All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of California, San Francisco. To facilitate comparison to prior work on the role of osteocytes in osteoarthritis, outcomes were analyzed in 16 week old mice (19, 20). Thirteen-week-old male and female FVB/NJ mice (The Jackson Laboratory, #001800, IMSR_JAX:001800) were acclimated to the University of California, San Francisco Laboratory Animal Resource Center (LARC) facility with 67°CF-74°CF, 30-70% humidity, a 12-hr light/dark cycle, and free access to water and irradiated standard chow (LabDiet 5058- PicoLab Rodent Diet 20) for a minimum of two weeks prior to experimental studies (19, 20, 24). At thirteen weeks, mice were randomly assigned for subcutaneous implantation with recommended placebo pellets (Innovative Research of America, cat# NG-111) or slow-releasing prednisolone (GC) pellets (2.1 mg/kg/d, 90-day release, cat# NG-151) for 21 days. Beginning the day after GC pellet implantation, mice received subcutaneous injections (5 days/week) of either vehicle (2% heat-inactivated FBS, 1mM HCl, 150mM NaCl) or rat parathyroid hormone 1-34 (PTH (1–34)) (80 µg/kg; Bachem Cat# H-5460), prior to euthanasia using an IACUC-approved standard procedure of carbon dioxide inhalation at 16 weeks of age.

Right femurs were dissected free of muscle, fixed in 10% neutral buffer formalin (NBF) for 3 days at 4°C, stored in 70% ethanol and scanned using a Scanco µCT50 scanner with x-ray potential of 55 kVP, current 109 μA, and 6W, at a voxel size (resolution) of 10μm, and 500ms integration time, as previously described (17, 25). Bone structural parameters were analyzed by manually contouring 100 slices of the trabecular (Tb) bone compartment (300µm proximal to epiphyseal plate) below the growth plate or cortical (Ct) compartment at mid-diaphysis using a Scanco analytic software. Table 1 shows standard µCT parameters (26) for male (n=4-6/group) and female (n=6-7/group) mice.

Unfixed left femurs (n=4-8/group) were subjected to three-point bending at mid-shaft to assess mechanical properties using a Bose Electroforce 3200 (RRID: SCR_019752) test frame (27). Briefly, bones were hydrated in 1X phosphate-buffered saline (PBS) at room temperature and placed on 2 lower supporting jigs (8mm apart) with the anterior side facing down. The test probe was placed at the mid-point between the 2 supporting jigs to create bending with a displacement rate of 10 µm/s. Mechanical properties of stiffness, yield force, and ultimate force were calculated from load-displacement curves using a custom MATLAB (RRID: SCR_001622) script as previously described (27, 28). Material properties of elastic modulus, yield stress, ultimate stress was calculated from µCT measurements of left femurs from 16-week-old male (n=2-7/group) and female (n=5-6/group) mice using the femur cross-section diameter and moment of inertial (Imin/Cmin and Imin) and equations from Turner et al. and Jepsen et al. (28, 29).

RNA was extracted from female humeri (n=4 mice/group) after removal of epiphysis and bone marrow to assess transcriptomic profiles of osteocyte-enriched cortical bone. Briefly, the dissected bones were flash frozen in liquid nitrogen and homogenized in QIAzol Lysis Reagent (Qiagen cat #79306), and total RNA was extracted using the RNeasy mini kit (Qiagen cat#74106) according to the manufacturer’s instructions. Direct mRNA counts were determined using an automated Nanostring nCounter Mx system (RRID: SCR_021712) (30, 31) with a custom probe set for 94 mouse skeletal genes in the UCSF CCMBM Skeletal Biology and Biomechanics Core. Analysis of expression profiles was performed using the nSolver Analysis Software (RRID: SCR_003420) and nCounter Advanced Analysis Software and normalized with seven housekeeping genes (Gapdh, Rpl19, Gilz (Tsc22d3), bone sialoprotein (Ibsp), beta-2 microglobulin (B2m), beta actin (Actb), Serpine2). Highly significant gene expression fold changes were determined by unpaired t-tests between experimental groups.

Osteocyte-like MLO-Y4 cells (provided by L. Bonewald, RRID: CVCL_M098) were maintained in alpha-MEM supplemented with 2.5% fetal bovine serum, 2.5% bovine calf serum, and 1% penicillin-streptomycin and grown on rat tail collagen type 1 (0.16 mg/ml) coated plates. MLO-Y4 cells were treated with 0.1µM or 1µM dexamethasone with or without 50 nM rat parathyroid hormone 1-34 [PTH(1-34)] for 24 hours (n=3 biological replicates/group and 2 independent experiments). RNA was extracted for real-time quantitative PCR (qPCR), using iQ SYBR Green Supermix (BioRad) on a Biorad CFX96 Touch Real-Time PCR Detection System (RRID: SCR_018064). Gene expression levels were normalized to the housekeeping gene Gapdh. Additional details for primers are provided in the Supplementary Table 1 . Fold change was determined using the delta-delta CT method (32). A one-way ANOVA was used for statistical analysis.

