Abstract

   Osteoarthritis (OA) treatment is limited by the lack of effective
   nonsurgical interventions to slow disease progression. Here, we
   examined the contributions of the subchondral bone properties to OA
   development. We used parathyroid hormone (PTH) to modulate bone mass
   before OA initiation and alendronate (ALN) to inhibit bone remodeling
   during OA progression. We examined the spatiotemporal progression of
   joint damage by combining histopathological and transcriptomic analyses
   across joint tissues. The additive effect of PTH pretreatment before OA
   initiation and ALN treatment during OA progression most effectively
   attenuated load-induced OA pathology. Individually, PTH directly
   improved cartilage health and slowed the development of cartilage
   damage, whereas ALN primarily attenuated subchondral bone changes
   associated with OA progression. Joint damage reflected early
   transcriptomic changes. With both treatments, the structural changes
   were associated with early modulation of immunoregulation and
   immunoresponse pathways that may contribute to disease mechanisms.
   Overall, our results demonstrate the potential of subchondral
   bone-modifying therapies to slow the progression of OA.
     __________________________________________________________________

   PTH treatment before joint loading attenuated load-induced
   osteoarthritis through directly acting on both bone and cartilage.

INTRODUCTION

   Concomitant changes in articular cartilage and the underlying
   subchondral bone are key contributors to osteoarthritis (OA)
   pathogenesis ([42]1). The close mechanical and biological relationship
   between cartilage and subchondral bone suggests that alterations in
   subchondral bone properties, including changes in bone mass, stiffness,
   and remodeling, influence the health of the overlaying cartilage
   ([43]2) and that tissue cross-talk is an important factor in OA disease
   ([44]3).

   Changes in subchondral bone mass and stiffness can modify both the
   stresses experienced by articular cartilage and the susceptibility to
   develop OA disease ([45]4). Clinical studies reported conflicting
   results relating bone mass and OA pathology, with several studies
   demonstrating a positive relationship between bone mass and OA disease
   ([46]5, [47]6) and other work indicating that low bone mass is
   associated with increased cartilage damage and accelerated progression
   toward OA ([48]7). In numerous preclinical studies, low bone mass or
   loss of bone mass before OA initiation exacerbated load-induced OA
   development ([49]8–[50]11). However, these studies examined the role of
   bone mass by comparing mouse strains with different intrinsic bone mass
   ([51]8) or altered estrogen signaling ([52]9–[53]11), both of which
   also may directly affect cartilage and other joint tissues, thereby
   affecting OA susceptibility.

   OA development also is associated with abnormal and increased
   subchondral bone remodeling, which likely plays an important role in OA
   pathology ([54]12, [55]13). Following OA initiation, a decoupling of
   normal bone resorption and formation occurs following an established
   pattern ([56]12, [57]13). In early-stage OA, an initial increase in
   bone resorption driven by increased osteoclast activity causes loss of
   subchondral bone mass, which is then followed by an increase in bone
   formation driven by increased osteoblast activity that ultimately
   results in subchondral bone sclerosis as OA progresses ([58]14,
   [59]15). Inhibiting these changes in bone remodeling immediately after
   OA initiation can reduce cartilage damage and OA progression in
   preclinical models, suggesting that early changes in bone remodeling
   are important to OA development ([60]16–[61]18).

   The contributions of predisease subchondral bone properties, bone
   remodeling following disease initiation, and their interactions in OA
   development remain unclear. Most previous studies investigating the
   relationship between subchondral bone and OA development focused on
   tissue-level structural changes ([62]17, [63]18). While informative,
   tissue-level analyses provides a limited view of complex molecular
   interactions and joint tissue cross-talk contributing to OA
   development. Identifying time-dependent molecular changes associated
   with tissue-level outcomes across joint tissues during OA development
   would provide unique insight into the role of subchondral bone in
   disease development and the contribution of multiscale tissue
   interactions to disease progression. Recent genomics analyses of mouse
   whole knee joints ([64]19, [65]20) and articular cartilage
   ([66]21–[67]24) provided insight into the disease mechanisms underlying
   the development of OA. However, these studies did not address how
   changes in bone properties affect other joint tissues and modify
   disease progression. Given the multifactorial nature of OA, studies
   that examine concomitant changes in bone and different joint tissues
   are needed to understand organ-specific alterations during OA
   progression.

   In the present study, we modulated subchondral bone properties in mice
   using two Food and Drug Administration–approved treatments for
   osteoporosis to examine the roles of subchondral bone mass, stiffness,
   and remodeling in OA development, integrating histopathological and
   transcriptomics analyses of multiple joint tissues. Parathyroid hormone
   (PTH), an anabolic osteoporosis treatment that increases bone mass and
   stiffness ([68]25, [69]26), was used to increase subchondral bone mass
   in vivo before OA initiation. Subchondral bone remodeling was reduced
   during OA progression using alendronate (ALN), an anticatabolic
   osteoporosis treatment that inhibits the resorptive capacity of
   osteoclasts ([70]27, [71]28). We subjected mice to daily tibial
   loading, a well-established model to initiate OA-like pathology and
   study load magnitude– and time-dependent changes in different tissues,
   including cartilage, bone, and lymph node ([72]8, [73]18,
   [74]29–[75]31).

   We hypothesized that, under conditions of repetitive loading,
   increasing subchondral bone mass before OA initiation via PTH
   pretreatment and reducing subchondral bone remodeling via ALN treatment
   will attenuate the progression of OA. Specifically, we expected PTH
   pretreatment and ALN treatment to slow the development of load-induced
   cartilage damage observed histologically and subchondral bone loss.
   Furthermore, we expected attenuation in tissue damage to be associated
   with early-stage transcriptomic changes in cartilage, bone, and lymph
   nodes.

RESULTS

PTH pretreatment improved cartilage health and increased bone mass

   We first determined the effects that PTH pretreatment had on both bone
   and cartilage baseline properties. We examined pretreatment-only
   groups, in which mice were euthanized after 8 weeks of PTH or vehicle
   (VEH) pretreatment ([76]Fig. 1A). Limbs were analyzed for tissue-level
   changes using histology, microcomputed tomography (microCT),
   immunohistochemistry (IHC), and nanoindentation. RNA isolated from
   cartilage, bone, and lymph nodes of PTH- or VEH-pretreated groups was
   used for RNA sequencing (RNA-seq) analyses.

Fig. 1. PTH pretreatment alone altered cartilage and subchondral bone.

   [77]Fig. 1.
   [78]Open in a new tab

   (A) Mice underwent 8 weeks of PTH or VEH pretreatment to assess
   systemic effects. (B) Representative histology and IHC images of
   pretreated cartilage. Scale bar, 200 μm. In cartilage, PTH pretreatment
   (C) increased thickness and reduced histological score and (D) reduced
   MMP-13 expression but did not alter Runx2 expression. (E) PTH
   pretreatment increased cartilage stiffness. Each shape represents a
   single mouse (n = 3 per group, indentation speed = 1 μm/s). (F) PTH
   pretreatment did not alter subchondral plate thickness but increased
   bone volume fraction in the epiphysis. (G) Representative TRAP and IHC
   images of pretreated bone. (H) PTH pretreatment increased presence of
   osteoclasts and active osteoblasts and reduced sclerostin expression in
   osteocytes. (I) Alluvial plot and Venn diagram representation of the
   differentially expressed genes (DEGs) between PTH and VEH pretreatment
   in cartilage, cortical bone, cancellous bone, and lymph node [false
   discovery rate (FDR) < 0.05, fold change (FC) > 2]. (J) Heatmap
   representation of selected genes related to cartilage health and
   hypertrophy in cartilage and osteoblast, osteoclast, and osteocyte
   markers in cancellous bone. Data were z-score normalized. Both right
   and left limbs included. (K) Volcano plots representing DEGs in
   cartilage and cancellous bone. *FDR < 0.05 (L) Representation of the
   top five up-regulated and down-regulated Biological Processes enriched
   (adj.p < 0.05) for all significant genes in cartilage. *Bars indicate
   mean + SD. Individual data points (n = 7 per group) are overlaid. *P <
   0.05 (t test).

