Two-way ANOVA was performed with interactions between genotype and age. N = 6 to 15 per group. A post hoc Bonferroni test was computed across genotypes for each time point to determine significance (*P < 0.05; **P < 0.01). N = 6 to 13 per group and skeletal location.

Fig. 6. Chrna1 cKO impaired the anabolic response to mechanical loading.

Adult mice were subjected to tibial loading, and bones were excised 12 days later. μCT measured tibial geometry for females (A to F) and males (G to J) to compare nonloaded limbs (white background) with loaded limbs (shaded gray background). Female Chrna1 cKO mice decreased trabecular bone measures (B-D) at baseline. Likewise, male Chrna1 cKO mice had a decreased midshaft bone phenotype (H to J). Female trabecular bone was recovered with loading (B to D). There was no change in male midshaft metrics with loading (H to J). (K to P) Fluorochrome labeling with calcein followed by alizarin red was used for osteoblast-derived indices of dynamic histomorphometry. Relative metrics were computed by subtracting nonloaded limbs from loaded limbs. Indices were calculated for both the endosteal and periosteal surfaces. While some load-induced formation differences were observed, the overall changes in bone formation rates between control and cKO groups were modest. Periosteal mineralizing surface was decreased in female cKO mice and increased in male cKO mice (K). Student t tests were performed between cKOs and Cre-negative controls to determine significance (P < 0.05). (P < 0.05). N = 8 to 22 per group and assay. Relative formation parameters for loading effects were calculated for each mouse by subtracting the nonloaded (left tibia) values from the loaded (right tibia) values. *P ≤ 0.05; **P ≤ 0.01.

Osteocyte-targeted Chrna1 deletion impairs bone anabolic response to mechanical challenge

To assess the impact of osteocyte cholinergic signaling on bone mechanotransduction, we studied tissue changes in response to in vivo loading over time. Mice received in vivo tibial loading to peak calibrated strain magnitudes of −2250 με three times per week for 2 weeks. Postmortem analyses of loaded and nonloaded tibia included μCT and dynamic histomorphometry. The anabolic impact of mechanical loading was present but mild in control mice (fig. S4). As noted above, nonloaded baseline conditions in tibiae from Chrna1^{∆Ocy/∆Ocy} females exhibited reduced trabecular bone mass at the metaphysis compared to Cre-negative littermate controls. This phenotype was rescued in loaded bones (Fig. 6, A to E, loaded), with trabecular total volume (Trab. TV; Fig. 6F, loaded) even exceeding that of Chrna^+/+ controls by +4.2%. Nonloaded tibiae from Chrna1^{∆Ocy/∆Ocy} males had normal trabecular bone but exhibited a midshaft phenotype with reduced cortical bone area and moment of inertia. Unlike females, anabolic loading did not rescue the phenotype in male mice (Fig. 6, G to J, gray background). Relative dynamic histomorphometry indices were assessed by subtracting nonloaded values from loaded limbs. Chrna1 cKO led to a slight decrease in periosteal formation rates compared to controls in female mice. By contrast, this measure was increased in male cKO mice (Fig. 6K). No other differences were found in load-induced formation markers for either sex (Fig. 6, K to P).

Osteocyte-targeted Chrna1 deletion affects toughness in male femurs

We next set out to determine whether there would be changes to the whole-bone mechanical properties at skeletally maturity. Femurs from 16- to 18-week mice underwent monotonic three-point bending to failure. Female Chrna1^{∆Ocy/∆Ocy} mice were not different than Cre-negative littermate controls (Fig. 7, A to D). Conversely, male Chrna1^{∆Ocy/∆Ocy} mice had greater plastic deformation, indicating that their bones were more ductile and resistant to mechanical failure. Male Chrna1^{∆Ocy/∆Ocy} mice had 228% increased post–yield displacement (Fig. 7F), 29.1% reduced fracture force (Fig. 7G), and +98.7% increased failure energy (Fig. 7H) compared to Chrna1^+/+ controls. Fourier transform infrared (FTIR) tissue composition analysis found no changes to the mineral-to-matrix ratio, collagen maturity, or mineral crystallinity for cKO compared to Cre-negative controls for either sex (Fig. 7, I to J). Together, these results firmly support a role for osteocyte ACh signaling in formation and/or regulation of bone material properties in male mice.

Fig. 7. Chrna1 cKO increased femur ductility and toughness in males.

(A to D) Mechanical properties were assessed with load-displacement curves from monotonic loading to failure. There were no differences in female bones. (E to H) Male Chrna cKO mice had increased post–yield deflection and total work to failure, indicating more ductile tissue properties. FTIR was used to evaluate bone compositional properties from a subset of left femora from mice used for mechanical testing for females (I) and males (J). No differences were found between the control and cKO groups. Student t tests were performed with cKO and Cre-negative controls to determine significance (P < 0.05) for each metric. N = 3 to 15 per group and assay. **P ≤ 0.01; ***P ≤ 0.001.

Osteocyte-targeted Rapsn cKO leads to site-specific skeletal abnormalities

Rapsyn clusters cholinergic ion channels on the cell membrane, allowing for coordinated receptor/channel function and more potent responses. We repeated the same experiments performed above with Chrna1 cKOs in mice lacking Rapsn in osteocytes (fig. S1). To study growth and development, mice with Rapsn cKO had DEXA scans of the whole body, hindlimbs, and lumbar spine over time. We found that cKO mice of both sexes largely formed skeletons with normal BMD compared to Cre-negative littermate controls (Fig. 8, A to F). Males exhibited a spike in BMD at 6 weeks in the lumbar spine that resolved and normalized in subsequent weeks (Fig. 8E). We also evaluated tibia morphology using μCT analysis (tables S3 and S4). We observed decreases in bone volume indices in female bone, including cortical volume (−5.1%; Fig. 9A), cortical total volume (−8.3%; Fig. 9B), and the total volume (−8.3%; Fig. 9E). Rapsn cKO mice had increased BMD compared to controls (+1.3%; Fig. 9D). Male Rapsn^{∆Ocy/∆Ocy} mice had increased bone volume density at the midshaft (+1.2%; Fig. 9H) compared to Cre-negative littermate controls but exhibited no other differences. Again, we see here sexually dimorphic impacts of ACh signaling disruption in osteocytes (tables S8 to S10).

