Journal of Cell Science 113, 493-506 (2000) 493 Printed in Great Britain © The Company of Biologists Limited 2000 JCS0705

Control of assembly by syndecan-2

Carmen M. Klass, John R. Couchman and Anne Woods* Department of Cell Biology and Cell Adhesion and Matrix Research Center, University of Alabama at Birmingham, Birmingham, AL 35294, USA *Author for correspondence (e-mail: [email protected])

Accepted 19 November 1999; published on WWW 19 January 2000

SUMMARY

Extracellular matrix (ECM) deposition and organization is β1 . The matrix assembly defect was at the cell maintained by transmembrane signaling and play surface, since S2∆S cells also lost the ability to rearrange major roles. We now show that a second transmembrane laminin or fibronectin substrates into fibrils and to bind component, syndecan-2 proteoglycan, is exogenous fibronectin. Transfection of activated αIibαL∆β3 pivotal in matrix assembly. Chinese Hamster Ovary (CHO) integrin into α5-deficient CHO B2 cells resulted in cells were stably transfected with full length (S2) or reestablishment of the previously lost fibronectin matrix. truncated syndecan-2 lacking the C-terminal 14 amino However, cotransfection of this cell line with S2∆S could acids of the cytoplasmic domain (S2∆S). No differences in override the presence of activated integrins. These results the amount of matrix assembly were noted with S2 cells, suggest a regulatory role for syndecan-2 in matrix but those expressing S2∆S could not assemble laminin or assembly, along with previously suggested roles for fibronectin into a fibrillar matrix. The loss of matrix activated integrins. formation was not caused by a failure to synthesize or externalize ECM components as determined by metabolic Key words: Syndecan, Heparan sulfate proteoglycan, Laminin, labeling or due to differences in surface expression of α5 or , Integrin

INTRODUCTION association site and growth of fibrils in vivo (Hocking et al., 1994; Sechler et al., 1997). Treatment of cells with antibodies Regulation of ECM structure and composition is crucial to against α5 inhibited matrix assembly (Akiyama et al., 1989; tissue integrity. However, the molecular processes that control Fogerty et al., 1990), and overexpression or reduction of α5β1 deposition of ECM are still poorly understood. Many studies in CHO cells resulted in increased fibronectin matrix assembly have concentrated on the mechanisms of insertion and (Giancotti and Ruoslahti, 1990) or its loss (Schreiner et al., organization of soluble fibronectin into a disulfide-bonded 1989), respectively. Matrix formation could be restored in insoluble pericellular fibrillar network. This is a multistep CHO B2 cells that lack α5 integrin by transfection with α5 (Wu process, which is under cellular control, leading to et al., 1993) or activated αIIbβ3 (lacking the GFFKR region of characteristic patterns of fibrillar arrays dependent on cell type αIIb; Wu et al., 1995b). Transfection of these cells with α3β1 (Bultmann et al., 1998; Christopher et al., 1997; Sechler et al., also increased fibronectin deposition, possibly indirectly 1997 and references therein). When soluble fibronectin is through increased deposition of entactin (Wu et al., 1995a). added to cells, it first binds to cell surface sites in a However, although expression of α4β1 and αvβ1 integrins in deoxycholate-soluble and reversibly bound pool. Over time, CHO B2 cells allowed them to adhere to fibronectin, they did this becomes a deoxycholate-insoluble fibrillar matrix not assemble a matrix (Wu et al., 1995c; Zhang et al., 1993). multimerized via cell-mediated disulfide exchange (McKeown- Although the ‘cell-binding’ domain of fibronectin Longo and Mosher, 1984; Chen and Mosher, 1996) or factor (particularly the III-10 module) is needed for fibrillogenesis, XIIIa-mediated cross-linking (Barry and Mosher, 1989). heparin-binding domains of the molecule are also required Endogenous fibronectin may be similarly polymerized, with (Bultmann et al., 1998; Christopher et al., 1997; Hocking et al., initial deposition acting as a nucleus for fibrillogenesis, 1994; McDonald et al., 1987; Schwarzbauer, 1991; Woods et forming small aggregates that move centripetally over the cell al., 1988). Addition of the amino-terminal 70 kDa region of surface, enlarging to form small fibrils and eventually larger fibronectin, which contains both heparin- and collagen-binding fibers around the cells (Ljubimov and Vasiliev, 1982; Sechler activities, will block the binding and assembly of intact et al., 1997). fibronectin to putative ‘matrix assembly sites’ (Dzamba et al., The major integrin involved in the deposition of a fibrillar 1994). Dimeric recombinant fibronectin molecules containing matrix in many cells is α5β1 (Akiyama et al., 1989; Fogerty et this region are incorporated into matrix fibrils when added to al., 1990; Wu et al., 1993), which binds to the RGD recognition cells (Sottile and Wiley, 1994). The 70 kDa binding activity site in fibronectin, leading to exposure of a fibronectin self- appears to reside in type I repeats 1-5 (i.e. the heparin-binding 494 C. M. Klass, J. R. Couchman and A. Woods activity), but its binding and incorporation into fibrils requires CHO-K1 (LeBaron et al., 1988) and CHO B2 cells were cultured in prior occupancy of α5β1 by the ‘cell’-binding domain (Dzamba Ham’s F12 medium (Mediatech, Charlottesville, VA) containing 10% et al., 1994; Fogerty et al., 1990) confirming the primary role fetal calf serum (FCS, Intergen, Purchase, NY). Vectors pcDNA3 and of this integrin. pcDNA3.1Zeo were purchased from Invitrogen (San Diego, CA). Cell surface can augment the interaction Bovine plasma fibronectin and vitronectin were prepared as of cells with fibronectin by glycosaminoglycan binding previously described (Miekka et al., 1982; Yatogho et al., 1988). All antibodies were diluted in phosphate-buffered saline (PBSÐ: to heparin-binding domains (reviewed by Carey, 1997; (8.0 g NaCl, 0.2 g KCl, 0.2 g KH2PO4, 1.15 g Na2HPO4 per liter, pH Woods and Couchman, 1998). Indeed, cells that lack 7.2). Rabbit polyclonal antibodies against laminin-1 from EHS-tumor glycosaminoglycan synthesis cannot adhere normally to and against bovine plasma fibronectin have been previously described fibronectin (LeBaron et al., 1988) nor establish a fibronectin (McCarthy et al., 1993). F-Actin was detected with Texas Red- matrix (Chung and Erickson, 1997). The four-member conjugated phalloidin (Molecular Probes, Eugene, OR) at a dilution syndecan family of heparan sulfate proteoglycans are a second of 1:1000. Paxillin antibodies (dilution 1:100) were purchased from class of transmembrane receptors for ECM components Zymed (San Francisco, CA). The mouse monoclonal antibody against (reviewed by Bernfield et al., 1992; Carey, 1997; Woods and hamster α5 integrin, PB1 (Brown and Juliano, 1988), was a generous gift from Dr R. L. Juliano, University of North Carolina, Chapel Hill, Couchman, 1998). Syndecan-4 (David et al., 1992b; Kojima et β al., 1992) may augment integrin-mediated cell adhesion to NC. Monoclonal antibodies against integrin 1 subunits were isolated from culture supernatants of 7E2 hybridoma obtained from the ECM and determine cytoskeletal organization (Longley et al., Developmental Studies Hybridoma Bank maintained by the 1999; Woods and Couchman, 1998), which correlates with Department of Pharmacology and Molecular Sciences at Johns matrix formation (Ali and Hynes, 1977; Woods et al., 1988; Hopkins University School of Medicine, Baltimore, MD, and the Wu et al., 1996). Syndecan-4 may signal through Department of Biology at the University of Iowa, Iowa City, IA, under kinase C (Baciu and Goetinck, 1995; Oh et al., 1997a,b), and contract number N01-HD-2-3-3144 from NICHD. A polyclonal protein kinase C activity is also needed for matrix assembly antibody against a glutathione-S-transferase (GST; Pharmacia (Somers and Mosher, 1993). However, any role for syndecans Biotech, Pescataway, NJ) fusion protein containing the extracellular in matrix assembly has not been investigated. and transmembrane domains of rat syndecan-2 (designated S2∆R; Syndecan-4 is only a minor component in cells, including Fig. 1) was raised in a New Zealand White rabbit as previously fibroblasts that produce ECM (Kim et al., 1994). The major described (Koff et al., 1991). The antiserum was affinity purified for immunofluorescence experiments (Koff et al., 1992) and shown to be syndecan in fibroblasts is syndecan-2 (David et al., 1992b, specific for syndecan-2 by ELISA, western blotting and 1993), whose function has not been well investigated. immunostaining (data not shown). A chicken antibody against the Syndecan-2 core protein can be phosphorylated both in vitro same construct was made by AVES Laboratories, Inc. (Tigard, OR) by protein kinase C (Oh et al., 1997c; Prasthofer et al., 1995), and the purified IgY was used at 1:75 dilution. Primary antibodies and in vivo (Itano et al., 1996). The phosphorylation site(s) were detected with F(ab′)2 fragments of goat anti-mouse, anti-rabbit, appear to be serine197 and serine198. All syndecans have highly or anti-chicken conjugated to fluorescein isothiocyanate (FITC) or conserved cytoplasmic domains, with constant regions Texas Red (Cappel, Durham, NC) diluted 1:50. proximal (C1) and distal (C2) to the membrane, separated by DNA and plasmid construction a region (V) unique to each family member (Bernfield et al., 1992; Woods and Couchman, 1998). In syndecan-4, the V Full-length cDNA constructs for wild-type and truncated syndecan-2 in pcDNA3 vectors (Invitrogen) were constructed in collaboration region mediates protein kinase C binding and activation (Oh et with Drs J. T. Gallagher and G. J. Cowling (Paterson Cancer Institute, al., 1997a,b), and may, therefore, be responsible for the role of Manchester, UK). Fig. 1 shows the cytoplasmic domain of syndecan- this syndecan in integrin-dependent adhesion. In syndecan-2, 2 and its closest homologue, syndecan-4. The C1 and C2 homology the V region forms the phosphorylation site (Oh et al., 1997c). regions, the V region unique to each syndecan and the sequences for We, therefore, reasoned that this site in syndecan-2 may play the partial (S2∆S, S4∆I) and full (S2∆R) truncations of the a role in augmenting integrin-mediated matrix assembly. cytoplasmic domains are indicated. A mammalian vector was cDNA constructs encoding full length syndecan-2 core protein, constructed by GC clamp polymerase chain reaction of rat liver partial cytoplasmic truncation (to serine197) or total syndecan-2 (clone 17; Pierce et al., 1992) in Bluescript vector cytoplasmic domain deletion, were expressed in CHO-K1 cells (Stratagene, La Jolla, CA), digested with the appropriate restriction and effects on matrix assembly were monitored. We also used endonucleases, and ligated into pcDNA3 (Invitrogen) using the 5′EcoRI and 3′XbaI sites. The oligonucleotides used for PCR of the constructs encoding equivalent syndecan-4 core and syndecan-2 cDNA were 5′-ggggaattcctcctatcttctctaggc-3′ and 3′- show that syndecan-2 specifically controls the assembly of gggtctagaagagacactaagtggag-5′. Syndecan-2 cDNA encoding both fibronectin and laminin matrices. truncated cytoplasmic domains were constructed using the identical 5′ primers but the following modified 3′ primer containing a new stop codon (underlined) adjacent to the XbaI site of pcDNA3: 3′- gggtctagattagcggtacaccaacaacag-5′ (S2∆R: full truncation) and 3′- MATERIALS AND METHODS gggtctagattaggacggtttgcgttctcc-5′ for the truncation ending at serine197 (S2∆S). A partially truncated syndecan-4 construct ending Supplies, cell culture, antibodies and matrix purification at isoleucine169 (designated S4∆I) was generated as described All general chemicals were from Sigma Chemical (St Louis, MO). (Longley et al., 1999). All mutated constructs were confirmed by Coverslips and cultureware were from Fisher (Altlanta, GA). CHO B2 sequencing. Syndecan-2 was also ligated into the pcDNA3 vector in cells, which are α5β1-deficient (Schreiner et al., 1989), and CHO B2 the antisense orientation. A 5′ fragment of 150 bp was excised using cells transfected with activated αIIbαL∆β3 which has the α subunit the polylinker HindIII site and an internal HindIII site. Excision added cytoplasmic membrane proximal sequence GFFKR deleted (O’Toole a unique BamHI site to the 5′-end, the construct was religated into et al., 1994) were kind gifts from Dr C. Wu (UAB, Birmingham, AL) pcDNA3 vector cleaved with HindIII, and fragment orientation was and Dr M. H. Ginsberg (The Scripps Research Institute, La Jolla, CA). confirmed using the BamHI restriction digests. Syndecan-2 controls matrix assembly 495

