Shed proteoglycans in tumor stroma

Cancer cell behavior is not only governed by tumor cell-autonomous properties but also by the surrounding tumor stroma. Cancer-associated fibroblasts, blood vessels, immune cells and the extracellular matrix of the tumor microenvironment have a profound influence on tumor progression. Proteoglycans control various normal and pathological processes, modulating cell proliferation and motility, cell-matrix interactions, immune cell recruitment and angiogenesis. They are major mediators of cancer cell behavior though a dynamic interplay with extracellular matrix components. During cancer progression, their altered expression can promote the activation of several signaling cascades regulating crucial functional properties of cancer cells. Notably, the function of cell surface proteoglycans can be altered by ectodomain shedding, which converts membrane-bound coreceptors into soluble paracrine effector molecules. In this review, we highlight the importance of proteoglycans and their soluble counterparts in cancer progression and the consequences of their interactions with the adjacent stroma. The dynamic interplay among shed proteoglycans and proteolytic enzymes has a significant impact both on tumor cells and their surrounding stroma, with important implications for the diagnosis of this disease and for novel therapeutic approaches. Graphical Abstract Syndecan shedding. The mechanism of shedding involves the proteolytic cleavage of their ectodomain near the plasma membrane by metzincin enzymes, such as metalloproteinases. N-acetylglucosamine-alpha-L-iduronic acid/beta-D-glucuronic acid (HS) chains can be additionally cleaved by heparanase. Syndecan core protein can be further processed by intramembrane enzymatic cleavage. Syndecans are in a dynamic interplay with the extracellular matrix and several receptor-tyrosine-kinases (RTKs) and various growth factors, for which they act as co-receptors, thus mediating numerous signaling pathways

The tumor stroma-a permissive environment for cancer progression Tumorigenesis and tumor progression are complex events that occur when physiological control mechanisms within a given cell are inactivated as a result of a series of gene mutations (Vogelstein and Kinzler 1993). Indeed, in their highly influential conceptual review BHallmarks of cancer^, Hanahan and Weinberg (2000) included the properties of evasion of apoptosis, self-sufficiency in growth signals, insensitivity to antigrowth signals and unlimited proliferative potential as selfautonomous properties that enable and promote the malignant growth of tumor cells. However, even this initial concept makes clear that cancer progression not only depends on the tumor cell itself but also on its microenvironment. For example, tumor cells interact with their environment to stimulate the process of angiogenesis, the outgrowth of blood vessels from the existing vasculature, in order to be supplied with nutrients and oxygen and they need to interact with and overcome their extracellular matrix (ECM) environment in order to move to distant sites in the process of metastasis (Hanahan and Weinberg 2000). In an update of their concept, the authors stress the importance of tumor cell interactions with stroma cells and with immune cells of the tumor microenvironment, as the tumor has to evade destruction by the immune system and may even influence immune cells in a way that promotes tumor growth (Hanahan and Weinberg 2011). Whereas these publications nicely delineate key cell biological events governing tumorigenesis and cancer progression, the idea of a permissive environment that promotes tumor growth actually has historic roots. As early as the 1850s, the German pathologist Virchow described a possible influence of inflammatory cells and of aberrant ECM biosynthesis on tumor growth, whereas the English surgeon Paget coined the Bseed and soilĥ ypothesis, according to which tumor cells Bcan only live and grow if they fall on congenial soil^, aptly describing an important role of the tumor stroma in the pathogenetic process (Paget 1889). Nearly a century later, several of these elements were placed into the context of a pathophysiological process when Harold Dvorak described tumors and their surrounding stroma as Bwounds that do not heal^ (Dvorak 1986). In fact, the granulation tissue that is formed during the physiological process of wound healing includes the key elements of the reactive stroma or Bdesmoplastic stroma^surrounding solid tumors. These include: cancer-associated fibroblasts (CAFs), inflammatory cells, cells constituting blood vessels (endothelial cells, pericytes) and the ECM, which is synthesized and remodeled by both the tumor cells and their cellular microenvironment. Moreover, the ECM can also be derived from plasma proteins escaping from leaky tumor blood vessels (Dvorak 2015). The mutual interactions and crosstalk between tumor cells and the stroma promote tumor progression in various ways. Under the influence of growth factors secreted by tumor cells, including transforming growth factor (TGF) beta, epidermal growth factor (EGF), basic fibroblast growth factor (FGF-2), or vascular endothelial growth factor (VEGF), cancer cells change the properties of their microenvironment (Mueller and Fusenig 2004). Fibroblasts are converted to large spindle-shaped CAFs characterized by the expression of alpha-smooth-muscle actin, an increased proliferative index and a high rate of collagen biosynthesis, all of which result in altered ECM deposition and composition (Lazard et al. 1993). This altered ECM has a profound influence on the properties of the tumor cells, as exemplified by the increased synthesis of tenascin C by CAFs, which promotes tumor cell motility and metastasis by providing a low-adhesive substrate (De Wever et al. 2010;Lange et al. 2007). Recent results indicate that altered ECM biosynthesis within the tumor microenvironment also modulates the properties of tumorinitiating cells, or cancer stem cells, a particularly therapyresistant subpopulation of cancer cells that share similarities with both tumor cells and stem cells (Greve et al. 2012). The aberrant expression of tenascin C, of proteoglycans, including decorin, biglycan, versican and syndecan-1 and of the glycosaminoglycan hyaluronan (the ligand for CD44) have been discussed as having important signaling functions in the cancer stem cell (Bourguignon et al. 2016;Fanhchaksai et al. 2016;Farace et al. 2015;Ibrahim et al. 2013;Jachetti et al. 2015). Moreover, CAFs secrete proteases that degrade the ECM and free cytokines from their ECM storage, thus facilitating tumor angiogenesis and metastasis (Olumi et al. 1999;Sieuwerts et al. 1998). Further, CAFs secrete cytokines such as insulin-like growth factor (IGF) and hepatocyte growth factor (HGF, Bscatter factor^), which increase tumor cell survival and promote cell motility in a vicious feed-forward cycle (Mueller and Fusenig 2004). With respect to the interaction with inflammatory cells, tumor cells secrete cytokines and chemokines that attract various groups of leukocytes, including tumor-associated macrophages (TAMs; Coussens and Werb 2002). This process may at first appear counter-intuitive, as leukocytes can attack and destroy tumor cells. However, the recruitment of immune cells also facilitates tumor growth and metastasis via secretion of matrix-degrading enzymes such as matrix metalloproteinases (MMPs), urokinase-type plasminogen activator (uPA) and heparanase (Coussens and Werb 2002;Gotte and Yip 2006). In addition, TAMs, monocytes, mast cells and neutrophils secrete a vast array of cytokines that exert a proangiogenic effect on endothelial cells and pericytes by inducing signaling events via the cell surface receptor tyrosine kinases and serine/threonine kinases expressed by these cell types. Furthermore, the cancer cells themselves promote tumor angiogenesis by secreting these cytokines, as a prerequisite for the enhanced supply of the tumor with nutrients and for providing access to the vasculature, thus facilitating metastatic spread (Carmeliet and Jain 2000;Rapraeger 2013). In summary, these data clearly indicate that interactions between tumor cells and the surrounding stroma make a considerable contribution to tumor cell survival, tumor growth and metastatic spread. Remarkably, all of the functions of the tumor stroma mentioned above are modulated by glycoproteins of the proteoglycan family, as will be outlined in detail below.

Proteoglycans-multifunctional modulators of cell physiology
Proteoglycans are important functional components of stroma-and cancer-derived ECMs and cell surfaces (Sofeu Feugaing et al. 2013;Theocharis et al. 2010). They are composed of a core protein substituted with one or more covalently linked carbohydrate chains of the glycosaminoglycan (GAG) type. GAGs form long chains of linear repetitive disaccharide units of an amino sugar and one uronic acid ; Fig. 1). GAGs of particular relevance for this review are the heparin-related heparan sulfate (HS; N-acetylglucosamine-alpha-L-iduronic acid/beta-D-glucuronic acid), chondroitin sulfate (CS; N-acetyl-D-galactosamine-D-glucuronic acid) and dermatan sulfate (DS), which is derived from CS by C5-epimerization of the beta-D-glucuronic acid residue. As the names indicate, the basic disaccharide pattern of these GAG chains is subject to post-translational modifications with sulfate residues. Such modification occurs as a non-template-driven enzymatic process during biosynthesis in the Golgi apparatus, thus providing the GAG chain not only with a high negative charge but also with a high degree of structural diversity and functional heterogeneity (Theocharis et al. 2010). This is exemplified by the HS 3-O-sulfotransferase HS3ST2, which is epigenetically silenced in numerous tumor entities and its activity has been shown to modulate signaling via the mitogen-activated protein kinase (MAPK) and Wntpathways in human breast cancer cells (Vijaya Kumar et al. 2014). The various modifications that occur in proteoglycan structure are cell-and tissue-specific but they also depend on the cell stimuli and differentiation stage (Theocharis et al. 2015b). According to their localization, proteoglycans are classified into four major families, including intracellular, cell surface, pericellular and extracellular members. Each family can be further divided into subgroups according to the following criteria: gene and protein homology, modular composition, protein core properties and molecular size (Iozzo 1998;Iozzo and Schaefer 2015;Schaefer and Schaefer 2010). In addition to their contribution to the necessary mechanical structure for the cellular components embedded in the stroma, proteoglycans can functionally interact with various ECM components and matrix-associated proteins, including growth factors, growth factor receptors and cytokines (Iozzo and Sanderson 2011). Because proteoglycans bind numerous secreted proteins and cell surface co-receptors, they can serve several morphogens and signaling molecules enabling them to interact with other