Proteoglycans and more – from molecules to biology

Abstract

In this article the organization and functional details of the extracellular matrix, with particular focus on cartilage, are described. All tissues contain a set of molecules that are arranged to contribute structural elements. Examples are fibril-forming collagens forming major fibrillar networks in most tissues. The assembly process is regulated by a number of proteins (thrombospondins, LRR-proteins, matrilins and other collagens) that can bind to the collagen molecule and in many cases remain bound to the formed fibre providing additional stability and enhancing networking to other structural networks. One such network is formed by collagen VI molecules assembled to beaded filaments in the matrix catalysed by interactions with small proteoglycans of the LRR-family, which remain bound to the filament providing for interactions via a linker of a matrilin to other matrix constituents like collagen fibres and the large proteoglycans, e.g. aggrecan in cartilage. Aggrecan is contributing an extreme anionic charge density to the extracellular matrix, which by osmotic effects leads to water retention and strive to swelling, resisted by the tensile properties of the collagen fibres. Aggrecan is bound via one end to hyaluronan, including such molecules retained at the cell surface, to form very large molecular entities that interact with other constituents of the matrix, e.g. fibulins that can form their own network. Other important interactions are those with cell surface receptors such as integrins, heparan sulphfate proteoglycans, hyaluronan receptors and others. Many of the molecules with an ability to interact with these receptors can also bind to molecules in the matrix and provide a bridge from the matrix to the cell and induce various responses. In pathology, there is an imbalance in matrix turnover with often excessive proteolytic breakdown. This results in the formation of protein fragments, where cleavage provides information on the active enzyme. Those fragments released can be specifically detected employing antibodies specific to the cleavage site and used to diagnose and monitor e.g. joint disease at early stages.

Introduction

The extracellular matrix of most tissues contains a set of related molecules interacting to form networks illustrated in Figure 1. Variability is to a large extent an effect of different assembly, partly regulated by a limited number of molecules more unique to a specific tissue. The major protein in any extracellular matrix is a fibril forming collagen, i.e. collagen type I in most tissues and collagen type II in cartilage related tissues. These collagens form fibrillar structures and the process is tuned by other fibrillar collagens, e.g. collagen type V with type I and collagen type XI with type II, present in amounts of less than 5% in the fibre (Kypreos et al. 2000; Eyre 2002; Wenstrup et al. 2004). Fibril formation is further regulated via a number of other extracellular matrix proteins. Examples are particularly members of the small leucine rich repeat protein family (SLRPs) exemplified by decorin (Danielson et al. 1997), biglycan (Schönherr et al. 1995), fibromodulin (Hedbom & Heinegård 1993) and lumican (Chakravarti et al. 1998) as well as the thrombospondins (Halasz et al. 2007), the matrilins (Wagener et al. 2005) and many others. Many of these molecules remain bound at the surface of the completed fibril thereby providing for interactions with other components of the matrix. One example is collagen type IX, which is quite specific for cartilage. It actually occurs covalently, bound to the main collagens of the fibres (Eyre et al. 2004), such that the major non-triple helical NC-4 domain plus its neighbouring col3 domain protrude out and allow for interactions with other matrix constituents (Vaughan et al. 1988). Thus the regulation of its assembly and the properties of the collagen network are regulated by similar mechanisms, but involving many proteins that can be expressed at different times and at different relative amount. Presently we have limited information on the fine details of this regulation.

Figure 1.

Schematic illustration of molecular constituents in cartilage and their arrangement into large multi-molecular assemblies. The different compositions and organizations at the cell surface with a number of receptors interacting with specific matrix molecules, at the interterritorial matrix closer to the cells and the interterritorial matrix at a distance are indicated.

There is an additional network of collagen type VI, particularly in weight bearing tissues such as cartilage carrying load (Ayad et al. 1989). In cartilage this network is localized close to the cells in the so-called territorial matrix. Also the assembly of this network is regulated by SLRPs (Wiberg et al. 2002) and these are retained in this beaded filament network to provide interactions with other elements in the tissue. The major non-collagenous component in tissues such as cartilage is the proteoglycans, also abundant in other pressure loaded tissues like the aortic wall. In cartilage some 5% of the wet weight is represented by aggrecan which is assembled to large aggregates by interactions with the polysaccharide hyaluronan (Hardingham & Muir 1972).