Fig. 8. Rapsn cKO produces normal skeletons during development.

Bone mineral content and density were measured using DEXA for the whole body, lumbar spine, and right hindlimb between 4 and 21 weeks of age. We found no differences between control and Chrna1 cKO groups in either female (A to C) or male (D to F) mice. Two-way ANOVA was performed with interactions between genotype and age. N = 3 to 7 per group. A post hoc Bonferroni test was computed across genotypes for each time point to determine significance (*P < 0.05).

Fig. 9. Rapsn cKO impaired bone formation in response to mechanical load.

Adult mice were subjected to tibial loading, and bones were excised 12 days later. μCT measured tibial geometry for females (A to F) and males (G to H) to compare nonloaded limbs (white background) with loaded limbs (shaded gray background). Female Rapsn cKOs mice at baseline had lower overall volume indices (A and B) but increased BMD (D). These results imply increased tissue hardness to compensate for lower volume/mass. These values are normalized with anabolic mechanical loading. (G) Male Rapsn cKO mice had increased tissue density at baseline compared to controls. (H) This difference resolved with loading, while trabecular tissue density increased. (I to P) Fluorochrome labeling was used for osteoblast-derived indices of dynamic histomorphometry. Relative metrics were computed by subtracting nonloaded limbs from loaded limbs. Indices were calculated for both the endosteal and periosteal surfaces. Scale bars, 200 μm or 50 μm (inset). Female Rapsn cKOs exhibited reduced bone formation rates (I to L), suggesting decreased mechanosensitivity. No differences were found between male control and cKO mice. Student t tests were performed between cKOs and Cre-negative controls to determine significance (P < 0.05). N = 4 to 12 per group and assay. *P ≤ 0.05; **P ≤ 0.01.

Osteocyte-targeted Rapsn deletion impairs bone anabolic response to mechanical challenge

To assess the impact of Rapsn on bone mechanotransduction, we studied tissue changes in response to in vivo loading. Mice received in vivo tibial loading to peak calibrated strain magnitudes of 2250 με three times per week for 2 weeks. Postmortem analyses of loaded and nonloaded tibia included μCT and dynamic histomorphometry. Nonloaded tibiae from Rapsn^{∆Ocy/∆Ocy} females exhibited deterioration in the metaphyseal cortex. Loading rescued this phenotype, returning all deficiencies to control values (Fig. 9, A, B, and E, gray background). However, loading was not beneficial overall. Rapsn cKO mice suffered decreased bone volume fraction (BV/TV) with anabolic loading compared to control mice (−4.1%; Fig. 9C). These results indicate an increase in the total volume (or cross-sectional area) of the bone during load-induced adaptation. In male Rapsn^ΔOcy/ΔOcy mice, loading only caused increased bone volume density in the metaphyseal trabeculae (+1.8%; Fig. 9H).

Relative dynamic histomorphometry indices were assessed by subtracting nonloaded values from loaded limbs. The anabolic impact of mechanical loading was present but mild in control mice (fig. S5). Rapsn^ΔOcy/ΔOcy females suffered from lower metabolic rates on the periosteum compared to Cre-negative littermate controls, with suppression of relative mineral apposition rate (Ps.rMAR) by −70.6% (Fig. 9I), relative mineralizing surface (Ps.rMS/BS) by −25.6% (Fig. 9K). Females also had reduced endosteal relative bone formation rate (−55.7%; Fig. 9L). In males, Rapsn^ΔOcy/ΔOcy mice showed no differences compared to Cre-negative littermate controls.

Osteocyte-targeted Rapsn cKO maintains or minimally increases whole-bone level mechanical competence in femurs

We next set out to determine whether there would be changes to the whole-bone mechanical properties at skeletally mature bone mass (Fig. 10). Femurs from 16- to 18-week mice underwent monotonic three-point bending to failure. Female mice did not show any differences in mechanical properties, whereas Rapsn^{∆Ocy/∆Ocy} males exhibited +10.9% higher ultimate force than controls (Fig. 10G). After measuring the tissue composition of a subset of femurs used for mechanical testing, we determined no changes to the mineral-to-matrix ratio, collagen maturity, or mineral crystallinity for cKO compared to Cre-negative controls (Fig. 10, I to J).

Fig. 10. Rapsn cKO did not alter whole-bone mechanical properties.

Mechanical properties were assessed with load-displacement curves from monotonic loading to failure. (A to D) There were no differences in female bones. (E to H) Male Rapsn cKO mice had increased ultimate and failure forces, indicating a slight improvement in tissue strength. FTIR was used to evaluate bone compositional properties from a subset of left femora from mice used for mechanical testing for females (I) and males (J). No differences were found between the control and cKO groups. Student t tests were performed with cKO and Cre-negative controls to determine significance (P < 0.05) for each metric. N = 5 to 12 per group and assay. *P ≤ 0.05; **P ≤ 0.01.