   PTH-pretreated mice had thicker tibial cartilage than VEH mice (P =
   0.023; [79]Fig. 1, B and C). Cartilage structure, assessed by the OARSI
   score, was improved with PTH pretreatment, which led to a more pristine
   cartilage surface with fewer fibrillations and increased proteoglycan
   content (P < 0.001; [80]Fig. 1C). PTH pretreatment decreased matrix
   metalloproteinase–13 (MMP-13) immunostaining in cartilage (P = 0.036)
   but did not alter runx-related transcription factor 2 (RUNX2) levels
   ([81]Fig. 1, B and D). Cartilage stiffness, measured by
   nanoindentation, also was increased with PTH pretreatment compared to
   VEH (P = 0.008; [82]Fig. 1E). Subchondral plate thickness ([83]Fig. 1F)
   and subchondral plate tissue mineral density (TMD), assessed by
   microCT, were not altered by PTH pretreatment. PTH pretreatment
   increased cancellous bone volume fraction in the epiphysis (P = 0.007;
   [84]Fig. 1F) and slightly decreased epiphysis TMD (P = 0.050). Bone
   remodeling in cancellous bone was increased in the PTH pretreatment
   group, as indicated by increased tartrate-resistant acid phosphatase
   (TRAP) (P = 0.029) and procollagen I (P = 0.043; [85]Fig. 1, G and H)
   immunostaining. PTH pretreatment also decreased the relative number of
   sclerostin-positive osteocytes (P = 0.015; [86]Fig. 1H). Overall, PTH
   pretreatment improved cartilage health and increased cancellous
   subchondral bone mass.

   In RNA-seq analyses of different joint tissues, tibial cartilage
   displayed the most pronounced gene expression changes in response to
   PTH ([87]Fig. 1I). Cartilage had 717 differentially expressed genes
   [DEGs; false discovery rate (FDR) < 0.05, fold change > 2], with 547
   genes with higher expression in the PTH-pretreated group and 170 genes
   with higher expression in the VEH-treated mice. Lymph nodes had a total
   of 55 DEGs, cancellous bone had 40 DEGs, and cortical bone had 2. The
   expression of major genes with known roles in cartilage health and
   chondrocyte hypertrophy was not different in VEH- and PTH-pretreated
   groups ([88]Fig. 1, J and K). Within cancellous bone, comparison of PTH
   and VEH groups did not identify changes in genes associated with
   osteoblast, osteoclast, or osteocyte activity ([89]Fig. 1, J and K). On
   the basis of biological pathway enrichment analysis, relative to
   VEH-treated mice, the top five biological processes activated by PTH
   pretreatment in cartilage included changes in gene pathways associated
   extracellular matrix organization and several pathways associated with
   morphogenesis ([90]Fig. 1L). The top five biological pathways
   suppressed by PTH pretreatment included changes in gene pathways
   associated with immune processes, including natural killer cell
   degranulation and neutrophil degranulation ([91]Fig. 1L). No
   differentially enriched pathways were found in cortical bone,
   cancellous bone, or lymph nodes after 8 weeks of pretreatment. Overall,
   our transcriptomic analysis demonstrated PTH alters the expression of
   genes associated with cartilage homeostasis.

PTH pretreatment attenuated load-induced cartilage damage

   Next, we examined whether the development of load-induced OA was
   influenced by the baseline subchondral bone mass or by bone remodeling.
   We applied daily mechanical loading for 1, 2, or 6 weeks to mice
   pretreated for 8 weeks with VEH or PTH to modulate baseline subchondral
   bone mass, followed by VEH or ALN treatment to inhibit bone remodeling,
   concomitant with loading for the duration of the loading period
   ([92]Fig. 2A). A linear mixed-effects model was used to statistically
   assess the effects of loading (contralateral control versus loaded),
   duration (1, 2, or 6 weeks), pretreatment (PTH versus VEH), treatment
   (ALN versus VEH), and their interactions on each cartilage and
   subchondral bone measure.

Fig. 2. PTH pretreatment attenuated load-induced cartilage damage.

   [93]Fig. 2.
   [94]Open in a new tab

   (A) Experimental design. (B) Representative histology images of
   cartilage damage following loading. Scale bar, 250 μm. (C) Cartilage
   damage increased with loading and the severity of load-induced
   cartilage damage was attenuated by PTH pretreatment. (D) Posterior
   cartilage thickness decreased with loading. Bars indicate mean + SD.
   Individual data points (n = 8 per group) are overlaid. Δ indicates
   difference between loaded and control limb measures within a single
   animal. P < 0.05 (linear mixed-effects model with Tukey’s post hoc
   test) indicated on vertical axis for effects of ^βDuration,
   ^γPretreatment, ^κDuration × Treatment.

   Consistent with our previous studies using this loading protocol, daily
   cyclic compressive loading induced cartilage damage at all load
   durations in all pretreatment and treatment groups (P < 0.001; [95]Fig.
   2B), with increasing cartilage damage scores with longer load durations
   (P < 0.001). PTH pretreatment attenuated load-induced cartilage damage
   (P < 0.001; [96]Fig. 2, B and C) throughout the entire joint, whereas
   inhibition of bone remodeling with ALN did not alter load-induced
   cartilage damage as measured by OARSI score. Cartilage damage was most
   severe in the posterior portion of the tibial plateau, but the effect
   of PTH pretreatment and ALN treatment on load-induced cartilage damage
   was similar throughout the joint. Whole-joint mean cartilage thickness
   was not altered by loading, pretreatment, or treatment. Localized
   cartilage thickness in the posterior region of the joint decreased with
   loading in a load duration–dependent manner (P < 0.001; [97]Fig. 2D).
   ALN treatment attenuated the load-induced decrease in cartilage
   thickness at 6 weeks. Overall, PTH pretreatment attenuated load-induced
   cartilage damage, and ALN treatment had a more limited effect on
   cartilage damage after loading.

