Regulated cell proliferation and growth does not only require the interaction of growth factors with their cellular receptors as a prerequisite for the completion of the cell cycle. Cell proliferation is also linked with changes in the interactions of cells with their macromolecular environment, with cell migration and differentiation. Hence, the extracellular matrix surrounding the cells is of similar importance for growth control as are the interactions between soluble, growth-regulating molecules and their cellular receptors.
The extracellular matrix is a complex alloy of macromolecules capable of self-assembly, predominantly via non-covalent bonds. It is composed predominantly of collagens, non-collagenous glycoproteins, elastin, hyaluronan and proteoglycans. This matrix, however, is not only a scaffold for the cells of a given tissue. It serves also as a reservoir for growth factors and cytokines and modulates their activation status and turnover. By interacting with matrix molecules the growth factors may become sequestered from their signalling receptors, they may become activated, e.g., by proteolytic processing, and they may be presented to the cells in a manner significantly altering their bioactivity. In addition, there is increasing evidence that extracellular matrix molecules may exhibit a direct signalling function, either via interactions with matrix receptors like integrins or via signalling through growth factor receptors themselves (see, for example, Ornitz, 2000; Schönherr and Hausser, 2000; Couchman et al., 2001; Yamaguchi, 2000 for recent reviews).
All classes of extracellular matrix molecules may now be considered as macromolecules involved in growth control. However, the “information network” in which the proteoglycans of the extracellular matrix are involved, has intensively been studied only recently. Proteoglycans of the extracellular matrix may be divided into several families (see Iozzo, 1998; Lander and Selleck, 2000; Sugahara and Kitagawa, 2000; Yamaguchi, 2000 for recent reviews). The lecticans are multidomain proteoglycans containing an N-terminal globular domain, which can interact with hyaluronan, a C-terminal selectin-like domain and chondroitin sulfate side chains. Keratan sulfate chains may also be present. Family members are aggrecan, versican, neurocan, and brevican, the latter two being characteristic components of neural tissue. Providing tissues with resilience against compressive forces was long considered as the main function of the lecticans. A second family comprises the so-called small proteoglycans/glycoproteins, whose core protein is composed of leucine-rich repeat structures (hence the acronym SLRPs), which presumably allow the molecules to adopt a horseshoe-like structure well suited for protein-protein interactions. Well known family members are among others decorin, biglycan, fibromodulin, lumican, and keratocan. As proteoglycans, they carry most often either chondroitin/dermatan sulfate or keratan sulfate chains. In the past, they were considered primarily as organizers of collagenous networks. Testicans represent a third family of matrix chondroitin/heparan sulfate proteoglycans. However, since their function is still elusive, they will not be covered in this review. Other matrix proteoglycans like phosphacan, neuroglycan, and others have not been grouped into families, and they will be mentioned in the text as indicated. There is also a growing number of matrix molecules, which may or may not be linked with glycosaminoglycan chains, depending on the developmental stage or due to not yet understood regulatory factors. They are grouped together as so called part-time proteoglycans, represented for example by CD 44, macrophage colony-stimulating factor (CSF-1), amyloid precursor protein and collagens type IX, XII, XIV, and XVIII. Three different proteoglycans are found predominantly, but not exclusively, in basement membranes. Perlecan and agrin are most often heparan sulfate proteoglycans, while bamacan has been described as a chondroitin sulfate proteoglycan.
Two families of proteoglycans, which carry predominantly or exclusively heparan sulfate chains, are the membrane-associated proteoglycans of the syndecan and glypican families. Membrane-associated heparan sulfate proteoglycans represent the most intensely studied proteoglycans involved in growth control. This is due to the fact that most growth factors are able to interact with defined heparan sulfate domains with high affinity and may form ternary structures with heparan sulfate and their signalling receptors (see e.g., Bernfield et al., 1999; Lin and Perrimon, 2000; Rapraeger, 2000, for recent reviews) In this review, we focus on extracellular matrix proteoglycans and do not take into account the wealth of information about the role of cell surface-associated proteoglycans in growth control. Furthermore, since the heparan sulfate chains of perlecan and agrin may also serve to present growth factors to membrane receptors, we consider in this review only core protein-mediated functions of these proteoglycans.
