Syndecans: transmembrane modulators of adhesion and ...dm5migu4zj3pb.cloudfront.net/manuscripts/12000/12802/JCI...can-2 interaction may be through the C1 domain. Again, since the C1

The Journal of Clinical Investigation | April 2001 | Volume 107 | Number 8 935

PERSPECTIVE SERIES

E. Helene Sage, Series Editor

Matricellular proteins

Syndecans are transmembrane heparan sulfate proteo-glycans (HSPGs) that are present on most cell types.HSPGs have been known for some time to regulate avariety of biological processes, ranging from coagula-tion cascades, growth factor signaling, lipase bindingand activity, cell adhesion to ECM and subsequentcytoskeletal organization, to infection of cells withmicroorganisms. They are complex molecules, withspecific core protein to which a variable number of gly-cosaminoglycan (GAG) chains are attached. Not onlythe number of chains varies; although syndecans main-ly bear heparan sulfate GAGs, some have additionalchondroitin/dermatan sulfate chains. Furthermore,heparan sulfate chains can vary in length, epimeriza-tion of glucuronic acid to iduronic acid, overall sulfa-tion of the chains, and position of sulfation of themonosaccharides.

Early studies relied on structurally nonspecificapproaches such as competition with heparin or treat-ment with heparinases or chlorate to remove GAGchains or prevent their sulfation, respectively. Theseapproaches defined a number of interacting ligands forheparan sulfates including insoluble ECM componentsand soluble growth factors and cytokines, and theyhelped to implicate HSPGs in a variety of biologicalprocesses (1–5). Cell surface HSPGs, including synde-cans, appear to play modulatory roles, such as present-ing growth factors to their primary receptors or increas-ing the infectivity of viruses by interacting with theirprimary receptors. In addition, these molecules canmodify adhesion mediated by the integrins, the majorfamily of receptors for ECM. This last function estab-lishes a parallel between the syndecans (as well as othercell surface HSPGs) and the secreted matricellular pro-teins discussed elsewhere in this Perspective series.

This is an exciting time to be in the field of syndecanresearch. Recently, there has been a surge of interest inthe role of HSPGs and in identifying both the specificsequences of monosaccharides and sulfation patternsinvolved in ligand binding. With the cloning of variouscore proteins, it has become possible to study the fea-tures of these proteins that affect the ligand-bindingability and subsequent biological responses (reviewed inref. 5; Table 1). Now that tools are available to analyzeindividual protein cores, and techniques are becoming

available to determine the sequence of their GAGchains, it is becoming evident that the protein and car-bohydrate components of these molecules each playspecific biological roles. Recent advances in “sequenc-ing” of GAG chains present on individual proteoglycanshave begun to allow structure to be correlated withfunction in these complex molecules. For example, adecasaccharide containing 2-O-sulfated iduronic acidand 6-O-sulfated N-acetylglucosamine is needed topotentiate the interaction of FGF-2 with its receptor. Incontrast, a less specific sequence of seven to eight N-sul-fated disaccharides containing iduronic acid (with orwithout 2-O-sulfation) serves to bind the higher affini-ty heparin-binding domain of fibronectin (HepII; ref. 5).This Perspective will concentrate on the syndecan fam-ily of transmembrane HSPGs and their roles in adhe-sion to ECM and subsequent matrix modification. Inaddition, it will address how syndecans may be pivotalin coordinating growth factor and adhesion signalingmechanisms. For discussion of some recent advancesconcerning the roles of cell surface proteoglycans inother fundamental biological processes, see refs. 2, 5, 6,and 7 and references therein.

Structure of syndecansThe core proteins of the four mammalian syndecanshave divergent ectodomains, with substantial sequenceconservation seen only at sites of GAG substitution.They are normally substituted with three to five heparansulfate chains, although some members can be addi-tionally substituted with chondroitin or dermatan sul-fate. Three of the syndecans (1, 2, and 4) have beenshown to bind ECM molecules. As with growth factorbinding, interactions with the ECM appear to bethrough the GAG chains, and since these heparan sul-fate GAGs are very similar overall (reviewed in ref. 5), itmight be expected that all cell surface HSPGs will bindligands such as matrix molecules or growth factors tothe same extent. Since this is clearly not the case(reviewed in ref. 1), other interactions involving regionsof the core protein must affect the outcome of GAGchain binding to these ligands (1). Ligand binding alsoappears to be determined by specific sulfation/epimer-ization patterns that result in regions of high (Sdomains) and low (N domains) modification along the