matrix components and thus modulating their activities and affecting cell-matrix dynamics (Theocharis et al. 2010(Theocharis et al. , 2015a. As a result of their structural complexity, proteoglycans contribute to several processes that are crucial for homeostasis, differentiation and tissue morphogenesis, participating in various normal and pathological processes, such as wound repair, inflammation and tumor development (Elenius et al. 2004;Frantz et al. 2010;Lu et al. 2011). In cancer, the altered expression of proteoglycans in the stroma affects cancer cell signaling leading to the disruption of critical processes during tumor progression (proliferation, adhesion, migration, invasion, angiogenesis and metastasis; Bouris et al. 2015;Gialeli et al. 2013;Piperigkou et al. 2016;Tsonis et al. 2013). Only one intracellular proteoglycan, serglycin, is involved in the formation of secretory granules and in the synthesis of various ECM components. Serglycin is expressed in hematopoietic and non-hematopoietic tumors and is reported to promote the migration and invasion of low-aggressive breast cancer cells (Korpetinou et al. 2013(Korpetinou et al. , 2014(Korpetinou et al. , 2015. Proteoglycans, such as versican, located in the ECM of cancer cells, can increase their proliferation and motility through modulating the adhesion of cancer cells to ECM (Iozzo and Sanderson 2011;Skandalis et al. 2011;Wight et al. 2014). Moreover, as a member of the small leucine-rich matrix proteoglycans, decorin is a key player in tumor growth and progression; it inhibits angiogenesis and induces tumor cell arrest (Sofeu Feugaing et al. 2013). Of note, soluble decorin is considered to be antioncogenic and anti-metastatic (Csordas et al. 2000;Neill et al. 2012Neill et al. , 2016. With regard to cell surface proteoglycans, syndecans are well established as being the major Fig. 1 Syndecan shedding. The mechanism of shedding involves the proteolytic cleavage of their ectodomain near the plasma membrane by metzincin enzymes, such as metalloproteinases. Heparan sulfate (HS) chains can be additionally cleaved by heparanase. Syndecan core protein can be further processed by intramembrane enzymatic cleavage. Syndecans are in a dynamic interplay with the extracellular matrix and several receptor-tyrosine-kinases (RTKs) and various growth factors, for which they act as co-receptors, thus mediating numerous signaling pathways representatives of the transmembrane HS proteoglycans. They have important roles in several biological processes, such as wound healing, development, stem cell differentiation, inflammation and tumorigenesis (Gotte 2003;Ibrahim et al. 2014b). In cancer, syndecans are considered to be cell surface mediators, as they modulate cell-cell and cellmatrix interactions (Lim et al. 2015;Manon-Jensen et al. 2010). Although they are expressed in several cancer types, each syndecan member has a diverse role based on the type and the stage of cancer, acting as promoters or inhibitors of tumor progression (Barbouri et al. 2014;Nikolova et al. 2009). During the last few years, research interest has focused on the field of epigenetics, as the expression of syndecans in cancer is reported to be regulated by specific microRNAs (Asuthkar et al. 2014;Ibrahim et al. 2012Ibrahim et al. , 2014aIbrahim et al. , 2014b. Another transmembrane proteoglycan is the melanoma-associated chondroitin sulfate proteoglycan (nerve/glial antigen 2; NG2), which promotes vascularization, cell survival, migration and adhesion and modulates the signal transduction of fibroblast growth factor receptor (Cattaruzza et al. 2013a(Cattaruzza et al. , 2013bGarusi et al. 2012;You et al. 2014). Betaglycan also belongs to the superfamily of transmembrane proteoglycans and is a modulator of epithelial-to-mesenchymal transition, acting as a co-receptor for TGFβ members (Diestel et al. 2013). Moreover, this proteoglycan is necessary for reproduction and fetal growth, although it is generally considered as a tumor suppressor (Bernabeu et al. 2009;Bilandzic and Stenvers 2011). Taking into consideration the above data, we can plausibly suggest that, because of their well-established involvement in cancer progression, proteoglycans are of relevance as potential targets for cancer therapy. In many cases, these proteoglycans are expressed by the tumor cells themselves; however, as will be discussed below, an aberrant and functionally relevant expression of selected proteoglycan members has also been observed in the tumor stroma (Ahmed Haji Omar et al. 2013;Farnedi et al. 2015;X. Li et al. 2014;Szarvas et al. 2014;Takahashi et al. 2012). Intriguingly, in the case of several cell-surface-bound proteoglycans, the proteolytic (or, in some cases, lipolytic) mechanism of ectodomain shedding is capable of converting the membrane-bound form of the proteoglycan into a soluble paracrine effector molecule (Nam and Park 2012). As one can imagine, shedding can have a profound effect on the tumor microenvironment, as formerly membrane-bound proteoglycans at the tumor or stroma cell surface might no longer be able to act as co-receptors for a wide range of ligands in this cell type, although they might either enhance or competitively inhibit processes such as proteolysis, angiogenesis and cytokine signaling within the paracrine diffusion range of the ectodomain. Before we discuss selected examples of shed proteoglycans in the tumor stroma and their functional implications, the basic mechanisms of proteoglycan ectodomain shedding are briefly presented and discussed below.
