Destroy to Rebuild: The Connection Between Bone Tissue Remodeling and Matrix Metalloproteinases

Bone is a dynamic organ that undergoes constant remodeling, an energetically costly process by which old bone is replaced and localized bone defects are repaired to renew the skeleton over time, thereby maintaining skeletal health. This review provides a general overview of bone’s main players (bone lining cells, osteocytes, osteoclasts, reversal cells, and osteoblasts) that participate in bone remodeling. Placing emphasis on the family of extracellular matrix metalloproteinases (MMPs), we describe how: (i) Convergence of multiple protease families (including MMPs and cysteine proteinases) ensures complexity and robustness of the bone remodeling process, (ii) Enzymatic activity of MMPs affects bone physiology at the molecular and cellular levels and (iii) Either overexpression or deficiency/insufficiency of individual MMPs impairs healthy bone remodeling and systemic metabolism. Today, it is generally accepted that proteolytic activity is required for the degradation of bone tissue in osteoarthritis and osteoporosis. However, it is increasingly evident that inactivating mutations in MMP genes can also lead to bone pathology including osteolysis and metabolic abnormalities such as delayed growth. We argue that there remains a need to rethink the role played by proteases in bone physiology and pathology.


INTRODUCTION
Bone is a hard, dense, rigid form of highly specialized connective tissue making up the skeleton of vertebrates. Bone protects internal organs, supports body structures, and aids in locomotion (Maffioli and Derosa, 2015). In addition, bone provides an environment for hematopoiesis (i.e., formation and development of blood cells) in the bone marrow, and acts as a homeostatic reservoir of calcium, phosphorus, insulin-like growth factors, transforming growth factor-β, and cytokines. Bone buffers the blood against drastic pH changes, thus detoxifying the circulation from heavy metals (Rauner et al., 2012). Bone develops by intramembranous ossification (e.g., bone of the clavicle, some skull bones), endochondral ossification (e.g., the appendicular and axial skeleton) or pseudo-metamorphic ossification (Rauner et al., 2012).
Bone remodeling is a complex process involving the sequential resorption of bone tissue and deposition of new bone at the same site (Kerschan-Schindl and Ebenbichler, 2012). Together with bone structure, geometry, size, and density, remodeling determines bone's overall mechanical properties (e.g., the strength) (Mosekilde et al., 1993;Jiang et al., 1997;Ikeda et al., 2003;Shahnazari et al., 2009) as well as enables the repair of damaged bone and the adaption of bone to changing biomechanical forces (Kerschan-Schindl and Ebenbichler, 2012).
We review here the prevailing view of the bone remodeling process with an emphasis on well-accepted and newly emerging roles played by matrix metalloproteinases (MMPs) and cysteine proteinases in this process. Finally, we review the increasing number of instances in which inactivating mutations in MMP genes are found to lead to bone pathology including osteolysis and metabolic abnormalities such as delayed growth.

GENERAL OVERVIEW ON THE CYCLE OF BONE REMODELING
The bone remodeling process consists of four distinct consecutive phases spanning over 3-6 months (Datta et al., 2008).
The first phase of bone remodeling is known as the 'activation phase' and can be triggered by mechanical and nutritional stress on the bone as well as by hormones (e.g., parathyroid hormone, estrogen) (Parra-Torres et al., 2013). As described in Table 1, terminally differentiated osteocyte cell is a key player in the activation phase (Rauner et al., 2012;Parra-Torres et al., 2013).
The second phase lasts 8-10 days (Teitelbaum, 2007) and is called the 'bone resorption phase' -a process by which large multinucleated osteoclast cells break down old bone organic matrix impregnated with minerals (e.g., calcium phosphate nanocrystals), as described in Table 2.
The third 'reversal' phase connects osteoclastic bone tissue resorption and osteoblastic bone tissue formation (Delaisse, 2014) and lasts 7-14 days (Pettit et al., 2008;Hienz et al., 2015). After departure of the osteoclast from a cavity in bones undergoing resorption, which is a resorptive lacuna known as the Howship's lacuna, bone lining cells occupy the Howship's lacuna and clean it (Everts et al., 2002). The cleaning process occurs by enwrapping and digesting non-mineralized collagenous proteins protruding from the bone surface left by osteoclasts. This cleaning process is a requirement for the subsequent deposition of a first layer of collagen along the Howship's lacuna (Everts et al., 2002). Four types of osteoclastderived coupling factors stimulate bone formation during the reversal phase: (i) Matrix-derived factors including transforming growth factor-β, bone morphogenetic protein-2, platelet-derived growth factor, and insulin-like growth factors, which are released during bone tissue resorption, (ii) Osteoclast-secreted factors, including cardiotrophin-1, sphingosine-1-phosphate, collagen triple helix repeat containing 1, and complement factor 3a, (iii) Osteoclast membrane-bound factors such as EphrinB2 and Semaphorin D,and (iv) Structural changes brought about by the osteoclast on the bone tissue surface (Sims and Martin, 2014). Reversal cells originating from pre-osteoblast cells (Andersen et al., 2013) colonize the osteoclast-eroded surface and respond to osteoclast-derived messages and coupling factors along with fibroblast-like cells covering the surface of bone (known as bone lining cells), osteoblast precursors, and canopy cells (Delaisse, 2014;Sims and Martin, 2014;Lassen et al., 2017;Pirapaharan et al., 2019).
The fourth phase of the bone remodeling cycle is 'formation, ' when mononucleate osteoblast cells synthesize new bone organic matrix formed by collagen fibers and non-collagenous proteins (e.g., bone sialoprotein, osteopontin, osteocalcin, proteoglycans) that later becomes surrounded and impregnated with mineral deposit mainly in the form of calcium hydroxyapatite. A summary of osteoblastogenesis, the roles played by osteoblasts during this last phase, and the fate of osteoblasts is described in Table 3.

