2.3. Family of type XV and XVIII collagens

2.3.1.1. Type XV collagen -specific features

The complete primary structure of the α1(XV) chain has been elucidated for human and mouse. The human α1(XV) chain consists of 1388 residues, which include a 25-residue putative signal peptide, and it has a calculated molecular mass of 141,92 kDa. The following domains can be discerned: a 530-residue N-terminal NC domain; a 577-residue frequently interrupted collagenous region; and a 256-residue C-terminal NC domain. The collagenous region consists of nine COL domains separated by eight NC domains (Kivirikko et al., 1994; Muragaki et al., 1994). The corresponding mouse α1(XV) chain is 1367 residues in length, with a 25-residue putative signal sequence, and has a molecular mass of 140,52 kDa. The 507-residue collagenous region is flanked by N and C-terminal NC domains of 579 and 256 residues, respectively. There are two less COL domains in mouse than in human, and the collagenous region in the mouse polypeptide is divided into seven COL domains separated by six NC domains (Hägg et al., 1997a). The second half of the N-terminal NC domain diverges in human and mouse. A 41-residue sequence is found in mouse, but is not included in the human polypeptide. Overall, the human and mouse α1(XV)-chains are highly homologous. They both contain a tsp-1 homology region in the N-terminal NC domain, and they are similar with respect to the lengths of their NC and COL domains and the positions of short interruptions in the collagenous sequence. Furthermore, both chains have eight conserved cysteines and a number of putative sites for N-linked and O-linked glycosylation, most of which are conserved.

To determine the molecular mass of full-length type XV collagen, Hägg et al. (1997b) prepared a baculovirus expression construct and produced full-length α1(XV) chains in insect cells. The recombinant α1(XV) chain had a molecular mass of 200-kDa on SDS-PAGE, which was greater than the mass predicted from the sequence. Several laboratories have employed SDS-PAGE gel electrophoresis of total protein homogenates followed by Western blotting to further investigate the biochemical properties of type XV collagen protein. This has resulted in some heterogeneity in the literature regarding the size of type XV collagen protein in various cell and tissue sources, most likely due to difficulties in preparing protein samples of such large proteins. In HeLa cells, 125-kDa α1(XV) monomers have been observed (Myers et al., 1992). Hägg et al. (1997b) reported a fuzzy, high-molecular weight band in human heart and kidney homogenates, which was reduced to 100 and 110-kDa polypeptide chains. This was in agreement with the results obtained earlier by Myers et al. (1996), who reported a 116-kDa type XV collagen monomer in placenta and colon. The biochemical properties of type XV collagen have been extensively studied by Li et al. (2000). Li and coworkers demonstrated in various human tissues and human rhabdomyosarcoma cells that upon chondroitinase treatment, a diffuse smear of ≥ 400 kDa resolves into 250/225-kDa core protein forms, indicating that type XV collagen is a chondroitin sulfate proteoglycan. By using antibodies against N and C-terminal NC domains in combination with chondroitinase and/or collagenase digestions, they could further demonstrate that most of the GAG-chains are attached to the N-terminal NC domain and that the 250/225 kDa forms differed in their C-terminal NC domains. Furthermore, they showed that although there are a total of eight cysteines present, only two cysteine residues in the collagenous region participate in the interchain disulfide links.

The biochemical properties of the type XV collagen endostatin homologue region are discussed in 2.3.3.3.