A primer on proteoglycan ectodomain shedding Cell membrane proteoglycans have crucial roles in many steps of tumor progression, as their interactions with the surrounding stroma have been correlated with differential proteoglycan effects in cancer cells. Several membrane proteoglycans are known to undergo controlled enzymatic cleavage of their ectodomain by certain proteases, the so-called sheddases. Shedding is a mechanism by which the intact extracellular domain of a membrane protein is converted into a soluble molecule. Notably, shedding does not result in the complete degradation of the ectodomain but rather leaves it intact, allowing for functional ligand interactions of this domain as a paracrine effector molecule (Manon-Jensen et al. 2010;Nam and Park 2012). The shedding of membrane-bound proteoglycans constitutes an important and clearly controlled posttranslational procedure that regulates several normal and pathological conditions. It can be a critical mechanism for the onset of infectious and non-infectious diseases; therefore, the development of reliable tools to measure the presence of shedding proteoglycans in such conditions is crucial. With the exception of the glycosylphosphatidyl-inositol (GPI)-anchored glypicans, which can be shed by a phospholipase (Traister et al. 2008), all proteoglycans of the tumor stroma are shed by proteases. Shedding usually occurs at a juxtamembrane site and is mediated by members of the MMP family (matrilysin, collagenases, gelatinases), by membrane-type metalloproteinases (MT1-MMP, MT3-MMP) and members of the ADAM (a disintegrin and metalloproteinase) and ADAM-TS (ADAM with thrombospondin motif) families (ADAM-10, ADAM-17 [TACE], ADAMTS-1, ADAMTS-4). In addition, serine proteases such as plasmin and thrombin have been shown to induce shedding of proteoglycans (Manon-Jensen et al. 2010;Nam and Park 2012). An overview of relevant sheddases can be found in Table 1. Shedding of proteoglycans can be regulated at several levels. Regulatory mechanisms include the induction or repression of sheddases and protease inhibitors, such as tissue inhibitors of metalloproteinases (TIMPs) and the modulation of enzyme activity via the binding of sheddases and protease inhibitors to glycosaminoglycan chains either stabilizing these proteins in an active or inactive conformation or protecting them from degradation (Bernfield et al. 1999;Elenius et al. 2004). Whereas shedding can occur constitutively, it can be substantially enhanced by exogenous stimuli, including growth factors, chemokines, bacterial virulence factors, cell stress, phorbol esters, trypsin, insulin and the HS-degrading enzyme, heparanase (Bernfield et al. 1999;Manon-Jensen et al. 2010;Nam and Park 2012). We will discuss selected examples in the following section, which addresses various members of shed proteoglycans in the tumor stroma.

Shed proteoglycans in the tumor stroma
Recent studies have categorized cell surface proteoglycans into various families. Membrane-bound proteoglycans include syndecans-1 to -4, the GPI-anchored glypicans-1 to -6, the TGF beta Type III receptor betaglycan, melanomaassociated chondroitin sulfate proteoglycan (MCSP/NG2) and phosphacan / protein tyrosine phosphatase gamma (PTPgamma). Syndecans and glypicans are HS proteoglycans, NG2 and phosphacan are CS proteoglycans, whereas the GAG chains of betaglycan are composed of CS and HS chains (Iozzo and Schaefer 2015). Notably, members of all of these proteoglycan families are known to be substrates for sheddases. In this section, we will present the process of ectodomain shedding by using selected examples and discuss their pathophysiological relevance within the tumor stroma.