MATRIX METALLOPROTEINASES: MODULATORS OF BONE REMODELING
Matrix metalloproteinases are a family of at least 24 highly homologous, multi-domain enzymes (Figure 1) with the capacity to degrade virtually all extracellular matrix components including collagen, aggrecan, elastin, and fibronectin (Lu et al., 2011;Fernandez-Patron et al., 2016).

Origin of osteocytes
Differentiation stages: (i) From mature osteoblasts to preosteocyte type I with dendritic projections formation; (ii) from preosteocyte type I to preosteocyte type II with cytoskeleton rearrangement; (iii) from preosteocyte type II to preosteocyte type III (mature osteocyte trapped within the mineralized bone matrix) with canaliculae formation (Hirao et al., 2007;Paiva and Granjeiro, 2017).
Osteocytes are good mechanosensors (i.e., they detect changes of mechanical stimuli) in bone tissue which serve to sense and respond to alterations produced when a bone is mechanically loaded. Such alterations may be physical deformation of the bone matrix, fluid flow shear stress generated by variations in canalicular fluid flow and electrical streaming potentials (Bonewald and Mundy, 1990;Mundy, 1993;Manolagas, 2000;Miyauchi et al., 2000;Bonewald and Johnson, 2008;Datta et al., 2008;Parra-Torres et al., 2013;Takemura et al., 2019). Mechanical strain signal is converted into a cellular response (i.e., biochemical signals) with the participation of membrane proteins (such as CD44, connexins, integrins, and ion channels) and downstream mediators of intracellular signaling (such as guanine regulatory proteins, mitogen activated protein kinase, cyclic adenosine monophosphate, inositol triphosphate, and intracellular calcium) (Rawlinson et al., 1996;Burger and Klein-Nulend, 1999;Mikuni-Takagaki, 1999;Miyauchi et al., 2000;Gu et al., 2001;Alford et al., 2003;Kapur et al., 2003;Plotkin et al., 2005;Rubin et al., 2006;Miyauchi et al., 2006). On the other hand, bone remodeling is also controlled by upregulation of RANKL and sclerostin in response to a decrease in mechanical signals (Parra-Torres et al., 2013).The precise signaling biochemical pathways (e.g., Wnt/β-catenin) and regulatory mechanisms that may mediate adaptive responses activated by mechanical loading and unloading in bone remain to be completely delineated (Dallas et al., 2013;Parra-Torres et al., 2013).
Other consequences of osteocyte activities on bone remodeling Retraction of the bone lining cells (elongated mature osteoblasts) on the endosteal surface (which is a thin layer of cell-rich connective tissue), and also digestion of the underlying collagenous membrane by collagenases (Murray et al., 1995;Karsdal et al., 2001;Datta et al., 2008;Kerschan-Schindl and Ebenbichler, 2012 of an intact pro-peptide accounts for the latency of pro-MMPs, which can be overriden through the activation of a "cysteine-switch" mechanism (Van Wart and ). The pro-peptide contains a cysteine residue that prevents catalytic activity when it is coordinated with a Zn(II)-ion in the catalytic domain (Springman et al., 1990;Van Wart and Birkedal-Hansen, 1990). The cysteine-Zn(II) interaction can be disrupted by alkylating compounds such as the organomercurial 4-aminophenylmercuric acetate as well as by serine proteases and other MMPs such as membrane-type MMPs, which act at the cell surface to which they anchor through their transmembrane domain/short cytoplasmic tail or by glycosylphosphatidylinositol linkage (Bonnans et al., 2014). MMP autolysis is another mechanism of activation mediated by allosteric perturbation of the inactive proenzyme (Springman et al., 1990;Van Wart and Birkedal-Hansen, 1990; Origin of osteoclasts Differentiation stages: Hematopoietic stem cell precursors differentiate into monocyte and macrophage, and then they fuse into end-differentiated multinucleated (bone resorbing) cells (Tanaka et al., 1993;Quinn et al., 1998;Roodman, 1999;Udagawa et al., 1999;Holmbeck and Szabov, 2006;Bar-Shavit, 2007;Bruzzaniti and Baron, 2006). Osteocyte apoptosis is thought to contribute to the recruitment of osteoclast precursors by diminishing the secretion of osteocyte-derived factors (e.g., TGF-β) that have inhibitory effect on osteoclast formation (Heino et al., 2002;Aguirre et al., 2006).

