Heparan sulfate proteoglycans (HSPGs) have been shown to regulate signaling in many systems and are of increasing interest in cancer. While these are not the only sugars to drive melanoma metastasis, HSPGs play important roles in driving metastatic signaling cascades in melanoma. The ability of these proteins to modulate ligand–receptor interactions in melanoma has been quite understudied. Recent data from several groups indicate the importance of these ligands in modulating key signaling pathways including Wnt and fibroblast growth factor (FGF) signaling. In this review, we summarize the current knowledge regarding the structure and function of these proteoglycans and their role in melanoma. Understanding how HSPGs modulate signaling in melanoma could lead to new therapeutic approaches via the dampening or heightening of key signaling pathways.
Melanoma is the most aggressive form of skin cancer and is responsible for over 75% of skin cancer deaths. The molecular mechanisms underlying the progression of melanoma from a superficial tumor on the surface of the skin (radial growth phase) to a metastatic state are not well understood. The process of metastasis is a complex one, involving the coordination of signals from within the tumor and from the surrounding microenvironment. How these signals are regulated and how their regulation is governed by external cues are critical to the understanding of cancer cell metastasis.
Regulation of signal transduction occurs in many fashions and may be genetic, epigenetic, or posttranslational. One of the key regulatory mechanisms of signaling is the modulation of ligand availability, which can be mediated by heparan sulfate proteoglycans. Heparan sulfate proteoglycans (HSPGs) are glycoproteins that belong to a large family of molecules that are divided based on the structure of their glycosaminoglycan (GAG) chains. The GAG chains are the targets of many ligands and have the capacity to decrease or increase intracellular signals by controlling ligand delivery to the receptor. They also act as co-receptors and therefore, if targeted through pharmacological intervention, may provide a mechanism for modulating signaling pathway intensity. Members of this proteoglycan family include the HSPGs, chondroitin sulfate proteoglycans (CSPGs), dermatan sulfate proteoglycans (DSPGs), and the keratan sulfate proteoglycans (KSPGs). In melanoma, the CSPGs have been an area of interest for the last decade or so, owing to the discovery of melanoma chondroitin sulfate protein (MCSP). This molecule, also known as high molecular weight melanoma-associated antigen (HMW-MAA), is a potent tumor antigen and target for immunotherapy. In humans, MCSP is encoded by the CSPG4 gene and is an integral membrane protein expressed by malignant melanoma cells (Luo et al., 2006; Pluschke et al., 1996). Melanoma chondroitin sulfate protein signaling in all stages of melanoma (radial and vertical growth phase, and metastases) promotes erk1/2 activation and increases the epithelial to mesenchymal transition (EMT) in radial-growth-phase tumors. It also enhances the expression of c-met and HGF in melanoma leading to a more aggressive subset of melanoma cells (Yang et al., 2009). More recently, some very elegant experiments have shown that targeting MCSP along with cd20 in a small subset of melanoma cells (2% of total) results in tumor eradication (Schmidt et al., 2011), challenging current approaches to melanoma therapy. We refer the interested reader to an excellent review on MCSP and its role melanoma (Campoli et al., 2010). As such, chondroitin sulfates have been extensively studied, and several excellent reviews exist on these proteins, so we will not address them in this review. Conversely, very little is known about DSPGs in melanoma, and KSPGs such as lumican are just emerging as potentially significant molecules involved in melanoma cell adhesion and migration. Because more is known about HSPGs in melanoma, but they have historically been an area of less focus than CSPGs, we will focus on the role of HSPGs in melanoma. This review will explore the studies that have contributed to the understanding of HSPG modulation of signal pathways and how that understanding may contribute to providing potential therapeutic targets to treat melanoma and other types of cancer.
Heparan sulfate proteoglycans
Structure and function
Heparan sulfate proteoglycans are ubiquitously expressed. They consist of a core protein, either transmembrane or membrane-anchored, with GAG side chains attached (Figure 1A). GAG binding of extracellular ligands modulates intracellular signaling by enhancing the formation of ligand–receptor complexes (Bishop et al., 2007). Heparan sulfate proteoglycans can also regulate ligand turnover and immobilize ligands via attachment to their GAG chains. They can also regulate ligands by cell surface shedding (Figure 1B). This results in reduced ligand availability at the cell surface. Heparan sulfate proteoglycans can also sequester proteins within secretory vesicles and link extracellular matrix (ECM) proteins together (Bishop et al., 2007).
Heparan sulfate chains are targets for many effector molecules. These include the following: cell–cell adhesion molecules, protease inhibitors, growth factors and their receptors, degradative enzymes and ECM proteins (Zimmermann and David, 1999). There are two main classifications of HSPGs: cell surface HSPGs and ECM-associated HSPGs. The ECM-associated HSPGs include molecules such as multidomain perlecan (that helps maintain the endothelial barrier function), agrin (required for the formation of neuromuscular junctions) and collagen XVIII (a form of collagen that can be cleaved to form endostatin, an anti-angiogenic agent). The cell surface HSPGs include the syndecans, glypicans, and the minor cell surface HSPGs. Minor forms of membrane HSPG include β-glycan and the V-3 isoform of CD44. β-Glycan is also known as transforming growth factor beta receptor III (TGFBR3) and is primarily involved in TGF-β-signaling (Andres et al., 1989, 1992). The HSPGs that this review will focus upon are the cell surface HSPGs: the syndecan family and the glypican family. These are the HSPGs that most commonly affect downstream signal transduction through interaction with a wide variety of ligands.
The syndecans are single-pass transmembrane proteins. The first syndecan discovered was named syndecan 1 and found to be capable of binding components of the ECM to epithelial cells. Following this initial discovery, other structurally related syndecans were discovered that could bind directly to membrane phospholipids via glycosyl phosphatidyl inositol (GPI) linkage (David et al., 1990). The glypicans are linked to the cell membrane via a GPI anchor (Bernfield et al., 1999; Bishop et al., 2007). Owing to this GPI linkage, these distinct HSPGs were named glypicans and, like syndecans, other structurally related glypicans were soon discovered (Bernfield et al., 1999). Syndecans and glypicans bind ligands with high affinity (Bernfield et al., 1992), and this is based on the sequence of the glycosaminoglycan (GAG) side chains, which in turn are determined based on the biosynthesis of the chains.