Syndecans
Syndecans are prototypical transmembrane HS proteoglycans that have several functions relevant to tumor progression (Fig. 1). For example, they are involved in the regulation of cell proliferation by acting as a co-receptor for several growth factor receptors; they act as cell adhesion molecules and modulate proteolysis, chemokine action, angiogenesis and cancer stem cell function (Ibrahim et al. 2013;Manon-Jensen et al. 2010;Yip et al. 2006). Notably, all syndecans are shed in vitro and in vivo (Bernfield et al. 1999). Proteolytic cleavage of syndecans is mediated by plasmin, thrombin, MMPs and ADAMTs (Kim et al. 1994; Table 1) and can be modulated by interfering with several signaling cascades including the protein kinase C (PKC), the protein tyrosine kinase (PTK) and the MAP kinase pathways (Bernfield et al. 1999). Substitution of the glycine residue at position 245 of syndecan-1 with a leucine has previously been demonstrated to reduce the extent of MT1-MMP mediated shedding by ∼ 50 % in human fibrosarcoma cells (Endo et al. 2003). On the other hand, the ectodomain shedding of murine syndecan-1 and -4 is inhibited by TIMP-1 and TIMP-2, revealing that the responsible sheddase belongs to the ADAM family of metalloproteinases (Fitzgerald et al. 2000). Syndecan shedding has bimodal effects under several pathological conditions in vivo (Fig. 2). Soluble syndecan-1 has been shown to be required for the formation of chemotactic gradients in a lung inflammation model in mice and may be invvolved in the regulation of leukocyte-endothelial interactions and angiogenesis (Gotte 2003;Gotte and Echtermeyer 2003). Moreover, shedding of the syndecan-3 ectodomain appears to modulate feeding behavior and body weight in mice (Reizes et al. 2003). On the other hand, the overexpression of soluble syndecan-1 delays wound repair and induces abnormal blood vessel morphology, because of the increased amount of soluble syndecan-1 observed in skin wound fluids (Elenius et al. 2004). These large amounts have been shown to increase elastolytic activity, which may also be of relevance in the tumor stroma. Once syndecan-1 is shed from the cell surface, the remaining transmembrane fragment (cCTF) undergoes intramembrane proteolysis by γ-secretase and is cleaved by the proteasome (Pasqualon et al. 2015). This fragment is able to inhibit cell migration and invasion in A549 lung cancer cells through the phosphorylation/activation of Src kinase, focal adhesion kinase and Rho-GTPase, whereas it blocks syndecan-1dependent lung tumor cell migration and invasion. Moreover, lung tumor formation of A549 cells in mice is reduced by the overexpression of the syndecan-1 cytoplasmic fragment, enhancing the crucial role of syndecan-1 in tumor cell migration. The cleavage sites for syndecan shedding seem to be conserved and have been identified in the case of human and murine syndecan-1 and for syndecan-4 (Manon-Jensen et al. 2010; Wang et al. 2005). The potential cleavage sites have been suggested to be located within the 15-amino acid juxtamembrane region near the plasma membrane and confirmation is available that the shedding is a membrane surface event. In the case of murine syndecan-1, for example, the cleavage site of syndecan-1 is located between Ala243 and Ser244, nine amino acids from the membrane, as determined by domain-swapping experiments and mass-spectrometric sequencing (Wang et al. 2005). Clearly, several functions of syndecans in an oncological context are mediated by the membrane-bound form. Moreover, most of these functions have been shown to be tumor-cell-autonomous. However, Fig. 2 Overview of various cell surface proteoglycans known to be shed, including transmembrane (syndecans, betaglycan, nerve/glial antigen 2 (NG2) and phosphacan) and glycosylphosphatidyl-inositol (GPI)-anchored (glypicans). The glycosaminoglycan (GAG) chain and the major cleavage sites for selected sheddases are depicted, as exemplified for the syndecan-1 core protein. The GPI-anchored glypicans are shed by phospholipases, whereas the other depicted transmembrane proteoglycans are cleaved by a wide range of proteases (see Table 1). Moreover, autocrine and paracrine effects induced by each proteoglycan are shown as they affect important stromal functions in angiogenesis, tumor cell proliferation and survival, cell adhesion, migration and metastatic behavior. Shed ectodomains are deposited in the extracellular matrix where they can influence the function and interactions of large structural matrix glycoproteins, growth factors and cytokines.
