Integrins are defined as cell adhesion structures, which are important in development and pathological processes. Integrins play critical roles in signaling, migration and survival of different cells (recently reviewed in depth by Barczyk et al. [5] and Iwamoto and Calderwood [6]). The signaling by these receptors is considered bi-directional, involving outside-in and inside-out signaling [5]. Integrins function as heterodimers, which in humans include one of 18 distinct α subunits and one of eight distinct β subunits, and are type I transmembrane glycoproteins with large extracellular and short cytoplasmic domains [6]. Four different integrin heterodimers (α₁β₁, α₂β₁, α₁₀β₁, and α₁₁β₁) have been demonstrated to bind collagen. In particular, α₁β₁ and α₂β₁ integrins have been most extensively studied. These integrins bind to both collagen types I and IV, however their affinities differ: α₁β₁ has a higher affinity for collagen type IV, while α₂β₁ preferentially binds to collagen type I [5, 7]. Integrin α₂β₁ has been reported to be one of the main collagen binding integrins present in bone and is critical for bone resorbing cells, osteoclasts. In these cells, α₂β₁ integrins affect the attachment of the cell to the bone surface and help form a sealing zone around the area to be resorbed, allowing formation of the localized highly acidic environment necessary for bone degradation [7–9].

The two discoidin domain receptors, DDR1 and DDR2 are receptor tyrosine kinases activated specifically by fibrillar collagens I–III and V, but not by individual α-chains, denatured collagen, de-glycosylated, or degraded collagens [10]. A distinct characteristic of these receptors is their slow and sustained activation upon stimulation. Binding of the discoidin receptors to triple helical collagen leads to tyrosine autophosphorylation with unique activation kinetics, which is followed by receptor internalization [11, 12]. Imbalance or dysregulation of DDR1 has been implicated in the development of diseases such as fibrosis, atherosclerosis, arthritis, and cancer [10]. DDR1-null mice are predisposed to osteoarthritis and temporomandibular joint disorder [13], but are protected from atherosclerosis and smooth muscle mineralization [14]. DDR1 function can be controlled by ADAM10-mediated ectodomain shedding [15]. DDR2 is involved in pathological scarring processes such as wound healing, arthritis, and cancer [16, 17]. DDR2-deficient mice exhibit dwarfism due to reduced proliferation of chondrocytes [18].

The leukocyte receptor complex (LRC) consists of a large group of cell surface receptors, essential for a diverse number of immune functions, including antiviral immunity, autoimmunity, and response to grafts [19]. A typical characteristic of these receptors is the occurrence of pairs of antipathetic receptors, which bind to the same ligands but generate opposing signaling responses [20]. The majority of LRC receptors are primarily expressed by immune cells and play diverse roles in modulating their activity [19, 20]. The stimulatory receptors have short cytoplasmic tails and generate positive signals through immunoreceptor tyrosine-based activation motifs (ITAM) present on the required adapter proteins, FcRγ, DAP10, and DAP12. The inhibitory receptors are characterized by long cytoplasmic tails, which contain immunoreceptor tyrosine-based inhibitory motifs (ITIM) [19, 20]. Of interest, collagen has been recently demonstrated to act as a ligand for a number of stimulatory and inhibitory receptors in this family, including osteoclast associated receptor (OSCAR), GPVI, and LAIR-1 [21–23]. The structural basis for collagen recognition by the immune receptors has been investigated in a number of studies [24–27], however some controversy regarding the alignment of collagen-recognition sites among different receptors exists.

OSCAR and GPVI are stimulatory receptors that are activated by collagen. OSCAR is particularly important for osteoclast differentiation, as it acts as a critical co-stimulatory receptor for osteoclast formation and function [23, 26–28]. GPVI is mainly found in platelets and binds to collagen during the process of blood coagulation [21]. The binding of collagen to OSCAR or GPVI results in the recruitment of ITAM-containing FcRγ chains. For OSCAR, the main downstream effect of activation is the initiation of calcium signaling, which is critically important for activation of a key osteoclastogenic transcription factor, nuclear factor of activated T-cells (NFAT) c1. Activation of GPVI leads to the binding of Syk to the FcR-γ chain, which causes an activation of Syk proteins and tyrosine phosphorylation. At the same time, phospholipase C γ2 (PLCγ2) also becomes activated [29, 30].