Syndecans: transmembrane modulators of adhesion and matrix assembly

Anne WoodsDepartment of Cell Biology, University of Alabama at Birmingham, Volker Hall 203A, 1530 3rd Avenue S., Birmingham, Alabama 35294-0019, USA. Phone: (205) 934-1548; Fax: (205) 975-9956; E-mail: [email protected].

chain length. The complete sequencing of the specificGAG chains for any particular syndecan remains to beachieved but will yield extremely important information.

Syndecan core proteins are small compared with theirlarge GAG chains (1–5). These proteins contain fourconserved tyrosine residues, and syndecans can be tyro-sine phosphorylated, although the residues involvedand the biological consequences of this modificationare unclear. Their transmembrane domains are highlyhomologous, and their short cytoplasmic tails have tworegions of high homology proximal and distal to themembrane (C1 and C2), with an intervening sequence(V region) that is specific to individual syndecans. Thishas led to the speculation that certain functions arecommon to all syndecans, with others specific for indi-vidual family members. In terms of common function,all syndecans have a COOH-terminal FYA sequence thatcan interact with PDZ domain containing proteins,which implies a role in protein-protein interactions. Todate, three PDZ-domain proteins have been shown tointeract with syndecans (Figure 1): syntenin (8),CASK/LIN (9), and synectin (10). In addition, a recent-ly identified binding partner, synbindin (11), containsa sequence with limited homology to a PDZ domain.

Initially it was thought that interactions of synde-cans with PDZ domain proteins resulted in the for-mation of a submembraneous scaffold that connectsto the cytoskeleton. However, recent evidence pointsto a possible role of these syndecan-binding proteinsin trafficking and/or sorting of proteins to or fromspecific membrane areas (11–13). In particular, itappears these proteins may regulate clustering of syn-decans (11). Since the FYA motif is common to all foursyndecans, it might be expected that some competi-tion for binding may occur. Syndecans, however, arehighly regulated in their expression, both in develop-ment and in a cell type–specific manner. In general,syndecan-1 is the major syndecan in epithelial cells,syndecan-2 in fibroblasts, and syndecan-3 in neuronaltissue, although more than one syndecan can beexpressed in the same cell type. Interestingly, synde-can-4, although a minor component, is present in arange of cell types, including fibroblasts, epithelial,and smooth muscle cells, perhaps indicating a specif-ic role for this transmembrane proteoglycan.

The C1 domains of syndecans (Figure 1) are also high-ly homologous. Syndecan-3 binds the heparin-bindinggrowth-associated molecule HB-GAM via its GAGchains. Binding results in axonal extension, concomi-tant with an interaction of the C1 domain with a com-plex that includes c-src and the syndecan-3 substratecortactin (reviewed in refs. 3–5). Cortactin interacts withthe microfilament submembraneous cytoskeleton, and,perhaps indirectly, the microtubule system. Ezrin, amember of the ERM (ezrin, radixin, moesin) family,binds syndecan-2 (14). ERM proteins are also postulat-ed to link membrane receptors to the cortical actin

meshwork. Since previous studies indicate that ERMsbind to membrane-proximal basic amino acidsequences of transmembrane proteins, the ezrin-synde-can-2 interaction may be through the C1 domain.Again, since the C1 domains are highly conserved, itmight be expected that this interaction could occurwith the other syndecans. A third protein that can inter-act with a C1 region is syndesmos (15). Syndesmosbinds the C1 region of syndecan-4, but here the inter-action appears to be limited to syndecan-4, since the dis-tinctive V region of this syndecan is also required.