2.3.1.2. Type XVIII collagen -specific features

Initially, three independent groups reported mouse cDNA clones encoding a new collagen chain, which was designated as α1(XVIII) (Abe et al., 1993; Oh et al., 1994a; Rehn & Pihlajaniemi, 1994). Subsequently, the primary structure of α1(XVIII) has been completely elucidated for human and mouse (Oh et al., 1994b; Rehn et al., 1994; Saarela et al., 1998a, and see below), C. elegans and Xenopus leavis (Ackley et al., 2001; Elamaa, H., Peterson, J., Pihlajaniemi, T., & Destrée, O., submitted), and partially for chick (Halfter et al., 1998). The major difference compared with type XV collagen is that type XVIII collagen exists as three variants differing in their N-termini (Muragaki et al., 1995; Rehn & Pihlajaniemi, 1995), whereas only one form is known for type XV collagen. The variant forms originate from the use of two alternative promoters and alternative splicing of the ensuing transcripts. The shortest variant of the mouse α1(XVIII) chain consists of 1315 residues with the following structure: a 25-residue putative signal peptide; a 301-residue N-terminal NC domain with tsp-1 homology motif; a 674-residue collagenous sequence; and a 315-residue C-terminal NC domain. The collagenous sequence consists of ten COL domains separated by nine NC domains (Oh et al., 1994b; Rehn et al., 1994). The two longer variants are 1527 and 1774 residues in length with 517- and 764-residue N-terminal NC1 domains, respectively. Moreover, the longest variant contains a 110-residue sequence with 10 cysteines, termed the fz-motif (Rehn & Pihlajaniemi, 1995) or the CR-domain (Muragaki et al., 1995), which is homologous to the wingless ligand binding domain of frizzled proteins and in several other proteins (Rehn et al., 1998).

Only two variants have been reported for the human α1(XVIII) chain encoding 1516- and 1336-residue polypeptides, thus the longest variant characterized by the fz-motif has not been identified at cDNA-level despite extensive search (Saarela et al. 1998a). However, the sequences coding for the fz-motif have been found in the human genomic DNA (Elamaa pers. commun; GenBank, NT 001151, locusID 80781). These results suggest low expression levels for this variant in humans. Interestingly, the C. elegans homologue for type XV and XVIII collagens lacks the fz-motif, but instead contains the netrin receptor unc-40 motif (Ackley et al., 2001, and see later in 2.3.3.5.), whereas the frog, Xenopus leavis, has the fz-motif (Elamaa, H., Peterson, J., Pihlajaniemi, T., & Destrée, O., submitted).

The two human α1(XVIII) chains have different signal peptides of 23 and 33 residues, respectively, and 493 and 303-residue NC1 domains corresponding to the NC1-517 and NC1-301 domains of the mouse (Saarela et al., 1998a). The collagenous region in the human chain is 688 residues in length and consists of ten COL domains, as does the mouse polypeptide. The C-terminal NC domain in humans is 312 residues and shares the greatest homology with the mouse, especially in the extreme 184 residues representing the endostatin fragment (85% identity and 99% homology). The human α1(XVIII) transcript was found to undergo alternative splicing affecting a 43-residue stretch at the junction of the NC1 domain and the beginning of the collagenous sequence.

The sequence-derived molecular masses are 182, 156, and 134-kDa for the mouse α1(XVIII) variants (Rehn and Pihlajaniemi, 1995) and 154 and 136-kDa for the human variants (Saarela et al., 1998a). The α1(XVIII) chain from mouse ES cell extracts migrates in the SDS-PAGE with a molecular mass 200-kDa (Muragaki et al., 1995). Halfter et al. (1998) detected a high molecular weight smear of ~300 kDa in chick vitrous body, meninges, amnion, and kidney extracts, which reduced to a core protein of 180-kDa upon heparitinase treatment, indicating that type XVIII collagen is a heparan sulfate proteoglycan. Saarela et al. (1998b) reported similar findings in human kidney extracts using heparin lyase II and III digestions.

The biochemical properties of the proteolytic fragment of type XVIII collagen, endostatin, are discussed in detail in 2.3.3.3.

2.3.2.1. Type XV collagen -specific features

Northern blot analysis of adult human (Kivirikko et al., 1995; Myers et al., 1996) and mouse (Hägg et al., 1997a) tissues has indicated that type XV collagen has a prevalent and widespread distribution. Strong hybridization signals have been seen in the human heart, skeletal muscle, placenta, testis, ovary, small intestine and colon, and moderate signals have been seen in the kidney, pancreas, lung, and prostate. Weak signals have been detected in the spleen, whereas the brain and liver were negative. The results obtained by Northern blot analysis of mouse tissues were essentially the same, but low levels of α1(XV) transcripts could be also detected in brain and liver.