PTH pretreatment was chondroprotective and modulated load-induced RUNX2 and
MMP-13 levels

   We examined changes in cartilage protein levels at 1 week after
   initiation of daily loading using IHC analyses to identify the
   interactions between loading, PTH pretreatment, and ALN treatment on
   cartilage degradation. Cartilage tissue area with positive MMP-13
   immunostaining, a collagenase with critical roles in type II collagen
   degradation ([98]32), was decreased by both PTH pretreatment (P =
   0.013) and ALN treatment (P = 0.002; [99]Fig. 3, A and B). Load-induced
   changes in MMP-13 immunostaining depended on pretreatment (P = 0.011).
   MMP-13 levels increased with loading in VEH-pretreated groups but
   decreased in PTH-pretreatment groups. RUNX2 is a transcription factor
   with central roles in chondrocyte maturation, hypertrophy, endochondral
   ossification ([100]33), and MMP-13 transcriptional activation ([101]34)
   and has both anabolic and catabolic roles in OA cartilage ([102]35).
   Tissue area with positive RUNX2 immunostaining was altered by loading
   (P = 0.035). RUNX2 levels increased with loading in VEH-pretreated
   groups but decreased with loading in PTH-pretreated groups (P = 0.001;
   [103]Fig. 3, A and C). ALN treatment did not affect RUNX2
   immunostaining in cartilage. Tissue area with positive type 1 PTH/PTHrP
   receptor (PTH1R) immunostaining, the primary PTH receptor on
   chondrocytes that decreases with OA progression ([104]36), was not
   altered by PTH pretreatment (P = 0.051), ALN treatment or loading (fig.
   S1, A and B). Tissue area with positive sclerostin immunostaining, a
   Wnt inhibitor expressed by chondrocytes and implicated in OA
   progression ([105]37), was decreased with PTH pretreatment (P = 0.003)
   and ALN treatment (P = 0.003; [106]Fig. 3, A and D). PTH pretreatment
   also prevented the load-induced increases in sclerostin immunostaining
   in cartilage (P = 0.022). Overall, PTH pretreatment attenuated
   load-induced increases in cartilage degrading enzymes and chondrocyte
   hypertrophy.

Fig. 3. PTH pretreatment and ALN treatment modulated load-induced protein
expression changed in cartilage.

   [107]Fig. 3.
   [108]Open in a new tab

   (A) Representative IHC images of MMP-13, Runx2, and sclerostin
   expression in cartilage following 1 week of loading. Scale bar, 200 μm.
   (B) MMP-13 expression decreased with PTH pretreatment and ALN
   treatment. PTH pretreatment prevented the load-induced increase in
   MMP-13 expression. (C) PTH pretreatment prevented the load-induced
   increase in Runx2 expression. (D) Sclerostin expression decreased with
   PTH pretreatment and ALN treatment. PTH pretreatment prevented the
   load-induced increase in sclerostin expression. Bars indicate mean +
   SD. Individual data points (n = 4 per group) are overlaid. P < 0.05
   (linear mixed-effects model with Tukey’s post hoc test) indicated on
   vertical axis for effects of ^αLoad, ^γPretreatment, ^δTreatment,
   ^εLoad × Pretreatment, ^λPretreatment × Treatment.

RNA-seq confirmed that articular cartilage is mechanosensitive and uncovered
transcriptional responses to PTH and ALN

   To further define cartilage responses, we performed RNA-seq analyses in
   RNA isolated from articular cartilage following 1 and 2 weeks of
   loading, from mice without and with pretreatment with PTH and treatment
   with ALN. First, we evaluated cartilage gene expression changes
   associated with load-induced OA-like pathology in the VEH-treated
   group, comparing load and control limbs (VEH-VEH comparison). We found
   a time-dependent increase in cartilage DEGs in response to loading,
   with 213 DEGs after 1 week of loading and 388 DEGs after 2 weeks of
   loading, with 150 overlapping DEGs ([109]Fig. 4A). The expression of
   genes related to both cartilage catabolic and anabolic processes,
   including Mmp13, Spp1, Col9a1, Col27a1, and Gdf5, increased with
   loading after 1 week ([110]Fig. 4, B and C), consistent with both
   catabolic and anabolic responses associated with OA ([111]38). Sox9
   expression was down-regulated at 2 weeks after loading, potentially
   indicating the loss of phenotypic stability of chondrocytes and a shift
   toward hypertrophic-like conversion ([112]39). Load-induced changes in
   gene expression were less obvious after 2 weeks of loading (fig. S2),
   suggesting that gene expression changes leading to cartilage damage
   happen early during load-induced OA progression.

Fig. 4. Loading, pretreatment, and treatment induced transcriptomic changes
in cartilage.

   [113]Fig. 4.
   [114]Open in a new tab

   (A) Venn diagram representation of the DEGs between 1 and 2 weeks of
   loading in cartilage for each treatment group (logFC > 1.5, FDR <
   0.05). (B) Volcano plot representing DEGs for loaded relative to
   control limb VEH-VEH cartilage at 1 week (FDR < 0.05). (C) Heatmap
   representation of selected specific cartilage health and hypertrophic
   related genes for VEH-VEH cartilage. Data were z-score normalized. (D)
   Representation of the top five up-regulated and down-regulated
   Biological Processes enriched (adj.p < 0.05) for all significant genes
   comparing loaded and control VEH-VEH cartilage. (E) Volcano plot
   representing DEGs for loaded relative to control samples in PTH-ALN
   cartilage at 1 week (FDR < 0.05). (F) Heatmap representation of
   selected specific cartilage health and hypertrophic related genes for
   PTH-ALN cartilage. Data were z-score normalized. (G) Representation of
   the top five up-regulated and down-regulated Biological Processes
   enriched (adj.p < 0.05) for all significant genes comparing PTH-ALN
   loaded and control cartilage.

   Top enriched biological pathways after 1 week of loading included
   collagen catabolic processes, ossification involved in bone remodeling,
   and endochondral bone ([115]Fig. 4D), consistent with the proposed
   reactivation of developmental processes and hypertrophy-like conversion
   of OA chondrocytes ([116]40). At 2 weeks of loading, enrichment was
   found in pathways associated with extracellular matrix organization,
   endochondral ossification processes, and myofibril organization (fig.
   S2), consistent with the proposed loss of phenotypic regulation of OA
   chondrocytes ([117]40), including fibroblast-like changes in late
   disease stages ([118]41). Overall, the early transcriptomic responses
   of cartilage to loading correlated to the structural damage observed at
   later time points.

   Next, we examined changes in gene expression in cartilage in response
   to loading with PTH pretreatment and ALN treatment. We found an overall
   increase in DEGs in all treatment groups and time points relative to
   the VEH-VEH treated group, except for decreased DEGs in the PTH-ALN
   group at 2 weeks ([119]Fig. 4A). In the VEH-ALN group, we identified a
   time-dependent increase in DEGs, with 1064 genes after 2 weeks relative
   to 741 genes after 1 week, with 291 overlapping DEGs. Similarly, the
   PTH-VEH group also displayed an increase in DEGs over time, with 1072
   DEGs at 2 weeks versus 370 DEGs at 1 week and 217 overlapping DEGs. In
   contrast, the PTH-ALN treatment had more genes differentially expressed
   after 1 week of loading (1201 genes) relative to the 2-week time point
   (270 genes), with 107 of these genes shared between time points. Two
   hundred fifty-four of the DEGs present at 1 week in the PTH-ALN group
   were then present at 2 weeks in the VEH-VEH group. Together, these
   results suggest that the combined PTH-ALN treatment shifted the
   temporal responses of cartilage to loading.

   Because the PTH-ALN cotreatment group displayed the most pronounced
   tissue-level changes across cartilage and bone, we performed additional
   analyses of the pretreatment and treatment effects in this group. With
   combined PTH-ALN treatment, the expression of genes traditionally used
   to define healthy articular chondrocytes and cartilage anabolism,
   including Acan, Col2a1, and Sox9, was increased in the loaded limb
   after 1 week ([120]Fig. 4, E and F). We also identified a concomitant
   increase in the expression of genes related to cartilage catabolism and
   chondrocyte hypertrophy, including Col10a1, Mmp13, and Spp1. After 1
   week of loading, extracellular matrix organization and cartilage
   development were among the top five up-regulated biological processes
   with loading ([121]Fig. 4G). The five most down-regulated biological
   processes were all associated with antigen presentation via major
   histocompatibility complex class II and T cell receptor signaling.