S2: RMRKKDEGSYDL GERKPSSAAYQ KAPTKEFYA Treatment of cells with conditioned medium C1 V C2 Conditioned medium was collected every other day from S2∆S transfected CHO-K1 cells, centrifuged (1,000 g, 10 minutes) and S2∆S: RMRKKDEGSYDL GERKPS stored at +4¡C. This was used daily for two passages to replace medium in cultures of CHO-K1 stably transfected with the empty pcDNA3 vector, before staining for laminin and fibronectin as described below. ∆ S2 R: R Immunohistochemical analysis and microscopy To monitor endogenous matrix assembly cells were grown on glass coverslips (10 mm diameter; Fisher) for 1-2 days (sparse cultures) or 3-5 days (confluent cultures) in culture medium. In some cases, S4: RMKKKDEGSYDL GKKPIYKK APTNEFYA exogenous bovine fibronectin (260 µg/ml) was added 18 hours after seeding, before fixation and staining at 3 days. In other experiments, cells were incubated on coated coverslips in culture medium for 24 hours. For experiments on reorganization of substrates into a fibrillar S4∆I: RMKKKDEGSYDL GKKPI matrix, cells were grown on laminin or fibronectin coated coverslips for 4-24 hours. Coating of coverslips with 5 µg of laminin (Sigma) or EC TM CP vitronectin, 10 µg of fibronectin or 100 µg of type I collagen (Vitrogen 100, Collagen Biomaterials, Palo Alto, CA) and blocking with 0.1% Fig. 1. Cytoplasmic domain sequences of wild-type and bovine serum albumin have been previously described (Woods and truncated syndecan-2. All syndecan cytoplasmic domains have high Couchman, 1994). Cultures were fixed (30 minutes) with 3% homology in membrane proximal and distal regions (C1 and C2) glutaraldehyde (Sigma) for phase contrast microscopy or 3.5% freshly with a unique variable (V) region in the center. EC, extracellular prepared paraformaldehyde in PBS− (20 minutes) for fluorescence. domain; TM, transmembrane domain; CP, cytoplasmic domain; S2 Permeabilization, when indicated, was by incubation (10 minutes) with syndecan-2; S4, syndecan-4. 0.1% Tween-20 (Bio-Rad, Richmond, CA) in PBS−. Labeling with primary and secondary antibodies, or with Texas Red-phalloidin was as previously described (Woods and Couchman, 1994). For double labeling, cells were incubated with a mixture of chicken anti-rat Transfections and selection for high expressors syndecan-2 and rabbit anti-fibronectin, followed by FITC goat anti- CHO-K1 cells were transfected either with empty pcDNA3 vector chicken Ig and Texas Red goat anti-rabbit IgG. Cells were monitored (designated pcDNA3) or pcDNA3 vector containing wild-type using a Nikon Optiphot microscope and photographed with Ilford HP5 syndecan-2 (S2), the partially (S2∆S) or fully (S2∆R) truncated film. Where amounts of extracellular labeling were compared, constant syndecan-2, antisense syndecan-2 (S2A), or truncated syndecan-4 time of exposure, film development and print conditions were used. (S4∆I). 7.5 µg plasmid cDNA prepared by miniprep were mixed with 15 µg lipofectamine (Life Technologies, Gaithersburg, MD) in 2.5 ml Biosynthetic labeling of cells and pulse chase of serum-free medium and incubated with CHO-K1 cells as per experiments suppliers instructions. Several independent stable transfectants for pcDNA3 or S2∆S cells (6×106 cells per well) at passage 3-5 were each construct were isolated over three passages in the presence of seeded in 6-well plates. After attachment and spreading for 2 hours, G418 (Life Technologies) at 500 µg/ml active gentamycin and gave cells were starved overnight in Hanks’ balanced salt solution identical results. Expression levels decreased with passage, so all (Mediatech, Charlottesville, PA). The next morning cells were pulsed experiments except Fig. 9 were performed at passages 3-6 from initial with 100 µCi [35S]methionine/cysteine (ICN Biomedical Flow transfection or fluorescence activated cell sorting (FACS) selection Laboratories) for 1 hour, and ‘chased’ with Ham’s F12 medium plus (see below). CHO B2 cells were transfected with the S2∆S construct 10% FBS for 0, 30, 60, 90 minutes or 24 hours to allow synthesis, using the pcDNA3 vector and selected with G418. Since αIIbαL∆β3- export and incorporation into any ECM which was formed. Very little transfected CHO B2 were already G418 selected, the S2∆S construct ECM is visible by immunofluorescence at 90 minutes. The was subcloned into pcDNA 3.1 vector containing a Zeocin resistance supernatants were collected and the cell layers lysed with RIPA-buffer cassette using EcoRI- and XbaI-restriction sites. (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% NP-40, 0.1% SDS, Transfected cell lines showed phenotypic drift over time, and CHO 0.5% sodium deoxycholate, 0.2 mM PMSF, 1 mM benzamidine-HCl). B2 transfection was inefficient, so high expressors were selected for Supernatants and cell lysates were immunoprecipitated (Woods and some experiments (Figs 2, 10) by FACS. Confluent cell monolayers Couchman, 1994) with rabbit anti-laminin or anti-fibronectin. were treated for 1 minute with 3 ml of cell dissociation buffer (Life Polypeptides were resolved by 3-15% SDS-PAGE reducing gradient Technologies). All further steps were performed at +4¡C. Cells were gel, and gels were impregnated with ENHANCE (NEN), dried, and suspended in 1 ml of Ham’s F12 plus 10% FBS, centrifuged at 1000 exposed to film for 5 days at –80°C before development. Fluorographs g (10 minutes), washed twice with PBS− plus 5% FBS and incubated were quantified using the IP Lab Gel software by Macintosh for with rabbit anti-syndecan-2 antiserum (diluted 1:50) for 45 minutes. densitometry and gel analysis. Data values are displayed in pixels. Cells were washed twice with PBS− plus 5% FBS, centrifuged (1000 g, 10 minutes), incubated with FITC-conjugated F(ab′)2 fragment of Measurement of cell surface syndecan-2 and integrins goat anti-rabbit IgG (Cappel, diluted 1:50), and washed three times To examine syndecan-2 cell surface expression, FACS analysis was as above. Cells incubated only with secondary antibody served as performed as described above. For integrin β1 cell surface expression controls. FACS was performed by the Flow Cytometry Core Facility levels, cells were harvested with 0.05% trypsin, washed with culture of the UAB Multipurpose Arthritis and Musculoskeletal Disease medium, followed by FACS buffer (1× PBS−, 5% FCS, 0.1% sodium Center using a Becton Dickinson FACS caliber machine and Cell azide) and incubated with mouse monoclonal antibodies 7E2 or PB1 Quest computer program. Cells comprising the 5% highest expressors for 45 minutes. Cells were washed three times with FACS buffer, were collected in Ham’s F12 medium and 10% heat-inactivated FBS, incubated with FITC-conjugated anti-mouse IgG for 45 minutes, grown to confluency, and stocks were frozen at passage 2-3 after washed 3 times with FACS buffer, and fixed in 1% paraformaldehyde selection. in FACS buffer. 496 C. M. Klass, J. R. Couchman and A. Woods