The in situ hybridization study of 20-gestational week human fetus indicated that the main producers of type XV collagen are mesenchymally derived cells, especially fibroblasts, muscle cells, and endothelial cells, although certain epithelial cells also produce it (Kivirikko et al., 1995). Briefly, all muscular tissues, namely skeletal, cardiac, and smooth muscle tissue, labeled for type XV collagen. Type XV collagen was also detected in vasculature in general, since strong labeling was seen in all endothelial cells and smooth muscle cells around larger arteries. The epithelial cells of the lower part of the nephron in developing kidney, as well as those in the developing alveolar structures of lung, were positive.

The deposition of type XV collagen protein has been studied by immunostainings in human (Myers et al., 1996; Hägg et al., 1997b) and mouse (see III) tissues. Type XV collagen was localized to the BM zone in many tissues, and was especially prominent around capillaries and skeletal muscle cells. It was also detected around smooth muscle cells, lipocytes, and Schwann cells, as well as in some epithelial BM zones. It was not found in all BM zones, since some epithelial BMs, such as those of the developing fetal alveoli, and some tubular BMs of kidney, lacked type XV collagen. Nor was it restricted to BM zones, as some could be observed in the fibrillar collagen matrix of e.g. skin and placenta (Myers et al., 1996; Hägg et al., 1997b). Electron microscopy was used to determine the extent of association of type XV collagen with BM (Myers et al., 1996). It was localized to the outermost aspect of the lamina densa of the trophoblastic BM facing the adjacent interstitium, whereas in the endothelial BM it was distributed throughout its entire thickness. The interstitial localization was restricted, as only those collagen fibers close to the BM contained type XV collagen. Collectively, the Northern blot, in situ, and immunofluorescence data were suggestive of developmentally regulated expression of type XV collagen in some organs, such as kidney and lung, and in the vasculature (Kivirikko et al., 1995; Hägg et al., 1997a). This matter has been thoroughly addressed in the study of type XV collagen in murine development (see III).

2.3.2.2. Type XVIII collagen specific features

Northern blotting and in situ hybridization have been used to study the expression of type XVIII collagen transcripts in tissues. A distinguishing feature of type XVIII collagen is its prominent expression in liver, although high levels can also be found in other tissues, such as kidney and lung, and moderate levels in skeletal muscle and testis. Brain, heart, and spleen contain markedly lower levels (Abe et al., 1993; Oh et al., 1994a and b; Rehn & Pihlajaniemi, 1994; Saarela et al., 1998a). Furthermore, the three N-terminal variants are expressed in a tissue-specific manner (Muragaki et al., 1995; Rehn & Pihlajaniemi, 1995; Saarela et al., 1998a). The high expression in the liver is mainly due to the occurrence of mRNAs encoding long variants, especially the NC1-517/NC1-497 variant, but also to some extent of the NC1-764 variant characterized by the fz-motif. The other tissues, however, contain the short NC1-301/NC1-303 as the main form. Furthermore, in situ hybridization revealed that the long variant is mainly produced by hepatocytes, whereas main producers of the short variant are endothelial and epithelial cells (Saarela et al., 1998b). The type XVIII collagen protein is found in virtually all BM zones with strong association in vascular BMs (Muragaki et al., 1995; Halfter et al., 1998; Saarela et al., 1998b, Miosge et al., 1999). The long variant has a restricted distribution in the liver sinusoids and occurs in only minor amounts elsewhere, whereas the short variant is found in most conventional BMs, including blood vessels and various epithelial structures. Briefly, type XVIII collagen is found e.g. in the Bowman´s capsule, glomeruli and tubuli of the kidney, around the alveoli and bronchi of the lung, in the peripheral nerves, and at the epidermal-dermal junction of the skin (Saarela et al., 1998b). Ultrastructurally, type XVIII collagen localizes to the matrix side of the lamina densa in various ocular BMs (Fukai, N., Eklund, L., Marneros, A.G., Oh, S.P., Keene, D.R., Tamarkin, L., Li, E., Pihlajaniemi, T., & Olsen B.R., submitted). In addition, heavy gold-labeling was seen in those vitreal collagen fibrils that appear to be inserted in the inner limiting membrane.