The V regions of syndecans are conserved betweenspecies but differ between syndecans 1–4. Thus, the Vregion of syndecans may determine specific biologicalroles. Few binding partners have been elucidated forthese regions. The syndecan-4 V region is required forinteraction of the C1 region with syndesmos, but it doesnot seem to interact directly with this ligand, suggest-ing that the V-region alters the conformation of the C1region in a way that favors the interaction with this lig-and. The syndecan-4 V region binds to phosphatidyli-nositol 4,5 bisphosphate (PIP2), as well as to the cat-alytic domain of the serine/threonine kinase proteinkinase Cα (PKCα), resulting in a superactivation of thisenzyme (reviewed in refs. 1–5). Superactivation of PKCαactivity requires oligomerization of the V region or thewhole core protein. All syndecans oligomerize to formhigher-order complexes, and oligomerization mayunderlie their biological roles. The self-interaction ofthe cytoplasmic domains has only been investigated forsyndecan-4 so far. Short synthetic peptides having thesequence of syndecan-4 V region form dimers with aparallel twisted clamp structure (reviewed in refs. 3, 4).Dimerization is stabilized by PIP2, which promotes fur-ther oligomerization.

936 The Journal of Clinical Investigation | April 2001 | Volume 107 | Number 8

PERSPECTIVE SERIES



Figure 1Schematic diagram of the known interactions of syndecan core proteins.The GAG chains on the ectodomain are indicated by white lines, and thecytoplasmic domain is divided into the C1, V, and C2 regions. Moleculesknown to interact with different sites within the core protein are indicat-ed. CBD, cell binding domain; TM, transmembrane domain.

Syndecans in adhesion and matrix assemblySyndecan-1. Syndecan-1 was the first HSPG of thisfamily to be identified and cloned. It is mainly limit-ed to epithelial cells, but it is also found in condens-ing mesenchyme during development and in pre-Blymphocytes and plasma cells. Consistent with a rolein adhesion, syndecan-1 is present in a basolateral dis-tribution in epithelia, and it appears to regulateepithelial morphology (reviewed in refs. 2, 3, 5), sincetransfection of epithelial cells with antisense mRNAfor syndecan-1 results in an epithelial-mesenchymalswitch and activates cells to invade collagen gels. Theattendant loss of E-cadherin in these cells suggests acoordinate regulation of syndecan-1 and E-cadherin,and indeed, reduced E-cadherin expression can alsoresult in decreased syndecan-1 production.

Early studies indicated that syndecan-1 may beinvolved in cell adhesion to ECM. Transfection of syn-decan-1 into Schwann cells, which normally lack thissyndecan, increases spreading and promotes the for-mation of focal adhesions and stress fibers. Althoughsyndecan-1 codistributes with the microfilament sys-tem at the membrane during spreading and becomesdetergent insoluble, it does not localize at focal adhe-sions. Indeed, there has been only one report of synde-

can-1 being present in focal adhesions. Interestingly,detergent-insolubility, which is usually taken to indi-cate a linkage to the cytoskeleton, does not require thepresence of the cytoplasmic domain; a very recent study(7) indicates that syndecan-1 transmembrane domainsassociate with detergent-insoluble lipid rafts as part ofa specialized type of endocytosis.

Although syndecan-1 is present in immature B cellsand mature plasma cells, it is not found in B lympho-cytes in the circulation, again suggesting that syndecan-1 controls adhesion to ECM. Studies with myelomacells that lack syndecan-1 and, therefore, provide a nullbackground for investigation of transfected constructs,indicate that syndecan-1 does indeed increase adhesionto collagens and limit invasion of collagen gels(reviewed in refs. 1–4). Both syndecan-1 and -4 have thisability, but transfection with glypican-1 (a glycophos-phoinositol–linked (GPI-linked), nontransmembraneHSPG) does not increase adhesion, suggesting that thepresence of GAG chains is not sufficient for this activi-ty and that the syndecan-1 and -4 cytoplasmic domainsmight be required. However, further experiments usingglypican-syndecan chimeras disproved the need for syn-decan cytoplasmic domains, since the ectodomain ofsyndecan-1 tethered to a GPI anchor was effective.


PERSPECTIVE SERIES



Figure 2Focal adhesions and stress fiber formation. (a) Syndecan-4 is diffusely distributed in the membrane, as is the phospholipid PIP2 (yellow). PKCα (redovals) is cytosolic under these conditions. (b) Following interaction with ligand (blue), syndecan-4 becomes clustered into forming focal adhesions.This clustering is stabilized by PIP2 and allows binding and superactivation of PKCα, leading to the phosphorylation of focal adhesion proteins. Theactin microfilament system is condensed at focal adhesions, inducing the formation of stress fibers. Shed forms of syndecan-4 may interfere with thecell’s ability to reorganize its cytoskeleton in response to adhesive interactions. The function of syndecan-4, either in its membrane-bound form or insolution, is analogous to that of a number of matricellular proteins described elsewhere in this Perspective series.