Quantitation of syndecan-2 at low levels in CHO-K1 cells, distributed mainly diffusely To quantitate syndecan-2 levels in cells transfected with the antisense over the surfaces (Fig. 2A) as previously seen for rat embryonic syndecan-2 construct, identical numbers of CHO-K1, S2 and S2A cells fibroblasts (Woods and Couchman, 1994). It did not appear to were grown for 3 days to confluency in 100 mm dishes. Cells were correlate in double staining with the distribution of washed with PBS+ and scraped into 1 ml of 100 mM NaCl, 1 mM extracellular matrix in CHO-K1 (see Fig. 4A) or rat embryo CaCl2, 50 mM HEPES, pH 7.0 (David et al., 1992a) containing 1 mM fibroblasts (not shown). CHO-K1 transfected with wild-type benzamidine-HCl, 1 mM phenylmethylsulfonyl fluoride, 10 mM N- ∆ µ (S2) or partially truncated (S2 S) syndecan-2 had variably ethylmaleimide and 10 g/ml leupeptin, centrifuged at 800 g and increased cell surface labeling, which was pronounced in resuspended in 50 µl of the above buffer. Samples for digestion were FACS-selected cells (Fig. 2B,C). This was present in a linear incubated with 2.5 mU heparitinase (EC.4.2.2.8) and 5 mU ∆ chondroitinase ABC (EC.4.2.2.4) (Seikagaku America) overnight at punctate pattern. In S2 S cells, ruffles, blebs, and other surface 37°C. After addition of 5× SDS sample buffer with 2-mercaptoethanol, protrusions, which are increased, were intensely labeled (Fig. samples were boiled for 3 minutes. Samples were applied to a 1.5 mm 2C). Cells transfected with the full (S2∆R) truncation-construct thick SDS-PAGE gel, electrotransferred at 75 V for 2.5 hours, and also showed some increased labeling, but, even after FACS immunoblotted onto Bio-Rad nitrocellulose membrane as previously selection, phenotypic drift occurred rapidly within 3-4 described (Woods and Couchman, 1994). The membranes were fixed − passages leaving mainly very low expressors (Fig. 2D) Cells for 30 minutes with PBS containing 0.05% glutaraldehyde and transfected with syndecan-2 antisense (S2A) vector showed ° blocked in TBS plus 5% non-fat dried milk overnight at +4 C. Primary very little labeling for syndecan-2 at the cell surface (Fig. 2E). antibody was affinity-purified rabbit anti-pan-syndecan antibody (0.5 There was no difference in immunofluorescent labeling or µg/ml) generated against the carboxy-terminal 10 amino acids of syndecan-1. This recognizes all syndecan core proteins (Ott and FACS analysis for syndecan-2 in CHO-K1 cells transfected Rapraeger, 1998) and was a generous gift from Dr A. C. Rapraeger with empty vector pcDNA3 (not shown). Increased cell surface (University of Madison, Wisconsin). After washing in TBS plus 0.1% expression of syndecan-2 was seen in the S2, S2∆S and S2∆R Tween-20, bound antibody was visualized with affinity-purified goat cells by FACS analysis (Fig. 3). Although the range of anti-rabbit IgG conjugated to horseradish peroxidase (dilution 1:5000) expression was large within populations, the mean (Bio-Rad), and chemiluminescence (ECL, Amersham) as per fluorescence of cells transfected with any syndecan-2 sense manufacturer’s instructions. Since endogenous syndecan-2 levels in construct was always well above that of control pcDNA3- CHO cells are very low, blots were exposed for 30 minutes. transfected cells or wild-type CHO-K1 (Fig. 3). Therefore, Densitometric analysis of syndecan-2 core protein was normalized to constructs were expressed at the cell surface, although the cytokeratin levels determined by reprobing the same blots. Blots were increase in S2∆R cells was less than that in S2∆S or S2 cells stripped (100 mM 2-mercaptoethanol, 2% SDS, 62.5 mM Tris-HCl, pH 6.7) for 30 minutes at 50°C, washed twice with PBST, and checked (Figs 2, 3). Since endogenous syndecan-2 levels were already for lack of residual label by chemiluminescence. After reblocking, the very low in CHO-K1 cells, quantitative immunoblots of cell blot was reprobed with monoclonal mouse anti-cytokeratin antibody lysates of CHO-K1 and S2A cells were performed instead of (Zymed, San Francisco, CA) diluted 1:50 for 1 hour, followed by goat FACS to demonstrate downregulation of syndecan-2 in the anti-mouse IgG conjugated to horseradish peroxidase (dilution 1:5000) S2A cells (Fig. 3D). The extracellular domains of syndecans (Bio-Rad) as secondary antibody, and chemiluminescence (ECL, differ between species, but their cytoplasmic domains are Amersham). The experiment was repeated three times. Signals were highly homologous (Bernfield et al., 1992). Due to limited assessed using the Image Analysis Software of model GS-670 reactivity of the polyclonal anti-rat syndecan-2 (against ± densitometer (Bio-Rad). Results are shown as mean standard error. extracellular and transmembrane domains) to endogenous Quantitation of cell surface laminin hamster syndecan-2, immunoblotting was performed with a Identical numbers of CHO-K1, S2 and S2∆S cells were grown to polyclonal antibody that recognized all syndecan cytoplasmic confluency (4 days) in two 100 mm dishes per cell line. Cell domains (Ott and Rapraeger, 1998). Samples were treated with monolayers of one dish per cell line were surface-labeled with 0.5 heparitinase and chondroitinase ABC to ensure complete mg/ml EZ-Link Sulfo-NHS-LC-Biotin (Pierce, Rockford, IL) in digestion of glycosaminoglycan chains. Fig. 3D shows that PBS− for 2 hours at +4°C. Then biotinylated and corresponding non- syndecan-1 was the major syndecan in CHO-K1 cells, with biotinylated cell layers were scraped into equal volumes of ice-cold minor amounts of syndecan-2 and syndecan-4. Syndecan-2 RIPA buffer (see above). Total protein content of the non-biotinylated core protein, represented by an approximately 48 kDa cell lysates were measured using the BCA Protein Assay Kit (Pierce) polypeptide as seen in previous studies (Ott and Rapraeger, µ and 500 g of each corresponding cell lysate were transferred into a 1998; Pierce et al., 1992), was reduced in S2A cells, but 1.5 ml Eppendorf tube. Immunoprecipitations were performed using increased in S2 cells. Quantitative analysis normalizing to rabbit anti-laminin serum (see above) after a preclearing step with preimmune rabbit serum. Polypeptides were resolved by 3-15% SDS- cytokeratin is shown in Fig. 3D. PAGE reducing gradient gel and transferred to nitrocellulose membranes (Bio-Rad, see above). After blocking, biotinylated Matrix assembly in CHO-K1 and transfected cells laminin chains were visualized with horseradish peroxidase CHO-K1 or pcDNA3 cells assemble an extracellular fibrillar conjugated streptavidin (dilution 1:2000) and chemiluminescence matrix that contains both laminin and fibronectin. The (Amersham) as per the manufacturer’s instructions. Results were distributions of laminin and fibronectin were similar in non- quantitated as above. transfected cells, and altered coordinately with transfection, so results are shown here only for laminin. In sparse cultures of RESULTS CHO-K1 or pcDNA3 (not shown) cells, a three-dimensional network of matrix fibers and smaller fibrils were visualized by Syndecan-2 expression in wild-type and transfected staining non-permeabilized cells with anti-laminin (Fig. 4A). CHO-K1 cells Additional label resembling the distribution of endoplasmic Immunofluorescent labeling showed that syndecan-2 is present reticulum/Golgi was evident in permeabilized cells (Fig. 4B). Syndecan-2 controls matrix assembly 497