Thus, homologous collagen types XV and XVIII exhibit both similarities and differences in their expression patterns and tissue distributions. They are often present in the same tissues, as is the case in capillaries, and are in fact produced by the same cells. Nevertheless, type XV collagen has a more restricted distribution in the epithelial BM zone, but predominates in the skeletal and cardiac muscle. In contrast to type XVIII, type XV collagen is more often found in the fibrillar collagen matrix. On the other hand, type XVIII collagen is a prominent component of liver, from which type XV is virtually lacking.

2.3.3.1. Roles of type XV and XVIII collagens in fibrosis and cancer

Both type XV and XVIII collagens have been implicated in fibrotic processes, type XV collagen in the context of kidney fibrosis (Hägg et al., 1997b) and type XVIII collagen in fibrotic liver (Musso et al., 1998). Kidney biopsies from patients with glomerular diseases originating from various pathological processes demonstrated pronounced accumulation of type XV collagen immunostaining in areas of interstitial fibrosis and in the thick-walled, sclerotic capillaries of diabetic glomeruli (Hägg et al., 1997b). These areas have been previously shown to accumulate at least type IV collagen and laminin (Falk et al., 1983; Kim et al., 1991; van den Born et al., 1995). In further studies, type XV collagen was found to be a sensitive indicator of glomerular damage, and was detectable already before collagen IV accumulation in mildly affected, border-line cases of mesangial glomerulonephritis (Hägg, P.M., Hägg, P.O, Autio-Harmainen, H., & Pihlajaniemi, T., unpublished results). The recent study of moderately differentiated colonic adenocarcinomas indicates that type XV collagen is less prevalent in the epithelial BM zones of malignant glands than type IV collagen and laminin, and like the other two proteins studied, shows an apparent shift in distribution from BM/BM zone into the tumor interstitium. These results collectively suggest a role in the invasive process for this collagen (Amenta et al., 2000).

The prominent expression of type XVIII collagen in liver raises the obvious question of its biological importance in liver diseases, e.g. in cirrhosis and carcinomas. Indeed, in 1998, Musso and coworkers demonstrated that type XVIII collagen is associated with BM remodeling and sinusoidal capillarization in liver cirrhosis. Type XVIII collagen was shown to be accumulated in cirrhotic nodules, where it formed thick deposits along the capillarized sinusoids. Interestingly, colocalization with an endothelial marker indicated that newly formed capillary sprouts produced type XVIII collagen, which deposited in BM-like cuffs around them. This finding is noteworthy in light of the potential role of type XVIII collagen endostatin in the regulation of angiogenesis (see 2.3.3.3.). Moreover, type XVIII collagen expression was shown to increase differentially in cells expressing this collagen as the fibrosis progressed. The increase in expression in activated stellate cells/myofibroblasts was 13-fold in the active and 2-fold in the quiescent stage of the disease, whereas in hepatocytes it was only 2-fold in both stages, suggesting that activated stellate cells/myofibroblasts are a major site of collagen XVIII expression in fibrosis. Further, these two cell types express different variants of collagen XVIII, since hepatocytes have been shown to express exclusively the long or NC1-493 variant, whereas BM-producing cells in liver express the short or NC1-303 variant (Saarela et al., 1998b; Musso et al., 2001a). Recently, the two variants were shown to be present in a strikingly tissue-specific manner in primary and metastatic liver cancers (Musso et al., 2001a). Tumor hepatocytes were shown to be the major source of the long variant in human hepatocellular carcinomas (HCCs). However, the short variant was identified in the stromal compartment of tumor tissues in both primary (HCCs), and metastatic liver cancers (cholangiocarcinomas and colorectal cancer metastases). It was produced by both tumor and stromal cells in cholangiocarcinomas, and by stromal cells in colorectal cancer. Interestingly, the expression level of the long variant has been shown to correlate with the tumor size in HCCs (Musso et al., 2001b). Low collagen XVIII expression in hepatocytes was associated with larger tumor size, whereas tumors expressing high collagen XVIII levels were small and had low microvessel density. This indicates that a decrease in type XVIII collagen expression is associated with angiogenesis in primary liver cancer. In summary, the presence of type XVIII collagen in neoplastic tissue has been postulated to provide an endogenous source of endostatin, which could be released by proteases into the tumor microenvironment and affect the tumor progression and angiogenesis (see 2.3.3.3. for further discussion).