Roles played by osteoblasts
Once osteoclasts have created a resorption cavity and detached from the bone surface, osteoblasts move into the cavity to initiate bone formation (Datta et al., 2008). Osteoblasts synthesize and lay down new unmineralized bone matrix (osteoid), which is subsequently mineralized (e.g., forming hydroxyapatite) over a period of about 20 days. Osteoblasts also synthesize and secrete the bone matrix proteins osteopontin, osteocalcin, bone sialoprotein, proteoglycans, and alkaline phosphatase (Baron, 1989;Roach, 1994;Ducy et al., 2000;Datta et al., 2008;Hienz et al., 2015). Why is the synthesis of non-collagen proteins necessary? These non-collagenous bone matrix proteins help to coordinate matrix mineralization and are essential for cellular adhesion (such as chemoattractant activity by osteocalcin), and regulation of cell activity (such as the osteopontin-and osteonectin-displayed cell activities) during coupling of bone resorption and formation (Robey, 1989;Raynal et al., 1996;Hienz et al., 2015). There is another function of osteoblasts that is worth highlighting. Osteoblasts also inhibit the ability of osteoclasts to degrade osseous tissue (Datta et al., 2008).
Osteoblast fate Bone-forming osteoblasts become encased in the mineralized matrix surrounding them, turning into osteocytes that gradually stop synthesizing osteoid (i.e., the newly formed unmineralized organic bone matrix) (Datta et al., 2008;Rauner et al., 2012). Osteocytes are evenly distributed throughout the bone matrix which enables contact with osteoblasts and vasculature (Kamioka et al., 2001;Plotkin et al., 2002;Zhao et al., 2002;Plotkin et al., 2008). Osteocytes not only facilitate mechanosensation as described in Table 1, but also control bone structure (amount and quality) through mineralization inhibitors such as dentin matrix protein-1, fetuin-A, and Wnt inhibitor (Poole et al., 2005;Feng et al., 2006;Coen et al., 2009;Liu et al., 2009;Rauner et al., 2012). Although it was thought that osteocytes remain inactive until the next bone remodeling cycle (Mikuni-Takagaki, 1999;Kamioka et al., 2001;Zhao et al., 2002;Knothe-Tate et al., 2004;Datta et al., 2008), it is now accepted that osteocytes constantly remodel the surrounding extracellular matrix (Yee et al., 2019). Another fate of osteoblasts is to become bone lining cells, which cover the freshly formed endosteal bone surface thus forming a physical barrier to avoid the process of osteoclast adhesion and bone resorption.
These observations suggest that bone remodeling may not be solely defined by the balance/imbalance between MMPs and TIMPs. Rather, other molecules expressed and released in the settings of bone physiology and pathology such as RECK (Paiva and Granjeiro, 2014) and some acute phase reactants (alpha 2-macroglobulin, fibrinogen) may regulate/dysregulate MMP activity in inflammatory conditions thus perturbing the normal bone remodeling process (Cook et al., 2018;Sarker et al., 2019). A consequence implied by the latter notion is that MMPs, ADAMs and ADAMTS molecules may be released from bone or non-bone tissues to influence bone remodeling through autocrine and paracrine actions. In other words, MMPs likely circulate bound to non-classical inhibitors (such as acute phase reactants) being recruited to sites of active bone remodeling, where local substrates act as chemoattractants and local activators (other proteases, reactive oxygen species) activate them.
The aforementioned levels of regulation effectively dissociate MMP expression from MMP activity (e.g., since overexpression of endogenous MMP inhibitors would effectively reduce MMP activity). Current biochemical techniques for assessing MMP activity are non-reliable. However, as research requires a proxy, MMP expression is often used as a surrogate (albeit incorrectly) for MMP activity. There remains an urgent need for highly sensitive, specific, and robust methods for assessing the activity potential of individual MMPs such that therapeutic strategies can be designed to specifically reduce the activity of overactive MMPs (i.e., those whose activity levels are above baseline) or to increase the activity of underactive MMPs (i.e., those whose activity levels are below baseline).