Ectodomain shedding reduces the available binding sites for such ligands at the cell surface, resulting in altered cytokine responses. Oval left In the case of syndecan-1 (Sdc-1), shedding is facilitated by degradation of the HS chains by heparanase. Shed syndecan-1 and betaglycan can be transported in exosomes, a means of transferring proteins and microRNAs from tumor cells to stroma cells. Finally, syndecan-1 ectodomains can reach the nucleus and modulate gene transcription (for details, see Theocharis et al. 2015b). Whereas phosphacan is a shed proteoglycan, very little is known so far about its function in the tumor stroma (MCSP melanoma-associated chondroitin sulfate proteoglycan, EMT epithelial to mesenchymal transition, TGF transforming growth factor, HS heparan sulfate, CS chondroitin sulfate, FGF fibroblast growth factor, NK natural killer, ADAMT a disintegrin and metalloproteinase with thrombospondin motif, GF growth factors) indications have also been obtained for important roles of syndecans in the tumor stroma and for soluble syndecan ectodomains in the stroma, which can be either generated by cleavage from the tumor cell surface or obtained from other cell types such as CAFs, endothelial cells, or leukocytes. Indeed, syndecan-1 is expressed and dysregulated in the stroma of several types of cancers, including breast carcinoma, colon cancer, gastric cancer, bladder cancer and squamous cell carcinoma (Ahmed Haji Omar et al. 2013;Hashimoto et al. 2008;Mukunyadzi et al. 2003;Szarvas et al. 2014;Yang et al. 2002). An aberrant stromal expression has also been noted for syndecan-2 in various tumor entities (Ahmed Haji Omar et al. 2013;Farnedi et al. 2015;Hrabar et al. 2010). Tumor promotion is counted among its main functions. This may be correlated with the existence of its soluble form in some cancers (Miles and Sikes 2014;Nikolova et al. 2009). Of note, the expression levels of syndecan-1 distributed from epithelium to stroma are higher by far in breast cancer tissue compared with normal tissue (Lofgren et al. 2007). Although the contribution of soluble syndecan ectodomains is difficult to assess in these histopathological studies, functional studies do indeed point at a mechanistic role for soluble syndecans in tumor progression. For example, soluble syndecan-1 promotes the growth of myeloma tumors in vivo and tumors producing soluble syndecan-1 ectodomain have been shown to grow significantly faster than tumors expressing syndecan-1 on the cell surface (Yang et al. 2002). Moreover, breast cancer cells overexpressing soluble syndecan-1 show higher cell proliferation rates and higher matrigel invasiveness compared with cells overexpressing a non-cleavable, constitutively membrane-bound form of this proteoglycan (Nikolova et al. 2009). This change in cell behavior can be attributed to a downregulation of the anti-metastatic homotypic cell adhesion molecule E-cadherin and a downregulation of the MMP inhibitor TIMP-1 in cells overexpressing soluble syndecan-1. Moreover, MT1-MMP produced by stromal fibroblasts has recently been revealed to release syndecan-1 ectodomain as a paracrine mediator by the direct cleavage of syndecan-1 at the surface of these cells (Table 1). CAFs stimulate breast cancer cell growth through MT1-MMP enzymatic activity. Together, these data pinpoint the novel role of fibroblastderived MMPs and their proteoglycan targets in stromal-epithelial signaling in carcinomas. A mechanistically highly relevant study employing a tumor cell/stromal fibroblast coculture system has demonstrated that stromal syndecan-1 mediates breast cancer cell proliferation in vitro and in vivo and that this regulation requires the HS chains and the presence of the stromal cell-derived factor 1 (SDF1) and fibroblast growth factor 2 (FGF-2; Su et al. 2008). Strikingly, the syndecan-1 expressed by stromal fibroblasts has apparently a profound effect on matrix organization within the tumor stroma and can thus be expected to modulate tumor cell motility and signaling via integrins (Yang et al. 2002). However, the potential roles of syndecans are not restricted to cancerassociated fibroblasts. Data from a syndecan-1 transgenic mouse skin wound model suggest an influence of the HS chains of syndecan-1 on the proteolytic milieu, as soluble syndecan-1 increases elastolytic activity in the wound fluid, leading to delayed repair and abnormal angiogenesis (Elenius et al. 2004). Indeed, several studies have demonstrated that syndecans are capable of binding proteases, including various members of the MMP family (Manon-Jensen et al. 2010). Whereas these data are in accordance with Dvorak's paradigm of tumors as wounds that do not heal (Dvorak 1986), additional indications have been found for a role of syndecan-1 in tumor angiogenesis: syndecan-1 is a modulator of integrins relevant for this process (Rapraeger 2013) and heparanaseaccelerated shedding of syndecan-1 ectodomains has been shown to modulate angiogenesis in myeloma via the binding of the proteoglycan to VEGF (Purushothaman et al. 2010). Apart from these processes, data in animal models suggest a role for syndecan-1 in leukocyte recruitment, thus including potential roles in endothelial and inflammatory cells. For example, the lack of syndecan-1 on leukocytes enhances integrin-ICAM-1 interactions and leads to increased leukocyte recruitment during inflammation, whereas lack of heparan sulfate on endothelial cells interferes with leukocyte recruitment (Kumar et al. 2015). Similar findings can be expected regarding leukocyte recruitment in the tumor stroma; however, no functional studies have been performed so far. Mechanistically, a striking finding has been the observation that shed syndecan-1 can be transported to the nucleus in which it can influence transcription via the inhibition of histone acetylation, thus broadening the functional impact of the molecule (Stewart et al. 2015) (Fig. 2). In summary, these data indicate that the soluble ectodomains of syndecans have a profound influence on the tumor microenvironment and, thus, these domains emerge as potential therapeutic targets.