2.3.3.2. Skeletal myopathy and cardiovascular defects in type XV collagen deficient mice

The generation of a mouse line deficient in type XV collagen has proven to be a valuable tool in defining the biological role of type XV collagen. The Col15a1-deficient mice developed symptoms in their skeletal muscles, microvessels, and heart (Eklund et al., 2001). After 3 months of age, the Col15a1^-/- mice showed progressive histological changes in their skeletal muscles, such as muscle cell degeneration, macrophage infiltration, and increased regeneration. These changes are frequently encountered in muscle diseases. This, together with the finding that Col15a1-deficient mice are more vulnerable to exercise-induced muscle damage than the control mice, suggests a role for type XV collagen in providing mechanical stability between cells and ECM in skeletal muscle. Previously, disruption of this linkage due to the mutations in the components of dystrophin-associated glycoproteins-laminin α2 axis (Campbell, 1995), the integrin α7 subunit (Mayer et al., 1997), and type VI collagen (Jöbsis et al., 1996; Bonaldo et al., 1998) has been shown to be implicated in the etiology of muscle diseases. Despite the potent antiangiogenic role of collagen XV endostatin-like region, the development of the vasculature was normal in the null mice. However, electron microscopy revealed structural changes such as collapsed capillaries and endothelial cell degeneration in the heart and skeletal muscle, which suggests a structural role for type XV collagen also in the microvessels. Furthermore, functional changes were observed in the Col15a1^-/- heart, such as diminished inotrophic response and damage after increased cardiac workload. Both of these changes mimic early or mild heart disease, possibly resulting from impaired perfusion (Eklund et al., 2001).

2.3.3.3. Roles of type XV and XVIII collagens in angiogenesis

Angiogenesis, the formation of new blood vessels from pre-existing ones, is an important biological process during development, tissue growth and regeneration, and in pathological conditions such as tumor growth and metastasis. It is controlled by a concerted action of various activators and inhibitors, several of which are known (Hanahan & Folkman, 1996; Risau 1995 and 1997; Beck & D"Amore, 1997; Folkman, 2001). In 1997, O"Reilly and coworkers identified a novel inhibitor of angiogenesis, endostatin, which turned out to be a proteolytically cleaved fragment of the C-terminal NC1 domain[1] of type XVIII collagen. Endostatin was shown to inhibit endothelial cell proliferation and migration, induce endothelial apoptosis, and, when used as the recombinant bacterial product, to suppress tumor-induced angiogenesis, resulting in the regression of tumor growth in several experimental tumors in mice (Boehm et al., 1997; O"Reilly et al., 1997; Dhanabal et al., 1999a and b; Yamaguchi et al., 1999). Somewhat surprisingly, the mice lacking collagen XVIII did not show any overt defects in their vascular development, although some abnormalities in blood vessel formation were observed at specific locations, supporting the role of collagen XVIII/endostatin in the regulation of angiogenesis at least locally (Fukai, N., Eklund, L., Marneros, A.G., Oh, S.P., Keene, D.R., Tamarkin, L., Li, E., Pihlajaniemi, T., & Olsen B.R., submitted). Briefly, the Col18a1^-/- mice showed delayed regression of blood vessels in the vitreous, resulting in the abnormal outgrowth of retinal vessels. The persistence of hyaloid vessels in the Col18a1^-/- mice was speculated to lead to high local oxygen levels, which would in turn prevent the normal upregulation of VEGF, known to be implicated in the oxygen–dependent control of vessel outgrowth (Yancopoulos et al., 2000).