The inhibitory receptor LAIR-1 was also shown to be activated by collagen [31]. Another inhibitory receptor G6b-B may also act as a collagen-binding receptor, although the existing evidence is weaker [32]. Both LAIR-1 and G6b-B were first identified to be expressed on megakaryocytes and platelets and to negatively regulate their function [22, 33]. LAIR-1 was also found to be present during osteoclastogenesis and to inhibit this process [34]. LAIR-1 is activated by triple helical collagen, specifically when it encounters the triplet (GPO)₁₀ (glycine-proline-hydroxyproline)₁₀, also known as “collagen related peptide” [31]. LAIR-1 contains two ITIMs, which upon phosphorylation recruit SHP-1 and SHP-2 phosphatases. These phosphatases directly dephosphorylate Syk, Zap70, and PLCγ, preventing ITAM-mediated stimulation of protein kinases and calcium signaling [22, 31]. G6b-B was suggested to be activated by collagen fragments, such as the collagen-related peptide, likely relevant to the microenvironment of damaged epithelium [32, 33]. G6b-B contains one ITIM as well as a newly described immunoreceptor tyrosine-based switch motif (ITSM) [33, 35]. In contrast to ITIM, which generally signals through activation of phosphatases, ITSM interferes with ITAM-mediated signaling by using adaptor molecules. An important characteristic of this motif is the ability to switch between stimulatory and inhibitory signals and to bind SHP1, SHP2, SHIP, and p85 [36]. In platelets, G6b-B interferes with positive signaling induced by collagen binding to GPVI [37].

The urokinase plasminogen activator receptor-associated protein (uPARAP/Endo180), a member of the mannose receptor family of type I transmembrane glycoproteins, is a multi-domain transmembrane glycoprotein. Characteristically this family of proteins includes an N-terminal, cysteine-rich/ricin B like domain, a fibronectin type II domain, and a series of 8–10 C-type lectin-like domains. This mesenchymal cell surface receptor has an important function in collagen internalization [38–40]. In addition, uPARAP/Endo180 was shown to aid in the initial adhesion of fibroblasts to collagen and to accelerate the migration of these cells on a fibrillar collagen matrix [38, 39, 41, 42]. In bone, this receptor is highly expressed on osteoblasts and osteocytes at sites of endochondral and intramembranous ossification during development [38].

Thus, multiple receptor families can bind collagen and induce a variety of cellular effects. Although the repertoire of cellular responses affected by collagen receptors appears to be similar, with adhesion, migration and survival being prominent on the list, it is interesting that the receptors can bind to different forms of collagen, including large triple helical fragments, matrix-incorporated collagen fibrils, and small collagen fragments. Next, we will consider the process of collagen turnover and identify the physiological stages during which different forms of collagen can act as effectors of receptor-mediated signaling.

Two classes of receptors, including those of LRC and of mannose-receptor family, have been shown respond to triple-helical collagen which is not necessarily incorporated in the matrix (Figure 1). In particular, the uPARAP/Endo180 receptor was found on early osteoblast precursors as well as actively matrix-producing osteoblasts [72, 73]. While this receptor has been implicated in osteoblast recruitment to the remodeling sites [73], this function can likely be attributed only to early precursors, but not to mature cells. It is thus possible that additional regulation may be exerted by this receptor present on mature osteoblasts. From the LRC family, OSCAR and LAIR-1 have been shown to be expressed by osteoclasts [28, 34]. It is possible that LAIR-1-mediated inhibition of osteoclastogenesis contributes to the prevention of premature activation of resorption at the sites of freshly laid down osteoid. It is also interesting to speculate that formation of the two fragments, NTP and CTP, during procollagen processing may generate soluble and thus longer-reaching signals for the receptors activated by smaller collagen fragments. Abnormal signaling though collagen receptors can also potentially contribute to the pathophysiology of OI, as it commonly results in abnormal collagen modification and thus would significantly alter receptor-ligand interactions.