Whether increased adhesion turns out to be due tointeractions of the ectodomain or transmembranedomains of syndecan core proteins with other mem-brane proteins remains to be determined.

Syndecan-2. Syndecan-2 is the major syndecan infibroblasts. Previous studies have shown that it can bephosphorylated on serine both in vivo and in vitro andthat its V region is a substrate for PKC and PKA(reviewed in ref. 16). Syndecan-2 shows a general diffusedistribution in the membrane and does not localize tofocal adhesions. However, if it is overexpressed in Chi-nese hamster ovary (CHO) cells, it induces increasedspreading, with the cells adopting a more epithelialmorphology (17). This results in an increase in stressfibers within the cells, but these do not terminate atfocal adhesions. In contrast to a direct role in adhesion,syndecan-2 appears to regulate matrix deposition (16).

Formation of a pericellular matrix of fibrillarfibronectin requires the participation of activated inte-grins as well as syndecan-2. Transfection of CHO-K1 cellswith full-length syndecan-2 has little effect on fibrilassembly of endogenous fibronectin or laminin (16).However, constructs that either lack the cytoplasmicregion entirely or are truncated in the middle of the Vregion inhibit fibril formation. Crucially, these mutantforms of syndecan-2 have no effect on integrin expressionor the synthesis or export of fibronectin or laminin. Cellsexpressing truncated syndecan-2 still attach and spreadon substrates of fibronectin, similar to control cells, butthey fail to remodel the substrate-adsorbed fibronectininto fibrils around their edges. Similarly, soluble exoge-nous fibronectin was incorporated into the matrix ofcontrol cells, but not those expressing the truncated con-structs. Both activation of integrins and an intactcytoskeleton are required for fibril assembly. Since fibrilformation is prevented even in cells transfected with con-stitutively active integrins, it appears that disregulationof the cytoskeleton by the truncated syndecan-2 is suffi-cient to block this process. Focal adhesions, althoughslightly reduced in number, appear to contain the normalrepertoire of components in transfected cells, indicatingthat expression of the truncated syndecan-2 may reducefibril formation through an abnormality in the sub-membraneous cortical meshwork, which has been impli-cated in fibril nucleation and initial growth. The findingthat ERM proteins may bind syndecan-2 may support arole for this HSPG in cortical microfilament function.

The molecular mechanisms underlying the lack ofmatrix assembly in cells carrying the truncated synde-can-2 is still under investigation, but the V region ofsyndecan-2 is a substrate for PKC, the activity of whichis needed for matrix assembly (reviewed in ref. 16).Both serines in the V region of syndecan-2 can bephosphorylated, and the extent of phosphorylation isdictated by the degree of oligomerization. As discussedbelow, oligomerization of syndecan-4 appears to con-trol its role in cell adhesion, so it is tempting to specu-

late that self-association of syndecan-2 regulatesmatrix assembly, possibly by altering the phosphory-lation status of syndecan-2. Experiments underway totransfect cells with syndecan-2 in which these serineresidues are mutated will help to determine if phos-phorylation of one or both serines is involved in itsrole in matrix assembly.

Other studies in fibroblasts have indicated possibleroles for syndecan-2 in adhesion. Overexpression ofsyndecan-2 and syntenin increases fibroblast spread-ing, and the two proteins colocalize at the membraneand in intracellular vesicles (8). Similarly, Kusano et al.(18) showed that downregulation of syndecan-2 candiminish microfilament bundle formation in responseto fibronectin, mimicking the effect of treatments thatblock heparan sulfate function. The authors conclud-ed that syndecan-2 and α5β1 integrin interactions areneeded for bundle formation, a situation analogous tothat seen with syndecan-4 (see below).