   Overall, these loading-based differences were less evident after 2
   weeks of loading (fig. S2). Similar analyses were performed on the
   VEH-ALN and PTH-VEH groups (figs. S2 and S3). Together, our analyses
   suggest that the combination of PTH pretreatment and ALN treatment
   leads to more rapid and potent anabolic responses and decreased
   catabolic responses in articular chondrocytes in response to loading,
   which concur with the reduced structural damage in cartilage observed
   in this treatment group.

ALN treatment slowed osteophyte development

   Osteophyte formation is a hallmark of OA disease ([122]42). We used
   established histological methods to assess osteophyte formation, size,
   and maturity ([123]43) to determine whether PTH pretreatment or ALN
   treatment altered ectopic bone formation resulting from load-induced
   OA. Consistent with the OA-like pathology previously reported in this
   mouse model ([124]8, [125]29), medial tibial osteophytes formed with
   loading in all groups ([126]Fig. 5A), and osteophyte size and maturity
   increased with load duration (P < 0.001; [127]Fig. 5, A to C).
   Osteophyte size decreased with both PTH pretreatment (P = 0.035) and
   ALN treatment (P = 0.021; [128]Fig. 5B), whereas osteophyte maturity
   decreased with ALN treatment (P < 0.001; [129]Fig. 5C) but was not
   altered by PTH pretreatment. In ALN-treated groups, osteophytes
   remained partially cartilaginous at 6 weeks after initiation of
   loading, whereas osteophytes were fully calcified at 6 weeks in
   VEH-treated groups. Overall, our findings confirm that ALN treatment
   slowed osteophyte development, as indicated by decreased osteophyte
   size and maturity following loading, whereas PTH pretreatment only
   reduced osteophyte size.

Fig. 5. ALN treatment attenuated bone changes.

   [130]Fig. 5.
   [131]Open in a new tab

   (A) Representative histology images of osteophyte formation following
   loading. Scale bar, 500 μm. (B) Both PTH pretreatment and ALN treatment
   decreased the width of osteophytes formed in loaded limbs. (C) ALN
   treatment slowed calcification of osteophytes at 2 and 6 weeks. (D)
   Neither PTH pretreatment or ALN treatment altered the load-induced
   decreased in subchondral plate thickness. (E) ALN treatment attenuated
   load-induced decreased in epiphysis TMD. (F) Representative IHC images
   of TRAP (osteoclasts) and procollagen I (active osteoblasts) in bone
   following 1 week of loading. Scale bar for TRAP and procollagen I, 200
   μm. (G) ALN treatment decreased presence of osteoclasts. (H) ALN
   treatment decreased presence of active osteoblasts. Bars indicate mean
   + SD. Individual data points (n = 8 per group) are overlaid. Δ
   indicates difference between loaded and control limb measures within a
   single animal. P < 0.05 (linear mixed-effects model with Tukey’s post
   hoc test) indicated on vertical axis for effects of ^βDuration,
   ^γPretreatment, ^δTreatment, ^κDuration × Treatment, ^λPretreatment ×
   Treatment, ^εLoad × Pretreatment.

ALN treatment systemically increased subchondral bone mass

   Next, we examined subchondral bone changes in the cortical subchondral
   plate and the epiphysis using microCT analysis to determine the effects
   of PTH pretreatment, ALN treatment, and loading. Consistent with
   previous results ([132]8), subchondral cortical plate thickness
   decreased with mechanical loading (P = 0.001; [133]Fig. 5D). ALN
   treatment increased subchondral plate thickness (P = 0.009; fig. S4)
   but did not alter the load-induced loss of subchondral plate thickness
   ([134]Fig. 5D). PTH pretreatment did not alter subchondral plate
   thickness or the change in thickness with loading. Epiphysis bone
   volume/total volume (BV/TV) decreased with mechanical loading (P <
   0.001; [135]Fig. 5E). ALN treatment increased BV/TV (P < 0.001; fig.
   S4) and attenuated the load-induced decrease in BV/TV (P < 0.001;
   [136]Fig. 5E). PTH pretreatment increased BV/TV (P = 0.004; fig. S4)
   but did not alter the response to mechanical loading. Overall, loading
   decreased subchondral bone mass and ALN treatment attenuated cancellous
   subchondral bone changes, whereas PTH pretreatment did not alter
   load-induced bone changes.

ALN treatment reduced bone remodeling

   To understand the impact of loading, PTH pretreatment, and ALN
   treatment in bone cells, we performed immunostaining of markers for
   osteoclasts (TRAP), osteoblasts (procollagen I) and osteocytes
   (sclerostin). TRAP staining increased in cells in the epiphyseal bone
   with PTH pretreatment (P = 0.003), in agreement with our assessment of
   changes at baseline, and decreased with ALN treatment (P < 0.001;
   [137]Fig. 5, F and G). Load-induced changes in TRAP^+ cells depended on
   PTH pretreatment (P = 0.040) and ALN treatment (P < 0.001): Tibial
   loading induced an increase in TRAP^+ immunostaining in the VEH-treated
   groups, but pretreatment with PTH or treatment with ALN prevented the
   load-induced changes. The presence of active osteoblasts, examined
   through procollagen I IHC, was decreased with ALN treatment (P < 0.001)
   but not altered by PTH pretreatment or loading ([138]Fig. 5, F and H).
   The relative number of sclerostin positive osteocytes was decreased by
   both PTH pretreatment (P < 0.001) and ALN treatment (P < 0.001) but not
   altered by loading (fig. S5). Together, these findings suggest that ALN
   attenuated bone remodeling and that both PTH pretreatment and ALN
   treatment altered load-induced changes in osteoclasts presence, likely
   reducing load-induced bone resorption.

PTH pretreatment and ALN treatment attenuated transcriptional responses to
loading in bone

   To further evaluate the responses of bone to loading without or with
   PTH pretreatment and ALN treatments, we performed RNA-seq analyses of
   RNA isolated from cortical and cancellous metaphyseal bone at 1 and 2
   weeks after initiation of loading. Analyses of the VEH-VEH group showed
   that the number of load-induced DEGs was similar at 1 and 2 weeks (280
   versus 288; [139]Fig. 6A), with 18 DEGs common to both time points. The
   five most suppressed biological processes in cortical bone after
   loading included positive regulation of T cell mediated immunity,
   neutrophil activation, and granulocyte activation ([140]Fig. 6, B to
   D). Cancellous bone was less responsive to loading, with only 18 DEGs
   at 1 week and no DEGs at 2 weeks ([141]Fig. 6G).

Fig. 6. Loading, pretreatment, and treatment induced transcriptomic changes
in bone and lymph node.

