MMP14 cleaves PTH1R in the chondrocyte-derived osteoblast lineage, curbing signaling intensity for proper bone anabolism

Bone homeostasis is regulated by hormones such as parathyroid hormone (PTH). While PTH can stimulate osteo-progenitor expansion and bone synthesis, how the PTH-signaling intensity in progenitors is controlled is unclear. Endochondral bone osteoblasts arise from perichondrium-derived osteoprogenitors and hypertrophic chondrocytes (HC). We found, via single-cell transcriptomics, that HC-descendent cells activate membrane-type 1 metalloproteinase 14 (MMP14) and the PTH pathway as they transition to osteoblasts in neonatal and adult mice. Unlike Mmp14 global knockouts, postnatal day 10 (p10) HC lineage-specific Mmp14 null mutants (Mmp14ΔHC) produce more bone. Mechanistically, MMP14 cleaves the extracellular domain of PTH1R, dampening PTH signaling, and consistent with the implied regulatory role, in Mmp14ΔHC mutants, PTH signaling is enhanced. We found that HC-derived osteoblasts contribute ~50% of osteogenesis promoted by treatment with PTH 1–34, and this response was amplified in Mmp14ΔHC. MMP14 control of PTH signaling likely applies also to both HC- and non-HC-derived osteoblasts because their transcriptomes are highly similar. Our study identifies a novel paradigm of MMP14 activity-mediated modulation of PTH signaling in the osteoblast lineage, contributing new insights into bone metabolism with therapeutic significance for bone-wasting diseases.

2 titration of PTH signaling in the osteoblast lineage, contributing new insights into bone metabolism with therapeutic significance for bone-wasting diseases.