The other major studies on syndecan-2 have been inthe nervous system, where CASK/LIN distribution cor-relates with both syndecan-2 and -3 in both a temporaland spatial manner (9). CASK, in addition to bindingto the C2 domain (Figure 1) of syndecans via its PDZdomain, has a protein 4.1 binding site that could serveto link the HSPG to the cortical cytoskeleton. In trans-fection studies, syndecan-2 promoted the early matu-ration of dendritic spines in rat hippocampal neurons(19), and this required the FYA motif in the C2 region.However, truncated forms of syndecan-2 lacking theFYA motif still accumulated in the normal location,implying that the FYA motif was not needed for local-ization but for the subsequent biological effect. Thecorrect distribution of syndecan-2 was lost when thecytoplasmic domain was truncated to only three amino


PERSPECTIVE SERIES



Table 1Known ligands for HSPGs, where sulfation/epimerization patterns havebeen determined

Ligand Biological effect

ECM molecules

laminin cell adhesionfibronectin cell adhesion

Growth factors/cytokines

FGF-1 and -2 dimerization, presentationto high affinity receptors

HGF sequestration of growth factorIL-8 sequestration of cytokine

PDGF-AA storage and secretion of growth factorPlatelet factor 4 stimulates factor activity

Protease inhibitors

Antithrombin III increases inhibitor activity

Details and original references can be found in ref. 5.

acids. Hence some region of the C1, V, or C2 subdo-main proximal to the FYA motif is responsible for cor-rect trafficking and targeting.

Recent experiments have identified a new syndecan-2binding partner, synbindin, which has a region thatbears limited homology to a PDZ-domain (11). It showsgreater homology with yeast proteins involved in theregulation of trafficking. Synbindin distribution is sim-ilar to that of syndecan-2 in the central nervous system,and both are concentrated in dendritic spines in maturehippocampal neurons in culture. Interestingly, the FYAsequence of syndecan-2 is needed for clustering of syn-bindin in dendritic spines (11), and this region isrequired for accelerated spine maturation induced bysyndecan-2, but not for syndecan-2 localization (19).These findings lead to the speculation that recruitmentof synbindin by syndecan-2 helps determine traffickingto maturing spines and postsynaptic structures.

Syndecan-3. This syndecan, unlike the others, appearsto bind ECM components poorly. Indeed, it may func-tion more in growth factor control of adhesion. Itsheparan sulfate chains can bind HB-GAM (also termedpleiotropin), which results in axonal extension(reviewed in refs. 3, 5). Concomitant with binding is theactivation of c-src and tyrosine phosphorylation of thecortical microfilament meshwork protein cortactin,leading to the formation of a complex containing syn-decan-3, c-src, cortactin, and tubulin. Competitionexperiments indicate that the interaction is throughthe C1 domain of syndecan-3, which is intriguing inlight of the homology between the C1 domains of thevarious syndecans. It has not been reported whetherbinding of other growth factors to other syndecans caninduce the formation of similar complexes and signal-ing cascades, but if this response is limited to syndecan-3, it may indicate that the V region can modulate theinteractions of C1 regions.

Syndecan-4. Syndecan-4 is the only syndecan that is awidespread component of focal adhesions (reviewed inrefs. 2–5). Different integrins are present in focal adhe-sions dependent on the substrate to which the cells areattached, but syndecan-4 is a common component irre-spective of which matrix component is used. The inser-tion of syndecan-4 into focal adhesions requires theactivity of PKC, as does spreading and focal adhesionformation, and its overexpression in CHO-K1 cellsincreases spreading and results in a more fibroblasticmorphology and increased formation of focal adhe-sions and stress fibers (17). Conversely, transfectionwith a truncated construct of syndecan-4 that termi-nates within the V region reduces spreading and focaladhesion and stress fiber formation (17). Again,changes in adhesion do not appear to be through alter-ations in integrin levels, which indicates that, as withsyndecan-2 and matrix assembly, the cytoplasmic tailof syndecan-4 modulates integrin-mediated processes.