DISCUSSION

Here, we report evidence of direct physical interaction between embedded osteocytes and cholinergic nerve fibers in vivo. These exciting findings imply a functional interaction between the peripheral nervous system and osteocyte networks. Osteocytes use Ca²⁺ signaling to translate tissue level mechanical load magnitude information into biological signals (30). nAChRs are Ca²⁺ channels that help to regulate membrane potential in muscles, thus determining one aspect of muscle activation sensitivity (54, 55). Given that osteocytes are primarily mechanosensors and the role for nAChRs in Ca²⁺ signaling, we suggest a novel signaling axis wherein cholinergic nerves regulate some aspects of osteocyte mechanosensitivity. This paradigm is supported by studies reporting anabolic impacts of cholinergic therapeutics on fracture risk (34, 35) and bone mass (36, 37). Moreover, our work is corroborated by work from others showing that osteocytes respond to ACh stimulation (56) and express nAChRs genes (38) that are differentially expressed after mechanical loading (44). Our data test the impact of osteocyte-targeted conditional deletion of ACh receptor components. In addition, we use novel intravital imaging approaches, providing insights into live osteocyte functional outcomes in vivo within their physiological context, where endocrine and paracrine influences are naturally present. These data work to fill a critical knowledge gap in basic bone biology. For most tissue systems, we understand the homeostatic balance achieved by parasympathetic and sympathetic signaling. In bone, however, much more is known about the role of sympathetic signaling (57–59) than parasympathetic and/or cholinergic signaling. The data offered here show evidence that there is a functional link between osteocytes and cholinergic signaling, suggesting a new physiological link that could be important for bone homeostasis and, thus, novel therapeutic targets.

An important consideration includes the spatial patterning of neurons in the bone and their relation to bone cells. Using a tissue-clearing protocol produced by our laboratory, we successfully cleared whole bones for 3D characterization of skeletal innervation. Cavities that acted as entry points into the marrow space for large nerve fibers were found to penetrate various regions of the cortical shell, with higher densities at the distal and proximal ends compared to the midshaft. The embedded position of these nerves suggests opportunity for neurotransmitter communication with osteocytes. Moreover, we showed plausible interactions between cholinergic nerves and osteocytes. The nature of those interactions remains an open question. It is possible that direct synaptic formation between osteocytes and Ach nerves exist. Alternatively, osteocyte-nerve interactions may be facilitated by paracine signaling within the osteocyte lacunocanalicular system. Osteoblasts express both choline acetyltransferase and vesicular acetylcholine transporter (39, 60), indicating a possibility of osteoblast-derived Ach signaling to osteocytes. Notably, this last point is limited by the relatively small number and surface position of osteoblasts. The exciting initial functional insights reported here motivate future studies aimed at defining the details of this physical interaction as well as outlining the nature of the functional behavior at those sites. One meaningful limitation to these data is that only larger positively stained bodies are currently resolved. Future work should include larger load-bearing long bones, such as the femur and tibia.

We achieved conditional osteocyte-targeted deletion of Chrna1 and Rapsn, which caused impairments to bone structure and bone anabolism in a sexually dimorphic manner. Female Chrna1 cKOs showed deteriorated microarchitecture tibial trabecular bone, while female Rapsn cKO showed reductions cortical bone. Male Chrna1 cKO mice suffered from smaller cortical shells at the midshaft in the tibia with increased ductility and toughness in the femurs. Broad phenotypic changes in BMD during skeletal development were not observed. The site- and sex-specific impact of our cKO models highlights the complexity of the osteocyte-ACh signaling axis and implies redundant regulatory mechanisms. This is common for other key features of bone mechanotransduction like membrane hemichannels (61, 62) and integrins (23, 63). High-resolution μCT data were collected at the end of our experiment from adult skeletally mature mice. AChE levels peak in early development to guide endochondral ossification before drastically declining, maintaining low levels for most of the life (60). Hence, it is possible that more drastic or different phenotypic changes are present in earlier developmental stages than what we tested here. Further investigation including earlier ages (<4 weeks) is needed to interrogate periods of elevated developmental cholinergic signaling. Likewise, it is important to follow these studies up with data from aging mice. Reduced ACh and increased AChE levels may occur in the elderly and are both associated with age-related diseases such as dementia (64, 65). Thus, how cholinergic signal affects bone through aging is also an important line of study that needs to be followed.

In males, increased mechanosensitivity clustered in the mid-physiological strain range, whereas females exhibited reduced mechanosensitivity at lower strain magnitudes. Factors determining osteocyte mechanosensitivity in vivo remain an active area of research. Key molecular players identified predominantly from in vitro studies include ion channels, integrins, connexins, and extracellular matrix organization (66–69). The prevailing hypothesis is that these elements collectively establish individual cellular thresholds for mechanosensitivity (70). Thus, it is plausible that deletion of Chrna1 or Rapsn alters the balance of osteocyte mechanosensitivity and/or modifies the strain threshold required to trigger cellular responses. Future research should explore the precise influence of osteocyte cholinergic receptor function on specific mechanotransduction pathways to clarify these relationships. An important implication of our findings is that modulation of cholinergic signaling can finely tune osteocyte mechanosensitivity, presenting a potentially powerful strategy for developing targeted clinical therapies for bone metabolic diseases.

Mechanotransduction is a multifaceted process involving the interplay of ion channels, integrins, proteoglycans, and ligand receptors, including ACh receptors. Our results suggest that disruption of ACh receptor signaling alters the balance among these components, leading to modified osteocyte responses to mechanical loading. However, the use of constitutive knockout models presents interpretive limitations due to potential compensatory or developmental adaptations. Hence, establishing direct mechanistic links remains challenging. In the work reported here, we show that deletion of Ach components Chrna1 and Rapsn in mouse osteocytes leads to strain-dependent differences in osteocyte mechanotransduction. We suspect these results to reflect cellular adaptation to the deficiencies induced. This phenomenon has precedent in the literature, including osteocyte-specific cKO of integrins (71), pannexins (72), and connexins (73). Future studies should incorporate inducible osteocyte-specific knockouts and temporally controlled pharmacological interventions to dissect the acute versus adaptive roles of ACh signaling. These approaches would provide higher temporal resolution and allow for a more precise understanding of the causal contributions of cholinergic pathways to osteocyte mechanotransduction. With this work, we introduce an emerging signaling axis that suggests that ACh can act on osteocytes in vivo to alter bone mass.