There are two classes of proteoglycans, which at present appear to be capable to directly affect growth. Members of the lectican family of proteoglycans represent one class, the other one is presently represented only by two members of the small proteoglycan family containing leucine-rich motifs in their core proteins. From other reports, it became evident that in some systems growth-related effects appear to be mediated via the glycosaminoglycan chain regardless of the biological properties of the core proteins themselves. In these cases, the abundance of a given saccharide structure at the place of action is obviously of greater importance than the fine-tuning of the expression of the proteoglycan itself. It should also be stressed that for the understanding of a proteoglycan's role in growth control, it is certainly an over-simplification when only the interaction with a cell surface receptor is considered. The temporal and spatial regulation of proteoglycan expression and the possible interactions with other molecules, such as cell adhesion molecules, have to be taken into account as well. Hence, the multifaceted extracellular milieu created by the matrix macromolecules together with the multitude of soluble growth mediators, which are contained within this matrix, yield a level of complexity that cannot be traced back to the interaction of a proteoglycan with a signalling membrane receptor alone. It is, therefore, not surprising that divergent results may be obtained when correlations between proteoglycans and growth are made in a different biological context (see, for example, Bandtlow and Zimmermann, 2000, for a review).
Lecticans are characterized by the presence of globular domains separated by a glycosaminoglycan attachment region. The N-terminally located G1 domain binds hyaluronan. The C-terminus contains a selectin-like domain, which is also called G3 domain regardless of whether or not a G2 domain is present. The G3 domain is composed of two EGF-like repeats, a lectin-like motif and a complement binding protein-like motif. The C-type lectin motif of this domain mediates binding to other extracellular matrix proteins, thereby contributing to the formation of a network being permissive for cell growth (Evanko et al., 1999; Olin et al., 2001). In addition, the EGF-like repeats make the lecticans candidates for proteoglycans directly being involved in growth control (Fig. 1). It is important to note, however, that there are no published studies, which show a direct interaction of an EGF-like repeat of a lectican with one of the members of the EGF-receptor family expressed at the surface of a growth factor-responsive cell.
Most of the studies, concerning growth and lecticans have focussed on versican. Versican is highly expressed by fast-growing cells (Zimmermann et al., 1994; Bode-Lesniewska et al., 1996). Accordingly, it was observed that over-expression of versican stimulates fibroblast and chondrocyte proliferation (Zhang et al., 1998a, 1999) as well as the cells of the dermal papilla (Kishimoto et al., 1999) which signal the hair growth process. Stable transfection of versican constructs without the EGF-like motifs significantly interfered with its growth promoting property. However, at least in some cases the G1 domain plays a supportive role due to the capability of this domain to destabilize adhesive contacts of the cells (Yang et al., 1999; Zhang et al., 1999). Specifically the growth-promoting property of versican for chondrocytes may be of physiological and pathophysiological relevance since early stages of osteoarthritis are characterized by increased chondrocyte proliferation as well as by an increased versican expression (Cs-Szabó et al., 1997). As proliferation and differentiation reflect events occurring sequentially and not simultaneously in a single cell system, it has also been shown that the EGF-like motifs of the G3 domain are responsible for the inhibitory effect of versican on mesenchymal chondrogenesis during development (Zhang et al., 1998b). It has been proposed that versican constructs lacking the EGF-like motifs of the G3 domain enhance differentiation through misassembly of actin filaments, i.e., through changes in the cytoskeletal structure (Zhang et al., 2001).