Biochemical experiments (reviewed in refs. 1–5) shed

some light on how syndecan-4 may affect adhesion(Figure 2). Syndecan-4 fusion proteins, as well as shortsynthetic peptides with the sequence of the V region,can potentiate the activity of PKCα in vitro, even whenthe enzyme has been activated by diacylglycerol andphosphatidylserine. This effect requires oligomeriza-tion of the fusion proteins or peptides, and minor sub-stitutions in peptide sequence abolish both potentia-tion and oligomerization. Similarly, PIP2, which alonecan partially activate PKCα, augments oligomerization,and a ternary mixture with syndecan-4 results in anenzymatic activity higher than that seen with normalmediators. The presence of syndecan-4 and PIP2 alsoresults in the conventional PKCα isoform becomingcalcium-independent. PKC can be coclustered with syn-decan-4 in live cells, and both molecules are found ascomponents of focal adhesions. In addition, PKCα canbe both coimmunoprecipitated with syndecan-4 andcan bind to syndecan-4 peptide in vitro. However, thepresence of a ternary complex has not yet been identi-fied in vivo. A ternary complex may be important sincemany of the focal adhesion or cortical meshwork com-ponents are substrates for PKCα and have their con-formation altered by PIP2. Integrin ligation increasesthe levels of PIP2, which may stabilize syndecan-4oligomers induced by interaction with heparin-bindingdomains of matrix molecules, thus allowing PKCα tobind and phosphorylate specific components at nas-cent focal adhesions.

Early studies that indicated a need for both integrinligation and an interaction with the heparin-bindingdomain of fibronectin for stress fiber and focal adhesionformation, have recently been confirmed and extended(3). One study showed that, although fibronectin-nullfibroblasts can attach and spread on recombinant frag-ments of fibronectin that contained the RGD integrin-binding sequence, they fail to form focal adhesionsunless triggered by addition of heparin-binding domainsor by the clustering of syndecan-4 with antibody. Phar-macological activation of Rho GTPase can circumventthe need for heparin-binding domains or antibody clus-tering, whereas its inhibition prevented the induction offocal adhesion formation. Another study confirmed thatCHO cells that are incapable of glycanating proteogly-cans due to an enzyme defect lack focal adhesions unlesssyndecan-4 is overexpressed, which may force theoligomerization of the syndecan-4 core proteins in theabsence of heparan sulfate chains. Finally, fibroblastsfrom mice that lack syndecan-4 cannot respond to theheparin-binding region of fibronectin when spread onthe integrin-binding fragment (20). They do, however,respond to intact fibronectin, implying that sufficientconcentration, or a difference in presentation, may influ-ence biological activity.

Although syndecan-4 knockout mice appear normal,they do, as with syndecan-1 knockout mice, showdefects in wound healing (21), where responses to


PERSPECTIVE SERIES



growth factors and ECM regulate migration, division,and wound closure. Since transfection with differentsyndecan constructs result in such different effects(e.g., on adhesion or on matrix assembly), it is difficultto envisage how syndecan family members may substi-tute for each other, although this possibility has notbeen tested. Growth factor stimulation may circum-vent signaling through syndecans and, indeed, treat-ment with some growth factors can promote focaladhesion formation in cells on the integrin-bindingfragment of fibronectin. These observations lead to afinal interesting concept: Syndecans are involved inboth growth factor binding and interaction with ECMmolecules. It is now becoming clear that signaling fromadhesion through integrins and signaling throughgrowth factor binding can converge to coordinate cel-lular responses. Thus, in the absence of one set of sig-nals, the other may circumvent the missing cascades.

Three separate indications of convergence of signal-ing through syndecans on adhesion- and growth fac-tor–mediated responses should be highlighted. First,serum-starved quiescent 3T3 cells show phosphoryla-tion of the serine in the C1 region of syndecan-4. Thismodification, which can be reversed rapidly uponaddition of FGF-2, controls syndecan-4 oligomeriza-tion, interaction with PIP2, and ultimately the abilityto activate PKCα (reviewed in refs. 3–5). Second, syn-decan-3 binding to HB-GAM initiates a src-mediatedsignaling cascade, concomitant with a trigger for axon-al extension. Third, cells lacking focal adhesions whenspread on the integrin-binding fibronectin can be pre-vented from responding to heparin-binding domainsof fibronectin via syndecan-4 by inhibition of theGTPase Rho. The family of Rho, Rac, Cdc42 and RasGTPases is often involved downstream of many sig-naling cascades resulting from growth factors or adhe-sion and can regulate cytoskeletal organization, thusdetermining morphology. Furthermore, syndecanscan be shed from the cell surface, effectively providinga soluble competitor for cell surface syndecans, whichnot only retain capacity to bind growth factors, butalso can regulate protease activity, thus affectingmatrix stability (2, 3, 5). The ability of cell surface orsoluble syndecans to participate in a range of funda-mental metabolic activities endows these moleculeswith broad regulatory roles in cell-matrix andcell–growth factor interactions.