   [142]Fig. 6.
   [143]Open in a new tab

   (A) Venn diagram representation of the DEGs between 1 and 2 weeks of
   loading in metaphyseal cortical bone for each treatment group (logFC >
   1.5, FDR < 0.05). (B) Volcano plot representing DEGs for loaded
   relative to control limb VEH-VEH metaphyseal cortical bone at 1 week
   (FDR < 0.05). (C) Heatmap representation of selected specific
   osteoblast, osteoclast, and osteocyte marker genes in metaphyseal
   cortical bone. Data were z-score normalized. (D) Representation of the
   top five up-regulated and down-regulated Biological Processes enriched
   (adj.p < 0.05) for all significant genes comparing loaded and control
   VEH-VEH metaphyseal cortical bone. (E) Volcano plot representing DEGs
   for loaded relative to control limb PTH-ALN metaphyseal cortical bone
   at 1 week (FDR < 0.05). (F) Heatmap representation of selected specific
   osteoblast, osteoclast, and osteocyte marker genes in PTH-ALN
   metaphyseal cortical. Data were z-score normalized. (G) Venn diagram
   representation of the DEGs between 1 and 2 weeks of loading in
   metaphyseal cancellous bone for each treatment group (logFC > 1.5, FDR
   < 0.05). (H) Venn diagram representation of the DEGs between 1 and 2
   weeks of loading in lymph nodes for each treatment group (FC > 2, FDR <
   0.05).

   Next, we examined responses to the PTH pretreatment and ALN treatment
   relative to the VEH-VEH group. Overall, fewer load-induced DEGs were
   present in the cortical bone of ALN and PTH groups relative to the
   VEH-VEH group. The VEH-ALN group had two DEGs at 2 weeks after
   initiation of loading and none after 1 week ([144]Fig. 6A). The PTH-VEH
   group had 92 DEGs at 1 week after loading and 17 DEGs at 2 weeks, with
   seven genes shared between time points. In contrast, the PTH-ALN group
   displayed more DEGs after 2 weeks of loading (160 genes) compared to 1
   week (132 genes), with only four overlapping DEGS. Similar to the
   findings in the VEH-VEH group, the cancellous bone was less responsive,
   and we only identified 52 load-induced DEGs at 1 week in the PTH-ALN
   group ([145]Fig. 6G).

   Because the tissue-level analyses showed more prominent changes in the
   PTH-ALN group, we performed additional analyses in this dataset. In
   cortical bone, genes related to Wnt signaling were down-regulated after
   1 week of loading in the PTH-ALN group ([146]Fig. 6, E and F).
   Expression of genes related to osteoblasts and osteoclasts in
   cancellous bone had no clear trend with loading in the PTH-ALN group.
   No biological pathway was enriched in the cortical or cancellous bone
   in the PTH-ALN treatment group at 1 week. Similar results were found in
   the VEH-ALN and PTH-VEH groups (figs. S6 and S7).

Loading induced transcriptomic changes in lymph nodes that attenuated over
time

   To examine immune responses, we performed RNA-seq analyses in the
   inguinal lymph nodes after 1 and 2 weeks of loading. Comparison of the
   VEH-VEH groups showed rapid responses to loading, with 491 DEGs at 1
   week that quickly attenuated to 46 DEGs at 2 weeks and 15 DEGs shared
   between time points ([147]Fig. 6H). In the VEH-ALN, a small number of
   DEGs were present at 1 week after initiation of loading (eight genes)
   and none present after 2 weeks. In contrast, DEGs were only present
   after 2 weeks of loading in the PTH-VEH group (325 genes). No DEGs were
   present in the PTH-ALN group. Enriched biological pathways were present
   in the VEH-VEH and VEH-ALN groups at 1 week but not in the PTH-VEH or
   PTH-ALN groups (fig. S8).

   The estimated fractions of B cells, CD4^+ T cells, and CD8^+ T cells,
   as calculated by CIBERSORT using gene expression profiles ([148]44),
   were altered by ALN treatment (fig. S9). The fraction of CD8^+ T cells
   increased with ALN treatment following 1 week of loading (P = 0.033).
   Following 2 weeks of loading, the fraction of B cells decreased with
   ALN treatment (P = 0.018), and the fraction of CD4^+ T cell increased
   with ALN treatment (P = 0.009). Overall, ALN treatment during loading
   modulated the presence of immune cells in the lymph node, but PTH had
   little effect.

DISCUSSION

   In the present study, we examined the role of subchondral bone
   properties in OA development and progression by determining the effect
   of bone-modulating therapies in attenuating damage severity. We also
   sought to understand the spatiotemporal progression of joint damage in
   OA disease by combining histopathologic and transcriptomic analyses.
   Our findings partially support our hypothesis as both bone-modulating
   therapies approved to treat osteoporosis slowed aspects of OA
   progression. The additive effect of PTH pretreatment before OA
   initiation combined with ALN treatment during damage progression most
   effectively attenuated load-induced pathology across the whole joint.
   PTH pretreatment slowed the development of cartilage damage and may
   have shifted chondrocyte phenotype, whereas ALN treatment attenuated
   osteophyte formation and subchondral bone changes associated with OA
   progression. We cannot distinguish whether decreased OA development
   with PTH pretreatment specifically resulted from increased predisease
   subchondral bone mass or, more likely, from a combination of increased
   bone mass and direct effects on predisease cartilage properties.

   Consistent with our previous reports ([149]8, [150]18, [151]29,
   [152]30), daily cyclic tibial loading induced joint damage in a
   duration-dependent manner. Using PTH pretreatment before loading and
   ALN treatment concurrent with loading altered the progression of
   OA-like pathology. PTH-pretreated mice had increased cancellous bone
   mass, fewer cartilage surface fibrillations, increased proteoglycan
   staining, and decreased MMP-13 immunostaining before loading. With
   loading, PTH pretreatment decreased the severity of load-induced
   cartilage damage and prevented the load-induced increase of
   matrix-degrading enzymes (MMP-13) and markers of chondrocyte
   hypertrophy (RUNX2) in cartilage. On the basis of these results, the
   PTH-induced prevention of cartilage damage was mediated, at least in
   part, by reduced RUNX2 and MMP-13 levels, consistent with reports in
   MMP-13 and RUNX2 knockout mice ([153]43, [154]45). Furthermore, our
   findings suggest that PTH had direct effects in cartilage, similar to
   prior work examining PTH treatment following osteochondral defects or
   collagenase-induced cartilage damage ([155]46, [156]47). PTH
   pretreatment also decreased osteophyte growth but did not slow
   osteophyte maturity or alter the subchondral bone response to loading.
   PTH pretreatment increased bone remodeling and inhibited sclerostin
   production by osteocytes, a potential mechanism through which PTH has
   an anabolic effect on bone ([157]48).

   In contrast, ALN treatment concurrent with daily loading attenuated
   OA-related changes in the subchondral bone, including inhibited bone
   remodeling, slowed osteophyte development, and attenuated cancellous
   subchondral bone loss following loading, but had little effect on the
   progression of load-induced cartilage damage. We previously have
   observed a similar reduction in osteophyte maturity with ALN treatment
   concurrent with daily loading ([158]8). In addition, multiple studies
   examining the role of bone remodeling in surgically induced
   posttraumatic OA models reported both a similar reduction of osteophyte
   development and attenuation of subchondral bone loss when bone
   remodeling is inhibited with bisphosphonate treatment ([159]17,
   [160]49). In contrast to our findings, ALN treatment immediately
   following a single bout of loading, which initiated OA, reduced
   cartilage degeneration ([161]18, [162]50). However, the lack of
   chondroprotection with ALN treatment in the present study is similar to
   previous results when ALN or another bisphosphonate treatment was
   concurrent with daily tibial loading ([163]8), occurred during
   spontaneous OA development in guinea pigs ([164]51), or administered to
   clinically diagnosed patients with OA ([165]52). Together, these
   findings suggest that ALN may be chondroprotective when given
   immediately following a single initiating event that leads to
   subsequent posttraumatic OA development rather than in chronic or
   spontaneous OA phenotypes.