As versican is also expressed in non-neuronal cells of the central nervous system, it is not surprising that correlations exist between axonal growth and versican expression. Proliferation of astrocytoma cells, too, was stimulated by the G3 domain of versican. Deletion of the EGF-like motifs, however, produced a dominant-negative effect on astrocytoma cell proliferation. This could be explained by an inhibition of versican secretion through the mutants, thus corroborating the biological relevance of versican-mediated growth stimulation at least in these types of cells (Wu et al., 2001). In contrast to these findings, the observation was that the versican splice variant V2, which is predominantly present in myelinated fiber tracts of the mature brain, exhibits potent inhibitory effects on axonal growth although the variant contains the EGF-like motif (Schmalfeldt et al., 2000). Indirect evidence suggests that brevican, another brain-specific lectican with an EGF-like motif, is also inhibitory for axonal growth (Milev et al., 1998a). Thus, it appears that the size of the core protein of lecticans may play a crucial role in neurite growth control.
A completely different mode of action leading to an inhibition of neurite outgrowth has been described for neurocan (Margolis and Margolis, 1997; Li et al., 2000). An N-terminal fragment of neurocan is able to interact specifically with the neurocan receptor, a membrane-intercalated glycosyltransferase (N-acetylgalactosamine-phosphate transferase), thereby co-ordinately inhibiting N-cadherin- and β1-integrin-mediated cell adhesion, possibly via the non-receptor tyrosine kinase Fer. This, in turn, prevents cell and neurite migration across boundaries.
Small proteoglycans with leucine-rich repeats
The involvement of decorin in growth control has been discussed on the basis of the proposal that decorin becomes up-regulated together with some other matrix and intracellular proteins in quiescent cells compared with logarithmically growing cells (Coppock et al., 1993; Mauviel et al., 1995; Laine et al., 2000). However, a complex regulation of the quiescence-induced genes with respect to cell adhesion and cell type was noted subsequently (Coppock et al., 2000). Work in our own laboratory failed to demonstrate an up-regulation of decorin in quiescent normal skin fibroblasts. In CHO cells, recombinant expression of decorin was accompanied by an inhibition of cell proliferation (Yamaguchi and Ruoslahti, 1988). The effect was considered to result from the proposed capability of decorin to inhibit the activity of TGF-β (Yamaguchi et al., 1990). While this explanation may indeed serve to explain the effect in part and/or in selected cell lines, decorin can also directly influence cell behavior in an autocrine or paracrine manner. The discovery that a fibril-associated proteoglycan serves not only as an organisator of the extracellular matrix, but is also a signalling molecule led to a novel level of complexity in cell matrix interactions. Simultaneously, it became apparent how cell-type specific the signalling pathways might be and to what large extent the final effects may depend on the cellular environment.
In A431 squamous carcinoma cells, decorin suppresses the malignant phenotype (Santra et al., 1995) by activating the EGF receptor (Santra et al., 1997) through its dimerization and increased phosphorylation (Moscatello et al., 1998). This leads to an activation of the MAPK pathway (Moscatello et al., 1998), a mobilization of calcium stores (Patel et al., 1998) and an up-regulation of p21WAF1/CIP1 (De Luca et al., 1996; Santra et al., 1997), an inhibitor of cyclin-dependent kinases being involved tumor suppression (el-Deiry et al., 1993). Upon prolonged interaction between decorin and the EGF receptor, e.g., after stable expression of decorin in tumor cells, the signalling pathway becomes desensitized by a reduction of the number of receptor molecules and by an abrogation of the degree of EGF receptor tyrosyl phosphorylation (Csordás et al., 2000). The signal-transducing role of decorin may not rely on interactions with the EGF receptor. Other members of the EGF receptor family may play a role. Thus, in breast carcinoma cells decorin seems to activate ErbB4, which in turn blocks the phorphorylation of heterodimers containing either ErbB2 or ErbB3 (Santra et al., 2000).