Introduction
Developing and maintaining appropriate proportions of cells of the osteoblast lineage and other cell types (e.g., osteoclasts, endothelial cells, adipocytes, stem cells) in the bone marrow are crucial for healthy bones. These cell types together maintain bone mass via concerted activities of bone building (anabolic) by osteoblasts versus bone resorption (catabolic) by osteoclasts, with input from factors produced by osteocytes to orchestrate the remodeling process (1).
Through this continuous remodeling, and balanced anabolic and catabolic activity, a homeostatic condition is achieved (2). This homeostasis is regulated by growth factors and cytokines produced within bone and systemic factors such as parathyroid hormone (PTH) and estrogens (3,4). An imbalance in proportions of these cells in bone impairs bone resorption and formation, which can result in diseases of bone mass such as osteopenia, osteoporosis, and osteopetrosis (5,6).
Long bones form by endochondral ossification, a process in which chondrocytes differentiate, proliferate, mature and become hypertrophic, forming a cartilaginous growth plate that coordinates longitudinal bone growth and is recapitulated in fracture repair (7,8). Hypertrophic cartilage is remodeled by matrix metalloproteinases (MMPs) (reviewed in (9). The chondrocyte differentiation cascade is coordinated by many transcription factors and signaling pathways, including reciprocal signaling via the parathyroid hormone-related protein (PTHRP)-Indian hedgehog (IHH) feedback loop (10,11). Resting chondrocytes secrete PTHRP, that binds to parathyroid hormone 1 receptor (PTH1R) expressed by pre-hypertrophic chondrocytes, acting to delay their differentiation to hypertrophic chondrocytes (HCs) (12,13). HCs specifically produce collagen type X (Col10a1), down-regulate PTH1R and undergo distinct phases of cell enlargement for bone elongation (14,15). Recent work has shown that osteoblasts in trabecular bone are derived from HCs in the growth plate and from osteoprogenitors in the perichondrium which accompany invading blood vessels (15)(16)(17). HCs have been shown to contribute to the full spectrum of cells in the osteoblast lineage in endochondral bone, trabeculae, endosteal/endocortical bone, including osteocytes, and, to a minor degree, bone marrow stromal cells and adipocytes (15,(17)(18)(19)(20). HC transformation to osteoblasts also occurs in bone healing (7,15,17). The functional importance of the HC lineage has been demonstrated in mice, which have reduced bone mass as a consequence of ablating β-catenin and Irx3/5 genes specifically in HCs (19,21). However, it is not known whether the HC-derived osteogenic lineage is molecularly distinct from the non-HC lineage (contributed by the perichondrium, periosteum, bone marrow mesenchymal stem cells), which specific MMPs act at the chondro-osseous junction to facilitate the transition of HCs to osteoblasts and what physiological contribution(s) they make to maintaining bone homeostasis and anabolism in response to extrinsic signals.
Here, using single-cell transcriptomics, we characterized the different osteoblast populations in endochondral bone and found that their transcriptomes are broadly similar to that of non-HC derived osteoblasts. We found HC-derived osteoblasts activate membrane-type 1 metalloproteinase 14 (MMP14) and the PTH/PTH1R pathway upon transition into sub-chondral bone. Using mouse mutants and biochemical approaches, we identified the cleavage of PTH1R by MMP14 in HC-derived osteoblasts as a novel mechanism that curbs the intensity of PTH signaling. Like their non-HC-derived counterparts, HC-derived osteoblasts respond to exogenous PTH treatment, and contribute significantly to the bone anabolic response. MMP14 modulation of PTH signaling intensity controls the differentiation and survival of osteoblasts and thereby the anabolic response to PTH in building bone mass.
We identified osteoblasts as HC-derived if they transcribed tdTomato, using the alignment to the tdTomato sequence and a surrogate sequence for it (WPRE (26, 27), Methods, Fig. S1, A, E-G). We annotated the different cell populations based on the hundreds of differentially expressed signature genes (Extended-Data Table   1) associated with each cell population and by reference to the literature (Fig. S1).
Immature osteoblasts represent an important early stage in the HC to osteoblast transition and 18.1% of them were found to express tdTomato and thus were annotated as HC-derived ( Fig. 1C and Fig. S1H). We used well established signature genes (23-25) such as Pth1r, Col10a1 and Grem1 to annotate the different populations as Pre-HCs, HCs and immature osteoblasts respectively  Table 2) (16) and confirms the lineage continuum of HCs and osteoblasts.
Given the need for HCs to be released from the cartilage matrix as they move across the chondro-osseous junction, MMP genes are candidates, especially those expressed at that region and in the adjacent sub-chondral bone. Amongst the Mmps, Mmp13 was expressed in both the HC and immature Ob populations ((15, 23) Fig. 1C, Extended-Data Table 1), consistent with the literature (28). By contrast, Mmp14 (encoding a membrane-anchored MMP (29) was not expressed in HCs but was enriched in immature osteoblasts (Fig. 1C, Fig.S1A, Extended-Data Table 1). Consistent with the scRNAseq results, immunostaining for MMP14 detected no protein in HCs and strong expression in LacZ-labeled HC-descendant cells, especially those located at the chondro-osseous junction (Fig. 1D) (identified by co-staining with LacZ in C10Cre; Rosa26-LacZ mice (15)).
We next investigated whether the molecular characteristics of the osteogenic signatures in HC-derived osteoblasts persisted by maturity at P56 (Fig. 1, E-J and  Table 1). Using the same cutoff for tdTomato expression, 23.6% of the P6 and 12.1% of the P56 osteogenic cells (combining immature and mature Obs) were HC-derived (Fig. 1H). Overall, the HC-derived osteogenic populations in young post-natal and mature mice are highly comparable in molecular signatures, cell-type and lineage composition, consistent with a hypertrophic origin of osteogenic cells (Fig. 1J). The variations in relative percentages may also be related to the limits in sensitivity in detecting tdTomato mRNAs and differences in the ease of releasing deeply embedded cells within the bone matrix versus more superficially located ones. However these frequencies are in agreement with the broad ranges (18-60%) reported previously for HCderived osteoblasts from a range of developmental and postnatal ages (15,17,30).