The majority of the collagen receptors are assumed to be activated by collagen present in mature matrix. The evidence for a physiological significance of such interactions is strongest for members of the integrin receptor family, which are well-known for their substrate-recognition roles [81]. Positive responses, such as support of cell adhesion, survival, migration, and proliferation result from collagen interacting with integrin receptors α1β1 and α2β1 [5], receptor tyrosine kinases DDR1 and DDR2 [12], LRC member OSCAR [23], and mannose family receptor uPARAP [38]. Inhibitory cellular effects have been observed upon stimulation of LAIR-1 with collagen [22, 32]. It is particularly interesting, that osteoclasts express both the stimulatory collagen receptor OSCAR [28] and an inhibitory receptor LAIR-1 [34]. However, if we assume that OSCAR and LAIR-1 are expressed at different stages of osteoclastogenesis, we can attempt to reconcile how contradictory collagen signals can be perceived by these cells. We can speculate that LAIR-1 is present on early precursors, and that its role is to prohibit osteoclast differentiation on immature matrix, while OSCAR is expressed later during osteoclastogenesis and is engaged by the matrix-incorporated collagen to support osteoclast formation on the correct substrate. While extensive studies on the role of OSCAR during osteoclast formation have been performed [23, 26, 28], much less is known about the temporal and spatial aspects of regulation of osteoclastogenesis by LAIR-1. In diseases associated with abnormal collagen synthesis, such as OI, as well as abnormal mineralization, for example osteomalacia due to calcium and phosphate deficiency, the structure of the mature tissue matrix is altered, which may potentially result in changes in receptor-ligand binding for collagen receptors, contributing to the pathophysiology of these disorders.

Cathepsin K, a member of the C1 peptidase family, is expressed by osteoclasts at the cell surface adjacent to the bone and was shown to be critical for matrix breakdown in the resorption compartment during bone remodeling [86, 87]. The protease works optimally at low pH and has been shown to degrade type I collagen [85]. Both the α₁ and α₂ collagen type I chains are cleaved by cathepsin K [88]. The protein cleavage site of papain-like cysteine proteases, including cathepsin K, is determined by the amino acids occupying the two positions before (N-terminal to) the cleavage site and one or two positions after the cleavage site. Generally, a hydrophobic side chain such as valine, leucine, or proline is found in the second residue before the cleavage site, whereas the amino acid directly before the cleavage site is usually glutamic acid, alanine, or glycine. Finally, in the position after the cleavage site, the most common amino acids are glycine, glutamic acid, and/or isoleucine [87, 89]. Overall, it has been found that cathepsin K performs a cleavage after helical cross-linking residues, which in collagen type I is most likely to occur after glycine due to high prevalence of GXY repeats [90]. Cathepsin K was shown to cleave substrates with proline in the second amino acid position after the cleavage site [91], which is critical for its ability to target collagen [87, 92].

Since cathepsin K is the only papain-like cysteine protease capable of cleaving triple helical collagen, it is of significant interest as a pharmaceutical target [90, 93]. Structural analysis revealed that for cathepsin K to demonstrate its collagenase activity, a dimer has to form an oligomeric complex with a glycosaminoglycan and dock onto a collagen molecule with its central grove [92, 94, 95]. The presence of glycosaminoglycans allows access to the triple helix, leading to the cutting of the fibril into smaller sub-fibrils and a simultaneous release of glycosaminoglycans. This process eventually results in a progressive unfolding of the fibrils, making them less stable and therefore accessible for further degradation [96]. Cathepsins that do not exhibit collagenase activity, such as cathepsins L, V, S, and B have only a limited effect on the fibril structures [96]. In addition to bone, cathepsin K is also expressed in hematopoietic, epithelial, and fibroblast cells, and was shown to play a role in arthritis, obesity, schizophrenia, bone metastases, and various other pathological conditions [89].

In humans, mutations in the cathepsin K gene were shown to underlie the skeletal disorder pycnodysostosis, which is characterized by osteopetrosis, bone fragility, short stature, acrosteolysis of the distal phalanges, delayed cranial suture closure, clavicular dysplasia, and dental abnormalities [97, 98]. In animal models, cathepsin K deficiency results in an osteopetrotic bone phenotype [99–101] and has been shown to predispose mice to lung fibrosis [102], abnormal airway morphologies [103] and deficiencies in learning and memory aptitudes [104]. Even though the osteoclast numbers were not changed in cathepsin K deficient mice, large areas of demineralized bone matrix underlying the ruffled borders of osteoclasts were frequently found, suggesting that only the degradation of organic bone matrix is impaired in these animals [99, 100]. Once cathepsin K cleaves the triple helix, it unwinds and becomes available for degradation by any protease with gelatinolytic activity. Interestingly in cathepsin K knockout mice, osteoclastic resorption, though severely impaired, was still present [99] due to involvement of other proteases with collagenase activity, such as the matrix metalloproteinases [101].