Fig. 2. Syndecan-2 cell surface distribution in CHO-K1 and transfected cells. CHO-K1 cells (A) contain little syndecan-2, which is mainly diffuse over the whole cell surface with some brighter spots, possibly representing aggregates. S2 (B) and S2∆S (C) cells showed a linear punctate pattern together with intense labeling of ruffles, blebs and other surface protrusions that are out of the plane of focus (arrows in C). S2∆R (D) cells showed punctate staining (arrows) while S2A (E) cells showed reduced labeling for syndecan- 2 in comparison to CHO-K1. (B-D) FACS-selected cells to demonstrate syndecan-2 distributions when expressed at high levels. Cells were not permeabilized. Bar, 10 µm.

labeling pattern (Fig. 4D) was similar to that of CHO-K1 cells. In contrast, S2∆S cells showed an almost complete lack of extracellular labeling for laminin (Fig. 4E) or fibronectin (not shown) except for areas of cell damage during processing, although staining of permeabilized S2∆S cells (Fig. 4F) revealed intracellular label similar to that of CHO-K1 and S2 cells. Overexpressing the S2∆R construct in non-selected cell populations showed only a slightly reduced fibrillar matrix (Fig. 4G). However, analysis of these cells showed few that stably expressed high levels of this construct. Most cells were low expressors (Figs 2, 3). The same reduction in laminin and fibronectin deposition by high expressors of S2∆R (detected by double immunostaining of syndecan-2 and matrix glycoproteins; data not shown) was seen as for S2∆S. It has been shown that syndecans with a totally truncated cytoplasmic domain may be abnormally processed, while partially truncated constructs may not be (Miettinen et al., 1994). Therefore, only the S2∆S cells were monitored further. Cells expressing a syndecan-4 core protein that is truncated in its V region (S4∆I) retained the ability to assemble a normal matrix (Fig. 4H). In very dense cultures of CHO-K1 (Fig. 5A) or S2 (not shown) cells, an elaborate ECM was assembled, but S2∆S cells established only patchy, sparse, non-fibrillar deposits of laminin and fibronectin-containing matrix at the basal surface with a few fibrils between cells (Fig. 5B). S2A cells also had a greatly reduced, but not absent, fibrillar matrix (Fig. 5C). This intermediate result is consistent with an approximate one third reduction in detectable syndecan-2 (Fig. 3D). Thus, transfection with truncated syndecan-2 or antisense-syndecan- 2 constructs reduces endogenous matrix assembly. Double staining of CHO-K1, S2 and S2∆S cells showed no colocalization of fibronectin and syndecan-2 (not shown). Quantitation of biotinylated, cell surface laminin showed equal amounts in CHO-K1 and S2 cells, but S2∆S cells had 50-65% less (Fig. 6). Cells use specific integrin receptors when adhering to different extracellular matrix glycoproteins and vitronectin receptors are normally used by cells cultured in the presence of serum (Fath et al., 1989). To determine if ligation of various integrins could override the S2∆S matrix defect, CHO-K1 and S2∆S cells were seeded onto laminin-1, fibronectin, vitronectin and type I collagen substrates. Both cell lines spread very well on fibronectin and vitronectin, but poorly on laminin-1 and S2 cells had a more epithelial cell morphology, but the amount type I collagen. Staining for endogenously-derived laminin or of either laminin (Fig. 4C) or fibronectin (not shown) fibronectin (not shown) showed that ECM deposition was assembled into extracellular matrix fibrils appeared normal. similar in each case with the S2∆S cells being unable to deposit The distribution pattern was slightly altered, however, with a a significant amount of fibrillar ECM irrespective of substrate. matrix, that was restricted to cell-cell contact areas with It was also possible that the S2∆S cells secreted a factor that reduced fibrils on top and underneath the cells. Their internal indirectly caused proteolysis or internalization of matrix 498 C. M. Klass, J. R. Couchman and A. Woods

D

Fig. 3. Syndecan-2 expression levels in CHO-K1 and transfected cells. (A-C) FACS analysis was performed with CHO-K1 (solid lines), S2 (A), S2∆S (B), and S2∆R (C) cells (dotted lines). All the transfected cells showed increased labeling for syndecan-2. Negative control (second antibody only) is shown on the left of each graph. (D) Immunoblotting of CHO-K1, S2A and S2 cells digested with heparitinase and chondroitinase with pan-syndecan antibody visualized syndecan-2 core protein (≈48 kDa) reduced in S2A cells and increased in S2 cells with respect to CHO-K1 cells. Syndecan-1 and -4 core proteins are also arrowed. Undigested S2 cells showed a high molecular weight smear typical for intact proteoglycans. The 48 kDa core proteins were subjected to densitometric analysis normalized to cytokeratin levels (lower panel). The experiment was performed three times (mean ± standard error). components. However, similar degrees of proteolytic cleavage cells after 30 minutes. However, both S2∆S and pcDNA3 cells of laminin were noted for pcDNA3 and S2∆S cells (see below). had similar cellular laminin and fibronectin content at 90 In addition, incubation of control pcDNA3 cells with minutes (Fig. 7). Since a constant amount of antibody was used conditioned medium from S2∆S cells did not affect subsequent at each time point, the ratio of cold to radiolabeled material matrix assembly even after continuous treatment for two decreased with time, resulting in increased cpm being passages. immunoprecipitated. These increases were similar in both cell types, further indicating similar processing and export. Matrix synthesis and export in normal and Proteolytically cleaved fragments of laminin and fibronectin transfected CHO-K1 cells were present in immunoprecipitates of both S2∆S and pcDNA3 To determine if decreased matrix assembly by the S2∆S cells cell supernatants. This pattern of laminin degradation is typical was due to decreased laminin and fibronectin synthesis or for cleavage by plasmin (Liotta et al., 1981). When the laminin export, pulse chase immunoprecipitation experiments were or fibronectin contents in supernatants and the corresponding performed. cell lysates were quantified, there were no significant S2∆S cell lysates contained less laminin (Fig. 7A) but more differences between pcDNA3 and S2∆S cells (Fig. 7A,B). At fibronectin (Fig. 7B) than vector-only transfected pcDNA3 these early time points, matrix fibrils are not visible by Syndecan-2 controls matrix assembly 499