The X-ray crystal structures of recombinant human and mouse endostatins have been determined, demonstrating a compact globular folding pattern related to that of the carbohydrate recognition domain of C-type lectins, and the presence of a fourfold ligated zinc ion (Ding et al., 1998; Hohenester et al., 1998; 2000). The parent 38-kDa NC1 domain has been shown to assemble non-covalently into a trimeric structure. Each monomer is composed of three segments; an N-terminal association domain responsible for trimerization, a C-terminal globular 20-kDa endostatin (ES) domain, and a hinge region, which is characterized by the presence of several protease-sensitive sites (Sasaki et al., 1998).

Obviously, a lot of scientific effort has been put into the study of the mechanisms underlying the antiangiogenic activity of endostatin. As mentioned earlier, type XVIII collagen has a prominent location in vascular BM zones (Muragaki et al., 1995; Saarela et al., 1998b), where it most likely is immobilized to some kind of network. In addition, there is evidence of this collagen or its fragments in circulation. Recently, Musso et al. (2001a) reported the presence of the long variant and two lower-molecular weight polypeptides of collagen XVIII in human plasma, where others have previously demonstrated circulating endostatin (Ständker et al., 1997; John et al., 1999). There are also reports of the presence of proteins with mobility similar to NC1 domain in various tissue homogenates (Sasaki et al., 1998; 2000), whereas the aorta- and skin-derived forms correspond in size to endostatin (Miosge et al., 1999). It has been suggested that both matrix-bound and partially processed circulating forms of type XVIII collagen could function as a readily available reservoir of endostatin, contributing to the homeostatic control of angiogenesis (O"Reilly et al., 1997; Sasaki et al., 1998).

Recent data have provided evidence concerning the enzymes involved in the processing of type XVIII collagen. Metalloproteinases have been implicated in the initial generation of fragments of type XVIII collagen, which are further cleaved by elastase and/or by cathepsin L to release endostatin (Wen et al., 1999; Felbor et al., 2000; Ferreras et al., 2000; Heljasvaara, R., Parikka, M., Luostarinen, J., Nyberg, P., Rehn, M., Sorsa, T., Salo, T. & Pihlajaniemi, T., submitted), the latter cleavage site being located within the hinge region (see above). Furthermore, the processing of endostatin from the parent NC1 changes the binding activity for extracellular ligands. Recombinant endostatin has been shown to bind to heparin, fibulin-1 and -2, and nidogen-2 (Miosge et al., 1999; Sasaki et al., 1998; 2000), the latter three interactions most likely occurring also in vivo, as shown by close colocalization in the immunogold double staining of elastic fibers of aortic media (Miosge et al., 1999). The NC1 domain, on the other hand, binds heparin, sulfatides, laminin-1, laminin-1-nidogen-1 –complex, and perlecan (Sasaki et al., 1998; 2000).

A few studies have provided insight into the possible mechanisms of endostatin action, some of them conflicting, suggesting that various mechanisms may be employed. The affinity of endostatin to heparin suggests a functional significance (O"Reilly et al., 1997; Sasaki et al., 1998). It has been hypothesized that endostatin may function as an antagonist to basic fibroblast growth factor (bFGF) by competing with it for heparan sulfate coreceptors (Sasaki et al., 1998), which are known to play a key role in bFGF controlled signaling in the promotion of angiogenesis (Friesel & Maciag, 1995). This has gained support in recent site-directed mutagenesis assays demonstrating that the inhibition of bFGF-2-induced chorioallantoic membrane (CAM) angiogenesis by endostatin is dependent on its heparin-binding property (Sasaki et al., 1999). Accordingly, increased apoptosis of cultured endothelial cells in the presence of endostatin was shown to depend on tyrosine-kinase signaling through Shb adaptor protein and the heparin-binding site of endostatin (Dixelius et al., 2000). On the other hand, Yamaguchi et al. (1999) showed that endostatin could inhibit VEGF-induced endothelial cell migration despite the lack of a heparin-binding domain.