Our anabolic mechanical loading studies revealed that cKO-driven deficiencies in female bone morphology could be rescued with tibial loading. Counterintuitively, Rapsn^{∆Ocy/∆Ocy} females also had suppressed bone formation and mineral apposition rate, indicating an alternative recovery mechanism. Since the present study only measured osteoblast-derived indices, we cannot disregard the possibility that alterations in osteoclast bone resorption are involved. ACh stimulation of MLO-Y4 cells increases the Receptor activator of nuclear factor kappa-Β ligand (RANKL) and Osteoprotegerin (OPG) expression ratio 60-fold (56). Adding either mecamylamine, a noncompetitive antagonist that inhibits all known nAChRs, or d-Tubocurarine, a nondepolarizing competitive antagonist that acts particularly on muscle nAChRs, ablates the response. Pyridostigmine, an AChE inhibitor that is not blood-brain barrier permeable (i.e., solely peripheral acting), increases RANKL/OPG serum levels and reduces femoral bone volume (74). Thus, nAChR activation on osteocytes may produce signals for osteoclast activity, implying that Rapsn cKO could inhibit bone resorption. In contrast to females, any reductions in bone mass in the Chrna1^{∆Ocy/∆Ocy} males remained significantly lower even with loading, highlighting fundamentally distinct ACh-mediated bone metabolism regulation between the sexes.

Our three-point bending mechanical tests indicated that Chrna1 deletion in male osteocytes produced more ductile bone; however, material analysis of the bone matrix did not indicate any changes to overall composition. These data suggest that any changes in mechanical behavior seen in male femurs likely come from tissue arrangement. Further, they suggest that an impact of osteocytes on tissue arrangement requires appropriate ACh signaling. Our studies here are limited in that we have used a constitutive model, which means that any effect will be present through growth and development of bone tissue. To understand this phenomenon better, it would be helpful to use a temporally controlled cKO model in the future to allow bone tissue to develop normally.

Sex differences from cKOs imply different roles for ACh signaling, corroborating evidence that cholinergic signaling has a greater impact in females than in males. The negative bone measures associated with Alzheimer’s, which in general reduces ACh levels in the body, are mostly found in women and not men (31–33, 74). In general, women are more susceptible to both bone loss (i.e., postmenopausal osteoporosis) and dementia. A potential underlying mechanism for these sex-specific variations may involve the interplay with estrogen, as highlighted by the increased risk of osteoporosis with estrogen depletion. Our findings suggest a role for estrogen in modulating cholinergic regulation of bone, a concept that warrants further in-depth investigation. Our results corroborate other studies on nAChRs in bone that found sex differences. For example, α7 global knockout does not produce changes in males (75), but in females, results in increased cortical area and thickness and reduced calcium content (76) as well as reductions in osteoclast count, RANKL serum levels, and increases to OPG serum levels (77). α9 global knockout in females leads to an increased number of apoptotic osteocytes (78). α2 global knockouts in females suffer from reduced trabecular bone volume and increased osteoclast count as well as resorption activity (36). Additional studies are needed to interrogate the potential for these additional subunits as osteocyte mechanotransduction regulators and as sex-specific regulators of bone mechanotransduction more broadly.

We expect the osteocyte cholinergic receptor to have a similar conformation as muscle nAChRs based on the component gene expression (38). These structures are heteropentameric with the stoichiometry (α1)₂β1γδ (75). The α1-ε and α1-δ interfaces act as the binding sites for ACh, and the deletion of key hydrophobic regions to α1 abolish receptor function; thus, α1 knockout in osteocytes presumably inhibits nAChR function, although there may be stoichiometric substitution of the α subunit that we are not aware of that provide altered/limited receptor function. Rapsyn deletion permits healthy individual receptor formation while reducing overall signaling potency on account of increased receptor mobility. Rapsyn is also known to assist in faster recovery rates (79), meaning that deletion may result in decreased signaling efficiency and effect. Future studies should include inducible deletion and pharmacological blockage of cholinergic signaling in bone to prevent any potential subunit substitutions. Follow-up RNA analyses would be helpful to identify any substitutions that occur with constitutive deletion.

Chrna1 and Rapsn cKOs are unlikely to ablate cholinergic signaling entirely in the osteocyte as other cholinergic receptors may be present (38, 56). Despite this, we have seen definitive evidence that osteocyte cholinergic receptors are important for bone development and response to mechanical challenges. This study did not trace the cholinergic source. This is important because ACh may be derived from a neuronal [e.g., parasympathetic/sympathetic nervous system (36, 37)] or nonneuronal [e.g., osteoblasts/bone lining cells (37, 39)] source, which has various biological implications and may also influence future therapeutic strategies. An important advance demonstrated in this study was the use of targeted genetic modification of ACh signaling in osteocytes specifically, rather than systemic pharmacological disruption or global knockout. Thus, we can show that ACh signaling to osteocytes influences bone morphology and mechanoadaptation. This indicates an important new signaling axis, potentially implicating autonomic nervous system modulation of osteocyte activity.

Given the escalating prevalence of osteoporosis, investigating osteocyte-based therapeutics is a critical next step to advancing treatments beyond what is currently possible. Cholinergic neuromodulators such as AChE inhibitors are already commonly used to treat neurodegenerative diseases such as Alzheimer’s and have been gaining traction as potential therapeutics for bone diseases such as osteoporosis. To get closer to clinical translation, there is still a much-needed understanding of how cholinergic signaling affects bone (specifically in osteocytes) and how this effect alters bone mechanotransduction. Our research is helping to elucidate the mechanism of cholinergic signaling in osteocyte mechanobiology and suggests a new signaling axis between brain and bone with far-reaching implications in bone biology that may be vital to treating bone disease.

MATERIALS AND METHODS

All methods are approved by Cornell Institutional Animal Use and Care Committee; protocol 2020-0046.