Proteoglycans

Studies of proteoglycans accelerated in the 1950s. At this time, Helen Muir demonstrated that the linkage of the glycosaminoglycan chondroitin sulfate was to serine residues in the protein core (Muir & Jacobs 1967). Martin Matthews studied distribution of the glycosaminoglycan chains along the core protein and coined the term doublets of chondroitin sulfate chains where he implied that the these were linked some 10 amino acids apart and this doublet was separated by some 50 amino acids (Mathews & Lozaityte 1958). Partridge had data to indicate that there were two different types of glycosaminoglycans bound to the same protein core (Partridge et al. 1961). Although this earlier work provided the basis for subsequent studies, only the data presented on the linkage to serine and the suggestion that there are two types of glycosaminoglycan chains in the same proteoglycan molecule have survived time. Later, work by Karl Meyer (Anderson et al. 1963) and his coworkers on one hand and Lindahl & Roden (1966) on the other hand non-ambiguously demonstrated a covalent O-glycosidic linkage between serine on the protein core and the reducing end at the linkage sequence of xylose-galactose-galactose to the chondroitin sulfate chain proper of the some 50 times repeated disaccharide of glucuronic acid-N-acetyl galactosamine with a 4 or 6 O-sulfate to form the chondroitin sulfate chain.

A very important step forward in the study of proteoglycans and extracellular matrix came with their work of Hascall and Sajdera, who introduced a novel extraction procedure based on denaturing solvents such as 4 M guanidine-HCl to extract most of the molecules except those cross-linked collagens. Their procedure was further combined with the purification of the large proteoglycan, aggrecan, using CsCl-density gradient centrifugation (Sajdera & Hascall 1969). They then applied their approach to demonstrate that the proteoglycans formed a specific aggregate with a second component. It was later demonstrated by Hardingham and Muir that this component was hyaluronan (Hardingham & Muir 1972) which together with the link protein formed the large aggregates (Hascall & Sajdera 1969). The binding to a minimal decasaccharide sequence of hyaluronan could be shown to be mediated by a globular domain (G1) at one end of the molecule and stabilized by the second protein, the link protein binding both hyaluronan and the G1 domain of aggrecan (Heinegård & Hascall 1974). When the complete sequence of aggrecan was unravelled it became clear that the link protein and the G1 domain were very homologous (Doege et al. 1987). The second G2 domain separated from the G1 by only a short stretch of amino acids also shares a homologous sequence with a G1 domain, but lacks ability to bind to hyaluronan. To date no function has been ascribed to this domain. Subsequently, using rotatory shadowing and electron microscopy on one hand and molecular cloning on the other hand a globular domain (G3) at the other C-terminal end of the molecule was demonstrated (Wiedemann et al. 1984; Halberg et al. 1988). This has homology with the C-type lectin but to date no distinct carbohydrate binding has been identified. It has more recently been shown that aggrecan via this domain can interact with certain matrix proteins containing EGF-repeats. These include the fibrillins, the fibulins as well as tenascins (Day et al. 2004). These molecules in themselves can form higher order networks. Therefore aggrecan can be complexed in the aggregates via its N-terminal domain on one hand and interact with other assemblies of macromolecules via their C-terminal.

Interestingly, a point mutation in the lectin domain which has proven to be ablating the interaction with fibulins, leads to chondritis dissicans with severe joint destruction at young age (Eva-Lena Stattin, Fredrik Wiklund, Karin Lindblom, et. al. Manuscript).