Roles of MMPs Associated to Bone Development and Remodeling
The biochemical actions of MMPs are intimately linked to their cells of origin. Table 4 describes cell-specific roles of MMPs in physiological bone remodeling. Osteoclast-mediated bone resorption in calvaria and long bones requires normal enzymatic activity of MMPs and cysteine proteinases such as Maintain bone and cartilage health by their normal proteolytic activity. Everts et al., 1992;Meikle et al., 1992;Mattot et al., 1995;Apte et al., 1997;Johansson et al., 1997;Bord et al., 1998;Jimenez et al., 1999;Filanti et al., 2000 Control bone tissue remodeling at the levels of osteocyte viability and activities, osteoclast recruitment and function, bone matrix solubilization, coupling of bone resorption and formation, osteoblast recruitment and survival, cell-extracellular matrix interaction, and cell-cell interaction.
cathepsin K whose deficiency impairs bone remodeling (Everts et al., 1999;Delaisse et al., 2003). This is evidenced in osteoclasts from patients with pycnodysostosis (an osteopetrosis-like bone disease related to loss-of-function mutations in the cathepsin K gene) and osteoclasts from cathepsin K-deficient mice which are unable to efficiently digest organic bone matrix, resulting in large, mineral-free areas of bone matrix (Everts et al., 1998(Everts et al., , 2009). Cysteine proteinases synthesized and used by the different osteoclasts for bone matrix digestion (Everts et al., 2006) can degrade intramembranous bones as well as osteoclast-derived MMPs (Everts et al., 2009). Cysteine proteinases are secreted to act in the low pH environments formed by osteoclasts in the resorption sites, with MMPs degrading the rest of the bone matrix when the pH increases (Everts et al., 1998) as well as contributing to the digestion of fibrillar, non-mineralized collagen in Howship's lacunae abandoned by osteoclast cells (Everts et al., 2002). These complementary and overlapping contributions of the MMP and cysteine proteinase families make the process of bone tissue remodeling both complex and robust. The involvement of MMPs in bone remodeling has become clear with the aid of animal models such as MMP-deficient mice, which show a variety of bone abnormalities ( Table 5). Impaired bone tissue remodeling in Mmp2 −/− mice (Table 5, row 2) is characterized by a reduced number of osteoblasts and osteoclasts, disruption of the canicular network exacerbating osteocyte death, disruption of the MMP-2-osteopontin-bone sialoprotein axis, and promotion of osteolysis (Martignetti et al., 2001;Inoue et al., 2006;Mosig et al., 2007;Malaponte et al., 2016). MMP-9-deficient mice show alterations in cartilage-bone replacement during endochondral ossification (Vu et al., 1998) ( Table 5, row 3). This phenotype may be explained by an inefficient degradation of the cartilage matrix, which leads to a diminished bioavailability of extracellular matrix-derived vascular endothelial growth factor and consequently effects osteoclasts and endothelial cells movement into the cartilage (Ortega et al., 2010). Bone tissue modeling and remodeling processes are altered in MMP-13 deficient mice ( Table 5, row 4) (Inada et al., 2004;Stickens et al., 2004;Ortega et al., 2005). MMP-14 deficiency ( Table 5, row 5), which is associated with high lethality, results in the most drastic skeletal phenotype among MMP-deficient mice (Holmbeck et al., 1999;Zhou et al., 2000). Double gene-deficient mice lacking at least one MMP gene have been engineered and their bone phenotype have been studied. For instance, double-knockout mice lacking MMP-2 and uPARAP/Endo180 (endocytic receptor of collagen and collagen fragments for degradation in the lysosomes) show reduced bone mineral density, short long bones, and poor trabecular bone quality (Madsen et al., 2013). MMP-8 and MMP-13 double-deficient mice have abnormal growth plate as well as augmented metaphyseal trabecular bone mineral density (Inada et al., 2001(Inada et al., , 2002Stickens et al., 2004). Double knockout mice lacking MMP-9 and MMP-13 exhibit expanded growth plates, disorganized hypertrophic chondrocyte zone, increased number of end-differentiated hypertrophic cells, and

MMPs as Sheddases
Beyond the direct degradation of extracellular matrix substrates (e.g., collagen), MMP-mediated cleavage of substrates can lead to the release (shedding) into the extracellular matrix of soluble fragments of cell membrane-anchored receptor ligands. This extracellular event enables ligand-mediated activation of cognate receptors and elicits downstream intracellular signal transduction cascades which modify gene transcription and, ultimately, cell behavior. A prominent example pertinent to osteoblasts is the release of RANKL, which is the ligand of receptor activator of nuclear factor kappa B (RANK), by MMP-14. This MMP-14/RANKL/RANK/signal transduction axis regulates osteoblastogenesis and osteoclastogenesis, making MMP-14 crucial for normal bone formation (Bonfil et al., 2007;Thiolloy et al., 2009;Sabbota et al., 2010;Bonfil and Cher, 2011). The ligand shedding activity of