Female right femur/tibia joints were dissected free of muscle, fixed in 10% neutral buffered formalin (NBF), decalcified in 10% EDTA, dehydrated, and embedded with knee joints positioned at a 45 angle in paraffin as previously described (19, 20). Coronal sections (7μm) of the knee joints were obtained using a microtome (Leica Microsystems, Buffalo Grove, IL), followed by standard dewaxing and hydration protocols (19, 20) before various histological staining described below. All brightfield images were obtained on a Nikon Eclipse E800 microscope (RRID: SCR_020326).

Knee joints sections were stained with the Safranin O/Fast Green using the protocol adapted from University of Rochester (33) with the following modifications: Weigert’s Iron Hematoxylin incubation for 3 mins, brief water rinse and differentiation in 1% acid-alcohol for 15 secs, stain with 0.02% Fast Green for 5 mins, differentiation with 1% acetic acid for 30 secs, rinse with water and incubation in 1% Safranin-O for 10 mins, prior to mounting with mounting media.

Osteoarthritis scoring of Safranin O/Fast Green-stained coronal sections (n=4/group) was performed by three blinded graders using the OARSI (34) and modified Mankin (35) scoring system. To maintain a consistent region of interest of the knee, sections with visible anterior cruciate ligament (ACL) and posterior cruciate ligament (PCL) were used for grading. Quantification of the whole knee joint was obtained using 10X and stitched 20X images to assess each quadrant of the knee joint (femur, tibia, lateral, medial). Mean scores across all blinded graders were obtained and the mean scores were averaged within each experimental group.

Bone resorption activity in the knee joint was observed using the tartrate-resistant acid phosphatase (TRAP) Leukocyte Acid Phosphatase staining kit (Sigma cat# 387) following the manufacturer’s instructions with slight modifications. Briefly, sections were post-fixed for 30 secs in Fixative Solution, rinsed in water, and incubated with a mixture of Fast Red Violet (Sigma cat#F3381) and Fast Garnet GBC Base Solution for 1 hour at 37°C in the dark. Slides were then rinsed in water and counterstained with 0.02% Fast Green (Sigma cat# F3381) and mounted. For quantification of bone resorption parameters, one image (20X) of the subchondral bone per quadrant of the knee joint (femur, tibia, medial, lateral) was evaluated. A total of 4 images per animal (n=4-5 mice/group) were analyzed by a blinded grader using the open source image analysis software TrapHisto (36) to measure the Osteoclast Surface per Bone Surface (Oc.S/BS %) and the Number of Osteoclasts per Tissue Volume (N.Oc/TV mm^-2). The mean of these parameters was averaged per quadrant of the knee for each animal and averaged within each experimental group to acquire mean total, medial and lateral joint values.

The lacunocanalicular network of the subchondral bone in the knee was visualized by Ploton silver nitrate stain as previously described (17, 19, 20, 37). Briefly, right knee joint sections were stained in a fresh mixture of 50% silver nitrate and 1% formic acid in 2% gelatin with a 2:1 ratio for 55 mins in the dark and then counterstained with Cresyl Violet. For consistency, sections with visible ACL and PCL were chosen for staining. Four high-resolution images (100X) per knee joint subchondral bone quadrant (femur, tibia, medial, lateral) were used for quantitative analysis. ImageJ (RRID: SCR_003070) was used by a blinded grader to quantify lacunar number and lacunae size for a total of sixteen images per animal (n=4 mice group) by converting to a binary image, manually contouring each lacunae, and measuring with the Analyze Particles feature. Mean values were obtained per quadrant of the knee per animal and were then averaged within each experimental group.

All data are represented as mean ± standard deviation (SD) or standard error mean (SEM) as appropriate for each assay, as stated in the figure legends. For in vivo data, the number of samples per group is denoted as “n”, while in vitro data, n indicates the number of independent experiments/biological replicates. GraphPad Prism (GraphPad Software version 10) was used for all statistical analysis and statistical significance required a p-value ¾ 0.05.