The glycosaminoglycan chains of the core protein are some 100 chondroitin sulfate chain distributed along the CS1 and CS2 domains (Heinegård et al. 1985; Doege et al. 1991). Interestingly, using an antibody to the specific epitope of some chondroitin sulfate chains particularly prominent in the growing individual (the 846-epitope) it could be shown that the chains on the core protein is are heterogeneous and differs in different diseases such as in rheumatoid arthritis where there are several antibody binding sites in the most N-terminal part of the CS1 domain, while in reactive arthritis, there are one single antibody binding to the very C-terminal part of the CS2 domain (Månsson, Mörgelin, Poole, Saxne and Heinegård, unpublished work). In other studies, Bayliss and coworkers demonstrated that the aggrecan character, such as the ratio of 4 to 6 sulfate changes chains considerably over age (Bayliss et al. 1999). The other type of glycosaminoglycan, keratan sulfate, is made up of repeat disaccharides of sulfated N-acetyl-glucosamine and galactose linked via and an N-acetyl-galactosamine residue to serine and threonine in the protein backbone (Bray et al. 1967). The short but quite variable keratan sulfate chain contains 22–30 disaccharide units and most of the more than 30 chains are located to a hexapeptide repeat domain with a higher proline content just N-terminal of the CS domains (Antonsson et al. 1989). The number of such repeat varies from some 20 in the bovine to four in the rat. Interestingly the rat does not produce keratan sulfate and all the potential sites are instead substituted by a short O-glycosidically linked oligosaccharides corresponding to the linkage domain (Lohmander et al. 1996). Indeed in the other species such as the bovine, during foetal life the aggrecan contains only the oligosaccharides and no keratan sulfate, while the number of these glycosaminoglycan chains increases with age (Inerot & Heinegård 1983).

The major function of the aggrecan molecule is to provide an extremely high fixed charge density, central to creating the high osmotic environment necessary to retain water in the tissue. This is mainly contributed by the chondroitin sulfate chains. Although there is quite a variability of the fine structure of the chondroitin sulfate chains, there is little knowledge of specific functions or interactions. One exception appears to be the somewhat oversulfated chondroitin sulfate chains in perlecan, a proteoglycan found in cartilage (French et al. 1999) but typically associated with basement membranes. These CS-chains, like CS-E, have the ability to accelerate the assembly of collagen fibres from collagen molecules (Kvist et al. 2006). The functional consequences of the change from a high 4-sulfate to a high 6-sulfate over age has not been unravelled.

The keratan sulfate rich region isolated from aggrecan by trypsin digestion has been shown to bind with high affinity for collagen (Hedlund et al. 1999). Interestingly, using immunogold technology it was shown that this particular domain is colocalized in the tissue with the so-called gap regions of the collagen fibres particularly in the pericellular and territorial matrix of cartilage.

One of the first interactions demonstrated for aggrecan is that to the cartilage specific protein CMP (Paulsson & Heinegård 1979), later renamed matrilin-1 when the other three family members were discovered (Wagener et al. 2005). This interaction even appears to result in a covalent linkage, however not yet defined (Hauser & Paulsson 1994).

Aggrecan is quite specific for cartilage, but can also be expressed in other tissues particularly those under high compressive load. Examples are the tendon (Koob et al. 1992) and also the vascular wall, particularly in arteriosclerosis (Ström et al. 2004) where the tissue probably experiences very high shear stress.

Non-collagenous matrix proteins

Matrilins

Cartilage contains a number of non-collagenous proteins that provide important functions to the tissue for tissue assembly and for maintaining tissue properties and function. There are differences between the compositions of cartilaginous tissues. One example is the matrilins, where matrilin-1 (originally CMP) is abundant in tracheal cartilage (Paulsson et al. 1984), while not in articular cartilage or other weight bearing cartilaginous tissues like the intervertebral disc (Paulsson & Heinegård 1982). On the other hand matrilin-3 is found in articular cartilage (Klatt et al. 2000). The matrilins are multimeric proteins with three or four subunits. They contain as the key functional domain one or two von Willebrand factor motifs, which typically bind to various other proteins including triple helical collagen (for references see Wagener et al. 2005). An important high affinity interaction is that with biglycan or decorin (Wiberg et al. 2003) that in turn bind to collagen type VI beaded filaments (Wiberg et al. 2001). In cartilage, the matrilin appear to be interspersed between collagen type VI via decorin/biglycan on one side and collagen type II fibres on the other hand, but is also found binding procollagen molecules as well as aggrecan (Wiberg et al. 2003). Notably such complexes have been isolated from the cartilaginous tissue of the rat chondrosarcoma. The SLRP in the complex with matrilin also binds to the N-terminal globular domain of the collagen type VI. Thus, the combination of decorin or biglycan with a matrilin provides a linker module connecting the collagen fibre network to other structures. Other roles of the matrilins include binding to the triple helical collagen domains of collagen type IX as well as binding to COMP (acronym for ‘cartilage oligomeric matrix protein’) (Budde et al. 2005). All these molecules thus appear to have roles in the collagen fibre network and are likely to provide interactions important for its stability and indeed crosslinking various structural elements such as collagen fibres, the beaded filaments of collagen type VI and aggrecan, all important for tissue integrity, schematically illustrated in Figure 1.