A somewhat different situation was encountered in endothelial cells, which do not express decorin except during angiogenesis (Järveläinen et al., 1992; Schönherr et al., 1999). Using the endothelial cell line EA.hy 926 in an in vitro model of angiogenesis, it could be shown that adenovirally mediated expression of decorin is sufficient for angiogenesis-related phenomena to occur and for the prevention of apoptosis in the absence of fetal calf serum. In EA.hy 926 cells other receptors than the EGF receptor must be involved in decorin signaling, because the signal cannot be blocked with tyrphostin AG1478, a specific inhibitor of the EGF receptor (Schönherr et al., 2000). Also calcium was not involved in signal transfer, but the phosphorylation of Akt (protein kinase B) was required to mediate some of the effects, as judged from the effects of expressing a dominant negative form of Akt (see below). Within a few hours upon decorin addition cyclin A became transiently up-regulated. As in epithelial cells decorin expression was followed subsequently by an induction of p21WAF1/CIP1 and additionally of p27KIP1. Akt, however, did only influence the expression of p21WAF1/CIP1 and cyclin A, and the mechanism(s) of induction are presently not understood. An increase in the two cyclin dependent kinase inhibitors was not only observed in endothelial cells over-expressing decorin, but also in arterial smooth muscle cells which were retrovirally induced to synthesize decorin and in addition stimulated to proliferate. In this experimental set up also a decrease in DNA synthesis by decorin was found (Fischer et al., 2001). Decorin expression in EA.hy 926 cells influenced also the NF-κB signal transduction system. p65, an active component of NF-κB, was translocated to the nucleus, and the cytoplasmic concentration of I-κBα decreased. Figure 2 shows a compilation of the different modes of signalling of decorin which have been observed until today. However, not all of these pathways may be activated in different types of cells.
From the studies of carcinoma cells one would like to conclude that decorin-deficient animals might develop spontaneous tumors frequently. This is, however, not the case. It could be shown instead, that decorin deficiency is permissive for lymphoma tumorigenesis in mice predisposed to cancer due to p53 mutations (Iozzo et al., 1999).
A further level of complexity becomes apparent from studies of periodontal ligament fibroblasts from decorin-deficient mice (Häkkinen et al., 2000). In accordance with the growth-inhibitory role of decorin, there was an about twofold higher number of cells in the periodontal ligament, and the growth of cultured ligamental fibroblasts was suppressed by decorin addition. However, blocking the EGF receptor tyrosine kinase activity did not prevent decorin-elicited growth suppression, and p21WAF1/CIP1 appeared not to be involved in the signal transduction pathway.
Biglycan, another member of the family of small leucine-rich proteoglycans, which is closely related to decorin may also directly affect signal transduction during growth and differentiation, although direct proof is lacking. It has been shown that sprouting endothelial cells increase the synthesis of biglycan before they start to secrete decorin (Järveläinen et al., 1992). However, biglycan does not activate the EGF receptor (Moscatello et al., 1998), but it has a nuclear localization sequence and has been found in the nuclei of neuronal cells (Liang et al., 1997), where it may be directly involved in the regulation of cell survival and proliferation.
A considerable body of evidence exists about the inhibitory activities of chondroitin sulfate on axonal growth upon injury of the central nervous system. As the three lecticans versican, neurocan, and brevican are widely expressed in the nervous system, it is not surprising that neurite growth-inhibitory activity has been attributed to these proteoglycans (McKeon et al., 1999; Niederost et al., 1999). However, unrelated proteoglycans like phosphacan (McKeon et al., 1999) and NG2 (Fidler et al., 1999) express the same activity. In case of NG2, a proteoglycan expressed by astrocytes, it was clearly shown that the glycosaminoglycan chain is required for the manifestation of the growth-inhibitory effect (Fidler et al., 1999). It should be noted that in none of these studies an attempt was made to identify the glycosaminoglycan receptor through which the effects are mediated.
Independent from an effect of glycosaminoglycans on the growth of neurites, it had also been shown that at least in vitro chondroitin sulfate is required for neurite survival (Koops et al., 1996; Kappler et al., 1997). This function had first been attributed to biglycan, which is expressed by astrocytes and glial cells. However, it could be demonstrated that the core protein was not required to maintain neurite survival and that different glycosaminoglycans exhibit the same biological activity, albeit at different doses.
Another recent study showed convincingly growth-related phenomena playing a role during wound healing and inflammation (Penc et al., 1999). It was demonstrated that specifically dermatan sulfate—and not heparan sulfate, as expected—served as activator of endothelial cells via the NF-κB pathway and the induction of ICAM-1. It appears, therefore, that perhaps dermatan sulfate may be similarly to heparan sulfate an information-bearing matrix component, which may modulate the biological activity of cells via signalling receptors.