We also assessed how similar HC-derived and non-HC-derived cells are. While between HC-and non-HC-derived cells (Fig. S1J). This suggests a high degree of similarity between these two sources of immature osteoblasts. Interestingly, both sources expressed progenitor cell (Lepr and Grem1 (31) and mesenchymal cell (Prrx1, Twist1, Pdgfrb) markers (Fig. 1C), which is consistent with a recent report (30) suggesting these cells pass through progenitor and mesenchymal states.
Comparing the immature and mature HC-and non-HC-derived osteoblasts combined, we detected 48 differentially expressed genes , of which 22 were expressed higher in the HC-derived cells, including some mature osteoblast markers, such as Ifitm5, Col22a1, and Smpd3; and 26 genes were expressed higher in the non-HC-derived cells, such as Postn, Igfbp5, Col3al (Fig. S1J). This small set of differentially expressed genes may reflect heterogeneity and differences in maturities between osteoblasts of different origins, the significance of which will be investigated in future studies. Overall, there was a broad similarity in transcriptomic characteristics for the HC-derived and non-HC-derived osteoblasts.
Global MMP14 is required for proper translocation of HC-derivatives to trabecular bone Mmp13 and Mmp14 are candidate facilitators for the translocation of HCs to the subchondral space. Mmp13 is expressed in late HCs (15,23) as well as osteoblasts. Chondrocyte-specific Mmp13 knockout mice display abnormal growth plates with disrupted terminal hypertrophy and increased trabecular bone (28), which could reflect an impact on the HC-Ob lineage continuum. By contrast MMP14 is expressed specifically at the chondro-osseous junction in immature HCderived osteoblasts but not in HCs themselves ( Fig. 1, C, D). Amongst MMPs, Mmp14 null mutants display the most severe skeletal phenotypes, including loss of trabecular bone, over-activity of osteoclasts, impaired angiogenesis and reduced calvarial ossification (29,(32)(33)(34), consistent with expression of Mmp14 by many cell types, such as skeletal progenitors, osteoblasts, osteoclasts, endothelial 9 cells and bone marrow stromal cells (Fig. S2A) (34)(35)(36)(37)(38). The impaired bone formation in Mmp14-deficient mice and the expression of MMP14 in HC-derived osteoblasts raised the possibility that MMP14 regulates their differentiation. The impact of chondrocyte-specific knockout of Mmp14 on trabecular bone formation is not known. Given MMP14's pivotal role in trabecular bone formation, we focused on Mmp14 and asked whether dysfunction in lineage progression of HCderived cells was responsible for the severe bone deficit in Mmp14 knockout mice.
We used the C10Cre;Rosa26-YFP reporter to lineage trace HC-derivatives in global Mmp14 knockout mice, and found an increased accumulation of HCdescendants at the chondro-osseous junction compared to WT controls ( Fig. S2B and C). There was also a small increase in proliferating HC-derivatives (marked by 5-Ethynyl-2'-deoxyuridine, Edu) in the region immediately below the chondroosseous junction (Fig. S2C). However there were fewer proliferating HCderivatives in further distal regions below the chondro-osseous junction, consistent with their accumulation there and impaired translocation into the trabecular bone global knockouts and was intrinsic to HC-derived cells themselves or involved extrinsic influences from other cell types, we genetically inactivated Mmp14 in HC-descendants by generating HC-specific conditional Mmp14 Flox (Mmp14 F/F ) mutants using C10Cre (abbreviated as Mmp14 ΔHC ) ( Fig. 2A). We found removing Mmp14 in HC-descendants did not recapitulate the accumulation of HC-descendants at the chondro-osseous junction observed in Mmp14 -/mice (Fig. S2C). We used tamoxifen mediated pulse-labelling and chasing of Mmp14 deficient HCdescendants using C10Cre-ERT;Mmp14 F/F ;RtdT to assay the dynamics of their translocation into the subchondral space. We found the localization and distribution of HC-descendants were unaffected in Mmp14 ΔHC mice (Fig. 2B). These results suggest the aberrant stagnation of HC-derivatives at the chondro-osseous junction was not intrinsic to a defect in these cells themselves but might be a consequence of MMP14 deficiency elsewhere, for example in the invading vascular cells and/or the accompanying osteoprogenitors from the perichondrium that co-migrate with blood vessels to populate the primary spongiosa (16,39). Vascular invasion is required for bone formation (39,40). We therefore examined the vascular capillaries in Mmp14 -/mice and found that both vascular density and endothelial cell count (measured using the marker ENDOMUCIN) were decreased compared with control mice (Fig. S2D). In vitro and in vivo studies have shown Mmp14 is important for angiogenesis, neovessel formation and migration (29,41). The abnormal vascularization in Mmp14 null mice may therefore be a major contributor to the compromised translocation of HC-derived cells to the trabecular bone. To test if the number of osteoclasts was affected in Mmp14 ΔHC mutants, we performed Tartrate-resistant acid phosphatase (TRAP) staining and quantitated the number of osteoclasts in trabecular bone (Fig. S3C). In line with reports showing MMP14 deficiency does not affect osteoclastogenesis (42,43), the number of osteoclasts was comparable between Mmp14 ΔHC mutants and controls at P10, suggesting the increased number of HC-derived osteogenic cells, not reduced resorption, was a major contributor for increased bone mass at this stage.
Overall these results are in line with previous reports showing a combined inactivation of MMPs is required for osteoclast dysfunction (43).
A small proportion (less than 2%) of WT HC-descendants have been shown to become bone marrow adipocytes (19,20). We found the total frequency of HCderived bone marrow adipocytes in Mmp14 ΔHC mice was unchanged compared with control, suggesting MMP14 does not influence the adipogenic fate choice of HC-derived cells. However non-HC derived adipocytes in Mmp14 ΔHC were substantially fewer than in controls, suggesting a non-cell-autonomous impact of MMP14 deficiency on signals from HC-derived cells on marrow adipogenesis in Mmp14 ΔHC mice (Fig. S3D). Such non cell-autonomous effects could be mediated by the effect of MMP14 deficiency on the collagenous microenvironment that coordinates adipogenesis (44,45).