Implications for diseaseThe pathological roles of syndecan have not beenextensively studied, partly due to a lack of specific tools,such as antibodies that recognize core proteins. How-ever, many studies are now ongoing, and a few arebeginning to be reported in the literature (reviewed inref. 5). Thus, during wound healing, both syndecan-1and syndecan-4 are increased in expression, and in theircleavage and release in soluble form. The expression of

some growth factors, such as FGF-1, -2 and -7, is alsoincreased in wound healing, and these factors requireHSPG for full activity. Increased expression of synde-can-1 and -4 is seen during dermal and mucosal repair,and during response to arterial injury. Mice deficientin syndecan-1 show delayed wound healing. Further-more, since fibrosis can be thought of as misregulatedwound healing in which deposition progresses too far,too early, or with the wrong set of proteins, the effectsof syndecan-2 on matrix assembly, described above,may well contribute to many diseases involving afibrotic response.

Roles of syndecans in tumorigenesis are also underinvestigation. Tumorigenesis is associated with the lossof normal cell morphology, a gain of invasiveness,decreased adhesion, and abnormal proliferation con-trol. Syndecans have been implicated in the regulationof all these fundamental processes. Syndecan-1 isreduced when epithelial cells are transformed, and sev-eral studies now indicate that its decrease may be aprognostic marker in some tumors. Interestingly, arecent study of mice transgenic for Wnt-1 showed thatthey had increased levels of mammary tumor forma-tion. However, if the mice were additionally syndecan-1 deficient, tumor formation was reduced (see ref. 6and references therein).

ConclusionsSyndecans participate in a range of fundamental bio-logical processes. They are unique in that their GAGchains allow interactions with ECM proteins, growthfactors, metalloproteinases, lipases, and even pathogens.Improved sequencing techniques have led to the con-cept of highly specific interaction sites within regulateddomains along the polysaccharide chains. It will be fas-cinating to discover whether the GAG chains of indi-vidual syndecans contain differing microdomains thatregulate specific interactions. Syndecans not only func-tion at the cell surface, but, since their shed ecto-domains retain function, they may be a type of matri-cellular protein (22) that regulates both growth factorand matrix interactions.

In addition to the ligand-binding GAG chains, thetransmembrane core proteins of syndecans can alsohave biological activity. Some signaling functions ofsyndecans have already been identified and more aresure to follow. Those identified to date confirm the reg-ulatory roles of these molecules, in that they modifysignaling cascades initiated by primary receptors orincrease ligand interaction with primary receptor. Sofar, most core protein functions have been attributedto the cytoplasmic domain, but there is increasing evi-dence for a role for their ectodomains in adhesion. Thenext few years should be interesting, with increasedinterest in the field leading to new discovery. Do cellsurface syndecans physically associate with other sur-face molecules, such as integrins or growth factor


PERSPECTIVE SERIES



receptors, to modulate activity? If so, are the interac-tion sites extracellular, transmembrane, or intracellu-lar? Preliminary data (23) indicate association of syn-decan-4 with both cytoskeletal components andintegrins, and future studies on how these interactionsare controlled in development and disease will also bean exciting area of research.

AcknowledgmentsI would like to thank friends and colleagues for usefuldiscussions. This work was supported by NIH grantDK 54605.

1. Rapraeger, A.C. 2000. Syndecan-regulated receptor signaling. J. Cell Biol.149:995–997.

2. Bernfield, M., et al. 1999. Functions of cell surface heparan sulfate pro-teoglycans. Annu. Rev. Biochem. 68:729–778.

3. Couchman, J.R., Chen, L., and Woods, A. 2001. Syndecans and cell adhe-sion. Int. Rev. Cytol. In press.

4. Woods, A., and Couchman, J.R. 2000. Integrin modulation by lateralassociation. J. Biol. Chem. 275:24233–24236.

5. Tumova, S., Woods, A., and Couchman, J.R. 2000. Heparan sulfate pro-teoglycans on the cell surface: versatile coordinators of cellular func-tions. Int. J. Biochem. Cell Biol. 32:269–288.