Small leucine rich proteins/proteoglycans

An important set of molecules in most extracellular matrices are the small leucine rich proteins (for references see Schaefer & Iozzo 2008). The characteristic feature of these is a number of repeats of somewhat variable length of around some 25 amino acids with leucine residues at conserved locations. One class of three of these proteins have six repeats and the others three classes have 10–11 repeats. The repeat region is in all cases flanked by disulphide loop structures. The one N-terminal contains two ‘crossed’ disulphide bonds while the one C-terminal contains a single disulphide with the exception of chondroadherin representing its own class. With the exception of this molecule all of the proteins have a variable N-terminal extension. In one class of SRLPs represented by decorin, biglycan and asporin this domain is very acidic in nature. In decorin and biglycan it contains one and two chondroitin sulfate/dermatan sulfate side chains such that these molecules are actually proteoglycans. Like all the other SLRPs, these molecules bind to fibril forming collagens via their leucine rich repeat domain (Hedbom & Heinegård 1993; Douglas et al. 2006; Zhang et al. 2006). Thus, it appears that the molecules bind the collagens during fibril formation and modify the rate and end stage of the process. Thus, in the decorin knockout mouse, the collagen fibres have very variable dimensions and are generally thicker than in the wild type mouse (Danielson et al. 1997), a result of the absence of the fibrillogenesis inhibitor decorin. Several of the SLRPs remain bound to the collagen fibril with their side chains protruding out possibly binding to neighbouring fibres (Scott & Orford 1981).

In addition, both biglycan and decorin appear to bind via their core protein to the beaded filaments of collagen type VI via the N-terminal part of this collagen (Wiberg et al. 2001). Interestingly, these proteoglycans bind already to the collagen type VI molecule and use their glycosaminoglycan chains to direct the assembly of the collagen into the beaded filaments (Wiberg et al. 2002). As discussed above, the molecules remain bound to the collagen filament and provide further interactions also with the matrilins, thereby cross-linking to other elements in the extracellular matrix.

Asporin also binds to collagen via its leucine rich repeat domain with high affinity (Kalamajski et al. 2009). This protein is interestingly much upregulated in the cartilage of early osteoarthritic patients (Lorenzo et al. 2001; Ikegawa 2008). The molecule has an N-terminal extension with a variable number from some 8 to 19 continuous aspartate acid residues. It has been shown that in the Asian populations those with 14 such residues have a higher incidence of osteoarthritis (Kizawa et al. 2005). Interestingly, the molecule binds calcium in an assay where none of select other SLRPs tested do (Kalamajski et al. 2009). This may have a role in the aberrant calcification in osteoarthritis.

The largest class of the SLRPs is the one where fibromodulin was the first defined member (Antonsson et al. 1993). Other members of this class include closely related lumican (Chakravarti et al. 1995), keratocan (Corpuz et al. 1996) and osteoadherin (Sommarin et al. 1998). All of these contain an N-terminal extension with one or several sulfated tyrosine residues. PRELP (Bengtsson et al. 1995), also a member of this class, on the other hand contains a basic cluster N-terminal domain that binds to heparin (Bengtsson et al. 2000). All of these proteins bind to fibrillar collagen via their leucine rich repeat region. Lumican and fibromodulin appear to bind to the same site and compete for binding, with fibromodulin showing somewhat higher affinity (Kalamajski & Oldberg 2009). Fibromodulin is highly negatively charged by containing one or two keratan sulfate chains linked to the protein core via the classical N-glycosidic linkage. Furthermore, the N-terminal extension contains up to nine sulfated tyrosine residues forming a cluster that together with a large number of acidic amino acid residues provides a very polyanionic domain (Önnerfjord et al. 2004). Interestingly this domain may simulate heparin in many interactions. In accordance, it has been shown to bind growth factors like FGF-2 and cytokines such as IL-10 and Oncostatin M. Other factors that bind are MMP-13 as well as the heparin binding proteins PRELP, Chondroadherin and collagen type IX, via its NC-4 domain (Tillgren et al. 2009). Binding is usually tight, with dissociation constants in the nanomolar range. Since this domain appears to be exposed when the fibromodulin binds via its leucine rich repeat domain to collagen, it is likely to serve important roles in cross linking collagen fibres as well binding and sequestering growth factors in the matrix to be released when it is degraded.