PTH1R is a substrate of MMP14
Next we sought to determine the underlying reason for the increased bone in Mmp14 ΔHC mice. Reduced calvarial osteogenesis and osteoclast overactivity in Mmp14 mutant mice were found attributable to cleavage of ADAM9, RANKL and extracellular matrix proteins by MMP14, regulating FGF, RANK and YAP/TAZ signaling (33,34,42). However, these findings are insufficient to account for the increased trabecular bone observed in Mmp14 ΔHC mice, suggesting a yet undiscovered mechanism could be the underlying cause of the phenotype.
Interestingly, MMP14 was shown to be a downstream target of PTH signaling in osteocytes (46). In a previous report, because cleavage of PTH1R can be inhibited by TIMP2, but not TIMP1 it was proposed that MMP-dependent cleavage of PTH1R causes reduced stability and degradation of PTH1R, raising the possibility that cleavage by a MMP might inhibit PTH signaling (47). However the identity of the responsible MMP was unknown. This collective evidence, including the coexpression of Mmp14 and Pth1r from scRNA data (Fig. 3A), strongly suggested a direct molecular link between MMP14 and PTH/PTH1R signaling pathway could be possible. Therefore, we hypothesized MMP14 can proteolytically process and perhaps inhibit PTH/PTH1R signaling. Knocking out MMP14 in HC-derived osteogenic progeny could remove its inhibitory effect on PTH pathway.
To test the hypothesis, we assayed the impact of co-expressing (Human influenza Given that rhMMP14 directly cleaves PTH1R-ECD into fragments with size around 20kDa and 15kDa, we propose at least one putative cleavage site exist around amino acid 55 to 65 in PTH1R. Computational prediction also suggest a possible cleavage site exist at amino acid around 61 (49). To test this, we incubated rhMMP14 with a synthetic peptide with amino acid sequence 55-67 from PTH1R.
Using mass spectrometry, we found a peak at 770 Da consistent with molecular mass of a VLQRPAS fragment, suggesting amino acid 61 is one of the cleavage sites of MMP14 (Fig. 3F).