MODULATION OF GROWTH FACTOR ACTIVITIES
Small leucine-rich proteoglycans
The biological implications of the interaction between small proteoglycans and the isoforms of TGF-β have been most intensely investigated. Starting with the observation that decorin expression by Chinese hamster ovary cells inhibits their proliferation via TGF-β inactivation (Yamaguchi et al., 1990), several attempts have been made to clarify the biological importance of the interplay between small proteoglycans, especially of decorin, and the cytokine. However, the cytokine may exhibit both growth promoting and growth inhibitory activities (see e.g., Sporn, 1999, for a review), and in most of the studies the role of TGF-β in matrix production and not in growth control was studied. There is no doubt that all the isoforms of TGF-β are interacting with the core proteins at least of decorin, biglycan and fibromodulin (Hildebrand et al., 1994), exhibiting dissociation constants in the nanomolar range. A centrally located decorin peptide was similarly active, and it could be shown that collagen-bound decorin, too, was able to interact with TGF-β (Schönherr et al., 1998). Controversy exists about the biological activity of the proteoglycan/TGF-β complex. In several models of fibrotic diseases (anti-Thy-1 glomerulonephritis, bleomycin-induced pulmonary fibrosis) decorin application improved considerably the course of the disease, regardless of whether the proteoglycan was administered parenterally, adenovirally or whether its synthesis was induced by gene transfer (Border et al., 1992; Isaka et al., 1996; Zhao et al., 1999; Kolb et al., 2001). Similarly, the growth response of arterial smooth muscle cells towards TGF-β was reduced upon retrovirally over-expression of decorin (Fischer et al., 2001) as was the chemotactic response of microglial cells in experimental rat glioma (Engel et al., 1999). Nevertheless, proliferation of decorin-over-expressing smooth muscle cells was not influenced after balloon injury (Fischer et al., 2000). Interestingly, no beneficial effects were seen when in pulmonary fibrosis biglycan instead of decorin was adenovirally induced (Kolb et al., 2001). On the other hand, it was shown that in several different experimental systems TGF-β remained unchanged in spite of an up to 10,000-fold molar excess of decorin (Hausser et al., 1994). Recent studies indicated that collagen-bound decorin may sequester the cytokine in the extracellular matrix (Markmann et al., 2000). As biglycan is located predominantly pericellularly, this proteoglycan would be much less suited to trap TGF-β in the extracellular matrix, and hence explain the unsuccessful treatments of fibrotic disorders with this proteoglycan. Studies in human diabetic nephropathy led to the conclusion that small proteoglycans might be able to remove the cytokine via the circulation or the urinary tract, again postulating that it is not the inactivation of the cytokine but the different localization of the TGF-β/decorin complex that may explain the beneficial effect of decorin application (Schaefer et al., 2001). Figure 3 shows the different possibilities of interaction of decorin/TGF-β complexes with extracellular matrixmolecules and/or cell surface receptors. In other studies, TGF-β/decorin complexes were not inhibiting but even of greater biological activity as the growth factor alone (Takeuchi et al., 1994; Riquelme et al., 2001). This may indicate that in given cells or in given biological constellations the proteoglycan/TGF-β complex may be more efficiently presented to the TGF-β signalling receptors than the free cytokine. The potential role of small proteoglycans in TGF-β activation has not yet been investigated.
Decorin may also exhibit therapeutic effects in inflammatory diseases, which are independent of its interaction with TGF-β. For example, apoptosis of mesangial cells, which is a hallmark of glomerular damage during inflammation, can be prevented by p27KIP1 (Ophascharoensuk et al., 1998). As decorin may induce this protein (see above), decorin administration in glomerulonephritis may also be beneficial because of the up-regulation of this cyclin-dependent protein kinase inhibitor.