In support, during cartilage breakdown, both in a model where it is induced by stimulation with IL-1 and in joint disease in vivo, fibromodulin is degraded such that the N-terminal sulfate domain is fragmented and released, (Heathfield et al. 2004) illustrated in Figure 2. The fragmentation is induced by MMP 13. As a consequence, interactions with other molecules are lost and any bound growth factor will be released. The major part of the molecule containing the collagen binding LRR-repeats is retained in the tissue much longer, possibly protected by the interactions with the collagen.

Figure 2.

Illustrations of some specific events in cartilage breakdown indicating specific cleavage sites and enzymes that are know to induce this cleavage also at the tissue level. The degradation of the molecules aggrecan, fibromodulin, collagen 9 and collagen 2 are indicated.

Lumican appears to share many of its functions with fibromodulin, except that its N-terminal part contains only a few (maximally 4) tyrosine sulfated residues and therefore does not provide the binding to the various growth factors via their heparin binding domains.

Osteoadherin, on the other hand contains up to six more closely spaced tyrosine sulfate residues in its N-terminal part and two in its C-terminal extension (Önnerfjord et al. 2004) and binds the growth factors as well as the heparin binding domains. Interestingly in contrast to the domain from fibromodulin it does not promote binding of MMP 13 (Tillgren et al. 2009).

Two of the SLRPs (chondroadherin and osteoadherin) bind to integrin cell surface receptors α2β1 (Camper et al. 1997) and αvβ3 (Wendel et al. 1998) respectively. The functional roles of osteoadherin are not documented, but it is of interest to note that the protein is present at the same sites as another bone protein, namely bone sialoprotein (BSP), and is found particularly enriched at the bone-cartilage interface (Ramstad et al. 2003). Chondroadherin is particularly enriched in cartilage, but also present in bone. The protein binds to the integrin on several cells and elicits intracellular signals in the form of tyrosine phosphorylation of factors like ERK (Haglund, Tillgren, Heinegård, unpublished work). In bone cells, i.e. osteoblasts, one downstream consequence of chondroadherin binding is a decreased production of cytokines produced by the cells to activate osteoclasts. Examples are IL-1, IL-6 and RANK ligand. An active peptide corresponding to the integrin binding site decreases bone resorption tested on bone slices. It is not known whether the protein will have the same effects on other cells, like those of the inflammatory system (Alamanou, Rucci, Camper, Kristoffer, Rufo, Gautvik, Teti and Heinegård, manuscript).

Two of the proteins within the SLRP family, i.e. chondroadherin and PRELP, contain a heparin binding domain (Tillgren et al. 2009). Via this domain the proteins can bind to cell surface receptors in the form of the syndecan heparan sulfate proteoglycans as well as to the glypican family of cell surface proteoglycans (Haglund, Tillgren and Heinegård, manuscript). Interestingly, ongoing work shows that the two heparin binding domains with different character of the clusters of basic amino acids show specificity for different cell types. This in-line, with the variable nature and potential for a large number of specific structures within the heparan sulfate glycosaminoglycan chains. These heparan sulfate chains have the potential to create cell specific binding sites for these two molecules that also contain other sites for interactions, exemplified by the domains tightly binding to collagen present in the extracellular environment (Månsson et al. 2001; Bengtsson et al. 2002). As a consequence, these proteins may have as one function to provide bridging from the extracellular environment to the cell. The heparin binding domain can upon binding to the cells directly elicit signals and specific activities. The chondroadherin heparan sulfate binding domain, can in analogy with the corresponding heparin binding domain of fibronectin, in concert with the integrin binding domain of the respective protein enlist in the formation of focal adhesion complexes and cell migration (Haglund, Tillgren and Heinegård, manuscript).

Another set of structures present in the extracellular matrix that can bind these heparin binding domains are represented by tyrosine sulfate domains, e.g. those on fibromodulin and osteoadherin (Tillgren et al. 2009). It is not clear whether there is a competition between binding to the cell surface proteoglycans and to the tyrosine sulfate domains.