Glypicans
The glypican family of cell surface proteoglycans consists in six members in mammals (Bernfield et al. 1999). The core protein of glypicans is attached to the cell surface through a GPI anchor. Several members of the glypican family influence tumor development and their expression is dysregulated in a context-dependent manner in various tumor entities (Ibrahim et al. 2014b;Theocharis et al. 2015b). For example, glypican-3 has been proposed as a prognostic marker for melanoma and human hepatocellular carcinoma (HCC; Haruyama et al. 2015). Notably, stromal expression of glypican-1 is increased in prostate cancer (Suhovskih et al. 2013) and in ovarian cancer, where it is a poor prognostic indicator for survival (Davies et al. 2004), indicating a possible important role in the tumor microenvironment. At the functional level, glypicans modulate numerous signaling pathways (FGF, Wnt, Hedgehog, bone morphogenetic protein [BMP]), thus regulating cell proliferation, cell survival, cell migration, cell differentiation and angiogenesis (Filmus and Capurro 2013). Similar to syndecans, glypicans can be shed, however, by necessity the molecular mechanism differs, as the HS proteoglycan is GPIanchored (Fig. 2). Glypican shedding includes the cleavage of their GPI anchor by phospholipases (Traister et al. 2008). Moreover, they can be dually processed by partial cleavage via proteases and lipases (Filmus and Capurro 2013). The high levels of shed glypican may arise either from the large amounts of this proteoglycan or from the increased enzymatic activity of proteases. Therefore, glypican-3 might be an important candidate antigen for immunotherapy of HCC, where it is strongly expressed. Various clinical approaches include the glypican-3 peptide vaccination and the anti-glypican-3 antibody in patients with melanoma and HCC (Ishiguro et al. 2008;Nakatsura et al. 2004). Moreover, the suppression of glypican-1 in pancreatic cancer leads to decreased cytotoxic effects of natural killer cells (Bloushtain et al. 2004;Hershkovitz et al. 2007).

Melanoma-associated chondroitin sulfate proteoglycan-nerve/glial antigen 2
NG2 is a large multidomain cell surface CS proteoglycan, which is of particular relevance for the tumor stroma, as it is expressed by mural cells (and is thus relevant for angiogenesis) and is widely used as a molecular marker for CAFs (Balzarini et al. 2012;Kim et al. 2015;Park et al. 2015). Important functions of NG2 include a role in pericyte recruitment during (tumor) angiogenesis (Gibby et al. 2012) and as a mediator of growth-stimulatory and pro-survival effects, which are attributable to activation of the Akt and betacatenin signaling pathways following the interaction of NG2 with its matrix substrate collagen VI (Iyengar et al. 2005;Cattaruzza et al. 2013a). Moreover, NG2 is capable of modulating FGF-2 signaling in perivascular cells (Cattaruzza et al. 2013b). Its role as a pericyte marker has recently been exploited in a therapeutic study in which a fusion peptide comprising an NG2-targeting peptide and tissue factor was utilized to induce tumor cell infarction in a xenograft model (Brand et al. 2016). Importantly, a clinicopathological study on head and neck cancer has revealed that NG2 expression in neoplastic cells and in the intralesional stroma is strongly associated with loco-regional relapse, emphasizing the relevance of aberrant NG2 expression in an oncological context (Farnedi et al. 2015). Moreover, a predictive value regarding metastasis formation has been assigned to NG2 in soft tissue carcinoma patients (Benassi et al. 2009). The juxtamembrane domain of NG2 contains several protease cleavage sites and the diffusible ectodomain can be detected in the blood stream of melanoma patients (Price et al. 2011). Furthermore, NG2 has been detected in unique glial cells receiving synaptic input from neurons. Interestingly, its enzymatic cleavage by ADAM-10 and γ-secretase yields an active ectodomain associated with the extracellular matrix, thus modulating neuronal networks (Sakry et al. 2014). Soluble NG2 has been detected in vitro and in vivo. Pharmacological inhibition of NG2 shedding in mice results in impaired behavior and sensory data that is correlated with various human diseases such as schizophrenia (Geyer et al. 2001), paving the way for similar therapeutic approaches in an oncological context.