MMP14 inhibits PTH/PTH1R signaling
Prior molecular and structural studies have demonstrated that the exon 2 encoding region of PTH1R, which harbors the cleavage site by MMP14, is dispensable for both binding to PTH/Pthrp and receptor function (50)(51)(52)(53). Since MMP14 has a wide range of substrates ranging from collagens to transmembrane ligands, we

HC-derived osteogenic cells respond to PTH which is enhanced in Mmp14 ΔHC
Having demonstrated MMP14 can directly cleave PTH1R to negatively regulate its In compound C10Cre;RtdT;C1-GFP mice, mature osteoblasts, pre-osteoblasts and HC-descendants were labeled by GFP, OSTERIX and RFP, respectively. We found many more HC-descendants in the trabecular region that co-expressed OSTERIX, a marker of osteoprogenitors, and C1-GFP in response to PTH treatment than in controls (Fig. 6A, B). These HC-descendants showed positive pCREB + RFP + staining confirming PTH signaling activity in HC-derivatives ( Fig.   6C). HC-derived cells constituted approximately 50% of all trabecular osteoblasts with and without PTH treatment (Fig. 6D). HC-derived osteoblasts therefore represent a significant osteogenic population directly contributing to half of the total osteoblast population and capable of responding to PTH treatment in the same way as non-HC-derived osteoblasts (Fig. 6D). In Mmp14 ΔHC mice, the number of RFP+ and RFP+GFP+ and RFP+OSX+ cells further increased compared to control mice treated with PTH, consistent with the heightened response to PTH treatment in Mmp14 deficient HC-derived cells (Fig. 6D).
By contrast with enhanced osteogenesis in trabecular bone, BV/TV and cortical BMD were reduced in PTH-treated Mmp14 ΔHC mutants compared to PTH-treated controls (Fig. S6A). There was no significant increase in RFP+GFP+ HC-derived endosteal cells in Mmp14 ΔHC mutants. Although both HC-derived (RFP+GFP+) and non-HC derived (RFP-GFP+) endosteal cells responded to PTH treatment, the response to PTH was not enhanced in mutants, suggesting differences in the effect of MMP14 deficiency and PTH sensitivity of HC-derived endosteal cells compared to the non-HC-derived counterparts. Although the number of HCderived RFP+ osteocytes were also not different between controls and mutants, there was a significantly enhanced response of HC-derived osteocytes to PTH in the cortical bone of Mmp14 ΔHC mutants ( Fig. S6A and B). These results may be relevant to the interesting observation that PTH treatment results in increased trabecular bone in patients but had no effect on endo-cortical, cortical or periosteal bone (55). Whether the discrepancy in outcome of PTH treatment or MMP14 deficiency on cortical bone and trabecular bone is related to a more transient effect on PTH signaling on bone synthesis in the former than in the latter and /or remodeling differences, are questions for future study.
HC-descendants persist and contribute to the anabolic response to PTH in aged mice Although, unlike in humans, the growth plate of mice does not close, in adult mice by one year of age, the hypertrophic cartilage is vestigial and endochondral ossification diminishes substantially (56). To test whether the HC-derived osteogenic cells could also contribute to the response to PTH treatment in older mice, we treated 1 year old C10Cre;RtdT mice with PTH for 4 weeks (Fig. 7A-C (i-iii)). Like in young mice, HC-descendants at 1 year old responded to PTH  treatment by producing more osteogenic progeny (Fig. 7C(i-iii)), suggesting sustained ability of HC-derived osteoblasts to contribute to PTH-stimulated osteogenesis with aging. Taken together our findings implicate a molecular link between MMPs and PTH signaling in regulating the activity of HC-derived osteoblasts in postnatal and adult mice (Fig. 8).

Discussion
Recent studies have established the concept that in the growth plate, resting chondrocytes function as skeletal stem cells for continuous supply of proliferating and pre-hypertrophic chondrocytes and HCs (12,13). These cells form part of a continuum in which the HCs transition into the full osteogenic lineage, contributing to the formation of primary spongiosa and trabecular bone (9,15,(17)(18)(19) and participating in bone regeneration and healing (7,15,17,57  By contrast a pan osteoblast-specific or HC-derived osteoblast-specific targeted manipulation approach would have a significant advantage as MMP14 activity (and therefore PTH signaling) only needs to be manipulated in a single cell-type and it avoids systemic intermittent injection of PTH which would thereby ameliorate side effects.