MMP Reported involvement References
MMP-1 Abundant in the diaphysis and metaphyses of long bones being upregulated in arthritis. Gack et al., 1995;Wu et al., 2008;Rose and Kooyman, 2016 MMP-2 Required for maintenance of bone mineral density and strength and in bone development (e.g., by affecting intramembranous and endochondral ossification); however, deregulated MMP-2 expression is observed in the settings of metabolic syndrome, osteoporosis, osteonecrosis of the jaws, ligamentum flavum degeneration in lumbar spinal canal stenosis, as well as in bone pre-metastatic niche formation. Duerr et al., 2004;Suh et al., 2004;Tester et al., 2004;Durie et al., 2005;Lynch, 2011;Fernandez-Patron et al., 2016;Rose and Kooyman, 2016;Sugimoto et al., 2018 MMP-3 Overexpressed in osteoarthritis (in cartilage and the synovium) and also acts on primary tumor growth. Okada et al., 1992;Tester et al., 2004;Lynch, 2011;Paiva and Granjeiro, 2017 MMP-8 Modulates human dentin and remodeling, but its deregulation may exacerbate periodontitis although it may be protective against inflammatory arthritis. Sulkala et al., 2007;Cox et al., 2010;Mauramo et al., 2018 MMP-9 Participates in chondrocyte biology; specific processes in which the enzyme is involved are apoptosis of hypertrophic chondrocytes present in utero, bone development (e.g., by being highly active to angiogenesis in the growth plate), strength and toughness of bone, as well as the regulation of gene pathways responsible for osteoclastogenesis). In turn, MMP-9 overexpression contributes to sinonasal osteitis, rheumatoid arthritis, and degenerated lumbar disk tissues. Vu et al., 1998;Liang et al., 2016;Mittal et al., 2016;Ahrens et al., 1996;Li et al., 2017 Osteoporotic bone (vs. normal bone) tissues express higher MMP-9 levels. Zhao et al., 1997 Involved in secondary (metastatic) breast cancer in the bone (e.g., by promoting angiogenesis, regulating VEGF bioavailability, contributing to bone remodeling) or prostate cancer (e.g., by influencing bone osteoblastic and osteoclastic activity). Bergers et al., 2000;Mannello et al., 2005;Pego et al., 2018 MMP-13 Required for bone development; it participates in the transition from cartilage to bone at the growth plates of long bones and in the remodeling of bone spicules. In turn, MMP-13-mediated degradation of articular cartilage exacerbates osteoarthritis. Inada et al., 2004;Stickens et al., 2004;Page-McCaw et al., 2007;Holmbeck et al., 1999;Mittal et al., 2016;Rose and Kooyman, 2016 In linking osteoarthritis to metabolic syndrome, the presence of adiponectin positively correlates with the presence of membrane-expressed PGE synthase and MMP-13. Francin et al., 2014 Overexpressed in congenital spondyloepiphyseal dysplasia which results in early development of osteoarthritis.

Rose and Kooyman, 2016
In addition to typical bone collagen matrix degradation, MMP-13 regulates bone resorption in periodontal disease through osteoclast differentiation (by inactivating galectin-3, an inhibitor of osteoclastogenesis) and osteoclast activation (by activating osteoclast-secreted pro-MMP-9 and favoring RANKL and TGF-β1 signaling). Nannuru et al., 2010;Pivetta et al., 2011;Cavalla et al., 2017 In breast cancer resulting from bone metastasis, MMP-13 deregulation may alter osteoblast morphology and bone resorption through differentiation of pre-osteoclasts, osteoclast activation, and osteolysis. Stickens et al., 2004;Page-McCaw et al., 2007;Shah et al., 2012 MMP-14 Contributes to bone development (endochondral and intramembranous ossification) and remodeling. Extracellular matrix remodeling by MMP-14 influences cell shape inducing the formation of a complex between MMP-14 and beta1-integrin, which activates the Rho/GTPase cascade leading to nuclear translocation of YAP and TAZthis series of signaling events is necessary for mesenchymal stem cells commitment during development. Palmitoylation (i.e., addition of 16-carbon palmitate to proteins) enables MMP-14 to anchor to cell membrane. This post-translational modification of MMP-14 has a major impact on bone development and bone tissue metabolism likely through influencing MMP-14 correct membrane localization and also decreasing the expression of osteocalcin and vascular endothelial growth factor in osteoblasts and chondrocytes. In turn, MMP-14 is critical for osteoclast resorption thus contributing to the pathogenesis of osteoporosis. Holmbeck et al., 1999Holmbeck et al., , 2003Zhou et al., 2000;Liao et al., 2004;Hienz et al., 2015;Tang et al., 2013;Paiva and Granjeiro, 2014;Song et al., 2014 Involved in bone cancer metastasis acting alongside MMP-1 and MMP-11. McGowan and Duffy, 2008;Rowe and Weiss, 2009;Paiva and Granjeiro, 2017 MMP-3 and MMP-9 Contribute to cartilage endplate degeneration. Zhang et al., 2018 MMP-2, MMP-9, and MMP-13 In experimental glucocorticoid-induced osteoporosis and osteocytic osteolysis, these three enzymes are upregulated in the trabecular bone of the metaphysis whereas MMP-2 and MMP-13 are expressed in the cortical bone diaphysis. Sun et al., 2016 VEGF, vascular endothelial growth factor; PGE, prostaglandin E synthase; RANKL, receptor activator of NF-kappa B ligand; TGF transforming growth factor; MMP, matrix metalloproteinase; YAP, Yes-association protein; TAZ, transcriptional coactivator with PDZ-binding motif.
MMPs influences the propensity to cancer metastasis and bone disease. For instance, MMP-14-mediated shedding of RANKL and downstream activation of RANK in the left supraclavicular lymph node cells of the prostate stimulates the non-receptor tyrosine kinase, SRC, to effectively increase the migration of prostate tumor cells which can metastasize  (Bord et al., 1999;Sobue et al., 2001;Huang et al., 2002;Geoffroy et al., 2004;Haeusler et al., 2005;Sahebjam et al., 2007;Shen et al., 2010;Miller et al., 2017).