   PTH pretreatment combined with ALN treatment concurrent with loading
   had an additive effect with protections against both load-induced
   cartilage degeneration provided by PTH pretreatment and subchondral
   bone changes provided by ALN treatment. To better understand the
   spatiotemporal progression of joint damage and the effects of PTH
   pretreatment and ALN treatment on OA progression, we examined
   transcriptomic changes across cartilage, bone, and lymph nodes at early
   time points after loading ([166]Fig. 7). The combined PTH pretreatment
   and ALN treatment increased the number of DEGs in cartilage at 1 week
   relative to other treatment groups, indicating an early surge in
   biological activity ([167]Fig. 7A). Increased expression of genes
   associated to cartilage homeostasis ([168]38), including Acan, Sox9,
   and Col2a1, suggest that the PTH-ALN cotreatment led to early
   reparative responses in cartilage that correlated to decreased tissue
   damage at later time points. In contrast, in the VEH-VEH group,
   increased cartilage structural damage and osteophyte formation observed
   at 6 weeks of loading corresponded with an early (1 week) increase in
   the expression of pathways involved in matrix remodeling and
   ossification in cartilage. Differential gene expression also was
   compared to OA-related genes from genome-wide associated studies (GWAS)
   ([169]53). Genes identified by GWAS were no longer significantly
   enriched in the cartilage of the PTH-ALN group after 2 weeks of loading
   as compared to the VEH-VEH and other treatment groups ([170]Fig. 7B).

Fig. 7. Cross tissue visualization of loading, pretreatment, and treatment
induced transcriptomic changes.

   [171]Fig. 7.
   [172]Open in a new tab

   (A) Alluvial diagrams of the DEGs between 1- and 2-week loaded limbs
   across tissues for each treatment group (FC > 2, FDR < 0.05). (B) Bar
   plot showing enrichment for OA-related genes calculated from GWAS in
   cartilage for each treatment group (odds ratio, FDR-corrected *P < 0.05
   indicates significant enrichment). (C) Venn diagram representation of
   the DEGs between treatment groups for each tissue type following 1 or 2
   weeks of loading (FC > 2, FDR < 0.05).

   Comparison of our differential gene expression and pathway analyses to
   recent single-cell RNA sequencing (scRNAseq) findings in cartilage
   ([173]54–[174]57) allowed us to draw further insight into differential
   expression associated with loading and treatment. In the PTH-ALN group,
   cartilage had up-regulation of pathways related to extracellular matrix
   organization and collagen fibril organization, both associated with
   changes in Col2a1 and Comp gene expression. scRNAseq analysis of human
   OA cartilage defined a chondrocyte phenotype, reparative chondrocytes,
   identified by enrichment for both extracellular matrix signaling and
   collagen fibril organization pathways and expression of Col2a1 and Comp
   ([175]56). In contrast, up-regulation of matrix remodeling and
   ossification pathways both, associated with Col10a1 and Jun expression,
   in the VEH-VEH group corresponded with gene expression profiles found
   in hypertrophic chondrocytes, which are known to contribute to
   catabolic metabolism and OA progression ([176]54, [177]56). Together,
   these findings suggest that PTH pretreatment combined with ALN
   treatment may influence chondrocyte phenotype, shifting gene expression
   profiles toward a reparative phenotype and away from a hypertrophic
   phenotype, thereby protecting cartilage from degeneration.

   While immune responses and regulation of inflammation in OA primarily
   occur through the synovial tissue signaling ([178]58), recent
   transcriptomic-based studies have identified alterations in genes and
   signaling pathways related to immune signaling in chondrocytes
   ([179]54–[180]57, [181]59). Pathways containing genes linked to T cell
   receptor signaling and antigen presentation were down-regulated in
   cartilage at 1 week following loading in the PTH-ALN and PTH
   pretreatment-only groups. Similarly, PTH pretreatment before loading
   also down-regulated expression of pathways containing genes linked to
   immune cell regulation. Chondrocytes have immunomodulatory capacities
   in vitro and are capable of suppressing activation, proliferation, and
   effector function of CD4^+ and CD8^+ T cells in coculture ([182]60).
   Both CD4^+ and CD8^+ T cells are present in the synovium during OA
   progression and, through secretion of proinflammatory cytokines,
   contribute to cartilage degeneration and subchondral bone changes
   ([183]61, [184]62). However, the implications of suppressed T cell
   receptor signaling and antigen presentation for chondrocytes embedded
   in the extracellular matrix, where immune cell infiltration is limited,
   requires additional investigation. In scRNAseq, a chondrocyte
   phenotype, regulatory chondrocytes, was associated with early-stage OA
   progression and characterized by the expression of pathways related to
   T cell and B cell receptor signaling and antigen presentation
   ([185]54). Regulatory chondrocytes play a regulatory role in OA
   progression and, given high expression of immune-specific markers found
   in this chondrocyte phenotype, are even speculated to have
   immune-related capacities within the cartilage extracellular matrix
   ([186]54, [187]57). Suppression of the characteristic regulatory
   chondrocyte pathways with PTH-ALN treatment may suggest that modulation
   of regulatory chondrocyte signaling attenuates OA progression
   ([188]57). Overall, PTH pretreatment suppressed gene expression related
   to immune signaling pathways in cartilage. Our interpretation of these
   findings in relation to PTH effects on immune regulation and joint
   inflammation with OA progression remains limited by our lack of
   synovial tissue analyses.

   PTH pretreatment and ALN treatment also altered load-induced
   differential gene expression in the lymph nodes and metaphyseal bone,
   giving further insight into the impact of treatment across the whole
   joint. PTH-ALN treatment eliminated an early increase in differential
   gene expression observed in the lymph node with VEH-VEH treatment,
   suggesting an early modulation of load-driven immune responses
   ([189]Fig. 7, A and C). The association between early-stage
   differential gene expression responses in the lymph node and increased
   OA development over time could be of interest from a diagnostic
   biomarker perspective and requires further investigation. ALN treatment
   also modulated the proportion of B cells and CD8^+ T cells inferred
   from gene expression profiles in the lymph node. The proportion of
   CD8^+ T cells in the lymph node increased at 1 week with ALN treatment.
   Bisphosphonates induce γδ T cells and their regulation of inflammatory
   cytokines ([190]63, [191]64) potentially altering the inflammatory
   environment within the joint. We were unable to examine the proportion
   of γδ T cells in the inguinal lymph node using CIBERSORT in the present
   study. Both PTH pretreatment and ALN treatment altered the
   transcriptomic response with loading in cortical and cancellous bone.
   However, for transcriptomic analyses, bone samples were collected from
   the metaphysis, not the subchondral bone plate directly beneath the
   cartilage, which limits our ability to draw direct conclusions about
   the subchondral bone response and cross-talk between bone and cartilage
   in OA progression.