The synthesis of heparan sulfate (HS) is a three-step process. First, region formation of the tetrasaccharide linker results in an area for attachment of the HS chain to the protein. Second is the generation of the polysaccharide chain, and third, the enzymatic modification of the chain resulting in saccharide sequences responsible for protein binding and structural organization (Bernfield et al., 1999; Esko and Selleck, 2002; Lindahl and Li, 2009). The occurrence of syndecans or glypicans without their GAG chains only occurs in newly synthesized molecules to which the chain has not yet been added (Carey, 1997). It is because of this that syndecans and glypicans are known as full-time proteoglycans. Part-time proteoglycans are proteins that occur both in a proteoglycan form and in a non-proteoglycan form without GAG chains. Examples of part-time proteoglycans include thrombomodulin, a mediator of endothelial anticoagulant defenses, and neuroglycan C (NGC). NGC exists in a proteoglycan form (a single CSPG GAG) in the developing cerebellum of the mouse and exists in a non-proteoglycan form in the mature cerebellum (Aono et al., 2000).
The formation of the tetrasaccharide linkage region of the chain is the initial step. The linkage region –Ser-Xyl-Gal-Gal-GlcA– is formed by xylose transfer from UDP xylose to serine-specific hydroxyl residues on the HSPGs core protein (Bernfield et al., 1999; Lindahl and Li, 2009). This linkage region is identical in both HS and chondroitin sulfate chains. The main difference between these chains is that chondroitin sulfate chains have a higher net negative charge than the HS chains, but the HS chains have greater binding affinity for most extracellular ligands (Carey, 1997).
Second, once attachment of the linkage region to the core protein has occurred, α-N-acetylglucosaminyltransferase I adds an N-acetylglucosamine (GlcNAc) moiety to the other end of the linkage region, committing the chain to HS synthesis. Chain elongation can then occur through the addition of alternating GlcNAc moieties and glucuronic acid molecules (Carey, 1997). Chain length varies according to cell type and core protein structure; however, the exact mechanism of chain elongation termination is mostly unknown. Third, once chain assembly is complete, it consists of 50–150 disaccharides, and these are then all subject to individual enzymatic modifications (Bernfield et al., 1999). Because not all of the substrate is modified, final chain structures can be quite diverse and of differing lengths. Initial modification is carried out by N-deacetylase/N-sulfotransferase replacing the GlcNAc N-acetyl group with sulfate groups on GlcNAc clusters, resulting in modified and unmodified regions of GAG chains (Aikawa and Esko, 1999; Orellana et al., 1994). Following this modification, d-glucuronic acid residues are epimerized by glucuronyl C-5 epimerase to l-iduronic acid units, and the modified disaccharides receive the bulk of the subsequent O-sulfations (Bernfield et al., 1999; Esko and Selleck, 2002; Li et al., 1997).
These modifications ensure that GAG chains from different cell types with the same core protein have structural differences, such as their O-sulfation pattern and variation in domain structure. These differences alter the ligand affinity for the GAG chain causing an increase in binding between the ligands and the sulfated domains. This increase in affinity extends the entire length of the chain, thus increasing the ligand-binding area. The proximity of GAG chains to the cell surface, expression of the GAG chains, and the GAG chain turnover rate is determined by the core proteins to which they are bound (Bernfield et al., 1999) (see Figure 1). Because of their length, GAG chains also result in a spatial increase in the range of binding area (Carey, 1997) providing increased area for ligand binding. This allows for ligand to be captured from a distance, maximizing the amount of ligand that can be bound by the HSPG.
Regulation: GAG chain modification
The synthesis and expression of GAG chains on HSPGs are regulated by a variety of enzymes, including sulfatases, glycosyltransferases, heparinases, and heparanases. In this section, we will describe these enzymes and their effects on GAG chains.
Two tumor suppressor genes, EXT1 (exostosin 1) and EXT2 (exostosin 2), are glycosyltransferases involved in HSPG biosynthesis, suggesting that modulation of the GAG chains is important in maintaining normal function of HSPGs. However, the precise role of the GAG chains in tumorigenesis remains unclear (Tumova et al., 2000). The EXT proteins are associated with HS biosynthesis, catalyzing GlcNAc and GlcA transfer to nascent chains (Zak et al., 2002). The EXT genes are mutated in the autosomal dominant disease, hereditary multiple exostoses (HME), also known as multiple hereditary exostoses. This disease is characterized by the development of osteochondromas (bony outgrowths with cartilaginous caps) at the growth plates of long bones. Most HME patients have frameshift or missense mutations in EXT1 or EXT2, resulting in truncated forms of the proteins. Even though mutations in EXT genes result in HME, the exact changes that occur in HS synthesis leading to ectopic bone growth remain unknown. However, changes in HS expression around the area of exostoses could have profound effects on chondrocyte growth and differentiation. HSPGs act as co-receptors for growth factors such as bone morphogenetic protein (BMP) and fibroblast growth factor (FGF) and affect signaling of wingless-type MMTV integration site family (Wnt) and Hedgehog (Hg) proteins. Thus, the lack of HSPG GAG chains as observed in HME could affect normal signaling, resulting in ectopic bony outgrowths (Zak et al., 2002).