MMP-Generated Neoepitopes
The proteolytic action of MMPs on extracellular matrix macromolecules can result in the exposure of neo-epitopes (i.e., unique bioactive MMP-generated fragments). Compared to healthy subject controls, patients with ankylosing spondylitis (which is a form of arthritis that causes inflammation of the vertebrae) show significantly higher levels of different neoepitopes such as C1M, C2M, C3M, C4M, C5M, C6M, and C7M from collagen type I, II, III, IV, V, VI, and VII Genovese and Karsdal, 2016). Some of these neoepitopes have been combined (e.g., C2M, C3M, and C6M) for diagnostic purposes . IPEN341-342FFGV is an MMP cleavage site which could be useful as diagnostic and prognostic makers for osteoarthritis (Bay-Jensen et al., 2011). Similarly, other MMP-generated neo-epitopes derived from collagen type II (e.g., C2C, C2M, C-terminal telopeptide of type II collagen (CTX-II), and TIINE) hold biomarker potential for osteoarthritis Qvist et al., 2010;Karsdal et al., 2011).

Over-Overexpression of MMPs
Over-expression of MMPs is frequently reported in arthritis (Burrage et al., 2006;Tokito and Jougasaki, 2016). Collagenolytic MMPs (such as  are expressed in the arthritic joint and likely participate in the degradation of cartilage type II collagen, while MMP-3, -7, and -9 can degrade aggrecan leading to joint destruction (Puliti et al., 2012;Tokito and Jougasaki, 2016). Such a pathological mechanism has been proposed for MMP-3 and MMP-13 in degenerative joint disease in the elderly (Neuhold et al., 2001;Troeberg and Nagase, 2012;Jackson et al., 2014;Pap and Korb-Pap, 2015). Other contributions to osteoarthritis from activities related to MMP-3 include MMP-3-mediated activation of MMP-1 and MMP-13 (Mancini and di Battista, 2006;Tokito and Jougasaki, 2016). In rheumatoid arthritis, MMP-14 is greatly expressed in fibroblast-like synoviocytes and macrophages, and it could be an effector to cartilage destruction (Pap et al., 2000;Sabeh et al., 2010). MMP-1 and MMP-3 likely participate in cartilage destruction in rheumatoid arthritis and osteoarthritis (Burrage et al., 2006;Fiedorczyk et al., 2006;Tokito and Jougasaki, 2016). As a result, MMP overexpression could be therapeutically targeted in arthritis (Tokito and Jougasaki, 2016). Whether reducing MMP expression (or activity) levels provides a clinical benefit is unclear. In experimental models, many synthetic MMP inhibitors have shown positive effects (Ishikawa et al., 2005). At the clinical level, however, all efforts with MMP inhibitors to block the damaging activity of MMPs in arthritis and other nonneoplastic conditions were regrettably unsuccessful (Burrage et al., 2006;Tokito and Jougasaki, 2016). Reasons for these failures include: (i) deficient clinical trial designs (Burrage et al., 2006), (ii) unwanted characteristics of MMP inhibitors (side effects including musculoskeletal pain, low oral bioavailability, short in vivo half-lives, and lack of selectivity [Iyer et al., 2012;Fields, 2015;Tokito and Jougasaki, 2016]), (iii) inability of MMP inhibitors to infiltrate the cartilage/bone/synovial interface (Burrage et al., 2006), (iv) neglect of the highly complex functions served by MMPs in physiological and disease states (Iyer et al., 2012;Li et al., 2013;Sawicki, 2013) and (v) broad tissue distribution and substrate promiscuity exhibited by MMPs and their substrates (Burrage et al., 2006;Tokito and Jougasaki, 2016). To date, there remains a need for highly selective MMP inhibitors and for better information on the disease-specific substrates, which could be therapeutically targeted as shown by recent studies with MMP-13 in osteoarthritis  as well as for more efficient and reliable techniques to sensitively measure condition-specific MMP activity potential (not just MMP expression levels).