   Our study has strengths and limitations. OA is a multifactorial disease
   caused by changes occurring in multiple tissues simultaneously and
   interdependently. By examining both tissue-level and transcriptomic
   changes across multiple tissues, we correlated early-stage
   transcriptomic changes with subsequent tissue-level structural damage
   and joint degeneration. However, bulk RNA-seq limited our ability to
   directly attribute signaling responses to specific cell populations.
   Because we used contralateral limbs as our loading controls, we also
   cannot identify whether systemic transcriptomic changes occurred
   because of loading or OA development. In addition, because of the
   challenge of simultaneously dissecting cartilage and adjacent
   subchondral bone from the same joint, we examined transcriptomic
   changes in the metaphyseal cortical and cancellous bone located below
   the growth plate, away from the joint surface. At the metaphysis, the
   response to loading likely was more anabolic and less reflective of OA
   development compared to subchondral bone directly beneath cartilage.
   Specifically, epiphyseal cancellous bone mass decreased with
   load-induced OA development in groups without ALN treatment, which
   likely would be evident in the transcriptomic response. Our
   transcriptomic results indicated suppression of immune-related pathways
   in cartilage with PTH pretreatment. Most prior work examining the role
   of joint inflammation and immune modulation in OA has focused on the
   role synovial tissue and synovium-cartilage cross-talk ([192]56,
   [193]65). In the present study, we uniquely examined transcriptomic
   changes in the lymph node to gain insight into the immune response but
   were unable to collect the synovium for RNA-seq limiting our
   mechanistic conclusions regarding the role of inflammation and immune
   modulation in PTH pretreatment and ALN treatment attenuating OA
   progression. Last, performing transcriptomic and IHC analysis at all
   time points would have provided additional insights throughout the
   course of the disease. However, given the large size of our study, we
   chose to prioritize these measures at early time points to identify
   gene and protein expression alterations that predated late-stage
   structural damage.

   In conclusion, our results indicated that therapies targeting the
   subchondral bone can attenuate the progression of load-induced OA
   pathology and that a shift in chondrocyte phenotype likely plays a
   mechanistic role in the efficacy of these therapies in joint tissues.
   In addition, early transcriptomic signatures of cartilage and lymph
   nodes can potentially serve as predictors of tissue-level changes and
   OA pathology at later time points, improving our understanding of the
   multiscale relationships driving load-induced OA. Together, our
   findings suggest the treatment potential of subchondral bone-modifying
   therapies to slow the progression of early OA and the diagnostic
   potential of early transcriptomic signatures to detect signs of OA
   before visible radiographic changes.

MATERIALS AND METHODS

Treatment conditions and cyclic mechanical loading

   PTH [PTH (1–34), 40 μg/kg per day, subcutaneous ([194]66)] or VEH
   (saline, same volume, subcutaneous) pretreatment was administered for 8
   weeks (5 days/week) to 18-week-old male C57BL/6 J mice (202 animals in
   total). Following pretreatment, cyclic mechanical loading was applied
   to the left tibia at a peak load of 9 N for 0 (pretreatment only, no
   loading; [195]Fig. 1A), 1, 2, or 6 weeks ([196]Fig. 2A) ([197]8,
   [198]29, [199]30). Concurrent with loading, pretreated mice received
   either ALN [73 μg/kg per day, intraperitoneal ([200]18)] or VEH
   (saline, same volume, intraperitoneal) treatment (5 days/week;
   [201]Fig. 2A). Mice were placed under general anesthesia during loading
   (2% isoflurane, 1.0 liter/min, Webster). Loading was applied to the
   left tibiae at 4 Hz for 1200 cycles (5 min) per day, 5 days/week
   ([202]8, [203]29, [204]30). The right tibiae served as contralateral
   controls. Following completion of the loading periods, mice were
   euthanized.

   In a subset of mice following 1, 2, and 6 weeks of loading, knee joints
   were harvested and fixed in 4% paraformaldehyde (PFA) overnight at 4°C
   for tissue-level analysis (n = 7 to 8 per group). In the
   pretreatment-only group, right limbs were harvested and fixed in 4% PFA
   overnight at 4°C, and left limbs were frozen in phosphate-buffered
   saline (PBS)–soaked gauze immediately following dissection for
   nanoindentation analysis. In a separate group of mice following 0, 1,
   or 2 weeks of loading, different tissues (cartilage, inguinal lymph
   nodes, and metaphyseal cancellous and cortical bone, after bone marrow
   removal) were retrieved. Cartilage samples were stabilized in RNAlater,
   while inguinal lymph node and bone samples were snap frozen in liquid
   nitrogen for for RNA isolation and bulk RNA-seq analyses (n = 4 to 8
   per group). RNA was extracted from tibial cartilage, metaphyseal
   cortical bone, cancellous bone, and inguinal lymph node tissues using
   protocols previously described ([205]67, [206]68) (n = 4 to 8/ea).
   Samples were not pooled. All experimental procedures were approved by
   the Institutional Animal Care and Use Committee at Cornell University
   and by the United States Army Medical Research and Development Command
   (USAMRMC) Animal Care and Use Review Office and are reported following
   ARRIVE guidelines ([207]69) and best practices to implement and report
   experiments using the tibial loading model ([208]70).

Nanoindentation of cartilage

   Nanoindentation was used to assess the effects of PTH pretreatment on
   cartilage mechanical properties. Left limbs of PTH or VEH
   pretreatment-only groups (n = 3 per group) were defrosted, and the
   femur, ligaments, tendons, and the meniscus were dissected to expose
   the tibial cartilage. Tibial plateaus with exposed cartilage were glued
   to the surface of a microscope slide (Loctite 409, Henkel Corp., Rocky
   Hill, CT). During indentation, tibial cartilage was maintained in PBS
   to prevent dehydration. Nanoindentation was performed on the medial and
   lateral tibial condyles with a spherical tip (borosilicate particle,
   radius = 5 μm, Au coating, purchased from Novascan Technologies) using
   an atomic force microscope (Asylum-MF3D-Bio-AFM-SPM), as previously
   described ([209]71). Indentation was repeated at seven to nine
   different locations per tibia.

Microcomputed tomography

   MicroCT scans were used to assess subchondral bone morphological
   differences with PTH pretreatment and ALN treatment and in response to
   cyclic loading ([210]Fig. 1A). After fixation in PFA, tissues were
   washed and transferred to 70% ethanol. Intact knee joints were scanned
   with an isotropic voxel resolution of 10 μm (μCT35, Scanco,
   Bruttisellen, Switzerland; 55 kVp, 145 μA, 600-ms integration time).
   Thickness and TMD of the subchondral cortical plate were measured at
   the proximal tibia. Bone volume fraction (BV/TV) and TMD were assessed
   in the tibial epiphysis. In all regions, bone was isolated by manually
   contouring the desired volume of interest (VOI). The VOI for the
   subchondral bone plate included the region of cortical bone beginning
   at the proximal end of the tibia and extended distally to the start of
   the cancellous bone in the epiphysis. The VOI of the epiphysis included
   cancellous bone distal to the subchondral plate and proximal to the
   growth plate.

Histological analysis

   Following microCT scanning, knee joints were decalcified in 10% EDTA
   for 2 weeks, dehydrated in an ethanol gradient, and embedded in
   paraffin for histological analysis. Coronal sections of 6-μm thickness
   were obtained from posterior to anterior using a microtome (Leica
   RM2255, Wetzlar, Germany). Cartilage damage was assessed with Safranin
   O/Fast Green staining of the medial and lateral tibial plateaus at
   90-μm intervals throughout the joint using OARSI histological scoring
   performed by a blinded researcher ([211]72). OARSI scores are reported
   as a whole joint mean. Cartilage thickness was measured on three
   representative Safranin O/Fast Green stained sections (posterior,
   middle, and anterior) on both medial and lateral tibial plateaus as
   previously described ([212]8, [213]18, [214]29, [215]30). Cartilage
   thickness is reported as a whole-joint mean and a posterior-region
   mean.