Betaglycan
Betaglycan (the proteoglycan form of TGFbeta receptor type III) has been identified as a suppressor of cancer progression in several epithelial tumors Hempel et al. 2007;Sun and Chen 1997), although its function is context-dependent in some tumor entities (Criswell et al. 2008). Mechanistically, betaglycan plays an important role as a receptor for TGFbeta and TGFbeta family members, mediating signaling and leading to a suppression of the transcription factor nuclear factor kappa B of downstream targets such as MMP2, and to a modulation of Rho-GTPase-mediated signaling cascades involved in cytoskeletal remodeling (Bilandzic et al. 2014;Oh et al. 2013;You et al. 2009). In accordance with a potential tumor suppressor role, betaglycan expression is downregulated in CAFs in oral squamous cell carcinoma (Meng et al. 2011). Notably, betaglycan inhibits heterotropic adhesion between myeloma cells and bone marrow stromal cells, indicating an important function in the tumor microenvironment (Lambert et al. 2011). Whereas betaglycan can be transported from tumor cells to fibroblasts via exosomes, thereby inducing their differentiation to a myofibroblast-like-state (Webber et al. 2010), some functions of this proteoglycan depend on the shedding of its ectodomain, which is mediated by MT1-MMP (Velasco-Loyden et al. 2004). By using betaglycan mutants that displayed either reduced or increased shedding rates, Elderbloom et al. (2014)

Outlook: prognostic and therapeutic considerations
The generation of soluble proteoglycan ectodomains within the tumor stroma has a considerable impact on several relevant non-cancer cell-autonomous processes, including the regulation of the proteolytic milieu, of cytokine signaling, of matrix remodeling as a prerequisite for metastasis and of angiogenesis (Fig. 2). Based on these important functions, the question arises as to whether this knowledge can be utilized in a translational or clinical setting. One possibility is the detection of proteoglycan ectodomain levels in the serum as a diagnostic tool. Elevated levels of soluble proteoglycans in biological fluids, such as blood serum, have been detected in patients with sepsis, ischemia or reperfusion injury, or graft-versus-host disease and in various cancer types (Joensuu et al. 2002;Kliment et al. 2009;Rehm et al. 2007;Seidel et al. 2003;Steppan et al. 2011). For example, the presence of the soluble syndecan-1 in the blood stream is indicative of poor prognosis in several malignancies (Joensuu et al. 2002;Lovell et al. 2005;Szarvas et al. 2014;Vassilakopoulos et al. 2005). A diagnostic or prognostic value for additional shed proteoglycan ectodomains in hepatocellular carcinoma has also been attributed to glypican-3 (Haruyama et al. 2015) and NG2 (Lu et al. 2015). Finally, in vivo studies have correlated the presence of shed syndecan-1 with the inflammatory response (Hayashida et al. 2008;Q. Li et al. 2002), with potential relevance to leukocyte recruitment to the tumor stroma. These data indicate that soluble proteoglycan ectodomains that have been released into the blood stream are indeed promising diagnostic markers that might aid therapeutic decisions in a clinical context. Given the functional importance of shed proteoglycan ectodomains in the tumor stroma, the inhibition of shedding might be a worthwhile therapeutic approach that may have an impact not only on the tumor cells (which may more easily develop therapeutic resistance because of their genomic instability and cancer stem cell properties) but also on its microenvironment. Therefore, a highly promising next step will be the development of specific shedding inhibitors in clinical trials for the focused approach of several pathological situations. Some potential inhibitors that have been recently discussed in the context of syndecan shedding include MT1-MMP inhibitors (GM6001, BB94, TIMP-2), MMP-7 inhibitors (TIMP-1), the FGF-2 antagonist PTX3 and all-trans retinoic acid (Choi et al. 2013). Whereas more translational research in this area is warranted, the inhibition of shedding might contribute to the exciting therapeutic concept of tumor stroma normalization, which aims at transforming the cancer microenvironment into a non-permissive one with the aim of slowing down or even reverting tumor progression (Mueller and Fusenig 2004).

Compliance with ethical standards
Conflict of interest The authors declare that no conflicts of interest exist.