COMP (Thrombospondin 5)

The extracellular matrix contains a set of molecules referred to as the thrombospondins containing five family members for references see (Adams & Lawler 2004). The typical representative is COMP, which will be discussed in greater detail as an example of these family members. COMP is primarily present in cartilage although the molecule can be found in other tissues, particularly those experiencing pressure load (Smith et al. 1997; Wong et al. 1999). It has been shown that dynamic load of cartilage explants will induce an upregulation of the production of the protein.

COMP consists of five identical subunits held together very close to their N-terminal by a coiled-coil domain, reinforced by disulphide bridges. The protein is modular and contains EGF repeats as well as calmodulin type units (Oldberg et al. 1992). The C-terminal end contains a globular domain which functions to bind a number of other matrix molecules. Among those bound, the fibrillar collagens type 1 and 2 have high affinities with a dissociation constant in the order of 10⁻⁹ (Rosenberg et al. 1998). The COMP molecule can thus via its five subunits at the same time bind five collagen molecules. Although there are four binding sites along the collagen, the COMP molecule can not span between any two. One consequence of this binding is that the COMP very rapidly brings five collagen molecules in close proximity and facilitates their interactions in enhancing and accelerating collagen fibril formation (Halasz et al. 2007). Interestingly, the COMP molecule does not bind to the collagen fibril and it appears that it leaves the forming fibre at an early stage. Thus, it represents a catalyst that serves to bring the collagen molecules together during early fibril formation. One consequence of this is that high COMP concentrations relative to the collagen will lead to occupancy of most binding sites on the collagen by a single COMP molecule and in effect an inhibition of collagen fibril formation, illustrated in Figure 3. This seems to be a situation in the very common joint disease osteoarthritis, where production and presence of COMP in the tissue is very much elevated, while collagen production is low. Thus the imbalance between building blocks produced is one factor that may hamper the effectiveness of tissue repair processes.

Figure 3.

COMP acts as a catalyst enhancing collagen fibril formation at low relative concentration by crossbridging and keeping the collagen molecules together, while acting as an inhibitor at high concentration relative to collagen by saturation of sites with single molecules precluding crossbridging. The coordinated synthesis of building blocks is a prerequisite for assembly of tissue structures.

COMP has other interaction partners, where one is collagen type IX. Binding is to anyone of the non-collagenase domains surrounding the three triple helical ones. Again, binding has high affinity with a KD of 10⁻⁹ (Thur et al. 2001). This interaction actually provides for a potential for COMP to bind to collagen fibres, but indirectly so via the collagen type IX molecules that are exposed at the fibre surface (Vaughan et al. 1988). Another tight interaction is to a matrilin (Mann et al. 2004), which in turn have collagen binding activity. As a consequence the COMP molecule appears to have one function in matrix assembly facilitating the collagen fibre formation, while in the adult cartilage with little collagen turnover the molecule primarily appears to have a role in the stability of the collagen network. In support of the role of COMP in collagen fibre formation, the levels are very much elevated in fibrotic conditions such as scar formation (Smith et al. 1997) and systemic sclerosis of the skin (Hesselstrand et al. 2008) at the same time as it’s not detectable in the normal skin.

Tissue in disease, particularly the articular cartilage in joints

The increased knowledge of the molecular composition of tissues like cartilage has provided an important background for studies of alterations in disease. In studies of fragmentation of molecules present in the extracellular matrix, one strategy is to induce tissue breakdown by factors, e.g. cytokines, which trigger the cells to secrete proteolytic enzymes which accomplish fragmentation of molecules in the tissue extracellular matrix, exemplified in Figure 2. This approach, applied to cartilage shows a sequence of events where initially aggrecan is fragmented and released, followed by a set of molecules including COMP, fibromodulin and later collagen type IX whereafter the major collagen fibres network is degraded and removed (Goldberg et al. 1995; Heathfield et al. 2004). This has proven valuable to discern the sequence of events that occurs. When the amino acid sequences surrounding the cleavage site have been identified, it becomes possible to search for those enzymes that can accomplish the very specific fragmentation. Examples of the successful approach is the identification of specific cleavages is in aggrecan, and the identification of the enzymes ADAMTS-4 and 5 (Lohmander et al. 1993; Glasson et al. 2005; Little et al. 2005) which accomplishes cleavages at four specific sites along the molecule. Specific antibodies have been raised toward the new termini created by these cleavages, referred to as neo epitope antibodies (Hughes et al. 1995). Although the COMP cleavages have not been established in detail, it appears that the primary cleavage site resides close to the linking coiled-coil domain in the N-terminal part (Dickinson et al. 2003). Cleavages affecting fibromodulin, primarily occurring bound at the collagen fibril surface, are initially releasing the N-terminal domain with a tyrosine sulfate clusters. One consequence of this is impaired interactions with molecules on neighbouring collagen fibres, such as the NC-4 domain of collagen type IX as discussed above. Specific neo epitope antibodies are available both to the N-terminal of the larger fragment initially retained bound at the collagen fibres in the tissue and to a smaller fragment that is released (Önnerfjord, Heathfield, Heinegård, unpublished work) (e.g. to synovial fluid from cartilage). The cleavage site of collagen type II has been very well established and antibodies to the neo epitopes produced are available.