   Osteophyte formation in the medial tibia was assessed from
   Safranin-O/Fast Green–stained sections in loaded joints. Osteophyte
   size was defined as the medial-lateral width of the osteophyte at its
   widest point, measured from the medial edge of the epiphysis to the
   edge of the ectopic bone. Osteophyte maturity was scored on a scale of
   0 to 3 (0 = no osteophyte, 1 = primarily cartilaginous, 2 = mixture of
   cartilage, and bone, or 3 = primarily bony structure) as previously
   described ([216]43).

Immunohistochemistry

   To investigate the impact of PTH pretreatment, ALN treatment, and
   mechanical loading in cartilage integrity, MMP-13, RUNX2, PTH1R, and
   sclerostin levels were measured in tibial cartilage by IHC analyses.
   All cartilage IHC was completed in the pretreatment-only and 1-week
   loaded groups (1 slide per limb, n = 4 mice per group). Sections were
   deparaffinized, rehydrated, and incubated in citric acid buffer for
   antigen retrieval at 60°C for 1 hour. Sections were incubated with 3%
   H[2]O[2] solution for 10 min to quench endogenous peroxidase activity
   and then blocked with protein block from rabbit-specific horseradish
   peroxidase/3,3′-diaminobenzidine (HRP/DAB) (ABC) detection kit
   (ab64261, Abcam, Cambridge, UK) for 5 min. Sections were incubated
   overnight at 4°C overnight with anti-rabbit MMP-13 (1:50, ab39012),
   RUNX2 (1:50, ab23981), PTH1R (1:100, ab180762), or sclerostin (1:50,
   ab63097) polyclonal antibodies (Abcam, Cambridge, UK) or incubated with
   recombinant rabbit immunoglobulin (IgG) (1:50, ab172730) for negative
   controls. Sections were then treated with biotinylated goat anti-rabbit
   IgG(H + L) followed by streptavidin peroxidase for 10 min each
   [rabbit-specific HRP/DAB (ABC) detection kit, ab64261, Abcam,
   Cambridge, UK]. Last, sections were incubated with DAB chromogen
   solution [50X DAB chromogen with DAB substrate at 1:50 dilution,
   rabbit-specific HRP/DAB (ABC) detection kit, ab64261, Abcam, Cambridge,
   UK] for 5 min. Counterstaining was performed using hematoxylin for 1
   min. Positive cells were stained brown. The total area of positive
   immunostaining in the articular cartilage of the medial and lateral
   tibial plateaus was calculated and normalized to cartilage area (ImageJ
   software, NIH), as described ([217]73).

   Bone remodeling and Wnt signaling were assessed in the epiphysis and
   metaphysis by IHC analysis in the pretreatment-only and 1-week loaded
   groups. Sections were stained for TRAP or procollagen I as previously
   described ([218]74). The number of positively stained osteoclasts
   (TRAP) and osteoblasts (procollagen I) were quantified and normalized
   to measured bone surface (OsteomeasureXP v3.2.1.7). Osteoclast number
   was quantified for two slides per limb (n = 7 to 8 per group), and
   osteoblast number was quantified for one slide per limb (n = 4 per
   group). Sections were stained for sclerostin in bone following the same
   protocol as described above for cartilage. The number of sclerostin
   positive osteocytes were quantified and normalized to the total number
   of osteocytes (1 slide per limb, n = 4 per group).

RNA sequencing

   All methods were performed with sterile reagents under sterile
   conditions. Tissues were dissected as previously described ([219]24,
   [220]31, [221]67). Briefly, tibial cartilage was microdissected from
   the proximal subchondral plate and isolated under a dissecting
   microscope while submerged in RNAlater. Cartilage samples were stored
   in RNAlater, and metaphyseal cancellous and cortical bone ([222]Fig.
   1A) and inguinal lymph node samples were flash frozen in liquid
   nitrogen. Trizol (Life Technologies, Carlsbad, CA) was added, and
   samples were pulverized (Tissue Lyser II, Qiagen, Germantown, MD, USA)
   for 6 min at 30 Hz in tubes containing 2.8-mm ceramic beads. RNA was
   isolated using a phenol-chloroform extraction (Qiagen RNeasy kit,
   Qiagen, Germantown, MD, with deoxyribonuclease digest) ([223]67). All
   samples had an RNA quality number ≥ 6.0 (3730xl DNA Analyzer, Applied
   Biosystems, Waltham, MA, USA). 3′ RNA-seq was performed at the Cornell
   Genomics Core (NextSeq 500, Illumina, San Diego, CA) ([224]75), which
   produces a single sequence per transcript to limit expression bias from
   longer transcripts.

   Transcripts were aligned to the mm10 genome [STAR ([225]76)]. Each
   sample contained over 1 million of uniquely mapped reads to the mouse
   genome. Genes with less than three counts per million were not included
   for analysis. Differential expression between loaded and control limbs
   was determined using a paired design [edgeR ([226]77)] and defined as
   genes with a fold change >2 or <−2 with a < 0.05 FDR. To examine
   biological processes, the log[2]fold change ranked gene list from the
   differential gene expression analysis was used to compute biological
   processes enriched with loading (clusterProfiler, Gene Ontology
   geneset, p.adj <0.05).

   Selected tissue-specific genes with well described roles in the
   homeostasis and OA-related pathology were examined. To examine these
   individual genes, the normalized gene counts [reads per million mapped
   reads calculated in DESeq2 with Trimmed Mean of M-values correction
   ([227]78)] were standardized between loaded and control limbs for each
   group.

CIBERSORT for immune cell analysis

   CIBERSORT, a deconvolution-based algorithm that uses standardized gene
   expression data to calculate cell fractions based on gene expression
   profiles ([228]44), was used to estimate proportion of immune cells in
   the inguinal lymph node at 0, 1, and 2 weeks. The gene composition of
   each cell was determined by calculating the expression level of each
   gene in each immune cell, with a mouse-specific gene archive containing
   eight types of immune cells ([229]79). Normalized gene counts were used
   as inputs.

Statistical analysis

   Histological scores, cartilage thickness, IHC measures, and bone
   parameters of PTH and VEH pretreatment-only (nonloaded) limbs were
   compared using a Student’s t test. Cartilage stiffness of PTH and VEH
   pretreatment-only limbs were compared using a linear mixed-effects
   model with a fixed effect of pretreatment and a random mouse effect to
   account for repeated measurements with the same mouse limbs. The
   effects of loading, duration, pretreatment, and treatment were
   determined using a linear mixed-effects model with fixed effects of
   loading (contralateral control versus loaded), duration (1 week versus
   2 weeks versus 6 weeks), pretreatment (PTH versus VEH) and treatment
   (ALN versus VEH), and a random mouse effect. Post hoc analysis was
   performed using Tukey’s test for significant interaction effects and t
   test for individual effects. Normality was confirmed visually using
   histograms and QQ plots of the residuals. Significance was set at P <
   0.05.

Acknowledgments