Fig. 4. Matrix assembly in normal and transfected CHO-K1 cells in sparse culture. Cells were grown for 1-2 days and stained for laminin. In non-permeabilized cells (A,C,E) only extracellular matrix is revealed (arrows); in permeabilized cells (B,D,F,G,H) additional intracellular label is visible: CHO-K1 (A,B) and S2 (C,D) cells showed similar results. S2∆S cells lacked external label (E) but had internal labeling (F). Arrowhead in E indicates a damaged cell. S2∆R (G) showed some reduction in extracellular laminin (see text). S4∆I (H) cells assembled laminin into a normal matrix. Bar, 10 µm. immunofluorescence (not shown), so the lysate represents mainly intracellular material plus any that is cell surface associated. The supernatant represents exported material. To further confirm a lack of extracellular matrix in S2∆S cells, these and pcDNA3 cells were ‘chased’ for 24 hours to allow any ECM deposition, then briefly trypsinized to remove any extracellular label. Immunoprecipitation of duplicate lysates before and after trypsinization showed a loss of laminin and fibronectin from pcDNA3, but not S2∆S cells (not shown). S2∆S-cells have lost the ability to rearrange laminin or fibronectin substrates Since laminin and fibronectin synthesis and export appears similar between S2∆S and CHO-K1 or pcDNA3 cells, but S2∆S cells cannot form these into a fibrillar matrix, cell surface events may be disrupted in S2∆S cells. To test this, CHO-K1, S2, S2∆S and S2A cells were plated for 4 hours on coverslips coated with laminin or fibronectin and immunostained with anti-laminin or anti-fibronectin to detect the ability of these cells to rearrange substrates into a fibrillar matrix (Woods et al., 1988; Christopher et al., 1997). Spreading on either substrate did not differ with cell type. CHO-K1 (Fig. 8A) and S2 cells (Fig. 8B) reorganized either laminin (not shown) or fibronectin substrates into thick fibrils, especially at the cell edges and between adjacent cells (arrows). The S2 cells were more effective at substrate reorganization than wild-type CHO- K1 cells, whereas S2∆S cells (Fig. 8C) almost completely lacked this ability. Substrates remained uniform, even after 24 hours, with dark areas representing the spread cells. S2A cells also had a reduced capacity to rearrange laminin or fibronectin into fibrils (Fig. 8D) with only a few fibrils (arrowed) being formed. All cell types could remove some of the substrate (evidenced as dark streaks), but only the CHO-K1, S2 and some of the S2A cells, but not the S2∆S cells, redistributed this into fibers. Exogenous fibronectin partially restores matrix assembly in S2∆S cells CHO-K1 cells can incorporate exogenously provided fibronectin into a matrix (Wu et al., 1995b). Addition of bovine plasma fibronectin to confluent cultures of wild-type CHO-K1 (Fig. 9A) and S2∆S cells increased matrix assembly. However, the S2∆S matrix was less substantial (Fig. 9B). When syndecan-2 levels were monitored in these S2∆S cells, most showed no labeling, but some intermediate (arrowhead) and high expressors (arrow) were visible (Fig. 9C). The dual image further showed that those cells expressing intermediate or high levels of truncated syndecan-2 at the cell surface were virtually unable to accumulate a pericellular matrix. Only those cells staining weakly for syndecan-2 core protein could establish a matrix (Fig. 9D). This suggests that exogenous fibronectin in 500 C. M. Klass, J. R. Couchman and A. Woods

Fig. 5. Matrix assembly in transfected CHO-K1 cells in dense culture. CHO-K1 cells (A) assembled a network of matrix fibrils, whereas S2∆S (B) and S2A (C) cells established only a patchy, sparse matrix with only a few fibrils (arrows). Cells were not permeabilized. Bar, 10 µm. high concentration may trigger ‘nucleation’ of matrix fibrils, than the mere presence of matrix molecules and activated and that this ‘nucleation’ is deficient in S2∆S cells. Laminin integrins are required for matrix assembly. Fibronectin matrix was not restored in S2∆S cells following fibronectin fibrillogenesis occurs centripetally in cultured cells (Ljubimov addition (not shown), indicating the independent nature of and Vasiliev, 1982), and requires an intact microfilament laminin and fibronectin fibril formation. cytoskeleton (Ali and Hynes, 1977; Wu et al., 1995b). Previous studies (Woods et al., 1988) had indicated a common The S2∆S defect is not due to reduced cell surface requirement for both cell-binding and heparin-binding integrins or lack of integrin activation fibronectin domains for stress fiber and focal adhesion Assembly of laminin and fibronectin matrices in CHO-K1 cells is mediated by α6β1 and α5β1 integrins (Furtado et al., 1992; Wu et al., 1993). However, FACS analysis with anti-α5 and - β1 integrin (data not shown) showed no differences between CHO-K1 and S2∆S cells in cell surface expression levels, even after FACS selection for high S2∆S expression. An α5β1 integrin-deficient CHO cell line termed CHO B2 cannot assemble a regular fibronectin matrix (Schreiner et al., 1989; Wu et al., 1993). However, these cells did assemble a laminin matrix (Fig. 10A), albeit not as well organized as in CHO-K1 cells (Fig. 4A), possibly due to reduced cell spreading. This ∆ confirms that fibronectin and laminin matrix assemblies are regulated independently. Transfection of CHO B2 cells with the S2∆S construct decreased laminin matrix formation (Fig. 10B) with extracellular laminin organized not as fibrils but only as amorphous aggregates. Thus, truncated syndecan-2 suppressed the fibronectin-independent assembly of a laminin matrix. Transfection of CHO B2 cells with activated αIIbβL∆β3 integrin (O’Toole et al., 1994) did not change the laminin matrix (Fig. 10C), but restored fibronectin assembly (Fig. 10E), through the expression of a constitutively activated fibronectin-binding integrin (Wu et al., 1995c). However, when cells expressing activated integrin were cotransfected with the S2∆S construct, laminin (Fig. 10D) and fibronectin (Fig. 10F) matrix assembly were both reduced, suggesting that S2∆S expression could largely override even the effects of activated integrins. Cytoskeletal changes in S2∆S cells The above experiments indicated that cell surface events other

Fig. 6. Quantitation of cell surface laminin in CHO-K1, S2 and S2∆S cell lysates. Immunoprecipitated, cell surface biotinylated laminin was detected by SDS-PAGE, electrophoretic transfer and streptavidin S ∆ binding. In S2∆S cells all laminin chains were decreased, whereas CHO-K1 and S2 cells had equivalent levels. Experiments were S2 performed twice with similar results. Data are shown as a percentage relative to CHO-K1 ± s.d. Syndecan-2 controls matrix assembly 501

∆ ∆

Fig. 7. Failure of ECM assembly by S2∆S cells is not due to a defect in synthesis or export of ECM proteins. Fluorograms from one of three pulse chase labeling experiments followed by immunoprecipitation of laminin (A) and fibronectin (B) are shown. Upper panels represent pcDNA3 cells; lower panels are S2∆S cells. Lanes 1, 3, 5 and 7 represent ECM molecules retained in the cell lysates after 0, 30, 60 and 90 minutes. Lanes 2, 4 and 6 represent laminin or fibronectin exported into the supernatants after 30, 60 and 90 minutes of ‘chase’. Lane 8 represents control immunoprecipitation of cell lysate after a 90 minute ‘chase’ without addition of primary antibody. Laminin and fibronectin were quantified by densitometry (lower panels in A and B), and were the mean of duplicates of one of three experiments. formation and for matrix assembly. We, therefore, tested a much less organized microfilament system than S2∆S cells, whether the reduction in matrix assembly correlated with a control of matrix assembly seems not to be solely dictated by reduction in cytoskeletal organization. Phase contrast the cytoskeleton. Consistent with this, double labeling (not microscopy (Fig. 11A-C) showed that S2∆S cells (B) spread shown) indicated that microfilament bundles were not colinear less well than CHO-K1 cells (A), and exhibited continuous with the linear punctate staining for syndecan-2 (Fig. 2). ruffling. The morphology of S2∆S cells resembled that of cells expressing a truncated syndecan-4 construct (S4∆I, Longley et al., 1999; Fig. 11C). However, interference reflection DISCUSSION microscopy (Woods et al., 1992) indicated that 60% of S2∆S cells could still form focal adhesions. This is reduced in respect The V region of the cytoplasmic domain of syndecan-2 core to CHO-K1 cells (70-75%) but higher than in S4∆I cells, where protein appears to be crucial for fibronectin and laminin matrix only 26% of the population form these structures. Phalloidin formation. Laminin is usually described as a basement membrane labeling showed stress fibers present in CHO-K1 (Fig. 11D), component, but it is found as fibrils in connective tissue (Timpl with some reduction in S2∆S cells (E), but this reduction was and Brown, 1996), and in cultured cells it can form a three- not as pronounced as in S4∆I cells (F). Labeling with anti- dimensional fibrillar matrix indistinguishable from that of paxillin confirmed the presence of focal adhesions in CHO-K1 fibronectin. Both fibronectin and laminin can bind entactin (Wu (G) and S2∆S (H) cells, but their lack in S4∆I cells (I). Since et al., 1988), so fibronectin and laminin matrices may be the S4∆I cells can still form a fibrillar matrix (Fig. 4H), but have interconnected. However, in the case of CHO B2 and S2∆S cells, 502 C. M. Klass, J. R. Couchman and A. Woods