Whole-mount tissue clearing and immunofluorescent staining

We used a modified BABB tissue clearing approach. Mice were perfused with zinc formalin (contains 3.7% formaldehyde), and bones were dissected and fixed overnight in zinc formalin. Decalcification occurred over 5 days in 7.5 pH buffered 20% EDTA at 37°C with daily solution changes and decolorized with 15% hydrogen peroxide for less than 1 hour. Immunofluorescent staining occurred before further proceeding. Samples are blocked overnight with 10% donkey serum (Sigma-Aldrich, D9663-10ML) and 3% bovine serum albumin in Phosphate-Buffered Saline with Tween (PBST). Primary antibodies are incubated at 1:100 dilution in a blocking buffer for 48 hours. Primary antibodies used were rabbit anti–tyrosine hydroxylase antibody (Sigma-Aldrich, AB152) and goat anti–vesicular ACh transporter antibody (Sigma-Aldrich, ABN100). Samples were washed in PBST overnight. Secondary antibodies were then incubated at 1:200 dilution in PBST for 48 hours. Secondary antibodies used were Alexa Flour 488 donkey anti-goat antibody (Thermo Fisher Scientific, A11055) and Alexa Flour 546 donkey anti-rabbit antibody (Thermo Fisher Scientific, A10040). Samples went through a final wash in PBST overnight. Tissue clearing was then finalized with dehydration in 2-ethoxyethanol for 4 hours, followed by three changes of acetonitrile between 4 hours or overnight at room temperature. Bones were then cleared in benzyl alcohol for 4 hours before being placed in a final refractive index matching solution of benzyl benzoate (refractive index 1.5) for long-term storage and imaging.

Light-sheet fluorescent microscopy

Imaging of tissue-cleared third metatarsals (MT3s) was done using a Light-Sheet microscope (LaVision BioTec) at the Cornell Institute for Biotechnology Resource Center. Bones were mounted to clear hematocrit tubes using cyanoacrylate, and then the tubes were attached to a custom-made mount for the microscope. Samples were completely submerged in a container filled with refractive index–matching solution (benzyl benzoate and any residual benzyl alcohol; refractive index 1.5). Imaging was performed with two channels at excitation wavelengths of 488 nm (6 to 8% power, emission collected with band-pass filter of 525/50) and 561 nm (9 to 11%, emission collected with band-pass filter 620/60) using an Olympus 2x CDC objective with a 3.5 collar and Dynamic Background Extraction (DBE) correction for high refractive index (RI). Two light sheets illuminated the sample in series from either the right or left directions. Images were acquired with a dynamic horizontal focus processing that blended the two images into one image, optimally reducing the scattering compared to a single light sheet. Both light sheets were calibrated and aligned in the z plane before imaging. To image an entire bone, mosaic tiling with a 10% overlap and 1-by-3 tiles were used with 3.2× zoom, sheet NA 1.20, and sheet width 40%. The final pixel resolution was 0.94 μm by 0.94 μm by 4 μm. Files were saved in TIF OME format for each tile and channel, comprising six raw data images. Images are stitched using BigStitcher in FIJI [National Institutes of Health (NIH)]. Stitched images were converted into Imaris (Oxford Instruments) files. Channel shifting, 100-μm radius background subtraction, and 1.2 gamma correction are applied in Imaris. Last, using the filaments tool in Imaris, large nerve fiber-like objects were manually traced and quantified using Imaris metrics.

In vivo loading of the third metatarsal

Our group uses an innovative approach to observe live osteocyte cell and molecular level fluorescent signals in vivo. In addition, we can collect these observations with simultaneous tissue level mechanical loading (fig. S6). This approach is an extremely powerful approach for studying basic mechanobiology in vivo. Before loading, calibrations were conducted on cadaveric mouse MT3 with digital image correlation to determine position-controlled loading regimes with corresponding strain values. Hind paws from three to six mice for each mouse group were amputated and stored in saline-soaked gauze at −20°C until calibration. For calibration, the paws were thawed at room temperature in Dulbecco’s Phosphate-Buffered Saline (DPBS). The MT3 was surgically isolated as described elsewhere, and a 1-mm diameter cylindrical fulcrum pin was inserted just underneath, separating the bone from any muscle fascia and soft tissue (30). A bracket for the two side pins was then positioned over the paw and connected with a screw to the load cell and actuator. Calibration was performed for each genotype according to previously published work (30).

For experiments, mice were initially placed under anesthesia in a small container with a 1 liter/min flow rate and 3% isoflurane. The mice were then transferred onto the loading device stage with a heating pad and head covering for more direct anesthesia, where the isoflurane was maintained at 1.5 to 2%, at which point surgery in the same manner as with calibration took place on the left hind paw. With a live mouse, it was important to avoid the rupture of blood vessels and to remove any tendon that resides longitudinally across the metatarsals. The fulcrum pin and bracket were attached accordingly. The paw was placed in a room-temperature DPBS bath for the duration of the experiment.

Nerve-osteocyte proximity analysis in vivo

A total of six ChAT strain mice (four males and two females) were used in the study, all 10 months old. For labeling, the osteocytes in the MT3 bone, argenine glycine asparagine (RGD)–functionalized integrin targeting nanoparticles (RGD C-Dots) (15 μl of 10 μM) were injected via subcutaneous method in the right and left paws, respectively. Following injection, 5-min incubation was maintained for RGD C-Dots. Imaging was performed under anesthesia using a two-photon microscope (Thorlabs Bergamo II, Newton, NJ) and a 20× water immersion objective with a long working distance (XLUMPLFLN, Olympus) to assess fluorescence signals, with specific observations on ChAT expression and RGD C-Dots localization. The RGD C-Dots fluorescence was examined at multiple wavelengths, with stronger signals detected at 1020 nm.