It has been well established that the GAG chain sulfation pattern can influence the signaling events on the cell surface by increasing or decreasing ligand affinity for the HSPG (Fromm et al., 1997; Herndon et al., 1999; Norgard-Sumnicht and Varki, 1995; Sun et al., 1989). The endosulfatases SULF1 and SULF2 are responsible for cleaving specific 6-O-sulfate groups within the GAG chains. A recent study has demonstrated that these enzymes are overexpressed in various human cancer types and can be associated with progression and prognosis. Specifically, SULF1 expression was shown to be increased in T prolymphocytic leukemia, acute myeloid leukemia, and renal carcinoma and was associated with a poor prognosis in lung adenocarcinoma, whereas SULF 2 was overexpressed in patients presenting with high-grade uveal melanoma and in patients presenting with colorectal carcinoma. Both endosulfatases were overexpressed in brain, breast, head and neck, renal, skin, and testicular cancers (Bret et al., 2011). In addition, a recent study has demonstrated that that extracellular HSulf-1 may function as a negative regulator of proliferation and invasion through suppression of Wnt/β-catenin signaling in gastric cancer (Li et al., 2011). In addition to sulfatases, HSPGs can be cleaved by two additional enzymes: heparanase and heparinase. Obviously with all these studies demonstrating HSPG biosynthesis as a contributory factor to cancer progression, significant investigation into the effects of heparinases has also been carried out [reviewed in (Arvatz et al., 2011)].
Heparanase and heparinase
Heparanase is an enzyme that acts both at the cell surface and within the ECM. Its main function is to degrade polymeric HS molecules into shorter chain length oligosaccharides (Hulett et al., 1999; Vlodavsky et al., 1999). The similarly named heparinase, a heparin lyase, is an enzyme that catalyzes the chemical reaction that causes eliminative cleavage of polysaccharides containing 1,4-linked d-glucuronate or l-iduronate residues and 1,4-alpha-linked 2-sulfoamino-2-deoxy-6-sulfo-d-glucose residues, thus cleaving off the GAG chains (Hovingh and Linker, 1970). Both heparanase and heparinase can affect tumor growth and invasion. Treating cancer cells with heparanase-1, which cleaves heparin-like regions, results in an increase in both tumor growth and metastatic dissemination (Roy and Marchetti, 2009). It has been demonstrated that heparanase shows selective localization in metastatic melanoma and plays an important role in regulating in vivo growth and progression, further emphasizing that therapies designed to block heparanase activity may aid in controlling this type of cancer (Murry et al., 2005). Treatment of cancer cells with heparinase, however, seems to have an opposing effect. Treating myeloma cells with heparinase III, which more specifically cleaves HSPGs (i.e., unsulfated d-glucorinic acid, heparan-sulfate-like regions), inhibits their metastatic capacity (Yang et al., 2007). We have also shown that treating melanoma cells with heparinase III inhibits the delivery of Wnt5A ligand to its receptor, by cleaving syndecan 1 and 4 backbones. This results in a decrease in the invasive capacity of melanoma cells (O’Connell et al., 2009b). These data highlight the importance of the syndecans and their regulatory enzymes in regulating signals leading to invasiveness in melanoma.
Syndecans can be grouped into subfamilies according to their transmembrane domain and the cytoplasmic domain (Carey, 1997; Zimmermann and David, 1999). Syndecans 2 and 4 form one subfamily, and syndecans 1 and 3 are part of a second subfamily. Membrane microdomain localization is dictated by the transmembrane domain of syndecans. This has been demonstrated in both syndecan 1 and syndecan 4, members of different subfamilies, where syndecan 1 polarizes to the basolateral epithelial cell surface and syndecan 4 localizes to focal adhesions (Bernfield et al., 1999; Hayashi et al., 1987; Rapraeger et al., 1986; Woods and Couchman, 1994). This is important as the localization of the HSPG can have significant effects on morphogen/growth factor gradient formation and can ultimately contribute to attenuating downstream signaling.
Structure of syndecans
The HSPG syndecans consist of GAG chains, a core protein containing a transmembrane domain and a cytoplasmic domain (which contains subdomains, C1, C2, and V). The syndecans are type 1 transmembrane HSPGs, consisting of four members in vertebrates. Syndecan 1 is found mainly in epithelial cells, syndecan 2 in fibroblasts, and syndecan 3 in neuronal cells, and syndecan 4 has widespread expression (Zimmermann and David, 1999). Sites for shedding and surface binding have been identified on the extracellular domain (Rapraeger, 2001). The extracellular domain also contains a region for GAG chain attachment close to the N-terminus consisting of consecutive serine–glycine sequences (2–3) that are surrounded by acidic and hydrophobic residues. All syndecans can undergo proteolytic cleavage from the cell surface, and the exogenous cleavage of syndecans by thrombin or plasmin occurs at a mono or dibasic sites, within five amino acid residues adjacent to the membrane (Bernfield et al., 1999).
The cytoplasmic region of syndecans contains three major domains denoted as C1, C2, and V. The C1 and C2 domains are two highly conserved regions that occur proximal and distal, respectively, to the membrane and are common to all syndecans. The C1 region is rich in basic residues, and in the syndecan 4 core protein, phosphorylation of the invariant serines of the C1 region occurs. The homology between the C1 regions of syndecans and other single-pass transmembrane proteins suggests that this region is involved in downstream intracellular signaling. These C regions are separated by a V or variable region, which is unique to each individual syndecan (Zimmermann and David, 1999) but has remarkable homology across species. This homology provides a unique opportunity for in vivo studies and potential therapy. Targeting the V region of syndecan 1 using a compound-based approach (such as a chemical inhibitor or blocking antibody) in a mouse model may demonstrate effects that are comparable in humans.
Functions of the syndecans
The types of syndecans and their roles differ depending on location and the variety of effector molecules present. Ligand binding of syndecans is specific; however, a large number of molecules bind to syndecans, including growth factors and cell–cell adhesion molecules (Carey, 1997).
Syndecan 1 is involved in microfilament cross talk and cell behavior. In mouse mammary epithelial cells, syndecan 1 co-localizes at the basolateral surface, with intracellular microfilaments, and this function is linked to the V region of syndecan 1 (Bernfield et al., 1999). Syndecans 1 and 4 are shed into human acute dermal wound fluid where they modify the proteolytic activity and balance of the fluid. The ectodomain of purified syndecan 1 prevents the inhibition of cathepsin G, present in the wound fluid, by α1-antichymotrypsin and the inhibition of elastase by α1-proteinase inhibitor (Kainulainen et al., 1998). Cathepsin G is a neutral serine proteinase that has been shown to inhibit the human tissue factor pathway inhibitor-2 (TFPI-2). TFPI-2, in turn, plays a significant role in reducing the invasive behavior of human amelanotic melanomas (Konduri et al., 2000). Thus, if syndecan 1 can also prevent the inhibition of cathepsin G, it also may promote increased invasion in melanoma cells. Indeed, we have shown that syndecan 1 is upregulated in highly metastatic melanoma cells, and knockdown of syndecan 1 inhibits invasion (O’Connell et al., 2009b).