6. Alexander, C.M., Hinkes, M.T., and Bernfield, M. 2000. Syndecan-1 isrequired for Wnt-1 induced tumorigenesis but not for morphogenesisof mouse mammary epithelia. Nat. Genet. 25:329–332.

7. Fuki, I.V., Meyer, M.E., and Williams, K.J. 2000. Transmembrane and cyto-plasmic domains of syndecan mediate a multi-step endocytic pathwayinvolving detergent-insoluble membrane rafts. Biochem. J. 351:607–612.

8. Grootjans, J.J., et al. 1997. Syntenin, a PDZ protein that binds syndecancytoplasmic domains. Proc. Natl. Acad. Sci. USA. 94:13683–13688.

9. Hsueh, Y.P., et al. 1998. Direct interaction of CASK/LIN-2 and syndecanheparan sulfate proteoglycan and their overlapping distribution in neu-

ronal synapses. J. Cell Biol. 142:139–151.10. Gao, Y., Li, M., Chen, W., and Simons, M. 2000. Synectin, syndecan-4

cytoplasmic domain binding PDZ protein, inhibits cell migration. J. Cell.Physiol. 184:373–379.

11. Ethell, I.M., Hagihara, K., Miura, Y., Irie, F., and Yamaguchi, Y. 2000. Syn-bindin, a novel syndecan-2-binding protein in neuronal dendritic spines.J. Cell Biol. 151:53–67.

12. Fialka, I., et al. 1999. Identification of syntenin as a protein of the apicalearly endocytic compartment in Madin-Darby canine kidney cells. J. Biol.Chem. 274:26233–26239.

13. Wang., L.-H., Kalb, R.G., and Strittmatter, S.M. 1999. A PDZ protein reg-ulates the distribution of the transmembrane semaphorin, M-SemF. J.Biol. Chem. 274:14137–14146.

14. Granes, F., Urena, J.M., Rocamora, N., and Vilaro, S. 2000. Ezrin linkssyndecan-2 to the cytoskeleton. J. Cell Sci. 113:1267–1276.

15. Baciu, P.C., et al. 2000. Syndesmos, a protein that interacts with the cyto-plasmic domain of syndecan-4, mediates cell spreading and actincytoskeletal association. J. Cell Sci. 113:315–324.

16. Klass, C.M., Couchman, J.R., and Woods, A. 2000. Control of extracellu-lar matrix assembly by syndecan-2 proteoglycan. J. Cell Sci. 113:493–506.

17. Longley, R.L., et al. 1999. Control of morphology, cytoskeleton andmigration by syndecan-4. J. Cell Sci. 112:3421–3431.

18. Kusano, Y., et al. 2000. Participation of syndecan 2 in the induction ofstress fiber formation in cooperation with integrin α5β1: structuralcharacteristics of heparan sulfate chains with avidity to COOH-termi-nal heparin-binding domain of fibronectin. Exp. Cell Res. 256:434–444.

19. Ethell, I.M., and Yamaguchi, Y. 1999. Cell surface heparan sulfate pro-teoglycan syndecan-2 induces the maturation of dendritic spines in rathippocampal neurons. J. Cell Biol. 144:575–586.

20. Ishiguro, K., et al. 2000. Syndecan-4 deficiency impairs focal adhesionformation only under restricted conditions. J. Biol. Chem. 275:5249–5252.

21. Echtermeyer, F., et al. 2001. Delayed wound repair and impaired angio-genesis in mice lacking syndecan-4. J. Clin. Invest. 107:R9–R14.

22. Bornstein, P. 1995. Diversity of function is inherent in matricellular pro-teins: an appraisal of thrombospondin. J. Cell Biol. 130:503–506.

23. Greene, D.K., Woods, A., Tumova, S., and Couchman, J.R. 2000. Synde-can-4 associations with the microfilament cytoskeleton and integrins.Mol. Biol. Cell 11:349a. (Abstr.)


PERSPECTIVE SERIES



Documents

Syndecans: transmembrane modulators of adhesion and ...dm5migu4zj3pb.cloudfront.net/manuscripts/12000/12802/JCI...can-2 interaction may be through the C1 domain. Again, since the C1