TUNEL labeling and Edu incorporation assay
Proliferation and apoptosis assays were described in Yang et.al. (15). Edu labeling was performed with intraperitoneal injections of Edu at 150 mg/kg body weight and the mice were sacrificed 2 hours later. Labeled cells in paraffin sections were detected using Click-iTR EdU imaging kits from Invitrogen. in situ terminal deoxynucleotidyl transferase deoxyuridin triphosphate nick end labeling (TUNEL) assay was performed using in situ cell death kit from Roche (Basel, Switzerland).

Quantitation of cells
Fluorescence images were processed with ImageJ (developed by National Institutes of Health, USA). For young mice at P3 or P10, the number of cells was counted manually. For mice aged 2 months, images were taken using Carl Zeiss LSM confocal 800 system (Carl Zeiss, Oberkochen, Germany). Overlap of DAPI with RFP (tdTomato), GFP or OSTERIX was considered as one single cell as in

Extraction of protein from mouse trabecular bone
Trabecular bones were harvested from the Mmp14 -/and littermate control mice, followed by homogenizing bones in RIPA with a bullet blender for 15 minutes at 4 ℃. After leaving on ice for 10 minutes, the samples were centrifuged at 13, 000 × g for 15 min at 4 ℃. Next, the supernatants were collected and processed for Bio-Rad protein assay for protein concentration determination.
In vitro cleavage of rPTH1R and peptide fragments from PTH1R

Western blot analysis
Western blotting was performed as described previously (34).

Co-immunoprecipitation assay
Co-immunoprecipitation assay was performed as previously described (34). We confirmed that tdTomato and WPRE are linearly correlated but with enhanced detectability in the latter (Fig. S1 C and D). As such, we used WPRE as a surrogate for tdTomato. We assigned a cutoff value of WPRE>14 to match our experimental determination of the frequency of tdTomato-positive HCs ( (15,19)~75%), which also corresponds to a saddle point in the WPRE distribution ( Fig. S1 B-E). This cutoff was used to separate the osteogenic cells into HCderived osteoblasts and non-HC-derived osteoblasts (Fig. 1 C, K, N). A kernel logistic regression model was trained to predict the boundaries between HCderived osteoblasts and non-HC-derived osteoblasts, using the KLR package in R (64) (Fig. S1F). To isolate endochondral cells, cells with fewer than 1,600 genes or clusters expressing Ptprc (CD45) were excluded. Data were further processed with Seurat (v3. approach (Fig. 1J). Gene-ontology analyses were performed using GSEA (gseamsigdb.org).

Statistics
Results were represented as mean ± SD. Statistical evaluation was done by nonparametric 2-tailed Student's t test using GraphPad Prism version 8 for Microsoft (www.graphpad.com). ONE-WAY ANOVA for multiple testing was used for Fig. 4, Fig. 5. P < 0.05 was considered significant.          Pre-HC

HC-pre-Ob
HC-pre-Ob HC-pre-Ob HC-pre-Ob  Upon PTH binding, PTH1R generates cAMP signals to activate cAMP response element binding protein (Creb). Activation of PTH1R causes upregulation of MMP14. MMP14 in turn cleaves PTH1R to inhibit its signaling activity. In the chondrocyte lineage, prehypertrophic chondrocytes (pre-HC) differentiate to HCs, translocate to the subchondral space and subsequently to HC-pre-osteoblasts contributing to trabecular, endosteal and endocortical bone and can respond to PTH treatment. Together, MMP14 moderates the HC derivatives' response to PTH ensuring a controlled supply of osteoblast precursors and thereby bone anabolism. Obs, osteoblasts, PHZ, Hypertrophic zone. HZ, Hypertrophic zone. TB, Trabecular bone.