Micro-computed tomography (µCT) identified sex-dependent differences in the effect of GC, PTH(1-34), and combined GC+PTH(1-34) treatments on bone phenotypes ( Figure 1 ; Table 1 ). At 16 weeks of age, male mice, regardless of treatment type, showed no significant changes in either trabecular (Tb) ( Figures 1A–D ) or cortical (Ct) ( Figures 1E–H ) bone parameters by the drug treatments versus vehicle controls, as visualized in the 3D-reconstructed images ( Figure 1I ) and their quantifications ( Figures 1A–H ). In contrast, female mice treated for 21 days with GC showed significant increases in Tb fraction (Tb.BV/TV) ( Figure 1A ) and number (Tb.N, Figure 1B ), with a complementary decrease in spacing (Tb.Sp, Figure 1D ). GC treatment caused loss of Ct bone in female mice ( Figures 1E–H ), similar to what we and others previously reported (6, 18, 38, 39), revealing the trabecular versus cortical region-specific effects of GC. In 16 week old female mice, intermittent PTH (1–34) treatment caused the anticipated anabolic response with significantly elevated Tb.BV/TV ( Figure 1A ), Tb.N ( Figure 1B ), and Tb.Th ( Figure 1C ), and reduced Tb.Sp ( Figure 1D ). Combined GC and PTH(1-34) treatment significantly increased Tb bone parameters relative to female controls ( Figures 1A–D ), with even greater increases in Tb.BV/TV than each treatment alone ( Figure 1A ). However, combined GC and PTH(1-34) did not mitigate GC-induced Ct bone loss ( Figures 1E–H ).

Mechanical testing by three-point bending showed that male femurs treated with PTH(1-34), relative to those treated with GC, have significantly increased yield force, but this effect is absent when GC and PTH(1-34) are combined ( Figures 1J–L ). Similar trends are present in females, with PTH-dependent increases in stiffness and ultimate force relative to bone from GC-treated mice ( Figures 1J–L ). As in males, PTH(1-34) does not overcome the effect of GC on mechanical properties in female bone ( Table 2 ). Material properties of male or female bones were unaffected by GC or PTH(1-34) ( Figure 1 ).

We evaluated the effect of GC, PTH(1-34), and GC+PTH(1-34) treatment on gene expression from osteocyte-enriched humeri using Nanostring nCounter assay and a custom probe set of 96 mouse genes important in skeletal biology, including bone, cartilage, tendon, and muscle. By directly measuring mRNA, this assay provides increased sensitivity across a range of conditions ( Supplementary Figures 2A–H ). Volcano plots show regulation of several genes associated with bone remodeling in osteocyte-enriched bones across all treatment groups from female ( Figures 2A–C ) and, to a lesser extent, from male mice ( Supplementary Figure 1 ). We previously reported that a 7-day GC treatment downregulates Mmp2 (18), which is recapitulated with 21-day treatment of GC ( Figure 2D ). In addition, as anticipated based on prior reports (38, 40), GC reduced mRNA levels of osteocrin (Ostn), osteoprotegerin (Tnfrsf11b), gap junction alpha 1 protein (Cx43) (Gja1), while increasing mRNA levels for tartrate resistant acid phosphatase (Acp5) and cathepsin K (Ctsk) ( Figure 2D ), confirming the efficacy of GC in these conditions. We previously reported that a 7-day GC treatment suppressed bone remodeling genes implicated in PLR (18), however here we observe that a longer 21-day GC treatment significantly upregulates several PLR-related genes including Acp5, Mmp13, Atp6v0d2, Ctsk ( Figure 2D ).

As expected based on prior reports of PTH(1-34) induction of Phex (41, 42) and Wnt4 (43), both genes are enriched in bone from the PTH (1–34) treated group ( Figure 2E ). Other PTH(1-34)-suppressed genes (Sost, Dmp1, Osteocalcin) (44–46) and PTH(1-34)-induced genes (Tnfrsf11a (Rank), Tnfrsf11b (Opg)) (43, 47) were not differentially expressed in these conditions. As we had hypothesized, PTH(1-34) also increased mRNA levels for several PLR-related genes (Acp5, Ctsk, Atp6v0d2), as well as Tnfrsf11a (Rank) ( Figure 2E ). The combined GC + PTH(1-34) treatment led to upregulation of Tnfrsf11a and the same PLR-related genes (Acp5, Ctsk, Atp6v0d2) as individual treatments ( Figure 2F ). Indeed, of the 21 genes in this panel that are significantly regulated by GC+PTH(1-34), relative to vehicle treated cells, all but 2 (Foxo1 and Igf1r) are regulated in the same manner by GC or PTH(1-34) alone, with 7 regulated by both stimuli ( Figures 2D–F , red bars). Overall, analysis of gene expression in these conditions suggests that GC and PTH(1-34), alone or combined, shift bone toward a more catabolic state.