Fig. 8. S2∆S cells cannot rearrange laminin or fibronectin substrates. Immunostaining with anti-fibronectin shows rearrangement by 4 hours of the fibronectin substrate by CHO-K1 cells (A). This was even more evident with S2 cells at 4 hours (arrows). In contrast, S2∆S cells (C) did not reorganize the fibronectin substrate even after 24 hours, and S2A cells (D) showed reduced matrix reorganization after this time. Cells were not permeabilized. Fibrils are arrowed. Bar, 10 µm.

of matrix assembly. This was shown by immunofluorescence, where only short thin fibrils of laminin or fibronectin were detected, even in dense cultures. It was evident in several stably transfected lines of CHO-K1 cells, where most cells were low expressors of the truncated syndecan-2, and in cultures of the 5% highest expressors selected by FACS. Immunoprecipitation of biotinylated cell surface laminin confirmed an up to an almost two-thirds reduction in extracellular laminin. Inability of S2∆S cells to form an endogenous matrix was not due to a defect in synthesis of either fibronectin or laminin, nor to decreased export of matrix components. Reduced matrix was also not due to proteolysis by secreted enzymes, since conditioned medium from S2∆S cells had no effect on cells containing empty vector, and the pattern of laminin proteolysis was similar in pcDNA3 and S2∆S culture supernatants. The decrease in matrix assembly was also not due to a lack of cell interaction with the matrix, since S2∆S and S2A cells attached and spread similarly to CHO-K1 or pcDNA3 cells on matrix- coated substrates, and they could remove the substrate coat around them (Grinnell, 1986; Christopher et al., 1997). However, unlike CHO-K1 cells, S2∆S cells were unable to form matrix fibrils with removed substrate fibronectin or laminin. S2A cells also showed a greatly reduced capacity to reorganize substrates, consistent with an approximate one third fibronectin and laminin matrix fibrils are regulated independently, reduction in syndecan-2 core protein levels. When exogenous, but both depend on syndecan-2 proteoglycan. This suggests that soluble fibronectin was provided to S2∆S cells, some matrix syndecan-2 operates independently of the integrins involved. was then formed, but only by cells expressing low levels of Deletion of the C-terminal 14 amino acids of the S2∆S construct. Those cells with high surface levels remained cytoplasmic domain of syndecan-2 led to a marked diminution free of matrix. The data suggest that matrix ‘nucleation’ at the

Fig. 9. Exogenous fibronectin can partially restore a fibrillar matrix. Addition of bovine plasma fibronectin increased the matrix of CHO-K1 cells (A) and to a lesser extent of S2∆S cells (B). Cells in B were also labeled for syndecan-2 expression (C), and the double image is shown in D. The dual image (D) revealed fibronectin matrix only on the lower syndecan-2 expressors, while high expressors (arrowed in C) showed no matrix assembly. (C) Overexposed image to show limited labeling of low expressors. The high (arrow) and intermediate expressors (arrowhead) are indicated. Fibronectin was labeled with rabbit anti-fibronectin and Texas Red-conjugated goat anti-rabbit under identical conditions for A and B. Syndecan-2 was visualized with chicken anti-syndecan-2 ectodomain followed by FITC goat anti-chicken. Cells were passage 9 after FACS selection, when phenotypic drift has occurred, and were not permeabilized. Bar 10 µm. Syndecan-2 controls matrix assembly 503

Integrins are required for both endogenous fibronectin matrix assembly (Fogerty et al., 1990; Wu et al., 1993) and for incorporation of exogenous fibronectin into deoxycholate- resistant matrices (Wu et al., 1995c). Much research has shown that fibronectin fibrillogenesis normally requires interaction of α5β1 with the RGD-containing domain of fibronectin, specifically the type III repeat 10 (reviewed by Bultmann et al., 1998; Schwarzbauer, 1991; Sechler et al., 1997; Wu et al., 1993). However, S2∆S cells showed no alteration in surface levels of α5 and β1 integrin. Interactions of α5β1 with the RGD site do not seem to be sufficient for fibrillogenesis and secondary sites may be needed both at the cell surface and within fibronectin. The existence of a ‘matrix assembly receptor’ has been postulated for some time, but has been elusive. The studies reported here indicate that syndecan-2 may be this ‘receptor’, confirming previous indications for a role for heparan sulfate proteoglycans (Chung and Erickson, 1997). Syndecans bind the heparin-binding domains of matrix molecules (reviewed by Bernfield et al., 1992; Carey, 1997), including fibronectin and laminin. Lack of interaction with the amino-terminal heparin-binding (Hep I) domain of fibronectin results in the formation of only short, thin fibrils (McDonald et al., 1987; Schwarzbauer, 1991) and the first five type I repeats (I1-5) and first type III repeat (III1) have all been implicated in matrix assembly (McKeown-Longo and Mosher, 1985; Quade and McDonald, 1988; Sottile et al., 1991). Other studies indicate a role for type III repeats 12-14 of the Hep II high affinity heparin-binding site near the C terminus of the molecule (Bultmann et al., 1998) and our previous studies showed that both the integrin-binding domain and either the Hep I or Hep II domains were needed for reorganization of an exogenous fibronectin substrate into fibrils (Woods et al., 1988). A ‘synergy’ site PHSRN in repeat III9 is needed for adhesion to fibronectin through α5β1 (Aota et al., 1994), and for efficient incorporation of recombinant fibronectins into fibers (Sechler et al., 1997). Fibrillogenesis initiates at the cell surface where receptor clustering provides a nucleus of fibronectin ‘activated’ by integrin binding into an unfolded form capable of self-association. This gives rise to short fibrils at the cell surface by 1 hour that increase in size and number. Additional fibronectin-fibronectin interactions generate deoxycholate-insoluble fibers over 16 hours forming a three- dimensional network of larger fibers extending away from the cell surface. Absence of the synergy site delays initiation of fibril formation, resulting in only very sparse short fibrils even at 16 hours, thus preventing fiber assembly. The need for the synergy site can be circumvented by activation of α β with ∆ 5 1 Fig. 10. Overexpression of S2 S can override the effect of activated antibodies or Mn2+. We, therefore, monitored whether the integrin. CHO B2 cells (A) assembled a laminin matrix, but CHO B2 presence of activated integrins would prevent the effect of transfected with S2∆S (B) did not. Transfection of CHO B2 with transfection with truncated syndecan-2. However, S2∆S constitutively activated αIIbαL∆β3 integrin resulted in no change in laminin deposition (C) but a restoration of fibronectin matrix appears to override even activated integrins, since CHO B2 α α ∆β assembly (E). Cotransfection of activated αIIbαL∆β3 and S2∆S led to cells expressing activated IIb L 3 that normally allows a reduction in deposition of both laminin (D) and fibronectin (F). fibronectin matrix assembly (Wu et al., 1995c) established only Cells were not permeabilized, so only cell surface matrix (arrows) is a sparse non-fibrillar matrix, if cotransfected with the truncated seen. Bars, 10 µm. syndecan-2 construct. The cytoskeleton may be involved in matrix fibrillogenesis. cell surface may lie at the heart of the S2∆S phenotype, and Treatment of cells with cytochalasin to disrupt the actin since this was not restricted to a single matrix component, microfilament system prevents fibronectin matrix assembly syndecan-2 may be involved in signaling responses common even in cells expressing activated integrin (Wu et al., 1995c). to fibrillogenesis. The requirements for both integrin- and heparin-binding 504 C. M. Klass, J. R. Couchman and A. Woods