Images containing nerve fibers and osteocytes were processed by first confirming file paths and loading corresponding binary masks. Osteocytes were identified by converting images to the hue, saturation, and value (HSV) color space and applying a dual-range threshold to isolate red-stained osteocyte structures, followed by contour extraction and centroid calculation. A secondary mask was applied to remove image regions dominated by background noise. Osteocytes overlapping nerve fibers or these masked noisy regions were excluded to enhance quantification accuracy. In addition, osteocytes located within 10 μm of each other were filtered to prevent duplicate counting. Nerve fiber boundaries were extracted from binary masks, and the shortest Euclidean distance from each osteocyte centroid to the nearest nerve fiber boundary was calculated using OpenCV’s cv2.pointPolygonTest. Pixel distances were converted to micrometers based on image scale bars. Each osteocyte was assigned to its nearest nerve fiber, and data including osteocyte coordinates, nerve assignments, and distances were recorded. Last, osteocytes were visually labeled and color-coded according to their nearest nerve fiber for validation purposes.

Given the large osteocyte sample size (n > 200), the Kolmogorov-Smirnov test was used to assess normality of data distributions separately for males and females (80). Nonparametric distributions (P < 0.05) were identified in both groups; therefore, statistical comparisons between males and females were conducted using the Mann-Whitney U test. Statistical significance was defined as P < 0.05.

Intravital imaging of calcium dynamics with two-photon microscopy

In vivo fluorescence signals of GCaMP6f from Ca²⁺ dynamics in osteocytes were visualized with a two-photon microscope (Thorlabs Bergamo II, Newton, NJ) and a 20× water immersion objective with a long working distance (XLUMPLFLN, Olympus). Osteocytes were imaged at the dorsal midshaft region of the MT3, directly above where the fulcrum pin was placed beneath the bone, and beneath the periosteal surface, verified with visible autofluorescence from the bone surface. Cyclic haversine position-controlled loading at 1 Hz was applied for 60 s to previously calibrated peak strains of 250, 500, 1000, 2000, and 3000 με. These levels encompass the range of strains that have been reported during physiological activities from in vivo strain gage studies, with strains up to 2000 με characteristic of habitual activities, while strains on the order of 3000 με are seen in extreme activities (4, 81). A 15-min wait time to account for osteocytes’ refractory period was instigated between each imaging/loading session (82). The region of interest (ROI) was 512–by–512 pixels with 2× zoom, leading to a pixel size of 0.545 μm by 0.54 μm. Excitation was at 920-nm wavelength from a tunable Ti-Sapphire laser (Coherent Chameleon Discovery NX) with a power selected such that the final output from the objective was between 30 and 35 mW. A photomultiplier with a 490- to 560-nm band-pass filter was used for detection with a voltage gain of 0.1 to 12 V. Time series images were acquired by averaging three frames with a sampling rate of 29.4 Hz, leading to a final frame rate of 9.8 fps. Across an approximate 150-s period, 1500 8-bit TIF images were collected, with 60 s of nonloading, 60 s of loading, and 30 s of resting.

MT3 loading image analysis

Intensity measurements were performed by postprocessing images using ImageJ. In-plane movements in the x and y were removed using a template-matching plugin. ROIs were selected with a semiautomated process. Gaussian blur with a sigma of 1 was applied to smooth edges. Automated size exclusion was performed, filtering ROIs for only 150 to 650 pixel sizes. Last, users would manually make final adjustments to ROIs by eye, deleting and adding ROIs as needed that were not properly captured by thresholding. An ROI indicating the background autofluorescence was included in an area where no cells were present. The average pixel intensity for each ROI was collected across all 1500 time series frames and output as a CSV.

In MATLAB, ROIs that exceeded a cutoff of the mean background intensity plus thee SDs were included in the final analysis as fluorescent osteocytes. The intensities for each cell of interest were detrended linearly to correct for photobleaching and normalized to the mean intensity for that cell over 40 s before loading. Peaks and troughs were identified using islocalmax and islocalmin in MATLAB with a prominence of 0.01. Amplitudes across the nonloaded and loaded intervals were computed by subtracting coupled troughs from peaks. Changes in amplitude fluorescent intensity were computed as follows: ∆F/F = (F_load − F_nonload)/F_nonload. F_load indicates the amplitude during loading, and F_nonload indicates the amplitude without loading. Responding cells were defined as cells showing a >25% increase in ∆F/F. ∆F/F was then plotted for responding cells only. Of 9222 ROIs (indicating cells with the possibility of repeats across different loading strains), 25 ROIs were removed as outliers with a change in fluorescent intensity greater than what we determine here as the dynamic range (>13) of GCaMP6f with repeated stimulation (83).

Generation of Chrna1 and Rapsn cKO mouse lines

Chrna1 and Rapsn deletion from the mouse genome was conducted by purchasing tm1a-targeted embryonic stem cells for both Chrna1 and Rapsn, from EUCOMM, which were expanded and injected into blastocysts, returned to pseudo-pregnant females, and screened for germline transmission in pups. Both Chrna1 and Rapsn germline deletion were lethal, consistent with previous studies (55, 84), so we took steps to generate Cre-ready conditional alleles (tm1c) for both loci (fig. S1). Briefly, the correctly targeted mice harbor LacZ and Neo cassettes in intron 2 (for Rapsn) or 3 (for Chrna1) flanked by flippase recognition target (FRT) sequences. The Neo cassette was flanked by loxP, and a third loxP was introduced after the final exon in the targeted sequence (exon 4 for Chrna1 and exon 5 for Rapsn). To convert the tm1a allele to a tm1c conditional floxed allele, tm1a mice were crossed to Rosa26Flp mice [Jax stock 012930; described elsewhere (85)] to induce germline recombination of the FRT sites and delete the LacZ/Neo/5’loxP insert, leaving one lonely FRT site and the targeted exon(s) flanked by loxP. These tm1c mice were then crossed to the Dentin matrix protein 1 Cre (10-kb Dmp1-Cre; Jax stock 023047) to induce conditional deletion of Chrna1 exon 4 or Rapsn exons 3 to 5 (both target regions house crucial exons) in osteocytes and late-stage osteoblasts. The Dmp1-Cre line has been described elsewhere (86). Chrna1 and Rapsn lines were validated for correct targeting by long-range PCR using primer pairs spanning the homology arms at both 5′ and 3′ ends of the targeting region. Both mutant mouse lines were on a pure C57BL/6J background. Female and male mice with Dmp1-Cre-positive cKOs in osteocytes are labeled Chrna1 cKO and Rapsn cKO, with littermate Dmp1-Cre–negative controls Chrna1^+/+ and Rapsn ^+/+ used as the respective reference.