Endostatin has been shown to contain zinc (Ding et al., 1998; Hohenester et al., 2000). In 1998, Boehm et al. (1998) showed that the coordination of a zinc-atom in the N-terminal region of endostatin is essential for its activity. This finding was not supported by Yamaguchi et al. (1999), who showed that despite mutations in Zn-binding region, endostatin could inhibit VEGF-induced endothelial cell migration and regress tumor growth. The first evidence of endothelial cell-specific receptors of endostatin came from Rehn et al. (2001). They showed that recombinantly produced endostatin interacts with α₅- and α_v -integrins on the surface of the human endothelial cells, and this interaction is of functional importance in vitro, as immobilized endostatin acts as the integrin agonist and supports endothelial cell survival and migration. In soluble form, however, it functions as the integrin antagonist and inhibits endothelial cell function.

Recent evidence suggests a specific function for the trimeric NC1 domain of collagen XVIII in inhibiting ECM-induced morphogenesis. The NC1 domain was shown to act as a motility-inducing factor in the regulation of ECM-dependent morphogenesis of endothelial and other cells, and this activity requires ES domain oligomerization (Kuo et al., 2001). Furthermore, its cleavage product, endostatin monomer, was shown specifically to block this function, constituting a negative autoregulatory loop on the action of NC1. Mechanistically similar findings have been reported by Ackley et al. (2001) in C. elegans (also see 2.3.3.5.), in which the trimeric NC1 domain of collagen XVIII homologue, CLE-1, could promote neuronal motility, whereas the ES monomers inhibited this activity.

Since the identification of endostatin, the obvious question concerning the high sequence homology encompassing the C-terminal NC1 domains of type XV and XVIII collagens (referred to NC1-XV and NC1-XVIII here, see 2.3.1.) was whether the corresponding region in type XV collagen could have similar activity. Indeed, in 1999, Ramchandran and coworkers reported the characterization of restin (termed endostatin-XV here), which had some of the functional properties previously assigned to endostatin. Like endostatin, its type XV collagen -homologue could suppress the growth of tumors, although contrary to endostatin, there was no regression on tumor size. Further, it could inhibit the migration of endothelial and also some non-endothelial cells, thus differing from the highly specific activity of endostatin on endothelial cells. It did not have any effect on the proliferation of endothelial cells. Also, endostatin-XV exists as a proteolytic fragment in circulation and has been isolated from human plasma (John et al., 1999). As predicted from the homologous primary structures, the crystal structure of endostatin-XV is nearly identical to that of endostatin-XVIII (Sasaki et al., 2000). Contrary to endostatin-XVIII, its XV-homologue does not bind either heparin or zinc, which is consistent with the lack of conservation in these sites in endostatin-XV. Both NC1 domains assemble noncovalently into trimeric structures containing N-terminal association- and C-terminal endostatin domains separated by a hinge region, the latter being considerably shorter in NC1-XV and therefore less vulnerable to proteolytic cleavage. Both collagen fragments were also able to inhibit angiogenesis in the CAM-assay, but in a strikingly different manner depending on the cytokine used. Both NC1-XV and endostatin-XV inhibited VEGF-stimulated angiogenesis, which was not affected by the corresponding collagen XVIII fragments. Instead, FGF2-induced angiogenesis could be efficiently inhibited by only NC1-XV, whereas of the two collagen XVIII fragments, only endostatin-XVIII possessed that activity. Furthermore, collagen XV and XVIII showed similar, although not identical, binding repertoires for ECM proteins. NC1-XV was shown to bind strongly to fibulin-2 and nidogen-2, and about 100-fold less to fibulin-1, nidogen-1, laminin-1-nidogen-1 complex, and perlecan. Endostatin-XV exhibited nearly indistinguishable binding properties. As described earlier, endostatin-XVIII, when compared to NC1-XVIII, was a weaker ligand for all these ECM proteins, except for the fibulins (Sasaki et al., 2000). In contrast to NC1-XVIII, oligomerized NC1-XV did not show a migration-inducing effect on endothelial cells, indicating that the motogenic properties are highly specific for collagen XVIII, at least in the experimental set-up used (Kuo et al., 2001). The lack of overt defects in the vascular development of the collagen XVIII/endostatin null mice (Fukai, N., Eklund, L., Marneros, A.G., Oh, S.P., Keene, D.R., Tamarkin, L., Li, E., Pihlajaniemi, T., & Olsen B.R., submitted) was not explained by compensation by its collagen XV homologue, as the vascular development in the double Col15a1-Col18a1 knock-outs also did not show any overt phenotype (Ylikärppä, R., personal communication). This suggests that, although several lines of evidence demonstrate a role for both endostatins in controlling angiogenesis, the mechanism behind it is far more complex.