Syndecan 2 is involved in granular macrophage colony-stimulating factor (GM-CSF) signaling. GAG chains are bound by GM-CSF, resulting in signaling through α- and β-chains of a cytokine receptor. This activates kinases such as JAK2 but also results in receptor binding to syndecan 2, leading to tyrosine phosphorylation of the cytoplasmic domain. The impact of this syndecan phosphorylation remains unknown, but the formation of the receptor–ligand–syndecan complexes suggests a role for syndecan core proteins in the signaling pathway (Rapraeger, 2001). Syndecan 2 has also been shown to play a role in morphogenesis by activating PKA via neurofibromin and PKA consequently phosphorylating Ena/VASP, promoting filopodia and spine formation in human embryonic kidney cells and hippocampal neurons (Lin et al., 2007). The syndecan 2 extracellular domain has recently been shown to be a novel ligand for the protein tyrosine phosphatase receptor CD148, and this study demonstrated a novel pathway for β1-integrin-mediated adhesion in cellular processes such as angiogenesis and inflammation (Whiteford et al., 2011). The cytoplasmic domain of syndecan-2 has been shown to regulate colon cancer cell migration via interaction with syntenin-1 (Lee et al., 2011). Syntenin-1 is a member of the tetraspanins that are clustered in specific microdomains in the cell membrane. They are also known as tetraspanin-enriched microdomains, and they regulate the functions of associated transmembrane receptors (Latysheva et al., 2006).
Syndecan 3 (also known as N-syndecan) is the largest syndecan molecule. It has a mass of 43 kDa consisting of a 405 amino acid core protein, a 33 amino acid cytoplasmic domain, and a 25 amino acid transmembrane domain. The extracellular domain is 347 amino acids with a secretory signal sequence domain at the N-terminus. HS GAG chains are potential sites for extracellular ligand interactions and the T/S/P (threonine, serine, and proline) domain, which is present only on syndecan 3, is a potential functional domain for cell–cell interactions. The T/S/P domain resembles domains present in mucin-like proteins that have O-oligosaccharides linked to the T and S residues (Kosher, 1998).
Syndecan 3 is highly expressed during the early stages of limb development. Syndecan 3 is also highly expressed during precartilage condensation of the skeletal elements, suggesting a mediatory role for syndecan 3 in cell–matrix and cell–cell interactions during the onset of chondrogenesis (Kosher, 1998). Syndecan 3 gene expression is associated with proliferating chondrocytes (Kirsch et al., 2002). It is possible that syndecan 3 plays a role in the modulation of heparin binding to signaling molecules involved in mediating the functions of the perichondrium. Signaling molecules with heparin-binding domains at their N-terminus include BMP-2, BMP-4, and BMP-7, suggesting that HSPGs at the cell surface and in the ECM regulate BMP activity during skeletogenesis and chondrogenesis (Kosher, 1998).
Syndecan 4 is integrated into focal adhesions that contain integrins, suggesting a role in ECM adhesion signaling. The V cytoplasmic region of syndecan 4 undergoes dimerization with PIP2 (phosphatidylinositol-4,5-bisphosphate), which allows the exposed side of the dimer to bind to the catalytic region of protein kinase Cα (Rapraeger, 2001). Recently, it has been shown that syndecan 4 signaling inhibits apoptosis and controls NFAT activity during myocardial damage and remodeling. Additional studies have shown that syndecan 4 signaling plays an important role in the inflammatory response and granulation tissue formation (Matsui et al., 2011). Syndecan 4 has also been shown to play a role in bone pathophysiology. Syndecan 4 controls osteoarthritis progression and cartilage breakdown by controlling the activation of ADAMTS-5 through direct interaction with the protease (Echtermeyer et al., 2009). Additional control is achieved through regulating mitogen-activated protein kinase (MAPK)-dependent synthesis of matrix metalloproteinase-3 (MMP-3). In addition, syndecan-4, along with syndecan-2, plays a role in osteoblast cell adhesion and survival mediated by a tissue transglutaminase–fibronectin complex known as TG2 (Wang et al., 2011b).
The glypicans are a family of C-terminal GPI-anchored cell surface HSPGs. They control growth and development by modulating ligand–receptor interactions at the cell surface (Fransson, 2003). This interaction also occurs via GAG chains. The major differences between glypicans and syndecans are that glypicans do not have a transmembrane domain and the GAG chains of glypicans reside much closer to the cell surface.
Structure of glypicans
Glypicans have 14 conserved cysteine residues located mainly near the N-terminus, with some in the central domain and GAG chains are attached to the cysteine residues located in the central domain. All glypicans are 60–70 kDa in size, and removal of the central domain results in a substitution of heparin sulfate to chondroitin sulfate (Chen and Lander, 2001; Fransson, 2003). There are six glypican HSPGs identified to date in mammals denoted as glypican 1–6, two glypicans in Drosophila known as Dally forms and one in zebrafish called Knypek (Esko and Selleck, 2002).
The N-terminal sequence of glypicans is responsible for translocation to the endoplasmic reticulum, and the C-terminal sequence is involved in initial membrane attachment and the subsequent GPI anchorage (Fransson, 2003). The GPI linkage area of the glypicans generally associates with ordered regions of the membrane. The GPI anchors mediate turnover of cell surface components by rapid endocytosis and degradation via lysosomes. However, in fibroblasts and endothelial cells, once internalization has occurred, the proteoglycan and ligand can be recycled back to the cell surface (Bernfield et al., 1999).