MMP Gene Polymorphism
A nucleotide polymorphism, by which an additional guanine creates an ETS transcription factor binding site (5 -GGA-3 ) at position 1607 in the promoter sequence of the MMP-1 gene, has been related to bone mineral density (BMD) (Rutter et al., 1998). This polymorphism is associated with increased transcription of the MMP-1 gene and elevated MMP-1 activity. Among 819 postmenopausal Japanese women, BMD (e.g., D50, D100) for the distal radius had a lower value in women with the GG/GG genotype (47.9%) than in those with other (e.g., G/GG [41.9%], G/G [10.3%], G/G + G/GG [52.1%]) genotypes. A -1562C3 thymine polymorphism in the MMP-9 gene has been related to BMD in a populationbased study (1114 Japanese men and 1087 women). It seems that the T allele (e.g., in men with CT or TT genotypes) of MMP-9, which shows greater transcriptional activity than the C allele (e.g., in men with CC genotype), is linked to decreased bone mass, and has a predominant effect on BMD (Zhang et al., 1999;Yamada et al., 2004). A single nucleotide polymorphism rs17576 may be involved in the pathogenesis of lumbar disk herniation (Jing et al., 2018); while the G allele of rs17576 appears to correlate with more severe stages of disk degeneration.

MMP Deficiency and Insufficiency in Humans
Having discussed the roles of MMPs under physiological and pathological conditions, we will next discuss how their deficiency and insufficiency relates to bone metabolic abnormalities.
MMP-2 gene deficiency leads to a rare human skeletal disorder 1 , which was first reported in consanguineous Saudi Arabian families, and is characterized by severe bone alterations (Martignetti et al., 2001). Osteolytic and metabolic changes linked to MMP-2 deficiency affect tarsal, carpal, and phalangeal bones, cause severe arthropathy, osteoporosis, fibrous nodules, distinctive craniofacial defects such as exophthalmos, brachycephaly, and flattened nasal bridges and dwarfism (Al-Aqeel et al., 2000;Al-Mayouf et al., 2000;Al-Aqeel, 2005;Mosig et al., 2007;Page-McCaw et al., 2007;Castberg et al., 2013). This complex syndrome is currently categorized as a form of Torg syndrome and results from homoallelic mutations in the gene for MMP-2 located at 16q12-21 (Martignetti et al., 2001;Liang et al., 2016). A Tyr codon in the MMP-2 prodomain is replaced with the Y244X stop codon and an Arg is replaced with a His (R101H) in the cysteine-containing domain (PRCGNPD substituted by PHCGNPD). The R101H mutation is suggested to perturb coordination of Cys102 to the catalytic Zn(II) domain, consequently activating intracellular pro-MMP-2 and leading to its auto-degradation (Kennedy et al., 2005;Krane and Inada, 2008). A homoallelic missense mutation in the catalytic Zn(II) domain (E404K) has been revealed in Winchester syndrome (another variant of multicentric osteolysis) (Zankl et al., 2005). These rare Torg and Winchester arthritic syndromes together with others (such as multicentric osteolysis with nodulosis and arthropathy [known as MONA]) belong to a general family of hereditary autosomal dominant and recessive skeletal disorders with progressive bone loss and joint destruction (Al-Mayouf et al., 2000;Martignetti et al., 2001;Al-Aqeel, 2005;Zankl et al., 2005;Rouzier et al., 2006;Mosig et al., 2007;Tuysuz et al., 2009). Similar to MMP-2, a homozygous dominant mutation (Ser substituted by Phe [F56S]) in the pro-region domain of MMP-13 also results in a bone development disorder known as spondyloepimetaphyseal dysplasia-Missouri type (Kennedy et al., 2005) 2 . This disorder, which appears to spontaneously resolve by adolescence, is characterized by anomalous modeling of long bones, mild defects in epiphysis, moderate to severe changes in the metaphysis morphology, pear-shaped vertebrae, femoral and tibial bowing, genu varum deformities, and osteoarthritis. While the biochemical mechanisms linking MMP-13 to these bone abnormalities remain unclear, the phenotype of MMP-13 deficiency could be due to a late exit of chondrocyte cells from the growth plate (Kennedy et al., 2005).
MMP-14 is widely considered one of the physiological activators of MMP-2 as it converts pro-MMP-2 into mature MMP-2 at the cell surface (Fernandez-Patron et al., 2016). An MMP-14 homoallelic mutation (T > R replacement in the signal peptide domain) destabilizes the interaction (e.g., recognition and binding) of the MMP-14 signal peptide with the signal recognition particle complex, thus affecting MMP-14 targeting to the plasma membrane (Evans et al., 2012). This MMP-14 homoallelic mutation causes an apparent deficiency of biochemically active MMP-14 at the cell membrane which impairs pro-MMP-2 activation and causes a condition of MMP-2 activity deficiency with Winchester syndrome (Evans et al., 2012) 3 .