Heparan sulfate proteoglycans
As mentioned above, this review does not cover the wide spectrum of growth factors/growth factor receptors, chemokines and cytokines that are able to interact with specific sequences of heparan sulfate chains. However, there are now several examples that implicate a role of the core proteins of extracellular matrix heparan sulfate proteoglycans in tissue remodelling and growth. It had been shown recently that the core protein of perlecan specifically promotes the mitogenic activity of FGF-7 (keratinocyte growth factor) and also interacts with FGF-binding protein (Mongiat et al., 2000, 2001).
Phosphacan, a chondroitin sulfate proteoglycan of nervous tissue which also represents the extracellular domain of a receptor-type tyrosine phosphatase, exhibits high affinity binding to FGF-2 (KD about 6 nM) via its core protein (Milev et al., 1998b). Since in contrast to versican and brevican phosphacan and the neonatal isoform of neurocan are up-regulated significantly in glial scar upon cerebral injury, it had been hypothesized that the latter two proteoglycans are nevertheless growth inhibitory and contribute to axonal regenerative failure in the chronic glial scar (McKeon et al., 1999). The unrelated transmembrane chondroitin sulfate proteoglycan NG2, which is expressed by immature progenitor cells in several developmental lineages, also binds through its core protein FGF-2 and additionally PDGF-AA with high affinity (Goretzki et al., 1999). Neuroglycan, too, is a transmembrane protein being predominantly expressed in the brain and carrying chondroitin sulfate chains in immature tissue. Although its role in growth and differentiation of brain is not yet precisely understood, it had been suggested that it is involved in the differential adhesion and synaptogenesis of the climbing and parallel fibers with the Purkinje cell dendrites (Aono et al., 2000). The complexity of the effects of the single proteoglycan on axonal growth has nicely been exemplified in case of agrin, which is a major component of several basement membranes and enriched in Schwann cell basal laminae. Despite the fact that mixed agrin/laminin-2 substrates inhibited neurite outgrowth, agrin may have nevertheless a supportive role in the development of axonal pathways, possibly as a binding component for growth factors (mediated by the heparan sulfate chains) and adhesive proteins (Halfter et al., 1997; Cotman et al., 1999). A stop signal is provided at the surface of the myofibrils (Chang et al., 1997).
Recent studies provided evidence that the glycosaminoglycan moiety of versican, i.e., chondroitin sulfate chains, as well as dermatan sulfate are able to bind certain chemokines, thereby negatively regulating the activation of integrins and Ca2+ mobilization in secondary lymphoid tissue and opposing the effects of heparan sulfate (Hirose et al., 2001).
Colony-stimulating factor-1 (M-CSF) represents an interesting example where the proteoglycan form of the growth factor may become immobilized in tissues via the glycosaminoglycan chain (Ohtsuki et al., 1993). Furthermore, glycosaminoglycan chain removal and further proteolytic processing of CSF-1 seem to be required to obtain full growth factor activity (Partenheimer et al., 1995). Hence, the glycosaminoglycan moiety in this particular molecule serves to mask the growth factor's activity and simultaneously to create a means for storage in the matrix.
As essential components of the extracellular matrix proteoglycans participate in all universal biological phenomena being related to differentiation, maintenance of tissue organization and organ shape. In a broad sense one can therefore propose that all alterations in the expression pattern of proteoglycans may have consequences for cell-cell- and cell-matrix-interactions and, hence, also for growth control. However, in case of proteoglycans not solely the rates of biosynthesis and turnover that can be regulated. An important aspect concerns the type, number, molecular mass, and domain structure of the glycosaminoglycan chains themselves. In spite of an extensive literature on changes of expression patterns of proteoglycan core proteins and of glycosaminoglycan structures in different tissues with regard to development, ageing and disease, unifying features have not yet emerged, and the respective problems cannot be discussed in the present review. Integrative proteomics and glycomics will eventually allow a deeper understanding of the problem how proteoglycans as tissue organizers play a role in growth control in addition to their direct action as signalling molecules and their interactions with growth factors.
Work in the authors' laboratory has been supported by the Deutsche Forschungsgemeinschaft (SFB 297, Project A7; SFB 492, Project A6) and by the Interdisciplinary Centre for Clinical Research at the Medical Faculty of the University of Münster (Project A2).