Functions of the glypicans
The underlying mechanisms of glypican function remain unknown. Glypican GAG chains can bind collagen, antithrombin, fibronectin, and FGF in vitro depending on which cell type the glypican resides on. There is also evidence that the glypicans can bind to the Wnt proteins and have effects on both canonical and non-canonical Wnt signaling (Filmus et al., 2008; Ho and Kim, 2011; Song et al., 2005; Wu et al., 2010).
Glypicans have large globular heads present above the GPI anchor that may prevent big or insoluble molecules from accessing the GAG chains, but allow for the passage of small molecules. This trafficking of molecules suggests that glypicans are capable of self-regulation (De Cat and David, 2001). It has been further suggested that glypicans are orphan receptors, as they have no ‘canonical’ ligand, for dual docking ligands in which the GAG chains provide a molecular rope between ligand and receptor. Because glypicans have no direct interaction with the intracellular environment, signaling must be performed by transmembrane molecules associated with glypicans (De Cat and David, 2001; Simons and Toomre, 2000). It has been proposed that this occurs via lipid raft clustering, resulting in endocytosis, but the exact mechanism(s) by which glypicans signal remains unknown.
Syndecans versus glypicans
The fact that there are two distinct families of cell surface HSPGs implies that they carry out different functions. However, the fact that the biosynthesis of the GAG chains is the same for both syndecans and glypicans suggests that the functions are at least related. Glypicans have cysteine-rich ectodomains, which likely form globular proteins owing to disulfide bond formation between cysteine residues. In contrast, syndecans have ectodomains rich in proline, suggesting a more extended structure (Bernfield et al., 1999). The HSPG ectodomains undergo cell surface shedding by proteolytic cleavage of the core protein. The shed syndecan ectodomains are present in body fluids, and these shed proteins have physiological roles. Glypican ectodomains are found in cell culture media, but the physiological role of glypican shedding remains unclear (Bernfield et al., 1999; Kato et al., 1998; Penc et al., 1999; Subramanian et al., 1997).
The association of the GPI link of glypicans and the transmembrane domain of syndecans with the cell surface results in different roles for each HSPG on the same cell. Glypican-associated internalization results in both ligand and receptor recycling, whereas syndecan internalization is thought to result in lysosomal degradation of both HSPGs and bound ligand. However, when the HSPG is degraded or recycled, it does not necessarily mean the entire complex of ligand, receptor, and HSPG follow the same route. If the ligand is released from the GAG chain prior to internalization, the complexes may follow different internalization pathways. Similarly, signaling differences occur between the two HSPGs. Syndecans contain a cytoplasmic domain, resulting in association with signaling molecules, whereas glypicans require association proteins to facilitate signaling (Bernfield et al., 1999).
HSPGs and cancer
Heparan sulfate proteoglycans play critical roles in the pathogenesis of various diseases. For example, faulty initiation of HSPG biosynthesis has also been linked to cardiac defects including heart malformations, including mitral valve prolapse, ventricular septal defect, and bicuspid aortic valve (Baasanjav et al., 2011). Heparan sulfate proteoglycans have also been implicated in diabetes (Chen et al., 2010; Wang et al., 2011a), and interestingly, in diabetic patients, alterations in HSPGs ultimately have effects on heart and skeletal muscles (Strunz et al., 2011), perhaps linking these two observations. One disease that has extensive HSPG involvement is cancer, and for the purpose of this review, we will focus on the roles of HSPGs in cancer and specifically melanoma.
Previous studies have suggested that HSPGs be used as cancer biomarkers. Glypican 3 has been shown to be a diagnostic and prognostic marker and also suggested to be an ideal target for therapy of hepatocellular carcinoma (Honsova et al., 2011; Zou et al., 2010). Glypican 3 is upregulated in human hepatocellular carcinoma (Suzuki et al., 2010) under the control of sulfatase 2, which also protects the cells from apoptosis (Lai et al., 2008, 2010). Glypican 1 is present on pancreatic cancer cells and surrounding fibroblasts, but not present on normal pancreatic cells (Kleeff et al., 1998). The growth of cancer cells may be regulated by glypican 1 binding to growth factors such as fibroblast growth factor 2 (FGF 2). Recent studies have shown that glypican 5 binds to both Hedgehog (Hg) and Patched 1 (Ptc1) through its GAG chains. This interaction ultimately inhibited Hg signaling through glypican 5 competing with Ptc1 for Hg binding. Interestingly, the GAG chains of glypicans 5 displayed a significantly higher degree of sulfation than those of glypican 3, and more importantly, glypican 3 was shown to bind to Hh through its core protein but not its GAG chain (Capurro et al., 2008; Li et al., 2011). Glypican 3 has been reported to be overexpressed in melanoma and may be a useful biomarker for the diagnosis of melanoma. It has been suggested that glypican 3 may also have a role in immunotherapy or tumor prevention based on a study showing GPC3 was highly immunogenic in mice and elicited effective anti-tumor immunity along with no evidence of autoimmunity (Nakatsura and Nishimura, 2005). However, recent studies have suggested that glypican 3 is not expressed in melanoma (Kandil et al., 2009), and others have claimed expression in only about 40% of early stage melanomas (Ikuta et al., 2005). The latter study suggested glypican 3 was a useful biomarker if used in combination with secreted protein, acidic and rich in cysteine (SPARC).
Syndecan 1 is one of the better-studied HSPGs in cancer. This is because of the ability of syndecan 1, along with E-cadherin and integrins, to control cell differentiation. Syndecan 1 expression has been shown to be a prognostic marker of human endometrial cancer. Loss of epithelial and induction of stromal syndecan-1 expression may be associated with tumor progression, and stromal syndecan-1 expression can serve as an indicator of poor prognosis (Hasengaowa et al., 2005). In addition, the shedding of syndecan 1 by stromal fibroblasts stimulates human breast cancer cell proliferation. This was shown to be via FGF2 activation (Su et al., 2007). The ectodomain of syndecan 1 has been shown to regulate matrix-dependent signaling in human breast carcinoma (Burbach et al., 2004). Relatively, little is known about the role of syndecans 2 and 3 in cancer. Syndecan 2 has been shown to be mainly associated with the formation of osteochondromas and chondrosarcomas (Birch and Skerry, 1999; Hameetman et al., 2007), and it has been shown that high canonical Wnt signaling can repress syndecan-2 in osteosarcoma cells (Dieudonne et al., 2010). In addition, syndecan 2 has been shown to suppress MMP-2 activation, which subsequently causes the suppression of metastasis in Lewis lung carcinoma 3LL (Munesue et al., 2007). Syndecan 2 also has some association with melanoma pathology (see below).