The matrix metalloproteinases (MMP), peptidase family M10, is composed of zinc-dependent endopeptidases, also known as the metzincin superfamily. The main task of MMPs is to degrade extracellular matrix proteins, as well as a number of bioactive molecules [87, 105, 106]. Within the MMP family of endopeptidases, MMP-1, MMP-8, MMP-13, and MMP-14 exhibit collagenase activity and are specifically capable of degrading triple-helical fibrillar collagens [106, 107]. MMP-1 and MMP-13 are produced by osteoclasts and play important roles in bone matrix degradation. After the triple helix is separated, MMP-2 and MMP-9 can further cleave collagen chains working as gelatinases. The MMP family members with collagenase activity require at least three structural components to successfully achieve their function: (i) a hemopexin-like C-terminal domain, (ii) a linker or hinge region between the catalytic and hemopexin domain, and (iii) a specific peptide loop in the catalytic domain [108, 109]. Collagen type I is cleaved by MMP-1, -8, and -13 at a characteristic site located between Gly₇₇₅/Ile₇₇₆, three quarters of the distance from the N-terminus, leading to the formation of two fragments—a larger fragment of ¾ and a smaller ¼ fragment [107, 110, 111].

MMP-1, also known as interstitial collagenase or fibroblast collagenase, is involved in the breakdown of extracellular matrix in normal physiological processes, such as embryonic development, reproduction, and tissue remodeling, as well as in disease processes, such as arthritis and cancer metastasis [106, 112, 113]. Osteoblasts have been demonstrated to produce MMP-1, which may affect their differentiation [114]. In addition, MMP-1 was shown to be upregulated in response to mechanical loading [115]. However, it is not clear if MMP-1 activity toward collagen type I is important for osteoblasts. While MMP-1 is not generally found to be expressed by osteoclasts, there is evidence that it may be involved in bone resorption in pathological conditions [113, 116, 117].

MMP-13 is involved in the degradation of extracellular matrix for tumor invasion and metastasis [84] and is a critical collagenase MMP for osteoclastic bone resorption [86]. Mutations in MMP-13 cause metaphyseal anadysplasia 1, which includes the Missouri type of spondyloepimetaphyseal dysplasia, a spectrum of diseases characterized by defective growth and severe skeletal changes that resolve spontaneously with age [118, 119]. This protease is more aggressive to collagen type II than to type I, therefore the effects are more prominent in the joints [120, 121]. In animal models, overexpression of MMP-13 produces arthritis with cartilage erosion [122]. Knockout of MMP-13 shows temporary anomalies in cartilage resorption in long bone growth and fracture healing. While these animals have a normal lifespan and sufficient fertility, microscopic analyses of the skeletal system verified profound defects in the growth plate cartilage with a clear intensification in the hypertrophic chondrocyte zone and a delay in primary ossification [121, 123].

The main gelatinase expressed and released by osteoclasts is MMP-9 [124]. MMP-9 participates in dissolution of bone collagens working inside the sealing zone located underneath the osteoclast, in concert with collagenases MMP-13 and cathepsin K [125]. MMP-9-deficient animals exhibit a skeletal development phenotype, and chondrocyte apoptosis [126]. In addition, vascularization and ossification of cartilage is significantly delayed [126]. The role of MMP-9 in bone resorption is not clear, however it was demonstrated that in the absence of MMP-9 osteoclastic recruitment is delayed and that MMP-9 is required for osteoclast invasion into the discontinuously mineralized hypertrophic cartilage [127]. Mutations in MMP-9 result in metaphyseal anadysplasia 2, which is phenotypically indistinguishable from metaphyseal anadysplasia 1 due to mutations in MMP-13 [119]. A dominant, more severe phenotype of metaphyseal anadysplasia was found to be associated with deactivation of both MMP-13 and MMP-9 [119].

Physiological degradation of collagen can result in production of shorter triple-helical proteins, along with single strand fragments of different molecular weight [86, 128]. Such fragments can potentially activate the receptors of LRC, GPVI, LAIR-1, and G6b-B, as well as uPARAP/Endo180. Since LAIR-1 is expressed by osteoclasts [34] and uPARAP/Endo180 by osteoblasts [129], it is possible that such signals mediate temporal and spatial coordination of collagen matrix formation and degradation. In addition, formation of circulating fragments, such as those used as biomarkers of bone resorption [130], raises a possibility of long-distance signals generated during collagen degradation. In pycnodysostosis and metaphyseal anadysplasia, mutations in cathepsin K and MMP-9 and MMP-13 result in altered collagen degradation, thus producing different collagen fragments, potentially interfering with collagen receptor signaling.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.