Monitoring disease

In joint disease, as a consequence of the fragmentation of specific matrix molecules, some of the fragments are released and can be assayed, initially in the surrounding synovial fluid, but they with eventually reach the circulation. Over time, assays have been developed to non-specified fragments and successfully used to verify that indeed they are useful in monitoring the activity of the tissue destroying process. Unfortunately, applications using readily available serum samples are hampered by high background levels of the fragments formed as a result of the normal adaptive tissue turnover. The neoepitopes formed by disease specific fragmentations offer a potential for new assays with a higher level of specificity for the disease process. The prerequisite, however, is that the cleavage in pathology is different from that in the normal turnover. In this case, the aggrecan cleavage is caused by the ADAMTS enzymes are the same in normal turnover as in pathology, such that high background can be expected from normal turnover. In principle the same is true for fibrillar collagen turnover, but in this case in the normal adult individual the turnover of collagen is so slow (in excess of 100 years) that there is little contribution and the breakdown in disease contributes the bulk of the fragments observed. Interestingly, it appears that the fragmentation of fibromodulin and collagen type IX is induced by MMP 13 (Heathfield et al. 2004; Danfelter et al. 2007), in a process that appears more specific to disease. In preliminary experiments, it appears that assays for new epitopes show a high level of specificity for the disease process, with no or minor contribution from other normal cartilages.

It is important to bear in mind that some of the fragments released are likely to have biological activities, which when they are presented in a new environment may affect the overall disease process. An old observation in the clinic is that the inflammation causing many of the symptoms in both rheumatoid arthritis and osteoarthritis recedes upon joint replacement, which primarily removes all the cartilage in the joint. It can be hypothesized that one or several factors released from this tissue can activate or accelerate the inflammation. Indeed, studies of the leucine rich repeat proteins, has shown that fibromodulin in particular (Sjöberg et al. 2005), but also osteoadherin and lumican can very efficiently activate the classical pathway of the complement cascade (Sjöberg et al. 2009), an important player in inflammation. There are differences such that e.g. biglycan and decorin do not provide this activation but rather function as inhibitors (Groeneveld et al. 2005; Sjöberg et al. 2009). As a consequence of release of such active fragments to the synovial fluid, the inflammation may increase, with more production of catabolic cytokines that in turn activates further tissue breakdown. This represents the hallmark of a vicious circle helping to perpetuate the process in chronic disease.

One consequence of the increasing knowledge of processes in the tissues in pathology, strengthened by an understanding of molecular functions is new opportunities to develop new therapeutics such as proteinase inhibitors that are specific for the disease process, without the side effects on normal tissue turnover that have caused problems in the past. It is interesting to note that MMP 13, originally only identified as a collagenase, shows an activity to non-triple helical structures in a gelatinous activity. Interestingly a new set of MMP 13 inhibitors have been recently identified having apparently different specificity (Baragi et al. 2009) than the early inhibitors primarily targeting the collagen breakdown. It remains to be shown whether they are effective in preventing e.g. fibromodulin breakdown.

The development over years of our understanding of the extracellular matrix biology, initially mapping of molecules present followed by studies of their functions and subsequently evolving into studies of alterations in disease have provided new interesting openings in therapeutic endeavours This development is particularly important in disease like those affecting the joint where particularly in the most common osteoarthritis there is not yet any disease modifying therapy currently.