Fig. 11. Morphology of CHO-K1, S2∆S cells and S4∆I cells. Phase contrast microscopy shows well spread CHO-K1 cells (A), and some reduction in spreading of S2∆S cells (B). This is more pronounced with S4∆I (C) cells. Stress fibers are seen by labeling with phalloidin in CHO-K1 (D) and S2∆S cells (E). Although stress fibers in S2∆S cells are less substantial, the cytoskeleton is more organized than in S4∆I cells (F). Paxillin staining shows focal adhesions in CHO-K1 (G) and S2∆S cells (H), but their lack in S4∆I cells (I). Bar, 10 µm. domains of fibronectin for matrix fibrillogenesis, or the were seen in S2∆S populations before FACS selection, where reorganization of substrates into fibrils, parallels a similar need modest increases in expression levels were noted. This argues for both domains for focal adhesion and stress fiber formation that neither increased ectodomain expression levels, nor in primary fibroblasts (Woods et al., 1988). Lysophosphatidic potential glycanation changes in ectodomains due to acid increases both stress fiber formation and fibronectin overexpression, are responsible for the definitive phenotype of fibrillogenesis (Zhong et al., 1997), and protein kinase C the S2∆S cells. The partial truncation mutant S2∆S terminates activity is needed for both events (Somers and Mosher, 1993; at serine197 in the V region of syndecan-2 core protein. Woods and Couchman, 1992). However, increased spreading Synthetic peptides encompassing the syndecan-2 V region induced by phorbol ester treatment did not result in increased (LGERKPSSAAYQK) can be phosphorylated by protein matrix assembly (Sechler et al., 1997), indicating that the two kinase C at both serine197 and serine198, with the extent of events do not necessarily correlate. This is confirmed by our phosphorylation varying with concentration (Oh et al., 1997c), present studies where both de novo matrix assembly and the and other studies have shown syndecan-2 phosphorylation in reorganization of surface bound matrix were disrupted in S2∆S vitro (Prasthofer et al., 1995) and in vivo on serine (Itano et cells, but the cytoskeleton was not dramatically altered. al., 1996). Phosphorylation of the V region may regulate matrix Additionally, transfection with a partially truncated syndecan- assembly. This would be compromised in both the S2∆S and 4 construct (S4∆I) had little or no effect on matrix assembly, S2∆R constructs, but only the S2∆S cells exhibited an almost even though these cells showed dramatically reduced focal complete lack of matrix assembly. However, the S2∆R adhesion and stress fiber formation. construct may be more rapidly turned over at the cell surface The defect in matrix assembly did not occur if cells were as seen with syndecans totally lacking a cytoplasmic domain transfected with a construct where the cytoplasmic domain was (Miettinen et al., 1994), thus allowing normal functioning fully deleted (S2∆R) except at very high expression levels, nor of endogenous syndecan-2. All syndecans self-associate did transfection with full length syndecan-2 increase matrix (Bernfield et al., 1992; Carey, 1997; Oh et al., 1997a,b; Woods assembly even at high expression levels. In contrast, changes and Couchman, 1998), and the association of the partially Syndecan-2 controls matrix assembly 505 truncated core proteins with those of endogenous normal distinct integrins in focal contacts is determined by the substratum syndecan-2 may prevent the functioning of whole clusters of composition. J. Cell Sci. 92, 67-75. syndecans aggregated by ligand. The positive requirement for Fogerty, F. J., Akiyama, S. K., Yamada, K. M. and Mosher, D. F. (1990). Inhibition of binding of fibronectin to matrix asssembly sites by anti-integrin syndecan-2 in matrix formation is emphasized by the loss of (α5β1) antibodies. J. Cell Biol. 111, 699-708. matrix assembly in cells expressing antisense for syndecan-2. Furtado, G. C., Cao, Y. and Joiner, K. A. (1992). Laminin on Toxoplasma Since syndecan-4 V region binds protein kinase C and gondii mediates parasite binding to the beta 1 integrin receptor alpha 6 beta modulates its activity (Oh et al., 1997a,b), the possibility 1 on human foreskin fibroblasts and Chinese hamster ovary cells. Infect. Immun. 60, 4925-4931. emerges that the V regions of syndecans function specifically Giancotti, F.G. and Ruoslahti, E. (1990). Elevated levels of the α5β1 by controlling different kinase systems. fibronectin receptor suppress the transformed phenotype of Chinese hamster ovary cells. Cell 60, 849-859. This work was supported by grant Kl 1122-1.1 (C.M.K.) from the Grinnell, F. (1986). Focal adhesion sites and the removal of substratum-bound German Research Foundation (Deutsche Forschungsgemeinschaft) fibronectin. J. Cell Biol. 103, 2697-2706. and NIH grants P60 AR20614 to the Multipurpose Arthritis and Hocking, D. C., Sottile, J. and McKeown-Longo, P. J. (1994). Fibronectin’s Musculoskeletal Diseases Center at UAB and RO1 DK 54605 (A.W.) III-1 module contains a conformation-dependent binding site for the amino- and The Cell Adhesion and Matrix Research Center at UAB. terminal region of fibronectin. J. Biol. Chem. 269, 19183-19187. Itano, N., Oguri, K., Nagayasu, Y., Kusano, Y., Nakanishi, H., David, G. and Okayama, M. (1996). Phosphorylation of a membrane-intercalated proteoglycan, syndecan-2, expressed in a stroma-inducing clone from a REFERENCES Lewis mouse sarcoma. Biochem. J. 315, 925-930. Kim, C. W., Goldberger, O. A., Gallo, R. L. and Bernfield, M. (1994). Akiyama, S. K., Yamada, S. S., Chen, W. T. and Yamada, K. M. (1989). Members of the syndecan family of heparan sulfate proteoglycans are Analysis of fibronectin receptor function with monoclonal antibodies: Roles expressed in distinct cell-, tissue-, and development-specific patterns. Mol. in cell adhesion, migration, matrix assembly, and cytoskeletal organization. Biol. Cell 5, 797-805. J. Cell Biol. 109, 863-875. Koff, A., Cross, F., Fisher A., Schumacher, J., Leguellec, K., Philippe M. Ali, I. U. and Hynes, R. O. (1977). Effects of cytochalasin B and colchicine and Roberts, J. M. (1991). Human cyclin E, a new cyclin that interacts on attachment of a major surface protein of fibroblasts. Biochim. Biophys. with two members of the CDC2 family. Cell 66, 1217-1228. Acta 471, 16-24. Koff, A., Giordano, A., Desai, D., Yamashita, K., Harper, J. W., Elledge, Aota, S., Nomizu, M. and Yamada, K. M. (1994). The short amino acid S., Nishimoto, T., Morgan, D. O., Franza, B. R. and Roberts. J. M. sequence Pro-His-Ser-Arg-Asn in human fibronectin enhances cell-adhesive (1992). Formation and activation of a cyclin E-cdk2 complex during the G1 function. J. Biol. Chem. 269, 24756-24761. phase of the human cell cycle. Science 257, 1689-1694. Baciu, P. C. and Goetinck, P. F. (1995). Protein kinase C regulates the Kojima, T., Shworak, N. W. and Rosenberg, R. D. (1992). Molecular cloning recruitment of syndecan-4 into focal contacts. Mol. Biol. Cell 6, 1503- and expression of two distinct cDNA-encoding heparan sulfate proteoglycan 1513. core proteins from a rat endothelial cell line. J. Biol. Chem. 267, 4870-4877. Barry, E. L. R. and Mosher, D. F. (1989). Factor XIIIa-mediated cross- LeBaron, R. G., Esko, J. D., Woods, A., Johansson, S. and Höök, M. (1988). linking of fibronectin in fibroblast cell layers. J. Biol. Chem. 264, 4179- Adhesion of glycosaminoglycan-deficient Chinese hamster ovary cell 4185. mutants to fibronectin substrata. J. Cell Biol. 106, 945-952. Bernfield, M., Kokenyesi, R., Kato, M., Hinkes, M. T., Spring, J., Gallo, Liotta, L. A., Goldfarb, R. H. and Terranova, V. (1981). Cleavage of laminin R. L. and Lose, E. J. (1992). Biology of the syndecans. Annu. Rev. Cell by thrombin and plasmin: Alpha thrombin selectively cleaves the beta chain Biol. 8, 365-393. of laminin. Thomb. Res. 21, 663-673. Brown, P. J. and Juliano, R. L. (1988). Monoclonal antibodies to distinctive Ljubimov, A. V. and Vasiliev, J. M. (1982). Distribution of fibronectin- epitopes on the α and β subunits of the fibronectin receptor. Exp. Cell Res. containing structures on the surface of lamelloplasm and endoplasm of 177, 303-318. fibroblasts: hypothesis of receptor-mediated assembly of fibronectin Bultmann, H., Santas, A. J. and Pesciotta Peters, D. M. (1998). Fibronectin structures. Biol. Int. Rep. 6, 105-112. fibrillogenesis involves the Heparin II binding domain of fibronectin. J. Biol. Longley, R. L., Woods, A., Fleetwood, A., Cowling, G. J., Gallagher, J. T. Chem. 273, 2601-2609. and Couchman, J. R. (1999). Control of morphology, cytoskeleton and Carey, D. J. (1997). Syndecans: multifunctional cell-surface co-receptors. migration by syndecan-4. J. Cell Sci. 112, 3421-3431. Biochem. J. 327, 1-16. McCarthy, K. J., Bynum, K., St. John, P. L., Abrahamson, D. R. and Chen, H. and Mosher, D. F. (1996). Formation of sodium dodecyl sulfate- Couchman, J. R. (1993). Basement membrane proteoglycans in glomerular stable fibronectin multimers. Failure to detect products of thiol-disulfide morphogenesis: chondroitin sulfate proteoglycan is temporally and spatially exchange in cyanogen bromide or limited acid digests of stabilized matrix restricted during development. J. Histochem. Cytochem. 41, 401-414. fibronectin. J. Biol. Chem. 271, 9084-9089. McDonald, J. A., Quade, B. J., Broekelmann, T. J., LaChance, R., Forsman, Christopher, R. A., Kowalczyk, A. P. and McKeown-Longo, P. J. (1997). K., Hasegawa, E. and Akiyama, S. (1987). Fibronectin’s cell adhesive Localization of fibronectin matrix assembly sites on fibroblasts and domain and an amino-terminal matrix assembly domain participate in its endothelial cells. J. Cell Sci. 110, 569-581. assembly into fibroblast pericellular matrix. J. Biol. Chem. 262, 2957-2967. Chung, C. Y. and Erickson, H. P. (1997). Glycosaminoglycans modulate McKeown-Longo, P. J. and Mosher, D. F. (1984). Mechanism of formation fibronectin matrix assembly and are essential for matrix incorporation of of disulfide-bonded multimers of plasma fibronectin in cell layers of tenascin-C. J. Cell Sci. 110, 1413-1419. cultured human fibroblasts. J. Biol. Chem. 259, 12210-12215. David, G., Bai, X. M., Van der Schueren, B., Cassiman, J.-J. and Van den McKeown-Longo, P. J. and Mosher, D. F. (1985). Interaction of the 70,000- Berghe, H. (1992a). Developmental changes in heparan sulfate expression: mol-wt amino-terminal fragment of fibronectin with the matrix-assembly In situ detection with mAbs. J. Cell Biol. 119, 961-975. receptor of fibroblasts. J. Cell Biol. 100, 364-374. David, G., Van der Schueren, B., Marynen, P., Cassiman, J.-J. and Van Miekka, S. I., Ingham, K. C. and Menache, D. (1982). Rapid methods for den Berghe, H. (1992b). Molecular cloning of amphiglycan, a novel isolation of human plasma fibronectin. Thromb. Res. 27, 1-14. integral membrane heparan sulfate proteoglycan expressed by epithelial and Miettinen, H. M., Edwards, S. N. and Jalkanen, M. (1994). Analysis of fibroblastic cells. J. Cell Biol. 118, 961-969. transport and targeting of syndecan-1: Effect of cytoplasmic tail deletions. David, G., Bai, X. M., Van der Schueren, B., Marynen, P., Cassiman, J.- Mol. Biol. Cell 5, 1325-1339. J. and Van den Berghe, H. (1993). Spatial and temporal changes in the O’Toole, T. E., Katagiri, Y., Faull, R. J., Peter, K., Tamura, R. N., expression of fibroglycan (syndecan-2) during mouse embryonic Quaranta, V., Loftus, J. C., Shattil, S. J. and Ginsberg, M. H. (1994). development. Development 119, 841-854. Integrin cytoplasmic domains mediate inside-out signal transduction. J. Cell Dzamba, B. J., Bultmann, H., Akiyama, S. K. and Peters, D. M. (1994). Biol. 124, 1047-1059. Substrate-specific binding of the amino terminus of fibronectin to an integrin Oh, E.-S., Woods, A. and Couchman, J. R. (1997a). Syndecan-4 complex in focal adhesions. J. Biol. Chem. 269, 19646-19652. proteoglycan regulates the distribution and activity of protein kinase C. J. Fath, K. R., Edgell, C.-J. and Burridge, K. (1989). The distribution of Biol. Chem. 272, 8133-8136. 506 C. M. Klass, J. R. Couchman and A. Woods