Syndecan 4, in addition to its role in melanoma progression (discussed below), has been implicated as an independent indicator of breast carcinoma (Lendorf et al., 2011). Syndecan 4 is also involved in chemokine-mediated migration, as syndecan 4 can regulate SDF-1/CXCL12-mediated invasion of human epithelioid carcinoma (Brule et al., 2009). Syndecans 1 and 4 are involved in RANTES/CCL5-induced migration and invasion of human hepatoma cells (Charni et al., 2009).
Growth factor–dependent signaling modulated by HSPGs has been implicated in primary tumor growth and angiogenesis (Bishop et al., 2007). For example, HSPGs are critical for endostatin binding and are low affinity receptors for these potent anti-angiogenic proteins (Karumanchi et al., 2001). On binding to α5β1 and an HSPG co-receptor, there is an association between endostatin and caveolin-enriched lipid rafts. This induces the Src pathway ultimately inactivating RhoA leading to the disassembly of actin stress fibers and focal adhesions. This results in altered endothelial locomotion and capillary morphogenesis, ultimately causing angiostasis (Iozzo, 2005). HSPGs have also been shown to play a role in the immune recognition of tumor cells. NKp44, a natural cytotoxicity receptor on human NK cells, can bind to HSPGs (Hershkovitz et al., 2007). In this study, various cancer cell lines were used, including human prostate adenocarcinoma, melanoma, and cervical adenocarcinoma. NKp44 showed HS-dependent binding to tumor cells that could be blocked with an antibody to HS. Binding of NKp44 to heparin was observed, and soluble heparin/HS enhanced the secretion of IFN-γ, a marker of T-cell activation. The putative heparin/HS-binding site of NKp44 was mutated, and tumor cell recognition of the mutated NKp44 proteins was significantly reduced. This study is of note as it provides evidence that HSPGs can be used to facilitate tumor cell identification by immune cells. This in turn may have implications for immunotherapy.
HSPG modulation of melanoma metastasis
Heparan sulfate proteoglycans have been significantly understudied in melanoma yet what is known clearly demonstrates their importance (see Table 1). A study using melanoma cells illustrated the importance of HSPG GAG chains as promoters or inhibitors of tumor growth and metastasis. It was shown that growth promoting and growth inhibiting sequences are contained within the GAG chains and that the dynamic balance between these distinct polysaccharide populations regulates specific intracellular signal transduction pathways (Liu et al., 2002). For example, it has been shown in B16F10 cells that the receptor for a synthetic laminin alpha-1 chain c-terminal peptide, a metastasis-promoting sequence in laminin, is a HSPG, and that the GAG chains on that HSPG are functionally important in the cell–peptide interaction. Injection of the synthetic laminin peptide in the tail veins of C57BL6/Nsd mice resulted in an increase in lung tumors and the formation of liver tumors in mice (Engbring et al., 2002).
Many different proteins have been shown to contribute to protein–HSPG interaction in different ways. The protein Epac1, an effector molecule of cAMP, has been demonstrated to modulate melanoma cell metastasis both through activation or overexpression of Epac1 and by causing the translocation of syndecan 2 to lipid rafts, modulated through the PIK3 pathway. Epac1 can also modulate the production of HS. Specifically, Epac1 induces HS production by increasing the expression and translation rate (determined by a [35S] methionine pulse labeling assay) of N-deacetylase/N-sulfotransferase-1 (NDST-1) possibly through the PI3K/Akt/GSK3β-pathway (Baljinnyam et al., 2009). Pleiotrophin (PTN) is an ECM heparin–binding growth association molecule that binds to HSPG GAG chains. Syndecan 3, as well as PTN, is present at high levels in neuronal cells and both play a role in neurite outgrowth. PTN is also expressed in the bone matrix where it is believed to promote osteoblast recruitment and migration through interactions with syndecan 3 (Rapraeger, 2001).
HSPGs and Wnt signaling
Pleiotrophin has been shown to be associated with the Wnt/β-catenin pathway, with PTN contributing to the stabilization of β-catenin (Weng and Liu, 2010). This suggests that PTN may be a contributing factor during the early stages of melanoma development, where Wnt/β-catenin signaling is increased. Pleiotrophin has also been shown to correlate with tumor progression and metastatic potential in melanoma (Seykora et al., 2003; Wu et al., 2005). Wnt proteins are also capable of binding directly to HSPGs and can regulate their expression. In dorsal mesodermal cells from Xenopus gastrulae, Wnt5A induces ubiquitination and degradation of syndecan 4 (Carvallo et al., 2010). Non-canonical Wnt signaling was shown to promote monoubiquitination of the variable region of SDC4 cytoplasmic domain, and mutation of these specific residues abrogates ubiquitination and results in increased SDC4 steady-state levels. The binding of Wnt5A to HSPGs has been shown to be through Wnt5A–GAG chain interactions. A recent bioactive screen further demonstrated the importance of GAG chain structure in melanoma metastasis. Low molecular weight heparins (LMWH) demonstrated that in addition to having varying affinities toward angiogenic growth factors (FGF2, VEGF, and SDF-1α), the modulation of specific sequences, especially the N-domains [-NS or NH(2) or NHCOCH(3)], has a major impact on the participation in a diverse range of biological activities and that the 6-O-sulfates critically affect the binding of a desulfated derivative to certain angiogenic proteins and its ability to inhibit P-selectin-mediated B16F10 melanoma metastases (Roy and Marchetti, 2009).