A missense homozygous mutation (g.16250T > A, which replaces His226 of the Zn(II) catalytic domain with Gln [p.H226Q]), in the MMP20 gene disrupts the metal-binding site and prevents MMP-20 proteolytic activity regarding enamel matrix proteins (Ozdemir et al., 2005) 4 . This mutation may lead to autosomal-recessive hypomaturation amelogenesis imperfecta, a group of inherited heterogeneous diseases that alter enamel development (amount, composition, structure) in humans (Kim J.W. et al., 2005). Another mutation in the intron 6 splice acceptor (g.30561A > T) that causes this disease is specifically characterized by pigmented teeth with a mottled and rough surface (Kim J.W. et al., 2005).
Partial loss of MMP activity or impaired MMP secretion can lead to MMP activity insufficiency. A pervasive cause 3 https://www.omim.org/entry/600754 4 https://www.omim.org/entry/604629 of MMP insufficiency can be medications with such MMP inhibitory actions including: (i) Statins (200 million prescriptions in the United States/year; 14 million prescriptions for lovastatin alone in 2014) 5 which can cause myositis and rhabdomyolysis (Luan et al., 2003;Thompson et al., 2003). (ii) Doxycycline (7 million prescriptions in 2014) 5 with side-effects including joint inflammation in humans and cardiac inflammation in mice (Berry et al., 2015). (iii) Therapeutic antibodies against MMPs and MMP inhibitor drugs for treating patients with rheumatoid arthritis, severely active Crohn's disease, and cystic fibrosis 6 . If these antibodies reduce MMP activity below baseline levels, they would cause MMP insufficiency with unpredictable consequences. Pharmacological MMP-inhibitors in Phase 3 clinical trials conducted during 1997 and 1998 in patients with advanced cancers led to an as of yet poorly understood, very severe inflammatory musculoskeletal syndrome (Zucker et al., 2000;Coussens et al., 2002). Another common cause of MMP insufficiency could be the pathological elevation of  (Datta et al., 2008;Fernandez-Patron et al., 2016;Paiva and Granjeiro, 2017;Cook et al., 2018). Other pathologies related to inactive/underactive MMPs are excessive inflammation, cardiovascular disorders, and metabolic dysregulation. MMP underactivity could also result from undesired side effects of common medications with MMP inhibitory actions (e.g., statins) (Cook et al., 2018). MSCs, mesenchymal stem cells; GFs, growth factors; RUNX2, runt-related transcription factor 2; RANKL, receptor activator of NF-kappa B ligand. endogenous MMP inhibitors (e.g., tissue inhibitors of MMPs, α-2-macroglobulin, RECK) (Mott et al., 2000;Oh et al., 2001;Nagase et al., 2006;Klein and Bischoff, 2011). In addition, there is fibrinogen, an acute phase reactant in arthritis, which our laboratory discovered recently to inhibit MMP-2 in a cohort of rheumatoid arthritis patients (Sarker et al., 2019).

SUMMARY
In summary, bone lining cells, osteocytes, osteoclasts, reversal cells, and osteoblasts are responsible for constant bone tissue remodeling (Figure 2). The activation of this multicellular unit and the intense communication between the bone cells is tightly regulated by mechanical stimuli, apoptosis, as well as systemic and local factors such as hormones and cytokines including RANKL, CSF-M, IL-3, and IL-6. Proteases of the MMP and cysteine proteinase families converge in the modulation of bone remodeling. Whereas proteolytic activity has long been thought to be required for the degradation of bone tissue in osteoarthritis and osteoporosis, inactivating mutations in MMP genes can also lead to bone pathology including osteolysis and metabolic abnormalities such as delayed growth. Thus, there remains a need to rethink the role played by proteases in bone physiology and pathology. More specific information related to bone remodeling and presumed pathways by which proteases, in particular MMPs, contribute to bone tissue remodeling in health and disease is provided in previous excellent reviews (Kini and Nandeesh, 2012;Rauner et al., 2012;Hienz et al., 2015;Liang et al., 2016;Mittal et al., 2016;Franco et al., 2017;Paiva and Granjeiro, 2017;Tauro and Lynch, 2018;Plotkin and Bruzzaniti, 2019).

AUTHOR CONTRIBUTIONS
EH and CF-P worked together on the conception, design, edition, revision, and approval of review manuscript.