Oh, E.-S., Woods, A. and Couchman, J. R. (1997b). Multimerization of the Woods, A. and Couchman, J. R. (1994). Syndecan-4 heparan sulfate cytoplasmic domain of syndecan-4 is required for its ability to activate proteoglycan is a selectively enriched and widespread focal adhesion protein kinase C. J. Biol. Chem. 272, 11805-11811. component. Mol. Biol. Cell 5, 183-192. Oh, E.-S., Couchman, J. R. and Woods, A. (1997c). Serine phosphorylation Woods, A. and Couchman, J. R. (1998). Syndecans: synergistic activators of of syndecan-2 proteoglycan cytoplasmic domain. Arch. Biochem. Biophys. cell adhesion. Trends Cell Biol. 8, 189-192. 344, 67-74. Woods, A., Johansson, S. and Höök, M. (1988). Fibronectin fibril formation Ott, V. L. and Rapraeger, A.C. (1998). Tyrosine phosphorylation of involves cell interactions with two fibronectin domains. Exp. Cell Res. 177, syndecan-1 and Ð4 cytoplasmic domains in adherent B82 fibroblasts. J. Biol. 272-283. Chem. 273, 35291-35298. Wu, C., Friedman, R. and Chung, A. E. (1988). Analysis of the Pierce, A., Lyon, M., Hampson, I. N., Cowling G. J. and Gallagher, J. T. assembly of laminin and the laminin-entactin complex with laminin chain (1992). Molecular cloning of the major cell surface heparan sulfate specific monoclonal and polyclonal antibodies. Biochemistry 27, 8780- proteoglycan from rat liver. J. Biol. Chem. 267, 3894-3900. 8787. Prasthofer, T., Ek, B., Ekman, P., Owens, R., Höök, M., Johansson, S. Wu, C., Bauer, J. S., Juliano, R. L. and McDonald, J. A. (1993). The α5β1 (1995). Protein kinase C phosphorylates two of the four known syndecan integrin fibronectin receptor, but not the α5 cytoplasmic domain, functions cytoplasmic domains in vitro. Biochem. Mol. Biol. Int. 36, 793-802. in an early and essential step in fibronectin matrix assembly. J. Biol. Chem. Quade, B. J. and McDonald, J. A. (1988). Fibronectin’s amino-terminal 268, 21883-21888. matrix assembly site is located within the 29-kDa amino-terminal domain Wu, C., Chung, A. E., McDonald, J. A. (1995a). A novel role for α3β1 containing five type I repeats. J. Biol. Chem. 263, 19602-19609. integrins in extracellular matrix assembly. J. Cell Sci. 108, 2511-2523. Schreiner, C. L., Bauer, J. S., Danilov, Y. N., Hussein, S., Sczekan, M. M. Wu, C., Keivens, V. M., O’Toole, T. E., McDonald J. A. and Ginsberg, M. and Juliano, R. L. (1989). Isolation and characterization of Chinese H. (1995b). Integrin activation and cytoskeletal interaction are essential for hamster ovary cell variantsdeficient in the expression of fibronectin receptor. the assembly of a fibronectin matrix. Cell 83, 715-724. J. Cell Biol. 109, 3157-3167. Wu, C., Fields, A. J., Kapteijn, B. A. and McDonald, J. A. (1995c). The Schwarzbauer, J. E. (1991). Identification of the fibronectin sequences role of α4β1 integrin in cell motility and fibronectin matrix assembly. J. Cell required for assembly of a fibrillar matrix. J. Cell Biol. 113, 1463-1473. Sci. 108, 821-829. Sechler, J., Corbett, S. A. and Schwarzbauer, J. E. (1997). Modulatory roles Wu, C., Hughes, P. E., Ginsberg, M. H. and McDonald, J. A. (1996). for the integrin activation and the synergy site of fibronectin during matrix Identification of a new biological function for the integrin alpha v beta 3: assembly. Mol. Biol. Cell 8, 2563-2573. initiation of fibronectin matrix assembly. Cell Adhes. Commun. 4, 149- Somers, C. E. and Mosher, D. F. (1993). Protein kinase C modulation of 158. fibronectin matrix assembly. J. Biol. Chem. 268, 22277-22280. Yatohgo, T., Izumi, M., Kashiwagi, H. and Hayashi, M. (1988). Novel Sottile, J., Schwarzbauer, J. Selegue, J. and Mosher, D. F. (1991). Five type purification of vitronectin from human plasma by heparin affinity I modules of fibronectin form a functional unit that binds to fibroblasts and chromatography. Cell Struct. Funct. 13, 281-292. staphylococcus aureus. J. Biol. Chem. 266, 12840-12843. Zhang, Z., Morla, A. O., Vuori, K., Bauer, J. S., Juliano, R. L. and Sottile, J. and Wiley, S. (1994). Assembly of amino-terminal fibronectin Ruoslahti, E. (1993). The alpha v beta 1 integrin functions as a fibronectin dimers into the extracellular matrix. J. Biol. Chem. 269, 17192-17198. receptor but does not support fibronectin matrix assembly and cell migration Timpl, R. and Brown, J. C. (1996). Supramolecular assembly of basement on fibronectin. J. Cell Biol. 122, 235-242. membranes. Bioessays 18, 123-132. Zhong, C., Kinch, M. S. and Burridge, K. (1997). Rho-stimulated Woods, A. and Couchman, J. R. (1992). Protein kinase C involvement in contractility contributes to the fibroblastic phenotype of Ras-transformed focal adhesion formation. J. Cell Sci. 101, 277-290. epithelial cells. Mol. Biol. Cell 8, 2329-2344.