To determine the direct actions of GC and PTH(1-34) on osteocytic activities, we cultured osteocyte-like MLO-Y4 cells with dexamethasone (DEX) with or without PTH(1-34) for 24 hours prior to RNA isolation. Real-time qPCR analysis confirmed the dose-dependent (0.1µM and 1µM) effects of DEX on glucocorticoid-inducible Atrogin1 and Murf1 gene expression ( Figures 3A, B ). Consistent with the previously reported DEX-dependent decrease in Mmp13 mRNA levels in cultured osteocytes (18), DEX suppresses Mmp13 expression in an osteocyte-intrinsic manner ( Figure 3C ). This result suggests that other osteocyte-independent factors may counteract the direct actions of GC on osteocytes to increase Mmp13 expression in osteocyte-enriched cortical bone in vivo ( Figure 2 ). PTH(1-34) did not mitigate suppression of Mmp13 expression by DEX ( Figure 3C ). These in vitro experiments along with the above in vivo studies highlight both cell-intrinsic and non-autonomous actions of GC and PTH(1-34) on osteocytes, and the inability of PTH(1-34) to rescue downregulated Mmp13 expression of GC on osteocytes.

Given that several of the GC and PTH(1-34) regulated genes can participate in bone resorption executed by either osteoclasts or osteocytes (12, 13, 48, 49), both of which can impact joint homeostasis (19, 50, 51), we sought to determine the effect of these treatments on articular cartilage and subchondral bone. Since microCT (µCT), mechanical testing, and gene expression analysis show greater sensitivity to GC and PTH(1-34) in females in these conditions, the remainder of this study focuses on female mice. The effect of GC and PTH(1-34) on the joint was evaluated in Safranin O/Fast green stained knee joint sections ( Figure 4A ) using standard OARSI ( Figure 4B ) (34) and modified Mankin Score ( Figure 4C ) grading systems (35). Across treatments, no signs of cartilage damage or early onset osteoarthritis were observed in 16-week-old female mice.

Among the catabolic genes induced by GC, PTH(1-34), and GC+PTH(1-34) is Acp5 (Trap), which can be expressed by osteoclasts or by osteocytes engaged in PLR (12, 52). TRAP staining was used to distinguish the cell populations associated with differential Acp5/Trap expression in subchondral bone of the female mouse knee ( Figure 5 ; Table 3 ). While abundant TRAP staining was detected on the surfaces of bony trabeculae, corresponding to osteoclasts ( Figures 5B–D ), relatively few TRAP-positive osteocytes were detected in any condition ( Figure 5A ). Quantitative analysis of the % osteoclast surface per bone surface (Oc.S/BS %) ( Figure 5D ) and number of osteoclast per tissue volume (N.Oc/TV mm^-2) ( Figure 5G ) revealed that GC significantly elevated TRAP activity in the medial subchondral bone, which contributed to the increase in total subchondral bone TRAP activity ( Figures 5B, E ). TRAP activity was unaltered by PTH(1-34) alone or in combination with GC ( Figures B–G ). The inability of PTH(1-34) to oppose GC-induced TRAP activity is consistent with their shared trabecular bone phenotype and Acp5 expression profile.

Both GC and PTH(1-34) regulate osteocytic PLR (12, 18) and the expression of genes implicated in this process, including Mmp13, Atp6v0d2, and Ctsk, as shown previously (13, 18, 53) and in Figure 2 . Disruption of the osteocyte lacunocanalicular network (LCN) is a hallmark of PLR suppression that results from GC treatment (18) or from osteocytic ablation of Mmp13 or Ctsk (13, 19, 49). In addition, long-term GC exposure induces osteocyte apoptosis (54, 55). Therefore, to examine the effect of GC and PTH(1-34), alone or in combination, on subchondral bone, osteocyte apoptosis and the LCN were examined histologically using terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) and Ploton silver nitrate stain, respectively. Though some apoptotic marrow cells, osteoclasts, and osteocytes were detected in each condition, the number of TUNEL-positive osteocytes was low and unchanged by GC or PTH(1-34), alone or in combination ( Supplementary Figure 3 ).

Silver staining permits qualitative analysis of canalicular organization ( Figure 6A ) and quantification of lacunar number ( Figure 6B ) and lacunae size ( Figure 6C ) were quantified in each subchondral bone quadrant of the knee. Unlike cortical bone, canalicular organization in trabecular bone is more variable, such that treatment-specific differences in canalicular integrity were not apparent. While GC-dependent differences in lacunar number or size were not observed, PTH(1-34) treatment showed the greatest effect on increased lacunar number in the femur medial compartment ( Figure 6B ) and decreased lacunar size in the tibia medial compartment ( Figure 6C ). The elevated number of lacunae and reduced average lacunar size observed with PTH(1-34) treatment is mitigated when combined with GC. This demonstrates that GC and PTH(1-34) effects on the osteocyte LCN in these conditions are mild, and that the modest effect of PTH(1-34) on lacunar size is blocked by exogenous GC.