Recent studies have demonstrated a relationship between Wnt signaling and the progression of malignancy in melanoma. Signaling via the canonical pathway, involving β-catenin, is associated with a less metastatic phenotype (Chien et al., 2009; Hoek et al., 2008), while activation of the non-canonical Wnt signaling pathway, involving Wnt5A, is associated with increased malignancy (Weeraratna et al., 2002). Wnt5A signaling leads to increases in metastasis caused by the upregulation of metastatic markers, such as CD44 and snail, and the promotion of an EMT (Dissanayake et al., 2007). In addition, signaling via Wnt5A can suppress the expression of melanoma antigens such as MART1, GP100, and DCT, effectively decreasing the immunogenicity of the melanoma cells. Studies indicate that Wnt5A binds to the receptor ROR2, which is then internalized via clathrin, and ultimately activates signaling via PKC and Ca2+ (O’Connell et al., 2010). This signaling leads to several additional events, such as loss of metastasis suppressors (kiss-1), changes in the cytoskeleton (filamin cleavage and actin re-organization), and the downregulation of melanoma differentiation antigens (Dissanayake and Weeraratna, 2008; O’Connell et al., 2009a). Importantly, the expression of HSPGs, specifically syndecans 1 and 4, is critical for the presentation of Wnt5A to its receptor, ROR2, and its subsequent internalization and signaling (O’Connell et al., 2010). Syndecan 1 and syndecan 4 expression correlates with Wnt5A expression and melanoma malignancy, and knockdown of syndecan 1 or 4 causes a decrease in cell invasion. Supporting this, when the GAG chains were cleaved from syndecans using heparinase, more Wnt5A was detected in the cell media, less at the cell surface, and thus, the metastatic potential of the cells decreased.
HSPGs and FGFs in melanoma
A significant number of studies have been performed on the FGF family of proteins and their effects on HSPGs. Syndecan 2 has been implicated in melanoma progression with its expression correlating with metastatic potential and increasing in the presence of FGF2, (Lee et al., 2009). Interestingly, FGF2 has been shown to decrease expression of syndecan 4, an effect that correlated with increased motility in melanoma (Chalkiadaki et al., 2009). This is in contradiction to data from our and other laboratories. This implies that HSPGs effects on the progression of melanoma are dependent on the type of ligand available and the expression of the core protein subtype. It has also been shown that chondroitin/dermatan sulfate-containing proteoglycans, likely in cooperation with HS, participate in metastatic melanoma cell FGF-2-induced mitogenic response, which establishes a central role of sulfated GAGs on melanoma growth (Nikitovic et al., 2008). These responses can be altered through GAG chain removal by heparanase. Extensive heparanase degradation of human metastatic melanoma cells inhibited FGF2 binding and stimulated extracellular signal-related kinase (ERK) and focal adhesion kinase (FAK) phosphorylation. Untreated cells did not demonstrate increased levels of phosphorylated (ERK) or FAK in response to FGF2. The presence of soluble heparanase–degraded HS enhanced FGF2 binding and ERK phosphorylation at low HS concentrations and inhibited it at higher concentrations. Importantly, cell exposure to heparanase modulated FGF2 induction of angiogenesis in melanoma. This study suggests a relevant mechanism for heparanase modulation of melanoma growth factor responsiveness and tumorigenicity (Reiland et al., 2006).
The role of HSPGs in modulating both Wnt signaling and FGF signaling melanoma has been linked through their interaction with the aging-associated protein Klotho. Klotho protein exists in two forms: a transmembrane form of Klotho that acts as a cofactor for FGF23 (Kurosu et al., 2006) and a secreted form that has a putative sialidase activity and is involved in the regulation of glycoprotein function at the cell surface (Cha et al., 2008). Klotho has been shown to affect the glycosylation status of glycoproteins by modifying N-linked glycans (Chang et al., 2005) and has both glucuronidase (Chang et al., 2005; Tohyama et al., 2004) and sialidase activity (Cha et al., 2008), which in turn have been shown to affect the glycosylation and expression of HSPGs. Recent studies have implicated Klotho sialidase activity in reducing syndecan-1- and syndecan-4-mediated Wnt5A signaling in melanoma cells (Camilli et al., 2011). This study demonstrated that Wnt5A and Klotho are inversely correlated and, in the presence of recombinant Klotho, Wnt5A internalization and signaling are decreased in high Wnt5A-expressing cells. These effects coincide with an increase in sialidase activity and decrease in syndecan expression, an effect that can be inhibited by using a sialidase inhibitor prior to Klotho treatment. These data suggest that Klotho can inhibit the internalization of Wnt5A and subsequent signaling via the de-sialidation of syndecans.
Heparinase, heparanase, and melanoma
A study in 2004 further demonstrated the importance of heparinase modulating HSPG expression in melanoma. It was shown that heparinase cleaves HS present on the cell surface of metastatic melanoma and specifically degrades the GAG chains of purified syndecan-1. In addition, syndecan-1 did not directly inhibit heparinase enzymatic activity, and the presence of exogenous syndecan-1 inhibited heparinase-mediated invasion. This study demonstrated that syndecan-1 is a degradative target of heparinase. This suggests that expression of syndecan-1 on the melanoma cell surface and its degradation by heparinase are important determinants in the control of tumor cell invasion and metastasis (Reiland et al., 2004). Although further study is needed, it is plausible that cleaving bioactive HS from the surface of one group of cells may result in delivery of bound ligand (already attached to the HS) to another set of cells nearby, thus increasing metastatic potential of the latter through an increasing concentration gradient. This is not unprecedented as one of the main functions of HSPGs is the formation of morphogen gradients (Yan and Lin, 2009).