Generation of osteocyte-targeted GCaMP6f mice with Chrna1 and Rapsn cKO

GCaMP6f is a recombinant protein construct containing calmodulin (CaM), the CaM-binding myosin light-chain kinase fragment (M13), and enhanced green fluorescent protein (EGFP). Bright EGFP fluorescence is observed following Ca²⁺ binding to CaM, causing a subsequent binding of CaM to M13 and a conformational change that results in increased fluorescence intensity. Mice exhibiting osteocyte-targeted expression of GCaMP6f were achieved by crossing Ai38 mice B6J.CgGt(ROSA)26Sortm95.1(CAGGCaMP6f)Hze/MwarJ (JAX Labs, strain no. 028865), which contain GCaMP6f DNA behind a Lox-STOP-Lox codon (83), with DMP1-Cre mice [B6N.FVB-Tg(Dmp1-cre)1Jqfe/ BwdJ; (JAX Labs, strain no. 023047)] (86). Both parent strains are on a C57BL/6 background. Previous studies with this mouse model used GCaMP3f and validated fluorescent expression in cortical osteocytes (30, 87). Our substitution with GCaMP6f was motivated by the fact that GCaMP6f was created as an ultrasensitive protein calcium sensor that outperforms previous GCaMP iterations with a variant that has faster kinetics and greater dynamic range (83).

Last, to obtain homozygous knock-in of GCaMP6f with simultaneous knock-out of either Chrna1 and Rapsn, we crossbred GCaMP6f mice with each cKO three generations deep to obtain GCaMP6f-Chrna1 cKO and GCaMP6f-Rapsn cKO mice. In this study, intact GCaMP6f mice were used as controls. For experiments, male and female mice were aged until skeletal maturity (16 to 18 weeks).

In vivo tibial loading

The axial tibial compression model was applied to Cre-positive and Cre-negative mice as previously described (88, 89). Before loading, calibration mice for each sex and genotype were euthanized at 17 weeks of age to collect strain measurements. Disarticulated hindlimbs previously stored at 20°C were used for strain gauge calibration. A single-element strain gauge (EA-06-015DJ-120; Vishay Precision Group, Malvern, PA, USA) was applied to the posterior midshaft surface of the tibia. The load-strain relationship was measured for each sample by applying step loading up to 11 N. A load-strain curve was derived from the simultaneous recording of the voltage output from the load cell and strain gauge. All tests were averaged within each genotype/sex to establish a calibration curve.

For in vivo anabolic loading, mice were anesthetized using isoflurane inhalation, and their right hindlimb (knee to paw) was loaded to a calibrated peak strain value of 2250 με using a haversine waveform (2 Hz, 180 cycles). Mice were given three bouts over 5 days with a day of rest between each bout. Mice were euthanized 12 days after the final bout.

Dual-energy x-ray absorptiometry

Bone mineral content (BMC) of the third to fifth lumbar vertebrae, the right leg, and the whole body was evaluated in vivo using DEXA. Mice were anesthetized via inhalation of 2.5% isoflurane (IsoFlo; Abbott Laboratories, North Chicago, IL) mixed with O₂ (1.5 liter/min) for approximately 8 min, including induction and scanning. The mice were placed prone on a specimen tray within the scanner. BMD was computed using raw BMC for each ROI. Scans were performed at 4, 6, 9, 13, 17, and 21 weeks of age. All BMC measures were normalized by body weight to eliminate the confounding effects of changing body size and weight during growth.

Microcomputed tomography

Tibiae were harvested 12 days after the final bout of tibial loading, wrapped in saline-soaked gauze, and stored in an airtight container at −20°C. Geometric properties of trabecular and cortical bone were evaluated using high-resolution μCT (Scanco Medical AG). Distal and mid-diaphysis 9-μm voxel size resolution scans were imported into Scion Image version 4.0.2 (Scion Corporation), where properties were calculated using standard and customized macros. Standard measurements included cortical area (mm²), maximum (Imax, mm⁴), and minimum (Imin, mm⁴) cross-sectional moments of inertia, polar moment of inertia (pMOI, mm⁴), total volume (TV, mm³), bone volume (BV, mm³), bone volume fraction (BV/TV, %), connectivity density (Conn.D, mm⁻³), structure model index, trabecular number (Tb. N, mm⁻¹), trabecular thickness (Tb. Th, μm), and trabecular separation (Tb. Sp, μm), BMC (grams), and BMD *mg/cm³).