2.3.3.4. Mutations in type XVIII collagen cause Knobloch syndrome

Knobloch syndrome, a rare genetic disorder, is characterized by high myopia, vitreoretinal degeneration with retinal detachment, macular abnormalities, and occipital encephalocele (Knobloch & Layer, 1971). By using the positional candidate approach, Sertie et al. (2000) identified a mutation in the COL18A1 gene as a cause of this syndrome. This mutation changes the invariant AG to AT in the acceptor site of the first intron of the short variant of COL18A1 gene, thus resulting in the creation of a stop codon in exon 4, and hence truncation of the α1(XVIII) polypeptides. Based on the clinical manifestations of Knobloch syndrome, the type XVIII collagen short form was suggested to play a major role in determining the retinal structure and in the closure of the neural tube. Indeed, the phenotypic consequences in the Col18a1^-/- mice support a structural role for type XVIII collagen at specific locations in the eye, and provide an explanation for the eye manifestations seen in the Knobloch patients (Fukai, N., Eklund, L., Marneros, A.G., Oh, S.P., Keene, D.R., Tamarkin, L., Li, E., Pihlajaniemi, T., & Olsen B.R., submitted). Strong type XVIII collagen immunogold labeling was demonstrated at sites where collagen fibrils in the vitreous are connected with the inner limiting membrane. This, together with the finding of reduction in the number of vitreous fibrils along the inner limiting membrane in the Col18a1^-/- mice, suggest a specific anchoring role for collagen XVIII between the vitreal collagen fibrils and the inner limiting membrane. Furthermore, the insufficient closure of the neural tube is consistent with the finding that type XVIII collagen NC1 domain affects cell migration (Ackley et al., 2001; Kuo et al., 2001, and see 2.3.3.3.).

2.3.3.5. Type XV/XVIII collagen in C. elegans

As the entire genomic sequence of the nematode Caenorhabditis elegans and fruit fly Drosophila melanogaster are now available in the Genbank (http://www.ncbi.nih.gov/PMGifs/Genomes/allorg.html#c), it is possible to search for the evolutionary conservation between distant phyla. Interestingly, among the twenty-three collagen types in mammals (see Myllyharju & Kivirikko, 2001 for reviews), only collagen IV and XV/XVIII are conserved within these genomes (Hutter et al., 2000). Ackley et al. (2001) has recently characterized cle-1 as the C. elegans homologue to vertebrate type XV/XVIII collagens. Although the exon/intron boundaries are in the identical positions in cle-1 and mammalian collagen XV and XVIII genes, and the amino acid sequence is equally similar to both collagens, a number of features support the conclusion that cle-1 is more similar to type XVIII collagen. Like mammalian collagen XVIII, CLE-1 exists as three variant forms, which exhibit tissue-specificities in their expression patterns, and contains a stretch of amino acids in the ES domain that has been shown to contribute to the heparin-binding property of endostatin. Although it does not have a wingless receptor fz-motif, it contains in an equivalent position an unc-40-motif related netrin receptor. The CLE-1 protein is broadly distributed in the BMs, being especially prominent in the nervous system. The deletion of the cle-1 NC1 domain caused multiple cell migration and axon guidance defects, which could be partly rescued by ectopic expression of the trimeric NC1 domain. In contrast, the addition of a monomeric ES domain did not result in the rescue effect, but caused cell and axon migration defects in wild-type worms, thus phenocoping the NC1 deletion. This data nicely supports the observation by Kuo et al. (2001) that in mammalian system, the trimeric NC1 and monomeric ES domains are functionally different, and that the ES domain functions as an autoregulator of the NC1 domain.