Recent studies identified the small GTPase GEF-H1 (guanine nucleotide exchange factor-H1) as a new component of a syndecan signaling complex that is differentially expressed in brain metastases of melanoma as compared to corresponding non-metastatic primary tumors. Knockdown of GEF-H1, SDC1, and SDC4 or treatment with exogenous heparanase decreased BMM cell invasiveness and GEF-H1 modulated small GTPase activity of Rac1 and RhoA in conjunction with heparanase treatment (Ridgway et al., 2010). However, additional studies have suggested that heparanase promotes melanoma cell migration and angiogenesis by releasing bioactive cell surface HS fragments which can stimulate melanoma cell migration in vitro and angiogenesis in vivo (Roy and Marchetti, 2009). This demonstrates how fine tuning the GAG chains with both heparinase and heparanase can lead to opposing effects in melanoma progression. These studies underscore the importance of understanding both the polysaccharide makeup of the cell and its protein complement and provide a framework for the development of polysaccharide-based anticancer molecules.
Therapeutic targeting of HSPGs
As HSPGs are not necessarily associated with any one particular signaling pathway, they present a unique potential target for therapy. The blocking of a single signal transduction pathway, if successful, ultimately leads to a complete abrogation of signaling through that receptor, whereas targeting proteoglycans provides an ability to attenuate signal intensity without completely shutting down the pathway. This would likely decrease toxicity to normal cells and be particularly useful for diseases that are associated with increases in receptor activity or expression, rather than mutation. However, the major problem with targeting proteoglycans is that it is not usually the protein core that modulates ligand delivery to signal pathway receptors, but the GAG chains. As a result, current HSPG therapeutic approaches include attempting to target key enzyme residues in the GAG biosynthesis pathway to pave the way for GAG therapeutics (Talhaoui et al., 2010). It has been demonstrated that the requirement for a tetrasaccharide linker (see Figure 1) polypeptide as a starting block to build a GAG chain can be bypassed in vitro and in vivo by xyloside derivatives (Lugemwa et al., 1996). Xylosides serve as substrates for β4GalT7, promoting initiation and polymerization of GAG chains, so targeting these molecules will prevent chain initiation. Such GAG precursors have demonstrated potential as antithrombotic drugs (Martin et al., 1996) and anti-amyloid agents (Kisilevsky et al., 2004). In addition, inhibitors of GAG synthesis have been proposed as chemotherapeutic agents (Belting et al., 2002; Mani et al., 1998; Schuksz et al., 2008).
Currently available compounds used to block GAGs include HMW cationic proteins or polypeptides rich in lysine and arginine residues (Wang and Rabenstein, 2006). Only a handful of synthetic heparin-binding molecules have been described and identified. In a study in 2001, 26 compounds were determined using HPLC-purified products, and 20 of the 26 showed significant inhibition of heparin binding to VEGF and/or bFGF. Eighteen of the 20 confirmed active compounds have a linear extended structure (Zhang et al., 2001). This study demonstrated an initial relationship between structure and activity, thus providing direction for further investigation of this type of heparin mimetic libraries. In addition, the design and evaluation of peptide foldamers (polymers with a tendency to adopt a specific conformation) have been described as a route of investigation (Choi et al., 2005). Synthetic heparin receptors based on phenyl boronates (a stable matrix with affinity to β-lactamase, Igase, and peroxidase) have shown a marked increase in affinity for heparin (Wright et al., 2005). Polycationic calixarenes have been shown to be able to recognize and neutralize heparin (Mecca et al., 2006). A calixarene is a cyclic oligomer based on a hydroxyalkylation product of a phenol and an aldehyde (Gutsche 1989), which have hydrophobic cavities that can hold smaller molecules. Other avenues have included positively charged polyamidoamine dendrimers, a class of non-viral vectors. Previous studies have shown that these dendrimers can form a complex with enoxaparin, a LMWH, providing a mechanism for pulmonary delivery of enoxaparin that is effective in preventing deep vein thrombosis (Bai et al., 2007). In addition, relatively HMW dyes such as methylene blue have been used to neutralize heparin and decrease bleeding owing to heparinization (Sloand et al., 1989). In a recent study, the simplest substrates of heparanase known to date were synthesized; hydrolysis by heparinase of the N-sulfated 4-nitrophenyl glycosyl glucuronide 24 and the N-sulfated 4-methylumbelliferyl glycosyl glucuronide 26 was detected (Pearson et al., 2011). Whether or not these substrates will present any form of potential therapy for cancer remains to be determined, but the authors suggest that glycosyl glucuronide 26 should serve as a useful template in the development of a rapid diagnostic test for heparanase activity.
One small molecule antagonist of HS, surfen (bis-2-methyl-4-amino-quinolyl-6-carbamide) has been studied for its potential as a therapeutic agent. Surfen binds to GAGs and neutralizes the anticoagulant activity of both unfractionated and LMWH and enzymes that act on heparin and HS as well as activities associated with HS (Schuksz et al., 2008). Unfortunately, it has a relatively high IC50 but provides a good starting point for the design and synthesis of more potent analogs, and it is tempting to suggest that surfen or its derivatives also could be used therapeutically for treating disorders exacerbated by unwanted HS interactions. These include chronic inflammation, infection, and tumor growth and angiogenesis (Fuster et al., 2007).
The engineered HS mimetic, M402, has shown some promise in inhibiting metastasis in melanoma. M402 is non-cytotoxic and designed to inhibit multiple factors implicated in tumor–host cell interactions, including VEGF, FGF2, SDF-1α, P-selectin, and heparanase. A single subcutaneous dose of M402 selectively accumulated in the primary tumor, decreased microvessel density, and effectively inhibited seeding of B16F10 murine melanoma cells to the lung in an experimental metastasis model (Zhou et al., 2011).
As the field of HSPG biochemistry moves forwards, so does the hope of finding an effective inhibitor of signal transduction pathways in cancer. These molecules are of the utmost importance in modulating signal transduction but have been significantly overlooked as targets for cancer therapy. This may be due to their involvement in such a vast array of signaling networks, their complexity, their multiple sites of interaction (protein and GAG chain), or a combination of these factors. Either way, more and more evidence is pointing to these molecules as modulators of signaling and as potential target molecules for disease therapy. In melanoma, HSPGs have been shown to be play a role in almost every process involved in metastasis, ranging from cell proliferation, to motility and invasion, to interactions with the tumor microenvironment. This makes the study of these molecules critical in the generation of a deeper understanding of melanoma metastasis and for opening up new avenues for melanoma therapy.