Bone histomorphometry

An intraperitoneal injection of calcein was given 1 day after the final bout of tibial loading, followed by an intraperitoneal injection of alizarin complexone 8 days later. The mice were euthanized 3 days after alizarin injection. The right and left tibias were harvested and placed in 10% neutral-buffered formalin for 2 days and then stored in 70% ethanol at 4°C. Tibiae and femora were dehydrated in graded alcohols, cleared in xylene, and embedded in methyl methacrylate. To measure load-induced bone formation, thick sections were cut from the tibia approximately 3 mm proximal to the tibiofibular junction and ground down with a polisher to approximately 50 μm (Multiprep Polishing System Allied High Tech Products). Tibial diaphysis sections were mounted unstained to visualize and read the calcein and alizarin labels administered at 17 and 18 weeks. The sections were imaged on a fluorescent microscope using filter sets that provide excitation and emission for the calcein and alizarin wavelengths. Digital images were imported into ImageJ (1.54f NIH, Java 1.8.0_322), and the following histomorphometric measurements (90, 91) were recorded for the periosteal and endosteal surfaces: total perimeter, single-label perimeter (sL.Pm), double-label perimeter (dL.Pm), interlabel thickness (Ir.L.Th), total bone area, and marrow area. The following results were calculated: mineral apposition rate (MAR = Ir.L.Th/8 days), mineralizing surface [MS/BS = (0.5 × sL.Pm + dL.Pm)/total perimeter × 100], and bone formation rate (BFR/BS = MAR × MS/BS × 3.65). Relative formation parameters (i.e., rMAR, rMS/BS, and rBFR/BS) for loading effects were calculated for each mouse by subtracting the nonloaded (left tibia) values from the loaded (right tibia) values.

Mechanical testing

Mice were euthanized at 16 to 18 weeks, and the right femurs were harvested and stored at −20°C in saline-soaked gauze until testing. On the day of testing, femurs were brought to room temperature in a saline bath for over 1 hour. Femurs were positioned posterior side down across the two lower supports of a three-point bending fixture. The femurs were loaded to failure in monotonic compression with a span length of 9 mm using a crosshead force, and displacement measurements were collected at every speed of 0.2 mm/s using a mechanical testing system (Bionix; MTS, Eden Prairie, MN, USA). Force and displacement measurements were measured using a 25-pound (11.3398 kg) load cell (SSM-25 Transducer Techniques, Temecula, CA, USA), and a linear variable differential transducer (a calibration curve spanning the expected force range was generated before testing). From the force versus displacement curves, ultimate force (newton), ultimate energy (millijoule), failure energy (millijoule), stiffness (newton per millimeter), yield force (newton), and post–yield displacement (millimeter) were calculated using standard equations (92). Before testing, the length of each femur was measured to the nearest 0.01 mm along the diaphyseal axis using digital calipers.

FTIR spectroscopy

FTIR spectroscopy was used to evaluate bone compositional properties. Left femora from mice used for mechanical testing were harvested and stored at −20°C in saline-soaked gauze until specimen preparation (93). Briefly, epiphyses of the femur were detached, and bone marrow was removed using a centrifuge. The pieces were defatted in 3 × 15 min solution changes of 100% isopropyl ether. Samples were lyophilized in a vacuum drier and made into powder using a CryoMill (770; SPEX SamplePrep, Metuchen, NJ, USA). Pellets were constructed by mixing 2.0 to 3.0 mg of powdered bone with dry potassium bromide to make a 200-mg mixture. Pellets were pressed with a 13-mm-diameter die and a maximum load of 10 tons using an evacuable pellet press (Pike Technologies, Fitchburg, WI, USA). FTIR spectra were collected at a spectral resolution of 4 cm⁻¹ over the spectral range of 800 to 2000 cm⁻¹ using an FTIR spectrometer (Spotlight 400; PerkinElmer Instruments, Waltham, MA, USA). Spectra were analyzed in which areas were defined between the following peaks for phosphate: 916 to 1180 cm⁻¹, amide I: 1596 to 1712 cm⁻¹, and carbonate: 852 to 890 cm⁻¹. The peak height ratios were defined as crystallinity: 1030/1020, acid phosphate: 1126/1096, and collagen maturity (cross-linking ratio): 1660/1690.

Statistical analysis

The effect of the cholinergic cKO on measurements of bone geometry, mineral density, mechanical properties, and metabolism was determined using a one-way analysis of variance (ANOVA) with group as the factor (Cre-negative littermate controls against Cre-positive experimental mice for the same sex) with a Tukey’s post hoc analysis followed by unpaired one-tailed t tests with alternative hypotheses as less than or greater than the control group depending on our hypothesis regarding the changing phenotypes. For DEXA BMD measurements, two-way ANOVA was performed with interactions between genotype and age. A Sidak post hoc test was computed across genotypes for each time point. Outliers greater than three SDs from the mean for each metric grouped by sex were removed. Statistical tests were conducted using R or Prism with α = 0.05.

The effect of the cholinergic cKO on the percent cells responding and change in fluorescence intensity was determined using a two-way ANOVA with a group (GCaMP controls versus GCaMP cKOs) and strain (250, 500, 1000, 2000, and 3000 με) as factors followed by unpaired one-tailed t tests with alternative hypotheses as less than or greater than the control group depending on our hypothesis regarding the changing phenotypes. Statistical tests were conducted using R or Prism with α = 0.05.

The mixed effects of sex and genotype within the dynamic histomorphometry, mechanical testing, μCT, and in vivo loading datasets was assessed using a mixed model run in RStudio version 2023. 12.1 + 402 using the lme4 package. Resulting tables are in the Supplementary Materials.

Sample size varied by group and assay. The sample sizes for each group (CHRNA: C, RASPN: R) are as follows: DEXA (C: 6 to 13, R: 3 to 7), dynamic histomorphometry (C: 8 to 14, R: 7 to 12), FTIR (C: 3 to 6, R: 5 to 7), mechanical testing (C: 9 to 14, R: 9 to 12), μCT (C: 9 to 22, R: 4 to 8), skeletal innervation (C: 3, R: 3 to 4), ΜΤ3 loading (GCaMP: 7 to 9, C: 9 to 10, R: 12 to 14).

Supplementary Materials

The PDF file includes:

Figs. S1 to S6

Tables S1 to S10

Legends for supplemental files S1 and S2

Other Supplementary Material for this manuscript includes the following:

Supplemental files S1 and S2